Seoul Journal of Business

Volume 7, Number 1 (December 2001)

A Comparative Analysis of ROE and Value-to-Price

based Trading Rules: Do Conventional Risk Factors
Matter?

Sam Han
Uniuersity of Illinois at Champaign,IL USA

Tony Kang
University of Illinois at Champaign, IL USA

Abstract
This paper examines the profitability of ROE and value to price (VP)
based trading rules. We find that the ROE based trading rule generates
significant hedge portfolio return over 12-month period after portfolio
formation. In addition, we find t h a t t h e ROE-based trading rule
significantly under-performs trading rules based on VP ratio, especially
over longer horizon. However, the result indicates that the profitability
of the ROE trading rule is by and large subsumed by the conventional
risk factors.

I. Introduction

Return on equity (ROE) h a s been one of the most popular

profitability measures among investors. It measures how the
stockholders fared during the year. In a n accounting sense, ROE
is a true bottom-line performance measure to the stockholders
since it measures how much profit is earned during the year for
every dollar in equity. To the extent that investors form their
expectations a b o u t firms' future profitability based on this
measure, the investors will purchase (sell) stocks that performed
well (poorly) in the previous year. In such cases, the ROE will
predict future stock returns. Nevertheless, the ability of ROE to
predict future stock returns has not been fully exploited in the
literature. This study attempts to provide a preliminary analysis
of this unexplored link.
In the literature the importance of investors' expectation about
future ROE is highlighted in the residual income valuation
model (Frankel and Lee 1998). Under this model, the market's
expectation about future ROE determines stock price. In the
accounting literature, identifying mis-priced stocks using
financial statement information such a s ROE is consistent with
the goal of fundamental analysis (Penman 1992). Given that
ROE is closely related to intrinsic value-to-price (VP hereafter)
ratio under the model, we adopt the VP ratio a s a benchmark in
our analysis.
This study extends prior literature on the residual income
valuation model by exploring the return predictability of a n
accounting ratio, i.e., ROE. Specifically, this study examines the
direct link between current ROE and future stock returns that
h a s not yet been examined in the literature. Although the return
predictability of ROE is to some extent implied in Frankel and
Lee (1998)'s analysis, their approach is indirect in that they use
current ROE to first predict future ROE, which is in turn used to
predict future stock returns. In contrast, the approach we take
in this study is a one-step procedure since it does not involve
predicting future ROE.
O u r a n a l y s e s s h o w t h a t ROE t r a d i n g r u l e g e n e r a t e s
statistically significant hedge portfolio returns u p to one year
after portfolio formation. However, the result suggests that the
profitability of ROE investment strategy is for the most part
subsumed by the conventional risk factors, i.e., the market, size,
and book-to-market. Further, the evidence indicates that ROE
clearly under-performs VP beyond one year after portfolio
formation.
This paper is organized as follows. In the next section, we
summarize the theory. In section 111, we discuss the portfolio test
procedures of ROE a n d VP strategy. Section IV contains a
description of data and sample. In section V, we report the
empirical results. We conclude in section VI.
 A Conzpnvutive Anulysis of ROE and Value-to-Price hased Trading Rules 43

11. Theory

This research is mainly linked to two streams of research in

accounting and finance. First, this study is closely related to the
fundamental analysis line of literature in accounting (Ou and
Penman 1989, Holthausen and Larker 1992, Stober 1992, Lev
and Thiagarajan 1993, Abarbanell and Bushee 1998). Ou and
Penman (1987) document abnormal return to a single summary
measure, called Pr that predicts future earnings. They devise a
LOGIT model for predicting changes in annual EPS one-year
ahead, using 28 financial statement variables chosen from a
wide set of 68 variables. They report over 8% abnormal return in
the first year after EPS predictions are made.
Holthausen and Larker (1992) and Stober (1992) extend Ou
and Penman (1989). Holthausen and Larker (1992) find that a
trading rule that exploits the correlation between current period
financial statement data and future returns (without any
hypothesis about future earnings) dominates Ou and Penman
(1989)'s strategy. In contrast, Stober (1992) finds that the
abnormal returns documented in Ou and Penman's study (1989)
persist for over six years beyond the earning prediction date. He
thus concludes that the financial statement variables used by
Ou and Penman (1989) can be seen as proxies for expected
returns.
Lev and Thiagarajan (1993) and Abarbanell and Bushee (1998)
explore the ability of "fundamental signals" to explain and
predict firm performance. The "fundamental signals" represent
the core set of information frequently employed by professional
financial analysts for fundamental analyses. The signals, first
introduced to the literature by Lev and Thiagarajan (1993),
include contemporaneous changes in accounts receivables,
inventories, gross margins, selling expenses, capital
expenditures, effective tax rates, inventory methods, audit
qualifications, and labor force sales productivity. Lev and
Thiagarajan (1993) find that these fundamental signals explain
contemporaneous stock r e t u r n s . Extending their study,
Abarbanell and Bushee (1998) find that the fundamental signals
can be used to generate abnormal returns. Their result suggests
44 Seoul Jo~trnnlof Busilzess

that the fundamental signals contain information about future

returns that is associated with future earnings news. Their
trading rule based on the fundamental signals earns an average
12-month cumulative size-adjusted abnormal return of 13.2
percent.
Second, this study is related to prior studies that examine the
ability of financial ratios to predict returns. These ratios include
book-to-market (Kothari and Shanken 1997, Rosenberg, Reid,
and Lanstein 1985), earnings-to-price (Basu 1977), dividend
yield (Campbell and Shiller 1988, Fama and French 1998), and
intrinsic value-to-price (Frankel and Lee 1998) ratio. Basu
(1977) finds that earnings-to-price (EP hereafter) ratio predicts
stock returns. Rosengerg, Reid, and Lanstein (1985), Fama and
French (1992), and Kothari and Shanken (1997) find that the
ratios of book-to-market value of equity predict stock returns.
Campbell and Shiller (1988) and Fama and French (1988) find
that dividend yield predicts stock returns since they reflect
information about expected returns. Frankel and Lee (1998)
document that intrinsic value-to-price (W)ratio outperforms BP
in predicting returns. l )
The theoretical cornerstone of this study is the following
dividend discounting model that shows firm value should equal
the discounted present value of expected future dividend a t
equilibrium.

where P i s value (or the "true" price), p is discount rate and D is

dividend

The dividend discount model is mathematically equivalent to

the following residual income valuation model often referred to
as the Edwards-Bell-Ohlson (EBO) model under some restrictive
assumptions (clean surplus and convergence conditions):

1) In contrast, Francis, Olsson, and Oswald (2000) find that the profitability of
VP strategy is subsumed by the conventional risk factors - the market, size,
and book-to-market.
 A Cornpararive Analysis of ROE and Value-to-Price based Trading Rules 45

where P is value (or the "true" price), B is book value, p is discount

rate2),and Xis net income

The above residual income model shows that price can be

expressed a s book value i n time t(Bt) p l u s t h e market's
expectation about future residual income, which is the second
term on the right hand side of the equation. Manipulating this
term using book value a t time t-1 yields the following equation,
which shows that the expectation about future ROE determines
the price (Penman 1996, Frankel and Lee 1998):

where ROEt = Xt/Bt.l

Due to various reasons, however, the observed market price PC

may deviate from t h e " t r u e " price P t . 3 ) Lee, Myers, a n d
Swaminathan (1999) give a good summary of why the observed
price may deviate from the true price. They note t h a t the
observed price may diverge from the true price due to (1) noise
t r a d i n g (Schiller 1 9 8 8 , Delong, Shleifer, S u m m e r s , a n d
Waldmann 1990); (2) uninformed trading (Wang 1993); or (3)
difficulties associated with measuring the true price. They argue
that the magnitude and the duration of the deviations depend on
the costs of arbitrage. In such cases, the process by which price
adjusts to intrinsic value requires time, and observed price does
not always perfectly reflect true price. In such a world, a s they
claim, a more realistic depiction of the relation between observed
price and true price is one of continuous convergence rather
than static equality. This is because in the long run, arbitrage
forces cause price to converge to value, but in the short run, the
existence of nontrivial arbitrage costs may prevent observed
price from converging to true price. A typical example of such
arbitrage cost is trading cost, which can take the form of bid-ask

2) p = 1 + r, where r is long-term return on equity

3) Or alternatively, the "intrinsic"value (Lee, Myers, and Swaminathan 1999)
46 Seoul Journal of B~lsiness

s p r e a d , commissions, t h e cost of selling s h o r t , or t h e

opportunity costs associated with implementing the strategy
(Bernard and Thomas 1989). Stoll and Whaley (1983) estimate
average trading costs for small a n d large firm stocks a t
approximately four to two percents on an annual basis. Thus, if
the potential arbitrage profit is below this threshold, then a
rational investor would not decide to engage in trading, and
observed price may deviate from true price.
Given that PP may be different from Pt a t times but should
converge over time, investors may be interested in finding Pt.
This is because if they have a better idea of Pt, they can take
investment positions based on this knowledge to generate
arbitrage profit. Let's consider the following example. If there is a
stock that will have high (low) ROE in the future but the market
does not fully impound this information, it will be mis-priced,
i.e., Pt f PP. This discrepancy between Pt and PF will give rise to
arbitrage opportunities to the investors. If Pt < PtO(P, > P?), the
investors are likely to profit by shorting (longing) the stock. To
the extent that ROE is informative about P, trading rule based
on ROE becomes an interesting research question.

111. Test Procedures

Equation ( 2 ) , (3) represent procedures for estimating a firm's

true value (PJ. The four main parameters needed for both ROE
and VP test are: r e t u r n on equity (ROEt), book value (Bt),
forecasted future net income (FNIL+I), and cost of capital (p).
Consistent with previous literature (Bernard 1994, Beaver and
Ryan 2000) ROEt is defined a s NIL/BVt.l, where NI and BV are
net income (annual Compustat #18) and book value (annual
Compustat #60) measured at per-share level, respectively. The
number of shares is Compustat item #25, adjusted for stock
splits and dividends.
The V P is calculated as in equation (4), similar to Frankel and
Lee (1998).4)The model is kept rather simple based on Liu,
Nissim, and Thomas (2000)'s findings that more complicated

4) The main difference is that they used I/B/E/S forecasts to derive future
ROE estimates. Instead, we directly use the median I/B/E/S consensus
earnings forecasts. See appendix A in Frankle and Lee (1998) for more
details.
 A Conzlmrutive Arzaly.~isof ROE a ~ i dValue-to-Price baser1 Trndirzg Rules 47

specification of intrinsic value estimate provides only trivial

improvement over more parsimonious ones.

where FNI is forecasted future net income

The above specification forecasts abnormal earnings for two

periods and takes the last period in perpetuity. For FNlt+, and
FN1i+2, we u s e t h e m e d i a n I / B / E / S c o n s e n s u s a n a l y s t s '
forecasts issued a t each fiscal year end.5) For t h e dividend
payout ratio that is necessary to forecast future dividend, we
calculate firm-specific payout ratio by dividing the common
stock dividends paid in the most recent year (Compustat Item
#21) by net income before extraordinary items (Compustat Item
#237). Similar to Frankel and Lee (1998), we divide dividends by
ten percent of book value to compute a n estimated payout ratio
for firms with negative earnings. Finally, we adopt two types of
cost-of-capital estimates.

where Rj,, is a n annual cost-of-capital estimate of firm j, RF, is risk-

free rate,6)and RPj is the industry-specific risk premium for firm j

The first is firm-specific rate using the capital asset pricing

model and the second one is industry specific cost-of-capital
(Fama and French 1997). Since Lee, Myers, and Swaminathan
(1999) show that the estimation of intrinsic value is improved
when time-varying component of cost-of-capital is used, we
report results based on the industry and time-specific cost-of-
capital estimates (Fama and French 1997). However, results are
comparable when the firm-specific rates are used.
In order to investigate the ability of current ROE to predict
future ROE, we rank firms based on ROEs at each fiscal year
end and track future ROEs up to three years. The mean ROEs in

5) In order to avoid t h e situation where t h e asymmetric distribution of

forecasted earnings affects the results, we use the median forecasts.
6) RF,,,is a n annualized rate of one-month T-bill rate obtained from Ibbotson
and Associate's 1999 Year Book
48 Seoul Journal of Business

each decile are reported in Panel A of Table 2, and they are

plotted in Panel B. To examine the return implications, we first
compute future buy-and-hold returns. These buy-and-hold
r e t u r n s are computed by compounding monthly r e t u r n s
obtained from CRSP monthly tape. To ensure that financial
statement information necessary to compute ROE has reached
the investors, we start calculating returns four months after the
fiscal year end.7) Thus, year 't' return period for a firm with
December fiscal year end starts in April of year t+l and ends in
March of year t+2. An advantage of examining buy-and-hold
(both raw and market-adjusted) return is that it captures long-
term investor experience relatively well compared to other
methods of calculating long-term returns. In this vein, Barber
and Lyon (1997) recommend the use of buy-and-hold return to
measure long-term stock return performance. We compute both
raw (BHR) and market-adjusted (BHAR) buy-and-hold returns in
the following manner:

where rU = raw stock return for firm i in the porffolio in month j, Rj =

market return (S&P500 Composite Index) in month j

To compute the market-adjusted buy-and-hold returns for

individual securities, we subtract the market return from the
return of a security during the corresponding period. Using
these individual returns, we compute BHAR at the portfolio level
by equally weighting the individual returns. With these, we
compute hedge portfolio return, which is the difference between
the return for portfolio 10 and the return for portfolio 1 . This
return summarizes the predictive ability of each ratio with
respect to future returns.
To adjust for risk, we estimate the following two models. The
first adjustment procedure involves Jensen alphas (Ibbotson

7) This holding period has been frequently examined in the accounting

literature, but is different than the ones commonly examined in the finance
literature. A typical holding period in the finance literature (i.e. Fama and
French 1993) starts from July of year t+l, regardless of the fiscal-year-end
month in year t, where as that in the accounting literature starts three to
four months after each fiscal year end.
 A Comparative Analysis o f ROE and Value-to-Price based Trading Rules 49

1975). This procedure uses the following time-series regression

for each portfolio.

where Rj,,is average raw monthly returns of the portfolio in calendar-

month t, RFt is risk free rate in month t (Ibbottson Associate's one-
month T-bill rate), and RM, i s market return for month j (value-
weighted CRSP market return)

The Pj measures the relative risk of each portfolio, and aj,

Jensen alpha, captures the excess return with respect to beta.
Under this specification, the intercept, by construction,
measures the monthly average excess return with respect to the
beta. Thus, for instance, the intercept of .Ol means that the
monthly average abnormal returns to the trading strategy under
examination is 1% after controlling for the market premium.
This specification assumes that investors use the Sharpe-
Lintner version of the capital asset pricing model in forming
expectations about future returns.
The second adjustment procedure is regarding firm size and
book-to-market ratio (Fama and French 1993). Since these two
variables are well-documented predictors of future stock
returns, we include these variables to the previous specification
and estimate the following three-factor model:

where SMB, is the difference between the return of portfolios of

"small" stocks and "big" stocks in month t, HML, is the difference
between the return of portfolios of "high book-to-market stocks and
"low" book-to market stocks in month t

Similar to the previous specification, i.e. equation (7), the

intercept aj in this equation measures the average monthly
abnormal returns of a portfolio with respect to the three factors.
If the profitability of ROE trading rule is a result of its
correlation with some of these risk factors, then the intercept aj
should not be significantly different from zero. A value of pj
greater (less) than one means that firms in portfolio j are, on
50 Seoul Jourr~nlof Business

average, riskier (less risky) than the market. A value of 3;. greater
(less) than zero indicates the influence of small (large) stocks in
the portfolio on the portfolio return. A value of q greater (less)
than zero indicates the influence of high (low) book-to-market
stocks in the portfolio on the portfolio return. Fama (1998),
among others, notes that this methodology h a s a n advantage
(over other methods of measuring long-term stock performance)
of being able to account for the cross-sectional correlations
a c r o s s securities t h r o u g h t h e time-series variation of t h e
monthly abnormal stock returns.

IV. Data and Descriptive Statistics

The evidence presented in this paper is based on a sample of

29,140 firm-year observations during 1982 to 1995 with (1)non-
h i s s i n g earnings a n d book value d a t a on 1999 Compustat
Annual Primary, Secondary, and Tertiary, Full Coverage, and
Merged Research Files; (2) necessary price and return data in
1 9 9 9 CRSP files; (3) one a n d two-year-ahead c o n s e n s u s
analysts' forecasts available on 1999 I/B/E/S files. The number
of yearly observations ranges between 1,598 (in 1982) and 2,763
(in 1995) during the sample period. To control for survivorship
bias: we do not exclude observations that do not have ROE or
return d a t a in the years subsequent to portfolio formation.
Thus, a n observation in the sample can have missing future
return data in one or all of the five subsequent years.
Descriptive statistics and correlations are shown in Panel A of
Table 1. Pearson (Spearman) correlation i s reported above
(below) the diagonal. As expected, ROE is positively related to VP
b u t negatively related to BP. The correlation between current
ROE and other financial ratios are generally significant a t 99%
confidence level. Panel B of Table1 reports a summary statistics
of the selected variables during the sample period. The mean
and the median ROE during the sample period approximate .10
and .13 respecti~ely.~) The mean of VP and BP are .7159 and
.7346. These figures are comparable t ~ t h onese reported in the
previous studies (Bernard 1994, Frankel and Lee 1998, and
Fama a n d French 1992). On average, there a r e 2 , 0 8 1

8) Similarly, Bernard (1994)had the mean ROE of . 1 1 in his sample.

 A Conzpamtive Analysis of ROE and Value-to-Price based Trading R~lles 51

Table 1. Correlations and Summary Statistics of Selected Variablesa

Panel A: Correlation among Selected Variablesb

ROE VP BP SIZE
ROE 0.0247 -0.1997 0.2107
(.0001) (.0001) (.0001)
VP 0.0506 0.3227 -0.0462
(.0001) (.0001) (.0001)

Panel B: Summary Statistics

Variables Mean Median Std. Dev. Maximum Minimum

ROE
VF'
BP
SIZE
3mo Ret
6mo Ret
Yrl Ret
Yr2 Ret
Yr3 Ret
aThis table is based on 29,104 observations during 1981-1996. ROE, VP,
and BP refer to book-return on equity, intrinsic value-to-price, and book-
to-price ratio, respectively. Firm size is measured a s the log of total
market capitalization at the beginning of the period
b The correlation reported above (below) the diagonal is Pearson (Spearman)
correlation. Reported in parentheses are p-values.

observations each year during the sample period.g) The mean

and the median beta in the sample are approximately 1.71 and
1.88. They are estimated using a t least 24 monthly returns. The
market risk premium of 8.4%, the mean of historical premium

9) Firms can have missing data in future years.

52 Seoul Journal of Business

during 1929- 1998 period (Ibbotson and Associates 1999), is

used. The mean and the median of the resulting firm-specific
cost-of-capital e s t i m a t e s a r e 14.33% a n d 15.83%. The
corresponding figures for the industry cost-of-capital estimates
are 12.39% and 12.33%, respectively.

V. Results and Discussions

Table 2 reports the evolution of future ROE in ROE portfolios.

The mean-reversion in ROES over the years is consistent with
previous studies (Bernard 1994, Beaver and Ryan 2000). The
hedge portfolio ROE (the top decile portfolio ROE less the bottom
decile portfolio ROE) decreases from about 73% in year zero to
24% by the end of year three. The mean reversion is stronger for
the extreme portfolios, consistent with competitive forces driving
out abnormally high profitability while dislocation costs "trap"
some firms with abnormally low profits (Bernard 1994).
Table 3 and Table 4 show the future return performance of
ROE and VP portfolio. The evidence indicates that although the
ROE hedge portfolio returns are significant up to one year after
portfolio formation, it under-performs the benchmark portfolio
thereafter. Over one year after portfolio formation, the ROE
portfolio generates hedge return of about 2.8%, while the VP
portfolio generates 5.0%. The VP portfolio return is comparable
to the ones reported in previous studies. For instance, the first
year VP portfolio hedge return is approximately 4-5%, about the
same a s Dechow, Hutton, and Sloan (1999) who report about 6%
in their sample.lo)
Beyond one year, the VP portfolio dominates the ROE portfolio,
and the ROE hedge portfolio return becomes less significant.
The superior return predictability of VP over longer horizon is
consistent with Frankel a n d Lee (1998). The VP portfolio
generates close to 9% for two years after the portfolio formation,
and about 7% for three years. Table 5 shows the results from

10) Frankel and Lee (1998) obtain similar level of return (about 3.4%) in the first
year. If Fama and French (1993) holding period is applied to my sample, the
VP hedge portfolio return increases across years, and it exceed 5% in the
first year.
 A Conzl~arativeAnalysis of ROE and Value-to-Price based Trading Rules 53

Table 2. The Evolution of Future ROE in ROE PortfoliosC

Panel A: The Evolution of Future ROE in ROE Portfolio

Portfolio YrO ROE Yrl ROE Yr2 ROE Yr3 ROE

Panel B: The Evolution of Future ROE in ROE Portfolio (Panel A P10tted)~

I year 0 Year 1 Year 2

Holding Period

c ROE is book return-on-equity. Decile 1 (10) refers to low (high) ROE

portfolios, respectively. YrO ROE, Yrl ROE, Yr2 ROE, Yr3 ROE refer to
ROE at Year 0, one-year, two-year, and three-year subsequent to portfolio
formation, respectively.
d In the graph, Pfo 1, Pfo 2 , ..., Pfo 10 refer to decile 1 (lowest ROE) to decile
10 (highest ROE) portfolio, respectively.
54 Seoul Jourrznl of Business

Table 3. Future Return Performance of ROE Portfolioseaf

Panel A: ROE Portfolio Future Raw Return

Portfolio mnROE 3moRET 6moRET YrlRET Yr2RET Yr3RET

Hedge 0.0173 0.0205 0.0281 0.0217 0.0268

(t-stat) (2.3926)***(1.8953)** (1.6545)** (1.3424)* (1.6788)**

Panel B: ROE Portfolio Future MarketlAdjusted Return

Portfolio mnROE 3moRET 6moRET YrlRET

1 -0.3386 0.0067 0.0066 0.0076
2 -0.0198 0.0136 0.0024 0.0223
3 0.0509 0.0080 0.0122 0.0176
4 0.0887 0.0033 0.0116 0.0171
5 0.1164 0.0084 0.0157 0.0347
6 0.1384 0.0070 0.0168 0.0240
7 0.1591 0.0137 0.0243 0.0376
8 0.1849 0.0145 0.0258 0.0393
9 0.2264 0.0148 0.0134 0.0268
10 0.3955 0.0199 0.0191 0.0273
Hedge 0.0132 0.0124 0.0196 0.0229 0.0329
(t-stat) (1.8333)**
(3.3514)***
(4.3946)***
(1.4313)* (2.1087)**
e ROE indicates book return-on-equity. MnROE refers to the mean ROE in
the decile. GmoRet, 1YrRet, 2YrRet,3YrRet refer to returns six-month, one-
year, two-year, and three-year subsequent to portfolio formation,
respectively. The portfolios are formed four-months after each fiscal year
end in order to ensure that financial statement information has reached
the investors.
f ***, ** and * signify one-tailed statistical significance at the 196, 5% and
10%levels, respectively.
,
 A Conzparative A17alysisof ROE and Value-to-Price based Trading Rules 55

Table 4. Future Return Performance of Value-to-Price (VP) Portfolio~g.~

Panel A: VP Portfolio Future Raw Return

Portfolio mnVP 3moRET 6moRET YrlRET Yr2RET Yr3RET

Hedge 0.0078 0.0089 0.0497 0.0876 0.0589

(t-stat) (2.1.814)**
(0.9993) (3.5981)***(6.0057)***
(4.0659)***

Panel B: VP Portfolio Future Market-Adjusted Return

Portfolio mnVP 3moRET 6moRET YrlRET Yr2RET Yr3RET ..'

Hedge -0.0008 -0.0013 0.0438 0.0932 0.0690

(t-stat) (-0.8008) (-0.5910) (3.2444)***(6.3856)***
(4.6000)***
g MnVP refers to the mean VP in the decile. GmoRet, lYrRet, 2YrRet, 3YrRet
refer to returns six-month, one-year, two-year, and three-year subsequent
to portfolio formation, respectively. The portfolios are formed four-months
after each fiscal year end in order to ensure that financial statement
information has reached the investors.
h ***, ** and * signify one-tailed statistical significance at the 1%. 5% and
10%levels, respectively.
56 Seoul Journal oj'Business

Table 5. Time Series Regressions of Monthly Excess Returnsi,J

-
1 -Factor Model: (Rj,< = aj + Pj (RM, - RFJ + ~ j , , (7)
3-Factor Model: ( R ~-J RFJ = oj+ Pj (RM, - R n + yj SMBt + cpj HMLt + E ~ , ,
(8)

Panel A: ROE Portfolio Regressions

Portfolio Model o
t
i 6 3;- (pj Adj .R2
High ROE 1-factor -0.002 1.20 0.860
(t-stat) (- 1.02) (32.09)***
3-factor 0.001 1.08 0.61 -0.22 0.956
(t-stat) (1.55) (44.98)*** (16.29)*** (-5.39)***
Low ROE 1-factor -0.002 1.27 0.507
(t-stat) (-0.559) (13.60)***
3-factor -0.002 1.26 1.65 0.45 0.768
(t-stat) (-0.669) (17.26)*** (14.25)*** (3.54)***

Panel B: VP Portfolio Regressions

Portfolio Model aj Pi 3;- (pj Adj.R2

High VP 1-factor 0.001 1.09 0.670
(t-stat) (0.34) (18.84)***
3-factor -0.001 1.15 1.01 0.57 0.864
(t-stat) (-.312) (27.14)*** (15.19)*** (7.92)***
Low VP 1-factor 0.003 1.08 0.507
(t-stat) (1.13) (16.12)***
3-factor 0.003 1.08 1.15 0.40 0.768
(t-stat) (1.58) (20.60)*** (14.16)*** (4.55)***
i The results are based on 180 monthly stock returns of the market, size,
and book-to-market factors from May 1982 to April 1997. The results are
based on 12-month holding investment strategy. Rj,t is equally-weighted
monthly stock returns of the portfolio in calendar-month t, RF,is risk free
rate in month t (Ibbottson Associate's one-month T-bill rate), RM, is
market return for month j (value-weighted CRSP market return), SMBt is
the difference between the return of portfolios of "small" stocks and "big"
stocks i n month t , and HML, i s the difference between the return of
portfolios of "high" and "low" book-to market stocks in month t. High (low)
ROE indicates top (bottom) decile ROE portfolio.
j ***, ** and * signify two-tailed statistical significance a t the 1%, 5% and
10% levels, respectively.
 A Comparative Analysis of ROE and Vahe-to-Price based Trading Rules 57

Table 6. Time Series Regressions of Monthly Excess Returns on

ROE Portfolios: Sub-sample Period A n a l y ~ i s ~ . ' . ~

1 -Factor Model: (R,,,- RFJ = 9 + pj ( R M ~- RFJ + j, , (7)

3-Factor Model: (R,,,- RFJ = '3 + (RM, - RFJ + 7;. SMB, + % HML, + Ej,t
(8)

Panel A: Pre- 1990 Period

Portfolio Model '3

High ROE 1-factor -0.003
(t-stat) (- 1.23)
3-factor 0.000
(t-stat) (0.17)
Low ROE 1-factor -0.007
(t-stat) (- 1.31)
3-factor -0.00 1
(t-stat) (-0.34)

Panel B: Post- 1990 Period

Portfolio Model '3 Pi 7;. Adj.R2

High VP 1-factor -0.003 1.27 0.639
(t-stat) (-0.79) (10.52)***
3-factor -0.000 1.12 0.81 -0.20 0.930
(t-stat) (-0.04) (19.04)*** (13.80)*** (-3.31)***
Low VP 1-factor 0.000 1.45 0.300
(t-stat) (0.01) (0.28)
3-factor -0.004 1.50 1.89 0.48 0.740
(t-stat) (-0.88) (8.09)*** (10.25)*** (2.47)**

k The results are based on 180 monthly stock returns of the market, size,
and book-to-market factors from May 1982 to April 1997. The results are
based on 12-month holding investment strategy. Rj,, is equally-weighted
monthly stock returns of the portfolio in calendar-month t, RFt is risk free
rate in month t (Ibbottson Associate's one-month T-bill rate), RM, is
market return for month j (value-weighted CRSP market return), S M B , is
the difference between the return of portfolios of "small" stocks and "big"
stocks in month t, and HML, is the difference between the return of
portfolios of "high" and "low" book-to market stocks in month t. High (low)
ROE indicates top (bottom)decile ROE portfolio.
1 ***, ** and * signify two-tailed statistical significance at the 1%, 5% and
10% levels, respectively.
m Pre- 1990 (Post-1990) period covers 1985-1991 (1992-1997).
58 Seoul Journal of Busirzess

estimating equation (7) and (8).The results indicate that the

profitability of ROE trading rule is by and large subsumed by the
conventional risk factors. The intercepts in both specifications
(Jensen alpha and the three-factor model alpha) are insignificant
in both top and bottom decile ROE portfolio.
More importantly, Table 5 shows interesting beta estimates for
the ROE portfolios. Looking down the beta column of ROE
portfolios in Panel A, we can see that low ROE portfolio h a s
higher betas than high ROE portfolio in both models. Given the
unexpected result for betas, we turn to additional analysis of
sub-sample period. Table 6 also shows that even in the sub-
sample period the high ROE portfolio betas are consistently
smaller than those of the low ROE portfolio. This result suggests
that the differences of returns among ROE based portfolios are
not due to differences in the conventional risk factors.

VI. Conclusion

Motivated by the residual income valuation model, this study

explores the return predictability of ROE and VP based trading
r u l e s . Although t h e ROE b a s e d hedge portfolio r e t u r n i s
significant up to one year after portfolio formation, our evidence
indicates t h a t the profitability of ROE trading rule is largely
explained by the conventional risk factors. Both Jensen's alpha
and the three-factor model alpha for 12-month holding period
are insignificant, suggesting that the profitability of ROE trading
rule is subsumed by the conventional risk factors. Beyond one
year after portfolio formation, the profitability of ROE trading
rule dissipates quickly, and the VP portfolio dominates ROE
portfolio u p to three years. However, the profitability of VP based
trading rule is also subsumed by the conventional risk factors.
One finding, which raises interesting questions about market
efficiency, is that unlike betas in VP portfolio, betas in high ROE
portfolio are smaller than those in low ROE portfolio. This result
is difficult to reconcile with the risk hypothesis t h a t all the
difference in returns are due to the difference in risks. There
may be another factor that affect the differences of returns other
t h a n conventional risk factors. We leave this issue to future
research.
 A Conzparative Atznlysis of ROE and Value-to-Price based Trading Rules 59

Reference:

Abarbanell J . , a n d V. Bernard, 2000. "Is t h e U. S . Stock Market

Myopic?" Working Paper, University of North Carolina a t Chapel
Hill.
Barber, B. M., and J . D. Lyon, 1996. "Detecting Long-Horizon Abnormal
Stock Returns: The Empirical Power a n d Specification of Test
Statistics." Journal of Financial Economics 43: 341-372.
Rasu S . , 1977. "The Investment Performance of Common Stocks in
Relation to Their Price-Earnings Ratios: A Test of the Efficient
Market Hypothesis." Journal of Finance (June): 663-682.
Beaver, W., and Ryan, S., 2000. "Biases and Lags in Book Value and
Their Effects on the Ability of the Book-to-Market Ratio to Predict
Book Return on Equity." Journal of Accounting Research (Spring):
127-148.
Bernard, L. V., 1994. "Accounting-Based Valuation Methods: Evidence
on the Market-to-Book Anomaly and the Implications for Financial
Statement Analysis." Working Paper, University of Michigan.
Bernard, L. V. and J . Thomas. 1989. "Post-Earnings Announcement
Drift: Delayed Price Reaction or Risk Premium?" J o u r n a l of
Accounting Research (Vol.27. Supplement): 1-48.
Campbell, J . P., and R. J . Shiller, 1988. "The Dividend Price Ratio and
Expectations of Future Dividends and Discount Factors." Review of
Financial Studies (1): 195-228.
Delong, J. B., A. Shleifer, L. H. Summers, and R. J . Waldmann, 1990.
"Noise Trader Risk i n Financial Markets." J o u r n a l of Political
Economy 98: 703-738.
Edwards, E., a n d Bell, P., 1961. "The Theory a n d Measurement of
Business Income." Berkeley, CA: University of California Press.
F a m a , E. F., 1998. "Market Efficiency, Long-Term R e t u r n s , a n d
Behavioral Finance." Journal of Financial Economics (49) 3: 283-
306.
Fama, E. F., and K. R. French, 1988. "Dividend Yields and Expected
Stock Returns." Journal of Financial Economics (22):3-25.
Fama, E. F., and K. R. French, 1993. "Common Risk Factors in the
Returns on Stocks and Bonds," Journal of Financial Economics 33,
3-56.
Francis, J . , Olsson, P., a n d Oswald, D., 2000. "Using Mechanical
Earnings a n d Residual Income Forecasts in Equity Valuation."
Working Paper, Duke University.
Frankel, R, and C. M. C. Lee, 1998. "Accounting Valuation, Market
60 Seoul Journal of Business

Expectation, a n d Cross-Sectional Stock Returns." J o u r n a l of

Accounting and Economics 25: 283-319.
Holthausen, R. W., and D. F. Larker, 1992. "The Prediction of Stock
R e t u r n s Using Financial Statement Information." J o u r n a l of
Accounting and Economics 15: 373-4 11.
Ibbotson, R., 1975. "Price Performance of Common Stock New Issues."
Journal of Financial Economics 2: 235-272.
Kothari, S. P., and J . Shanken, 1997. "Book-to-Market, Dividend Yield,
and Expected Market Returns: A Time-Series Analysis." Journal of
Financial Economics (44): 169-203.
Lee, C. M. C., J. Myers, a n d B. Swaminathan, 1999. "What is the
Intrinsic Value of the Dow?" The Journal of Finance 54: 1693-
1741.
Liu, J . , D. Nissim, a n d J . Thomas, 2000. "Equity Valuation Using
Multiples." Working Paper, Columbia Business School.
Ohlson J . , 1995. "Earnings, Book Values, and Dividend in Security
Valuation." Contemporary Accounting Research 11: 6 6 1-687.
P e n m a n , S . H . , 1 9 9 2 . " R e t u r n t o F u n d a m e n t a l s . " J o u r n a l of
Accounting, Auditing, and Finance, Fall: 465-483.
Penman, S. H., 1996. "The Articulation of Price-Earnings Ratios and
Market-to-Book Ratios and the Evaluation of Growth." Journal of
Accounting Research (Autumn): 235-259.
Rosenberg, B., K. Reid, and R. Lanstein, 1985. "Persuasive Evidence of
Market Inefficiency." Journal of Portfolio Management: 9- 17.
Schiller, R. J., 1984. "Stock Prices and Social Dynamics.'' The Brooking
Papers on Economic Activity 2: 457-570.
Stober, T. L., 1992. "Summary Financial Statement Measures a n d
Analysts' Forecasts of Earnings." J o u r n a l of Accounting a n d
Economics 15: 347-372.
Stoll, H., and R. Whaley., 1983. "Transaction Costs and the Small Firm
Effect." Journal of Financial Economics 12: 57-80.
Wang, J . , 1993."A Model of I n t e r t e m p o r a l A s s e t Prices U n d e r
Asymmetric Information." Review of Economic Studies 60: 249-
282.