can quantitative screening of uk smaller company portfolios result in superior investment returns
DESCRIPTION
My MBA Dissertation from Imperial College, London examining the relationship between fundamental factors (financial characteristics) and share price outperformance.TRANSCRIPT
IMPERIAL COLLEGE OF SCIENCE, TECHNOLOGY AND MEDICINE
(University Of London)
The Management School
Can Quantitative Screening of UK Smaller Companies Result in
Superior Investment Returns?
By
Ian J. Lancaster
A report submitted in partial fulfilment of the requirements for the MBA
degree and the DIC
September 1999
“Money machines don’t exist – not for long, anyway”
Brealey and Myers
“There is nothing so dangerous as the pursuit of a rational
investment policy in an irrational world”
John Maynard Keynes
“Its different this time”
Anon
Contents
CONTENTS
Page
LIST OF FIGURES vii
SYNOPSIS x
DECLARATION xi
ACKNOWLEDGEMENTS xi
1.0 INTRODUCTION 1
1.1 Background 3
1.2 Aims of this study 4
1.3 Data sources 4
1.4 Definitions 5
2.0 LITERATURE REVIEW 7
2.1 Introduction 8
2.2 The CAPM and efficient markets 9
2.3 The APT and efficient markets 11
2.4 Evidence of market inefficiencies 12
2.4.1 The price to earnings ratio 12
2.4.2 The price to book ratio 16
2.4.3 The price to sales ratio 20
i
Contents
2.4.4 The return on equity 21
2.4.5 Equity market value 22
Page
2.4.6 Other ratios associated with market inefficiency 23
2.4.6.1 Yield 23
2.4.6.2 Debt to equity ratio 23
2.4.7 Patterns in equity price reaction 26
2.4.7.1 Past share price performance 26
2.4.7.2 January effect 28
2.5 The information content of reported earnings 29
2.6 Evidence for the sustainability of accounting ratios 33
2.6.1 Return on equity 33
2.6.2 Earnings per share growth 33
2.7 The use of earnings forecasts 37
2.7.1 Earnings models 41
2.7.2 Cash flow models 42
2.7.3 Other predictive models 45
2.7.3.1 The T-Model 45
2.7.3.2 Jacobs and Levy Model 46
2.7.3.3 Brennan and Subrahmanyam Model 46
2.8 Are the market inefficiencies a proxy for risk? 47
2.9 Conclusion of the literature review 51
ii
Contents
Page
3.0 INTRODUCTION TO EQUITY VALUATION 60
3.1 The dividend discount model 60
3.1.1 The relationship between the DDM and accounting
ratios 62
3.1.2 Price to earnings 64
3.1.3 Price to book value 66
3.1.4 Price to sales 67
3.1.5 Debt to equity ratio 69
3.2 Cash flow valuation 70
3.3 Conclusion of the introduction to equity valuation 71
4.0 DETERMINANTS OF COMPANY FINANCIAL PERFORMANCE 73
4.1 Equity growth rate 73
4.2 Issues of cash flow 75
4.2.1 Depreciation charge 75
4.2.2 Working capital 75
4.2.3 Free cash flow 76
4.3 The value drivers 77
4.4 Operational gearing 81
4.5 Risk of failure 82
4.6 A test for correlation between accounting ratios and outperforming
companies 84
iii
Contents
Page
5.0 RANKED PORTFOLIO TESTS FOR MARKET INEFFICIENCY :
UK SMALLER COMPANIES 86
5.1 Tests for market inefficiencies 87
5.2 Methodology 89
5.2.1 Definition of FTSE Smaller Company Index 89
5.2.2 Data collection 90
5.2.2.1 Constituents of the proxy smaller company
index 90
5.2.2.2 Portfolio creation dates 92
5.2.2.3 Time period 96
5.2.2.4 Test portfolios 96
5.3 Consideration of the significance of the results 98
5.3.1 The Student’s t test 98
5.3.2 Portfolio risk 100
5.3.3 Stage of business cycle 102
5.4 Portfolio ranking, performance and analysis 103
5.4.1 The proxy universe 103
5.4.2 Price to earnings 105
5.4.2.1 Discussion 111
5.4.3 Price to book 114
5.4.3.1 Discussion 117
iv
Contents
Page
5.4.4 Price to sales 118
5.4.4.1 Discussion 121
5.4.5 Return on equity 122
5.4.5.1 Discussion 123
5.4.6 Cash EPS % of EPS 125
5.4.6.1 Discussion 126
5.4.7 Dividend Yield 127
5.4.7.1 Discussion 129
5.4.8 Share price momentum 129
5.4.8.1 Discussion 131
5.5 Summary of single factor tests 132
5.6 Combination portfolios 134
5.6.1 PER and momentum 135
5.6.1.1 Discussion 136
5.6.2 PSR and momentum 136
5.6.2.1 Discussion 137
5.7 Summary of combination portfolios 138
5.8 Conclusion of data analysis 139
v
Contents
Page
6.0 CONCLUSION AND FURTHER STUDY 143
6.1 Can the quantitative screening of UK Smaller Companies result in
superior investment returns? 143
6.2 Further study 145
7.0 REFERENCES 147
8.0 APPENDICES 158
8.1 Test of whether the proxy index companies perform in line with
the FTSE Smaller Company Index
8.2 Test of portfolios by single factor
8.3 Test of combination portfolios
8.4 Regression of ROE on Log PTBV
8.5 Test of high ROE, low PTBV screen
vi
List of Figures
LIST OF FIGURES
Page
Figure 2.01: Average Annual Return by PER Quintile. Compustat data
1956-1971 13
Figure 2.02: One-Year Total Return by PER Quintile. Compustat data
1968-1990 14
Figure 2.03: One-Year Total Return by PTBV Quintile. Compustat data
1963-1990 17
Figure 2.04: Average Excess Annual Return by Ranked MV and PTBV Quintile.
Compustat data 1982-1992 18
Figure 2.05: Price to Book Value against Beta 19
Figure 2.06: Compounded Annual Returns for Varying Balance Sheet Strength 25
Figure 2.07: Cumulative Return on a Neutral, Six-Month Relative Strength,
Winner/Loser Portfolio. Compustat data 1965-1989 27
Figure 2.08: Average Return 1926-1986 of 5-Year Relative Strength Stocks
by Calendar Months 28
Figure 2.09: Monthly Difference between Winner and Loser Portfolios at
Earnings Announcement Dates 31
Figure 2.10: Relative Subsequent EPS Growth in Quintiles of the EP Ratio 35
Figure 2.11: Earnings Surprise versus PE Ratio Compustat Data 1980-1994 37
Figure 2.12: Studies of the Accuracy of Analysts Earnings Estimates 38
Figure 2.13: Studies on the Accuracy of Analysts Earnings Estimates versus
Mechanical Earnings Models 39
vii
List of Figures
Page
Figure 2.14: Zero Net Investment, Long Low Accrual, and Short High Accrual.
Compustat data 1962-1992 44
Figure 2.15: The relationship between Beta and the PE Ratio 47
Figure 2.16: The Annual Returns of High vs. Low Volatility Portfolios 49
Figure 2.17: Annualised Excess Return for Investment Strategies 55
Figure 5.01: Market Capitalisation Cut-Off Points for the Proxy Smaller
Company Index 91
Figure 5.02: Year-End Distribution for December 1994 Proxy-Index 94
Figure 5.03: Portfolio Creation Dates 95
Figure 5.04: Portfolio Creation Dates 95
Figure 5.05: Portfolio Creation Dates 95
Figure 5.06: Portfolio Creation Dates 96
Figure 5.07: UK GDP Growth (% YOY) 6/1990 – 6/1999 102
Figure 5.08: Test of the Acceptability of Portfolios as a Proxy the FTSE
Smaller Company Index 104
Figure 5.09: Theoretical PE Ratios for High and Low Growth Companies 108
Figure 5.10: Performance of Portfolio Quartiles by PE Ratio 110
Figure 5.11: Performance of Portfolio Quartiles by PTBV Ratio 116
Figure 5.12: Performance of Portfolio Quartiles by PS Ratio 120
Figure 5.13: Performance of Portfolio Quartiles by ROE 123
Figure 5.14: Performance of Portfolio Quartiles by CEPS/EPS 126
viii
List of Figures
Page
Figure 5.15: Performance of Portfolio Quartiles by Dividend Yield 128
Figure 5.16: Performance of Portfolio by Share Price Momentum 130
Figure 5.17: Statistically Significant Results for the Single Factor Screens 132
Figure 5.18: Outperforming Single Factor Strategies 133
Figure 5.19: Performance of Portfolio by PER and Share Price Momentum 135
Figure 5.20: Performance of Portfolio by PSR and Share Price Momentum. 137
Figure 5.21: Performance of Combination Strategies 138
Figure 5.22: Sharpe Ratios of Outperforming Strategies 139
ix
Synopsis
SYNOPSIS
Academic research on stock market inefficiencies was surveyed to indicate which factors
might be relevant when looking for potential inefficiencies within the UK Smaller
Company Index.
The relationship between equity valuation, and the theoretical basis for company
financial performance was explored in order to lay the foundations for more complex
accounting based screens of UK Smaller Companies.
It was found that the quantitative screening of UK Smaller Companies by a simple
combination of an accounting ratio, and share price momentum resulted in an excess to
Index return of 5.2 % p.a. The trade off between the risk and return of this strategy was
superior to that of the FTSE Smaller Company Index. Nine other statistically significant
inefficiencies were recorded.
The results of the value screens for UK Smaller Companies are the opposite of the results
of previous research on the wider market index. This indicates that there are distinct
differences between profitable investment strategies for Smaller Companies as compared
to Larger Companies.
x
Declaration and Acknowledgements
DECLARATION
Plagiarism: the presentation of another person’s words, ideas, judgement or data as
though they were your own.
I have read the above definition of plagiarism. I am fully aware of what it means and I
hereby certify that, except where indicated, this project report is entirely my own work.
ACKNOWLEDGEMENTS
I would like to thank my supervisor, Dr. Colin Clubb, for his guidance and comments,
and the lesson that data always has a deeper meaning than the numbers alone would
suggest.
I would also like to thank my girlfriend, Amanda, for her support both before, and
throughout this course.
This project is dedicated to my father, like a true contrarian, he always supported the
underdog.
xi
Introduction
1.0 INTRODUCTION
1
Introduction
1.0 INTRODUCTION
Like modern day alchemists, portfolio managers attempt to refine the components of a
market index into a portfolio of equities that will outperform that index. There is
however, little evidence that the expensive process of fundamental research and careful
stock selection adds any value, in fact as a whole, managed portfolios underperform
(Malkiel (1995), Jensen (1968)). It could be argued that the complex interactions that
occur between economic forces, companies, management and those who value financial
assets is beyond human understanding. The inroads that chaos theory is making into the
world of financial theory seems to support this view. An interested reader should see
Fogler (1995) for an excellent synopsis of the emerging opportunities created by non-
linear forecasting models.
In the search for a leading edge in portfolio selection, it is becoming common for vast
databases to be interrogated for patterns between share prices and financial attributes. In a
similar vein, this study focuses on investigating patterns in the share price behaviour of
UK Smaller Companies. As UK Smaller Companies account for only 3.84% (31/8/99) of
the market capitalisation of the UK equity market, this data set is rarely examined by
academics.
For larger, more frequently traded companies, arbitrage would be expected to eliminate
valuation anomalies. However UK Smaller Companies are relatively under researched, so
2
Introduction
within this sub sector there may be the opportunity of achieving risk-adjusted excess
returns.
1.1 Background
The FTSE Smaller Company Index excluding investment trusts (FTSC X-IT), consists of
364 companies. These range in size from £30m to £406m. I worked within a team of 3
UK Smaller Company analysts on behalf of a large institution investor (assets of £15bn.,
£650m. in Smaller Companies). We considered that each individual could cover 35
companies sufficiently well to justify a holding within the portfolio, and follow the
fortunes of a further 20 companies for possible inclusion. Between three analysts, this
totaled only 165 companies. A key management problem was deciding which of the 364
index companies, should become part of the list of the 165 companies that were covered
by the analysts. It was therefore essential for a screening process to take place.
The investment screens were both quantitative and qualitative. Qualitative decisions
could only be made after meeting with the company management and exploring the
prospects for a company’s product or service. This was time consuming, and in any case,
it was not feasible to meet all of the companies in the Smaller Company Index. Due to
this, it was considered necessary to consider various quantitative screens in order to
eliminate certain ‘poor prospects’.
3
Introduction
1.2 Aims of this study
It is the primary aim of this study to establish whether it is possible to use quantitative
screens in order to achieve risk-adjusted excess returns for a portfolio of UK Smaller
Companies. A secondary aim is to increase the readers understanding of the interaction
between accounting ratios, equity valuation and the theoretical basis for equity
outperformance. In the Literature Review, academic research is surveyed in order to
establish whether market inefficiencies exist, and if so how best to exploit them. In the
portfolio screening tests of this study, a high-low ranking of companies by historic
accounting ratios is used to select portfolios that express a similar attribute. The total
return of these portfolios is then assessed against the total return of the FTSE Smaller
Company Index over the subsequent year in order to measure their performance.
1.3 Data Sources
Compustat is a US database of company account data and share prices from 1925 to
present day. It is the prime source of information for the academic literature.
Datastream is a UK database of financial and economic time series. The data set is used
in this study for accounting ratios and price information.
I/B/E/S is the Institutional Broker Estimate System. It is used extensively by Investment
Managers to track Analysts forecasts.
4
Introduction
1.4 Definitions
APT - Arbitrage Pricing Theory
Beta or β - The β coefficient of the CAPM
BV - Book value
BVP - Book value to price
BVPS - Book value per [ordinary] share
CAPM - Capital Asset Pricing Model
CEPS - Cash earnings per [ordinary] share
CFO - Cash flow from operations
DDM - The dividend discount model (Gordon growth model)
DER - The debt to equity ratio
DPS - Dividends per [ordinary] share
DYLD - Dividend yield
EP - Earnings yield (1/EPS)
EPS - Earnings per [ordinary] share
Excess Return - Provide a higher return (capital plus income) than the relevant stock
market index (capital plus income)
FTSC Index - FTSE Smaller Company Index
Micro-cap - Used to describe a very small quoted equity (sub £55m MV)
MV - Market value of equity
Net - After interest and tax
Outperform - Provide a higher return (capital plus income) than the relevant stock
market index (capital plus income)
5
Introduction
Payout ratio - DPS / EPS
PCEPS - Price to cash EPS
PEG ratio - Price to earnings ratio / EPS growth rate
PER - Price to earnings ratio
PS - Price to sales
PSR - Price to sales ratio
PTBV - Price to book value
Retention ratio- 1-Payout Ratio
ROCE - Return on capital employed
ROE - Return on ordinary shareholders equity
SPS - Sales per [ordinary] share
Underperform - Provide a lower return (capital plus income) than the relevant stock
market index (capital plus income)
6
Literature Review
2.0 LITERATURE REVIEW
7
Literature Review
2.0 LITERATURE REVIEW
2.1 Introduction
Early empirical research tested the association between the returns of portfolios against
the predicted returns of the CAPM framework (see Section 2.2, (Sharpe (1964), Lintner
(1965), and Black (1972)). Subsequently, Gibbons (1982) and Stambaugh (1982) tested
the hypothesis that expected returns were predicted solely by the Beta value. Following
the development of the APT (Ross (1976)), similar tests of this model were conducted.
Tests of the CAPM and APT models were based upon the theory that the excess return of
a portfolio was also dependent upon non-risk security characteristics. These included firm
size (Banz (1981)) book to market ratios, dividend yields, past sales growth, debt to
equity ratios (Bhandari (1988)), low share price (Stoll and Whaley (1983)), and earnings-
price ratios (Basu (1977,1983)). See also Black and Scholes (1974), Rosenberg, Reid and
Lanstein (1985), Lakonishok, Shleifer and Vishny (1994), and Haugen (1995).
In addition, relationships have been demonstrated between a reversal in long term returns
of three to five years (De Bondt and Thaler (1985)) and, in contrast, a tendency for short-
term (up to one year) performance to continue (Jegadeesh and Titman (1993), Asness
(1994)).
8
Literature Review
2.2 The CAPM and efficient markets
The Capital Asset Pricing Model (CAPM) has been an axiom of financial analysis since
its conception by Sharpe (1964) and Lintner (1965). The CAPM states that the expected
return of a stock is a function of the risk free rate of return plus the Beta (β) value of the
stock multiplied by the equity market risk premium.
Expected return = Rf +β (ERm – Rf)
Where Rf = The risk free rate of return (the return on long term bonds)
β A measure of the historic share price return to the historic market return
ERm = Expected return for the market
Beta is calculated from a regression of past stock returns on past market returns, and thus
describes the sensitivity of a stocks return to that of the market. The equity risk premium
is the compensation for the additional risk that investors take on when investing in the
stock market as opposed to a risk free asset. The risk premium is defined as the expected
return of the equity market minus the expected return from a risk free asset (i.e. a
government bond). As a stock with a high Beta has a high price variance compared to the
market, then consequently investors will require a higher return from the stock in order to
compensate for the higher risk (standard deviation of returns). By definition the market
has a β value of one, and government bonds a Beta value of zero.
9
Literature Review
A central principle of the CAPM is that arbitrage by investors (between theoretical and
actual value) is expected to keep all stocks at the correct expected return. In the most
literal of terms, the CAPM states that investors will be unable to achieve higher returns
for a particular portfolio, without taking on extra risk. Markets are therefore considered
efficient at equating risk with return. Brealey and Myers (1996) state that the CAPM’s
resilience is because “The Capital Asset Pricing Model captures [the idea of] risk in a
simple way.”
However critics of the CAPM theory point to evidence that market inefficiencies are
demonstrated to exist when Beta is controlled for. For example, if companies are sorted
into groups of similar Beta, and similar market capitalisation (MV), at a particular value
of Beta, the lower MV companies tend to outperform. If the outperformance is
compensation for the risk that the investors in low MV companies endure, then evidently
the risk is not captured by Beta alone. Various other inefficiencies are discussed in this
chapter and these suggest that the single risk factor of Beta is inadequate.
The CAPM theory has expanded to include varying Beta models. These are based upon
the theory that risk is not constant over time. Howton and Peterson (1998) have shown
that during bull (bear) markets, if Beta is allowed to vary, then Beta is significantly
positive (negative). However Ghysels (1998) warns that, in practice, the estimation errors
of varying Beta models can prove to be higher than stable Beta models.
10
Literature Review
2.3 The APT and efficient markets
The Arbitrage Pricing Theory (APT, Ross (1976)) splits stock risk into two components.
First, there is macroeconomic factor risk, that is the risk from economic or general factors
that influence equity return such as the oil price, inflation etc. (General factors may also
include potential inefficiencies such as the PER, PTBV etc). Second, there is the risk
from factors that are unique to a particular company, this is termed ‘noise risk’. The APT
assumes that, the noise risk can be reduced as a portfolio becomes more diversified.
Mathematically the APT can be expressed as,
Stock expected return, (Er) = a + b1 (Rp factor 1) + b2 (Rp factor 2) + … + noise
Where ‘a’ is the risk free rate of interest, ‘Rp factor n’ is the risk premium for economic
or general factor n (i.e. the excess return that a sample portfolio would have earned in the
past for taking on the particular factor risk), and ‘bn’ is the sensitivity of the return of a
stock to factor n. The value of ‘bn’ is calculated by regressing factor n on the historic
share price return.
The risk premium of a stock therefore depends upon the associated risk premium of each
factor and the sensitivity of the stock to each factor.
Stock risk premium = Er - Rf
= b1(Rp factor 1 – Rf) + b2(Rp factor 2 – Rf) + … + bn(Rp
factor n – Rf)
11
Literature Review
Where ‘Rf’ is the risk free rate of return (the return on government bonds).
There are no standard factors, so a standard model cannot be used to test for the
availability of excess returns vs. those predicted by the APT theory1. Chen et al. (1986)
have suggested that five macroeconomic factors are common, these are industrial
production, changes in default premium, shifts in the term structure of interest rates,
unanticipated inflation, and changes in the real rate of return. The factor specific nature of
APT makes it superior to the simple one factor (β) CAPM, however its mathematical
complexity means that, so far, the CAPM still dominates in terms of practical use.
2.4 Evidence of market inefficiencies
2.4.1 The price to earnings ratio
The price to earnings ratio (PER) is an established valuation screen. Its simplicity and
empirical success made it the cornerstone of early portfolio strategy. In his seminal book
“The Intelligent Investor” Benjamin Graham (1973) discusses the merits of low PER
stocks, but he warns against the use of low PER combined with low MV, because
“market neglect” [possibly due to stock illiquidity and lack of research] can result in low
PER and MV stocks remaining undervalued for some time.
1 Certain studies have concluded that relevant factors are: Yield spread (long govt. bond-30 treasury bill), Δ interest rate, Δ exchange rate, Δ real GNP forecast and Δ inflation forecast.
12
Literature Review
A number of studies (for example Basu (1977,1983) indicate that low PER stocks
outperform. In these studies it was hypothesised that a low PER effect could also be
associated with a size effect, so returns were compared against the size group of which a
company is a member. The results however, showed that the low PER effect continues
regardless of the size (MV) of the company.
Basu (1977) tested the low PER screen on 1400 companies from the NYSE between 1956
and 1971. He used the CAPM framework to show that low PER stocks earn excess
returns relative to the market. The results of the study are shown below.
Figure 2.01: Average Annual Return by PER Quintile. Compustat data 1956-1971
PER Quintile Average Annual Return BetaA (Highest) 9.3% 1.1121A* 9.6 1.0579B 9.3 1.0387C 11.7 0.9678D 13.6 0.9401E (Lowest) 16.3 0.9866
Source: Basu, S., 1977, “The Investment performance of Common Stocks in Relation to their Price-Earnings: A test of the Efficient Markets Hypothesis,” Journal of Finance, 32, pp. 663-682.
A* = Highest PER quintile excluding stocks with negative earnings.
One argument against the use of a low PER strategy is that low PER stocks outperform in
order to compensate investors for higher risk. This is partly refuted by the demonstration
that lower PER stocks, in fact, have lower Beta values (of course, this depends upon ones
acceptance of the CAPM). If Beta captures risk then it would seem, from this study, that
there is an inverse relationship between risk and reward. See Figure 2.01 above.
13
Literature Review
In one of the first papers to focus on contrarian investment strategies, Lakonishok et al.
(1994) examined the Compustat data over the period of April 1968 to April 1990, and
formed portfolios based upon ranked quintiles of PE ratios. The results are shown below.
Figure 2.02: One-Year Total Return by PER Quintile. Compustat data 1968-1990
Source: Dreman, D.,N., 1998, “Contrarian Investment Strategies: The Next Generation,” Simon and Schuster, New York, p. 154. Adapted from Lakonishok, J., Shleifer, A., and Vishny, R., 1994, “Contrarian Investment, Extrapolation, and Risk,” Journal of Finance, 49, pp. 1541-1578.
It can be seen that over the period of 1968 to 1990, a low PER strategy would have
outperformed by an average of 2.7% p.a.
In an extension of the preceding studies, Goodman and Peavy (1983) tested stocks with a
PER that was low relative to the PER of their industrial categorisation, the low PER
effect continued.
14
Literature Review
There are various arguments for the (Beta controlled) low PER phenomena.
First, is that the CAPM consistently underestimates the Beta values of low PER stocks
and overestimates the Beta values of high PER stocks; however as low PER stocks are
often large, stable (and consequently less ‘risky’) companies this argument is not
compelling.
Second, is that the tax considerations of investors cause the anomaly as investors are
adverse to the high yield of low PER stocks (Ball (1989)); however tax rates on capital
gain and income are now at similar levels and the effect continues.
Third, and most compelling, is that investors systematically pay too much for growth
(high PER) companies. The reassurance of being part of ‘the crowd’, pushes many
investors into the same limited number of growth companies, and consequently they
force the PE ratio to (theoretically) high levels. It is for this reason that those who follow
a low PER strategy are called ‘contrarian’ investors; they are happy to go against the
crowd.
To try and eliminate the possibility of the overvaluation of high PER stocks, Peters
(1991) tested the strategy of buying portfolios of low “PER / EPS Growth” companies.
This screen was applied in order to incorporate a basis of ‘reasonable value’ i.e. high
PER stocks could be held if the earnings growth rate was correspondingly high. The
lowest decile outperformed in 26 out of the 30 quarters measured, and provided a return
of 1,536% vs. 356% for the S&P 500 over the seven and a half year period.
15
Literature Review
Damodaran (1994) suggested that significant explanatory factors for a regression
calculation of the PE ratio in any one year, include the payout ratio, β, and the earnings
growth rate. He found that for the individual years of 1987 to 1991, the R2 ranged from
32% to 92%.
However the regressions are unstable year by year, quite possibly because investors
‘price in’ the effects of the business cycle on earnings. I.e. investors will pay a high PER
for a company at the low point of its earnings cycle and visa versa. Hence whilst the PER
regressions may demonstrate relationships, they are poor long-term predictors.
2.4.2 The price to book ratio
Rosenberg, Reid, and Lanstein (1985) looked at the period of 1973 to 1984 and found
that low PTBV portfolios provided an excess return of 36 bp. per month. In addition,
Fama and French (1992) further confirmed the relationship in a study of the period 1963
to 1990. They found that a low PTBV portfolio earned an average monthly return of
1.83% vs. an average monthly return of 0.3% for a high PTBV portfolio.
16
Literature Review
Figure 2.03: One-Year Total Return by PTBV Quintile. Compustat data 1963-1990
Source: Dreman, D.,N., 1998, “Contrarian Investment Strategies: The Next Generation,” Simon and Schuster, New York, p. 152. Reference: Fama, E.,F., and French K.,R., 1992, “The Cross-Section of Expected Returns,” Journal of Finance, 47, pp. 427-466.
The low PTBV quintile outperformed the market on average by 4.6% annually, whilst the
high PTBV quintile underperformed the market by 5.7%.
In order to counter the arguments of survivorship bias (often databases delete the history
of failed companies) in the Fama and French paper, De Silva (undated) used the
Compustat data for the 1982-1992 period (for which survivorship bias was not a
problem). In his study he segmented the sample to see whether there was an underlying
size effect. The results are illustrated below.
17
Literature Review
Figure 2.04: Average Excess Annual Return by Ranked MV and PTBV Quintile.
Compustat data 1982-1992
Source: Adapted from Haugen, R.,A., 1995, “The New Finance: The Case against Efficient Markets,” Prentice Hall, Englewood Cliffs, New Jersey, p. 5. Reference: De Silva, H., “What Underlies the Book-to-Market Effect,” Working Paper, Graduate School of Management, University of California, Irvine.
In the De Silva paper the size effect is quite evident, however the results have been
questioned as the small MV quintile may be based upon very small illiquid companies.
This study will incorporate a market value cut off in order to insure that a tradable
strategy is formulated.
Interestingly, Capaul, Rowley and Sharpe (1993) found that between 1981 and 1992, a
low PTBV portfolio earned excess returns in every international market that they
analysed.
18
Literature Review
Fama and French (1992) hypothesised that a low PTBV is a proxy for risk. This is
because low PTBV stocks are more likely to be in financial trouble. However during the
course of their study they found that lower PTBV stocks in fact had lower Beta values
(see also Basu (1977), Figure 2.01 above, in relation to low PER).
Figure 2.05: Price to Book Value against Beta
Source: Fama, E.,F., and French, K.,R., 1992, “The Cross Section of Expected Stock Returns,” Journal of Finance, 47, pp. 427-466.
Damodaran (1994) found that the most significant explanatory factors for the PTBV ratio
are the payout ratio and earnings growth rate. For the period of 1987 to 1991 the
predicted PTBV ratio was stable year on year i.e. the model had a stable R2 of between
86.0% to 88.5%.
There is a theoretical relationship between the return on equity (ROE) of a company and
the PTBV ratio at which it should be priced (see Section 3.1.3). To test this relationship,
Damodaran screened all NYSE stocks from 1981 to 1990 into “top 25% ROE – bottom
19
Literature Review
25% PBV” and “bottom 25% ROE – top 25% PBV” portfolios. Over the period the first
portfolio outperformed by 8.1% and the second portfolio underperformed by 6.8%.
2.4.3 The price to sales ratio
Senchack and Martin (1987) conducted one of the few tests of the performance of low
price to sales ratio (PSR) portfolios. They used the Compustat database to study the
returns of US stocks (AME and NYSE) between 1976 and 1984. Their conclusion was
that low PSR companies outperformed the market, but low PER was both dominant and
more consistent than low PSR.
Minard (1984) has argued that a low PSR strategy is superior to a low PER strategy as,
1) Sales are more stable over time.
2) The PSR can be used when companies are loss making.
3) That a low PER strategy is distorted by the elimination of high PER stocks that have
temporarily depressed earnings, and the selection of low PER cyclical stocks at the
peak of their earnings cycle.
In a study of the PS ratio that used regression analysis, Jacobs and Levy (1988) found
that a low PSR factor was significant in providing an excess return of 0.17% a month, it
also remained a significant factor when PER and MV were included.
20
Literature Review
There is a theoretical relationship between the PSR and net profit margin (see Section
3.1.4.). To exploit this relationship, for the period of 1981 to 1990, Damodaran (1994)
created “undervalued” portfolios consisting of the lowest quartile PSR stocks and highest
quartile net profit margin stocks, and “overvalued” portfolios consisting of the highest
quartile PSR stocks and lowest quartile net profit margin stocks. Damodaran’s
“undervalued” portfolio outperformed the S&P 500 by 6.2% and “overvalued” portfolio
underperformed by 2.0%.
In addition, for the period of 1987 to 1991, cross sectional regressions were performed
on the net profit margin, dividend payout ratio, Beta, and the growth rate in earnings.
The R2 ranged from 44% to 88%, however it is not stable over time.
2.4.4 The return on equity
Wilcox (1984) found that a strong cross sectional relationship existed between the Log of
price to book value (PTBV) and return on equity (ROE). The derived formula was,
Log PTBV = α + β1ROE
Where α and β1 are factors from a multiple regression of ROE on Log PTBV for a single
year.
The Wilcox PTBV – ROE model is based on the relationships that define the expected
growth rate of book equity (see Section 4.1). Wilcox suggests that analysts should use the
21
Literature Review
model to compute a forecast PTBV ratio based upon the consensus ROE. Stocks with a
lower PTBV ratio than that forecast by the model should be bought. For the period of
1976 to 1980, the Wilcox model demonstrated a lower mean squared error than a PER
regression model when used for predicting the value of US companies.
This model is useful, as unlike the earnings growth rate, past ROE has been shown to be
a good predictor of future ROE (see Babcock (1980), Paynor (1966) and Sections 2.6.1
and 2.6.2). However, the PTBV – ROE relationship does not hold for stocks were the
past ROE is not a good predictor of the future ROE, so it should only be used for
companies with a five-year standard deviation in ROE < 0.05.
2.4.5 Equity market value
Over extended periods, size (MV) has be shown to be a better predictor of expected
returns than Beta (Banz (1981), Reinganum (1981), Keim (1983), Fama and French
(1992,1993,1996)). Ball (1992) has suggested that the size anomaly is a candidate for an
expected return model as,
1) The information is publicly available, but the anomaly persists
2) Information processing costs are likely to decrease as market size increases
3) As there is a correlation between size and the yield component of expected return
then taxation of income becomes a factor
4) It is possible that Beta estimates are skewed by the non representative sample of
assets in equity indexes, Stambaugh (1982)
22
Literature Review
5) Bid-ask spreads are a decreasing function of size
6) The size effect is possibly as a consequence of short-interval Beta measurement,
Handa et al. (1989)
2.4.6 Other ratios associated with market inefficiency
2.4.6.1 Yield
The result of a high yield (DYLD) screen is counter-intuitive. It is assumed that high
yield stocks are predominantly in low growth industries, however high yield stocks tend
to outperform. The results of a study by Dreman (1998) on the Compustat 1500 data base
(largest 1500 US stocks) showed that for the 27 years ending December 1996, a high
quintile DYLD portfolio that was re balanced annually, provided average annual returns
of 16.1% vs. 14.9% for the market. [The second highest DYLD quintile is the best
performer, with an average annual return of 17.5%]. The downside protection of high
yield was also demonstrated in the study, as for all down quarters (bear markets) the high
DYLD strategy averaged a loss of only 3.8% vs. a loss of 7.5% for the market.
2.4.6.2 Debt to equity ratio
In general, a higher debt to equity ratio (DER) would be expected to result in greater risk
for the common equity holder. This is because when the DER increases, so does the
interest cost as a proportion of operating profits. As debt interest is in effect a fixed cost,
then if operating earnings were to fall (rise) the consequence of debt would be to amplify
profits negatively (positively) at the net level. However, Bhandari (1988) demonstrated
that after controlling for Beta and firm size, the expected common stock return is
23
Literature Review
positively related to the ratio of debt to equity. In an extension to this line of reasoning,
Ross-Healy and Sgromo (1993) showed that portfolio returns can be enhanced by
decreasing exposure to companies with balance sheet excess (defined as both under and
over leveraged), increasing exposure to companies with improving balance sheet
strength, and decreasing exposure to companies with deteriorating balance sheet strength.
The data was taken from S&P 500 companies in the period of 1983 to 1990 and the
results are shown below. The data shows that for a portfolio of companies that had
improving Balance Sheet strength, an excess return of 7.7% p.a. was achieved.
24
Literature Review
Figure 2.06: Compounded Annual Returns for Varying Balance Sheet Strength
Source: Adapted from, Ross-Healy, C., and Sgromo, E., 1993, “How to Beat the S&P 500 Index Using Credit Analysis Alone,” Journal of Portfolio Management, Winter, pp. 25-31, Exhibit one.
NB: A1 C denotes Strong Weak Balance Sheet
pA1 pB denotes balance sheet strength quartile promotion, eg pA1 indicates promotion from A to A1, pA indicates promotion from B to A.
dA dC denotes balance sheet strength quartile demotion, eg dA indicates demotion from A1 to A, dB indicates demotion from A to B.
The S&P 500 returned 9.78% compound during the period.
Whilst the Ross-Healy and Sgromo paper uses a specific ‘SAC’ score (developed by
Solvency Analysis Corporation), the issues can be related to other measures of financial
strength available in the UK. For example, Professor Richard Taffler, of the City
University Business School (1997) found the Z-Score rating of Syspas (UK) to be useful
in predicting company failure, see Section 4.5.
25
Literature Review
2.4.7 Patterns in equity price reaction
2.4.7.1 Past share price performance
De Bondt and Thaler (1985) propose a contrarian trading strategy that buys stocks that
have performed poorly over a 3 or 5 year period. Over the subsequent 3 or 5 year holding
period these stocks go on to outperform those that had performed well in the past period.
Other studies (Jegadeesh (1990)) have illustrated a similar short-term (week/month)
return reversal. However these findings are at odds with the medium term (less than 12
month) relative strength screens that are used by many investment professionals.
Jegadeesh and Titman (1993) posited that relative returns may be time variable, that is
that contrarian strategies may provide excess returns in the short and long term, and that
relative strength strategies may provide excess returns in the medium term. In their study
stocks were classified as winners (the 10% of stocks with the best returns over the
preceding six months), and losers (the 10% of stocks with the worse returns over the
preceding six months). The chart below shows the returns to the relative strength
strategy.
26
Literature Review
Figure 2.07: Cumulative Return on a Neutral, Six-Month Relative Strength,
Winner/Loser Portfolio. Compustat data 1965-1989
Source: Adapted from, Jegadeesh, N., and Titman, S., 1993, “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency,” Journal of Finance, 48, pp. 65-92. Table VII.
The strategy return is negative for periods of less than one month and of greater than
twelve months. It corresponds to the short-term contrarian view, the medium term
practitioner view and the long-term contrarian view. However the t-stats of the study
decline for the longer periods. The conclusion of the Jegadeesh and Titman study is
replicated in non-US markets for the period of 1957 to 1986 by Poterba and Summers
(1988). They found positive autocorrelation of stock prices for periods of less than twelve
months and negative autocorrelation for periods of greater than twelve months.
The most profitable Buy-Hold-Sell strategy [in the De Bondt and Thaler study] was
based upon test period returns of twelve months for a portfolio that was held for three
months, this strategy yielded an excess return of 1.31% per month.
27
Literature Review
Various explanations for price overreaction have been put forward. One is behavioural
insomuch as analysts and investors are systematically over (under) enthusiastic about
short-term winners (losers) (Dreman (1998). Another is based upon liquidity, there is
simply not enough stock in the system to accommodate the buying and selling pressure
post valuation relevant surprises (Grossman and Miller (1988) and Jagadeesh and Titman
(1995)).
2.4.7.2 January effect
In addition to the excess returns associated with earnings announcements, (detailed in
Section 2.5), there is also evidence of a ‘January effect’ (De Bondt and Thaler (1985),
Chopra et al. (1992)). In Figure 2.08 below, portfolios are ranked 1-20 in order of share
price return over the previous five years (portfolio 20 is high).
Figure 2.08: Average Return 1926-1986 of 5-Year Relative Strength Stocks by
Calendar Months
Source: Adapted from Chopra, N., Lakonishok, J., and Ritter, J.,R., 1992, “Measuring abnormal performance: Do Stocks Overreact?,” Journal of Financial Economics, 31, pp. 235-268, Table 3.
28
Literature Review
It can be seen that, in January, the return of underperforming stocks (over the previous
five years) is disproportionately high. An explanation for this could be that of the
‘window dressing’ of portfolios at the calendar year-end. Professional investors may do
this in order to present blue chip portfolios to clients for year-end fund reports. In
January, the subsequent ‘bottom fishing’ of these badly performing stocks results in their
outperformance.
2.5 The information content of reported earnings
In 1968, Ball and Brown were the first to record the observation that abnormal rates of
return followed earnings increases as reported at annual earnings announcements.
Subsequent US studies by Bernard and Thomas (1990) extended the research to quarterly
earnings announcements. They looked at over 100,000 announcements during the period
of 1974 to 1986 and concluded that,
1) A zero exposure, long-short portfolio, that each quarter takes a long position in the
top decile of earnings performers and a short position in the bottom decile, earns an
average excess return of +4.19% over the 60 days following the announcement. The
abnormal return is positive for 46 of the 50 quarters in the study.
2) The excess returns were a decreasing function of size
3) One sixth of the 60 day excess return is achieved in the first five days
4) Over 180 days the excess return is +7.74%.
5) There is no price drift post 180 days
29
Literature Review
They note that “a significant component of the predictable post-announcement abnormal
return occurs at the announcement of the following quarters earnings.” It seems that the
market ignores the information about future earnings contained in earnings
announcements. One reason put forward for the lagged share price response is that
analysts take up to three quarters to correct for an earnings surprise, (Abarbanell and
Bernard (1992)).
The Jegadeesh and Titman (1993) study, (see Section 2.4.7.1) also looked at the
performance of winner/loser portfolios in relation to earnings announcements (top and
bottom decile, ranked by six month share price performance relative to the index). The
performance of each group of stocks was measured over the two days before, and the day
of, the next quarterly results following portfolio formation. It was found that for seven
months winner stocks continued to outperform at earnings announcements. This implies
that the market failed to recognise that good earnings numbers foretold of more to come,
and likewise that earnings disappointment would continue. See below.
30
Literature Review
Figure 2.09: Monthly Difference between Winner and Loser Portfolios at Earnings
Announcement Dates
Source: Haugen, R.,A., 1990, “Modern Investment Theory,” Englewood Cliffs, N.J., Prentice Hall, p.21. Reference: Jegadeesh, N., and Titman, S., 1993, “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency,” Journal of Finance, 48, pp. 65-92. Table IX.
By the sixteenth month post the date of portfolio construction, the loser portfolio was
outperforming the winning portfolio. This implies that if the strategy of six-month
relative share price strength is to be used, then the investor should sell the portfolio after
eight months. Another strategy would be to take a position in the stocks that have
underperformed over fourteen months. I.e. the six-month test period, plus the eight
months during which the information contained in the earnings announcements are
absorbed by the market.
Hawkins, Chamberlin and Daniel (1984) put forward evidence that the revision of
earnings forecasts has informational value. Twelve-month risk adjusted returns of 14.2%
31
Literature Review
were available for investors in stocks with the largest upward revisions in consensus
earnings forecasts. Richards and Martin (1979) found that the revisions in the first quarter
contain more price sensitive information than revisions later in the year.
Mott and Coker (1993) investigated the effect of earnings surprise on the subsequent
performance of Smaller Companies. A positive surprise is defined by a Standardised
Unexpected Earnings (SUE) ratio of greater than one, where:
SUE = Actual EPS – Consensus EPS estimate Standard Deviation of Consensus
Their conclusion was that for Smaller Companies exhibiting a positive surprise, the
relative outperformance (from one-week post announcement) was +2.1% for the first
month and +12.9% for the first twelve months. Overall, ninety percent of these
companies outperformed over the following twelve months. The corresponding one
month and twelve month relative performance for negative surprise (SUE<1) was –0.9%
and –3.5% respectively. Ninety percent of negative surprise companies underperformed
over the following twelve months.
Analysts seem to fail to incorporate the information contained in announcements that
report an earnings surprise. In the Mott and Coker sample, in a reflection of the
inaccuracy of analysts earnings estimates (see Section 2.7), positive surprises were seen
20% of the time and negative surprises 26% of the time. In addition, for the positive
surprise sample, a repeat surprise was seen 34% of the time, and for the negative surprise
32
Literature Review
sample, a repeat surprise was seen 41% of the time. For positive and negative repeaters
the subsequent 12-month relative performance was 19.3% and –3.6% respectively. ‘Third
time’ positive or negative surprises were seen post the repeat surprise for 39% and 44%
of the time respectively. These observations of systematic error should prove useful in the
construction of screening models.
2.6 Evidence for the sustainability of accounting ratios
2.6.1 Return on equity
The Return on Equity has been shown to be a relatively consistent accounting ratio over
time (Wilcox (1984) Babcock (1980), Paynor (1966), see also Section 2.4.2).
2.6.2 Earnings per share growth
In a UK study of the period of 1951 to 1961, the statistician I.M.D.Little (1962, 1966)
found limited evidence that past earnings growth indicates future earnings growth. He
wrote that,
“My impression is that many stockholders, financial journalists, economists and
investors believe that past growth behaviour is some sort of guide to future growth…
For the privilege of holding these particular growth stocks, the investor has been
willing to forego a considerable amount of income…[in a belief that] firms which
have grown relatively better than others in the past will continue to do so in the
future”.
33
Literature Review
Little split a sample of stocks into two five-year sub periods. For the first period,
quartile groupings of low to high EPS growth were formed. The EPS growth for the
second period was then compared to that of the first. The average correlation of
earnings growth rates was just 0.02. There was therefore no consistency demonstrated
in persistence of respect of earnings growth. An updated study for the periods of 1981
to 1985 and of 1986 to 1990 supported the findings. Given the emphasis that
professional investors place on past EPS growth rates, this is an interesting finding. It
leads to the conclusion that the EPS growth rate experienced in the past, will not be a
useful investment screen.
However investors also screen by other quantitative and qualitative factors. Fuller et al.
(1993) tested whether they were able to use other data available to them in order to make
judgements about future growth prospects. During the period 1973-1990, Fuller ranked
stocks by the EP ratio (the inverse of the PER ratio) and then compared the subsequent
EPS growth over each of the next 8 years.
34
Literature Review
Figure 2.10: Relative Subsequent EPS Growth in Quintiles of the EP Ratio
Source: Fuller, R.,J., Huberts, L.,C., and Levinson, M.,J., 1993, “Returns to E/P Strategies; Higgledy Piggledy Growth; Analysts Forecast Errors; and Omitted Risk Factors,” Journal of Portfolio Management, Winter, Exhibit 6.
NB: The middle quintile grew EPS by 10.2% in the first year. Analyst forecast error was similar and low across quintiles.
Figure 2.10 shows that the earnings growth rate of low EP (high PER) stocks, is greater
than that of the earnings growth rate of the middle quintile stocks for some years into the
future. If it is assumed that a low EP ratio is a reflection of investor’s anticipation of high
35
Literature Review
future EPS growth, then the data suggest that investors can use fundamental analysis to
form reasonable judgements about the future EPS growth prospects for a company. With
reference to Figure 2.10, it seems that investors are able to judge the prospects of
premium earnings growth for a company, for around six years into the future.
Given the demonstrated underperformance of low EP stocks (see Section 2.4.1) and the
fact that low EP stocks achieve premium earnings growth, then it is interesting to
speculate as to why low EP stocks underperform. One explanation may be that the
valuation of these stocks is forced to too high a level for the achievable growth rate. This
is difficult to test for, as the growth expectation of investors cannot be measured directly,
it can only be inferred via the high EP ratio.
Another logical explanation as to why high PER stocks underperform, could be because
of the revision of over-optimistic forecasts by analysts. Figure 2.11, illustrates the finding
of Bauman and Miller (1997) that as PE ratios increase, then earnings surprise also
increases.
36
Literature Review
Figure 2.11: Earnings Surprise versus PE Ratio Compustat Data 1980-1994
Source: Bauman, W.,S., and Miller, R.,E., 1997, “Why Value Stocks Outperform Growth Stocks,” Journal of Portfolio Management, Spring, pp. 57-68, Exhibits 1 and 2.
NB: Earnings Surprise is defined as (reported earnings - forecast earnings)/standard deviation of forecasts.
The evidence suggests that high PE ratios tend to reflect past high EPS growth rates, and
as there is no correlation between past and future rates of earnings growth, then it is not
surprising that these high PER stocks systematically disappoint investors. The under
performance of high PER stocks could well be a combination of both over valuation, and
disappointment with ‘over estimated’ future earnings performance.
2.7 The use of earnings forecasts
By showing that companies that beat the consensus earnings forecast out performed the
market, Beaver, Clark, and Wright (1979) emphasised the advantage of being able to
predict large, unexpected, changes in earnings. This suggests that a forecast of earnings
per share in a valuation model would aid predictions of equity performance.
37
Literature Review
However there is evidence that consensus earnings forecasts from analysts have such a
high error rate that given the sensitivity of valuation models, they can add little useful
information, see below.
Figure 2.12: Studies of the Accuracy of Analysts Earnings Estimates
Study Period Number of Companies
Mean Error (%)
Stewart, Jr., 1973 1960-64 14 10 to 15Barefield and Cominsky, 1975
1967-72 100 16.1
Richards, 1976 1972 93 8.8Richards and Frazer, 1977
1973 213 22.7
Richards et al., 1977 1972-76 92 24.1Richards et al., 1977 1969-72 50 18.1
Mean estimate error over all studies 16.6
Source: Dreman, D.,N., 1998, “Contrarian Investment Strategies: The Next Generation,” Simon and Schuster, New York, p.99.
NB: Mean Error is defined as the absolute difference between the actual earnings and the forecast for the next quarter, as a percentage of the forecast.
In addition, the three studies illustrated below show that the earnings forecasts of analysts
for the next quarter are only slightly more accurate than those based upon mechanical
forecasts that simply rely on historic earnings.1
1 Three time-series models have been shown to be useful see: Bathke, A.,W., and Lorek, K.,S., 1984, “The Relationship between Time Series Models and the Security Markets Expectation of Quarterly Earnings,” The Accounting Review, 59, pp. 163-176. For a more recent paper examining time-series models, see Elgers and Murray (1992).
38
Literature Review
Figure 2.13: Studies on the Accuracy of Analysts Earnings Estimates versus
Mechanical Earnings Models
Mean Error (%)Study Analyst Group Analyst Forecasts Mechanical ModelsCollins and Hopwood, 1980
Value Line Forecasts 1970-75
31.7 34.1
Brown and Rozeff, 1979
Value Line Forecasts 1972-75
28.4 32.2
Fried and Givoly, 1982
Earnings Forecaster 16.4 19.8
Source: Damodaran, A., 1994, “Damodaran on Valuation,” John Wiley & Sons, New York, p81.
NB: Mean Error is defined as the absolute difference between the actual earnings and the forecast for the next quarter, as a percent of the forecast.
The evidence also suggests that the superior forecasting ability of analysts declines, as the
forecast period increases. O’Brien (1988) found that the forecasts of analysts were more
reliable than mechanical models for 1 and 2 quarters ahead, equal for 3 quarters ahead
and worse for 4 quarters ahead.
Even though forecasts may be inaccurate, the extent of disagreement between analysts, as
measured by the standard deviation of growth predictions, has been shown to be a useful
indicator as to the reliability of the consensus earnings estimates. The lower the standard
deviation, the more likely is the forecast to be accurate. See Givoly and Lakonishok
(1984).
39
Literature Review
Research from Strathclyde University concluded that 1) forecasts tend to err towards
optimism, 2) that the further ahead the time period, the greater the degree of error, and 3)
that forecasts made more than one period ahead were little more accurate than a naive
extrapolation of past earnings.
Over and above the normal difficulty of predicting future events, forecaster optimism
(and inaccuracy) possibly stems from the conflict of interests within Investment Banks.
Analysts are constrained by the fact that, as well as providing impartial advice and
forecasts for companies to institutional investors, the same companies are often corporate
clients. As corporate clients would rather a high share price to a low share price, the
Corporate Finance department will ensure that forecasts lead to the highest possible share
rating.
For Smaller Companies, the consequences of this conflict of interest can be magnified.
Due to the relatively low turnover of shares it would be reasonable to assume that
corporate, rather than institutional brokerage is the main driver of revenue. In addition,
securities analysts are paid in relation to the stock they sell rather than the accuracy of
their forecasts (Dorfman (1991)). The fact that only a few stock brokerage houses follow
Smaller Companies adds to forecast unreliability, as the corporate brokers forecast is
usually taken as a guide by other competing analysts. This reduces the benefit of a range
of independent analysis.
40
Literature Review
2.7.1 Earnings models
As discussed above, a simple extrapolation of the past earnings growth rate is not
sufficient. However certain studies have produced reasonably accurate forecasts of future
earnings from historic accounting data.
Ou and Penman (1989a) hypothesized that there is underutilised information about future
earnings contained in a variety of financial variables. They examined the Compustat
database for the period of 1965 to 1972 in order to form a strategy for investment during
the period of 1973 to 1983. A summary measure named ‘Pr’ was constructed from
published financial ratios. The Pr measure was used to indicate the direction of change of
one-year forward earnings. The forecast direction of change in earnings was then
compared to previous earnings (including a drift value that was based upon the last four
years earnings growth, cf. Section 2.6.2). Following the Ou and Penman strategy,
companies that were forecast to beat their ‘drift adjusted’ earnings were bought. Over a
two-year holding period, the average return of a portfolio that was made up of positive
earnings (long) and negative earnings (short) predictions was 7.0%. Out of the 34 (50%
of sample) attributes shown to be significant over one period, 26 of them were found to
indicate the same direction of forecast earnings for the second period.
Ou and Penman found that a high Pr was associated with prior earnings declines and
future price increases and vice versa. They also found that the results were improved
when the model was combined with a low PER strategy (Ou and Penman (1989b)).
41
Literature Review
Stober (1992) found that where brokers estimates are available (via I/B/E/S) the brokers
estimates are marginally more accurate than the prediction of Ou and Penman’s Pr
measure (72.8% vs. 70.9%, respectively). Interestingly the predictive performance of Pr
when the forecast of Pr and I/B/E/S estimates agree, rose to 78.2%, and when Pr and
I/B/E/S predictions disagree the analysts predictions are correct in 54% cases vs. 46% for
the Pr strategy. So the Pr model is best used, if at all, in combination with estimates from
analysts.
It should be noted that in promoting their own forecasting model, Holthausen and Larcker
(1992), demonstrate that the Pr strategy ‘breaks down’ in the 1983-1988 period.
2.7.2 Cash flow models
Sloan (1996) put forward the hypothesis that because investors “fixate on current
earnings”, they fail to use the information in accruals and cash flows that provides a good
forecast of future earnings. Sloan showed that companies with a high (low) level of
accruals experience negative (positive) future abnormal stock returns. He argued that the
value of cash flow from operations (CFO) when compared to net earnings is due to the
fact that aggressive accruals are not sustainable over long periods of time. This
information can be used to gain a position in a stock before the share price reacts when
the next earnings figures are announced. Sloans definition of accruals is,
42
Literature Review
Accruals = ( ΔCA – ΔCash ) – ( ΔCL – ΔSTD – ΔTP ) – Dep
| Current Assets | | Current Liabilities | | Depreciation |
Where ΔCA = change in current assets
ΔCash = change in cash/cash equivalents
ΔCL = change in current liabilities
ΔSTD = change in debt included in current liabilities
ΔTP = change in income taxes payable
Dep = depreciation and amortisation expense
In order to scale for size (MV) the accruals measure should be divided by Average Total
Assets.
The result of a ‘long’ bottom decile accrual, ‘short’ top decile accrual portfolio over a
one-year holding period is shown below.
43
Literature Review
Figure 2.14: Zero Net Investment, Long Low Accrual, and Short High Accrual.
Compustat data 1962-1992
Source: Sloan, R.,G., 1996, “Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future Earnings,” The Accounting Review, 71, July, pp. 289-315, Table 6.
The returns are relatively stable with an average size adjusted excess return of 10.4%
over the thirty-year study period.
In support of the above strategy, Bernstein (1993) states that,
“Analysts [should] relate CFO to reported net income, as a check on the quality of that
income. A company with a high net income and a low cash flow may be using income
recognition or expense accrual criteria that are suspect”
44
Literature Review
2.7.3 Other predictive models
2.7.3.1 The T-Model
Tony Estep (1987) developed the T-Model, this expresses expected stock return in three
terms—Growth, Cash flow yield, and Valuation change. It was demonstrated that 94% of
the return of a stock during 1982-1985 could be forecast if ROE, growth rate of
shareholders equity, and change in PTBV were known with hindsight. The Estep
proposition is that if the dependent variables are accurately forecast, then the model has
predictive ability.
The T-Score = g + ROE – g + PTBV * (1 + g) PTBV PTBV
| Growth | | Cash flow yield | | Valuation change |
Where,
T-Score = Total return
g = Growth in shareholders equity over the period
ROE = Net Income during the period / Opening Shareholders Equity
PTBV = Opening Market Value of firms common stock / Opening Shareholders
Equity
PTBV = Change in PTBV during the period
During the period of the study, portfolios with a high expected return according to the T-
Model, outperformed the S&P 500 by 9.8%, the portfolios had a standard deviation of
13.8% vs. 13.2% for the S&P 500.
45
Literature Review
2.7.3.2 Jacobs and Levy Model
Over the test period of 1982 to 1987, for a sample of 1000 stocks, Jacobs and Levy
(1988) looked at the prediction of the dividend discount model (DDM) return vs. a model
comprising of 25 equity attributes. They found that the DDM (see Section 3.1) was
insignificant when placed in a multivariate model with the 25 attributes. The attributes
that provided statistically significant abnormal returns were: PSR, short term tax loss,
neglect, relative strength, residual return reversal (-1 month, -2 months), trends in
analysts estimates (-1 month, -2 months), sigma, and earnings surprise (-2 months, -3
months) the R2 of the model was 43.93%.
They also tested PER, book/price, cash flow/price, yield, Beta, coskewness, size, earnings
growth consistency, low price, tax rate, trends in analysts estimates (-3 months), and
earnings surprise (-1 month). However these attributes were not found to be statistically
significant.
2.7.3.3 Brennan and Subrahmanyam Model
In addition to other attributes, Brennan and Subrahmanyam (1996) assessed liquidity as a
factor that effects returns. They used the dollar volume of trading as a reasonable proxy
for liquidity, and in one of the few studies to test the APT, they found the returns to
Volume, MV, PTBV, and 3, 6, and 12 month lagged share price performance to be
significant.
46
Literature Review
2.8 Are the market inefficiencies a proxy for risk?
A major debate in Finance has been as to whether the rational asset pricing paradigm
(CAPM or APT) holds. Are the accounting characteristics (discussed above), that
empirically provide excess returns, a proxy for risk, or do they explain predictable excess
stock returns at constant risk?
Damodaran (1994, see Section 3.1.2), rearranged the DDM relationship, to show that,
P0/E0 = (Payout ratio * (1 + gn))/ (Er-gn)
Where gn is the expected growth rate of earnings, and Er is the required rate of return.
Keeping other values constant, the relationship between PER and β can then be
demonstrated (from the CAPM relationship, via a change in Er).
Figure 2.15: The relationship between Beta and the PE Ratio
Source: Damodaran, A., 1994, “Damodaran on Valuation,” John Wiley & Sons, New York, p. 204.
47
Literature Review
It can be seen that, the PER will fall as β increases. This is because under the CAPM (see
Section 2.2) Er, the required rate of return, is a positive function of β. If the CAPM holds
then a low PER is a proxy for risk as defined by Beta. However this is not demonstrated
in empirical tests, in fact the relationship is reversed. See Figure 2.01, and Section 2.4.1.
Fama and French (1993, 1994, 1996) have theorised that the various market inefficiencies
are related to each other, and can be amalgamated into a three-factor model. The FF
model is defined as:
E(Ri) - Rf = bi [E(RM) – Rf] + si E(SMB) + hi E(HML)
Where E(Ri) - Rf, is the expected return of a portfolio in excess of the risk free rate. This
is dependent upon three factors. (1) the excess return of a broad market portfolio, (2) the
difference in return of a portfolio of small stocks and the return of a portfolio of large
stocks (SMB), and (3) the difference in return of a portfolio of high book-to-market
stocks and the return of a portfolio of low book-to-market stocks (HML). The historic
return of a low PTBV (high book-to-market) portfolio is around 20.5 percent per year
(see Figure 2.03), so for the stocks with a low EP, low cash flow per share and high sales
growth (which tend to have a high PTBV ratio), the expected return is lower. Fama and
French propose this model as an equilibrium pricing model and find that in a regression
of the three factors on price an R2 of 92% is demonstrated.
48
Literature Review
In collaboration with Baker, Haugen (1995) studied the relationship between volatility
and return. In this US study, Larger Companies were classified as those stocks in the
S&P 500 Index, and Smaller Companies were those in the Russell 2000 Index
(MV<$200m.). For the period of 1928 to 1992, Haugen formed portfolios that were
based upon stocks with (24 month historic) low and high volatility. Interestingly, the
results show that the low volatility stocks provided higher future annual returns.
Figure 2.16: The Annual Returns of High vs. Low Volatility Portfolios
Source: Haugen, R.,A., 1995, “The New Finance: The Case against Efficient Markets,” Prentice Hall, Englewood Cliffs, New Jersey, p. 86.
NB: Large Cap. (S&P 500) sample 1928-1992.Small Cap. (Russell 2000) sample 1979-1992.
The effect is even more pronounced for Smaller Companies, as measured by the Russell
2000 Index. These results are in direct contrast to the CAPM theory, which states that
higher rewards can only be achieved by taking higher risk (volatility).
49
Literature Review
Fuller et al. (1993) produced one of the few studies that make use of the BARRA
database. This sophisticated risk management system is widely used amongst investment
management companies. The BARRA system attempts to identify the sources of alpha
(excess return) in terms of fifty-five industry classifications and thirteen other potential
risk factors. However, Fuller et al. found that their sorted EP portfolios provided a better
explanation of stock performance than that of the complex BARRA model.
It would seem that investment risk is difficult to express in a simplified manner. This is
not surprising given the many factors that influence share prices. It may be that risk
occurs in different ways for different strategies. For example, low PER stocks may
outperform because investors are compensated for the risk of company failure. However
investors in high PER stocks would also seem to be taking risk, insomuch as they are
relying on a premium level of earnings growth, that is difficult (impossible) to forecast.
Both sectors then, suffer a different type of risk.
The Fama and French measure (above) is an extension of the CAPM model, and in a
sense, is a simplification of the APT model. Their measure has a high predictive ability
but, if remains to be seen whether irregularities are found in this model, either over time,
or over explanatory factors.
Until more complex models are developed that can consistently classify the risk of
holding a particular stock, it is perhaps best to express risk in simple terms, as the
variability of total return. All investors will understand that, in practice, the risk of
50
Literature Review
holding equities is a combination of the average return that they provide and the
variability of that return.
Consequently, this study will assess risk as the standard deviation of total return, and will
avoid the misleading simplification of using Beta values.
2.9 Conclusion of the literature review
Whilst many of the studies reported in this section are US based, it seems that value
strategies are also profitable in other international markets (Capaul et al. (1993)).
The strategy of value investing (that is, of investing in low rated stocks) was first
documented the 1930s (Graham and Dodd (1934)). One of first academic studies to
demonstrate the benefits of the low PER strategy was undertaken by Basu (1977,1983).
In 1994, Lakonishok et al. extended the work of Basu, and calculated that over a period
of 22 years a low PER investment screen provided, an average excess (to market) return
of 2.7% p.a.
Fama and French (1992) investigated the result of following a low PTBV strategy. They
found that over the period of 1963 to 1990, an excess (to market) return of 4.6% p.a.
would have been achieved.
Fewer studies have focused on the potential benefits of following a low PSR strategy, but
the results, nevertheless, have been impressive with excess returns of 2.04% p.a.
available over an 8 year study period (Jacobs and Levy (1988)).
51
Literature Review
In a test of a high yield screen Dreman (1988) found that over a 27 year period, these
stocks provided an average excess return of 1.2 % p.a.
The most striking aspect of the value strategies, is the magnitude of the outperformance
recorded. With the benefit of compounding over a 20 year period, an excess return of
2.7% or 4.6% p.a. for the PER or PTBV strategies respectively, results in phenomenal
total returns.
Strategies that exploited the patterns detected in share price behaviour were found to be
even more profitable. Jegadeesh and Titman (1993) were able to demonstrate a one-
month and long-term (three year or five year) share price return reversal, along with the
continuation of share price momentum over shorter time periods. A long-short portfolio
of the companies with the highest, and lowest, six-month relative share price momentum,
provided an excess return of 9.5% over the following twelve months.
The information value of earnings (growth) announcements was investigated by Bernard
and Thomas (1990). They found that a portfolio that takes a long position in the top
decile of earnings performers, and a short position in the bottom decile of earnings
performers, provided an excess return of 7.74% over a 180 day holding period.
52
Literature Review
Jegadeesh and Titman (1993) noted that those shares that outperformed at earnings
announcements also tended to be those with positive six-month share price momentum.
It has been shown that analysts are poor forecasters of company earnings (Damodaran
(1994), Dreman (1998)). In one of the few studies of Smaller Companies, Mott and
Coker (1993) reported that companies that reported an earnings surprise tended to
outperform by 12.9% over the following twelve months. Interestingly, a high proportion
of companies (34%) continued to surprise the market at future announcements. For these
‘positive repeaters’ the excess return over twelve months accelerated to 19.3%.
Bauman and Miller (1997) looked at earnings surprise by equity characteristics. They
found that earnings surprise tended to be negative for high PER companies and positive
for low PER companies. This provides a partial explanation as to why low PER stocks
outperform the market.
Hawkins, Chamberlin and Daniel (1984) put forward evidence that the revision of
earnings forecasts has informational value. Twelve-month risk adjusted returns of 14.2%
were available for investors in stocks with the largest upward revisions in consensus
earnings forecasts.
The above trading strategies are relatively simple. Greater rewards were thought to be
available for investors who could use the information within company accounts that was
deemed relevant to the future share price.
53
Literature Review
Ross-Healy and Sgromo (1993) devised a complex strategy that identified companies
with improving Balance Sheet strength. During the period of 1983 to 1990 this strategy
outperformed the market by 7.7% p.a.
The strategy of Ou and Penman (1989a) was to use published accounting data in order to
create a long-short hedge portfolio consisting of companies that were forecast to beat (or
not beat) their ‘drift adjusted’ earnings. Over a ten-year test period, the portfolio provided
a size adjusted excess return of 3.5% p.a.
A more profitable (and less complex) strategy was developed by Stober (1992). He
constructed a long-short hedge portfolio based upon accounting accruals. Over a period
of thirty years, this portfolio provided an average excess return of 10.4%.
Finally, the T-Model developed by Estep (1987) was shown to produce an excess return
of 9.8% p.a. However the time period of the test was relatively short (3 years), and other
researchers have had difficulty in replicating the results. Nevertheless, the use of the
PTBV ratio in a theoretical way is of interest.
A summary of the profitable investment screens found during the course of this Literature
Review is documented below.
54
Literature Review
Figure 2.17: Annualised Excess Return for Investment Strategies
NB: The strategies of Accounting accrual, Ou and Penman ‘Pr’, Share price relative strength, and Earnings growth are based upon long and short (hedge) portfolios, consequently they also benefit from the returns of the short position. The other strategies are long only.
One criticism of the studies is that they all rely on the integrity of Compustat. There is
bound to be little disagreement over the market inefficiencies as there is no comparative
data set. Consequently there is a debate about the accuracy of the data. One argument is
that there is a survivorship (selection) bias (Kothari et al. (1995), Breen and Korajczyk
(1994)). This is that, we may expect to see a correlation, for example between low PER
firms and outperformance, if those low PER firms were in financial distress at the time
that they were included in test portfolios and subsequently survived. If those companies
that had correspondingly low PERs, but went bankrupt, were removed from the database,
then there would be a bias in the results towards low PER companies. However a great
55
Literature Review
deal of effort has been spent on trying to clean up the data base (De Silva (undated)) and
as yet no major reversals in the findings have been reported.
The recorded excess return of certain strategies may also be as a consequence of the
miss-measurement of expected returns because, 1) the processing of financial information
is not costless and 2) that the prices recorded on Compustat are sometimes not prices that
could be traded at. In addition, transaction costs will erode a proportion of the returns.
These are not insignificant, for instance, the average bid-ask spread on the NYSE has
been calculated as 2.8% (Keim (1989)).
With reference to Figure 2.17 the returns that the strategies provide are far superior to
those achieved by traditional active Investment Management. It is fair to ask why, in a
world where outperformance is substantially rewarded in terms of large bonuses, have the
inefficiencies not been exploited and consequently arbitraged away?
One answer may be that whilst the idea of value strategies have been documented since
Graham and Dodd (1934), the computing power to ‘prove the case’ has been available for
considerably less time.
An additional explanation as to why Institutional Investors fail to exploit the
demonstrated inefficiencies, would be that their time horizons are so short that they
cannot take the risk of pursuing a value strategy that might take a number of years to
outperform. This is irrational given that Trustees should be seeking outperforming
56
Literature Review
strategies, but is not incomprehensible in a world where investment mandates are
reviewed yearly, if not more frequently.
It is also possible that the answer as to why market inefficiencies persist, lies in human
psychology. The behavioural scientist Amos Tversky (1995) commented,
“Rather than operating on rational expectations, people are commonly biased in several
directions: they are optimistic; they overestimate the chances that they will succeed, and
they overestimate their degree of knowledge, in the sense that their confidence far
exceeds their ‘hit rate’”,
and also that,
“Recognising our limited ability to predict the future, is an important lesson to learn”.
I.M.D.Little (1962, 1966) theorised that investors put undue weight on past events, and
that these are poor predictors of the future (see Section 2.6.2). The desire to be part of
‘The Crowd’ pushes investors towards well-known and proven companies. Trustees are
unlikely to question blue chip investments, whereas value stocks are often considered to
be the shares of companies that face severe problems. Detailed discussions of the field of
Behavioural Finance are to be found in Jacobs and Levy (1988), Hunter (1988), Thaler
(1993), and Wood (1995).
57
Literature Review
One of the most striking findings in the process of this Literature Review, was that of the
work of Fuller et al. (1993). They found that investors are able to identify, and award a
high PER to those stocks with superior earnings potential. However it is shown in other
work that the same high PER stocks, and investment funds as a group, tend to
underperform (Malkiel (1995), Jensen (1968), and Section 2.4.1). Consequently, it seems
that fund underperformance is due to a combination of overaggressive valuation of those
equities that are (rightly) considered to be superior earnings performers, and also of the
earnings disappointment that is a consequence of analyst forecast error (Damodaran
(1994), Dreman (1998), Figure 2.11).
Finally, there is the thorny issue of risk. Principally the studies calculate the risk-adjusted
return of a strategy, from the expected return as defined by the CAPM. However the
CAPM depends upon stock Beta values that have been shown to be poor predictors of
future return (Gibbons (1982), Stambaugh (1982), Fama and French (1993, 1994, 1996)).
Due to this, the results must be viewed with caution. Market capitalisation has been
demonstrated to be a better predictor of stock performance than Beta, so this has also
been used as a proxy for risk. But it is not clearly understood for what reasons MV is a
proxy for risk, and indeed how consistent this factor may be. (Note that in the UK, the
beginning of the underperformance of the FTSE Smaller Company Index relative to the
FTSE All-Share Index coincided almost exactly with the publication of articles about the
UK ‘Small Cap [outperformance] effect’ in the late 1980’s). Size as a measure of risk,
may prove to be as inconsistent as Beta.
58
Literature Review
However the Efficient Market Theory cannot be refuted until the complex subject of risk
has been fully understood. Even though so many ‘inefficiencies’ have been demonstrated
that it is temping to reject the theory merely on the weight of evidence alone, it is not yet
possible to precisely specify the risk that any of these strategies suffer, so this temptation
should be avoided.
It is interesting to consider that as soon as risk is properly understood, those strategies
that really do provide risk-adjusted excess returns will be exploited. When this occurs,
the anomaly will cease to exist, and the Efficient Market Theory will dominate once
more.
59
Introduction to Equity Valuation
3.0 INTRODUCTION TO EQUITY VALUATION
60
Introduction to Equity Valuation
3.0 INTRODUCTION TO EQUITY VALUATION
In the Literature Review, the various market inefficiencies that have been presented to
date were discussed. However the demonstration of the existence of a pattern between a
share price and an equity attribute, is not sufficient to prove the relationship. It is possible
that the occurrence of the pattern may be due to chance, or due to circumstances that may
not be repeated in the future. It is therefore essential for a proposed market inefficiency to
have some theoretical basis. The subjects of equity valuation, and company financial
performance are explored in this, and the following chapter.
3.1 The dividend discount model
The Dividend Discount Model (DDM) states that the value of a company is the sum of
the discounted dividend flows that the company will produce over its lifetime.
Company Value = D0 + D1___ + .… + Dn___
(1+r)1 (1+r)n
Where D is the forecast dividend in each time period, r is the discount rate and n is the
life time of the company or investment.
When the term structure of both interest rate and equity risk premium is flat, and earnings
growth is constant, the value of an equity can be calculated via a rearrangement of the
61
Introduction to Equity Valuation
perpetuity formula. This is called the Gordon Growth Model after Gordon (1962). It is
expressed below.
P = D0 * (1+g)/r-g
Where P = Share price, D0 = Current dividend per share, g = Growth rate of earnings per
share and r = the discount rate.
3.1.1 The relationship between the DDM and accounting ratios
Miller and Modigliani (1961) show that the DDM can be rewritten in terms of various
accounting ratios. This is due to the fact that 1) a constant dividend growth rate model is
mathematically similar to that of an annuity, and 2) that accounting ratios are interrelated.
As various accounting ratios are related to the dividend, a DDM valuation can be
expressed in terms of earnings (via a constant payout ratio), cash flow (via constant
depreciation rate, working capital adjustment and payout ratio), sales (via a constant net
margin and payout ratio) and book value (via a constant ROE and payout ratio).
Consequently, a high DDM valuation would be characterised by a high dividend (yield),
high earnings (i.e. low PER), high cash flow (low PCF), high sales (low PSR) or high
book value (low PTBV).
62
Introduction to Equity Valuation
In a study, Sorensen and Williamson (1980) found that a DDM ranked portfolio had a
16% excess return for undervalued stocks and 15% negative excess return for overvalued
stocks. This was supported by similar findings from Haugen (1990).
It would seem then, that some of the value-based inefficiencies that are discussed in the
Literature Review could be thought of as a failure of investors to price stocks correctly on
the basis of the DDM. However, this could also be seen in terms of risk. It is not
uncommon for the DDM to place over 80% of the value of a company on dividends
forecast more than five years ahead 1. Therefore, the reward for holding low PER stocks
(which are often indicated as undervalued by a DDM) could be for the risk that investors
take when trying to forecast so far into the future.
Given the complexity of data input into a DDM, and the fact that low PER and high yield
strategies outperform DDM strategies, then on its own, the DDM seems of little use as a
portfolio filter.
However it does provide a theoretical basis for the observed outperformance of high
yield, low PER, low PCF, low PSR and low PTBV stocks.
1 This is easy to calculate. From the MV of a quoted company, deduct the DDM valuation of the dividends expected over the next five years, the remainder is the DDM value of the company post year five.
63
Introduction to Equity Valuation
3.1.2 Price to earnings
The following derivations of the DDM in terms of accounting ratios are based upon the
work of Damodaran (1994).
From the constant growth DDM,
P0 =__DPS1__ (1)( r-gn )
Where, P0 is the Price at time 0, DPS1 is the dividend for next period, r is the cost of
equity (from the CAPM), and gn is the expected growth rate in equity and earnings.
If DPS1 is rewritten as,
DPS1 = EPS0 * Payout ratio1 * (1 + gn)
where EPS0 is the earnings per share at time 0, and Payout ratio1 is the payout ratio for
period 1, then equation (1) can be expressed as,
P0__ = Payout ratio1 * (1 + gn) (2)EPS0 (r-gn)
64
Introduction to Equity Valuation
And in terms of the EPS for the next time period as,
P0__ = Payout ratio1_EPS1 (r-gn)
This model can be extended into multiple time periods via the equation:
P0___ = Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) (3)EPS0 (r-g)
+ Payout ration * (1+g) n * (1+g n)__(r-gn) * (1+r)n
Where,
n = Number of years of high growth (and/or visibility of a particular growth rate)
g = Expected growth rate in equity and earnings in the first n years
gn = Expected growth rate in equity and earnings in the subsequent years
Payout ratio = Payout ratio for first n years
Payout ration = Payout ratio after n years
For companies with low/no dividend payments, the ratio of FCF to earnings can be
substituted for the payout ratio.
65
Introduction to Equity Valuation
3.1.3 Price to book value
From the two period growth DDM in terms of the PER, (equation (3))
P0___ = Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) (4)EPS0 (r-g)
+ Payout ration * (1+g) n * (1+g n)__(r-gn) * (1+r)n
If EPS0 is expressed as, EPS0 = BVPS0 * ROE, (this is from a simple accounting identity
where BVPS0 is the Book value per share in time period 0), and substituted into (4) then,
P0___ = ROE * Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) BVPS0 (r-g)
+ Payout ration * (1+g) n * (1+g n) (r-gn) * (1+r)n
The PTBV ratio is seen to be dependent upon the ROE. In other words, companies with a
high ROE should sell for a higher PTBV ratio than companies with a low ROE.
Therefore, if the DDM holds, then the stocks that investors should look for (i.e. those that
are undervalued), are stocks with a low PTBV ratio and a high ROE.
66
Introduction to Equity Valuation
3.1.4 Price to sales
The main benefit of the Price to Sales Ratio (PSR) is that it can be used to value
companies that are making a loss. In addition, the sales figure is not subject to the
vagaries of accounting policy to the same extent as are the PE and PTBV ratios. The PSR
is also more stable over time.
From the one period stable growth DDM in terms of the PER (equation (2)),
P0__ = Payout ratio1 * (1 + gn)EPS0 (r-gn)
Substituting in, EPS0 = Net Profit Margin0 * SPS0, (where SPS0 is the Sales per share in
period 0),
P0 = Net Profit Margin0 * SPS0 *_Payout ratio1 * (1 + gn) ( r-gn )
P0__ = Net Profit Margin0 * Payout ratio1 * (1 + gn) SPS0 ( r-gn )
If the profit margin is based upon the earnings in the next time period then,
P0__ = Net Profit Margin1 * Payout ratio1 SPS0 ( r-gn )
67
Introduction to Equity Valuation
For the two stage model,
P0_ = Net Profit Margin1 * Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) SPS0 (r-g)
+ Payout ration * (1+g) n * (1+g n) (r-gn) * (1+r)n
From the above relationships, it is evident that firms with higher net profit margins will
sell for higher PS ratios than firms with lower net profit margins. This is for two reasons,
one is direct, and the other is via the lower expected growth rate.
With reference to Section 4.1, the effect on the growth rate (g) is due to following
relationship,
Fundamental EPS Growth (g)= Retention ratio * ROE
= Retention ratio * (Net profit/Sales) * (Sales/BV of
Equity)
= Retention ratio * Net Profit Margin * (Sales/BV of
Equity)
As the Net profit margin is reduced the expected growth rate will reduce, unless Sales
increase proportionately.
68
Introduction to Equity Valuation
For an investment screen, companies with low PS ratios and high net profit margins
should be considered undervalued and companies with high PS ratios and low net profit
margins should be considered overvalued.
3.1.5 Debt to equity ratio
A high debt to equity ratio (DER) enhances a company’s valuation by increasing the
ROE. This can be demonstrated via the following relationship.
ROE = ROA + (DER * (ROA – i (1 – t))) (5)
Where,
ROA = Return on Assets (Pre-interest)
DER = Book Value of Debt / Book Value of Equity
i = Interest Expense
t = Tax rate on ordinary income
Equation (5) can then be substituted into,
Earnings Growth = Retention ratio * ROE
Earnings Growth = Retention ratio * [ROA + (DER * (ROA – i (1 – t)))]
69
Introduction to Equity Valuation
So, the higher the DER, the higher is the fundamental earnings growth of a company.
This provides a theoretical basis for the finding of Bhandari (1988), Section 2.4.6.2, that
companies with a high DER provided positive excess returns.
3.2 Cash flow valuation
In his book, Alfred Rappaport (1986) argues that managers and investors judge company
performance by looking at the wrong financial measures. He highlights some of the
shortfalls of earnings which are discussed below.
1) Rappaport argues that the PER as a theoretically sound valuation ratio is flawed. He
asserts that the leap from dividend, which is a cash measure, to earnings, which are an
accounting measure, is like comparing “apples and pears”. For example, earnings can
be effected by a change in accounting policy, i.e. a change in stock valuation from
LIFO to FIFO, however cash flow will not alter as a consequence.
The difference between cash flow and earnings stems from the fact that accountants
measure revenue from the point of sale and purchases from the point of invoice.
Whereas in reality, there is a delay until actual cash leaves or is received by a
business. This is why a company can be declared bankrupt, even though the P&L
account (and hence EPS) looks healthy. The difference in approach between cash
flow and EPS valuation is apparent for growth companies. Growth Companies would
normally have to increase inventory before cash from sales is received. In this case,
the cost of sales figure will understate the cash outflow for inventory. Obviously there
70
Introduction to Equity Valuation
is a counter balance in the increase in payables, however the miss-match of cash to
reported earnings will lead to EPS being greater than cash EPS for a growth
company.
2) Depreciation smoothes the deterioration in the book value of an asset over its
estimated life. However the extent of the charge is subjective. For example, the
estimated life of computer equipment has proved difficult to judge, with some now
arguing that IT spend should be fully written off at the date of purchase. Over time,
errors in predicting asset life can accumulate and prove misleading when certain
accounting measures (particularly Book Value) are assessed.
3) In addition, two companies may produce the same earnings growth however the
leverage of one company could be higher, and consequently the variance of expected
earnings would also be higher. A single measure of EPS does not account for the
increased risk.
Consequently, Rappaport argues that investments should be valued (and compared) on
the basis of discounted cash flow. This supports the findings in Section 2.7.2 where the
positive excess return of low accrual portfolios (Sloan (1996)) was noted.
3.3 Conclusion of the introduction to equity valuation
Valuation theory can justify the market inefficiencies that have been demonstrated for the
ratios of PER, PCEPS, PSR, PTBV, DER, and DYLD and of accounting accruals.
71
Introduction to Equity Valuation
In Chapter 5.0 these measures will be tested for potential market inefficiencies within the
sub sector of UK Smaller Companies.
72
Determinants of Company Financial Performance
4.0 DETERMINANTS OF COMPANY FINANCIAL
PERFORMANCE
73
Determinants of Company Financial Performance
4.0 DETERMINANTS OF COMPANY FINANCIAL PERFORMANCE
4.1 Equity growth rate
It is not sufficient just to seek out companies with a high Return on Equity (ROE). If the
return is high, but the ability to absorb cash (i.e. reinvest) is low, then the growth in value
will be lower than for a similar company that can reinvest more of its cash flow (at the
same ROE). The retention ratio (RR) can be used as an indication as to the investment
opportunities available for a company, the higher the RR, the more opportunities. The
achievable growth rate is defined as,
Achievable Growth Rate = ROE * RR
Where, RR = (EPS-DIV)/EPS
An additional consideration, is that a company may face an incremental Return on
Investment that is higher/lower than the historic rate. The incremental ROI (IROI) can be
calculated as below,
Incremental ROI = OP NI
Where OP = Change in Operating profit less interest, less tax over a period
NI = Net investment in working capital and fixed assets over a period
74
Determinants of Company Financial Performance
One of the benefits of calculating the IROI is that it is not distorted by out of date
Balance Sheet asset valuations, as is the ROE.
4.2 Issues of cash flow
4.2.1 Depreciation charge
Depreciation is a provision out of profit towards the cost of replacing an asset at the end
of its life. However the rate of depreciation is almost always determined by the
management, so it would be reasonable to be suspicious of companies where the
depreciation rate is substantially lower than that of the peer group.
4.2.2 Working capital
Over time, the increase in working capital (stock + debtors - creditors) should be
proportionate to the increase in sales. However for growth companies, expansion causes
stock and debtors to increase disproportionately. As accounting measures accruals, this
causes a mismatch, for example, earnings are recorded on sales that are not yet paid
(debtors), and assets are recorded when no contracted buyer has yet been found for them
(stock). In addition trade creditors are not reported in the P&L account, although these
are debts that must be paid. The P&L account can therefore mask the true cash position
of a company.
One test that is used in order to assess earnings manipulation is to compare EPS growth
over a period, with cash flow per share (CEPS) growth over the same period. A shortfall
75
Determinants of Company Financial Performance
in cash growth would be seen as a negative indicator. Also, an expanding working capital
to sales ratio would warn of a potentially unsustainable growth rate.
4.2.3 Free cash flow
The cash flow statement is split into three sections. These are net cash from Trading,
Financing, and Investing. Any surplus or loss on the sale of businesses or fixed assets is
deducted from the operating profit and included in the Investing account. Any cash from
operating activities that is left after financing activities is then available for capital
expenditure and dividends to equity shareholders. Capital expenditure is not segmented
further, but it has two distinct uses. First to grow the business organically, and second to
replenish worn out assets. The replenishment of assets is necessary in order to maintain
present cash flow, and is therefore essential spending. Free cash flow is that which is left
after this expenditure. Despite the reservations stated above, the depreciation charge is
often taken as a proxy for replenishment spend, as for an outside investor it is difficult to
assess the true need for replacement capital expenditure.
Free cash flow as a percentage of Pre-tax profit is a reasonable measure of the spare cash
resources that a company has after funding capital expenditure, dividends, tax, interest,
and working capital. A ‘high’ percentage indicates that a company has the flexibility to
invest in acquisitions or organic growth, or has the ability (if the free cash flow is
consistent) to service higher debt levels in order to grow.
76
Determinants of Company Financial Performance
Free cash flow can also be used as an indicator for potential corporate activity. For
example, if free cash flow is able to finance debt (at a rate of 2% above LIBOR, and with
repayment over 15 years) of greater than the market capitalisation of the company it is a
prime target for an MBO or takeover.
During the period 1992-1996, Philips (1998) found that seventy eight percent of
companies in the UK Engineering sector with free cash flow as a percentage pre-tax
profit of greater than 120% outperformed the market. In his analysis Philips created an
index of ‘average operating cash as a percentage of pre tax profit’ * ‘average return on
capital employed’, the higher the value, the better the prospects for company.
The results of tests of CEPS as a percentage of EPS for UK Smaller Companies are
reported in Chapter 5.0.
4.3 The value drivers
The Du Pont Formula splits the Return on Equity into components in order to quantify
the value drivers of a business.
Return on Equity = Attributable profit / Book Equity
= Attributable profit * Sales_ * Assets_____Sales Assets Book Equity
= Earnings Margin * Asset Turn * Equity Solidity
77
Determinants of Company Financial Performance
An increase in the Earnings Margin, Asset Turn, or Equity Solidity Ratio will increase
the ROE and hence the achievable growth rate of a company (see Section 4.1).
Although a higher earnings margin can compensate for a lower asset turn, the best
companies outperform on all measures. This is because high margins and high asset turns
generally provide high cash flow to 1) Invest in more high return assets, 2) Retire equity,
or 3) Pay high dividends.
The Drivers of Earnings Margin are:
Depreciation Turnover
This ratio may decline if previously owned assets are converted to leases. Whilst
operating leases are charged against operating profits, and finance leases are charged to
the interest account, in both cases Asset Turn would increase. If the ratio of depreciation
to turnover has reduced without a corresponding increase in Asset Turn then depreciation
rates may have been relaxed.
Staff costsSales
The above ratio is a useful measure of the service element of a business. High staff cost
to sales = high service = high margin = high PSR (see Section 3.1.4).
78
Determinants of Company Financial Performance
Sales______________Number of Employees
This is measure of how labour intensive the business is.
Operating profitEmployees
This is a measure of how much a business can charge for the time of its employees.
A business that is both labour intensive and has a high profit per employee is therefore
attractive.
Per capita profit Per capita sales
A ratio of greater than one can indicate a positive change in product or service mix. [Less
than one can indicate pursuit of turnover for its own sake].
The Drivers of Capital Efficiency are:
Sales_______________________________________ = Asset turnAverage of opening and closing Tangible fixed assets
Asset turn will rise if the capital spend of a business is producing an increase in sales
over and above the increase in assets. A static asset turn with higher sales is not
satisfactory if the increase in assets is essentially a requirement in order to carry on
trading.
79
Determinants of Company Financial Performance
Working CapitalSales
If there is a divergence between operating profit and cash inflow from operations, then
the cause is normally that sales growth requires further investment in working capital
(e.g. stocks and debtors).
Sales___________________________________ = Intangible asset turnAverage of opening and closing intangible assets
Acquisitive companies can seem to be more profitable than they are, for example, by
writing down assets to ‘fair value’, writing off the goodwill to reserves, and then selling
the stock at higher than carrying value, therefore inflating profit. Similarly over-provision
can be made for (not so) doubtful debtors. However working capital as a percentage of
sales will then be seen to increase, as both are replenished at normal values, and under
FRS 10 companies that buy-in turnover will show a deteriorating intangible asset turn.
Cost of goods sold________________ = Stock turnAverage of opening and closing stock
This indicates the speed at which a company can turn its inventory into saleable goods.
Average of opening and closing trade debtors = Debtor daysSales/365
The timely collection of debt indicates both good working capital management and the
strength of a company’s bargaining position.
80
Determinants of Company Financial Performance
Average of opening and closing trade creditors = Creditor daysCost of goods sold / 365
As long as supplier relationships are not damaged, the more interest free credit that a
company can gain, the better.
4.4 Operational gearing
In general the resilience of the profits of a company to volume and price pressures, is
determined by the margin structure of the profits. The resilience to volume declines
depends upon operational gearing, and resilience to price falls depends upon the
operating margin. However it is the relationship between fixed and variable costs, that
determines the degree of the resilience.
Resilience to volume declines:
The profit sensitivity to a decline in volume is determined by the relationship between
direct variable costs, and fixed overhead costs. However it is not just that fixed costs are
negative and that variable costs are positive. The most important ratio is the amount by
which gross profit covers net profit, that is the extent by which profits, after variable
costs, exceed the fixed overhead.
Traditionally if a company had a gross margin of 40% and a net margin of 20% then if
sales declined by 10% profits would be expected to fall by 20%. This is illustrated by the
relationship below,
81
Determinants of Company Financial Performance
Profit gearing = Gross margin * Change in sales Net margin
However in the short term “Gross margin” costs such as Labour are likely to be relatively
fixed. If the charge for Labour moves to below the calculation of gross margin then the
sensitivity of profits to sales declines increases greatly. In general, Labour costs could be
defined as 80-90% fixed, other costs 90% fixed, and depreciation and R&D, 100% fixed.
Resilience to price declines:
A 5% price fall for a company earning a net profit margin of 15% will result in the net
profit margin falling to 10%. So, the higher the net margin the less destructive is a given
price fall.
Depending upon the stage of the business cycle operational gearing can be either useful
or destructive. It may prove advantageous to use a screen for operational gearing as a
check on whether, for example, a value stock is on a low rating because it has a high level
of operational gearing and is about to enter a cyclical downturn.
4.5 Risk of failure
Syspas Z-Score:
Syspas looked at a sample of failed companies in order to see whether these companies
had similar accounting characteristics leading up to the point of failure. Companies
82
Determinants of Company Financial Performance
deemed ‘at risk of failure’ were given a negative Z-Score. The findings are summarised
below,
1) Turnover - At risk companies had 30% of the turnover of the sample
2) Profits - 44% report losses vs. 3% for strong firms
Profitable at risk companies had margins of 6% compared to 12%
for profitable strong firms
3) Assets - At risk companies spend less on fixed assets, typically 4.5% of
turnover vs. 7.5% of turnover for strong companies.
4) Capital Structure -
Healthy At Risk
Net Worth 47 13
Creditors 27 36
Long Term Debt 16 36
Provisions 8 3
Short Term Debt 2 12
Theory suggests that investing in high-risk (negative Z-Score) companies would provide
greater returns. However Professor Richard Taffler (1997) of the City University
Business School, found in a sample of all UK quoted companies over the period 1983-
1994, that those with a negative Z-Score underperformed the positive Z-Score portfolio
by an average of 4.7% p.a.
83
Determinants of Company Financial Performance
There were however dramatic periods of low Z-Score outperformance ( 6/1984 - 1/1987 )
and underperformance ( 1/1987 – 1/1992 ), reflecting a strong business cycle effect.
The Z-Score could be used in the process of portfolio, screening in order to make sure
that low PER or low PTBV stocks are not reflecting a risk of company failure.
4.6 A test for correlation between accounting ratios and outperforming companies
In their book “In Search of Excellence” Peters and Waterman (1982) attempted to define
the characteristics of a company with better than average growth prospects. They took a
list of companies that were regarded as superior by businessmen and screened them for
common attributes. These were found to be,
Attribute 1961-1980 Average
1) Rate of growth in corporate assets 21.78 %
2) Rate of growth in book value 18.43 %
3) Average ratio of market price to book value 2.46 x
4) Average return on corporate assets 16.04 %
5) Average return on book value 19.05 %
6) Average ratio of net income to sales 8.62 %
In a study of the subsequent performance of these companies, Clayman (1987) found that
the share prices of the group that she called the ‘unexcellent’ (the 39 companies with the
84
Determinants of Company Financial Performance
worse combination of the six characteristics) had risen by 297%. Whilst, the share prices
of the excellent companies had risen by only 181%. In addition, the rates of growth in
assets and book value nearly halved for the excellent group and reductions were seen in
other attributes as well.
Clayman’s study clearly demonstrates the concept of ‘reversion to the mean’ (as
discussed in Section 2.6.2). This strengthens the case for value based strategies as
creating a portfolio of the days best performing companies (in terms of accounting ratios)
is shown to prove unsuccessful.
85
Ranked Portfolio Tests: UK Smaller Companies
5.0 RANKED PORTFOLIO TESTS FOR MARKET
INEFFICIENCY: UK SMALLER COMPANIES
86
Ranked Portfolio Tests: UK Smaller Companies
5.0 RANKED PORTFOLIO TESTS FOR MARKET INEFFICIENCY:
UK SMALLER COMPANIES
5.1 Tests for market inefficiencies
In the chapters of the Literature Review, Introduction to Equity Valuation, and
Determinants of Company Financial Performance it was demonstrated that exposure to
the following factors would have produced portfolio returns in excess of the market
index.
1) Low firm Size
2) Low price to book
3) Low price to sales
4) High dividend yield
5) Mid-high debt to equity ratio
6) Low PE ratio
7) Low three/five year historic share price performance (positive return over 3-5 year
holding period)
8) High six month historic share price performance (positive return over one year
holding period)
9) High twelve month historic share price performance (positive return over three month
holding period)
10) Low accruals (scaled for MV)
87
Ranked Portfolio Tests: UK Smaller Companies
11) Top 10% of earnings performers
12) Positive earnings surprise
13) High upward revision in earnings forecasts
14) Ou and Penman (1989a) ‘Pr’ value of greater than 0
15) Attractive Estep (1987) valuation
In addition, Valuation Theory indicates that those companies with a high Return on
Equity and those companies that convert a high level of earnings into cash should
perform well.
The factors of positive earnings surprise and high upward revision in earnings forecasts
would seem worth investigating for UK Smaller Companies but unfortunately, it has not
been possible to gain access to the I/B/E/S earnings forecast data set. This is
disappointing given the high returns of these strategies. In the case of Ou and Penman’s
Pr measure and Estep’s T valuation, these are too complex to calculate given the time
frame of this study.
Therefore tests for possible inefficiencies within the FTSE Smaller Company Index, were
conducted on the following attributes.
1) Low price to book
2) Low price to sales
3) High dividend yield
88
Ranked Portfolio Tests: UK Smaller Companies
4) Low PE ratio
5) High six month historic share price performance (positive return over one year
holding period)
6) ROE
7) Cash EPS as a percentage of EPS
Plus, combinations of any strategy that was found to be successful.
5.2 Methodology
5.2.1 Definition of the FTSE Smaller Company (Ex IT) Index
The FTSE Smaller Company Index is made up of all companies in the FTSE All-Share
Index that are not in the FTSE 350 Index. The FTSE 350 Index consists of the 350 largest
(by market capitalisation) companies that are quoted on the London Stock Exchange, and
the FTSE All-Share Index consists of companies representing the top 98% (by market
capitalisation) of quoted companies. The FTSE Smaller Company (Ex IT) Index,
(referred to as the FTSE Smaller Company Index from here on) is the FTSE Smaller
Company Index after eliminating all Investment Trust companies. This is the usual
benchmark index for UK Smaller Company portfolio managers as they specialise in UK
equities, whereas the Investment Trusts that are included in the FTSE All-Share Index (of
which 82 are in the ‘Smaller Company’ market capitalisation band) can invest
internationally.
89
Ranked Portfolio Tests: UK Smaller Companies
5.2.2 Data collection
The data for this study was extracted from the Datastream database.
5.2.2.1 Constituents of the proxy smaller company index
In order to test the index-relative performance of quartiles of portfolios of companies
ranked by a particular accounting ratio, it was necessary to have a list of the constituents
of the benchmark index for past years. However it was found that FTSE International will
only make the data available on a commercial basis, and that Datastream does not hold
the FTSE Smaller Company Index constituent data.
Due to this, it was necessary to assemble a universe of stocks that could act as a proxy
index. The proxy index contained the stocks that would have been candidates for the
FTSE Smaller Company Index at a particular point in time.
Datastream holds the constituents of the FTSE All-Share index from the period 1994 to
date (as of 29/12 of each year). From this constituent list it is possible to construct a
reasonable proxy for the FTSE Smaller Company Index by eliminating companies below
and above the FTSE Smaller company threshold level at that date. These levels are based
upon market capitalisation, and are detailed below.
90
Ranked Portfolio Tests: UK Smaller Companies
Figure 5.01: Market Capitalisation Cut-Off Points for the Proxy Smaller Company
Index
Year-End Low Mkt. Cap. Company High Mkt. Cap. Company
1994 £8.46m Ross Group £285.79 Lex Service
1995 £10.10m Calderburn £470.60m Trinity Int.
1996 £4.92m NSW £354.83m Chiroscience
1997 £8.56m Automotive Hldgs. £370.83 Micro Focus
1998 £37.95m Fortune Oil £338.40m AMV
Source: FTSE International PLC.
The constituents of the FTSE All-Share Index for each year were down loaded into an
Excel file and ‘stripped’ in order to leave only companies whose market capitalisation
was between the above bands. At this point Investment Trusts were removed.
Next, three more sectors were eliminated. These were Pharmaceutical (Biotech)
companies, Oil and Gas exploration companies, and Real Estate companies. This was
because the above sectors are valued based upon their assets (or more correctly, their
projected assets). Often they do not have current earnings or sales, and consequently if
they were kept in the proxy index, they would distort the results.
By this process, a proxy universe, (constituents of a proxy Smaller Company Index) of
stocks was created on the 31st of December for the years of 1994 to 1998.
91
Ranked Portfolio Tests: UK Smaller Companies
5.2.2.2 Portfolio creation dates
Important requirements for test portfolio formation are,
1) Like for like ratios are compared.
The selection universe must consist of companies with the same financial year-end.
This is because the information value of, for example, the historic PE ratio erodes
over time. Consider the historic PE ratio of a company on the day before and the day
after it reports its earnings. On the day after the results are reported, the information is
current, an investor can assess the earnings of last year on that days share price.
However, on the day before the results, the ratio is almost a year out of date. On that
day investors will pay a higher historic PE ratio, as the company has had almost a
year to improve on its historic earnings. For this reason, at a particular date,
companies with different year-ends will be at a different point in their reporting cycle.
Selection rule: At a particular date select companies from a universe of companies
that share the same financial year-end.
2) Ratios are calculated at their most useful point in time
To have a reasonable chance of identifying market inefficiencies, ratio information
must be assessed when it is most relevant. For this exercise, this would be the day that
the results were reported.
92
Ranked Portfolio Tests: UK Smaller Companies
Selection rule: At a particular date select companies from a universe of companies
with the same financial year-end. The date should be the day that results were
reported.
3) The screening criteria must be available at the portfolio creation date.
If a company has a financial year-end in December, then full-year results will
normally be announced to the Stock Market in March of the following year. Thus, for
tests of portfolios consisting of companies with year-ends in December, ratios were
calculated using the stock price as at the first day of April of the following year. It is
assumed that all companies would have reported their results by three months after
their financial year-end.
Selection rule: At a particular date, select companies from a universe of companies
with the same financial year-end. The date for the share price should be three months
after the financial year-end.
4) Within the sample portfolios, company-specific risk must be minimised.
One problem with portfolio formulation by criteria is that the sample size reduces
fairly quickly. For the tests, it was considered that the minimum number of shares in a
portfolio should be eighteen. Although not ideal, (this is a function of the size of the
universe of around 450 companies and the aim to assess quartile ranked share price
performance), this size of portfolio is large enough to diversify away approximately
80% of the company specific risk. Due to the constraint of a minimum quartile size of
93
Ranked Portfolio Tests: UK Smaller Companies
eighteen stocks, only companies with a December and March year-end have sufficient
representation. See below.
Figure 5.02: Year-End Distribution for December 1994 Proxy-Index
Month of YE 1 2 3 4 5 6 7 8 9 10 11 12
No. of Cos. 18 5 86 27 11 30 7 6 34 9 6 115
5) A reasonable number of sample portfolios must be built in order to achieve statistical
significance.
The more samples that are taken, the more the data will reflect the characteristics of
the Population. Around thirty data points are regarded as sufficient in order to start to
draw statistically significant conclusions about the Population.
If the selection rules are adhered to with regard to point 4 above, and as the index
constituents are only available from December 1994, then only 8 data points would be
collected. I.e. Full-year results for March and December year-end companies, for a period
of four and a half years. In order to increase the sample size, it was thought reasonable to
include interim results, as these provide new valuation information. However, this still
results in only 15 data points. So, due to the limited data availability, it was considered
acceptable to create portfolios with a ‘six month from financial year-end’ time lag. In
support of this decision, recall the finding that it takes up to three months for analysts to
incorporate new information into forecasts (see Section 2.5). These lagged portfolios give
investors three months to disseminate the information reported in the Company Accounts
94
Ranked Portfolio Tests: UK Smaller Companies
(as opposed to portfolio formation on (or near to) the day of the results for the non-lagged
samples).
For companies with December and March year-ends the tables below illustrate the
portfolio formation dates, and the proxy index from which the companies are selected.
Figure 5.03: Portfolio Creation Dates
Constituents from Proxy Index at 31/12
December Year-end, Full-year results. Ratios at:
December Year-end, Interim results. Ratios at:
19941995 1/4/95 1/10/951996 1/4/96 1/10/961997 1/4/97 1/10/971998 1/4/98
Figure 5.04: Portfolio Creation Dates
Constituents from Proxy Index at 31/12
December Year-end, Full-year results. Lagged; Ratios at:
December Year-end, Interim results. Lagged; Ratios at:
19941995 1/7/95 1/1/961996 1/7/96 1/1/971997 1/7/97 1/1/981998 1/7/98
Figure 5.05: Portfolio Creation Dates
Constituents from Proxy Index at 31/12
March Year-end, Full-year results. Ratios at:
March Year-end, Interim results. Ratios at:
1994 1/1/951995 1/7/95 1/1/961996 1/7/96 1/1/971997 1/7/97 1/1/981998 1/7/98
95
Ranked Portfolio Tests: UK Smaller Companies
Figure 5.06: Portfolio Creation Dates
Constituents from Proxy Index at 31/12
March Year-end, Full-year results. Lagged; Ratios at:
March Year-end, Interim results. Lagged; Ratios at:
19941995 1/10/95 1/4/951996 1/10/96 1/4/961997 1/10/97 1/4/971998 1/10/98 * 1/4/98
* All share price performance tests are for the period of 1/10/98 – 1/8/99
5.2.2.3 Time Period
The measurement of share price performance against the FTSE Smaller Companies Index
was over the twelve months following the formation of the portfolio. It was considered
(in line with the Literature) that this was a sufficiently long period for share prices to
react to information contained in the financial ratios. In addition, fund managers are often
assessed over a rolling twelve-month period.
5.2.2.4 Test portfolios
With regard to Figures 5.03-5.06, it will be noted that the year for the proxy index
universe is sometimes different to that of the year of the formation of the test portfolio.
This was in order to match the period during which the performance of the portfolios was
measured, with the universe from which the stocks were selected. See, for example,
“March year-end, interim results, ratios at: 1/1/95”, the proxy index for this portfolio is
from 31/12/94, as this is nearer to the portfolio creation date than 31/12/95.
Portfolios were each allocated a code based upon Year of proxy portfolio, Month of
creation of test portfolio, Year of creation, Year of end of measurement period of return,
96
Ranked Portfolio Tests: UK Smaller Companies
Month of financial year-end, and whether the sample was taken after full-year, or interim
figures. If ‘L’ is within the code, then this signifies a portfolio that has been created with
a Lag, i.e. three months after figures were released.
For example, 9419596MINT, represents a portfolio consisting of companies from the
1994 proxy universe, with a March year-end, that have recently presented their interim
results; the portfolio is constructed in January 1995, and total return is measured from
January 1995 until January 1996.
For the portfolio screening tests, each list of possible portfolio constituents from the 30
date points was taken in turn; then companies were sorted in order of the attribute that
was being assessed. First a screen to eliminate outliers was made, then the portfolio was
ranked into quartiles. Over the subsequent year, the total return of each quartile was noted
and compared against the total return of the FTSE Smaller Company Index. This was
then repeated for other attributes.
In a laborious process, 300 different portfolios were constructed in order to measure the
total return for the 10 test characteristics, over the 30 time periods.
97
Ranked Portfolio Tests: UK Smaller Companies
5.3 Consideration of the significance of the results
It was necessary to assess the total return for the portfolios in two ways. First there was
the possibility that the data could mislead if a number of results distorted the calculated
mean value (i.e. they may not be representative of the true population). To quantify
whether the mean value was representative, the two tailed Student t test was used.
Second, as investors are concerned about both the absolute return of a particular strategy,
and the volatility (standard deviation) of the return, then some assessment had to be made
as to the trade off between return and the consistency of that return. This was evaluated
by calculating the Sharpe Ratio.
5.3.1 The Student’s t test
The Student t test was used to calculate whether the average performance of a particular
quartile was significantly different from zero. The Student t distribution was used
because the population standard deviation was not known and also because the sample
size was small (30 portfolios). In using the t distribution it was assumed that the
population was approximately normal.
The central (null) hypothesis of the test was that the portfolio did not outperform the
FTSE Smaller Company Index, i.e. the outperformance was zero. This is denoted as:
H0: μhe alternate hypothesis was that the performance of the portfolio was not
equal to zero, H1: μ≠. As the level of outperformance could be greater than, or equal
to, zero then a two-tailed test was appropriate.
98
Ranked Portfolio Tests: UK Smaller Companies
The level of significance of the test is the probability of accepting the alternate hypothesis
(in this case “that the characteristic outperforms the FTSE Smaller Company Index”)
when, in fact, the sample mean is not significantly different from zero. The higher the
level of significance is, the more sure that we can be in accepting the alternate
hypothesis. The tests were undertaken at 10, 5, 2, and 1% levels of significance (1% is
high, 10% is low). The process of the t test is detailed below.
The sample size was 30, so the test was for 29 ‘degrees of freedom’. (n=30, d.f.=29)
H0: μ
H1: μ≠
The Estimated standard error of the mean, σ = __ σ __ n
The sample mean was then ‘standardised’ by subtracting the hypothesised mean (H0:
μ), and dividing by the Estimated standard error of the mean, σ . The test statistic
is,
t = - μH0
σ
If the value of t was greater than + or - the critical value of t (that is the value of t from
the t distribution for a given level of significance and degrees of freedom) then the null
99
Ranked Portfolio Tests: UK Smaller Companies
hypothesis (in this case the ‘sample does not outperform’) was rejected. The critical
values for t in the tests were,
100
Ranked Portfolio Tests: UK Smaller Companies
To 10% significance: +/- 1.699
5% significance: +/- 2.045
2% significance: +/- 2.462
1% significance: +/- 2.756
5.3.2 Portfolio risk
A rational investor would prefer a higher level of return for a given level of risk.
However the trade off between risk and return is the subject of much debate. The Sharpe,
Jensen alpha (αp) and Treynor ratios are most commonly used to assess this relationship.
The Sharpe ratio is defined as,
Sharpe ratio = (Rp - Rf)/Rp)
Whilst the Jensen αp is defined as,
Jensen αp = p (Rm - Rf) - (Rp - Rf)
(This is derived from yearly regressions of the portfolio return)
And the Treynor ratio is defined as,
Treynor measure = (Rp - Rf)/p,
101
Ranked Portfolio Tests: UK Smaller Companies
Where for each ratio,
Rp - Rf = portfolio excess return
Rf = average risk free holding-period return,
Rm = market holding period-return,
Rp = average portfolio holding period-return,
Rp) = standard deviation of excess (to risk free) returns for portfolio p.
p = portfolio p’s index of systematic risk,
The Treynor ratio is dependent upon calculating a β value for the portfolio. In effect, the
portfolio return is standardised by the CAPM expected return. The Jensen alpha assesses
portfolio risk in a similar fashion. As there is an ongoing debate about the usefulness of β
(See the Literature Review), and as the calculation of 300 portfolio Beta values would be
time consuming, it was decided not to use the Treynor ratio or Jensen alpha to assess the
risk-reward relationship.
For the Sharpe ratio the excess return is defined as the portfolio return over the risk free
rate. The Sharpe ratio, scales excess return by the volatility (standard deviation) of the
excess return. This is a basic statistical measure, and is relatively easy to understand. The
higher the excess return that is achieved per unit of risk (standard deviation), the higher is
the Sharpe ratio. Due to its simplicity of understanding and calculation, this study
measures the risk-adjusted excess return of portfolios by their Sharpe ratio.. Datastream
does not construct a benchmark yield for one-year UK Government Bonds so the yield
for three-month Bonds is taken as the risk free rate.
102
Ranked Portfolio Tests: UK Smaller Companies
5.3.3 Stage of business cycle
This study focuses on the performance of Smaller Company shares during the period of
1/1/95 to 1/8/99. In the analysis of a data sample it is important to note any variables that
may have affected the results. To help achieve this, the year on year percentage change in
UK GDP during the study period is shown below in Figure 5.7.
Figure 5.07: UK GDP Growth (% YOY) 6/1990 – 6/1999
Source: Datastream
103
Ranked Portfolio Tests: UK Smaller Companies
On the chart, the trend growth of UK GDP is plotted at 2.25% p.a. This is, approximately,
the rate of real growth that the UK economy has enjoyed over the medium term. It can be
seen that during most of the period of this study the UK economy was growing at a rate
above its normal trend. Consequently the most rigorous interpretation of the results of the
tests would be that UK Smaller Companies perform in the way indicated, during a period
of above trend economic growth.
5.4 Portfolio ranking, performance and analysis
The total returns (capital plus income) of the individual portfolios formed for each test of
market inefficiency are detailed in Appendix 8.2.
5.4.1 The proxy universe
The first test to be conducted was to see whether the thirty proxy-index portfolios
performed similarly to the FTSE Smaller Company Index. If these did not perform
similarly, then the usefulness of the subsequent results would be limited to proxy
universes constructed in the same way, rather than to the FTSE Smaller Companies
Index.
104
Ranked Portfolio Tests: UK Smaller Companies
Figure 5.08: Test of the Acceptability of Portfolios as a Proxy the FTSE Smaller
Company Index
Portfolio Mean Outperformance (%)
Portfolio Mean Outperformance (%)
9419596MINT -6.85 9549596LMINT -4.59519697MINT -11.3 9579596LDFY 1.99549596DFY 2.9 95109596LMFY -6.69579596MFY -3.1 9519697LDINT -2.795109596DINT -0.4 9649697LMINT -11.79619798MINT -2.8 9679697LDFY -5.49649697DFY -7.6 96109697LMFY -7.19679697MFY -12.9 9619798LDINT 1.896109697DINT -1.8 9749798LMINT -6.69719899MINT 4.5 9779798LDFY 09749798DFY -4.9 97109798LMFY 1.19779798MFY -1.9 9719899LDINT 4.397109798DINT 3.7 9849899LMINT 4.89849899DFY -3.7 9879899LDFY -5.49879899MFY 1.9 981098To899LMFY 6.6
Mean (%) -2.5Standard Deviation 5.18
Null Hypothesis Outperformance =
0
Degrees of Freedom 29Estimated standard error of Pop. mean
2.02
Standardised Value -1.22
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
105
Ranked Portfolio Tests: UK Smaller Companies
The t statistic for the thirty portfolios with H0: μand H1: μ≠ is –1.22. This
indicates that to the 10% significance level, the null hypothesis of zero outperformance
against the FTSE Smaller Company Index cannot be rejected.
Therefore the thirty test portfolios are representative of the FTSE Smaller Company
Index.
5.4.2 Price to earnings
Datastream definition:
This is the ratio of the current share price divided by the earnings per share reported over
the last twelve months.
Earnings per share = Net Profit after Tax, minority interest and preference dividends,
(excluding pre tax extraordinary items, non-operating provisions
and transfers to tax-exempt reserves and any other items not
considered to normal trading activities of the company), divided by
the average number of shares in issue during the period, (adjusted
for scrip and rights issues).
Each test portfolio was sorted by the PE ratio of its constituents. Within the data set there
were found to be extreme values for the PE ratio. At the low end were companies that
were loss makers (i.e. with negative PE ratios), and at the high end were companies that
were making insignificant earnings per share relative to their share price. In order to
106
Ranked Portfolio Tests: UK Smaller Companies
screen the test portfolios so that a general investment rule could be established (if any),
then the outliers needed to be eliminated.
This is a difficult stage of data analysis as the researcher could be accused of data mining.
Data mining is when a sample is manipulated in order to achieve the results that are
required. One approach that is used to eliminate outliers is to delete all data points that
are above, and below, two standard deviations from the mean. The rule would result in a
portfolio that incorporates 95% of the data points. Although, even the choice of two
(rather than one, or three…) standard deviations from the mean could be questioned.
This study takes a different approach. Whilst wishing to be academically rigorous (i.e.
not wishing to be accused of data mining), the aim of this project is to create investment
rules that have a theoretical backing. A two standard deviation rule still leaves companies
in the data set with very low and very high PE ratios, these will tend to be companies in
financial distress, and the aim is to use the investment rules to invest in companies that
are acceptable to an institutional investor.
107
Ranked Portfolio Tests: UK Smaller Companies
With reference to Section 3.1.2, Damodaran has expressed the PE ratio as,
P0___ = Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) EPS0 (r-g)
+ Payout ration * (1+g) n * (1+g n)__ (r-gn) * (1+r)n
Where,
n = Number of years of high growth/visibility
g = Expected growth rate in equity and earnings in the first n years
gn = Expected growth rate in equity and earnings in the subsequent years
Payout ratio = Payout ratio for first n years
Payout ration = Payout ratio after n years
r = Cost of equity
In order to calculate upper and lower bounds for PE ratios a high-growth company and a
low-growth company were characterised.
108
Ranked Portfolio Tests: UK Smaller Companies
Figure 5.09: Theoretical PE Ratios for High and Low Growth Companies
Factor Low Growth High GrowthNumber of years of high growth/visibility 6 6ROE for first n years 10.6% 25%Payout ratio for first n years 33% 25%Expected growth rate in equity and earnings in the first n years
7% 19%
ROE after n years 10.6 13Payout ratio after n years 33% 25%Expected growth rate in equity and earnings in the subsequent years
7 10
Cost of equity 10.57 10.57Damodaran PE ratio 10.13x 53.29x
The ‘number of years of high growth/visibility’ is set at six years. This follows from the
work of Fuller et al. (1993) as discussed in Section 2.6.2. (They found that investors are
able to correctly judge the relative EPS growth rate of a company for up to six years
ahead.)
The ‘Expected growth rate in equity’ is calculated as Growth Rate = Return on Equity *
Retention ratio. For an explanation of this relationship, see Section 4.1. The retention
ratio is defined as 1- the payout ratio.
The cost of equity is the yield on a 10 year benchmark bond (@9/7/99) plus an assumed
equity risk premium (ERP) of 5%. (There is currently a debate as to whether the ERP is
declining from historic levels. For UK Equities, the ERP for the period of 1919-1993 was
6.14% 1. Five percent was considered a reasonable assumption).
1 BZW Equity and Gilt Study 1993
109
Ranked Portfolio Tests: UK Smaller Companies
Based upon the above assumptions the Damodaran PE ratio ranges from approximately
10x for a low-growth company to approximately 50x for a high-growth company. As 10x
is a theoretical rating for a low growth company, and because we wish to leave in the
sample companies that are ‘under priced’, then the low PER band is set at 7x historic
earnings. The high band is set at 50x historic earnings.
For the test, companies in each proxy index universe with a December or March year-end
and PE ratios of between 7x and 50x were sorted into ascending order of PER, and then
into PER quartiles. The total return of each quartile was then measured over the
following year and noted, along with the average size of the market capitalisation of the
quartile. This process was repeated for the 30 portfolio creation dates, in order to
calculate an average value.
110
Ranked Portfolio Tests: UK Smaller Companies
The results of the PE Ratio screen are summarised below.
Figure 5.10: Performance of Portfolio Quartiles by PE Ratio
Quartile FTSE SC Index
1 (Low)
2 3 4 (High)
Strategy Performance relative to the FTSC Index
Mean (%)
-4.29 -6.45 -3.44 3.87
Standard deviation
13.82 5.94 7.58 7.16
Estimated standard error of the population mean
2.57 1.10 1.41 1.33
t value -1.67 -5.84 -2.44 2.91Sharpe ratio 4.9 -8.5 12.9 59.6 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The t values for quartiles 2, 3, and 4 are highly significant. This means that the null
hypothesis of zero outperformance can be rejected. The null hypothesis of zero
outperformance can also be rejected for the first quartile (but only at the 10% level).
Over the period, the forth quartile of PE ratios outperformed the FTSE Smaller Company
Index by an average of 3.87%. This value is significant at a level of 1% (critical value of t
= +/- 2.46). The strategy outperformed in nineteen out of the thirty periods measured, so
positive returns were recorded for 63% of the time.
111
Ranked Portfolio Tests: UK Smaller Companies
The Sharpe ratio of the FTSE Smaller Company Index is 39.6 (this will remain constant
over all of the tests). The average Sharpe ratio of the forth quartile portfolios is 59.6, so
the benefit of the excess return of the strategy is not hindered by excessive volatility
relative to the volatility of the return of the FTSE Smaller Company Index.
Therefore, from the evidence above, it is statistically valid to state that “Investments
should be in made in companies with a PE ratio in the forth quartile of an ascending PER
ranking.”
5.4.2.1 Discussion of the PER screen
Interestingly, the result of the PER screen for UK Smaller Companies is the opposite of
the result of earlier studies. Graham (1973), Basu (1977,1983), Lakonishok et al. (1994),
and Goodman and Peavy (1983) all found that low PER stocks tend to outperform.
Whereas this study found that high PER stocks have outperformed, and that low PER
stocks have underperformed.
First the differences between previous studies and this must be addressed. Previous
studies have assessed all companies quoted on the stock market, whereas this study has
deliberately focussed on Smaller Companies. In earlier studies of the market as a whole,
it was noted that Smaller Companies tend to outperform, and that they also have lower
than market PE ratios; for this reason previous work has attempted to eliminate the ‘size
bias’ in results, by scaling for market capitalisation. This was done to counter the (pro-
CAPM) argument that low PER stocks outperformed in order to compensate investors for
112
Ranked Portfolio Tests: UK Smaller Companies
the higher default risk of investing in Smaller Companies. One would expect then, that
the results of this study would reflect the previous non-MV scaled finding that low MV,
low PER stocks outperform. Is it possible that there is a different size effect happening
within the Smaller Company sector?
In terms of the distribution of the PE ratio, the sample seems quite similar to the non-MV
scaled ‘all market’ studies. For the PER test in this study, the average MV of Q1, Q2, Q3
and Q4 was £137m, £157m, £165m and £162m respectively. The MV of Q4 is higher
than Q1, so in a similar fashion to the ‘all market’ studies, the PE ratio is a positive
function of size; however in this study Q4 outperforms, rather than Q1.
If the reason for the variance in quartile performance was due to risk, then in common
with past research, this study would be expected to show that a high PER strategy
underperforms. This is because investing in larger ‘Smaller Companies’ is considered
less risky by investors, for example, because of higher liquidity, see Brennan and
Subrahmanyam (1988), Section 2.7.3.3.
This study illustrates that in the case of Smaller Companies, investing in high risk/ low
MV / low PER companies is penalized: once again, reward does not equal the risk taken.
See also Figures 2.01 and 2.05.
However, this does not explain why the high PER stocks in the sample outperform.
113
Ranked Portfolio Tests: UK Smaller Companies
Another way of looking at the result is to say that investors are systematically paying too
low a PE, for a high PE ratio company. Relative to the market, over a twelve-month
period, the price of a high PER company is seen to increase. In ‘all market’ studies, it is
theorised that the low PER sector, outperforms because companies tend to revert to mean
performance, and that high PER stocks tend to underperform investors expectations. See
for example, Figures 2.10 and 2.11.
In this study, the outperformance of high PER stocks could be due to the earnings of
these companies surprising on the upside. It may be that the expected earnings growth for
the high PER sector was too low, or it could be that earnings growth remained higher
than investor’s original expectations.
During the period of this study the UK economy was growing fairly strongly, see Figure
5.07. The best (high PER) Smaller Companies would be expected to benefit from this
growth environment. During the period the rate of inflation was also reducing, companies
would be reporting high earnings growth, but in a climate of low inflation, investors may
have been cautious as to the sustainability of the high earnings growth demonstrated by
high PER companies.
If a less favorable economic environment were forecast, then PE ratios would normally
be expected to discount lower earnings in the future. However, as the economic
environment continued to be positive and above trend, then the market may have been
consistently ‘surprised’ by strong company earnings. Share prices would have to adjust
114
Ranked Portfolio Tests: UK Smaller Companies
upwards in order to maintain a constant PE ratio. The longer the boom continued, the
longer growth stocks (high PER stocks) remained attractive, and the longer the low PER
stocks remained unattractive. During the Literature Review, a long-term study of the
Smaller Company sector by PER was not found, so it is not possible to cross check the
results of this study. The ‘growth phase’ argument seems a reasonable explanation as to
the outperformance of high PER Smaller Companies, over the time period studied.
Until a longer-term data set is available (one that incorporates a number of business
cycles), the investment rule should be limited to “During periods of above trend
economic growth, invest in Smaller Company shares with a high PER”.
Rule One: During periods of above trend economic growth, invest in
companies with a PER in the highest 25% of all companies. Avoid companies
with a PER in the lowest 75% of all companies.
5.4.3 Price to book
Datastream definition:
This is the ratio of the current share price divided by the Book value per share.
Book value per share = Share capital reserves minus intangible and deferred assets,
divided by the average number of shares in issue during the
period, (adjusted for scrip and rights issues).
115
Ranked Portfolio Tests: UK Smaller Companies
A price to book value ratio of less 0.5x would mean that investors are willing to pay only
50 pence, or less, for each £1 of balance sheet assets. This would indicate that a company
is in financial distress. In order to select companies that were not in financial distress, the
lower bound of the price to book value ratio was set at 0.5x. To check whether this was a
reasonable level, the Damodaran value for the PTBV ratio was calculated for the low
growth company in Figure 5.09. The Damodaran formula for the PTBV ratio is discussed
in Section 3.1.3 and restated below.
P0___ = ROE * Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) BV0 (r-g)
+ Payout ration * (1+g) n * (1+g n) (r-gn) * (1+r)n
Where,
n = Number of years of visibility
g = Expected growth rate in equity and earnings in the first n years
gn = Expected growth rate in equity and earnings in the subsequent years
Payout ratio = Payout ratio for first n years
Payout ration = Payout ratio after n years
r = Cost of equity
Based upon the above assumptions, the Damodaran value for the PTBV of a Low Growth
company is 1.29x. The limit of 0.5x would therefore seem a reasonable indicator of a
116
Ranked Portfolio Tests: UK Smaller Companies
company in distress. There was no upper bound set for book value, as book value is
relatively stable over time. This means that (in contrast to the problem with high PE
ratios) there are few companies on extremely high PTBV ratios. NB: According to the
assumption in Figure 5.09, the price for a high growth company could be justified up to a
PTBV ratio of 18x.
The results of the PTBV screen are summarised below.
Figure 5.11: Performance of Portfolio Quartiles by PTBV Ratio
Quartile FTSE SC Index
1 (Low)
2 3 4 (High)
Strategy Performance relative to the FTSC Index
Mean (%)
-1.18 -3.70 -5.80 -2.66
Standard deviation
5.21 6.26 11.01 9.18
Estimated standard error of the population mean
0.97 1.16 2.04 1.70
t value -1.22 -3.19 -2.84 -1.56Sharpe ratio 31.3 10.7 -3.7 14.4 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The t statistics for quartile 1 and quartile 4 are not significant to 10 %. However quartiles
2 and 3 are significant to the 1% level.
117
Ranked Portfolio Tests: UK Smaller Companies
If the sample is assessed by bottom half of PTBV and top half of PTBV, the means are
-2.2 and -4.3, and the t statistics are –2.67 and –2.64 respectively. These figures are
significant to the level of 2%. The data shows that price to book value cannot be used as a
screen for investments to buy. Another basis for rejecting the PTBV as a screen is that the
Sharpe ratio for each quartile is below the Sharpe ratio for the market. Although as the
figures for quartiles 2 and 3 are highly significant, the PTBV screen could be used in the
sense that stocks with a PTBV in quartiles 2 and 3 should be avoided.
5.4.3.1 Discussion of the PTBV screen
The PTBV was found to be a useful screen by Rosenberg, Reid, and Lanstein (1985), and
Fama and French (1992). In addition, De Silva (Figure 2.04) scaled companies by MV
and found the low PTBV effect to be more pronounced in Smaller Companies.
It is difficult to interpret the PTBV test in this study, as only Q2 and Q3 are significant.
The underperformance of Q3 is greater than Q2, so it is possible that the data displays a
trend for higher PTBV ratio companies to underperform lower PTBV ratio companies,
but unfortunately the mean values for Q1 and Q4 are not significant to 10% so the trend
cannot be confirmed further. In addition the whole sample underperforms the market.
Theoretically, this is possible, as the test portfolios are equally weighted, and the FTSE
Smaller Company Index is Market Value weighted. (The reason that this study tests
equally weighted portfolios, is because Smaller Company investors do not track the index
in the same way as larger company investors. I.e. they are happy to take on more
company specific risk.)
118
Ranked Portfolio Tests: UK Smaller Companies
It is possible that at the level of MV for Smaller Companies, the results of the PTBV tests
are inconclusive, as investors do not see the PTBV ratio as a relevant valuation measure.
When valuing a small growth company the main concern is likely of intangible assets
rather than physical property. A market cap weighted test over a longer time period
would further aid understanding.
Rule Two: The tests do not indicate that PTBV is a relevant portfolio screen
for Smaller Companies.
5.4.4 Price to sales
Datastream definition:
This is the ratio of the current share price divided by sales per share.
Sales per share = Goods and services to third parties relating to the normal activities of
the company, (net of sales related taxes), divided by the average
number of shares in issue during the period, (adjusted for scrip and
rights issues).
The price to sales ratio is relatively stable over time, see Section 2.4.3. In addition its use
as a valuation tool benefits from the fact that sales are not subject to the accounting
adjustment of depreciation that distorts the PER and PTBV ratios, see Section 4.2.1.
119
Ranked Portfolio Tests: UK Smaller Companies
The Damodaran derivation for the PS ratio (see Section in 3.1.4) is given below.
P0___= Net Profit Margin1 * Payout ratio * (1+g) * (1- ((1+g) n /(1+r) n )) S0 (r-g)
+ Payout ration * (1+g) n * (1+g n) (r-gn) * (1+r)n
With a Net Profit Margin of 15%1, and a sustainable ROE of 10.6% as the required equity
return, (i.e. assuming mean reversion of the ROE at year six), the high growth company
in Figure 5.09, would be correctly valued at a price to sales ratio of 3.6x. It was therefore
considered appropriate to set an upper bound of 10x for the price to sales ratio in order to
screen out companies for which the prospect of meaningful sales can only be at some
time in the very distant future. (E.g. Internet Companies).
1 According to Company Refs. data the market median net margin is 5.5 %, with the upper quartile at 10.5%
120
Ranked Portfolio Tests: UK Smaller Companies
The results of the PSR screen are summarised below.
Figure 5.12: Performance of Portfolio Quartiles by PS Ratio
Quartile FTSE SC Index
1 (Low)
2 3 4 (High)
Strategy Performance relative to the FTSC Index
Mean (%)
-2.54 -7.24 -4.16 2.25
Standard deviation
8.62 6.60 10.52 8.25
Estimated standard error of the population mean
1.60 1.23 1.95 1.53
t value -1.59 -5.91 -2.13 1.47Sharpe ratio 16.3 -16.2 5.7 46.3 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The t statistics for Q2 and Q3 are significant to the level of 1 and 5 percent respectively.
Therefore both of these quartiles underperform the FTSE Smaller Company Index to a
significant degree. The underperformance of Q1 is not significant to 10%. Unfortunately
the outperformance of Q4 is also not significant to 10% (it is significant to 15%).
However due to the higher than market Sharpe ratio (46.3), this quartile will be
considered further in a screen of a combination of factors. Q4 PSR outperformed the
FTSE Smaller Company Index in 56% of periods.
121
Ranked Portfolio Tests: UK Smaller Companies
5.4.4.1 Discussion of the PSR screen
Senchack and Martin (1987) and Jacobs and Levy (1988) found that a low PS ratio was a
significant factor in determining stock outperformance. The results of the PSR test in this
study seem to indicate a trend in the other direction. High PSR outperforms (but not
significant to 10%), whereas Q2 PSR underperforms.
In a similar argument to that of the PER test (Section 5.4.2.1), high PSR stocks should
benefit from a positive economic environment. High PSR stocks will tend to have a high
net margin (see Section 3.1.4). As sales will tend to increase in a period of growth, each
additional £ of sales for a higher net margin company, will translate into a higher level of
profitability (than for a low net margin company). Consequently, the outperformance of
the high PSR sector could be due to the positive economic environment during the time
period of the study, rather than a discernable long-term trend for high PSR companies to
outperform.
Rule Three: During periods of above trend economic growth, avoid investing
in companies in the second and third quartiles of companies ranked by PSR,
(possibly) invest in Q4 PSR companies.
122
Ranked Portfolio Tests: UK Smaller Companies
5.4.5 Return on equity
Datastream definition:
ROE = Net Profit after Tax, minority interest and preference dividends (this excludes pre
tax extraordinary items, non-operating provisions and transfers to tax-exempt
reserves and any other items not considered to be normal trading activities of the
company), divided by equity capital and reserves (less intangibles, plus deferred
tax).
Companies with a Return on Equity of between 0 and 100 percent were included in the
test portfolios. This eliminates loss making companies and those companies that have a
ROE that is abnormally high (this maybe due to writing down assets excessively, etc).
The results of the ROE screen are summarised below.
123
Ranked Portfolio Tests: UK Smaller Companies
Figure 5.13: Performance of Portfolio Quartiles by ROE
Quartile FTSE SC Index
1 (Low)
2 3 4 (High)
Strategy Performance relative to the FTSC Index
Mean (%)
-4.55 -3.14 -4.08 -0.31
Standard deviation
7.91 6.36 9.43 11.78
Estimated standard error of the population mean
1.47 1.18 1.75 2.19
t value -3.10 -2.66 -2.33 -0.14Sharpe ratio 5.2 13.2 6.1 29.0 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The t statistics for Q1 and Q2 (the lower half of the sample) are significant to 1% and 2%
respectively. These quartiles underperform the FTSE Smaller Company Index. The same
is true of the Q3 sample (to 5%). The result for the forth quartile is not statistically
significant. This sample indicates that avoiding low ROE companies will be beneficial
for portfolio performance.
5.4.5.1 Discussion of the ROE screen
In Section 4.1, it was noted that the ROE is a driver of the rate of growth of earnings per
share. In Section 2.6.1 it was shown the ROE is also relatively stable over time.
Therefore it would be expected that high ROE companies would have high stable rates of
growth. In addition, in Section 5.4.3 it was shown that the PTBV and ROE ratios are
124
Ranked Portfolio Tests: UK Smaller Companies
mathematically related, accordingly if low PTBV stocks were to outperform, then low
ROE stocks should also outperform. The low PTBV strategy stems from the results of the
contrarian (mean reversion) strategies discussed in the Literature Review. However
because ROE is relatively stable, this could mean that mean reversion might not occur
within the time scale of this study. In addition, it seems (from Section 5.4.3) that the
PTBV ratio is also an unreliable screen for UK Smaller Companies.
The results indicate the opposite of previous research, at least to the level of Q1 and Q2
ROE underperformance. Once again this could be a reflection of a growth economy
during the period of the study. High ROE stocks achieve high EPS growth, see Section
4.1, these stocks would be highly rated and at some point contrarian theory suggests that
they would disappoint; but in the period of 1994-1999, they continued to perform well.
During the period, investors may have sold companies with a low ROE in order to buy
the high ROE companies that were benefiting most from the positive economic
environment. Another possibility is that whatever the environment, Smaller Company
investors do not wish to invest in low ROE companies. However to prove this statement,
data over a longer time period would be necessary.
Rule Four: During periods of above trend economic growth, avoid investing
in companies in the lowest 75% of companies ranked by ROE.
125
Ranked Portfolio Tests: UK Smaller Companies
5.4.6 Cash EPS % of EPS
Datastream definition of Cash EPS (CEPS):
CEPS = Net Profit after Tax, minority interest and preference dividends, (this excludes
pre tax extraordinary items, non-operating provisions and transfers to tax-
exempt reserves and any other items not considered to normal trading activities
of the company), + [depreciation + adjustment for assets sold + depreciation of
assets leased out] + deferred tax + overseas tax equalisation, divided by the
average number of shares in issue during the period, (adjusted for scrip and
rights issues).
CEPS_ * 100 = Datastream definition of CEPS (see above) * 100 EPS Datastream definition of EPS (see Section 5.4.2)
Cash EPS as a percentage of EPS measures how much of the accounting earnings of a
company are converted into cash. For the test samples, a figure of 100 indicates that all
earnings have been converted into cash, below 100 indicates that cash is being absorbed,
above 100 indicates that cash is being released.
126
Ranked Portfolio Tests: UK Smaller Companies
The results of the CEPS/EPS screen are summarised below.
Figure 5.14: Performance of Portfolio Quartiles by CEPS/EPS
Quartile FTSE SC Index
1 (Low)
2 3 4 (High)
Strategy Performance relative to the FTSC Index
Mean (%)
-3.05 -5.53 -1.86 -0.49
Standard deviation
7.95 8.47 8.13 7.56
Estimated standard error of the population mean
1.48 1.57 1.51 1.40
t value -2.07 -3.51 -1.23 -0.35Sharpe ratio 13.4 -2.5 21.1 34.1 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The t statistics for Q1 and Q2 are significant to 5% and 1% respectively. Companies that
do not convert EPS into Cash EPS tend to underperform the FTSE Smaller Company
Index over the following year. The results of Q3 and Q4 are not statistically significant
(to 10%).
5.4.6.1 Discussion of the CEPS % of EPS screen
Alfred Rappaport (1986), see Section 3.2, emphasised the importance of real cash, rather
than accounting profitability in company valuation. Sloan (1996) and Bernstein (1993),
see Section 2.7.2, put forward the theory, and demonstrated that, companies that had low
127
Ranked Portfolio Tests: UK Smaller Companies
levels of accruals tended to outperform. In addition, Philips (1998), see Section 4.2.3,
found that companies that generated high levels of cash as a percentage of profits tended
to outperform. All of the above studies noted the superiority of cash flow over accounting
earnings.
The test of cash EPS as a percentage of EPS indicates that a low conversion rate of
profits into cash tends to result in share price underperformance. The results for Q1 and
Q2 agree with previous research.
Rule Five: During periods of above trend economic growth, avoid investing in
companies in the lowest 50% of companies ranked by Cash EPS as a
percentage of EPS
5.4.7 Dividend Yield
Datastream definition:
Yield is calculated as the current share price, divided by dividends paid in the last
(rolling) twelve-month period.
It was not possible to draw a distinction between companies that do not pay a dividend
because of 1) financial distress or 2) a high re-investment rate (presumably as a
consequence of superior investment returns). So it was decided to eliminate zero yield
companies from the selection universe. There was no upper bound put on yield.
128
Ranked Portfolio Tests: UK Smaller Companies
The results of the DYLD screen are summarised below.
Figure 5.15: Performance of Portfolio Quartiles by Dividend Yield
Quartile FTSE SC Index
1 (Low)
2 3 4 (High)
Strategy Performance relative to the FTSC Index
Mean (%)
2.26 -2.29 -6.52 -5.07
Standard deviation
9.39 7.64 8.25 9.04
Estimated standard error of the population mean
1.74 1.42 1.53 1.68
t value 1.30 -1.61 -4.26 -3.02Sharpe ratio 42.1 19.4 -11.1 0.7 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The results for quartiles 3 and 4 are highly significant with t statistics that are significant
at the 1% level. Companies that have yields in the top half of yield ranking tend to
underperform the FTSE Smaller Company Index over the following year. The results for
quartiles 1 and 2 are not significant at the 10% level. The Sharpe ratio for Q1 is slightly
higher than that for that for the FTSE Smaller Company Index, however this figure
cannot be relied upon as the quartile outperformance may be due to chance.
129
Ranked Portfolio Tests: UK Smaller Companies
5.4.7.1 Discussion of DYLD screen
Dreman (1998) found that stocks with a high yield tend to outperform, see Section
2.4.6.1. In contrast, over the period studied in this test, high yield stocks have
underperformed.
As Miller and Modigliani (1961) showed, see Section 3.1, the dividend yield and PE ratio
are interrelated (via the payout ratio). Consequently, high yield stocks would tend to be
those stocks that are also low PER stocks. With reference to Section 5.4.2.1, the
underperformance of high yield stocks could be as a result of the positive economic
environment. Alternatively, the result may indicate that Smaller Company investors wish
to invest in companies with high a ROE and retention rate (for which a low dividend
yield could be a proxy).
Rule Six: During periods of above trend economic growth, avoid investing in
companies in the highest 50% of companies ranked by Dividend Yield.
5.4.8 Share price momentum
Definition of test ratio:
Price Change = Current share priceInitial Share Price
Index Change = Current FTSE Smaller Company Index ValueInitial FTSE Smaller Company Index Value
130
Ranked Portfolio Tests: UK Smaller Companies
Share price momentum = [Price Change – Index Change] x 100Index Change
Share price momentum was measured over the six months prior to the formation of the
test portfolio. Companies within the test portfolio were split into two groups, those with
negative momentum (underperformers) and those with positive momentum
(outperformers).
The results of the Momentum screen are summarised below.
Figure 5.16: Performance of Portfolio by Share Price Momentum
Momentum FTSE SC Index
Negative Positive
Strategy Performance relative to the FTSC Index
Mean (%)
-7.00 1.81
Standard deviation
8.96 6.34
Estimated standard error of the population mean
1.66 1.18
t value -4.21 1.54Sharpe ratio -11.3 50.6 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
131
Ranked Portfolio Tests: UK Smaller Companies
In the year following negative six-month share price momentum, companies
underperformed the FTSE Smaller Company Index by an average of 7%. As the mean
performance figure is significant to 1%, then this is clearly a factor to be avoided. The
positive momentum sector outperformed by 1.8% over the following year and the Sharpe
ratio (50.6) is higher than that of the FTSE Smaller Company Index (39.6), however the t
statistic is not significant to 10% (only to 13%). The positive momentum sector
outperformed in 21 out of 30 time periods (70%) and the negative momentum sector
underperformed for 25 out of 30 time periods (83%).
5.4.8.1 Discussion of the share price momentum screen
The tendency for six-month and twelve month share price momentum to persist, was
noted by Jegadeesh and Titman (1993), see Section 2.4.7.1. The results of the momentum
test indicate quite strongly that stocks with negative momentum should be avoided. The
degree of the under performance (7%), and the significance of the t Statistic (to 1%),
suggests that this screen may prove useful in a combination test.
Rule Seven: During periods of above trend economic growth, avoid investing
in companies with negative share price momentum over six months.
132
Ranked Portfolio Tests: UK Smaller Companies
5.5 Summary of single factor tests
The single factor strategies that displayed statistically significant t statistics are illustrated
below.
Figure 5.17: Statistically Significant Results for the Single Factor Screens
NB: The figures below (and above) the data columns indicate the level of significance of a two tailedStudent t test of the mean performance of the strategy.
In addition, to a level of significance of 1%, stocks with negative share price momentum
underperformed by 7.0%.
The objective of this study was to provide practical indications as to the investment
screens that could lead to portfolio outperformance. Although sixteen statistically
relevant inefficiencies have been demonstrated (by quartile), we cannot take practical
133
Ranked Portfolio Tests: UK Smaller Companies
advantage of any of the inefficiencies that indicate predictable underperformance,
because it is not possible to take short positions in the shares of Smaller Companies.
Therefore, the quartiles that underperformed would best be viewed as sectors to avoid.
The strategies that outperformed the FTSE Smaller Company Index, and that have a
Sharpe ratio of higher than the FTSE Smaller Company Index, are shown below.
Figure 5.18: Outperforming Single Factor Strategies
Strategy Q4 PER Q4 PSR + Momentum Q1 DYLD
Strategy Performance relative
to the FTSC Index (%)
3.87 2.25 1.18 2.26
Sharpe ratio 59.6 46.3 50.6 42.1
% of Time periods in which
the strategy outperforms
63 56 70 60
t value 2.91 1.60 1.54 1.3
NB: FTSE Smaller Company Index Sharpe ratio = 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The Q4 PER strategy has a high Sharpe ratio and a t statistic that is significant to 1%. The
next highest Sharpe ratios are for the + Momentum, followed by Q4 PSR, followed by
Q1 DYLD strategies. These are not significant to the level of 10%, however the t
134
Ranked Portfolio Tests: UK Smaller Companies
statistics of the + Momentum and PSR strategies are not too far away from the 10% level
(they are significant to the level of 13% and 15% respectively). It was decided to see
whether a combination of these strategies would enhance returns, over and above, the
returns of the individual strategies.
Dividend yield was not assessed further as 1) Statistically, it was the least significant and
2) It could be considered as a proxy for the PER.
5.6 Combination portfolios
The total returns of the individual portfolios formed for each combination test of market
inefficiency are detailed in Appendix 8.3.
As stated in Section 5.1.2.1, it was important to monitor the number of shares in each
portfolio to make sure that they had enough stocks in them to diversify away the majority
of stock-specific risk. When combination screens were run, the number of stocks that
satisfied the first criteria would be by definition only one quarter of the number of stocks
in the initial portfolio. The second screen would then reduce the sample further. For this
reason, quartiles could not be used in the combination tests, either for the initial criteria,
or for the reporting of results. This is as a consequence of the unavoidably small sample
size of the initial portfolios (e.g. there were only 80 stocks in 9419596MINT).
135
Ranked Portfolio Tests: UK Smaller Companies
5.6.1 PER and momentum
Stocks that had PE ratios within the top one third of a PER ranking were selected for the
initial screen. This group was chosen to reflect the outperformance demonstrated by the
top quartile PER strategy, and in order to satisfy the need to hold more than 18 stocks in
the test portfolio. Taking those stocks with positive momentum and those stocks with
negative momentum two test portfolios were then created.
The results of the PER-Momentum screen are summarised below.
Figure 5.19: Performance of Portfolio by PER and Share Price Momentum
Top Third PER FTSE SC Index
NegativeMomentum
PositiveMomentum
Strategy Performance relative to the FTSC Index
Mean (%)
-1.81 5.07
Standard deviation
10.35 8.24
Estimated standard error of the population mean
1.92 1.53
t value -0.94 3.31Sharpe ratio 20.3 63.4 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The outperformance of 5.07 % for the top third PER, positive momentum strategy is
significant to 1%. The strategy outperformed for 21 out of thirty periods, that is 70% of
136
Ranked Portfolio Tests: UK Smaller Companies
the time, and the Sharpe ratio is 63.4. The result for the top third PER, negative
momentum portfolio is not statistically significant to 10%.
5.6.1.1 Discussion of top third of PE ratio and positive momentum screen
The combination screen of PER and share price momentum, provides better results than a
screen of either of the two strategies taken individually (see Figure 5.2.2). The mean
outperformance is highly significant, and has consistently beaten the FTSE Smaller
Company Index (70% of time periods). Whilst the result may be limited to a positive
economic environment, see Section 5.4.2.1, when that environment persists, it would
seem that the strategy is far superior to that of traditional stock picking, see Malkiel
(1995) and Jensen (1968), Section 1.0.
Rule Eight: During periods of above trend economic growth, invest in
companies in the top third of companies ranked by PE ratio that have positive
six-month share price momentum.
5.6.2 PSR and momentum
Stocks that had a PS ratio within the top one third of the sample were selected on the
initial screen. Taking those stocks with positive momentum and those stocks with
negative momentum two test portfolios were then created.
The results of the PSR-Momentum screen are summarised below.
137
Ranked Portfolio Tests: UK Smaller Companies
Figure 5.20: Performance of Portfolio by PSR and Share Price Momentum.
Top Third PSR FTSE SC Index
NegativeMomentum
PositiveMomentum
Strategy Performance relative to the FTSC Index
Mean (%)
-4.09 5.20
Standard deviation
10.49 9.56
Estimated standard error of the population mean
1.95 1.77
t value -2.10 2.93Sharpe ratio 5.9 67.7 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
The t statistic for the top one third of stocks rated by PS ratio with positive six-month
share price momentum is significant to 1%. The portfolios outperformed the FTSE
Smaller Company Index by an average of 5.2%, and have a Sharpe ratio of 67.7. The
portfolios outperformed the market in 21 out of 30 periods (70%).
The t statistic for the negative momentum portfolios is significant to 5%. The mean
underperformance was 4.09 %, with the strategy underperforming in 19 out of 30 periods
(63%).
5.6.2.1 Discussion of top third of PS ratio and positive momentum screen
The result is highly significant, and has consistently beaten the FTSE Smaller Company
Index.
138
Ranked Portfolio Tests: UK Smaller Companies
Rule Nine: During periods of above trend economic growth, invest in companies in
the top third of companies ranked by PS ratio that have positive six-month share
price momentum. Avoid companies in the top third of companies ranked by PS
ratio that have negative six-month share price momentum.
5.7 Summary of combination portfolios
The statistically significant results from the combination portfolios are summarised
below.
Figure 5.21: Performance of Combination Strategies
Strategy Top 1/3 PER,
+ Momentum
Top 1/3 PS, +
Momentum
Top 1/3 PS, -
Momentum
Strategy Performance relative to
the FTSC Index (%)
5.07 5.20 -4.09
Sharpe ratio 63.4 67.7 5.9
% of Time periods that strategy
outperforms or underperforms
70 70 63
t value 3.31 2.93 -2.1
NB: FTSE Smaller Company Index Sharpe ratio = 39.6
Critical values for t: 10% significance: +/- 1.699@ 29 d.f. 5% significance: +/- 2.045
2% significance: +/- 2.4621% significance: +/- 2.756
139
Ranked Portfolio Tests: UK Smaller Companies
5.8 Conclusion of data analysis
Figure 5.22: Sharpe Ratios of Outperforming Strategies
NB: The figures above the data columns indicate the level of significance of a two tailed Student t test of the mean performance of the strategy.
With reference to Figures 5.18 and 5.21, the results of the combination portfolios show
that risk adjusted returns can be enhanced through screening a stock universe by more
than one factor.
Over 56% of time periods, the Q4 PSR strategy outperformed by an average of 2.25%
p.a., with a Sharpe ratio of 56.0. Over 60% of time periods the + Momentum strategy
outperformed by an average of 1.18% p.a., with a Sharpe ratio of 42.1. Together, the
140
Ranked Portfolio Tests: UK Smaller Companies
strategies outperformed for 70% of time periods by an average of 5.2% p.a., with a
Sharpe ratio of 67.7. The joint strategy increased return and risk-adjusted return.
The combination of an accounting ratio screen (PSR) and a momentum screen seems
helpful in picking the high PSR stocks that had the greatest potential for continued
outperformance. In addition, the underperformance of the high PSR and negative
momentum strategy indicated that within the high PSR group, there were certain stocks
that were being de-rated (as contrarian theory would suggest).
Momentum is demonstrated to be a useful indicator of those companies whose results
will disappoint or surprise investors over the following twelve months.
The top 1/3 PSR and positive momentum strategy beats the top 1/3 PER and positive
momentum strategy in terms of both average total return, and Sharpe ratio.
There are a few points that need to be noted before these strategies are followed.
1) A straight comparison between the single and combination strategies was not possible
as in the single factor test, the top quartile of stocks was held, rather than the top 1/3.
2) A disadvantage of including some three-month lagged portfolios, is that over the
three months post the results, the share price may have already reacted, and hence
outperformance may not be registered over the year-long holding period. This
141
Ranked Portfolio Tests: UK Smaller Companies
increases the likelihood of accepting the null-hypothesis that the selection criterion
does not outperform the FTSE Smaller Company Index. However as indicated in the
Literature Review (see Section 2.4.7.1) share price reaction tends to continue for up
to one year ahead.
3) In addition incorporating overlapping time periods could lead to a misleadingly high,
standardised t-statistic. This is because some of the same companies will have been
selected for both periods, and consequently the standard deviation of the mean
outperformance of the strategy will be reduced (i.e. some of the companies share
prices will perform in exactly the same way for a 9-month period). However a
cursory glance at the lagged and non-lagged portfolios, does not indicate as high a
degree of portfolio similarity as feared.
4) The test period of 1994 to 1999 does not incorporate enough business cycles to enable
a general strategy to develop. A particular strategy is only valid if the economic
environment is similar to that of the test period.
The above criticisms would be resolved if a longer time series of Smaller Company Index
constituents were available for analysis.
Nevertheless, the criticisms do not mitigate the results for the study, which are
statistically significant and indicate that a high level of outperformance can be achieved,
by following very simple rules.
142
Ranked Portfolio Tests: UK Smaller Companies
Two further tests were conducted. The first was a test of the Wilcox (1984) ROE - Log
PTBV relationship for which a high R2 between the two variables is demonstrated for UK
Smaller Companies (see Appendix 8.4).
The second (see Appendix 8.5) was a test of a low PTBV and high ROE combination
screen proposed by Damodaran (1994, Section 3.1.3). The results of following this
strategy indicate no significant outperformance.
143
Conclusion and Further Study
6.0 CONCLUSION AND FURTHER STUDY
144
Conclusion and Further Study
6.0 CONCLUSION AND FURTHER STUDY
Past studies have indicated patterns (market inefficiencies) in share price behaviour that
are predictable and profitable. The credibility of these patterns is enhanced as they have a
foundation in valuation theory. The contrarian strategies, the share price reversal, and the
momentum strategy, exist for the same reasons. These are 1) the desire of Investors to be
associated with success (consequently they overpay for stock), and 2) the failure of
Investors to accept that the performance of successful companies will eventually ‘revert
to the mean’. As these are basic human traits, which prove hard to counter, then there is
the possibility that the inefficiencies will continue; they are part of human nature.
6.1 Can the quantitative screening of UK Smaller Companies result in superior
investment returns?
In this study, the UK Smaller Company market has been shown to be inefficient. Seven
single factor quantitative screens, and three combination quantitative screens were
demonstrated to provide statistically significant excess returns.
During the period of the tests, through holding a portfolio of the top one third of
companies ranked by the PSR that had also outperformed the FTSE Smaller Company
Index over the preceding six months, it was possible to earn an excess return (vs. the
FTSE Smaller Company Index) of 5.2% p.a. The Sharpe ratio of the strategy was 67.7 vs.
39.6 for the Index. This means that the volatility of the total return of the strategy was not
excessive as compared to the volatility of the Index. Even in the unlikely event that the
145
Conclusion and Further Study
whole portfolio is turned over each year, (as some stocks will remain in the screen), the
bid-offer spread of 2.8% (Keim (1989) does not erode the gain excessively, in contrast
with many reported inefficiencies.
The results of the study are all the more interesting as the inefficiencies found in the PER,
PSR and DYLD screens go against the findings of previous research. There are three
possible reasons (or combinations of reasons) for this. First is that FTSE Smaller
Company Investors do not overpay for stock 1, second is that the financial performance of
Smaller Companies does not mean revert as quickly as Larger Companies. Third is that,
as the period of the study coincides with above trend economic growth, there is a
business-cycle-variant nature to whether high or low rated stocks outperform.
Given the high statistical significance of the reported results, reasons for the perverse
nature of Smaller Company share price performance warrant further investigation.
6.2 Further study
Due to the lack of I/B/E/S data the earnings based strategies that proved profitable in all-
market studies were not tested, these should be tested for UK Smaller Companies. Also
within the time frame of this study, it was not possible to test for all of the relationships
established in the review of the determinants of company financial performance. Now
that the relationships are understood, it would be interesting to see whether the share
price of companies that express the desired characteristics outperform the index.1 Even at a high PER, the super-normal earnings growth a Smaller Company may reduce the PER very quickly. Indeed this is one of the main reasons for investing in Smaller Companies.
146
Conclusion and Further Study
In addition, if a longer-term data set were available then a study of the performance of
ratio sorted portfolios during different points of the business cycle would further aid
understanding.
147
References
7.0 REFERENCES
148
References
7.0 REFERENCES
Abarbanell, J., and Bernard, V., 1992, “Tests of Analysts’ Overreaction/Underreaction to Earnings Information as an Explanation for Anomalous Stock Price Behavior,” Journal of Finance, 47, pp. 1181-1208.
Asness, C.,S., 1994, “The Power of past Stock Returns to Explain Future Stock Returns,” Manuscript, June.
Babcock, G.,C., 1980, “The Roots of Risk and Return,” Financial Analysts Journal, January/February.
Ball R., and Brown, P., 1968, “An Empirical Evaluation of Accounting Income Numbers,” Journal of Accounting Research, Fall, pp. 159-178.
Ball, R., 1992, “The Earnings-Price Anomaly,” Journal of Accounting and Economics, 15, pp. 319-345.
Banz, R.,W., 1981, “The relationship between return and market value of common stocks,” Journal of Financial Economics, 9, pp. 3-18.
Barefield, R.,M., and Cominsky, E.,E., 1975, “The Accuracy of Analysts’ Forecasts of Earnings per Share,” Journal of Business Research, 3, July, pp. 241-252.
Basu, S., 1977, “The investment performance of common stocks in relation to their price-earnings ratios: test of the efficient markets hypothesis,” Journal of Finance, 32, pp. 663-682.
Basu, S., 1983, “The Relationship between earnings Yield, Market Value, and Return for the NYSE Stocks”. Journal of Financial Economics, 12, June , pp. 129-156.
Bathke, A.,W., and Lorek, K.,S., 1984, “The Relationship between Time Series Models and the Security Markets Expectation of Quarterly Earnings,” The Accounting Review, 59, pp. 163-176.
Bauman, W.,S., and Miller, R.,E., 1997, “Why Value Stocks Outperform Growth Stocks,” Journal of Portfolio Management, Spring, pp. 57-68.
Beaver, W.,H., Clark , R., and Wright, W., 1979, “The Association between unsystematic Security Returns and the Magnitude of the Earnings Forecasting Error.” Journal of Accounting Research, .Autumn, pp. 316-340.
149
References
Beaver, W.,H., and Morse, D., 1978, “What determines Price-Earnings Ratios,” Financial Analysts Journal, July-August, pp. 65-76.
Beaver, W.,H., Kettler, P., and Scholes, M., 1969, “The Association Between Market Determined and Accounting Determined Risk Measures”, The Accounting Review, 44, pp. 654-682.
Bernard, V., and Thomas, J., 1990, “Evidence that Stock Prices do not fully reflect the Implications of Current Earnings for Future Earnings,” Journal of Accounting and Economics, 13, December, pp. 305-340.
Bernstein, L., 1993, “Financial Statement Analysis,” 5th. Ed., Homewood, IL: Irwin.
Bhandari, L.,C., 1988, “Debt/Equity Ratio and Expected Common Stock Returns: Empirical Evidence,” Journal of Finance, XLIII, 2, June, pp. 507-530.
Black, F., 1972, “Capital Market Equilibrium with Restricted Borrowing,” Journal of Business, 45, pp. 444-465.
Black, F., and Scholes, M., 1974, “The effects of dividend yield and dividend policy on common stock prices and returns,” Journal of Financial Economics, 1, pp. 1-22.
Black, F., Jensen, M., and Scholes, M., 1972, “The capital asset pricing model: some empirical tests.” In Jensen, M., (Ed.), “Studies in the Theory of Capital Markets,” Praeger Publishers, New York.
Brealey, R.,A., and Myers, S.,C., 1996, “Principles of Corporate Finance,” McGraw-Hill, New York, International Edition, pp. 184.
Breen, W.,J., and Korajczyk, R., 1994, “On Selection Biases in Book to Market Based Tests of Asset Pricing Models,” Working Paper, Northwestern Univ., USA.
Brennan, M.,J., and Subrahmanyam, A., 1996, “Market Microstructure and Asset Pricing: On the Compensation for Illiquidity in Stock Returns,” Journal of Financial Economics, 41, pp. 341-364.
Brown, L., D., and Rozeff, M., S., 1979, “Univariate Time Series Models of Quarterly Accounting Earnings per Share: a Proposed Model,” Journal of Accounting Research, pp. 178-189.
Campbell, J., Y., and Shiller, R., J., 1988, “Stock Prices, Earnings and Expected Dividends”, The Journal of Finance, 43, pp. 661-676.
Capaul, C., Rowley, I., and Sharpe, W.,F., 1993, “International Value and Growth Stock Returns,” The Financial Analysts Journal, 49, pp. 27-36.
150
References
Chan, L.,K.,C., Hamao, Y., and Lakonishok, J., 1991, “Fundamentals and Stock Returns in Japan”, Journal of Finance, 46, pp. 1739-1789.
Chen, H., Roll, R., and Ross, S.,A., 1986, “Economic Forces and the Stock Market,” Journal of Business, 59, pp. 383-404.
Chopra, N., Lakonishok, J., and Ritter, J.,R., 1992, “Measuring abnormal performance: Do Stocks Overreact?,” Journal of Financial Economics, 31, pp. 235-268.
Clayman, M., 1987, “In Search of Excellence: The Investor’s Viewpoint,” Financial Analysts Journal, May-June.
Clubb, C., and Charitou, A., 1999, “Earnings, Cash Flows and Security Returns over Long Return Intervals: Analysis and UK Evidence,” Journal of Business Finance and Accounting, 26,Nos. 3 & 4, pp. 283-312.
Collins, W., and Hopwood, W., 1980, “A Multivariate Analysis of Annual Earnings Forecasts Generated from Quarterly Forecasts of Financial Analysts and Univariate Time Series Models,” Journal of Accounting Research.
Connor, G., and Korajczyk, R., 1988, “Risk and Return in an Equilibrium APT: Application of a new test methodology,” Journal of Financial Economics, 21, pp. 255-290.
Connor, G., and Korajczyk, R., 1993, “A Test for the Number of Factors in an Approximate Factor Model,” Journal of Finance, 48, pp. 1263-1291.
Copeland, T.,L., Koller, T., and Murrin, J., 1998, “Valuation: Measuring and Managing the value of Companies,” John Wiley & Sons, New York.
Cragg, J.,G., and Malkiel, B.,G., 1968, “The Consensus and Accuracy of Predictions of the Growth of Corporate Earnings,” Journal of Finance 23, pp. 67-84.
Damodaran, A., 1994, “Damodaran on Valuation,” John Wiley & Sons, New York, pp. 198.
Davis, J.,L., 1994, “The Cross-Section of Realised Stock Returns: The Pre-COMPUSTAT Evidence,” Journal of Finance, 49, pp. 1579-1593.
De Bondt, W.,F.,M., and Thaler, R.,H., 1985, “Does the Stock Market Overreact?,” Journal of Finance, 40, pp. 793-805.
De Bondt, W.,F.,M., and Thaler, R.,H., 1990, “Do Security Analysts Overreact?,” American Economic Review, May, pp. 52-57.
151
References
De Silva, H., “What Underlies the Book-to-Market Effect,” Working Paper, Graduate School of Management, University of California, Irvine.
Deacon, E., B., 1976, “Distributions of some Financial Accounting Ratios”, The Accounting Review, Jan, pp. 90-96.
Dorfman, J.,R., 1991, “Analysts Devote More Time to Selling as Firms Keep Scorecard on Performance,” in “Heard on the Street,” Wall Street Journal, Oct. 29th, 1991.
Dreman, D.,N., 1998, “Contrarian Investment Strategies: The Next Generation,” Simon and Schuster, New York.
Easton, P.,D., Harris, T., S., and Ohlson, J., A., 1992, “Aggregate Accounting Earnings can explain Most of Security Returns: The case of Long Return Intervals,” Journal of Accounting and Economics, 15, pp. 119-142.
Elgers, P., and Murray, D., 1992, “The Relative and Complementary Performance of Analyst and Security-Price-Based Measures of Expected Earnings,” Journal of Accounting and Economics, 15, No. 2-3.
Estep, T,. 1987, “Security Analysis and Stock Selection: Turning Financial Information into Return Forecasts,” Financial Analysts Journal, July/August, pp. 34-43.
Ettredge, M., Toolson, R., Hall, S., and Na, C., 1996, “Behavior of Earnings, Stock Returns, Accruals, and Analysts’ Forecasts Following Negative Annual Earnings,” Review of Financial Economics, 5, No.2, pp. 147-162.
Fama, E.,F., and French, K.,R., 1992, “The Cross Section of Expected Stock Returns,” Journal of Finance, 47, pp. 427-466.
Fama, E.,F., and French, K.,R., 1993, “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics, 33, pp. 3-56.
Fama, E.,F., and French, K.,R., 1994, “Industry Costs of Equity,” Working Paper, Graduate School of Business, University of Chicago, Chicago, IL, revised July 1995.
Fama, E.,F., and French, K.,R., 1996, “Multifactor Explanations of Asset Pricing Anomalies,” Journal of Finance, Vol. LI, No. 1, March, pp. 55-84.
Fogler, H.,R., 1995, “Investment Analysis and New Quantitative Tools,” Journal of Portfolio Management, Summer, pp. 39- 48.
Freeman, R., Ohlson, J., Penman, S., 1982, “Book Rate of Return and Prediction of Earnings Changes: An Empirical investigation”, Journal of Accounting Research, Autumn, pt. 2, pp. 639-653.
152
References
Fried, D., and Givoly, D., 1982, “Financial Analysts Forecasts of Earnings: A Better Surrogate for Earnings Expectations,” Journal of Accounting and Economics 4, pp. 85-107.
Fuller, R.,J., Huberts, L.,C., and Levinson, M.,J., 1993, “Returns to E/P Strategies; Higgledy Piggledy Growth; Analysts Forecast Errors; and Omitted Risk Factors,” Journal of Portfolio Management, Winter
Ghysels, E., 1998, “On Stable Factor Structures in the Pricing of Risk/; Do Time-Varying Betas Help of Hurt ?,” Journal of Finance, 53, No.2, pp.549-573.
Gibbons, M.,R., 1982, “Multivariate tests of financial models: a new approach,” Journal of Financial Economics, 10, pp. 3-27.
Givoly, D., and Lakonishok, J., 1984, “The Quality of Analysts’ Forecasts of Earnings,” Financial Analysts Journal, 40, pp. 40-47.
Goodman, D.,A., and Peavy III, J.,W., 1983, “Industry Relative Price Earnings Ratios as indicators of Investment Returns,” Financial Analysts Journal, 39, pp. 60-66.
Gordon, M.,J., 1962, “The Investment, Financing and Valuation of a Corporation”, Homewood, Illinois: Richard D. Irwine, p.5.
Graham, B., 1973, “The Intelligent Investor,” Harper and Row, New York.
Graham, B., and Dodd, D.,L., 1934, “Security Analysis,” McGraw-Hill Book Company, New York.
Grossman, S., and Miller, M., 1988, “Liquidity and Market Structure,” Journal of Finance, 43, pp. 617-633.
Handa, P., Kothari, S.,P., and Wasley, C., 1989, “The Relation Between the Return Interval and Betas: Implications for the Size Effect,” Journal of Financial Economics, 23, pp. 79-100.
Haugen, R.,A., 1990, “Modern Investment Theory,” Englewood Cliffs, N.J., Prentice Hall.
Haugen, R.,A., 1995, “The New Finance: The Case against Efficient Markets,” Prentice Hall, Englewood Cliffs, New Jersey.
Hawkins, E.,H., Chamberlin, S.,C., and Daniel, W.,E., 1984, “Earnings Expectations and Security Prices”, Financial Analysts Journal 40, pp. 24-27, pp. 30-38, pp. 74.
153
References
Holthausen, R.,W., and Larcker, D.,F., 1992, “The Prediction of Stock Returns Using Financial Statement Information”, Journal of Accounting and Economics, 15, pp. 373-411.
Howton, S.,W., and Peterson, D.,R., 1998, “An Examination of Cross-Sectional Realised Returns using a Varying-Risk Beta Model,” The Financial Review, 33, pp. 199-212.
Hunter, J.,E., and Coggin, T.,D., 1988, “Analyst Judgement: The Efficient Market Hypothesis Versus a Psychological Theory of Human Judgement,” Organisational Behavior and Human Decision Processes, Vol. 42, Dec, pp. 284-302.
Itami, Hiroyumi, 1987, “Mobilising Hidden Assets”, Boston, Harvard University Press.
Jacobs, B., I., and Levy, K.,N., 1988, “Disentangling Equity Return Irregularities: New Insights and Investment Opportunities,” Financial Analysts Journal, 44, pp. 18-44.
Jegadeesh, N., 1990, “Evidence of Predictable Behavior of Security Returns,” Journal of Finance, 45, pp. 881-898.
Jegadeesh, N., and Titman, S., 1993, “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency,” Journal of Finance, 48, pp. 65-92.
Jegadeesh, N., and Titman, S., 1995, “Overreaction, delayed reaction, and contrarian profits,” Review of Financial Studies, 8, pp. 973-993.
Jegadeesh, N., and Titman, S., 1995, “Short-Horizon Return Reversals and the Bid-Ask Spread,” Journal of Financial Intermediation, 4, pp. 116-132.
Jensen, M., 1968, “The Performance of Mutual Funds in the Period 1945-1964,” Journal of Finance, 23, May, pp. 389-416.
Kahneman, D., and Tversky, A., 1982, “Intuitive Prediction: Biases and Corrective Procedures,” in Kahneman, D., Slovic, P., and Tversky, A., eds., “Judgement Under Uncertainty: Heuristics and Biases,” Cambridge University Press, London.
Keim, D.,B., 1983, “Size-Related Anomalies and Stock Return Seasonality,” Journal of Financial Economics, 12, pp. 13-32.
Keim, D.,B., 1989, “Trading Patterns, Bid-Ask Spreads and Estimated Security Returns: The case of Common Stocks at Calendar Turning Points,” Journal of Financial Economics, 25, pp. 75-98.
Kim, D., 1997, “A Reexamination of Firm Size, Book-to-Market, and Earnings Price in the Cross-Section of Expected Stock Returns,” Journal of Financial and Quantitative Analysis, 32, No.4, pp. 463-489.
154
References
Kothari, S.,P., and Shanken, J., 1995, “Book-To-Market, Dividend Yield, and Expected Market Returns: A Time-Series Analysis,” Working Paper, Univ. of Rochester.
Kothari, S.,P., Shanken, J., and Sloan, R.,G., 1995, “Another Look at the Cross-Section of Expected Stock Returns,” Journal of Finance, 50, pp. 185-224.
Lakonishok, J., Shleifer, A., and Vishny, R., 1994, “Contrarian Investment, Extrapolation, and Risk,” Journal of Finance, 49, pp. 1541-1578.
Lintner, J., 1965, “The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets,” Review of Economics and Statistics, 47, pp. 13-37.
Little, I.,M.,D., “Higgledy Piggledy Growth”, Institute of Statistics, Oxford (UK)Litzenberger, R., and Ramaswamy, K., 1979, “Dividends, Short-Selling Restrictions, Tax-Induced Investor Clienteles and Market Equilibrium,” Journal of Financial Economics, 7, pp. 163-196.
Lo, A.,W., and MacKinlay, A.,C., 1988, “Stock Prices Do Not Follow Random Walks: Evidence from a Simple Specification Test,” Review of Financial Studies, Vol. 1., Table 2.
Malkiel, B., 1995, “Returns From Investing in Equity Mutual Funds 1971 to 1991,” Journal of Finance, 50, June, pp. 549-573.
McTaggart, Kontes and Mankins, “The Value Imperative”
Miller, M., and Modigliani, F., 1961, “Dividend Policy, Growth, and the Valuation of Shares,” Journal of Business, October, pp. 411-433.
Miller, M.,M., and Scholes, M.,S., 1978, “Dividends and Taxes”, Journal of Financial Economics, 6, pp. 333-364.
Miller, M.,M., and Scholes, M.,S., 1982, “Dividends and Taxes: Some Empirical Evidence,” Journal of political Economy, 90, pp. 1118-1141.
Minard, L., 1984, “The Case Against Price/Earnings Ratios,” Forbes, February 13, pp. 172-176.
Mott, C.,E., and Coker, D.,P., 1993, “Earnings Surprise in the Small-Cap World,” Journal of Portfolio Management, Fall, pp. 64-75.
O’Brian, P., 1988, “Analyst’s Forecasts as Earnings Expectation,” Journal of Accounting and Economics 10, pp. 53-83.
155
References
O’Shaughnessy, J.,P., 1998, “What Works on Wall Street,” McGraw-Hill, N.Y.
Ou, J.A., & Penman, S.H., 1989, “Accounting Measurement, Price-Earnings Ratio, and the Information Content of Security Prices”, Journal of Accounting Research, Vol. 27, Supplement, pp111-144.
Ou, J.A., & Penman, S.H., 1989, “Financial Statement Analysis and The Prediction of Stock Returns”, Journal of Accounting and Economics, 11, pp. 295-329.
Peters, D.,J., 1991, “Valuing a growth stock,” Journal of Portfolio Management, 17, pp. 49-51.
Peters, T.,J., and Waterman, R.,H., 1982, “In Search of Excellence: Lessons From America’s Best run Corporations,” Harper and Row, New York.
Philips, H., 1998, “Acquisitions or returning cash,” Panmure Gordon, Feb.
Poterba, J.,M., and Summers, L.,H., 1988, “Mean Reversion of Stock Prices: Evidence and Implications,” Journal of Financial Economics, 22, October, pp. 27-59.
Rappaport A., 1986, “Creating Shareholder Value,” The Free Press, Macmillan Inc., New York, pp.
Rayner, A.,C., and Little, I.,M.,D., 1966, “Higgledy Piggledy Growth Again”, Basil Blackwell, Oxford (UK).
Reinganum, M.,R., 1981, “Misspecification of Capital Asset Pricing: Empirical Anomalies Based on Earnings Yields and Market Values,” Journal of Financial Economics, 9, March, pp. 19-46.
Richards, R.,M., 1976, “Analysts’ Performance and the Accuracy of Corporate Earnings Forecasts,” Journal of Business, 49, July, pp. 350-357.
Richards, R.,M., and Frazer, D.,R., 1977, “Further Evidence on the Accuracy of Analysts’ Earnings Forecasts: A Comparison Among Analysts,” Journal of Economics and Business, 29, Spring-Summer, pp. 193-197.
Richards, R.,M., and Martin, J.,D., 1979, “Revisions in Earnings Forecasts: How Much Response?,” Journal of Portfolio Management, 5, pp. 47-52.
Richards, R.,M., Benjamin, J.,J., and Strawser, R.,W., 1977, “An Examination of the Accuracy of Earnings Forecasts,” Financial Management, Fall.
Rosenburg, B., Reid, K., and Lanstein, R., 1985, “Persuasive Evidence of Market Inefficiency,” Journal of Portfolio Management, 11, pp. 9-17.
156
References
Ross, S.,A., 1976, “The Arbitrage Theory of Capital Asset Pricing,” Journal of Economic Theory, 13, pp. 341-360.
Ross-Healy, C., and Sgromo, E., 1993, “How to Beat the S&P 500 Index Using Credit Analysis Alone,” Journal of Portfolio Management, Winter, pp. 25-31.
Senchack Jr., A.,J., and Martin, J.,D., 1987, “The Relative Performance of the PSR and PER Investment Strategies,” Financial Analysts Journal, 43, pp.46-56.
Shanken, J., 1985, “Multibeta CAPM or Equilibrium APT?: A Reply,” Journal of Finance, 40, pp. 1189-1196.
Shanken, J., 1987, “Multivariate proxies and Arbitrage Pricing Relations: Living with the Roll Critique,” Journal of Financial Economics, 18, pp. 91-110.
Sharpe, W.,F., 1964, “Capital Asset Prices: a Theory of Market Equilibrium under Conditions of Risk,” Journal of Finance, 19, pp. 425-442.
Sloan, R.,G., 1996, “Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future Earnings,” The Accounting Review, 71, July, pp. 289-315.
Sorensen, E.,H., and Williamson, D.,A., 1985, “Some Evidence on the Value of the Dividend Discount Model,” Financial Analysts Journal 41, pp. 60-69.
Stambaugh, R.,F., 1982, “On the exclusion of assets from tests of the two-parameter model: a sensitivity analysis,” Journal of Financial Economics, 10, pp. 237-268.
Stewart Jr., S.,S., 1973, “Research Report on Corporate Forecasts,” Financial Analysts Journal, January-February, pp. 77-85.
Stewart, B., “The Quest for Value”, HarperCollins.
Stober, T.,L., 1992, “Summary Financial Statement Measures and Analysts’ Forecasts of Earnings,” Journal of Accounting and Economics, 15, pp. 347-372.
Stoll, H.,R., and Whaley, R.,E., 1983, “Transaction Costs and the Small Firm Effect,” Journal of Financial Economics, 12, June, pp. 57-79.
Taffler, R., 1997, “Enhancing Equity Returns with Z-Scores,” Professional Investor, July/August.
Thaler R.,H., ed., 1993, “Advances in Behavioral Finance,” Russell Sage Foundation, New York.
157
References
Tversky, A., 1995, “The Psychology of Decision Making,” in A. Wood (ed.), 1995, “Behavioral Finance and Decision Theory in Investment Management,” ICFA Continuing Education Series, pp. 2-6.
Vander Weide, J.,H., and Charlton, W.,T., 1988, “Investor Growth Expectations: Analysts Vs. History,” Journal of Portfolio Management 14, pp. 78-83.
Wilcox, J.,W., 1984, “The P/B-ROE Valuation Model,” Financial Analysts Journal, 40, pp. 58-66.
Wood, A.,S., ed., 1995, “Behavioral Finance and Decision Theory in Investment Management,” Association for Investment Management and Research, Charlottesville, USA.
158
Appendices
8.0 APPENDICES
159