The Ef cient Market Hypothesis and ItsCritics

Burton G Malkiel

A generation ago the ef cient market hypothesis was widely accepted byacademic nancial economists for example see Eugene Famarsquos (1970)in uential survey article ldquoEf cient Capital Marketsrdquo It was generally be-

lieved that securities markets were extremely ef cient in re ecting informationabout individual stocks and about the stock market as a whole The accepted viewwas that when information arises the news spreads very quickly and is incorporatedinto the prices of securities without delay Thus neither technical analysis which isthe study of past stock prices in an attempt to predict future prices nor evenfundamental analysis which is the analysis of nancial information such as com-pany earnings and asset values to help investors select ldquoundervaluedrdquo stocks wouldenable an investor to achieve returns greater than those that could be obtained byholding a randomly selected portfolio of individual stocks at least not with com-parable risk

The ef cient market hypothesis is associated with the idea of a ldquorandom walkrdquowhich is a term loosely used in the nance literature to characterize a price serieswhere all subsequent price changes represent random departures from previousprices The logic of the random walk idea is that if the ow of information isunimpeded and information is immediately re ected in stock prices then tomor-rowrsquos price change will re ect only tomorrowrsquos news and will be independent of theprice changes today But news is by de nition unpredictable and thus resultingprice changes must be unpredictable and random As a result prices fully re ect allknown information and even uninformed investors buying a diversied portfolio atthe tableau of prices given by the market will obtain a rate of return as generous asthat achieved by the experts

y Burton G Malkiel is Chemical Bank Chairmanrsquos Professor of Economics PrincetonUniversity Princeton New Jersey His e-mail address is bmalkielprincetonedu

Journal of Economic PerspectivesmdashVolume 17 Number 1mdashWinter 2003mdashPages 59ndash82

The way I put it in my book A Random Walk Down Wall Street rst published in1973 a blindfolded chimpanzee throwing darts at the Wall Street Journal could selecta portfolio that would do as well as the experts Of course the advice was notliterally to throw darts but instead to throw a towel over the stock pagesmdashthat isto buy a broad-based index fund that bought and held all the stocks in the marketand that charged very low expenses

By the start of the twenty- rst century the intellectual dominance of theef cient market hypothesis had become far less universal Many nancial econo-mists and statisticians began to believe that stock prices are at least partiallypredictable A new breed of economists emphasized psychological and behavioralelements of stock-price determination and they came to believe that future stockprices are somewhat predictable on the basis of past stock price patterns as well ascertain ldquofundamentalrdquo valuation metrics Moreover many of these economists wereeven making the far more controversial claim that these predictable patternsenable investors to earn excess risk adjusted rates of return

This paper examines the attacks on the ef cient market hypothesis and thebelief that stock prices are partially predictable While I make no attempt to presenta complete survey of the purported regularities or anomalies in the stock marketI will describe the major statistical ndings as well as their behavioral underpin-nings where relevant and also examine the relationship between predictability andef ciency I will also describe the major arguments of those who believe thatmarkets are often irrational by analyzing the ldquocrash of 1987rdquo the Internet ldquobubblerdquoof the n de siecle and other speci c irrationalities often mentioned by critics ofef ciency I conclude that our stock markets are far more ef cient and far lesspredictable than some recent academic papers would have us believe Moreoverthe evidence is overwhelming that whatever anomalous behavior of stock pricesmay exist it does not create a portfolio trading opportunity that enables investorsto earn extraordinary risk adjusted returns

At the outset it is important to make clear what I mean by the term ldquoef -ciencyrdquo I will use as a de nition of ef cient nancial markets that such markets donot allow investors to earn above-average returns without accepting above-averagerisks A well-known story tells of a nance professor and a student who come acrossa $100 bill lying on the ground As the student stops to pick it up the professor saysldquoDonrsquot bothermdashif it were really a $100 bill it wouldnrsquot be thererdquo The story wellillustrates what nancial economists usually mean when they say markets areef cient Markets can be ef cient in this sense even if they sometimes make errorsin valuation as was certainly true during the 1999ndash early 2000 Internet ldquobubblerdquoMarkets can be ef cient even if many market participants are quite irrationalMarkets can be ef cient even if stock prices exhibit greater volatility than canapparently be explained by fundamentals such as earnings and dividends Many ofus economists who believe in ef ciency do so because we view markets as amazinglysuccessful devices for re ecting new information rapidly and for the most partaccurately Above all we believe that nancial markets are ef cient because theydonrsquot allow investors to earn above-average risk adjusted returns In short we

60 Journal of Economic Perspectives

believe that $100 bills are not lying around for the taking either by the professionalor the amateur investor

What I do not argue is that the market pricing is always perfect After the factwe know that markets have made egregious mistakes as I think occurred during therecent Internet ldquobubblerdquo Nor do I deny that psychological factors in uencesecurities prices But I am convinced that Benjamin Graham (1965) was correct insuggesting that while the stock market in the short run may be a voting mechanismin the long run it is a weighing mechanism True value will win out in the endBefore the fact there is no way in which investors can reliably exploit any anomaliesor patterns that might exist I am skeptical that any of the ldquopredictable patternsrdquothat have been documented in the literature were ever suf ciently robust so as tohave created pro table investment opportunities and after they have been discov-ered and publicized they will certainly not allow investors to earn excess returns

A Nonrandom Walk Down Wall Street

In this section I review some of the patterns of possible predictability sug-gested by studies of the behavior of past stock prices

Short-Term Momentum Including Underreaction to New InformationThe original empirical work supporting the notion of randomness in stock

prices looked at measures of short-run serial correlations between successive stockprice changes In general this work supported the view that the stock market hasno memorymdashthat is the way a stock price behaved in the past is not useful indivining how it will behave in the future for example see the survey of articlescontained in Cootner (1964) More recent work by Lo and MacKinlay (1999) ndsthat short-run serial correlations are not zero and that the existence of ldquotoo manyrdquosuccessive moves in the same direction enable them to reject the hypothesis thatstock prices behave as true random walks There does seem to be some momentumin short-run stock prices Moreover Lo Mamaysky and Wang (2000) also ndthrough the use of sophisticated nonparametric statistical techniques that canrecognize patterns some of the stock price signals used by ldquotechnical analystsrdquo suchas ldquohead and shouldersrdquo formations and ldquodouble bottomsrdquo may actually have somemodest predictive power

Economists and psychologists in the eld of behavioral nance nd suchshort-run momentum to be consistent with psychological feedback mechanismsIndividuals see a stock price rising and are drawn into the market in a kind ofldquobandwagon effectrdquo For example Shiller (2000) describes the rise in the US stockmarket during the late 1990s as the result of psychological contagion leading toirrational exuberance The behavioralists offered another explanation for patternsof short-run momentummdasha tendency for investors to underreact to new informa-tion If the full impact of an important news announcement is only grasped over aperiod of time stock prices will exhibit the positive serial correlation found by

Burton G Malkiel 61

investigators As behavioral nance became more prominent as a branch of thestudy of nancial markets momentum as opposed to randomness seemed rea-sonable to many investigators

However several factors should prevent us from interpreting the empiricalresults reported above as an indication that markets are inef cient First while thestock market may not be a mathematically perfect random walk it is important todistinguish statistical signi cance from economic signi cance The statistical de-pendencies giving rise to momentum are extremely small and are not likely topermit investors to realize excess returns Anyone who pays transactions costs isunlikely to fashion a trading strategy based on the kinds of momentum found inthese studies that will beat a buy-and-hold strategy Indeed Odean (1999) suggeststhat momentum investors do not realize excess returns Quite the oppositemdashasample of such investors suggests that such traders did far worse than buy-and-holdinvestors even during a period where there was clear statistical evidence of positivemomentum This is because of the large transactions costs involved in attemptingto exploit whatever momentum exists Similarly Lesmond Schill and Zhou (2001) nd that standard ldquorelative strengthrdquo strategies are not pro table because of thetrading costs involved in their execution

Second while behavioral hypotheses about bandwagon effects and under-reaction to new information may sound plausible enough the evidence that sucheffects occur systematically in the stock market is often rather thin For exampleEugene Fama (1998) surveys the considerable body of empirical work on ldquoeventstudiesrdquo that seeks to determine if stock prices respond ef ciently to informationThe ldquoeventsrdquo include such announcements as earnings surprises stock splits divi-dend actions mergers new exchange listings and initial public offerings Fama nds that apparent underreaction to information is about as common as over-reaction and postevent continuation of abnormal returns is as frequent aspostevent reversals He also shows that many of the return ldquoanomaliesrdquo arise only inthe context of some very particular model and that the results tend to disappearwhen exposed to different models for expected ldquonormalrdquo returns different meth-ods to adjust for risk and when different statistical approaches are used to measurethem For example a study that gives equal weight to postannouncement returns ofmany stocks can produce different results from a study that weights the stocksaccording to their value Certainly whatever momentum displayed by stock pricesdoes not appear to offer investors a dependable way to earn abnormal returns

The key factor is whether any patterns of serial correlation are consistent overtime Momentum strategies which refer to buying stocks that display positive serialcorrelation andor positive relative strength appeared to produce positive relativereturns during some periods of the late 1990s but highly negative relative returnsduring 2000 It is far from clear that any stock price patterns are useful for investorsin fashioning an investment strategy that will dependably earn excess returns

Many predictable patterns seem to disappear after they are published in the nance literature Schwert (2001) points out two possible explanations for such apattern One explanation may be that researchers are always sifting through

62 Journal of Economic Perspectives

mountains of nancial data Their normal tendency is to focus on results thatchallenge perceived wisdom and every now and again a combination of a certainsample and a certain technique will produce a statistically signi cant result thatseems to challenge the ef cient markets hypothesis Alternatively perhaps practi-tioners learn quickly about any true predictable pattern and exploit it to the extentthat it becomes no longer pro table My own view is that such apparent patternswere never suf ciently large or stable to guarantee consistently superior investmentresults and certainly such patterns will never be useful for investors after they havereceived considerable publicity The so-called ldquoJanuary effectrdquo for example inwhich stock prices rose in early January seems to have disappeared soon after it wasdiscovered

Long-Run Return ReversalsIn the short-run when stock returns are measured over periods of days or

weeks the usual argument against market ef ciency is that some positive serialcorrelation exists But many studies have shown evidence of negative serialcorrelationmdashthat is return reversalsmdash over longer holding periods For exampleFama and French (1988) found that 25 to 40 percent of the variation in longholding period returns can be predicted in terms of a negative correlation with pastreturns Similarly Poterba and Summers (1988) found substantial mean reversionin stock market returns at longer horizons

Some studies have attributed this forecastability to the tendency of stockmarket prices to ldquooverreactrdquo DeBondt and Thaler (1985) for example argue thatinvestors are subject to waves of optimism and pessimism that cause prices todeviate systematically from their fundamental values and later to exhibit meanreversion They suggest that such overreaction to past events is consistent with thebehavioral decision theory of Kahneman and Tversky (1979) where investors aresystematically overcon dent in their ability to forecast either future stock prices orfuture corporate earnings These ndings give some support to investment tech-niques that rest on a ldquocontrarianrdquo strategy that is buying the stocks or groups ofstocks that have been out of favor for long periods of time and avoiding thosestocks that have had large run-ups over the last several years

There is indeed considerable support for long-run negative serial correla-tion in stock returns However the nding of mean reversion is not uniformacross studies and is quite a bit weaker in some periods than it is for other periodsIndeed the strongest empirical results come from periods including the GreatDepressionmdashwhich may be a time with patterns that do not generalize wellMoreover such return reversals for the market as a whole may be quite consistentwith the ef cient functioning of the market since they could result in part fromthe volatility of interest rates and the tendency of interest rates to be meanreverting Since stock returns must rise or fall to be competitive with bond returnsthere is a tendency when interest rates go up for prices of both bond and stocks togo down and as interest rates go down for prices of bonds and stocks to go up Ifinterest rates revert to the mean over time this pattern will tend to generate return

The Efcient Market Hypothesis and Its Critics 63

reversals or mean reversion in a way that is quite consistent with the ef cientfunctioning of markets

Moreover it may not be possible to pro t from the tendency for individualstocks to exhibit return reversals Fluck Malkiel and Quandt (1997) simulated astrategy of buying stocks over a 13-year period during the 1980s and early 1990s thathad particularly poor returns over the past three to ve years They found thatstocks with very low returns over the past three to ve years had higher returns inthe next period and that stocks with very high returns over the past three to veyears had lower returns in the next period Thus they con rmed the very strongstatistical evidence of return reversals However they also found that returns in thenext period were similar for both groups so they could not con rm that acontrarian approach would yield higher-than-average returns There was a statisti-cally strong pattern of return reversal but not one that implied an inef ciency inthe market that would enable investors to make excess returns

Seasonal and Day-of-the-Week PatternsA number of researchers have found that January has been a very unusual

month for stock market returns Returns from an equally weighted stock index havetended to be unusually high during the rst two weeks of the year The returnpremium has been particularly evident for stocks with relatively small total capital-izations (Keim 1983) Haugen and Lakonishok (1988) documented the highJanuary returns in a book titled The Incredible January Effect There also appear to bea number of day-of-the-week effects For example French (1980) documentssigni cantly higher Monday returns There appear to be signi cant differences inaverage daily returns in countries other than the United States (Hawawini andKeim 1995) There also appear to be some patterns in returns around the turn ofthe month (Lakonishok and Smidt 1988) as well as around holidays (Ariel 1990)

The general problem with these predictable patterns or anomalies however isthat they are not dependable from period to period Wall Street traders now jokethat the ldquoJanuary effectrdquo is more likely to occur on the previous ThanksgivingMoreover these nonrandom effects (even if they were dependable) are very smallrelative to the transactions costs involved in trying to exploit them They do notappear to offer arbitrage opportunities that would enable investors to make excessrisk adjusted returns

Predictable Patterns Based on Valuation Parameters

Considerable empirical research has been conducted to determine if futurestock returns can be predicted on the basis of initial valuation parameters It isclaimed that valuation ratios such as the price-earnings multiple or the dividendyield of the stock market as a whole have considerable predictive power Thissection examines the body of work based on time series analyses

64 Journal of Economic Perspectives

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

The way I put it in my book A Random Walk Down Wall Street rst published in1973 a blindfolded chimpanzee throwing darts at the Wall Street Journal could selecta portfolio that would do as well as the experts Of course the advice was notliterally to throw darts but instead to throw a towel over the stock pagesmdashthat isto buy a broad-based index fund that bought and held all the stocks in the marketand that charged very low expenses

By the start of the twenty- rst century the intellectual dominance of theef cient market hypothesis had become far less universal Many nancial econo-mists and statisticians began to believe that stock prices are at least partiallypredictable A new breed of economists emphasized psychological and behavioralelements of stock-price determination and they came to believe that future stockprices are somewhat predictable on the basis of past stock price patterns as well ascertain ldquofundamentalrdquo valuation metrics Moreover many of these economists wereeven making the far more controversial claim that these predictable patternsenable investors to earn excess risk adjusted rates of return

This paper examines the attacks on the ef cient market hypothesis and thebelief that stock prices are partially predictable While I make no attempt to presenta complete survey of the purported regularities or anomalies in the stock marketI will describe the major statistical ndings as well as their behavioral underpin-nings where relevant and also examine the relationship between predictability andef ciency I will also describe the major arguments of those who believe thatmarkets are often irrational by analyzing the ldquocrash of 1987rdquo the Internet ldquobubblerdquoof the n de siecle and other speci c irrationalities often mentioned by critics ofef ciency I conclude that our stock markets are far more ef cient and far lesspredictable than some recent academic papers would have us believe Moreoverthe evidence is overwhelming that whatever anomalous behavior of stock pricesmay exist it does not create a portfolio trading opportunity that enables investorsto earn extraordinary risk adjusted returns

At the outset it is important to make clear what I mean by the term ldquoef -ciencyrdquo I will use as a de nition of ef cient nancial markets that such markets donot allow investors to earn above-average returns without accepting above-averagerisks A well-known story tells of a nance professor and a student who come acrossa $100 bill lying on the ground As the student stops to pick it up the professor saysldquoDonrsquot bothermdashif it were really a $100 bill it wouldnrsquot be thererdquo The story wellillustrates what nancial economists usually mean when they say markets areef cient Markets can be ef cient in this sense even if they sometimes make errorsin valuation as was certainly true during the 1999ndash early 2000 Internet ldquobubblerdquoMarkets can be ef cient even if many market participants are quite irrationalMarkets can be ef cient even if stock prices exhibit greater volatility than canapparently be explained by fundamentals such as earnings and dividends Many ofus economists who believe in ef ciency do so because we view markets as amazinglysuccessful devices for re ecting new information rapidly and for the most partaccurately Above all we believe that nancial markets are ef cient because theydonrsquot allow investors to earn above-average risk adjusted returns In short we

60 Journal of Economic Perspectives

believe that $100 bills are not lying around for the taking either by the professionalor the amateur investor

What I do not argue is that the market pricing is always perfect After the factwe know that markets have made egregious mistakes as I think occurred during therecent Internet ldquobubblerdquo Nor do I deny that psychological factors in uencesecurities prices But I am convinced that Benjamin Graham (1965) was correct insuggesting that while the stock market in the short run may be a voting mechanismin the long run it is a weighing mechanism True value will win out in the endBefore the fact there is no way in which investors can reliably exploit any anomaliesor patterns that might exist I am skeptical that any of the ldquopredictable patternsrdquothat have been documented in the literature were ever suf ciently robust so as tohave created pro table investment opportunities and after they have been discov-ered and publicized they will certainly not allow investors to earn excess returns

A Nonrandom Walk Down Wall Street

In this section I review some of the patterns of possible predictability sug-gested by studies of the behavior of past stock prices

Short-Term Momentum Including Underreaction to New InformationThe original empirical work supporting the notion of randomness in stock

prices looked at measures of short-run serial correlations between successive stockprice changes In general this work supported the view that the stock market hasno memorymdashthat is the way a stock price behaved in the past is not useful indivining how it will behave in the future for example see the survey of articlescontained in Cootner (1964) More recent work by Lo and MacKinlay (1999) ndsthat short-run serial correlations are not zero and that the existence of ldquotoo manyrdquosuccessive moves in the same direction enable them to reject the hypothesis thatstock prices behave as true random walks There does seem to be some momentumin short-run stock prices Moreover Lo Mamaysky and Wang (2000) also ndthrough the use of sophisticated nonparametric statistical techniques that canrecognize patterns some of the stock price signals used by ldquotechnical analystsrdquo suchas ldquohead and shouldersrdquo formations and ldquodouble bottomsrdquo may actually have somemodest predictive power

Economists and psychologists in the eld of behavioral nance nd suchshort-run momentum to be consistent with psychological feedback mechanismsIndividuals see a stock price rising and are drawn into the market in a kind ofldquobandwagon effectrdquo For example Shiller (2000) describes the rise in the US stockmarket during the late 1990s as the result of psychological contagion leading toirrational exuberance The behavioralists offered another explanation for patternsof short-run momentummdasha tendency for investors to underreact to new informa-tion If the full impact of an important news announcement is only grasped over aperiod of time stock prices will exhibit the positive serial correlation found by

Burton G Malkiel 61

investigators As behavioral nance became more prominent as a branch of thestudy of nancial markets momentum as opposed to randomness seemed rea-sonable to many investigators

However several factors should prevent us from interpreting the empiricalresults reported above as an indication that markets are inef cient First while thestock market may not be a mathematically perfect random walk it is important todistinguish statistical signi cance from economic signi cance The statistical de-pendencies giving rise to momentum are extremely small and are not likely topermit investors to realize excess returns Anyone who pays transactions costs isunlikely to fashion a trading strategy based on the kinds of momentum found inthese studies that will beat a buy-and-hold strategy Indeed Odean (1999) suggeststhat momentum investors do not realize excess returns Quite the oppositemdashasample of such investors suggests that such traders did far worse than buy-and-holdinvestors even during a period where there was clear statistical evidence of positivemomentum This is because of the large transactions costs involved in attemptingto exploit whatever momentum exists Similarly Lesmond Schill and Zhou (2001) nd that standard ldquorelative strengthrdquo strategies are not pro table because of thetrading costs involved in their execution

Second while behavioral hypotheses about bandwagon effects and under-reaction to new information may sound plausible enough the evidence that sucheffects occur systematically in the stock market is often rather thin For exampleEugene Fama (1998) surveys the considerable body of empirical work on ldquoeventstudiesrdquo that seeks to determine if stock prices respond ef ciently to informationThe ldquoeventsrdquo include such announcements as earnings surprises stock splits divi-dend actions mergers new exchange listings and initial public offerings Fama nds that apparent underreaction to information is about as common as over-reaction and postevent continuation of abnormal returns is as frequent aspostevent reversals He also shows that many of the return ldquoanomaliesrdquo arise only inthe context of some very particular model and that the results tend to disappearwhen exposed to different models for expected ldquonormalrdquo returns different meth-ods to adjust for risk and when different statistical approaches are used to measurethem For example a study that gives equal weight to postannouncement returns ofmany stocks can produce different results from a study that weights the stocksaccording to their value Certainly whatever momentum displayed by stock pricesdoes not appear to offer investors a dependable way to earn abnormal returns

The key factor is whether any patterns of serial correlation are consistent overtime Momentum strategies which refer to buying stocks that display positive serialcorrelation andor positive relative strength appeared to produce positive relativereturns during some periods of the late 1990s but highly negative relative returnsduring 2000 It is far from clear that any stock price patterns are useful for investorsin fashioning an investment strategy that will dependably earn excess returns

Many predictable patterns seem to disappear after they are published in the nance literature Schwert (2001) points out two possible explanations for such apattern One explanation may be that researchers are always sifting through

62 Journal of Economic Perspectives

mountains of nancial data Their normal tendency is to focus on results thatchallenge perceived wisdom and every now and again a combination of a certainsample and a certain technique will produce a statistically signi cant result thatseems to challenge the ef cient markets hypothesis Alternatively perhaps practi-tioners learn quickly about any true predictable pattern and exploit it to the extentthat it becomes no longer pro table My own view is that such apparent patternswere never suf ciently large or stable to guarantee consistently superior investmentresults and certainly such patterns will never be useful for investors after they havereceived considerable publicity The so-called ldquoJanuary effectrdquo for example inwhich stock prices rose in early January seems to have disappeared soon after it wasdiscovered

Long-Run Return ReversalsIn the short-run when stock returns are measured over periods of days or

weeks the usual argument against market ef ciency is that some positive serialcorrelation exists But many studies have shown evidence of negative serialcorrelationmdashthat is return reversalsmdash over longer holding periods For exampleFama and French (1988) found that 25 to 40 percent of the variation in longholding period returns can be predicted in terms of a negative correlation with pastreturns Similarly Poterba and Summers (1988) found substantial mean reversionin stock market returns at longer horizons

Some studies have attributed this forecastability to the tendency of stockmarket prices to ldquooverreactrdquo DeBondt and Thaler (1985) for example argue thatinvestors are subject to waves of optimism and pessimism that cause prices todeviate systematically from their fundamental values and later to exhibit meanreversion They suggest that such overreaction to past events is consistent with thebehavioral decision theory of Kahneman and Tversky (1979) where investors aresystematically overcon dent in their ability to forecast either future stock prices orfuture corporate earnings These ndings give some support to investment tech-niques that rest on a ldquocontrarianrdquo strategy that is buying the stocks or groups ofstocks that have been out of favor for long periods of time and avoiding thosestocks that have had large run-ups over the last several years

There is indeed considerable support for long-run negative serial correla-tion in stock returns However the nding of mean reversion is not uniformacross studies and is quite a bit weaker in some periods than it is for other periodsIndeed the strongest empirical results come from periods including the GreatDepressionmdashwhich may be a time with patterns that do not generalize wellMoreover such return reversals for the market as a whole may be quite consistentwith the ef cient functioning of the market since they could result in part fromthe volatility of interest rates and the tendency of interest rates to be meanreverting Since stock returns must rise or fall to be competitive with bond returnsthere is a tendency when interest rates go up for prices of both bond and stocks togo down and as interest rates go down for prices of bonds and stocks to go up Ifinterest rates revert to the mean over time this pattern will tend to generate return

The Efcient Market Hypothesis and Its Critics 63

reversals or mean reversion in a way that is quite consistent with the ef cientfunctioning of markets

Moreover it may not be possible to pro t from the tendency for individualstocks to exhibit return reversals Fluck Malkiel and Quandt (1997) simulated astrategy of buying stocks over a 13-year period during the 1980s and early 1990s thathad particularly poor returns over the past three to ve years They found thatstocks with very low returns over the past three to ve years had higher returns inthe next period and that stocks with very high returns over the past three to veyears had lower returns in the next period Thus they con rmed the very strongstatistical evidence of return reversals However they also found that returns in thenext period were similar for both groups so they could not con rm that acontrarian approach would yield higher-than-average returns There was a statisti-cally strong pattern of return reversal but not one that implied an inef ciency inthe market that would enable investors to make excess returns

Seasonal and Day-of-the-Week PatternsA number of researchers have found that January has been a very unusual

month for stock market returns Returns from an equally weighted stock index havetended to be unusually high during the rst two weeks of the year The returnpremium has been particularly evident for stocks with relatively small total capital-izations (Keim 1983) Haugen and Lakonishok (1988) documented the highJanuary returns in a book titled The Incredible January Effect There also appear to bea number of day-of-the-week effects For example French (1980) documentssigni cantly higher Monday returns There appear to be signi cant differences inaverage daily returns in countries other than the United States (Hawawini andKeim 1995) There also appear to be some patterns in returns around the turn ofthe month (Lakonishok and Smidt 1988) as well as around holidays (Ariel 1990)

The general problem with these predictable patterns or anomalies however isthat they are not dependable from period to period Wall Street traders now jokethat the ldquoJanuary effectrdquo is more likely to occur on the previous ThanksgivingMoreover these nonrandom effects (even if they were dependable) are very smallrelative to the transactions costs involved in trying to exploit them They do notappear to offer arbitrage opportunities that would enable investors to make excessrisk adjusted returns

Predictable Patterns Based on Valuation Parameters

Considerable empirical research has been conducted to determine if futurestock returns can be predicted on the basis of initial valuation parameters It isclaimed that valuation ratios such as the price-earnings multiple or the dividendyield of the stock market as a whole have considerable predictive power Thissection examines the body of work based on time series analyses

64 Journal of Economic Perspectives

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

believe that $100 bills are not lying around for the taking either by the professionalor the amateur investor

What I do not argue is that the market pricing is always perfect After the factwe know that markets have made egregious mistakes as I think occurred during therecent Internet ldquobubblerdquo Nor do I deny that psychological factors in uencesecurities prices But I am convinced that Benjamin Graham (1965) was correct insuggesting that while the stock market in the short run may be a voting mechanismin the long run it is a weighing mechanism True value will win out in the endBefore the fact there is no way in which investors can reliably exploit any anomaliesor patterns that might exist I am skeptical that any of the ldquopredictable patternsrdquothat have been documented in the literature were ever suf ciently robust so as tohave created pro table investment opportunities and after they have been discov-ered and publicized they will certainly not allow investors to earn excess returns

A Nonrandom Walk Down Wall Street

In this section I review some of the patterns of possible predictability sug-gested by studies of the behavior of past stock prices

Short-Term Momentum Including Underreaction to New InformationThe original empirical work supporting the notion of randomness in stock

prices looked at measures of short-run serial correlations between successive stockprice changes In general this work supported the view that the stock market hasno memorymdashthat is the way a stock price behaved in the past is not useful indivining how it will behave in the future for example see the survey of articlescontained in Cootner (1964) More recent work by Lo and MacKinlay (1999) ndsthat short-run serial correlations are not zero and that the existence of ldquotoo manyrdquosuccessive moves in the same direction enable them to reject the hypothesis thatstock prices behave as true random walks There does seem to be some momentumin short-run stock prices Moreover Lo Mamaysky and Wang (2000) also ndthrough the use of sophisticated nonparametric statistical techniques that canrecognize patterns some of the stock price signals used by ldquotechnical analystsrdquo suchas ldquohead and shouldersrdquo formations and ldquodouble bottomsrdquo may actually have somemodest predictive power

Economists and psychologists in the eld of behavioral nance nd suchshort-run momentum to be consistent with psychological feedback mechanismsIndividuals see a stock price rising and are drawn into the market in a kind ofldquobandwagon effectrdquo For example Shiller (2000) describes the rise in the US stockmarket during the late 1990s as the result of psychological contagion leading toirrational exuberance The behavioralists offered another explanation for patternsof short-run momentummdasha tendency for investors to underreact to new informa-tion If the full impact of an important news announcement is only grasped over aperiod of time stock prices will exhibit the positive serial correlation found by

Burton G Malkiel 61

investigators As behavioral nance became more prominent as a branch of thestudy of nancial markets momentum as opposed to randomness seemed rea-sonable to many investigators

However several factors should prevent us from interpreting the empiricalresults reported above as an indication that markets are inef cient First while thestock market may not be a mathematically perfect random walk it is important todistinguish statistical signi cance from economic signi cance The statistical de-pendencies giving rise to momentum are extremely small and are not likely topermit investors to realize excess returns Anyone who pays transactions costs isunlikely to fashion a trading strategy based on the kinds of momentum found inthese studies that will beat a buy-and-hold strategy Indeed Odean (1999) suggeststhat momentum investors do not realize excess returns Quite the oppositemdashasample of such investors suggests that such traders did far worse than buy-and-holdinvestors even during a period where there was clear statistical evidence of positivemomentum This is because of the large transactions costs involved in attemptingto exploit whatever momentum exists Similarly Lesmond Schill and Zhou (2001) nd that standard ldquorelative strengthrdquo strategies are not pro table because of thetrading costs involved in their execution

Second while behavioral hypotheses about bandwagon effects and under-reaction to new information may sound plausible enough the evidence that sucheffects occur systematically in the stock market is often rather thin For exampleEugene Fama (1998) surveys the considerable body of empirical work on ldquoeventstudiesrdquo that seeks to determine if stock prices respond ef ciently to informationThe ldquoeventsrdquo include such announcements as earnings surprises stock splits divi-dend actions mergers new exchange listings and initial public offerings Fama nds that apparent underreaction to information is about as common as over-reaction and postevent continuation of abnormal returns is as frequent aspostevent reversals He also shows that many of the return ldquoanomaliesrdquo arise only inthe context of some very particular model and that the results tend to disappearwhen exposed to different models for expected ldquonormalrdquo returns different meth-ods to adjust for risk and when different statistical approaches are used to measurethem For example a study that gives equal weight to postannouncement returns ofmany stocks can produce different results from a study that weights the stocksaccording to their value Certainly whatever momentum displayed by stock pricesdoes not appear to offer investors a dependable way to earn abnormal returns

The key factor is whether any patterns of serial correlation are consistent overtime Momentum strategies which refer to buying stocks that display positive serialcorrelation andor positive relative strength appeared to produce positive relativereturns during some periods of the late 1990s but highly negative relative returnsduring 2000 It is far from clear that any stock price patterns are useful for investorsin fashioning an investment strategy that will dependably earn excess returns

Many predictable patterns seem to disappear after they are published in the nance literature Schwert (2001) points out two possible explanations for such apattern One explanation may be that researchers are always sifting through

62 Journal of Economic Perspectives

mountains of nancial data Their normal tendency is to focus on results thatchallenge perceived wisdom and every now and again a combination of a certainsample and a certain technique will produce a statistically signi cant result thatseems to challenge the ef cient markets hypothesis Alternatively perhaps practi-tioners learn quickly about any true predictable pattern and exploit it to the extentthat it becomes no longer pro table My own view is that such apparent patternswere never suf ciently large or stable to guarantee consistently superior investmentresults and certainly such patterns will never be useful for investors after they havereceived considerable publicity The so-called ldquoJanuary effectrdquo for example inwhich stock prices rose in early January seems to have disappeared soon after it wasdiscovered

Long-Run Return ReversalsIn the short-run when stock returns are measured over periods of days or

weeks the usual argument against market ef ciency is that some positive serialcorrelation exists But many studies have shown evidence of negative serialcorrelationmdashthat is return reversalsmdash over longer holding periods For exampleFama and French (1988) found that 25 to 40 percent of the variation in longholding period returns can be predicted in terms of a negative correlation with pastreturns Similarly Poterba and Summers (1988) found substantial mean reversionin stock market returns at longer horizons

Some studies have attributed this forecastability to the tendency of stockmarket prices to ldquooverreactrdquo DeBondt and Thaler (1985) for example argue thatinvestors are subject to waves of optimism and pessimism that cause prices todeviate systematically from their fundamental values and later to exhibit meanreversion They suggest that such overreaction to past events is consistent with thebehavioral decision theory of Kahneman and Tversky (1979) where investors aresystematically overcon dent in their ability to forecast either future stock prices orfuture corporate earnings These ndings give some support to investment tech-niques that rest on a ldquocontrarianrdquo strategy that is buying the stocks or groups ofstocks that have been out of favor for long periods of time and avoiding thosestocks that have had large run-ups over the last several years

There is indeed considerable support for long-run negative serial correla-tion in stock returns However the nding of mean reversion is not uniformacross studies and is quite a bit weaker in some periods than it is for other periodsIndeed the strongest empirical results come from periods including the GreatDepressionmdashwhich may be a time with patterns that do not generalize wellMoreover such return reversals for the market as a whole may be quite consistentwith the ef cient functioning of the market since they could result in part fromthe volatility of interest rates and the tendency of interest rates to be meanreverting Since stock returns must rise or fall to be competitive with bond returnsthere is a tendency when interest rates go up for prices of both bond and stocks togo down and as interest rates go down for prices of bonds and stocks to go up Ifinterest rates revert to the mean over time this pattern will tend to generate return

The Efcient Market Hypothesis and Its Critics 63

reversals or mean reversion in a way that is quite consistent with the ef cientfunctioning of markets

Moreover it may not be possible to pro t from the tendency for individualstocks to exhibit return reversals Fluck Malkiel and Quandt (1997) simulated astrategy of buying stocks over a 13-year period during the 1980s and early 1990s thathad particularly poor returns over the past three to ve years They found thatstocks with very low returns over the past three to ve years had higher returns inthe next period and that stocks with very high returns over the past three to veyears had lower returns in the next period Thus they con rmed the very strongstatistical evidence of return reversals However they also found that returns in thenext period were similar for both groups so they could not con rm that acontrarian approach would yield higher-than-average returns There was a statisti-cally strong pattern of return reversal but not one that implied an inef ciency inthe market that would enable investors to make excess returns

Seasonal and Day-of-the-Week PatternsA number of researchers have found that January has been a very unusual

month for stock market returns Returns from an equally weighted stock index havetended to be unusually high during the rst two weeks of the year The returnpremium has been particularly evident for stocks with relatively small total capital-izations (Keim 1983) Haugen and Lakonishok (1988) documented the highJanuary returns in a book titled The Incredible January Effect There also appear to bea number of day-of-the-week effects For example French (1980) documentssigni cantly higher Monday returns There appear to be signi cant differences inaverage daily returns in countries other than the United States (Hawawini andKeim 1995) There also appear to be some patterns in returns around the turn ofthe month (Lakonishok and Smidt 1988) as well as around holidays (Ariel 1990)

The general problem with these predictable patterns or anomalies however isthat they are not dependable from period to period Wall Street traders now jokethat the ldquoJanuary effectrdquo is more likely to occur on the previous ThanksgivingMoreover these nonrandom effects (even if they were dependable) are very smallrelative to the transactions costs involved in trying to exploit them They do notappear to offer arbitrage opportunities that would enable investors to make excessrisk adjusted returns

Predictable Patterns Based on Valuation Parameters

Considerable empirical research has been conducted to determine if futurestock returns can be predicted on the basis of initial valuation parameters It isclaimed that valuation ratios such as the price-earnings multiple or the dividendyield of the stock market as a whole have considerable predictive power Thissection examines the body of work based on time series analyses

64 Journal of Economic Perspectives

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

investigators As behavioral nance became more prominent as a branch of thestudy of nancial markets momentum as opposed to randomness seemed rea-sonable to many investigators

However several factors should prevent us from interpreting the empiricalresults reported above as an indication that markets are inef cient First while thestock market may not be a mathematically perfect random walk it is important todistinguish statistical signi cance from economic signi cance The statistical de-pendencies giving rise to momentum are extremely small and are not likely topermit investors to realize excess returns Anyone who pays transactions costs isunlikely to fashion a trading strategy based on the kinds of momentum found inthese studies that will beat a buy-and-hold strategy Indeed Odean (1999) suggeststhat momentum investors do not realize excess returns Quite the oppositemdashasample of such investors suggests that such traders did far worse than buy-and-holdinvestors even during a period where there was clear statistical evidence of positivemomentum This is because of the large transactions costs involved in attemptingto exploit whatever momentum exists Similarly Lesmond Schill and Zhou (2001) nd that standard ldquorelative strengthrdquo strategies are not pro table because of thetrading costs involved in their execution

Second while behavioral hypotheses about bandwagon effects and under-reaction to new information may sound plausible enough the evidence that sucheffects occur systematically in the stock market is often rather thin For exampleEugene Fama (1998) surveys the considerable body of empirical work on ldquoeventstudiesrdquo that seeks to determine if stock prices respond ef ciently to informationThe ldquoeventsrdquo include such announcements as earnings surprises stock splits divi-dend actions mergers new exchange listings and initial public offerings Fama nds that apparent underreaction to information is about as common as over-reaction and postevent continuation of abnormal returns is as frequent aspostevent reversals He also shows that many of the return ldquoanomaliesrdquo arise only inthe context of some very particular model and that the results tend to disappearwhen exposed to different models for expected ldquonormalrdquo returns different meth-ods to adjust for risk and when different statistical approaches are used to measurethem For example a study that gives equal weight to postannouncement returns ofmany stocks can produce different results from a study that weights the stocksaccording to their value Certainly whatever momentum displayed by stock pricesdoes not appear to offer investors a dependable way to earn abnormal returns

The key factor is whether any patterns of serial correlation are consistent overtime Momentum strategies which refer to buying stocks that display positive serialcorrelation andor positive relative strength appeared to produce positive relativereturns during some periods of the late 1990s but highly negative relative returnsduring 2000 It is far from clear that any stock price patterns are useful for investorsin fashioning an investment strategy that will dependably earn excess returns

Many predictable patterns seem to disappear after they are published in the nance literature Schwert (2001) points out two possible explanations for such apattern One explanation may be that researchers are always sifting through

62 Journal of Economic Perspectives

mountains of nancial data Their normal tendency is to focus on results thatchallenge perceived wisdom and every now and again a combination of a certainsample and a certain technique will produce a statistically signi cant result thatseems to challenge the ef cient markets hypothesis Alternatively perhaps practi-tioners learn quickly about any true predictable pattern and exploit it to the extentthat it becomes no longer pro table My own view is that such apparent patternswere never suf ciently large or stable to guarantee consistently superior investmentresults and certainly such patterns will never be useful for investors after they havereceived considerable publicity The so-called ldquoJanuary effectrdquo for example inwhich stock prices rose in early January seems to have disappeared soon after it wasdiscovered

Long-Run Return ReversalsIn the short-run when stock returns are measured over periods of days or

weeks the usual argument against market ef ciency is that some positive serialcorrelation exists But many studies have shown evidence of negative serialcorrelationmdashthat is return reversalsmdash over longer holding periods For exampleFama and French (1988) found that 25 to 40 percent of the variation in longholding period returns can be predicted in terms of a negative correlation with pastreturns Similarly Poterba and Summers (1988) found substantial mean reversionin stock market returns at longer horizons

Some studies have attributed this forecastability to the tendency of stockmarket prices to ldquooverreactrdquo DeBondt and Thaler (1985) for example argue thatinvestors are subject to waves of optimism and pessimism that cause prices todeviate systematically from their fundamental values and later to exhibit meanreversion They suggest that such overreaction to past events is consistent with thebehavioral decision theory of Kahneman and Tversky (1979) where investors aresystematically overcon dent in their ability to forecast either future stock prices orfuture corporate earnings These ndings give some support to investment tech-niques that rest on a ldquocontrarianrdquo strategy that is buying the stocks or groups ofstocks that have been out of favor for long periods of time and avoiding thosestocks that have had large run-ups over the last several years

There is indeed considerable support for long-run negative serial correla-tion in stock returns However the nding of mean reversion is not uniformacross studies and is quite a bit weaker in some periods than it is for other periodsIndeed the strongest empirical results come from periods including the GreatDepressionmdashwhich may be a time with patterns that do not generalize wellMoreover such return reversals for the market as a whole may be quite consistentwith the ef cient functioning of the market since they could result in part fromthe volatility of interest rates and the tendency of interest rates to be meanreverting Since stock returns must rise or fall to be competitive with bond returnsthere is a tendency when interest rates go up for prices of both bond and stocks togo down and as interest rates go down for prices of bonds and stocks to go up Ifinterest rates revert to the mean over time this pattern will tend to generate return

The Efcient Market Hypothesis and Its Critics 63

reversals or mean reversion in a way that is quite consistent with the ef cientfunctioning of markets

Moreover it may not be possible to pro t from the tendency for individualstocks to exhibit return reversals Fluck Malkiel and Quandt (1997) simulated astrategy of buying stocks over a 13-year period during the 1980s and early 1990s thathad particularly poor returns over the past three to ve years They found thatstocks with very low returns over the past three to ve years had higher returns inthe next period and that stocks with very high returns over the past three to veyears had lower returns in the next period Thus they con rmed the very strongstatistical evidence of return reversals However they also found that returns in thenext period were similar for both groups so they could not con rm that acontrarian approach would yield higher-than-average returns There was a statisti-cally strong pattern of return reversal but not one that implied an inef ciency inthe market that would enable investors to make excess returns

Seasonal and Day-of-the-Week PatternsA number of researchers have found that January has been a very unusual

month for stock market returns Returns from an equally weighted stock index havetended to be unusually high during the rst two weeks of the year The returnpremium has been particularly evident for stocks with relatively small total capital-izations (Keim 1983) Haugen and Lakonishok (1988) documented the highJanuary returns in a book titled The Incredible January Effect There also appear to bea number of day-of-the-week effects For example French (1980) documentssigni cantly higher Monday returns There appear to be signi cant differences inaverage daily returns in countries other than the United States (Hawawini andKeim 1995) There also appear to be some patterns in returns around the turn ofthe month (Lakonishok and Smidt 1988) as well as around holidays (Ariel 1990)

The general problem with these predictable patterns or anomalies however isthat they are not dependable from period to period Wall Street traders now jokethat the ldquoJanuary effectrdquo is more likely to occur on the previous ThanksgivingMoreover these nonrandom effects (even if they were dependable) are very smallrelative to the transactions costs involved in trying to exploit them They do notappear to offer arbitrage opportunities that would enable investors to make excessrisk adjusted returns

Predictable Patterns Based on Valuation Parameters

Considerable empirical research has been conducted to determine if futurestock returns can be predicted on the basis of initial valuation parameters It isclaimed that valuation ratios such as the price-earnings multiple or the dividendyield of the stock market as a whole have considerable predictive power Thissection examines the body of work based on time series analyses

64 Journal of Economic Perspectives

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

mountains of nancial data Their normal tendency is to focus on results thatchallenge perceived wisdom and every now and again a combination of a certainsample and a certain technique will produce a statistically signi cant result thatseems to challenge the ef cient markets hypothesis Alternatively perhaps practi-tioners learn quickly about any true predictable pattern and exploit it to the extentthat it becomes no longer pro table My own view is that such apparent patternswere never suf ciently large or stable to guarantee consistently superior investmentresults and certainly such patterns will never be useful for investors after they havereceived considerable publicity The so-called ldquoJanuary effectrdquo for example inwhich stock prices rose in early January seems to have disappeared soon after it wasdiscovered

Long-Run Return ReversalsIn the short-run when stock returns are measured over periods of days or

weeks the usual argument against market ef ciency is that some positive serialcorrelation exists But many studies have shown evidence of negative serialcorrelationmdashthat is return reversalsmdash over longer holding periods For exampleFama and French (1988) found that 25 to 40 percent of the variation in longholding period returns can be predicted in terms of a negative correlation with pastreturns Similarly Poterba and Summers (1988) found substantial mean reversionin stock market returns at longer horizons

Some studies have attributed this forecastability to the tendency of stockmarket prices to ldquooverreactrdquo DeBondt and Thaler (1985) for example argue thatinvestors are subject to waves of optimism and pessimism that cause prices todeviate systematically from their fundamental values and later to exhibit meanreversion They suggest that such overreaction to past events is consistent with thebehavioral decision theory of Kahneman and Tversky (1979) where investors aresystematically overcon dent in their ability to forecast either future stock prices orfuture corporate earnings These ndings give some support to investment tech-niques that rest on a ldquocontrarianrdquo strategy that is buying the stocks or groups ofstocks that have been out of favor for long periods of time and avoiding thosestocks that have had large run-ups over the last several years

There is indeed considerable support for long-run negative serial correla-tion in stock returns However the nding of mean reversion is not uniformacross studies and is quite a bit weaker in some periods than it is for other periodsIndeed the strongest empirical results come from periods including the GreatDepressionmdashwhich may be a time with patterns that do not generalize wellMoreover such return reversals for the market as a whole may be quite consistentwith the ef cient functioning of the market since they could result in part fromthe volatility of interest rates and the tendency of interest rates to be meanreverting Since stock returns must rise or fall to be competitive with bond returnsthere is a tendency when interest rates go up for prices of both bond and stocks togo down and as interest rates go down for prices of bonds and stocks to go up Ifinterest rates revert to the mean over time this pattern will tend to generate return

The Efcient Market Hypothesis and Its Critics 63

reversals or mean reversion in a way that is quite consistent with the ef cientfunctioning of markets

Moreover it may not be possible to pro t from the tendency for individualstocks to exhibit return reversals Fluck Malkiel and Quandt (1997) simulated astrategy of buying stocks over a 13-year period during the 1980s and early 1990s thathad particularly poor returns over the past three to ve years They found thatstocks with very low returns over the past three to ve years had higher returns inthe next period and that stocks with very high returns over the past three to veyears had lower returns in the next period Thus they con rmed the very strongstatistical evidence of return reversals However they also found that returns in thenext period were similar for both groups so they could not con rm that acontrarian approach would yield higher-than-average returns There was a statisti-cally strong pattern of return reversal but not one that implied an inef ciency inthe market that would enable investors to make excess returns

Seasonal and Day-of-the-Week PatternsA number of researchers have found that January has been a very unusual

month for stock market returns Returns from an equally weighted stock index havetended to be unusually high during the rst two weeks of the year The returnpremium has been particularly evident for stocks with relatively small total capital-izations (Keim 1983) Haugen and Lakonishok (1988) documented the highJanuary returns in a book titled The Incredible January Effect There also appear to bea number of day-of-the-week effects For example French (1980) documentssigni cantly higher Monday returns There appear to be signi cant differences inaverage daily returns in countries other than the United States (Hawawini andKeim 1995) There also appear to be some patterns in returns around the turn ofthe month (Lakonishok and Smidt 1988) as well as around holidays (Ariel 1990)

The general problem with these predictable patterns or anomalies however isthat they are not dependable from period to period Wall Street traders now jokethat the ldquoJanuary effectrdquo is more likely to occur on the previous ThanksgivingMoreover these nonrandom effects (even if they were dependable) are very smallrelative to the transactions costs involved in trying to exploit them They do notappear to offer arbitrage opportunities that would enable investors to make excessrisk adjusted returns

Predictable Patterns Based on Valuation Parameters

Considerable empirical research has been conducted to determine if futurestock returns can be predicted on the basis of initial valuation parameters It isclaimed that valuation ratios such as the price-earnings multiple or the dividendyield of the stock market as a whole have considerable predictive power Thissection examines the body of work based on time series analyses

64 Journal of Economic Perspectives

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

reversals or mean reversion in a way that is quite consistent with the ef cientfunctioning of markets

Moreover it may not be possible to pro t from the tendency for individualstocks to exhibit return reversals Fluck Malkiel and Quandt (1997) simulated astrategy of buying stocks over a 13-year period during the 1980s and early 1990s thathad particularly poor returns over the past three to ve years They found thatstocks with very low returns over the past three to ve years had higher returns inthe next period and that stocks with very high returns over the past three to veyears had lower returns in the next period Thus they con rmed the very strongstatistical evidence of return reversals However they also found that returns in thenext period were similar for both groups so they could not con rm that acontrarian approach would yield higher-than-average returns There was a statisti-cally strong pattern of return reversal but not one that implied an inef ciency inthe market that would enable investors to make excess returns

Seasonal and Day-of-the-Week PatternsA number of researchers have found that January has been a very unusual

month for stock market returns Returns from an equally weighted stock index havetended to be unusually high during the rst two weeks of the year The returnpremium has been particularly evident for stocks with relatively small total capital-izations (Keim 1983) Haugen and Lakonishok (1988) documented the highJanuary returns in a book titled The Incredible January Effect There also appear to bea number of day-of-the-week effects For example French (1980) documentssigni cantly higher Monday returns There appear to be signi cant differences inaverage daily returns in countries other than the United States (Hawawini andKeim 1995) There also appear to be some patterns in returns around the turn ofthe month (Lakonishok and Smidt 1988) as well as around holidays (Ariel 1990)

The general problem with these predictable patterns or anomalies however isthat they are not dependable from period to period Wall Street traders now jokethat the ldquoJanuary effectrdquo is more likely to occur on the previous ThanksgivingMoreover these nonrandom effects (even if they were dependable) are very smallrelative to the transactions costs involved in trying to exploit them They do notappear to offer arbitrage opportunities that would enable investors to make excessrisk adjusted returns

Predictable Patterns Based on Valuation Parameters

Considerable empirical research has been conducted to determine if futurestock returns can be predicted on the basis of initial valuation parameters It isclaimed that valuation ratios such as the price-earnings multiple or the dividendyield of the stock market as a whole have considerable predictive power Thissection examines the body of work based on time series analyses

64 Journal of Economic Perspectives

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

Predicting Future Returns from Initial Dividend YieldsFormal statistical tests of the ability of dividend yields (that is the ratio of

dividend to stock price) to forecast future returns have been conducted by Famaand French (1988) and Campbell and Shiller (1988) Depending on the forecasthorizon involved as much as 40 percent of the variance of future returns for thestock market as a whole can be predicted on the basis of the initial dividend yieldof the market index

An interesting way of presenting the results is shown in the top panel of Exhibit1 The exhibit was produced by measuring the dividend yield of the broad USstock market Standard amp Poorrsquos 500 Stock Index each quarter since 1926 and thencalculating the marketrsquos subsequent ten-year total return through the year 2001The observations were then divided into deciles depending upon the level of theinitial dividend yield In general the exhibit shows that investors have earned ahigher rate of return from the stock market when they purchased a market basketof equities with an initial dividend yield that was relatively high and relatively lowfuture rates of return when stocks were purchased at low dividend yields

These ndings are not necessarily inconsistent with ef ciency Dividend yieldsof stocks tend to be high when interest rates are high and they tend to be low wheninterest rates are low Consequently the ability of initial yields to predict returnsmay simply re ect the adjustment of the stock market to general economic condi-tions Moreover the use of dividend yields to predict future returns has beenineffective since the mid-1980s Dividend yields have been at the 3 percent level orbelow continuously since the mid-1980s indicating very low forecasted returns Infact for all ten-year periods from 1985 through 1992 that ended June 30 2002realized annual equity returns from the market index have averaged over15 percent One possible explanation is that the dividend behavior of US corpo-rations may have changed over time (Bagwell and Shoven 1989 Fama and French2001) Companies in the twenty- rst century may be more likely to institute a sharerepurchase program rather than increase their dividends Thus dividend yield maynot be as meaningful as in the past as a useful predictor of future equity returns

Finally it is worth noting that this phenomenon does not work consistently withindividual stocks (Fluck Malkiel and Quandt 1997) Investors who simply purchasea portfolio of individual stocks with the highest dividend yields in the market willnot earn a particularly high rate of return One popular implementation of such aldquohigh dividendrdquo strategy in the United States is the ldquoDogs of the Dow Strategyrdquowhich involves buying the ten stocks in the Dow Jones Industrial Average with thehighest dividend yields For some past periods this strategy handily outpaced theoverall average and so several ldquoDogs of the Dowrdquo mutual funds were brought tomarket and aggressively sold to individual investors However such funds generallyunderperformed the market averages during the 1995ndash1999 period

Predicting Market Returns from Initial Price-Earnings MultiplesThe same kind of predictability for the market as a whole as was demonstrated

for dividends has been shown for price-earnings ratios The data are shown in the

Burton G Malkiel 65

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

bottom half of Exhibit 1 The exhibit presents a decile analysis similar to thatdescribed for dividend yields above Investors have tended to earn larger long-horizon returns when purchasing the market basket of stocks at relatively lowprice-earnings multiples Campbell and Shiller (1998) report that initial PE ratios

Exhibit 1The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Dividend Yields (DP)

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

DP 6

1

DP 5 5

7

DP 5 5

1

DP 5 4

7

DP 5 43

DP 5 3

9

DP 5 3

6

DP 5 3

3

DP 5 30

DP 2

9

The Future 10-Year Rates of Return When Stocks are Purchased at AlternativeInitial Price-to-Earnings (PE) Multiples

0

2

4

6

8

10

12

14

Stocks Cheap Stocks Expensive

16

18Return ()

PE 9

9

PE 5 104

PE 5 115

PE 5 12

9

PE 5 14

4

PE 5 16

0

PE 5 174

PE 5 18

6

PE 5 20

1

PE 20

9

Source The Leuthold Group

66 Journal of Economic Perspectives

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

explained as much as 40 percent of the variance of future returns They concludethat equity returns have been predictable in the past to a considerable extent

Consider however the recent experience of investors who have attempted toundertake investment strategies based either on the level of the price-earningsmultiple or the dividend yield to predict future long-horizon returns Price-earnings multiples for the Standard amp Poorrsquos 500 stock index rose into the low 20son June 30 1987 (suggesting very low long-horizon returns) Dividend yields fellbelow 3 percent Price-earnings multiples rose into the low 20s The average annualtotal return from the index over the next 10 years was an extraordinarily generous167 percent Dividend yields again fell to 3 percent in June 1992 Price-earningsmultiples rose to the mid-20s The subsequent return through June 2002 was114 percent The yield of the index uctuated between 2 and 3 percent from 1993through 1995 and earnings multiples remained in the mid-20s yet long-horizonreturns through June 30 2002 uctuated between 11 and 12 percent Even fromearly December 1996 the date of Campbell and Shillerrsquos presentation to theFederal Reserve suggesting near zero returns for the SampP500 the index providedalmost a 7 percent annual return through mid-2002 Such results suggest to me avery cautious assessment of the extent to which stock market returns are predictableon this basis

Other Predictable Time Series PatternsStudies have found some amount of predictability of stock returns based on

various nancial statistics For example Fama and Schwert (1977) found thatshort-term interest rates were related to future stock returns Campbell (1987)found that term structure of interest rates spreads contained useful information forforecasting stock returns Keim and Stambaugh (1986) found that risk spreadsbetween high-yield corporate bonds and short rates had some predictive powerAgain even if some predictability exists it may re ect time varying risk premiumsand required rates of return for stock investors rather than an inef ciency More-over it is far from clear that any of these results can be used to generate pro tabletrading strategies Whether such historical statistical relations give investors reliableand useful guides to appropriate asset allocation is far from clear

Cross-Sectional Predictable Patterns Based on Firm Characteristicsand Valuation Parameters

A large number of patterns that are claimed to be predictable are based on rm characteristics and different valuation parameters

The Size EffectOne of the strongest effects investigators have found is the tendency over long

periods of time for smaller-company stocks to generate larger returns that those of

The Efcient Market Hypothesis and Its Critics 67

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

large-company stocks Since 1926 small-company stocks in the United States haveproduced annual rates of return over 1 percentage point larger than the returnsfrom large stocks (Keim 1983) Fama and French (1993) examined data from 1963to 1990 and divided all stocks into deciles according to their size as measured bytotal capitalization Decile one contained the smallest 10 percent of all stocks whiledecile ten contained the largest stocks The results plotted in Exhibit 2 show aclear tendency for the deciles made up of portfolios of smaller stocks to generatehigher average monthly returns than deciles made up of larger stocks

The crucial issue here is the extent to which the higher returns of smallcompanies represent a predictable pattern that will allow investors to generateexcess risk-adjusted returns According to the capital asset pricing model thecorrect measure of risk for a stock is its ldquobetardquomdashthat is the extent to which thereturn of the stock varies with the return for the market as a whole If the betameasure of systematic risk from the capital asset pricing model is accepted as thecorrect risk measurement statistic the size effect can be interpreted as indicating ananomaly and a market inef ciency because using this measure portfolios consistingof smaller stocks have excess risk-adjusted returns Fama and French (1993) pointout however that the average relationship between beta and return during the1963ndash1990 period was atmdashnot upward sloping as the capital asset pricing modelpredicts Moreover if stocks are divided up by beta deciles ten portfolios con-structed by size display the same kind of positive relationship shown in Exhibit 2On the other hand within size deciles the relationship between beta and returncontinues to be at Fama and French suggest that size may be a far better proxyfor risk than beta and therefore that their ndings should not be interpreted asindicating that markets are inef cient

The dependability of the size phenomenon is also open to question From themid-1980s through the decade of the 1990s there has been no gain from holdingsmaller stocks Indeed in most world markets larger capitalization stocks producedlarger rates of return It may be that the growing institutionalization of the marketled portfolio managers to prefer larger companies with more liquidity to smallercompanies where it would be dif cult to liquidate signi cant blocks of stockFinally it is also possible that some studies of the small- rm effect have beenaffected by survivorship bias Todayrsquos computerized databases of companies includeonly small rms that have survived not the ones that later went bankrupt Thus aresearcher who examined the ten-year performance of todayrsquos small companieswould be measuring the performance of those companies that survivedmdashnot theones that failed

Value StocksSeveral studies suggest that ldquovaluerdquo stocks have higher returns than so-called

ldquogrowthrdquo stocks The most common two methods of identifying value stocks havebeen price-earnings ratios and price-to-book-value ratios

Stocks with low price-earnings multiples (often called ldquovaluerdquo stocks) appear toprovide higher rates of return than stocks with high price-to-earnings ratios as rst

68 Journal of Economic Perspectives

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

shown by Nicholson (1960) and later con rmed by Ball (1978) and Basu (1983)This nding is consistent with the views of behavioralists that investors tend to beovercon dent of their ability to project high earnings growth and thus overpay forldquogrowthrdquo stocks (for example Kahneman and Riepe 1998) The nding is alsoconsistent with the views of Graham and Dodd (1934) rst expounded in theirclassic book on security analysis and later championed by the legendary USinvestor Warren Buffett Similar results have been shown for pricecash owmultiples where cash ow is de ned as earnings plus depreciation and amortiza-tion (Hawawini and Keim 1995)

The ratio of stock price to book value de ned as the value of a rmrsquos assetsminus its liabilities divided by the number of shares outstanding has also beenfound to be a useful predictor of future returns Low price-to-book is considered tobe another hallmark of so-called ldquovaluerdquo in equity securities and is also consistentwith the view of behavioralists that investors tend to overpay for ldquogrowthrdquo stocks thatsubsequently fail to live up to expectations Fama and French (1993) concludedthat size and price-to-book-value together provide considerable explanatory powerfor future returns and once they are accounted for little additional in uence canbe attributed to PE multiples Fama and French (1997) also conclude that thePBV effect is important in many world stock markets other than the United States

Such results raise questions about the ef ciency of the market if one acceptsthe capital asset pricing model as Lakonishok Shleifer and Vishny (1994) pointout But these ndings do not necessarily imply inef ciency They may simplyindicate failure of the capital asset pricing model to capture all the dimensions of

Exhibit 2Average Monthly Returns for Portfolios Formed on the Basis of Size 1963ndash1990

Small Stocks

Return

8

1 2 3 4 5 6 7 8 9 10 Size

Large Stocks

9

10

11

12

13

14

15

16

Source Fama and French (1992)

Burton G Malkiel 69

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

risk For example Fama and French (1993) suggest that the price-to-book-valueratio may re ect another risk factor that is priced into the market and not capturedby the capital asset pricing model Companies in some degree of nancial distressfor example are likely to sell at low prices relative to book values Fama and French(1993) argue that a three-factor asset-pricing model (including price-to-book-valueand size as measures of risk) is the appropriate benchmark against which anomaliesshould be measured

We also need to keep in mind that the results of published studiesmdash even thosedone over decadesmdashmay still be time-dependent and ask whether the returnpatterns of academic studies can actually be generated with real money Exhibit 3presents average actual returns generated by mutual funds classi ed by either theirldquogrowthrdquo or ldquovaluerdquo objectives ldquoValuerdquo funds are so classi ed if they buy stocks withprice-to-earnings and price-to-book-value multiples that are below the averages forthe whole stock market Over a period running back to the 1930s it does notappear that investors could actually have realized higher rates of return frommutual funds specializing in ldquovaluerdquo stocks Indeed the exhibit suggests that theFama-French (1993) period from the early 1960s through 1990 may have been aunique period in which value stocks rather consistently produced higher rates ofreturn

Schwert (2001) points out that the investment rm of Dimensional FundAdvisors actually began a mutual fund that selected value stocks quantitativelyaccording to the Fama and French (1993) criteria The abnormal return of such aportfolio (adjusting for beta the capital asset pricing model measure of risk) was anegative 02 percent per month over the 1993ndash1998 period The absence duringthat period of an excess return to the ldquovaluerdquo stocks is consistent with the resultsfrom ldquoactively managedrdquo value mutual funds shown in Exhibit 3

The Equity Risk Premium PuzzleAnother puzzle that is often used to suggest that markets are less than fully

rational is the existence of a very large historical equity risk premium that seemsinconsistent with the actual riskiness of common stocks as can be measuredstatistically For example using the Ibbotson data on stock returns from 1926through 2001 common stocks have produced rates of retain of approximately105 percent while high-grade bonds have returned only about 55 percent Ibelieve that this nding is the result of a combination of perceived equity risk beingconsiderably higher during the early years of the period and of average equityreturns being much higher than had been forecast by investors

It is easy to say 50 to 75 years later that common stocks were underpricedduring the 1930s and 1940s But remember that the annual average of almost6 percent growth in corporate earnings and dividends that has occurred since 1926was hardly a foregone conclusion during a period of severe depression and worldwar Indeed the US stock market is almost unique in that it is one of the few worldmarkets that remained in continuous operation during the entire period and themeasured risk premium results in part from survivorship bias One must be very

70 Journal of Economic Perspectives

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

careful to distinguish between expected risk premiums and such premiums mea-sured after the fact Fama and French (2002) argue that the high average realizedreturns result in part from large unexpected capital gains Economists such as Shillerhave suggested that during the early 2000s the expected equity risk premium wasif anything irrationally too low

Summarizing the ldquoAnomaliesrdquo and Predictable PatternsThe preceding sections have pointed out many ldquoanomaliesrdquo and statistically

signi cant predictable patterns in the stock returns that have been uncovered inthe literature However these patterns are not robust and dependable in differentsample periods and some of the patterns based on fundamental valuation mea-sures of individual stocks may simply re ect better proxies for measuring risk

Moreover many of these patterns even if they did exist could self-destruct inthe future as many of them have already done Indeed this is the logical reasonwhy one should be cautious not to overemphasize these anomalies and predictablepatterns Suppose for example one of the anomalies or predictable patternsappears to be robust Suppose there is a truly dependable and exploitable Januaryeffect that the stock marketmdash especially stocks of small companiesmdashwill generateextraordinary returns during the rst ve days of January What will investors doThey will buy on the last day of December and sell on January 5 But then investors nd that the market rallied on the last day of December and so they will need tobegin to buy on the next-to-last day of December and because there is so muchldquopro t takingrdquo on January 5 investors will have to sell on January 4 to takeadvantage of this effect Thus to beat the gun investors will have to be buying

Exhibit 3Reversion to the Mean Relative Performance of ldquoValuerdquo vs ldquoGrowthrdquo MutualFunds 1937ndashJune 2002

07

08

09Growth

Outperforming

ValueOutperforming

Average Annual ReturnGrowthValue

10611057

10

11

06

37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 00

Rel

ativ

e R

etur

n

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PennsylvaniaNote The exhibit shows the cumulative value of one dollar invested in the average ldquovaluerdquo funddivided by the same statistic calculated for the average ldquogrowthrdquo fund

The Efcient Market Hypothesis and Its Critics 71

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

earlier and earlier in December and selling earlier and earlier in January so thateventually the pattern will self-destruct Any truly repetitive and exploitable patternthat can be discovered in the stock market and can be arbitraged away willself-destruct Indeed the January effect became undependable after it receivedconsiderable publicity

Similarly suppose there is a general tendency for stock prices to underreact tocertain new events leading to abnormal returns to investors who exploit the lack offull immediate adjustment (DeBondt and Thaler 1995 Campbell Lo and Mac-Kinlay 1977) ldquoQuantitativerdquo investment managers will then develop trading strat-egies to exploit the pattern Indeed the more potentially pro table a discoverablepattern is the less likely it is to survive

Many of the predictable patterns that have been discovered may simply be theresult of data mining The ease of experimenting with nancial databanks of almostevery conceivable dimension makes it quite likely that investigators will nd someseemingly signi cant but wholly spurious correlation between nancial variables oramong nancial and non nancial data sets Given enough time and massaging ofdata series it is possible to tease almost any pattern out of most data sets Moreoverthe published literature is likely to be biased in favor of reporting such resultsSigni cant effects are likely to be published in professional journals while negativeresults or boring con rmations of previous ndings are relegated to the ledrawer or discarded Data-mining problems are unique to nonexperimental sci-ences such as economics which rely on statistical analysis for their insights andcannot test hypotheses by running repeated controlled experiments

An exchange at a symposium about a decade ago between Robert Shiller aneconomist who is sympathetic to the argument that stock prices are partiallypredictable and skeptical about market ef ciency and Richard Roll an academic nancial economist who also is a portfolio manager is quite revealing (Roll andShiller 1992) After Shiller stressed the importance of inef ciencies in the pricingof stocks Roll responded as follows

I have personally tried to invest money my clientrsquos money and my own inevery single anomaly and predictive device that academics have dreamedup I have attempted to exploit the so-called year-end anomalies and awhole variety of strategies supposedly documented by academic research AndI have yet to make a nickel on any of these supposed market inefciencies a truemarket inefciency ought to be an exploitable opportunity If therersquos nothinginvestors can exploit in a systematic way time in and time out then itrsquos veryhard to say that information is not being properly incorporated into stockprices

Seemingly Irrefutable Cases of Inef ciency

Critics of ef ciency argue that there are several instances of recent markethistory where market prices could not plausibly have been set by rational investors

72 Journal of Economic Perspectives

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

and that psychological considerations must have played the dominant role It isalleged for example that the stock market lost about one-third of its value fromearly to mid-October 1987 with essentially no change in the general economicenvironment How could market prices be ef cient both at the start of October andduring the middle of the month Similarly it is widely believed that the pricing ofInternet stocks in early 2000 could only be explained by the behavior of irrationalinvestors Do such events make a belief in ef cient markets untenable

The Market Crash of October 1987Can the October 1987 market crash be explained by rational considerations or

does such a rapid and signi cant change in market valuations prove the dominanceof psychological rather than logical factors in understanding the stock marketBehaviorists would say that the one-third drop in market prices which occurredearly in October 1987 can only be explained by relying on psychological consid-erations since the basic elements of the valuation equation did not change rapidlyover that period It is of course impossible to rule out the existence of behavioralor psychological in uences on stock market pricing But logical considerations canexplain a sharp change in market valuations such as occurred during the rst weeksof October 1987

A number of factors could rationally have changed investorsrsquo views about theproper value of the stock market in October 1987 For one thing yields onlong-term Treasury bonds increased from about 9 percent to almost 105 percent inthe two months prior to mid-October Moreover a number of events may rationallyhave increased risk perceptions during the rst two weeks of October Early in themonth Congress threatened to impose a ldquomerger taxrdquo that would have mademerger activity prohibitively expensive and could well have ended the mergerboom The risk that merger activity might be curtailed increased risks throughoutthe stock market by weakening the discipline over corporate management thatpotential takeovers provide Also in early October 1987 then Secretary of theTreasury James Baker had threatened to encourage a further fall in the exchangevalue of the dollar increasing risks for foreign investors and frightening domesticinvestors as well While it is impossible to correlate each dayrsquos movement in stockprices to speci c news events it is not unreasonable to ascribe the sharp decline inmid-October to the cumulative effect of a number of unfavorable ldquofundamentalrdquoevents As Merton Miller (1991) has written ldquo on October 19 some weeks ofexternal events minor in themselves cumulatively signaled a possible change inwhat had been up to then a very favorable political and economic climate forequities and many investors simultaneously came to believe they were hold-ing too large a share of their wealth in risky equitiesrdquo

Share prices can be highly sensitive as a result of rational responses to smallchanges in interest rates and risk perceptions Suppose stocks are priced as thepresent value of the expected future stream of dividends For a long-term holder ofstocks this rational principle of valuation translates to a formula

Burton G Malkiel 73

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

r 5 DP 1 g

where r is the rate of return DP is the (expected) dividend yield and g is thelong-term growth rate For present purposes consider r to be the required rate ofreturn for the market as a whole Suppose initially that the ldquorisklessrdquo rate of intereston government bonds is 9 percent and that the required additional risk premiumfor equity investors is 2 percentage points In this case r will be 11 percent (009 1002 5 011) If a typical stockrsquos expected growth rate g is 7 percent and if thedividend is $4 per share we can solve for the appropriate price of the stock index(P) obtaining

011 5$4P 1 007

P 5 $100

Now assume that yields on government bonds rise from 9 to 105 percent withno increase in expected in ation and that risk perceptions increase so thatstock-market investors now demand a premium of 25 percentage points instead ofthe 2 points in the previous example The appropriate rate of return or discountrate for stocks r rises then from 11 percent to 13 percent (0105 1 0025) and theprice of our stock index falls from $100 to $6667

013 5$4P

1 007

P 5 $6667

The price must fall to raise the dividend yield from 4 to 6 percent so as to raisethe total return by the required 2 percentage points Clearly no irrationality isrequired for share prices to suffer quite dramatic declines with the sorts of changesin interest rates and risk perceptions that occurred in October 1987 Of courseeven a very small decline in anticipated growth would have magni ed thesedeclines in warranted share valuations

This is not to say that psychological factors were irrelevant in explaining thesharp drop in prices during October 1987mdashthey undoubtedly played a role But itwould be a mistake to dismiss the signi cant change in the external environmentwhich can provide an entirely rational explanation for a signi cant decline in theappropriate values for common stocks

The Internet Bubble of the Late 1990sAnother stock market event often cited by behavioralists as clear evidence of

the irrationality of markets is the Internet ldquobubblerdquo of the late 1990s Surely the

74 Journal of Economic Perspectives

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

remarkable market values assigned to Internet and related high-tech companiesseem inconsistent with rational valuation I have some sympathy with behavioralistsin this instance and in reviewing Robert Shillerrsquos (2000) Irrational Exuberance Iagreed that it was in the high-tech sector of the market that his thesis could besupported But even here when we know after the fact that major errors weremade there were certainly no arbitrage opportunities available to rational investorsbefore the ldquobubblerdquo popped

Equity valuations rest on uncertain future forecasts Even if all market partic-ipants rationally price common stocks as the present value of all future cash owsexpected it is still possible for excesses to develop We know now with the bene tof hindsight that outlandish and unsupportable claims were being made regardingthe growth of the Internet (and the related telecommunications structure neededto support it) We know now that projections for the rates and duration of growthof these ldquonew economyrdquo companies were unsustainable But remember sharp-penciled professional investors argued that the valuations of high-tech companieswere proper Many of Wall Streetrsquos most respected security analysts including thoseindependent of investment banking rms were recommending Internet stocks tothe rmrsquos institutional and individual clients as being fairly valued Professionalpension fund and mutual fund managers overweighted their portfolios with high-tech stocks

While it is now clear in retrospect that such professionals were egregiouslywrong there was certainly no obvious arbitrage opportunity available One coulddisagree with the projected growth rates of security analysts But who could be surewith the use of the Internet for a time doubling every several months that theextraordinary growth rates that could justify stock valuations were impossible Afterall even Alan Greenspan was singing the praises of the new economy Nothing isever as clear in prospect as it is in retrospect The extent of the ldquobubblerdquo was onlyclear in retrospect

Not only is it almost impossible to judge with con dence what the properfundamental value is for any security but potential arbitrageurs face additionalrisks Shleifer (2000) has argued that ldquonoise trader riskrdquomdashthe risk from traders whoare attempting to buy into rising markets and sell into declining marketsmdashlimitsthe extent to which one should expect arbitrage to bring prices quickly back torational values even in the presence of an apparent bubble Professional arbitra-geurs will be loath to sell short a stock they believe is trading at two times itsldquofundamentalrdquo value when it is always possible that some greater fools may bewilling to pay three times the stockrsquos value Arbitrageurs are quite likely to haveshort horizons since even temporary losses may induce their clients to withdrawtheir money

While there were no pro table and predictable arbitrage opportunities avail-able during the Internet ldquobubblerdquo and while stock prices eventually did adjust tolevels that more reasonably re ected the likely present value of their cash ows anargument can be maintained that the asset prices did remain ldquoincorrectrdquo for aperiod of time The result was that too much new capital owed to Internet and

The Efcient Market Hypothesis and Its Critics 75

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

related telecommunications companies Thus the stock market may well havetemporarily failed in its role as an ef cient allocator of equity capital Fortunatelyldquobubblerdquo periods are the exception rather than the rule and acceptance of suchoccasional mistakes is the necessary price of a exible market system that usuallydoes a very effective job of allocating capital to its most productive uses

Other Illustrations of Irrational PricingAre there not some illustrations of irrational pricing that can be clearly

ascertained as they arise not simply after a ldquobubblerdquo has burst My favorite illus-tration concerns the spinoff of Palm Pilot from its parent 3Com Corporationduring the height of the Internet boom in early 2000 Initially only 5 percent of thePalm Pilot shares were distributed to the public the other 95 percent remained on3Comrsquos balance sheet As Palm Pilot began trading enthusiasm for the shares wasso great that the 95 percent of its shares still owned by 3Com had a market valueconsiderably more than the entire market capitalization of 3Com implying that allthe rest of its business had a negative value Other illustrations involve ticker symbolconfusion Rasches (2001) nds clear evidence of comovement of stocks withsimilar ticker symbols for example the stock of MCI Corporation (ticker symbolMCIC) moves in tandem with an unrelated closed-end bond investment fund MassMutual Corporate Investors (ticker symbol MCI) In a charming article entitled ldquoARosecom by Any Other Namerdquo Cooper Dimitrov and Rau (2001) found positivestock price reactions during 1998 and 1999 on corporate name changes when ldquodotcomrdquo was added to the corporate title Finally it has been argued that closed-endfunds sell at irrational discounts from their net asset values (for example Shleifer2000)

But none of these illustrations should shake our faith that exploitable arbitrageopportunities should not exist in an ef cient market The apparent arbitrage in thePalm Pilot case (sell Palm Pilot short and buy 3Com) could not be undertakenbecause not enough Palm stock was outstanding to make borrowing the stockpossible to effectuate a short sale The ldquoanomalyrdquo disappeared once 3Com spun offmore of Palm stock Moreover the potential pro ts from name or ticker symbolconfusion are extremely small relative to the transactions costs that would berequired to exploit them Finally the ldquoclosed-end fund puzzlerdquo is not really a puzzletoday Discounts have narrowed from historical averages for funds with assetstraded in liquid markets and researchers such as Ross (forthcoming) have sug-gested that they can largely be explained by fund management fees Perhaps themore important puzzle today is why so many investors buy high-expense activelymanaged mutual funds instead of low-cost index funds

The Performance of Professional Investors

For me the most direct and most convincing tests of market ef ciency aredirect tests of the ability of professional fund managers to outperform the market

76 Journal of Economic Perspectives

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

as a whole Surely if market prices were determined by irrational investors andsystematically deviated from rational estimates of the present value of corporationsand if it were easy to spot predictable patterns in security returns or anomaloussecurity prices then professional fund managers should be able to beat the marketDirect tests of the actual performance of professionals who often are compensatedwith strong incentives to outperform the market should represent the most com-pelling evidence of market ef ciency

A remarkably large body of evidence suggests that professional investmentmanagers are not able to outperform index funds that buy and hold the broadstock market portfolio The rst study of mutual fund performance was undertakenby Jensen (1968) He found that active mutual fund managers were unable to addvalue and in fact tended to underperform the market by approximately theamount of their added expenses I repeated Jensenrsquos study with data from asubsequent period and con rmed the earlier results (Malkiel 1995) Moreover Ifound that the degree of ldquosurvivorship biasrdquo in the data was substantial that ispoorly performing funds tend to be merged into other funds in the mutual fundrsquosfamily complex thus burying the records of many of the underperformers Exhibit4 updates the study I performed through mid-2002 showing that the return forsurviving funds is quite a bit better than the actual return for all funds includingfunds liquidated or merged out of existence Survivorship bias makes the interpre-tation of long-run mutual fund data sets very dif cult But even using data sets withsome degree of survivorship bias one cannot sustain the argument that profes-sional investors can beat the market

Exhibit 5 presents the percentage of actively managed mutual funds that havebeen outperformed by the Standard amp Poorrsquos 500 and the Wilshire stock indexesThroughout the past decade about three-quarters of actively managed funds have

Exhibit 4The Records of Surviving Funds Overstates the Success of Active Management

10

1970 rsquo71

rsquo72

rsquo73

rsquo74

rsquo75

rsquo76

rsquo77

rsquo78

rsquo79

rsquo80

rsquo81

rsquo82

rsquo83

rsquo84

rsquo85

rsquo86

rsquo87

rsquo88

rsquo89

rsquo90

rsquo91

rsquo92

rsquo93

rsquo94

rsquo95

rsquo96

rsquo97

rsquo98

rsquo99

rsquo00

rsquo01

rsquo02

June

0

20

30

Surviving funds

All funds

40

50R

ate

of r

etur

n

Includes those liquidated or merged out of existenceSource Lipper Analytic Services as compiled by The Vanguard Group

Burton G Malkiel 77

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

failed to beat the index Similar results obtain for earlier decades Exhibit 6 showsthat the median large-capitalization professionally managed equity fund has un-derperformed the SampP 500 index by almost 2 percentage points over the past 10-15- and 20-year periods Exhibit 7 shows similar results in different markets andagainst different benchmarks

Managed funds are regularly outperformed by broad index funds with equiv-alent risk Moreover those funds that produce excess returns in one period are notlikely to do so in the next There is no dependable persistence in performanceDuring the 1970s the top 20 mutual funds enjoyed almost double the performanceof the index During the 1980s those same funds underperformed the index Thebest performing funds of the 1980s similarly underperformed during the 1990s Amore dramatic example of the lack of persistence in performance is shown inExhibit 8 These mutual funds during 1998 and 1999 enjoyed three times theperformance of the index During 2000 and 2001 they did three times worse thanthe index Over the long run the results are even more devastating to activemanagers One can count on the ngers of one hand the number of professionalportfolio managers who have managed to beat the market by any signi cantamount Exhibit 9 shows the distribution of return Of the original 355 funds only ve of them outperformed the market by 2 percent per year or more

The record of professionals does not suggest that suf cient predictability existsin the stock market or that there are recognizable and exploitable irrationalitiessuf cient to produce excess returns

Exhibit 5Percentage of Large Capitalization Equity Funds Outperformed by Index Ending6302002

SampP 500 vs Large Cap Equity FundsWilshire 5000 vs Large Cap Equity Funds

6372

1 year

5664

3 years

7069

5 years

7974

10 years

Source Lipper Analytic ServicesNote All large capitalization mutual funds in existence are covered with the exception of ldquosectorrdquofunds and funds investing in foreign securities

Exhibit 6Median Total Returns Ending 12312001

Large Cap Equity FundsSampP 500 Index

10981294

10 years

11951374

13421524

15 years 20 years

Source Lipper Analytic Services Wilshire Associates Standard amp Poorrsquos and The Vanguard Group

78 Journal of Economic Perspectives

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

Exhibit 7Percentage of Various Actively Managed Funds Outperformed by BenchmarkIndex 10 Years to 123101

80

60

40

20

0

100

U S L arg e -C a p Equ ityvs Samp P 50 0

U S L arg e -C a p Equ ityv s W ilsh ire

500 0

Eu rop eanF u n ds vs

M S C IEu rope

Ja pa n eseF u n ds vs

M SC IJap a n

Em erg in gM a rke t Fu n d s

v s M SC IE M F

G lo bal B on dF un d s v sSa lo m on

In d ex

Source Lipper Analytic Services and Micropal

Exhibit 8Getting Burned by Hot Funds

Van WagonerEmrg GrowthRydexOTC FundInvTCW GalileoAGr EqInstlRS InvEmerg GrowthPBHGLarge Cap 20Janus Olympus FundVan Kampen Aggr GroAJanus MercuryPBHGSel EquityWMGrowthABerger New GeneratnInvJanus EnterpriseJanus VentureFidelity Aggr GrowthJanus TwentyAmer CentNew OpptyMorg Stan Sm Cap GroBVan Kampen Emerg GroATCW GalileoSC GroInstlBlackRockMdCp GroInstl

123456789

1011121314151617181920

Rank

1998ndash1999 2000ndash2001

Name

Fund

105529343927890198456772476707631762174777331722872227056690967646659656764876444

AverageAnnualReturn

11061103109810551078106110671057109710461107110110911105109010331102102110991009

Rank

2435423631234002261722903227032280422635233192258224596235402308923802230832241123596222702347722218

Average Fund Return 7672 23152

SampP 500 Return 2475 21050

AverageAnnualReturn

Source Lipper Analytic Services and Bogle Research Institute Valley Forge PA

The Efcient Market Hypothesis and Its Critics 79

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

Conclusion

As long as stock markets exist the collective judgment of investors will some-times make mistakes Undoubtedly some market participants are demonstrably lessthan rational As a result pricing irregularities and even predictable patterns instock returns can appear over time and even persist for short periods Moreoverthe market cannot be perfectly ef cient or there would be no incentive forprofessionals to uncover the information that gets so quickly re ected in marketprices a point stressed by Grossman and Stiglitz (1980) Undoubtedly with thepassage of time and with the increasing sophistication of our databases and empir-ical techniques we will document further apparent departures from ef ciency andfurther patterns in the development of stock returns

But I suspect that the end result will not be an abandonment of the belief ofmany in the profession that the stock market is remarkably ef cient in its utilizationof information Periods such as 1999 where ldquobubblesrdquo seem to have existed at leastin certain sectors of the market are fortunately the exception rather than the ruleMoreover whatever patterns or irrationalities in the pricing of individual stocksthat have been discovered in a search of historical experience are unlikely to persistand will not provide investors with a method to obtain extraordinary returns If any$100 bills are lying around the stock exchanges of the world they will not be therefor long

y I wish to thank J Bradford De Long Timothy Taylor and Michael Waldman for theirextremely helpful observations While they may not agree with all of the conclusions in thispaper they have strengthened my arguments in important ways

Exhibit 9The Odds of Success Returns of Surviving Mutual Funds vs SampP 500 1970ndash2001

0

10

13

24or less

20

30

40

50

11

23

28

22

34

21

29

0 to21

21

0 to11

17

1

3

2

1

3

1

4 ormore

86 Losers

Number ofEquity Funds

355158

19702001

197Nonsurvivors

50 MarketEquivalent

22 Winners

Source Data from Lipper Analytic Services and The Vanguard Group

80 Journal of Economic Perspectives

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

References

Ariel Robert A 1990 ldquoHigh Stock ReturnsBefore Holidays Existence and Evidence on Pos-sible Causesrdquo Journal of Finance December 455pp 1611ndash626

Bagwell Laurie Simon and John B Shoven1989 ldquoCash Distributions to Shareholdersrdquo Jour-nal of Economic Perspectives Summer 33 pp129ndash 40

Ball Ray 1978 ldquoAnomalies in RelationshipsBetween Securitiesrsquo Yields and Yield-SurrogatesrdquoJournal of Financial Economics July 6 pp 103ndash26

Basu Sanjoy 1983 ldquoThe Relationship Be-tween Earningsrsquo Yield Market Value and theReturns for NYSE Common Stocks Further Evi-dencerdquo Journal of Financial Economics June 121pp 129ndash56

Campbell John Y 1987 ldquoStock Returns andthe Term Structurerdquo Journal of Financial Econom-ics June 18 pp 373ndash 400

Campbell John Y and Robert J Shiller 1988ldquoStock Prices Earnings and Expected Divi-dendsrdquo Journal of Finance 433 pp 661ndash76

Campbell John Y and Robert J Shiller 1998ldquoValuation Ratios and the Long-Run Stock Mar-ket Outlookrdquo Journal of Portfolio ManagementWinter 24 pp 11ndash26

Campbell John Y Andrew W Lo and ACraig MacKinlay 1997 The Econometrics of Finan-cial Markets Princeton Princeton UniversityPress

Cooper Michael Orlin Dimitrov and PRRau 2001 ldquoA Rosecom by Any Other NamerdquoJournal of Finance June 15 56 pp 2371ndash388

Cootner Paul ed 1964 The Random Characterof Stock Market Prices Cambridge Mass MITPress

DeBondt Werner F M and Richard Thaler1985 ldquoDoes the Stock Market Overreactrdquo Jour-nal of Finance July 40 pp 793ndash 805

DeBondt Werner F M and Richard Thaler1995 ldquoFinancial Decision-Making in Marketsand Firms A Behavioral Perspectiverdquo in Hand-book in OR amp MS Volume 9 R Jarrow et al edsNew York Elsevier Science Chapter 13

Fama Eugene 1970 ldquoEf cient Capital Mar-kets A Review of Theory and Empirical WorkrdquoJournal of Finance 252 pp 383ndash417

Fama Eugene 1998 ldquoMarket Ef ciencyLong-Term Returns and Behavioral FinancerdquoJournal of Financial Economics 493 pp 283ndash306

Fama Eugene and Kenneth French 1988ldquoPermanent and Temporary Components ofStock Pricesrdquo Journal of Political Economy 962pp 246ndash73

Fama Eugene and Kenneth French 1992ldquoThe Cross-Section of Expected Stock ReturnsrdquoJournal of Finance June 47 pp 427ndash 65

Fama Eugene and Kenneth French 1993ldquoCommon Risk Factors in the Returns on Stocksand Bondsrdquo Journal of Finance 331 pp 3ndash56

Fama Eugene and Kenneth French 1997ldquoValue vs Growth The International EvidencerdquoJournal of Finance December 53 pp 1975ndash999

Fama Eugene and Kenneth French 2001ldquoDisappearing Dividends Changing Firm Char-acteristics or Lower Propensity to Payrdquo Journal ofFinancial Economics April 601 pp 3ndash 43

Fama Eugene and Kenneth French 2002ldquoThe Equity Premiumrdquo Journal of Finance 622pp 637ndash59

Fama Eugene and G William Schwert 1977ldquoAsset Returns and In ationrdquo Journal of FinancialEconomics November 52 pp 55ndash69

Fluck Zsuzsanna Burton Malkiel and RichardQuandt 1997 ldquoThe Predictability of Stock Re-turns A Cross-Sectional Simulationrdquo Review ofEconomics and Statistics 792 pp 176 ndash83

French Kenneth 1980 ldquoStock Returns andthe Weekend Effectrdquo Journal of Financial Econom-ics March 8 pp 55ndash 69

Graham Benjamin and David L Dodd 1934Security Analysis Principles and Techniques NewYork McGraw Hill

Graham Benjamin and David L Dodd 1965The Intelligent Investor New York Harper amp Row

Grossman Sanford J and Joseph E Stiglitz1980 ldquoOn the Impossibility of InformationallyEf cient Marketsrdquo American Economic Review703 pp 393ndash408

Haugen Robert A and Josef Lakonishok1988 The Incredible January Effect HomewoodDow Jones-Irwin

Hawawini Gabriel and Donald B Keim 1995ldquoOn the Predictability of Common Stock Re-turns Worldwide Evidencerdquo in Handbooks in Op-erations Research amp Management Science Volume 9R Jarrow et al eds Amsterdam Elsevier Sci-ence B V pp 497ndash544

Jensen Michael 1968 ldquoThe Performance ofMutual Funds in the Period 1945ndash64rdquo Journal ofFinance May 23 pp 389ndash 416

Kahneman Daniel and Mark W Riepe 1998ldquoAspects of Investor Psychologyrdquo Journal of Port-folio Management Summer 244 pp 52ndash 65

Kahneman Daniel and Amos Tversky 1973ldquoOn the Psychology of Predictionrdquo PsychologicalReview 80 pp 237ndash51

Kahneman Daniel and Amos Tversky 1979

Burton G Malkiel 81

ldquoProspect Theory An Analysis of Decision Un-der Riskrdquo Econometrica 472 pp 263ndash91

Keim Donald B 1983 ldquoSize-Related Anoma-lies and Stock Return Seasonality Further Em-pirical Evidencerdquo Journal of Financial EconomicsJune 12 pp 13ndash32

Keim Donald B and Robert T Stambaugh1986 ldquoPredicting Returns in Stock and BondMarketsrdquo Journal of Financial Economics Decem-ber 17 pp 357ndash90

Lakonishok Josef and S Smidt 1988 ldquoAreSeasonal Anomalies Real A Ninety-Year Per-spectiverdquo Review of Financial Studies Winter 14pp 403ndash25

Lakonishok Josef Andrei Shleifer and Rob-ert Vishny 1994 ldquoContrarian Investment Ex-trapolation and Riskrdquo Journal of Finance Decem-ber 49 pp 1541ndash578

Lesmond David Michael Schill and Chun-sheng Zhou 2001 ldquoThe Illusory Nature of Mo-mentum Pro tsrdquo Unpublished manuscript Tu-lane University

Lo Andrew W and A Craig MacKinlay 1999A Non-Random Walk Down Wall Street PrincetonPrinceton University Press

Lo Andrew W Harry Mamaysky and JiangWang 2000 ldquoFoundations of Technical AnalysisComputational Algorithms Statistical Inferenceand Empirical Implementationrdquo Journal of Fi-nance August 554 pp 1705ndash765

Malkiel Burton G 1973 A Random Walk DownWall Street New York W W Norton amp Co

Malkiel Burton G 1995 ldquoReturns from In-vesting in Equity Mutual Funds 1971 to 1991rdquoJournal of Finance June 502 pp 549 ndash72

Malkiel Burton G 2000 ldquoReview of Robert JShillerrsquos Irrational Exuberancerdquo Wall Street JournalApril 4

Miller Merton 1991 Financial Innovations andMarket Volatility Cambridge Blackwell

Nicholson S F 1960 ldquoPrice-Earnings Ratiosrdquo

Financial Analysts Journal JulyAugust 16 pp43ndash50

Odean Terrance 1999 ldquoDo Investors TradeToo Muchrdquo American Economic Review Decem-ber 895 pp 1279ndash298

Poterba James and Lawrence Summers 1988ldquoMean Reversion in Stock Returns Evidenceand Implicationsrdquo Journal of Financial Economics221 pp 27ndash59

Rasches Michael 2001 ldquoMassively ConfusedInvestors Making Conspicuously IgnorantChoices (MCI-MCIC)rdquo Journal of Finance 565pp 1911ndash927

Roll Richard and Robert J Shiller 1992ldquoComments Symposium on Volatility in USand Japanese Stock Marketsrdquo Journal of AppliedCorporate Finance 51 pp 25ndash29

Ross Stephen Forthcoming Princeton Lecturesin Finance 2001 Princeton Princeton UniversityPress

Samuelson Paul 1965 ldquoProof that ProperlyAnticipated Prices Fluctuate Randomlyrdquo Indus-trial Management Review Spring 6 pp 41ndash 49

Schwert G William 2001 ldquoAnomalies andMarket Ef ciencyrdquo in Handbook of the Economicsof Finance G Constantinides et al eds Amster-dam North Holland Chapter 17

Shiller Robert J 1981 ldquoDo Stock Prices Moveso Much to be Justi ed by Subsequent Changesin Dividendsrdquo American Economic Review 713pp 421ndash36

Shiller Robert J 1996 ldquoPrice-Earnings Ratiosas Forecasters of Returns The Stock MarketOutlook in 1996rdquo Unpublished manuscriptYale University

Shiller Robert J 2000 Irrational ExuberancePrinceton Princeton University Press

Shleifer Andrei 2000 Inefcient Markets AnIntroduction to Behavioral Finance New York Ox-ford University Press

82 Journal of Economic Perspectives