chapter 5 scott

8/10/2019 Chapter 5 Scott

1/65

5.1 OVERVIEW

There is a saying that "the proof of the pudding is in the eating."If the efficient markets theory and the decision theories

underlying it are reasonable descriptions of reality on average,

we should observe the market value. of securities responding in

predictable ways to new information.

This leads to an examination of empirical research inaccounting. espite the difficulties of designing experiments to

test the implications of decision usefulness, accounting research

has established that security market prices do respond to

accounting information. The first solid evidence of this security

market reaction to earnings announcements was provided by !alland !rown in 1 #$. %ince then, a large number of empirical

studies have documented additional aspects of securities market

response.


2/65

&n the basis of these %tudies, it does seem that accounting

information is useful to investors in helping them estimate the

expected values and risks of security returns. &ne has only tocontemplate the use of !ayes' theorem in (xample ).1 to see that if

accounting Information did not have information content there

would be no revision of beliefs upon receipt, hence no triggering of

buy*sell decisions. +ithout buy*sell decisions, there would be no

trading volume or price changes. In essence, information is useful if it leads investors to change their beliefs and actions. urthermore,

the degree of usefulness for investors can be measured by the excent

of volume or price change following release of the information.

This e-uating of usefutness to information content is called

the information approach to decision usefulness of financialreporting, an approach that has dominated financial accounting

theory and research since 1 #$, and has only within the last few

years yielded to a measurement approach, to be discussed in hapters

# and /. 0s we have seen in %ections ).$ and .$, the information

approach has been adopted by ma2or accounting standard3setting bodies. This approach takes the view that investors want to make

their own predictions of future security rerurns 4instead of having

accountants do it for them, as under ideal condition and will

"gobble up" all useful information in this regard. 0s mentioned,

empirical research has shown that at least some accounting


3/65

information is perceived as useful. urthermore, the information

approach implies that empirical research can help accountants to

further increase usefulness by letting market response guide them

as to what information is and is not valued br investors.

Theinformation approachto decision usefulnessis an approach to financial

reporting that recogni6es individual responsibility for predicting future firm

performance an that concentrateson providing usefull information for this puspos

The approach assumes securities market efficiency, recogni6ing that the markewill react to useful information from any source, including financial statement.

&ne must be careful, however, when e-uating usefulness with the

extent of security price change. +hile investors, and accountants,

may benefit from useful information, it does not follow that society

will necessarily be better off. Informarion is a very complex

commodity and its private and social values are not the same. &ne

reason is cost. inancial statement users do not generally pay directly

for this information. 0s a result, they may find information useful

even though it costs society more 4e.g., in the form of higher

product prices to help firms pay for generating and reporting the

information than the increased usefulness is worth. urthermore,

information affects people differently, re-uiring complex cost3benefit

radeoffs to balance the competing interests of different constituencies.


4/65

These social considerations do not invalidate the information

approach. 0ccountants can still strive to improve their competitive

position in the information marketplace by providing useful

information. 0nd, it is still true that securities markets will work

better to allocate scarce capital if security prices provide good

indicators of investment opportunicies. 7owever, what accountants

cannot do is claim that the best accounting policy is the one that

produces the greatest market response.

igure 5.1 outlines the organi6ation of thi%chapter.


5/65

5.8 OUTLINE OF THE RESEARCH PROBLEM

5.8.1 Rea on for Mar!et Re pon e

+e begin by reviewing the reasons why we would predict that

the market price of a firm's shares will respond to its financial

statement information. or most of this chapter we will confine

financial statement information to reported net income. Theintormation content of net income is a topic that has received

extensive empirical investigation. lnformation content of

other financial statement components wilt be discussed in

%ection 5./ and in hapter /.

onsider the following predictions about investor behaviour,in response to financial statement information9

I. investor have priot beliefs about a firms future performance, that

is, its dividends, cash flows, and*or earnings, which affect the

expected returns and risk of the firm's shares. These prior

beliefs will be based on all available information, including

market price, up to 2ust prior to the release of the firm's current

net income. (ven if they are based on publicly available

information, these prior beliefs need not all be the same,

because investors will differ in the amount of information they

have obtained and in their abilities to interpret it.


6/65

8. :pon release of current year's net income, certain investors

will decide to become more informed by analy6ing the income

number. or example, if net income is .high, &r higher than

expected, this may be good news. If so, investors, by means of

!ayes' theorem, would revise upward their beliefs about

future firm performance. &ther investors, who perhaps had

overly high expectations for what cunene net income should

be, might interpret the same ner income number as bad news.

). investors who have revised their beliefs about furure firm

performance upward willbe inclined to buy the firm's shares at

their current market price, and vice versa for those who have

revised their beliefs downward. Investors' evaluations of theriskiness of these shares may also be revised.

. +e would expect ro observe the volume of shares traded to

increase when che firm reports its net income. urthermore, chis

volume should be greater the greater are the differences in

investors' prior beliefs and in their Interpretations of the current;inan3 cial information. If the investors who interpret reported

net income as good news 4and hence have increased their

expectations of future performance outweigh those wh.o

interpret it as bad news, we would expect co observe an

increase in the market price of the firm's shares, and vice versa.


7/65

!eaver 41 #$ , in a classic study, examined trading

volume reaction. 7e found a dramatic increase in volume

during ;he week of release of earnings announcements. urther

details of !eaver's findings are included in -uestion of this

chapter. In the bal3 ance of this chapter we will concentrate on

market price reacrion. errecchia 41 / suggests that volume is noisier

than price change as a measure of decision usefulness of financial

statement information.

?ou will recogni6e chat the preceding predictions follow the

decision theory andefficienc markets theory of hapters ) and -uite closely. If these

theories are to have relevance to accountants, their predictions should

be borne out empirically. 0n empirical researcher could test these

predictions by obtaining a sample of firms that issue annual reports

and investigating whether the volume and price reacttons to good or bad news in earnings occur as the theories lead us to believe. This is

not as easy as it might seem, how3 ever, for a number of reasons, as we

will discuss next.

5.8.8 inding the


8/65


9/65


10/65

here 4com %ection .5 4(-uation .

0s described in %ection .5, the researcher will obtain past

data on @ 2, )nd @ < K and use regression analysis to estimate the

coefficients of the model. %uppose rhar chis yields

4H. 2 L E.EEE1 and M2 L E.$E, as shown in the figure, t Now, armed with this estima re of the market mode I for firm 2,

the researcher can con3 sulO The Wall Street Journal to find ;he day

of ;he fimP's current earnings announcement. all this day "day

E." %uppose that for day E the return on 4he ow Hones lndustrial

index was E.EE1.8 Then, the estimated market model for fimP 2

i.s used to predict the return EE finn 2's shares for this day. 0s

shown in igure 5.8., this expected return is E.EEE . Now

assume that the actt


11/65

day." This abnormal return is also interpreted as the rate of return on

finn 2's shares for day E (lfu r removing the influence of market3wide

factors. Note that this interpretation is consistent with (xample ).),

where we separated the factors thar affect share returns into market3

wide and firm3specific categories. The present procedure proM'idesan

operacional way to make this separation.

5.8. omparing @eturns and Income

The empirical researcher can now compare the abnormal share return

on day E as calculated above with the unexpected component of the

firm's current reponed net income. If this unexpected net income isgood news 4rhar is, a positive unexpected net income chen, given

securities market efficiency, a positive abnormal share return

constitutes evidence that investors on average are reacting favourably

to the unexpected good news in earnings. 0 similar line of reasoning

applies if the current earnings announcement is bad news.To Increase the power of the investigation, the researcher mal' wish

to also examine

a few days on either side of day &. It is possible, 4or example, that

the efficient market might learn of the good or bad earnings news a

day or two early. onversely, positive or negative abnormal returnsmay conc%cwe for a Hat' or &' ) sitet day * while (he mar+et


12/65

digests the informacion, although market effi iency implies that any

excess returns should die out -uickly. onse-uently, the sumrrung of

abnormal returns for a three3to3five3day narrow window around day

E seems more reasonable than exarruning day E only. It also helps

protect against the possibility that the date of publication of current

earnings in the financial media may not be a completely accurate

estimate of the date of rheir public availabihry.

If posirive and ne,ative abnormal returns surrounding good &f

bad earnings news are found to hold across a sample of firms, the

researcher may conclude that predictions based on the decision theory

and efficient securities market theory are supported. This would in

;urn support the decision usefulness approach to financial

accounting and reporting, because, if investors did not find the

reported net income information useful, a market response would

hardly be observed.

&f course, this methodology is not foolproof3a number of

assumptions and estima3 nons have co be made along rhe way. &ne

complication is that other firm3specific intor3 marion fre-uenrlv

comes along around the time of a firm's earnings announcement.

or example, if finn 2 announced a stock split or a change in its

dividend on the same day that it released its current earnings, it

would be hard to know if a market response was due to one or the

other. 7owever, researchers can cope with this by simply removingsuch firms from rhe sample.


13/65


14/65

its capital structure. If the estimated bera is different 4rom the

true beta, this affects the calcularion of abnormal return, posstblv

biasing the results of the investigation.There is a variety of ways to cope wirh this complication. or

example, it may be pos3

Mibk to get a "second opinion" on beta by esMim'iltPng it from

financial statement information rather than from marker data.

4This is considered in %ection /.5.1. 0lternatively, beta may beestimated from a period after the earnings announcement and

compared with the estimate from a period before the

announcement.

0lso, there are ways to separate market3wide and firm3

specific returns that ignore beta. or example, we can estimatefirm3specific returns by the difference between firm 2's stock

return during period E and the average return on its shares

over some prior period.

&r, we can take the difference between firm 2's return during

period E and the return on

the market portfolio for the same period. 0lternatively, as in

(aston and 7arris 41 1 , we can simply work with rotaR share

returns and not Sactor &ut market3wide returns at all.

The rationale for rhese simpler procedures is that there is no

guarantee that the mar3 ket model ade-uarelv captures the real


15/65

process generating share returns3the impact of estimation

risk EE the 0 < was discussed in hapter . To the extent

that the market model does not fully capture reality, its use may

introduce more error in estimating beta and abnormal returns

chan it reduces by removing market3wide returns and

controlling for risk. 0 further complication is that there is a

variety of market oonfclio return indices

available, of which the ow Hones Industrial 0verage is onlvone. +hich one should be

used

These issues were examined by !rown. and +arner 41 $E

in a simulation study. espite modelling and measurement

problems such as those 2ust mentioned, !rown and +arner concluded that, for monthly return windows, the market model3

based procedure outlined in %ection 5.8.) performed reasonably

well relative to the above alternatives. onse-uently, this is the

procedure we will concentrate on.

:5ing this ptocedure, it does appeal that the market reacts

to earnings information much as the theories predict. +e will

now review the firsr solid evidence of chis reaction. the famous

1"#$ !all and !rown study.


16/65

5.) T7( !0GG 0N !@&+N %T: ?

5.).1


17/65

!! examined a sample of 8#1 New ?ork %rock (xchange

4N?A( firms over nine years from 1 5/ to 1 #.5. They

concentrated on the information content of earnings, co theexclusion of other potentially informative financial statement

components such as li-uidity and capital structure. &ne reason

4or this, as mentioned earlier, was that earnings for N?%(

41E115 were typically announced in the media prior to

actual release of the annual report so that it was relatively easyto determine when the informacion first became publicly

available,

!!'s first 4ask was to measure the information content of

earnings, that is, whether reported earnings were greater than

what the market had expected 4VN , or less than expected4!N . &f course, this re-uires a proxy for the market's

expectation. &ne proxy they used was last year's acruai

earnings, from which it follows their unexpected earnin,s

is simply the change in earnings. % Thus, firms with earnings

higher than last year'. were

classified as &N, and firms with earnings lower than last year's

were classified as !N.

The next task was to evaluate the market return on the

shares of the sample firms near the time of each earnings

announcement. This was done according to the abnormal


18/65

returns procedure illustrated in igure. 5.8. The only difference

was that !! used monthly returns 4daily returns were not

available on databases in 1 #$ .

-nalo,ousl # co /i,ure 5.8, %:pp&7' rhar firm 0 reported irs

1 5/ earnings in ebruary

1 5$, and that these earnings were &N. %uppose also that rhe

return on the N?%( mar3 ket portfolio in ebruary 1 5$ was

E.EE1, yielding an ex peered firm 2 rerum of E.EEE . !! wouldthen calculate the actual return on firm 2 shares for ebruary

1 5$. %uppose this was E.EE1.5, yielding an abnormal return for

ebruary of E.EEE#. %ince finn 2's 1 5/ earn, ings were reported

in ebruary 1 5$ and since its shares earned E.EEE# over and

above the market in this moreh, one might suspect that thereason for the positive abnormal rerum was that investors were

reacting favourably to the VN information in earnings.

The -uestion then was9 +as this pattern repeated across the

sample The answer +4I% yes. If we rake all the &N earnings

announcements in the sample 4there were 1,8)1 , rne avera,eabnormal security market return in the month of earnings

release was strongly pos' irive. onversely, the average abnormal

return for the 1,1E bad news earnings announce, ments in the

sample was strongly negative. This provides substantial

evidence that the market did respond to the good or bad newsin earnings during a narrow window consisting of the month of


19/65

earnings announcement release.

0n interesting and important aspect of the !! study was that

they repeated their abnormal security market returns calculation

for a wide window consisting of each of the 11 months prior to

and the six months following the month of earnings release

4moruh E . !! calculated average abnormal returns for each

month of this l %3month window. The results are shown in

igure 5.), taken from !!.

The upper @rt of igure 5.) shows cumulative average

abnormal returns for the &N earnings announcement firms in

4he sampleU the bottom part shows the same for the !N

announcement tlrms. 0s can be seen, toe VN 4irms strongly

outperformed the total sample 4which approximates the market3wide return ' and the !N firms strongly under, performed, over

the 113rnonrh period leading up to the month of earnings

release.


20/65

5.).8 ausation >ersus 0ssociation

Nore char the returns are cumulative in the diagram. +hile there was asubstantial increase 4for &N and decrease 4for !N in average

abnormal returns in the narrow window con3 sisting of month E, as

uescSibed above, iguMe 5.) suggests that the market begar, to antic1

ipate the &N or !N as much as a year early, with the result that

returns accumulated steadily over the period. -s can be seen, if aninvesror could have bought the shares of all &N firms one year before

the good news was released and held them until the end of the monrh

of release, there would have been an extra return of about #W over

and above the

igure 5.) 0bnormal @eturns for VN and !N irms

1.n

1.1E

1.E$


21/65


22/65

Total sample * .

318

31E 3#

38 o

Z8 Z

Z#


23/65

market3wide return. %imilarly, an abnormal loss of over W

would have been incurred on a portfolio of !N firms bought one

year before the bad news was released.This leads to an important distinction between narrow and

wide window studies. If a security market reaction 4E

accounting information is observed during a narrow window of

a few days 4E4, in the case of !!, a month surrounding an

earnings announcement, it can be argued that thtO accountinginformation is rhe cause of the market reaction. The reason is

that during a narrow window there are relatively few firm3specific

events other than nee income to affect share returns. 0lso, if

other events do occur, such as stock splits or dividend

announcements, the affected firms can be removed from thesample, as mentioned. Thus, a narrow3window association

between security returns and accounting information suggests

that accounting disclosures are the S)UTee of new information to

investors.

(valuation of se l,rity returns over a wide window, however,opens them up to a host of other value3relevant events, or

example, a firm or have discovered new oil and gas reserves, be

engaged in promismg @B pro2ects, or have rising sales and

market share. 0s rhe market learns this information from more

rimely sources, such as media articles, firm announcements,conditions in the economy and industry, -uarterly reports,


24/65

etc., share price would begin to rise. This reflects the pardy

informative nature of secunry prices since, in an efficient

market, security prices reflect all available informanon. not

H:%t accounting information. Thus, firms that in a real sense are

doing well would have much

of 4he effect on their share prices anticipated by the efficient

market before the &N appears in the financial statements. That

is, because of recognition lag, prices leadearnings over a widewindow.

learly, this effect was taking place in the !! study. 0s a

result, it cannot be claimed chat reported net income causes the

abnormal returns during the 11 months leading up to month

EX The most that can be argued is ;hat net income and returnsare essocicred,That is, for wide windows, it is the real,

underlying, economic performance of the firm that generates

;he association, since both share price and 4with a lag net

income reflect real performance.

To pursue this "prices lead earnings" effect, suppose that wecontinue to widen the window! perhaps up ro several years. +e.

will find rhar rhe association between share returns and

earnings increases as the window widens. +hile historical

cost3based net income tends to lag behind the market in

reflecting value3relevant events, as the window is widened ;herelative effect of the lag decreases. &ver a long period of time


25/65


26/65

!ut a narrow window association, as !! found for month E,

provides stronger support for decision usefulness, since it

suggests that it is the accounting information that actuallycauses investor belief revision and hence security returns.

5.).) &utcomes of the %! %tudy

&ne of the most important outcomes of !! was that it

opened up a large number of additional usefulness issues. 0

logical next step is to ask whether the ma,J3ilud.e of

unexpected earnings is related co the ma,nitude of the securirv

market response3recall chat !!'s analysis was based only on the si,n of unexpected earnings. That is, rhe information content

of earnings in !!'s study was classified only inco &N or !N,

a fairly coarse measure.

The -uestion of magnitude of response was investigated,

for example, by !eaver, larke, and +right 4! + in 1 / .They examined a sample of 8/# N?%( firms with ecember )1

year3ends, over the 1E3year period from 1 #5 to 1 / . or each

sample firm. for each year of the sample period, they estimated

the unexpected earnings changes. They then used the market

model procedure described in %ections .5 and 5.8.) co esnmare

the abnormal securirv returns associated with. these unexpected


27/65

earnings changes.

:pon comparison of unexpected earnings changes with

abnormal security returns, ! + found that the greater the

change in unexpected earnings, the greater the security market

response. This resulr is consistent with the 0 < 4%ection

.5 and with the decision usefulness approach, since the larger

are unexpected earnings changes the more investors on average

will revise their estimates of future firm performance andresulting returns from their investments, other things e-ual."

0lso, since 1 #$, accounting researchers have %tudied

securities market response to net income on other stock

exchanges, in other countries, and for -uarterly earnings

reports, with similar results. The approach has been applied tostudy market response to the informacion contained in new

accounting standards, auditor changes, etc. 7ere, however, we

will concentrate on what is probably the most important

extension of !!, earnings response coefficients. This line of

research asks a different -uestion than ! +, namely, for a ,ivenamount of unexpected earnings, is the security market response

greater f&t ",omc firms than for others

5. (0@NINV% @(% &N%( &( I I(NT%


28/65


29/65

found, accountants can improve their understanding of how

accounting information is useful to investors. This. in rum, could

lead to the preparation of more useful financial statements.onse-uenrlv, one of the most important directions that

empirical financial accounting research took following the !! study

was the Identification and explanation of differential market

response to earnings information. This is called earnings response

coefficient 4(@e research. %

-n earnin,s response coefficient measures the extent of a securit 4s

a$normal mar+et return in response to the unexpected component of

reported earnin,! of the firmissuing5hal security.

5. .1 @easons for ifferential


30/65

earnings as an indicator of future firm performance and share returns,

the riskier these future returns are the lower investors' reactions to a

given amount of unexpected earnings will be.

To illustrate, think of a typical risk3averse, rational investor whose

utility increases in the expected value and decreases in the risk of the

return on his or her portfolio. %uppose that the investor, upon

becoming aware that a portfolio security has 2ust released on

earnings information, revises upwards the expected late of return

on this security, and decides to buy more of it. 7owever, if this

security has high beta, this will increase portfolio risk onse-uently,

the investor would not buy as much more as if the security was low

beta. In effect, the high beta acts as a brake on the investor's demand

for the &N seen riry. %ince all risk3averse, rarional, informed investorswill think chis way, the demand 4or the &N firm's shares will be

lower the higher is irs beta, other things e-ual. &f course, lower

demand implies a lower increase in market price and share return in

response to the &N, hence. a lower (@ .

(mpirical evidence of a lower (@ 4or higher3beta securities wasfound by ollins and =othari 41 $ and by (aston aod \mi2ewski

41 $ .

apital %tructure or highly levered firms, an increase, say, in

earnings 4before interest adds strength and safety to bonds andother outstanding debt, so that much of the good news in earnings


31/65

goes to the debtbolders racher chan the shareholders. Thus, the (@

for a highly levered firm should be lower than char of a firm with

littie or no debt, other things e-ual.

(mpirical evidence of a lower (@ for more highly levered firms

was reported by

haliwal, Gee. and argher 41 H .


32/65

Earnin" #$a%it& @ecall from %ection ).).8 that we define the

-uality 4i.e., the informativeness of earnings by the magnitude

of the main diagonal probabilities of the associated informationsystem. The higher these probabilities, the higher we would

expect the (@e co be, since investors are benet able to infer

future firm performance from current performance.

0s a practical matter, measurement of earnings -uality is less

clear, since information system probabilities are not directlyobservable and a sampling approach runs into problems of

estimation risk due to sampling error. 0n indirect approach,

discussed in %ection ).).8, is to infer earnings -uality by the

magnitude of analysts' earnings forecast revisions following

earnings announcements. 7owever, this 2ust raises the -uestionof whyanalysts revise

their forecasts more for some firms than others.

ortunately, orher dimensions of earnings -uality are

available, including the important concept of earnings

persistence. +e would expect that the (@ will be higher the

more the good or bad news in current earnings is expected co

persist into the future, since current earnings ;hen provide a

better indication of future firm performance. Thus, if currene &N

is due co the successful introduction of a new product or ost3

cutting by management, the (@ should be higher than if the


33/65


34/65

K ermanent, expected to persist indefinitely

K Transitory, affecting earnings in the current year but not

future years

K rice3irrelevanc, persistence of 6ero

The (@ s per dollar of unexpected earnings for these are 4l

Z @r OIP3 4where @4is the risk3free rate of interest under ideal

conditions , 1, and E respectively. 1E

In effect, there are three (@ s, all of which may be present in

the same income state3 ment. @T suggest that instead of tryingco estimate an average (@e, investors should acternpr to

identit? the three types separately and assign different (@ s

co each. In so


35/65

doing, rhev can identify the firm's permanent, or persistent, earning

power. This implies that accountants should provide lots of

classification and detail on the income state rnent.

To understand the (@ for permanent earnings, note that it can be

written as 1 Z 1M.

Thus, under ideal conditions, the market response to A1 of

permanent earnings consists of the current year's installment of A1

plus the present value of the perpetuity of future installments of If@ r .4This ignores riskiness of the future installments, which is appropri3

are if investors are risk3neutral or the permanent earnings are firm3

specific. +riting the

(@e this way also shows chat when earnings persist beyond the

currenr year, the magnitude of the (@ varies inversely with theinterest rate.

0nother aspect of (@ s is that their persistence can depend on

the finn's account3 ing policies. or example, suppose that a firm uses

current value accounting, sav for a cap3 ical asset! and char 4he fair

value of che asset increases $y73**. -ssum


36/65

%uppose instead 4hat the linn uses historical cost accounting for

the asset and that the annual increase in contribution margin is A .E .

Then there will be only A .E of &N in earnings this year. The

reason, of course, is that under historical cosc accounting the

A1EE increase in current value is brought into income only as it is

reali6ed. The efficient

market will recogni6e that the current A .E &N is only the "first

installment.9"O If ie regards the value increase as permanent and @4 L

1EW,the (@ will be II 41.1E*E.1E .

\ero3persistence income statement components can result 4rom

choice of accounting policy. %uppose, for example, that a firm

capitali6es a large amount of organi6ation cost sT.his could result in &N

on the current income statement, which is freed of the &%t%because

of their capualuacion. 7owever, assuming the organi6ation &%t%

have no salvage value, the market would not react to the u&N," that

is, its persistence is 6ero. 0s another example, suppose that a firm

writes off research cost scurrently in accordance with &00 . This could

produce !N in current earnings. 7owever, to the extent the market

perceives the research &%t%as having future value, it would react

positively co this !N so that persistence is negative. The possibility

of 6ero or negati ve persistence suggests once more the need for

detailed income statement disclosure, including a statement of

accounting policies.*^Y. second dimension of earnings -ualirv is accruals -ualiry. This


37/65

approach was proposed by e how and ichev 48EE8 . They pointed

out that net income is composed of9

9et income L cash now from oper9HSions9t ner accruills

where net accruals, which can be positive or negative, include

changes in non3cash work3ing capital accounts such as receivables,

allowance for doubtful accounts, inventories, accounts payable, etc.,

as well as amorti6ation expense. They then argued that earnings

-uality depends primarily on the -uality of working capical accruals,

since cash flow from operations is relatively less sub2ect to errors and

manager bias, and therefore of reasonably high -uality to start with.


38/65


39/65

suggested estimating the following regression e-uation9

where #.+ , is the change in net non3cash working capital for

the firm in -uestion for period r, that is, working capital

accruals. or example, in our illustracion above, if accounts

receivable and allowance for doubtful accounts are the only non3

cash working capital items, working capital has increased by

#.+ , L3 A8#E 4l.e., A EE 3 A# E in period r. This is an accrual

because net income includes this amount 4assuming the firm

recogni6es income at point of sale but ir has not yec been received

in cash.

&,XI is cash flow from operations in period r 33 1, erc., b o' bl'

and b 8 are constants to be estimated, and D, is the residual error

term, that is, the portion of total accruals not

explained by cash from operations.

0ccrual -uality, hence earnings -uality, is measured by the

magnitude of (" chat is, high (, indicates a &&T mateo between

current accruals r . we, and acroai operating ta.sh flow reali6ations.

(vidence thar firms' (@ s and share prices respond positively

to accrual -ualiry as measured by this procedure is reported by

rancis, Gafond, &lsson, and %chipper 48EE and 8EE5 and


40/65

(cker, rancis, =im, &lsson. and %chipper 48EE# .1)

'ro(th Opport$nitie The VN or !N in current earnings

may suggest future growth prospects 4or the finn, and hence a

higher (@ . &ne might rhink that since financial statements

contain a considerable historical cost component, net income

really cannot say much about the future growth of the firm.

7owever, this is not necessarily the case. %uppose that current net

income reveals unexpectedly high profitability for some of

the firm's recent investment pro2ects. This may indicate to the

market that rhe firm will en2oy strong growth in the future. &ne

reason, of course, is that to the extent the high


41/65


42/65

will have different expectations of a firm's next3 period earnings,

depending on their prior information and the extent of their abihties

ro evaluate financial statement information. 7owever, these

differences will be reduced to the extent that they draw on a common

information source, such as analysts' consensus forecasts, when

forming their expectations. onsider a firm's announcement of its

currenr earnings. epending on their expectartons, some investors

will regard this information as VN, others as !N, hence some will be

inclined ro buy and some to sell. 7owever, to the extent that

investors' earnings expectations were "close together," chey will put

the same mterprerarlon on the news. or example, if most invesrors

base their earnings expectation on the analysts' consensus forecast,

and current earnings are less than forecast, they will all regard this as

!N and will be inclined ;E sell rather than buy. Thus, the more

similar the earnings expectaricns the greater rhe effect of a dollar of

abnormal earnings on share price. In effect, rhe more precise are

analysts' forecasts the more similar are investors' earnings

expeccations and the greater the (@ , other things e-ual. or an

analysis of condirions under which the (@ is increasing in the

precision of analysts' earnings forecasts and how this precision is

affected by factors such as the num3 ber of analysts forecasting the

firm, see 0barbanell, Ganen, and >errecchia 41 5 .

The Informativeness of rice +e have suggested on several


43/65

previous occasions that market price itself is partially informative

about the future value of the firm. In particular, price is informative

about Si.e., leads earnings. @ecall that the reason is that market

price aggregates all publiclv known lnformation about the firm, much

of which the accounting system recogni6es with a lag. onse-uently,

the more informative is price, the less will be the information

content of current accounting earnings, other things e-ual. hence

the lower the (@e.


44/65

0 proxy for the informativeness of price I% firm si:e! since larger

firms are more in the news. 7owever, (aston and \mi2ewski

4l $ found that firm si6e was not a significant explanatory

variable for the (@ . The reason I% probably that firm si6e

proxies for other firm characteristics, such as risk and growth, as

much as it proxies for rhe informativeness of share price. &nce

these facrors are controlled for, any significant effect of si6e on

the (@ seems to go away. ollins and =othari 41 $ dealt

with si6e by moving the wide window over which security

returns were measured earlier in time 4or large firms, on the

grounds that share price is more informative for such firms.

They found rhar this substan3 tially improved the relationship

between changes in earnings and securiry returns. since a more

informative share price implies ;hat the marker anticipates

changes in earning power sooner. &nce this time shifting was

done, si6e appeared to have no explanatory power 4or the (@e.

5. .8 Implications of (@ @esearch

!e sure char you see the reason why accountants should be

interested in the market's response to financial accounting

information. (ssentially, the reason is that improved

understanding of market response suggests ways that they

can further improve the decision. usefulness of financial


45/65

statements. or example. lower informativeness of price for

smaller firms implies chat expanded disclosure for these

firms would be useful for investors, contrary to a common

argument that larger firms should have greater reporting

responsibilities.

0lso, the finding that (@ s are lower for highly levered

4irms supports arguments to expand disclosure of the nature and

magnitude of financial instruments, induding those that are

off3balance3sheet. If the relative %i6e of a firm's liabilities

affects rhe market's response to net income, then it is desirable

that all liabilities be disclosed.

The importance of growth opportunities to investors suggests,

for example, the desir3

abilirv of disclosure of segment information, since profitabi 7ty

information by segments would better enable investors to isolate

the profitable, and unprofitable, operarions of the firm. 0lso,

< B0 enables the firm ro communicate irs growth prospects,

as illusrrared in %ection .$.8.

inally, the importance of earnings persistence to the (@e

means that disclosure of

the componentsof net income is useful for investors. In sum, lots

of detail in the income statement, the balance sheet, and in

supplemental inforrnarton helps investors interpret the currentearnings number.


46/65

5. .)


47/65

it does not. Thus, obtaining a reasonable estimate of earnings

expectations is a crucial component of information approach

research.

:nder the ideal conditions of (xample 8.8, expected

earnings is simply accretion of discount on opening firm value.

+hen conditions are not ideal, however, earnings expec3 tations

are more complex. &ne approach is to pro2ect the time series

formed by the firm's past reported net incomes. chat is, to base

future expectations on past performance. 0 rea3 sonable

pro2ection, however, depends on earnings persistence. To see

this, consider the extremes of 1EEW persistent earnings and

6ero persistent earnings. Hf earnings are corn3 plerelv persistent.

expected earntngs for the current year are H:%tlast year'sactual earn3 ings. Then, unexpected earnings are estimated as

the dulnge from last year. This approach was used by !alland

!rown, as described in %ection 5.). If earnings are of 6ero

persistence, then there is no information in last year's earrungs

about future earnings, and all of cur3rent earnings are unexpected, That I%, unexpected earnings

are e-ual to the Hewlof cur3

rem year's earnings. This approach was used by !ill autious in

(xample ).1.

+hich extreme is closer to the truth This can be evaluated by the degree of corre3 lation between security returns and the


48/65

estimate of unexpected earnings, a -uestion exam3 ined by

(aston and 7arris 4l 1 . :sing regression analysis of a large

sample of :.%. firms

over the period 1"#"31"$#, they documented a correlation

between one3year security

returns and the change in net income, consistent with the

approach of !all and !rown. 7owever, there was an even

stronger correlanon between returns and the level of ner

income. unhermore, when both earnings changes and levels

were used, the two variables combined did a %ignificantly better

2ob of predicting returns than either variable sepa3 rately.

These results suggest that the truth is somewhere in the

middle, that is, both changes in and levels of net income are

components of the market's earnings expecca3 dons, where the

relative weighcs on the two components depend on earnings

persistence.

The foregoing discussion is based solely on a time series

approach, however. 0nother source of earnings expectations isanalysts' forecasts. These are now widely available for mosr

large firms. If analysts' forecasts are more accurate than time

series forecasts, they provide a bener estimate of earnings

expectations, since rational investors will presum3 ably use

the most accurate forecasts. (vidence by !rown,7agerman, Vnffin, and \rni2ewskr 41 $/ , who studied the


49/65


50/65

the average ignored how old ;he individual forecasts were. This suggests

that ;he timeli3ness of a forecast dominates the cancelllng3our3of3errors

effect of the average forecast.espite evidence that analysts' forecasts tend to be more accurate

;han forecasts

based on time series, ocher evidence, discussed by =othari 48EE1 ,

suggests that analysts' forecasts are optimistically biased. although the

bias may have decreased in recent years. Nevertheless. recent studies of the information content of earnings tend to base earnings expectations

on analysts' 4orecasrs.

Theory in ractice 5.1

isco %ystems Hnc.Is a lar,e provider

of network3 ing e-uipment, based

in %an Hose, alifornia. In 0vgust

8EE , Ii reieaMd flnandal resultsfor the -uarter ended Huly )E,

8EE . Its revenues increased by

8#W over the same -uarter of 8EE).

Its net income for the -uarter was

A1. bi2lionor 81 cents per share. a 1 W increase

over the same -uarter of 8EE). and

5W in excess of the average analysts'forecast of 8E cents,

>et. isco's share price fell almost

1$W to

A1$.8 following the

announcement This fall in priceseems contrYlry' to tile results of

!all and !rown and subse-uent

researchers. who have documented

a positive market response to good

earnings news. 7owever, certain balance sheet and supplemental


51/65

information was not so 3

favourable. or example,

inventory turnover declined to #.

from #.$ in 8.EE). gross margin

declined slightly. order backlog wasdown and, while revenue +as

growing, its rate of growth

appeared to be declining. 0lso.

several analysts. commented on an

increase in irwentories, 5':g3gesting lower earnings persistence

and accrual -uality to the extent

these inventorilO%would be slow in

se1lilng. urthermore, isco's (&,

com3 menting on me -uarter'sresults. mentioned that the firm's

customers were becoming more

4au3 tious about spending.

These n,egative slgna;s implied low

-uality and negative persistence for the good earnln-s news. probably


52/65


53/65

empirical research has demon3

seated a differential market

response depending on most

of these factors. These ernpiri3

cal results are really -uite

remarkable. irst, they have

overcome substantial stariscical

and experimental design

problems. %econd, they show

that the marker is, on average,

very sophisticated in its ability

to evaluate accounting

information. This supports the

theory of securities marker

efficiency and rhe decision

theories that underlie it.

inally, they sup3 port the

decision usefulness approach tofinancial reporting.

Indeed, the extent to

which historical cost3based

net income can provide "clues"

about future firm performance

may seem surprising. The key, of

course, is the information svsrern

probabilities, as shown in Table

).8. In effect, the higher the

main diagonal prob3 abilities, the

greater we would expect the (@

to be. This supports the 0%!'s

conrenrion in irs onceptual

ramework that investors'

expectations are based "at least

partly on evaluations of past

enterprise performance" 4%ection

).$ . 0s accountants gain a

better IlO9lderstanding oS ilwe"to;

;MpoIP%e to financial statement

information, their abiPitM'to provide useful information to

investors will further increase.


54/65

0 0>(0T 0!&:T T7( "!(%T" 0 &:NTINV &GI ?

To chis &7P , we have argued that accountants can be guided bysecurities market reaction in determining usefulness of financial

accounting information. rom this, lr is tempting to


55/65

conclude mac the best accounting policy is the one that produces

the greatest market price response, or example, if net income

reported by oil and gas firms under successful3efforts accounting

produces a greater market reaction than net income reported

under full3cost accounting, successful3efforts should be used,

because investors find it more useful.

7owever, we must be extremely careful about this

conclusion. 0ccountants may be better off to the extent that

they provide useful information to investors, but it does not

follow that society will necessarily be better off,

The reason is ;hat information has characteristics of a public

good. 0 public good is a good such that consumption by one

person does not descroy it for use by another. onsumption of a private good3such as an apple3eliminates its usefulness

for ocher consumers. 7owever, an investor can use the.

information in an annual report without eliminating its

usefulness to other investors. onse-uently, suppliers of public

goods mayhave trouble charging for these products, so that we often

wtrness them being supplied by

governmental or -uasi3governmental agencies3roads and

national defence, for example, ;f a firm tried to charge investors

for its annual report, ir would probably not attract manycustomers, because a single annual report, once produced, could


56/65

be downloaded to manv users. Instead, we observe governments

through securities legislation and corporations

acts, re-uiring firms to issue annual reports.

&f course, firms' annual reports ate not "free." roduction of

annual reports is costly. &ther, more significant, costs include

the disclosure of valuable informarion to competi3 tors and the

likelihood that manager operating decisions will be affected by

the amount of information about those decisions that has to be

released. or example, managers may curtail plans for

expansion if t&& much information about them has to be

disclosed.

Investors will eventually pay for these cosrs through higher

product prices. Nevertheless, investors perceive annual reporrs asfree, since the extent ro which they use them will not affect the

product prices they pay. 0lso, investors may incur &%t% 4E

inform themselves,

either directly, or indirectly by paying for analyst or other

informarlon services, Nevertheless, the basic "raw material" is perceived as free and

investors will do what any other rational. consumer will do

when prices are 4ow3Yonsume more of it, 0s a result, invest)rs

may perceiveaccounting inSYnmacionas useful eFvenr.ho5t,h;rom socie&y#s

standpoint the roS'S ofdtis i.nJormonon outwei,h tile $enefits to illves5( 3#s.0lso, as mentioned in hapter I, information affects


57/65

different people differently. Thus, informarion may be useful to

potential investors and competitors but current share3 holders

may be harmed by supplying it. 0s a result, the social value of

such information depends on both the benefits to potential

investors and competitors, and the o%t% to shareholders. %uch

cost3benefic tradeoffs are extremely difficult to make.

Think of information as a commodity, demanded by investors

and supplied by firms

rhrough accountants, !ecause of the public3good aspect of

information, we cannot rely on the forces of demand and supply

co produce the socially "righc' or first3best amount of pro3 duction,

as we can 4or private goods produced under competition. The

essential reason is that the price system does not, and probably

cannot, operate ro charge investors ;he full costs of the

information they use. onse-uently, from a social perspective

we cannot rely


58/65

on the extent of security market response to tell us which

accounung policies should be used 4or, e-uivalenrlv, "how much"

information to produce . ormal arguments to %:pp&7 chis

conclusion were given bv Vonedes and opuch 41 / .

+e will rerum to the -uestion of regulation of information

production in hapters 18 and 1). or now, the poinc co reali6e

is that it is still true that accountants can be guided

by market response to maintain and improve their competitive position as suppliers to the marketplace 4or information. It is also

true that securities markets will work better to the extent

security prices provide good indications of underlying real

investment opportuni3 ties. 7owever, these social considerations

do suggest that, as a general rule, accounting standard3setting bodies should be wary of using securities market response to

guide their decisions.

Interestingly, an exception to this rule seems to have occurred

with respect to standard setters' decisions to eliminate current

cost accounting for capital assets. % 0% )), which re-uired :.%.firms to report supplemental current &%tinformation for certain

assets, was

discontinued in 1 $#. iscontinuance was based in part on the

influential study by !eaver and Gandsman 41 $) , who failed to

find any incremental securities market reaction to current cost


59/65


60/65

earlier. espite the relevance of @@0 information 4%ection

8. , studies by


61/65

as announcemencs of discoveries. and analyst forecasts. 0lso.

the point in time that the market first becomes aware of che

@@0 inforrnarion is often unclear. or net income. media

publication of the earnings announcement provides a

reasonable event date. 7owever. given the inside nature of oil

and gas reserves information and its irnportance ;_ firm value,

analysts and others may work particularly hard to ferret it &:t in

advance of the annual report. If a reasonable event date for the

release of other financial statement

information cannot be found. return studies must use wide

windows, which are open to a large number of influences on

price in addition to accounting information.

7owever. there is an indirect approach to finding evidenceof usefulness that links other tnformarion ;_ the -uality of

earnings. To illustrate, suppose that an oil company reports high

earnings this year, but supplemental oil and gas information

shows that its reserves have declined substantially over the

year. 0n lnrerpreration of this informationis that the firm has used up its reserves to increase sales in the

short run. If %_, the -uality of current earnings is reduced. since

they contain ) non3persistent component rhar will dissipate if

new reserves are not found. Then, the market's reaction to the

bad news in the supplemental reserve information may be moreeasily found in a low (@e than in a direct reaction to the reserve


62/65

information itself. onversely, a higher (@e would be expected

if reserves had increased.

This approach was generali6ed by Gev and Thiagara2an 41 )

4GT. They identified

18 "fundamentals'' used by financial analysts in evaluating

earnings -uality. or example, one fundamental was the change

in inventories, relative ro sales. If inventories increase. this may

suggest a decline in earnings -uaOity3the firm may be entering

a period of low sales. or simply be managing its inventories less

effectively. &ther fundamentals include change in capital

expenditures. order backlog, and. in the case of an oil and gas

company, the change in its reserves.

or each firm in their sample, GT calculated a measure of

earnings -uality by assign3

ing a score of 1 or E to each of that firm's 18 fundamentals,

then adding rhe scores. or example, for inventories, a 1 is

assigned if that firm's inventories. relative to sales, are down

for the year. suggesting higher earnings -uality, and a E score isassigned if invento3 ries are up.

+hen GT added these fundamental scores as an additional

explanatory variable in an

(@e regression analysis. there was a substanual mcrease in

ability to explain abnormal security returns beyond theexplanatory power of unexpected earnings alone. This sug3


63/65

gests that the market, aided perhaps by analysts. is -uite

sophisticated in its use of balance sheet and supplementary

information. Instead of a direct reaction to this information, it

seems to use it to augment the information content of earnings.

5.$ &N G:%I&N% &N T7( IN &@


64/65

response to reported net income is impressive in terms of its

sophistication. @esults of empirical research in this area support the

efficient markets theory and underlying deci3 sion theories.

The market does not seem 4E respond to other financial

statement informanon as strongly as it does to earnings information.

The extent 4_ which the lack of strong market response to this

other information is due ro methodological difficulties, co low

reliability, to availability of alternative informacion sources, or co

failure of efficient markets theory itself is not fully understood,

although it may be char investors anticipate balance sheet and

supplementary information to fine3rune the (@e, rather than using

such information directly.

0s stated earlier, the approach 4E financial accounting theory

that e-uates the extent of security price change with information

concern and hence with decision usefulness is known as the

Information approach. The essence of this approach is that investors

are viewed as attempting to predict future returns from their

investments. They seek all rele3 vant information in this regard, not 2ust accounting information. To maximi6e their competitive

position as suppliers of information, accountants may then seek co

use the extent of security market response to vanous types of

accounting information as a guide to its usefulness to investors.

This motivates their interest in empirical research on decisionusefulness. urthermore, the more information accountants can move


65/65

chapter 5 scott

Documents