A comparison between the Abnormal Earnings Growth Model and the Residual Income

Valuation Model: is the AEGM better in predicting future returns than the RIVM?




Author: Danny Slewe
Student number: 0516104







Date and version: June 03, 2008. Final version.
Course: Bachelor Thesis
Supervisor: David Veenman













Bachelor of Science: Accountancy & Control
University of Amsterdam


Contents

Abstract..................................................................................................................1
1 Introduction..........................................................................................................1-3
2 The concept of the Residual Income Valuation Model (RIVM).................................. 3
2.1 From the Dividend Discount Model to the Residual Income Valuation
Model........................................................................................................... 3-6
2.2 The Clean Surplus Relationship (CSR) ...................................................... 6-8
2.3 What exactly is Residual Income? ............................................................. 8-9
3 Research conducted thus far to the Residual Income Valuation Model........................ 9
3.1 Are higher returns possible using the Residual Income
Valuation Model?.......................................................................................... 9-12
3.2 The Residual Income Valuation Model in relation to risk............................. 13-14
3.3 Other frequently used models compared with the Residual Income
Valuation Model............................................................................................ 14-17
3.4 Will greater valuation errors occur if companies do not apply
Clean Surplus accounting? ............................................................................. 17-18
4 The Abnormal Earnings Growth Model (AEGM)..................................................... 18
4.1 From the Dividend Discount Model to the Abnormal Earnings
Growth Model............................................................................................... 19-21
4.2 What is Abnormal Growth in Earnings exactly? .......................................... 21-22
4.3 The benefits and drawbacks of the Abnormal Earnings
Growth Model............................................................................................... 22-24
5 Empirical pilot study.............................................................................................. 24
5.1 Method.................................................................................................... 24-26
5.2 Sample.................................................................................................... 26-30
5.3 Results.................................................................................................... 30-35
6 Summary.............................................................................................................. 35-36
References............................................................................................................... 37-38



1

Abstract

In this thesis two accounting-based valuation models are compared. The first model described
in this thesis is the Residual Income Valuation Model (RIVM). Much research has been
conducted to this model. Those studies show that the RIVM is able to predict future returns,
while the stocks that are predicted to yield higher returns are not riskier than other stocks.
Furthermore, the studies show that the RIVM is useful when predicting future returns of
different types of companies in different countries with different accounting methods.
However, not much research has been conducted to a newer model, the Abnormal Earnings
Growth Model (AEGM). Possibly, this model is better in predicting future returns than the
RIVM because it does not rely on the clean surplus relationship (CSR), while the RIVM does.
This is investigated in this thesis by valuing 20 stocks that were listed on the AEX index on
December 31, 2003 and comparing returns for one, two and three years after the valuation
date. Different scenarios for both models are used in order to determine whether the AEGM is
better than the RIVM in predicting future returns. I find that for most scenarios the AEGM is
not better in predicting future returns than the RIVM.


1. Introduction

Different models are available when valuing a stock. Although most fundamental valuation
models are based on the dividend discount model (DDM), they produce different results in
practice (e.g. Francis et al., 2000). In this thesis the most widely used model (Ohlson, 2005,
p.323), the Residual Income Valuation Model (RIVM) discussed by Ohlson (1995), will be
compared with a more recent model, the Abnormal Earnings Growth Model (AEGM)
developed by Ohlson and Juettner-Nauroth (2005). The question whether the AEGM is better
in predicting future returns than the RIVM, will be answered in this thesis.
Both models will be derived from the DDM in this thesis, thus theoretically they
should produce the same results. Furthermore, the theoretical benefits and drawbacks of both
models will be outlined. In addition, studies conducted to both models (mainly RIVM because
not many studies have been devoted to the AEGM) will be reviewed. Finally, because not
much research has been conducted to the AEGM, I publish the results of my own empirical
pilot study using 20 stocks that were listed on the AEX index on December 31, 2003.
2

Because most investors are interested in future returns, mainly making a higher return than the
stock index, I will compare different scenarios of both models with the realized return of the
AEX index in order to determine which model is most useful when predicting future returns.
The valuation date will be December 31, 2003 and returns of up to three years after this date
will be compared. Consistent with Frankel and Lee (1998) the 20 stocks will be allocated to
five different portfolios (each portfolio consists of four stocks) based on their V/P ratio, where
V stands for the value calculated by using the RIVM or AEGM (depending on the scenario)
and P stands for the current stock price. I expect high V/P ratio stocks to yield higher returns
than low V/P ratio stocks because a high V/P ratio means that the estimated value is higher
than the current stock price, so the stock is probably undervalued. Consequently, the highest
V/P ratio portfolio needs to yield higher returns than the lowest V/P ratio portfolio in order to
have predictive power.
The literature review shows that the RIVM is able to predict future returns better than
other models based on the DDM (Francis et al., 2000, p.69), the AEGM is not investigated in
that paper, though. Furthermore, the literature review shows that the RIVM is better in
predicting future returns than the book-to-price multiple (Frankel and Lee, 1998, p.315), and
heuristic models like the price-to-earnings multiple (Bradshaw, 2004, p.42). Moreover, when
adjusting for risk, the RIVM still has predictive power (Ali et al., 2003, p.394).
However, there are some problems with the RIVM. The main problem is that the
model relies on the clean surplus relationship (CSR). This relationship means that all changes
in equity, except dividends (dividends include issues and repurchases of shares), need to flow
through the income statement first (Ohlson, 1995, p.667). The accounting of many companies
is not consistent with the CSR (Isidro et al., 2006, p.316). Consequently, whether the RIVM
can be used for those companies remains questionable, although Isidro et al. (2006) show that
the predictive power of the RIVM does not, or does only slightly, depend on whether
companies violate the CSR. Other potential problems can be found in section 4.3.
The AEGM does not need the CSR to hold. Since many companies violate the CSR
(Isidro et al., 2006, p.316), it is possible that the AEGM produce better results than the RIVM.
Moreover, not much research has been conducted to this model. Consequently, it is very
interesting to investigate which model is better able to predict future returns. Because
different versions of the AEGM have been derived from the basic model, those formulas will
also be described below and investigated in the pilot study.
3

While reading this thesis, it is important to keep in mind that residual income and abnormal
earnings are synonyms. In this thesis, those words will be used interchangeably to illustrate
some major points.
The remaining part of this thesis is structured as follows, in the next section the RIVM
will be derived from the DDM, the CSR will be outlined and the question what residual
income exactly is will be answered. In section 3, different studies conducted to the RIVM will
be discussed. In the fourth section, the AEGM will be derived from the DDM, different
versions of the AEGM will be presented, the meaning of abnormal growth in earnings and
growth in abnormal growth in earnings will be outlined and the main benefits and drawbacks
of the AEGM will be discussed. In section 5, the empirical pilot study will be published, first
the method will be outlined then the sample will be discussed and finally the results will be
reviewed. Finally, in section 6 the main findings and conclusions are summarized.


2. The concept of the Residual Income Valuation Model (RIVM)

In this section the RIVM will be explained. First, the model will be derived from the dividend
discount model (DDM); in order to do this, the clean surplus relationship will be introduced.
In the second section, the clean surplus relationship will be discussed in more detail. Finally,
the concept of residual income will be examined further.


2.1 From the Dividend Discount Model to the Residual Income Valuation Model

It is widely accepted that the intrinsic value of a share of a company is the discounted
expected future dividend per share (e.g. Lee et al., 1999, p.1693). That is, if all future
dividends would be discounted with the appropriate discount factor and then added together,
the theoretical value of a stock can be derived. The following formula is known as the
dividend discount model or DDM (see next page).




4



where
= value per share at time t
= dividends per share in year t
= cost of equity capital

If one assumes that all changes in equity flow through the income statement, except
dividends, dividends per share in year t can be derived from



where:
= book value of equity per share in year t
= earnings per share in year t
The explanation of other inputs can be found above.

The above described formula is often referred to as the clean surplus relationship (CSR)
(Ohlson, 1995, p.667). This formula can be rewritten as:



To derive the RIVM from this model, we substitute the CSR into the DDM, for simplicity we
just do a two-period valuation for year t, thus:



Substitution of equation (3) into equation (4) yields:


(1)
(2)
(3)
(4)
(5)
5



Then, rewrite equation (5) using the following mathematical equivalence:



Substitute equation (6) into equation (5):





Now the RIVM (assuming a company will exist forever) can be derived (Ohlson, 1995,
p.667):


where
= value per share at time t
= book value of equity per share in year t
= Residual Income (or Abnormal Earnings) =

= cost of equity capital

This model implicitly assumes that paying dividends reduces market value, but let current
earnings unaffected and that future earnings are reduced by paying dividends now. Those
assumptions are consistent with the dividend irrelevancy in the Modigliani and Miller (MM)
world (Ohlson, 1995, p.662). The RIVM has some advantages above the DDM, for instance
not all firms pay dividends so the DDM would be difficult to use for such a company,
earnings forecasts are widely available while dividend forecasts are not and the DDM is based
(7)
(8)
(6)
(9)
6

on the distribution of value, while the RIVM is based on the creation of value (Bradshaw,
2004, pp.6-7). Specifically, if an investor wants to use the DDM for a company with a zero
pay-out policy, he needs to forecast how much the shareholders will get if the company were
to be sold in year . That is, he needs to forecast the terminal value (see below). Because the
terminal value is the most uncertain component of the valuation, using the DDM for such a
company would probably yield noisy results. Empirical studies approve that the RIVM yields
better results than the DDM and that the terminal value represents a greater portion of the
valuation when using the DDM (Francis et al, 2000, pp.45-57). In section 3.3 the RIVM will
be more thoroughly compared with the DDM and other frequently used valuation models.
Because it is very difficult, if not impossible, to project future earnings in perpetuity,
the model is used differently in practice. When an investor wants to do a valuation, he
forecasts the earnings for a couple of years (the forecast horizon) and then he calculates a
terminal value (Penman, 1998, p.303).
For this purpose the RIVM can be rewritten as:



where

= assumed growth in Residual Income after year t+n.
= Terminal Value
The explanation of the other inputs can be found above.


2.2 The Clean Surplus Relationship (CSR)

As was stated in the first paragraph of this chapter, the clean surplus relationship is essential
for the RIVM. To investigate this relationship further, we need to make a distinction between
clean surplus on a total equity basis and on a per share basis, because if the CSR holds on a
total equity basis, it does not necessary mean it also holds on a per share basis (Ohlson, 2005,
p.324). The following formula applies on a total equity basis (see next page):


(10)
7






where
= Book value of equity in year t
= Net Income in year t
= Dividends paid in year t

According to the CSR on a total equity basis all changes in equity, except dividends
(dividends include issues and repurchases of shares), need to flow through the income
statement first (Ohlson, 1995, p.667). Unfortunately, many companies across different
countries violate the CSR (Isidro et al., 2006, p.316). Consequently, those companies have so-
called dirty surplus accounting items. Hence, some profits or losses flow directly into equity
of such a company. Examples of profits or losses that flow directly into equity include
goodwill write-offs, revaluations of assets, prior-year adjustments and currency revaluations.
In some countries, some of these dirty surplus items may no longer flow directly into equity.
For example, goodwill write-offs may no longer flow directly into equity in the UK and the
US. Unfortunately, there are enough possibilities left in the different countries to apply dirty
surplus accounting (Isidro et al., 2006, p. 310).
Because we need the clean surplus relationship to hold on a per share basis, another
problem arises. Recall equation (2), the CSR formula on a per share basis. If the number of
shares changes during a year, because of issues or repurchases of shares, the bvps changes
even if eps and dps are zero. Hence, even if a company applies the clean surplus relationship
on a total equity basis, it does not mean it also holds on a per share basis (Ohlson, 2005, pp.
326-327).
A solution for this problem would be to simply use eps as a plug. Specifically, the
CSR formula on a per share basis can be written as eps
t
= bvps
t
dps
t
(where bvps
t
=
bvps
t
bvps
t-1
). However, it is not generally accepted to calculate earnings per share this way.
Moreover, recall that RIVM stands for Residual Income Valuation Model. When the CSR is
used this way, the RIVM becomes a book value approach rather than an income valuation
approach to value equity (Ohlson, 2005, p.331).
(11)
8

Summarizing, the main problem with this model is that it tries to connect two
accounting measures, namely earnings and book values. To connect these two measures, the
clean surplus relationship is needed and because of this the above described problems arise
(Ohlson, 2005, p.330). More about the CSR and possible valuation errors that occur if the
CSR does not hold, can be found in section 3.4.


2.3 What exactly is Residual Income?

The above sections described how the RIVM can be derived from the DDM and why the CSR
is needed in order to do so. In this section, I will outline what residual income (or abnormal
earnings) exactly means.
To illustrate what residual income essentially means, recall that the formula for
residual income was RI
t+1
= eps
t+1
r
e
* bvps
t.
This can be rewritten as RI
t+1
= (ROE
t+1
r
e
)*
bvps
t
. Clearly, with this formula one can see that residual income is the income (loss) above
(below) the cost of equity capital of a company. Thus, if a company earns more than its cost
of equity capital, it generates abnormal earnings. If we now recall that the full RIVM formula
was: , it is easy to see that a company that does not generate
earnings above (below) its cost of equity capital, will be worth their current book value.
However, firms that will generate positive (negative) abnormal earnings will be worth more
(less) than their current book value. Another important point is made here, since it is
insufficient to rely just on bvps
t
if an investor wants to value a stock (Lee et al., 1999,
pp.1699-1700).
A company should not be able to generate positive nor negative abnormal earnings in
perpetuity. If a company, for example, can earn earnings above the cost of equity capital,
other companies will enter this industry and thus drive profits down. Additionally, if a
company earns less than its cost of equity capital, the company (or one of its competitors) will
leave the industry and thus drive profits up. Hence, in the long run abnormal earnings will be
driven to zero because of competitive forces. Other reasons for abnormal earnings to revert to
zero are an upper limit on constant growth and less than full earnings payout. Empirical
evidence approves this theory (Bradshaw, 2004, pp.8-20). Furthermore, empirical evidence
approves that firms with low (high) B/P values, have higher (lower) future ROEs (Frankel and
Lee, 1998, p. 287). However, as Giner and Iiguez (2006, p.178) point out, because of
9

accounting conservatism the book value of equity is generally lower than the market price.
Consequently, the growth in abnormal earnings should be positive if conservatism were to
occur in the future as well because of the undervalued book value of equity. Thus both
possibilities (a positive and an negative growth in abnormal earnings) have some support in
literature. As we will see in section 3, most papers use a neutral growth ratio of 0%, although
studies with both a negative and a positive growth rate will also be discussed. The three
possible growth rates will be investigated in the empirical pilot study, that can be found in
section 5, in order to determine which growth rate has most predictive power.


3. Research conducted thus far to the Residual Income Valuation Model

The theory behind the RIVM has now been intensively explained. But does the model work in
practice? And what about the CSR? I have concluded that this relationship is, to put it mildly,
not always followed; is the RIVM still useful for those companies? To answer those
questions, this chapter is divided into four sections. In the first section literature that
investigated whether the RIVM can be used to generate higher returns, will be reviewed. The
second section will be devoted to whether the higher returns are due to higher risk for those
stocks, or to mispricing of the market. In the third section other frequently used valuation
models will be compared with the RIVM, here will be determined which model works better
predicting future returns. Finally, the fourth section is completely dedicated to the clean
surplus relationship.


3.1 Are higher returns possible using the Residual Income Valuation Model?

Frankel and Lee (1998) performed a study aimed to determine whether the RIVM can explain
variations in stock prices and can predict future returns. In order to do this, stock prices of
18,162 firm years between 19791991 were calculated using the RIVM. To estimate future
earnings, analysts forecasts for three future years were used. After that period the terminal
value was calculated, using an expected growth in residual income of 0%. Cost of equity
capital was calculated using a constant discount rate, the capital asset pricing model (or
CAPM, for details see section 5.2) and the three-factor model described in Fama and French
(1997). In this model market beta, B/P (book value of equity divided by current stock price)
10

ratio and firm size are considered to be risk factors (Fama and French, 1997, p.156). Because
the three discount rates yielded approximately the same results, only the results for the three-
factor model are described (Frankel and Lee, 1998, pp.284-294).
The expected stock price using RIVM (called V
f
in the paper), was divided by the current
stock price (P). This V
f
/P ratio was then used to form five portfolios, the highest V
f
/P
portfolio was expected to do better than the lowest V
f
/P ratio (because a high V
f
/P ratio means
that the estimated value is higher than the current stock price, so the stock is probably
undervalued). Furthermore, portfolios based on a B/P ratio were formed to compare which
ratio yields a higher return. The realized returns were then calculated over the next 12, 24 and
36 month periods (Frankel and Lee, 1998, p.295).
The results show that both ratios have some predictive power, the highest B/P quintile
portfolio earns 18.6%, while the lowest quintile earns 13.7% (thus 4.9% less) over the next 12
months which is statistically significant at 1%. The highest V
f
/P quintile portfolio earns
16,9% over the next 12 months (note that this is less than the highest B/P quintile portfolio),
and firms in the lowest quintile earn 13.8% (thus 3.1% less), this difference is also statistically
significant at 1% (Frankel and Lee, 1998, pp.297-298).
For the 24 and 36 months period, the difference between the highest and the lowest
quintile is 8.2% and 15.1% respectively for the B/P ratio. However, the difference between
the returns for the highest and lowest quintile for the V
f
/P ratio is now higher. For 24 months,
the spread between the highest en the lowest quintile is 15.2% and for 36 months the
difference is 30.6%. Summarizing, the V
f
/P ratio is better in predicting long-term returns.
However, for the short term (12 months) period the B/P ratio can yield higher returns than
trading based on the V
f
/P ratio (Frankel and Lee, 1998, pp.297-299).
Because this paper uses analysts forecasts as one of the most important inputs of the
RIVM (earnings forecast), and analysts are overly optimistic (e.g. Bradshaw, 2004, p.11),
Frankel and Lee (1998) have proposed to account for this optimism. The researchers tried to
adjust the analysts earnings forecasts, using ex ante firm characteristics such as sales growth
and B/P ratio. A relation has been found between forecast errors and four firm characteristics
(details how to implement this technique can be found in Frankel and Lee, 1998, p.310). With
this information, a new earnings forecast has been made and implemented in the RIVM. This
PErr-based strategy yields higher returns for the 36 month holding period only if an
investor buys stocks that are both in the highest PErr quintile and the highest V
f
/P quintile
(and sells stocks that are both in the lowest PErr quintile and the lowest V
f
/P quintile), the
return is then 45,5%. However, if an investor trades based on the PErr strategy only, the
11

return will actually be lower than trading based on the V
f
/P strategy (Frankel and Lee, 1998,
pp.303-313).
Bradshaw (2004) also tried to determine whether a strategy based on the RIVM can yield
higher returns. In this study, also analysts earnings forecasts were used to estimate future
earnings. A forecast horizon of five years was taken and then a terminal value was calculated,
using an expected growth in abnormal earnings of -32% (based on the in section 2.3 described
theory that abnormal earnings will revert to zero, using this percentage after 10 years
abnormal earnings will approximately be zero) and cost of equity capital was calculated using
the three-factor model of Fama and French (1997) with an annualized risk free rate of the 30-
day treasury bill for the most recent month (Bradshaw, 2004, pp.7-10).
The results in het period 1994-1997 were generally in line with the results of Frankel
and Lee (1998). For the highest V/P quintile, returns were 15% higher in the one-year holding
period than for the lowest V/P quintile (compared with 3.1% in Frankel and Lee). For the
two-year holding period, the return was 24.5% higher (compared with 15.2% in the Frankel
and Lee paper). Finally, for the three-year holding period, the return was 25% higher
(compared with 30.6% in Frankel and Lee). Hence, both papers agree that the V/P strategy
works best for a long holding period. Percentage differences may have occurred because of a
different method of calculating terminal value and/or because of a longer forecast horizon in
the Bradshaw (2004) paper (five year, compared with a three year horizon in the Frankel and
Lee (1998) paper) (Bradshaw, 2004, p.42).
Lee et al. (1999) performed a slightly different study. This study aimed to show
whether different estimates of intrinsic value were able to track price valuations and/or able to
predict market returns for the 30 stocks listed on the Dow Jones index between 1963-1996.
V/P ratios with a lower standard deviation and with a faster reversion rate to the mean V/P
have a better tracking ability (less tracking errors). V/P ratios that have better intrinsic value
estimates, have greater predictive power (Lee et al., 1999, pp.1693-1698). Lee and
Swaminathan (1999) have published a less complicated version of this paper, some parts of
the text below will refer to that paper.
The study was conducted using the RIVM with different forecast horizons (3, 12 and
18 years) plus a terminal value estimate, with an expected growth in abnormal earnings of
0%. Cost of equity capital was computed using a constant discount factor, the CAPM and the
Fama and French (1997) three-factor model plus a short-term and a long-term risk free rate
(using monthly annualized one-month T-Bill rates and monthly annualized long-term
Treasury bond rates, respectively). Earnings were forecasted with both a historical earnings
12

forecast (using a time-series model) and analysts forecasts (after 1st January 1979, since
analysts forecasts were not available before). This way, 22 V/P scenarios were made plus 3
traditional market multiples (dividend yield, price-to-earnings ratio and price-to-book ratio).
Those 25 methods were then compared and reviewed based on the above described targets:
tracking ability and predictive power for the post 1979 period (Lee et al., 1999, pp.1701-
1706).
As one would expect, predictive power and tracking error are highly negatively
correlated. Thus, a scenario with high predictive ability, has, generally speaking, a low
tracking error (Lee and Swaminathan, 1999, p.17). During the period January 1979 to June
1996, the three traditional market multiples had a very high tracking error and a low
predictive ability. The two scenarios with a constant discount rate, did (almost) as bad as the
traditional market multiples (thus a high tracking error, combined with no significant
predictive power). Hence, traditional market multiples and the RIVM model with a constant
discount rate, were not useful in either predicting future stock returns or tracking price
variations. The scenarios with the long term risk free rate improved both the ability to predict
future returns and the tracking error. Moreover, using analysts forecasts further improved
those abilities. However, the best scenarios were those that used the short term risk free rate.
Using a longer forecast horizon and analyst forecast further improved the ability to predict
stock returns with a lower tracking error in most cases. When Lee et al. (1999) used the three-
factor model from Fama and French (1997) to estimate cost of equity capital, predictive
ability was slightly better, though this was offset by a greater tracking error (Lee et al., 1999,
pp.1726-1737).
Summarizing, three different papers in a (partly) different time period using different
forecast horizons and in some cases different growth in abnormal earnings after the terminal
year, conclude that higher returns are possible using the RIVM. Traditional market multiples
are not good in predicting those returns, or are only for a short period. Furthermore, predicting
returns more accurately is negatively correlated with tracking error (faster reversion to mean
V/P and lower standard deviation). The residual income valuation model works best using a
short term risk free rate, with analysts forecasts of future earnings and, in general, an as long
as possible forecast horizon. The finding that the V/P strategy works best for a long holding
period is puzzling, especially if those returns are due to mispricing of the market (Frankel and
Lee, 1998, p.315). Maybe information is not incorporated into market prices the way prior
evidence suggests. This question is open to future research.

13



3.2 The Residual Income Valuation Model in relation to risk

Thus far, we have concluded that, if the RIVM is used correctly (with the right inputs), firms
with a high V/P ratio will generate higher returns than firms with a low V/P ratio. Because
investors demand a higher return for firms with a high risk, it is possible that high V/P firms
are just riskier than low V/P firms. In that case, the higher returns have nothing to do with
mispricing of the market, but are simply due to higher risk (Frankel and Lee, 1998, p.316).
Recall that Frankel and Lee (1998) used a three-factor discount model with firm size,
market beta and B/P ratio. Because of this discount factor, the study already controls these
sources of risk. Additionally, Frankel and Lee (1998) find that in general high V
f
/P firms have
low market betas. Hence, if high V
f
/P firms are riskier than firms with a low V
f
/P ratio, this
risk is not due to firm size, sensitivity to the market or B/P ratio. Furthermore, because high
V
f
/P ratio firms earn their higher returns particularly in the long run, it does not make very
much sense that risk factors provide higher returns. Note that this is in contrast with the B/P
ratio, which yields higher returns in both the long and the short run and probably is a source
of risk according to Fama and French (1997, p.156). Hence, the higher return of high B/P
ratio firms is probably due to higher risk for those firms. However, high V
f
/P firms may still
prove to be riskier than low V
f
/P firms when accounting for other risk factors (Frankel and
Lee, 1998, p.316).
Ali et al., (2003) have conducted more research whether higher returns for high V
f
/P
ratio firms are due to higher risk of those firms, or to market mispricing. Essentially, the paper
is presented as a sequel to the Frankel and Lee (1998) paper. The study is performed exactly
the same way, except that this paper uses data from 1976 to 1997 where Frankel and Lee
(1998) use data from 1979 to 1997. This results in slightly different percentages (e.g. for the
36 months holding period, the difference between the return on the highest V
f
/P quintile firms
portfolio and the lowest V
f
/P quintile firms portfolio is 31.1%, where it is 30.6% in the paper
of Frankel and Lee (1998)). Furthermore, Ali et al., (2003) show that market capitalization is
significantly greater for the high V
f
/P quintile than for the low V
f
/P portfolio (Ali et al., 2003,
p.380). Because it is generally accepted that firms with a high market capitalization are less
risky than firms with a low market capitalization (e.g. Fama and French, 1997, p.156),
because of differences in information environment for small and large firms (Ali et al., 2003,
14

p.385), this result contradicts with the theory that high V
f
/P firms yield higher returns because
of higher risk for those firms.
As a test for the market mispricing hypothesis, Ali et al. (2003) investigated whether the
abnormal returns were realized during earnings announcement periods. If abnormal returns
are concentrated during earnings announcements, this probably means that the higher returns
are due to mispricing of the market and that this mispricing corrects during large amounts of
information flowing into the market. Ali et al. (2003) indeed found that more than the average
of the higher returns for the high V
f
/P portfolio is realized during earnings announcement
periods. Hence, this result is consistent with the market mispricing hypothesis (Ali et al.,
2003, pp.380-384).
In addition, Ali et al. (2003) have tested the risk hypothesis (according to this
hypothesis high V
f
/P firms yield higher returns because of higher risks for those stocks) using
a twelve risk factor model (for details of this model, see Ali et al., 2003, p.385). Furthermore,
Ali et al. (2003) controlled for those risk factors and determined whether the high V
f
/P ratio
portfolio still yield significantly higher returns than the low V
f
/P ratio quintile after
controlling for those risk factors. They conclude indeed that, after controlling for the twelve
risk factors, the high V
f
/P portfolio still yields higher returns than the low V
f
/P portfolio. This
is consistent with the market mispricing hypothesis rather than the risk hypothesis (Ali et al.,
285-394).
In summary, the higher return for high V
f
/P firms is probably not (completely) related
to higher risk factors for those firms. Hence, this return probably occurs (partly) because of
mispricing of the market, which corrects during time periods with large amounts of
information flowing into the market.


3.3 Other frequently used models compared with the Residual Income Valuation Model

The studies in the text above show that higher returns are possible using the RIVM, without
taking more risk. Although, this does mean that the RIVM is useful if an investor wants to
earn a higher return than the market, it does not rule out the possibility that other models are
even better than the RIVM in predicting future returns. In this section three fundamental
valuation models will be compared, in addition a heuristic model (based on market multiples)
will be reviewed. Finally, I will briefly discuss which model (a heuristic or a fundamental
model) is used most frequently by analysts.
15

Recall from chapter one that the RIVM is derived from the dividend discount model
(DDM). Additionally, another model can be derived from the DDM, the discounted free cash
flow model (FCF). In this model, sales are first converted into cash flows and then discounted
with the appropriate discount rate. A complete review of the FCF falls outside the scope of
this thesis, but can be found in Francis et al. (2000, p.49). Since the RIVM and the FCF can
be derived from the DDM, the three models should yield exactly the same intrinsic value
estimates (and thus should predict the same returns). Nevertheless, because of the different
inputs in the models, results from the models will differ in practice (Francis et al., 2000,
pp.45-46).
To forecast earnings, Francis et al. (2000) used a database called Value Line, with
analysts forecasts for the coming years. The cost of equity capital was estimated using the
CAPM, the risk premium was set on 6% with an equity beta set equal to the industry where
the company belongs. The risk free rate was set equal to the intermediate-term Treasury bond
rate minus the historical premium on Treasury bonds over Treasury bills. The three-factor
model of Fama and French (1997) has also been used, but the results were similar to the
CAPM results and are not reported. Finally, the forecast horizon was set on five years, after
that a terminal value was calculated using growth of abnormal earnings, dividends and free
cash flow of 0% and 4%. Since the results were almost the same, only the results of the 4%
scenario are discussed (Francis et al., 2000, pp.51-54).
Francis et al. (2000) used 2,907 firm-year observations in the period 1989-1993 to
determine which model estimates the stock price most accurate and is best in explaining the
variations in stock prices. They conclude that the RIVM model is best in both accuracy and
ability to explain variations in stock prices. According to Francis et al. (2000), the most likely
explanation for this result is that the RIVM consists of both a stock component (book value of
equity) and a flow component (abnormal earnings), whereas the other models only have a
flow component (dividends and free cash flow). Consequently, the terminal value (probably
the most difficult part to estimate, thus the noisiest component) represents a greater share of
the estimated stock price for the FCF and DDM than for the RIVM: 82%, 65% and 21%
respectively (Francis et al., 2000, pp.45-57).
It is possible, though, that because of the book value component in the RIVM, for
some companies the other models work better. Specifically, because of accounting rules high
R&D firms may not always capitalize their R&D. Consequently, these costs are directly
incurred and book value of equity may be different from market estimates (because the
market expects that the R&D will yield profits in the future). Furthermore, managers of firms
16

with many accruals may have an incentive to manage earnings, for example because of bonus
schemes (Healy, 1985, pp.90-92). Consequently, book value may differ from market
estimates in this case as well. Francis et al. (2000) have investigated these possibilities, and
concluded that even for high R&D firms and firms with high accruals, the RIVM still is
significantly better in both accuracy and explanatory power than the tow other investigated
models (Francis et al., 2000, pp.59-63).
Finally, Francis et al. (2000, p.65) conclude that forecasts of abnormal earnings are
more precise and predictable than forecasts of free cash flows. They are less precise than
forecasts of dividends, because companies are reluctant to change their dividend policy
(Francis et al., 2000, p.64). However, because of the in section 2.1 described problems with
the DDM and because Francis et al. (2000) concluded that the RIVM works better than the
DDM (see above), the RIVM still seems to be the better model.
As discussed in section 3.2, Lee et al. (1999) conclude that traditional market
multiples are not good in predicting returns, furthermore, they have a high tracking error.
Bradshaw (2000) further investigated a widely used ratio, based on those multiples. This ratio
is called the PEG ratio and can be found by dividing the P/E ratio by the forecasted long
term growth ratio (Bradshaw, 2004, p.21).
Bradshaw (2004, pp.24-25) finds that the V
peg
/P ratio, does have predictive power for
future returns only for very high V
peg
/P values. Consequently, one would expect that analysts
do not rely on the V
peg
/P ratio or other traditional multiples if they want to give an advice for
a stock. Instead, they should use the RIVM (based on the above described studies). Bradshaw
(2004) finds, though, that analysts forecasts are significant positively related to the V
peg
/P
ratio, thus analysts seem to estimate the value of a stock, using the PEG ratio. Furthermore, he
finds that analysts forecasts are nonrelated or even negatively related to the RIVM model.
Hence, if according to the RIVM, an investor should buy the stock (because the estimated
value is below the current stock price), analysts do not give a buy advice, or even give a
sell advice (Bradshaw, 2004, pp.16-22).

In summary, the residual income valuation model is, of the models investigated, the
best model to predict future returns. Other fundamental valuation models, such as the DDM
and the FCF, do have that ability, though to a lesser extend. The reason for this superiority of
RIVM, can probably be found in the book value component of RIVM. Consequently, terminal
value (the noisiest component of the valuation models) is a smaller component of the
estimated stock price. In firms where book values may differ from market estimates, such as
high R&D firms and firms with a high portion of accruals relative to net income, the RIVM
17

still yields better results than the other two models investigated. Furthermore, earnings are
easier (thus more precise) to forecast than free cash flows (Francis et al., 2000, pp.68-69).
Finally, heuristic models based on traditional multiples, do not have, or only have low,
predictive power. It is puzzling, given this information, that analysts seem to base their stock
recommendations on those traditional multiples and not on the RIVM (Bradshaw, 2004, p.34).


3.4 Will greater valuation errors occur if companies do not apply Clean Surplus
accounting?

We now have concluded that using the RIVM can yield higher returns, without taking more
risk and that the RIVM is better than other models in predicting returns even for companies
where book value may not reflect the expected market value. Nevertheless, it is still possible
that, since the RIVM needs the clean surplus relationship (CSR), the RIVM has greater
valuation errors for firms that have dirty surplus items. Furthermore, different countries apply
different accounting regimes. Hence, since some countries give more opportunities to have
dirty surplus items, the RIVM may prove less useful in those countries (Isidro et al., 2006,
pp.302-303).
Isidro et al. (2006) have recognized these potential problems of the RIVM.
Consequently, they have conducted a study in four different countries, aimed to show whether
greater valuation errors occur for companies with great dirty surplus flows and whether
greater valuation errors occur in countries where dirty surplus accounting appears more
frequently. Isidro et al. (2006) expected that for companies and countries with greater dirty
surplus flows, more valuation errors would occur (Isidro et al, 2006, pp.303-306).
The study was conducted for France, Germany, the U.K. and the U.S. from 1994 to
2001 (Isidro et al., 2006, p.303). Obviously, the RIVM was used to estimate value per share
for the studied companies. This estimated value per share was then subtracted by the current
stock price of each investigated company. Since Isidro et al. (2006) implicitly assume the
efficient market hypothesis to hold, which implies that the current stock price equals the real
value of a company, the greater the difference between estimated value and price, the greater
the valuation error became. Earnings per share were estimated using analysts forecasts, with
a two year forecast horizon. After that a terminal value was calculated, using growth in
abnormal earnings of the expected return on equity for the country and industry to where the
company belongs, multiplied by the expected country-industry retention ratio. Cost of equity
18

capital was estimated using the CAPM, with a risk free rate equal to the Treasury Bond rate
for the country to which the company belongs plus a risk premium of 5% times the equity
beta of the industry to which the company belongs (Isidro et al., 2006, pp.307-309) .
Isidro et al. (2006, p.324) did not find evidence for cross-country differences between
Germany, U.K. and U.S. The mean dirty surplus flows are higher in the U.K. than in the U.S.,
though, more valuation errors occur in the U.S. than in the U.K. For France, they did find
weak evidence at the cross-country level that because of higher dirty surplus flows, more
valuation errors occur. Moreover, Isidro et al. (2006, pp.325-334) did not find strong evidence
that higher dirty surplus flows result in more valuation errors. Only for the U.S. they did find
some weak evidence.
Finally, they found that in the European countries higher dirty surplus flows, provide
less valuation errors in some cases, this is probably due to goodwill-related items (Isidro et
al., 2006, pp.329-333). The reason that dirty surplus flows and valuation errors do not have
the predicted negative relation in general, suggests either that predictable cash flows from
current flows will only impact dividends in the distant future or that realized flows are totally
unrelated to expected future flows. Because of either of those explanations, current dirty
surplus flows are unimportant in present value terms (Isidro et al., 2006, p.340).
In short, only for the U.S. there is some weak evidence that companies with higher
dirty surplus flows provide more valuation errors. For other countries, the relationship is
sometimes even contrary, partly because of goodwill-related items. There is also some weak
evidence at the cross-country level that greater valuation errors occur in countries where dirty
surplus accounting appears more frequently. Isidro et al. (2006, p.341) conclude that dirty
surplus flows are probably not an important source of error when using the RIVM. Hence, the
RIVM can be used in different countries with different accounting regimes to value
companies, without having more forecast errors than in countries where the accounting
regime results in less dirty surplus items.


4. The Abnormal Earnings Growth Model (AEGM)

The AEGM will be explained in this section. In the first section, I will show how the AEGM
can be derived from the DDM. The second section will focus on what abnormal growth in
earnings and growth in abnormal growth in earnings exactly means. Finally, the third section
will outline the benefits and the drawbacks of this model.
19


4.1 From the Dividend Discount Model to the Abnormal Earnings Growth Model

Recall the dividend discount model (DDM) (equation (1) from section 2.1). To show how the
AEGM can be derived from the DDM, again consider a two period valuation. The appropriate
formula becomes:


Assume that the following equation holds, so that dividends per share can be rewritten as:



This can then be substituted in the DDM:



Rewriting this formula yields the following:



The AEGM formula is then defined as (Ohlson and Juettner-Nauroth, 2005, p.352):



where:
= value per share at time t.
= abnormal growth in earnings =

= earnings per share in year t + 1.

= cost of equity capital.
(12)
(13)
(14)
(15)
(16)
20


Note that, in contrast to the RIVM, this model does not need the clean surplus relationship to
hold (Palapu et al., 2007, p.316). This is because this model relies only on earnings and
dividends and therefore does not need to connect a book and a flow variable. The RIVM does
need the clean surplus relationship to hold, because it does try to connect book value and
earnings (Ohlson, 2005, p.331).
It is very difficult, if not impossible, to estimate earnings in perpetuity. Therefore, the
model is used differently in practice. When an investor wants to do a valuation, he forecasts
the earnings for a couple of years (the forecast horizon) and then he calculates a terminal
value (Penman, 1998, p.303). The AEGM formula becomes:



where
= assumed growth in abnormal growth in earnings (see section 4.2)
= Terminal Value
The explanation of the other variables can be found above.

From this basic model, two other formulas have been derived. Because it is interesting which
of those formulas works best in practice, these two models will also be described below and
investigated in the pilot study. Ohlson and Juettner-Nauroth (2005, p.355) describe the second
formula:



where:
=
An explanation of the other variables can be found above.


(17)
(18)
21



The third formula is described and used by Gode and Mohanram (2003, p.402):



All variables used in this equation are explained above.


4.2 What is Abnormal Growth in Earnings exactly?

In order to show what abnormal growth in earnings exactly mean, recall from the previous
section that the abnormal growth in earnings formula was:

=

If we assume that a company has a full pay-out policy, that is eps
t
= dps
t
, then the abnormal
growth in earnings is simply eps
t
eps
t-1
(Gode and Mohanram, 2003, p.402). This growth is
abnormal because if a company pays out all it earns, it cannot invest in its business.
Consequently, the company is expected to earn the same every year.
To illustrate this further, imagine a company that does not pay any dividends. The
expected next year earnings are eps
t+1
= eps
t
+ r
e
*eps
t
for this company. Clearly, this company
is able to invest all it earns. With a rate of return of r
e
for this company, the company is
expected to earn next year the same as this year plus the return on the reinvested earnings. As
Easton (2004) describes, these are economic earnings. Since accounting earnings (these are
the reported earnings) differ from economic earnings, a company is able to generate positive
abnormal growth in earnings when accounting was conservative. That is, when reported
earnings were lower than economic earnings (Easton, 2004, pp.79-80). Returning to the
previous example of a company with a full pay-out policy, such a company is able to generate
higher earnings in year t+1 if the accounting was conservative in year t (Ohlson, 2005, 336).
In the long run, the accounting earnings should correct for the difference between
accounting and economic earnings. This is called the growth in abnormal growth in earnings
(g
agr
). Theoretically, one would thus expect that g
agr
is negative (positive) for companies with
(19)
(20)
22

positive (negative) abnormal growth in earnings. However, because accounting earnings are
usually conservative and earnings are thus understated every year, the growth in abnormal
growth in earnings is usually positive. As Easton (2004, p. 85) shows, during the period 1981-
1999 the average growth in abnormal growth in earnings was 2,9% for the 22 investigated
U.S. firms.


4.3 The benefits and drawbacks of the Abnormal Earnings Growth Model

The previous two sections have outlined the basics of the AEGM with the different versions
of the model used in practice. However, why should we do such an effort if the RIVM is
already able, as showed in section three, to forecast returns? Therefore, the main benefits and
drawbacks of the AEGM will be outlined in this section. Although, there has not been
conducted much research to this model, Penman (2005) has compared the RIVM and one of
the versions of the AEGM. This study will also be described below.
One of the main benefits of the AEGM if compared to the RIVM, is that the AEGM
does not need the clean surplus relationship to hold (Ohlson, 2005, p.331). As we saw in
section 2.2, the CSR does not necessary hold on a per share basis if it holds on a total equity
basis. Specifically, if a company issues or repurchases shares, the CSR may hold on a total
equity basis, though, does not hold on a per share basis. Because the AEGM does not rely on
the CSR, this model can be used on a per share basis (Penman, 2005, p.370). Furthermore,
Ohlson (2005) argues that perfect balance sheets result in perfect earnings measures when the
CSR holds. However, perfect earnings measures do not automatically results in perfect
balance sheets because of the canceling error concept. Under this concept, if a constant is
deducted from the balance sheets, earnings do not change and thus remain perfect but the
balance sheet is not perfect anymore because of the deducted constant. Consequently, an
investor should focus on earnings and not on book values. Hence, an investor should use the
AEGM and not the RIVM (Ohlson, 2005, pp.333-334).
In addition, investors seem to rely more on earnings than on book value in practice.
Partly because book values may, as discussed above, not be stated perfectly when earnings are
(while if the balance is perfect the earnings are also always stated perfectly when the clean
surplus relationship holds). Furthermore, book values look back while earnings look forward
(and investors want to look forward). Because of those observations the model that is used by
investors, should also rely on earnings and their future growth and not on book values and
23

their future growth. The RIVM values a company by estimating growth in book value (as was
explained in section 2.2), while the AEGM tries to value a company using estimates of next
period earnings and future earnings growth. Consequently, an investor should use the AEGM
and not the RIVM (Ohlson, 2005, pp.334-342).
Another difference between the AEGM and the RIVM is the starting point. The
starting point of the RIVM is the book value of equity, while the starting point of the AEGM
is the capitalized next period earnings (eps
1
/r
e
). According to Ohlson (2005, p.331), the
capitalized next period earnings are a much better starting point than current book value.
Penman (2005, p.373) agrees with this point. However, because eps
1
is speculative and eps
1

could be anything (for example one could set eps
1
equal to ebit
1
), Penman (2005) thinks that
current book value is a better starting point. Furthermore, for some companies current book
value can be a very good anchor, think, for example, of an investor with investments in
companies where the balance is equal to fair value. In addition, one of the most uncertain
components of both models is the cost of equity capital. Because the AEGM capitalizes eps
1
,
this model automatically puts a lot of weight on this uncertain component, more than the
RIVM (Penman, 2005, pp. 372-376).
Finally, firms can increase their earnings (and earnings growth) by borrowing more
money, even though borrowing does not create value according to Modigliani and Miller
conditions (which are assumed in the Ohlson and Juettner-Nauroth (2005) paper). Hence,
when the AEGM is used a firm may be worth more if a company borrows more, although this
does not correspondent with the MM conditions (Penman, 2005, p.375).
As stated in the introduction of this section, Penman (2005) has conducted a simple
study, comparing the RIVM with the AEGM. He used two year ahead earnings forecasted by
analysts, assumed a long term growth rate of g
agr
=g
riv
=4% and a constant cost of capital of
10%. With those data, Penman (2005) estimated the value of all U.S. traded equities in the
period 1975-2002 by using both equation (10) for the RIVM (see section 2.1) and equation
(18) for the AEGM (explained in section 4.1). All forecasted values were then divided by the
current price. If the market is efficient (that is, the real value of a company is the same as the
current market price), V/P should be equal to one. Consequently, Penman (2005) assumes that
the best model has a median V/P ratio of 1.00. The median V
riv
/P ratio is exactly 1.0 in the
period 1975-2002 with the assumptions he made. However, the median V
agr
/P ratio is 2.02 in
the same period. Thus, based on this study, the RIVM seems to be better in explaining current
stock prices. Note that this does not necessary mean that the RIVM is also better in predicting
future returns (Penman, 2005, pp.376-378).
24

In summary, the main benefits of the AEGM are the no longer needed clean surplus
relationship, perfect balance sheets result in perfect earnings measures while the converse is
not necessary true, and finally investors rely more on earnings than on book values in
practice, hence the model they use should also rely more on earnings. Drawbacks of the
AEGM are the use of capitalized next year earnings as a starting point, because of this the
uncertain cost of capital is made very important and for some companies book value of equity
is a very good starting point. Furthermore, firms can increase their earnings by borrowing
more money, even though this does not increase value, according to the AEGM, though, this
does increase value. In addition, the study conducted by Penman (2005) shows that the RIVM
is better in explaining current stock prices than the AEGM.


5. Empirical pilot study

The empirical pilot study will be discussed in this section. First, I will describe the method of
the empirical study. Then, the stocks and other data I have used will be outlined. Finally, the
results of the study will be published and discussed.


5.1 Method

For this empirical pilot study 20 stocks that were listed on the Amsterdam Exchange index
(AEX index) on December 31, 2003 will be used (see sample for details). First, the
fundamental value (V) of those stocks will be estimated using both the RIVM and three
different versions of the AEGM (see section 4.1). For both models different scenarios will be
made, based on the discount factor (for both models), the growth in residual income (for
RIVM) and growth in abnormal growth in earnings (for AEGM). See section 5.2 (sample) for
details about the different scenarios. In addition, I will make another scenario based on the
book-to-price multiple (by dividing book value of equity by the stock price as of December
31, 2003) in order to determine whether this scenario is able to predict future returns.
The fundamental value of all scenarios (book value of equity for the book value
scenario) will then be divided by the stock price as of December 31, 2003. Further, the stocks
will be allocated to five portfolios based on their V/P (B/P) ratio. The four stocks with the
highest V/P ratio will be allocated to portfolio 1 and the four stocks with the lowest V/P ratio
25

will be allocated to portfolio 5. Then, the returns of all stocks and portfolios will be calculated
for one, two and three years after the valuation date (December 31, 2003). In addition, the
return of the AEX will be calculated in order to determine whether the different scenarios are
able to yield higher returns than the AEX index.
Because a high V/P ratio implies that the current stock price is too low, I expect
portfolio 1 to yield higher returns than portfolio 5. If a scenario is able to predict that portfolio
1 yields higher returns than portfolio 5 in at least two of the three investigated future years, I
will conclude that this scenario has low predictive ability. Moreover, if a scenario is also
able to produce higher returns than the AEX index in the investigated period for at least two
years and for at least two different investment strategies (explained below), I will conclude
that this scenario has high predictive ability, because in that case an investor could have
made higher returns by investing in high V/P (or B/P) ratio stocks than he could have made by
investing in the AEX index. Consequently, the scenario is valuable for the investor. However,
a portfolio can have high predictive ability only if it has also low predictive ability, because
otherwise an even higher return could have been made by investing in low V/P ratio stocks,
thus in that case it is probably only a coincidence that the high V/P portfolio yield higher
returns than the AEX index.
In order to determine whether a scenario has high predictive ability, four different
investment strategies are used. In the first investment strategy, equal portions (25%) of all
high V/P ratio stocks are bought. The second investment strategy buys an equal portion of all
high V/P ratio stocks and sells an equal portion of all low V/P ratio stocks (thus then half of
the portfolio consists of long positions and half of the portfolio consists of short positions).
The third investment strategy assumes that the highest V/P ratio stock also yields the highest
return. Consequently, 40% is invested in the highest V/P ratio stock, 30% in the second
highest, 20% in the third highest and 10% in the fourth highest V/P ratio stock. Finally, the
fourth investment strategy expects the highest V/P ratio stock to yields the highest return and
the lowest V/P ratio stock to yield the lowest return. Consequently, again 40% is invested in
the highest V/P ratio stock, 30% in the second highest, 20% in the third highest and 10% in
the fourth highest V/P ratio stock, furthermore 40% is sold in the lowest V/P ratio stock, 30%
in the second lowest, 20% in the third lowest and 10% in the fourth lowest V/P ratio stock
(again, then half of the portfolio consists of long positions and half of the portfolio consists of
short positions). Scenarios where the second and fourth strategies yield better returns than the
first and third strategies respectively, relatively to the AEX index, imply a very high
26

predictive power because the two highest (lowest) V/P ratio stocks, where 70% is invested
(sold) in, also yield the highest returns.
The actual return of the AEX index will then be subtracted from the actual return of
each investment strategy, in order to determine whether and by how much an investment
strategy was able to yield higher returns than the AEX. This will be done for one, two and
three years after the valuation date.


5.2 Sample

As stated earlier, for this empirical pilot study 20 stocks that were listed on the AEX index on
December 31, 2003 will be used. There were 25 stocks listed on the AEX at that time,
however, for some stocks there was not enough data available and some stocks disappeared
from the stock exchange before December 31, 2006 (recall that the return of up to three years
after the valuation date will be calculated, thus each stock needs to be listed on a stock
exchange until December 31, 2006 to be able to calculate the returns for up to three years).
Recall from section 2 (for the RIVM) and section 4 (for the AEGM) we need the
following data for each stock in order to do a valuation: book value of equity of 2002,
forecasted earnings, number of shares outstanding, dividend payout ratio (I will assume that
the dividend payout ratio will be the same in future years as it was in fiscal year 2002) and
cost of equity capital. Furthermore, we need an estimate of growth in residual income (g
riv
)
and growth in abnormal growth in earnings (g
aeg
) for each RIVM and AEGM scenario
respectively. Book value of equity of 2002, number of shares outstanding and dividend payout
ratio of 2002 can be found in the financial report of 2002 of each company. Forecasted
earnings for up to five years will be estimated by using the mean of the analysts forecasts that
were available on December 18, 2003 (for most companies at least 15 analysts estimated
future earnings, for some companies only three analysts forecasted future earnings though).
Cost of equity capital will be estimated by using the earlier mentioned capital asset pricing
model (CAPM). This formula is known as (see next page):





27



where:
= cost of equity capital
= risk free rate
= market risk premium (expected market return over the risk free rate)
= equity beta, risk of the stock relatively to the market (for example, each time
the market rises 1%, the stock rises 2%, this stock then has an equity beta of 2)

The risk free rate will be set on 4,33%, this was the long term interest rate in The Netherlands
in December, 2003 (ECB, 2008). The average future market return is difficult to estimate (as
is the market risk premium). I will set the expected future market return on 12% (the average
market return of the past 25 years was approximately 15%). Consequently, the market risk
premium will be set on 12 4,33 = 7,67%. Equity beta will be estimated by comparing the
return of the market for each month for the previous five years with the return of each stock of
each month for the previous five years. Recall from the previous section that different
scenarios will be made based on, among others, cost of equity capital. Consistent with Lee et
al. (1999), I will also use a short term risk free rate. This short term risk free rate will be set
on 2,13%, this was the 1 month euribor (euro interbank offered rate) in December, 2003 in
Europe (ECB, 2008).
Table 1 shows the companies that were listed on the AEX index on December 31,
2003. The first twenty stocks are included in the sample, the last five stocks are not included
because of the above described reasons. For the stocks included in the sample, some variables
used to determine the V/P ratio, have been published.









(21)
28

Table 1

Companies and some variables used to calculate the value according to the AEGM and RIVM
Company Name price 2003 bps 2002 dps 2003 eps 2003 ltg
e
ABN-Amro Holding NV 18,22 6,57 1,10 1,87 9,00% 1,35
Aegon NV 11,40 9,40 0,28 1,12 7,50% 1,72
Ahold NV, Koninklijke 5,99 1,68 0,00 0,12 9,40% 1,24
Akzo-Nobel NV 28,27 7,33 1,01 2,41 4,30% 1,00
ASML Holding NV 14,85 2,73 0,00 -0,32 15,00% 1,66
Buhrmann NV 6,59 12,68 0,01 0,20 10,00% 1,20
DSM NV, Koninklijke 36,85 50,93 1,07 2,10 7,00% 0,79
Fortis NV 15,59 4,20 0,75 1,51 10,00% 1,04
Getronics NV 1,65 0,30 0,00 -0,10 12,00% 2,35
Hagemeyer NV 2,10 8,48 0,00 -1,33 -40,00% 1,21
Heineken NV 30,78 6,49 0,41 2,07 9,15% 0,12
ING Groep NV 17,79 8,85 0,84 1,95 9,50% 1,41
KPN NV, Koninklijke 5,77 0,00 0,00 0,28 7,55% 1,04
Koninklijke Olie NV 39,43 28,83 0,61 3,23 2,63% 0,71
Numico NV, Koninklijke 22,20 1,04 0,26 0,74 8,00% 1,53
Philips Electronics NV, Koninklijke 23,15 10,58 0,11 0,45 10,00% 1,29
Reed Elsevier NV 9,37 2,75 0,31 0,20 10,00% 0,50
TPG NV 18,36 5,92 0,28 0,87 6,89% 0,70
Unilever NV 49,79 9,97 1,55 2,76 11,00% 0,38
Van der Moolen Holding NV 7,15 8,13 0,28 0,60 7,00% 1,00
Listed on AEX, though not included in sample:
Gucci Group NV
LogicaCMG NV
SBM Offshore NV
VNU NV
Wolters Kluwer NV
Notes:
All 25 companies stated in table 1 were listed on the AEX index on December 31, 2003. The first 20 companies are used in the sample, the last 5 companies are
not because not enough data was available for those companies, or the companies were no longer listed on a stock exchange on December 31, 2006. Price 2003
is the price per share on December 31, 2003, bps 2002 is the book value of equity per share at the end of fiscal 2002, dps 2003 are the estimated dividends per
share in 2003 based on the payout ratio in 2002 or the dps of 2002 divided by 6% of the total assets for companies with negative eps in 2002, eps 2003 are the
estimated earnings per share for fiscal 2003 using analysts' forecasts of earnings, ltg is the long-term growth in earnings forecasted by analysts, and e is the
equity beta, estimated by comparing the return of the market of each month for the previous five years with the return of each stock of each month for the
previous five years. All amounts are in euros, except ltg and e.

29


In addition to the different risk free rate scenarios, different scenarios will also be made based
on the growth in residual income (for the RIVM) and based on the growth in abnormal growth
in earnings (for the AEGM). For the RIVM five different scenarios will be made based on
different estimates of growth in residual income. Recall from section 2.3 that there are two
different points of view in literature. According to the first standpoint, abnormal earnings will
be driven to zero in the long run because of competitive forces. Alternatively, according to the
second standpoint the growth in abnormal earnings should be positive if conservatism were to
occur in the future as well because of the undervalued book value of equity (Giner and
Iiguez, 2006). Both standpoints will be investigated in this empirical pilot study.
The first RIVM scenario is based on the first standpoint in literature, the expected
growth in residual income is -50% for this scenario, which implies that abnormal earnings are
expected to fade away in 5 years. The second scenario is also based on the first standpoint in
literature as well as the Bradshaw (2004) paper, the expected growth in residual income will
be -32% for this scenario, which implies that abnormal earnings are expected to fade away in
10 years. The third RIVM scenario is based on multiple papers (e.g. Frankel and Lee, 1998),
the expected growth in residual income is 0% in this scenario. The fourth scenario is based on
the second standpoint of literature the expected growth in residual income is set on 2% for
this scenario. Finally, for the fifth scenario we need to distinguish companies with negative
residual income in the final forecasted year (fiscal year 2007 for this study) and companies
with positive residual income in that year. In this scenario I assume that it is not reasonable to
expect that negative residual income will grow. That is, I expect that companies with negative
abnormal earnings will not generate even more abnormal earnings in perpetuity. This is also
argued in Francis et al. (2000, p.53), according to this paper companies that generate
abnormal earnings need to pay-out negative dividends (that is, they need to issue shares),
however, since it is impossible to pay-out negative dividends in perpetuity, those companies
will eventually go bankrupt. Consequently, in the fifth scenario it is assumed that for
companies with negative abnormal earnings in 2007, abnormal earnings will grow with -50%
and for companies with positive abnormal earnings in 2007, abnormal earnings will grow
with 0%.
For the AEGM, three different scenarios will be made by using the formulas described
in section 4.1. Additionally, two different versions of equation (17) will be used because when
using this formula much of the estimated fundamental value of a company depends on the
final forecasted year. Possibly, analysts are not able to estimate earnings for five years ahead.
30

Consequently, the first version uses forecasts for five years ahead (until 2007) and the second
version uses analysts forecasts for four years ahead (until 2006). The RIVM does not have
this problem, because the expected fundamental value of a company does not depends that
much on the final year. Finally, two scenarios of each formula will be made by using a 0%
growth in abnormal growth in earnings and a 3% growth in abnormal growth in earnings,
consistent with Easton (2004, p. 85).


5.3 Results

Consistent with Frankel and Lee (1998, pp.284-294), the results showed that the ranking of
stocks by V/P ratio do not depend much on the estimated cost of equity capital. Therefore,
only the results of the long term risk free rate (4,33%) have been published. Moreover, for the
AEGM the ranking of the scenarios do not depend on the expected growth in abnormal
growth in earnings. Because of this, only the results of the 0% scenario will be published.
Consequently, there are now 10 scenarios left. Five scenarios based on the RIVM, four
scenarios based on (different versions of) the AEGM and one scenario based on the book-to-
price multiple (or B/P ratio).
Table 2 shows the results of the different scenarios. From left to right, the first
scenario is the scenario with the best results (highest predictive power) and the last scenario is
the scenario with the lowest predictive power. Note that the published returns are buy-and-
hold returns, thus the returns in the column 2 are the realized returns if an investor would have
bought a portfolio or index on December 31, 2003 and sold the stock on December 31, 2005
(2 years later).
After the publication some conclusions based on those results will be formulated,
possible future research areas will be discussed and I will give answer to the question whether
the AEGM is better in predicting future returns than the RIVM.

31

Table 2
Summary of results of the empirical pilot study performed to determine whether the AEGM is better in predicting future returns than the RIVM
Year 1 2 3 1 2 3 1 2 3 1 2 3
Scenario 5 RIVM (mixed g) Scenario 3 AEGM (Eq. 19) Scenario 1 RIVM (g= -50%) Scenario 2 RIVM (g= -32%)
S2: RetP1 - RetP5 20,3% 53,4% 85,8% 15,4% 21,4% 76,3% 6,4% 38,4% 44,8% 6,4% 38,4% 44,8%
S4: RetP1 - RetP5 25,2% 62,9% 119,4% 21,9% 47,1% 102,4% 19,8% 65,0% 105,3% 17,6% 54,5% 94,2%
Return AEX 14,8% 34,0% 78,8% 14,8% 34,0% 78,8% 14,8% 34,0% 78,8% 14,8% 34,0% 78,8%
Ret S1 - Ret AEX 1,3% 43,1% 31,1% 0,7% 29,5% 28,2% -3,9% 37,3% 9,2% -3,9% 37,3% 9,2%
Ret S2 - Ret AEX 5,5% 19,5% 7,0% 0,6% -12,5% -2,4% -8,3% 4,4% -34,0% -8,3% 4,4% -34,0%
Ret S3 - Ret AEX 6,3% 48,9% 50,9% 6,8% 47,1% 53,2% 4,3% 54,7% 44,5% 2,1% 44,2% 33,4%
Ret S4 - Ret AEX 10,5% 28,9% 40,6% 7,1% 13,2% 23,6% 5,0% 31,0% 26,6% 2,8% 20,5% 15,5%
B/P Scenario Scen. 2 AEGM (Eq.17 efeps4) Scenario 3 RIVM (g= 0%)
S2: RetP1 - RetP5 6,4% 38,4% 44,8% 12,4% -0,5% 20,0% 9,3% -9,0% 6,0%
S4: RetP1 - RetP5 4,6% 46,7% 57,3% 11,9% -5,7% 7,9% -2,1% -27,5% -25,1%
Return AEX 14,8% 34,0% 78,8% 14,8% 34,0% 78,8% 14,8% 34,0% 78,8%
Ret S1 - Ret AEX -3,9% 37,3% 9,2% -11,6% -0,7% -16,2% -3,0% 12,4% -10,1%
Ret S2 - Ret AEX -8,3% 4,4% -34,0% -2,3% -34,5% -58,8% -5,5% -42,9% -72,8%
Ret S3 - Ret AEX 2,0% 55,2% 41,1% -14,7% -13,4% -33,3% -9,3% -4,0% -27,3%
Ret S4 - Ret AEX -10,2% 12,8% -21,5% -2,9% -39,7% -70,9% -16,9% -61,4% -103,9%
Scen. 1 AEGM (Eq.17 efeps5) Scen. 4 AEGM (Eq. 18) Scenario 4 RIVM (g= 2%)
S2: RetP1 - RetP5 3,1% -4,5% -15,4% -3,2% -21,5% -15,8% 2,0% -26,7% -13,5%
S4: RetP1 - RetP5 0,1% -16,5% -23,7% -10,7% -40,5% -39,9% -5,6% -51,6% -56,7%
Return AEX 14,8% 34,0% 78,8% 14,8% 34,0% 78,8% 14,8% 34,0% 78,8%
Ret S1 - Ret AEX -3,3% 11,5% -3,4% -11,4% -9,1% -30,0% -10,3% -5,3% -29,6%
Ret S2 - Ret AEX -11,7% -38,4% -94,2% -18,0% -55,5% -94,6% -12,8% -60,7% -92,3%
Ret S3 - Ret AEX -7,3% -1,2% -14,4% -16,3% -20,7% -41,3% -12,3% -11,1% -35,1%
Ret S4 - Ret AEX -14,6% -50,4% -102,5% -25,5% -74,5% -118,7% -20,4% -85,6% -135,5%
Notes:
Table shows the buy-and-hold returns for ten different scenarios for 1, 2 and 3 years after the valuation date, which is December 31, 2003. Five scenarios are based on the RIVM, four scenarios are based on the AEGM and one scenario is based on the book value of
equity. For RIVM scenario 1 a growth in residual income of -50% is chosen, scenario 2 represents a growth rate of -32%, scenario 3 a growth rate of 0%, scenario 4 a growth rate of 2% and scenario 5 a mixed growth rate of -50% for companies with negative abnormal
earnings in the final forecasted year and 0% for companies with positive abnormal earnings in the final forecasted year. For AEGM scenario 1 equation (17) has been used with analysts forecasts of up to five years after the valuation date, for scenario 2 equation (17)
has also been used, though now with forecasts of up to four years after the valuation date, for scenario 3 equation (18) has been used and for scenario 4 equation (19) has been used, each AEGM scenario represents a growth in abnormal growth in earnings of 0%, all
AEGM equations are described in section 4.1. For the book value scenario, the book value of equity of fiscal 2002 has been used. All values have been divided by the stock price as of December 31, 2003 (see table 1). The four highest V/P (or B/P) ratio stocks formed
portfolio 1 (P1), the four lowest V/P ratio stocks formed portfolio 5 (P5). For each scenario four different investment strategies have been used. SX stand for strategy X. Consequently, S2: RetP1 RetP5 represents the strategy where 25% is invested (shorted) in
each P1 (P5) stock, so that the percentages in that row represents the average return of the four P1 stocks above/below the average return of the four P5 stocks in the investigated period. Row S4: RetP1 RetP5 represents the same, though now 40% is invested
(shorted) in the highest (lowest) V/P ratio stock, 30 % in the second highest (lowest), 20% in the third highest (lowest) and 10% is invested (shorted) in the fourth highest (lowest) V/P ratio stock. Return AEX represents the buy-and-hold returns of the AEX index for
1, 2 and 3 years after the valuation date. Ret S1 and S3 Ret AEX represents the buy-and-hold returns when 25% was invested in each P1 stock or 40% was invested in the highest V/P ratio stock, 30% in the second highest, 20% in the third highest and 10% in the
fourth highest V/P ratio stock respectively on the valuation date less the return of the AEX index. Ret S2 and S4 Ret AEX represents the returns of strategy 2 and strategy 4 (described above) less the return of the AEX index if the respective strategies were pursued
on the valuation date. The ranking of the scenarios is based on their predictive power, with the upper-left scenario being the scenario with the highest predictive power and the bottom-right scenario being the scenario with the lowest predictive power.
32

As can be concluded from table 2, five of the ten investigated scenarios are able to predict that
portfolio 1 (the high V/P ratio portfolio) yields higher returns than portfolio 5 (the low V/P
ratio portfolio) and are able to yield higher returns than the AEX index in at least two years
and for at least two different investment strategies. Hence, five out of the ten scenarios have
high predictive power. Two scenarios are only able to predict that portfolio 1 yields higher
returns than portfolio 5 for at least two years and thus have low predictive power (an investor
would invest in the AEX index rather than in the highest V/P ratio portfolio stocks since the
AEX yield higher returns). Three scenarios are neither able to predict which stocks yield
relatively high returns nor which stocks yield relatively low returns in at least two different
years, these scenarios have no predictive power.
Inconsistent with the findings of Francis et al. (2000), a comparison between the
results of the different RIVM scenarios show that the predictive ability of this model depends
heavily on the choice of the growth in residual income. The scenario with a growth rate of 2%
(which assumes that because of accounting conservatism residual income will grow in
perpetuity), has no predictive power. In addition, the scenario with a growth rate of 0% has
low predictive power. Thus an investor would not rely on this scenario because it yields lower
return than the AEX index in almost every year and for almost every strategy. However, the
scenarios with a negative growth rate all have high predictive power. Moreover, the scenario
with a negative growth rate of -50% for companies with negative residual income in 2007 and
neutral growth (0%) for companies with positive abnormal earnings in 2007 produces the best
results. These growth rates imply that companies with negative abnormal earnings will see
their abnormal earnings fade away in 5 years, while companies with positive abnormal
earnings will see their abnormal earnings fade away in year . This is the only scenario that is
able to yield higher returns than the AEX index in all years and for all strategies. Because this
is only a minor study, which investigates only 20 stocks listed on one index for one time
period, future research needs to be conducted in order to determine which growth rate
produces the best results. Nevertheless, this study points to a mixed or negative abnormal
earnings growth rate which assumes that because of competitive forces abnormal earnings
will be driven to zero.
When the results of the different AEGM scenarios are compared, only equation (19),
which was used by Gode and Mohanram (2003, p.402), has high predictive power. Equation
(18), which was used by Penman (2005) to determine whether the AEGM was better in
predicting current stock prices than the RIVM (see section 4.3), yields the lowest results and
does not have any predictive power at all. Furthermore, the two versions of equation (17)
33

produce very different results. The version with the lowest forecast horizon (where analysts
forecasts of up to 2006 are used) produces the best result, which still is low predictive power.
However, the version which uses analysts forecasts of up to 2007, has no predictive power.
This odd result, it is widely accepted that a longer time horizon produces better results and
recall from section 3.1 that Lee et al (1999, p.1735) showed that for the RIVM a longer time
horizon does produce better results in most cases, can probably be explained by the high
reliance of the AEGM on the last forecasted year in combination of the inability of analysts to
forecast future earnings a long time ahead. Another possible explanation is that the terminal
value of the AEGM is heavily influenced by the final year and because of this when the
earnings of the final year are not correctly estimated, the terminal value also is not correctly
estimated. However, since the estimated growth in abnormal growth in earning (which also
heavily influences the terminal value) does not have any effect on the ranking of the
scenarios, this seems to be an unlikely explanation.
A comparison between the results of the different models show that from the five
scenarios with high predictive power, three scenarios were made by using the RIVM, one
scenario was made by using the AEGM and also one scenario was made by using the B/P
ratio. Moreover, the best scenario was made by using the RIVM and the B/P ratio produced
the worst results of the scenarios with high predictive power. This is consistent with other
studies (e.g. Frankel and Lee, 1998) which also show that the RIVM is better in predicting
future returns than the B/P ratio. However, the scenario with the least predictive ability of all
scenarios was also made by using the RIVM. Consequently, it is important to estimate the
growth in residual income as correctly as possible because otherwise the AEX index yields
better results. Two AEGM scenarios have no predictive power and one RIVM and one
AEGM scenario have low predictive power. Overall, the results do not show that the AEGM
is better able to predict future returns than the RIVM. Alternatively, the results rather show
that the RIVM is better in predicting future returns than the AEGM. This is consistent with
the findings of Francis et al. (2000), a likely explanation is that the RIVM consists of both a
book value component and a flow variable, while the AEGM only consists of a flow variable.
Since flow variables are more uncertain than the book value component, the RIVM produces
better results (Francis et al., 2000, p.45-57)
Finally, the results of the different strategies will be compared. Because it is not very
useful to compare the results for the scenarios with no or low predictive power (because those
scenarios produce lower returns than the AEX index), I will only discuss the results of the
high predictive scenarios. Those results show that strategy 3, which invests 70% in the first
34

two V/P ratio stocks, yields the best results for each of the high predictive power scenario.
This means that the scenarios are not only able to predict which four stocks yield the best
returns, but are also able to predict which of those four stocks yields the best return.
Furthermore, consistent with the findings of Frankel and Lee (1998), the best results are being
made after a longer (two to three years) time horizon. Because the AEX index has risen
78,8% over three years in the investigated period, the strategies where stocks are bought in
the highest V/P portfolio and sold in the lowest V/P portfolio (strategy 2 and 4) yield lower
returns than the strategies where stocks are bought only. However, again in most cases the
strategy where 70% is invested in the two highest V/P ratio stocks and 70% shorted in the two
lowest V/P ratio stocks (strategy 4) produces better results than the strategy where 25% is
invested in each high V/P ratio stock and shorted in each low V/P ratio stock (strategy 2).
Consequently, the high predictive power scenarios are not only able to predict which stock
produces the highest return, but is also able to predict which stock yields the lowest return.
Future research is needed whether it is sensible to buy stocks only or to buy and sell stocks for
multiple time periods. Potentially, it is even possible to predict whether the index will
increase or decline in the future period by taking the weighted average V/P ratio of the index.
This can be investigated by looking to multiple time periods and compare the weighted
average V/P ratio of the AEX index and the future return of the different time periods.
Obviously, if it is possible to predict whether the index will increase or decline, it is probably
reasonable to follow one of the buy strategies when the index is expected to increase and
follow a neutral or sell strategy when the index is expected to decline.
Summarizing, the RIVM has the most high predictive power scenarios. Moreover, the
highest predictive power scenario has been made by using the RIVM. Three out of five RIVM
scenarios have high predictive power, while only one out of four AEGM scenarios has high
predictive power. Consequently, the RIVM seems to be better in predicting future returns than
the AEGM. It is important, though, to estimate growth in residual income as good as possible,
because otherwise the RIVM has no predictive power at all. The results of this study point to
a negative or mixed growth in residual income. However, since in this study only one time
period is investigated, more research is needed to determine which growth in residual income
can best be used when predicting future returns. When comparing the different strategies, the
buy strategies yield the highest returns. Furthermore, the strategies which invest (sell) 70% in
the highest (lowest) two V/P ratio stocks, yield better returns than the strategies where 50% is
invested (sold) in the highest (lowest) two V/P ratio stocks. More research is needed in order
to determine whether it is better to pursue a buy strategy or a buy/sell strategy. Furthermore,
35

more research needs to determine whether the scenarios are always (in multiple time periods)
able to predict which stock will yield the highest (lowest) return in the high (low) V/P ratio
portfolio and whether it is possible to predict whether a stock index will increase or decline in
future years.


6. Summary

In this thesis I have tried to give answer to the question whether the AEGM is better in
predicting future returns than the RIVM. The literature review shows that the RIVM is able to
predict future returns while high V/P ratio stocks, which produce higher returns, are not
significantly riskier than low V/P ratio stocks. Hence, the higher returns are probably due to
mispricing of the market. Furthermore, the literature review shows that the RIVM is better in
predicting future returns than other frequently used models. Moreover, the RIVM is able to
predict future returns for different kinds of companies, even for companies which do not
apply clean surplus accounting (although CSR is assumed in the RIVM) and companies in
countries with different accounting regimes.
Unfortunately, not much research has been conducted to the AEGM. This model does
not rely on the CSR and, since many companies do not apply the CSR, this model should be
able to predict future returns better than the RIVM, although both models can be derived from
the DDM. Only Penman (2005) has conducted a very small study to one of the variants of the
AEGM to investigate whether the AEGM is better in explaining current stock prices. He
found that the AEGM is not better in explaining current stock prices than the RIVM.
Because little research has been done to the AEGM, I have conducted my own
empirical pilot study. The results of this study show that the AEGM is, in general, not better
in predicting future returns than the RIVM. Moreover, when comparing different scenarios of
both models, the RIVM seems to be better in predicting future returns in most cases. Hence, it
seems that the drawbacks of the AEGM, described in section 4.3, are greater than the benefits.
For all scenarios with high predictive power, I find that the best results are made over a longer
time horizon of at least two years. Although both models do not rely much on the cost of
equity capital when using the V/P ratio to predict future returns, the RIVM does rely much on
the chosen growth in residual income. Because not much research has been conducted to this
growth ratio, it would be interesting to investigate which growth rate is best when predicting
future returns of multiple time periods. In this minor study, I find that a high negative growth
36

ratio for companies with negative residual income in the final forecasted year (the year before
the terminal value is calculated) and a 0% growth rate for companies with positive residual
income in the final forecasted year, results in the greatest predictive power of the model.
In summary, when an investor wants to value a company in order to gain high returns,
he should rely on the RIVM, even for companies with dirty surplus accounting items. The
AEGM is generally not better in predicting future returns than the RIVM. Future research to
the RIVM might result in even more predictive power for this model.



































37

References
Ali, A., Hwang, L.S., and Trombley, M.A. (2003). Residual-Income-Based Valuation Predicts
Future Stock Returns: Evidence on Mispricing vs. Risk Explanations. The Accounting
Review, 78, (2), 377-396.

Bradshaw, M.T. (2004). How Do Analysts Use Their Earnings Forecasts in Generating Stock
Recommendations? The Accounting Review, 79, (1), 25-50.

Easton, P. (2004). PE Ratios, PEG Ratios, and Estimating the Implied Expected Rate of
Return on Equity Capital. The Accounting Review, 79, (1), 73-95.

ECB (2008). Long term interest rate page ([Link]
html/[Link]#fn1), April, 30.

ECB (2008). Short term interest rate page ([Link]
node=bbn175&start=&end=&trans=N&dvfreq=M), April, 30.

Fama, E.F. and French K.R. (1997). Industry cost of equity. Journal of Financial Economics,
43, 153-193.

Frankel, R. and Lee, C.M.C. (1998). Accounting valuation, market expectation and cross-
sectional stock returns. Journal of Accounting and Economics, 25, 283-319.

Francis, J., Olsson, P., and Oswald, D. R. (2000). Comparing the Accuracy and Explainability
of Dividend, Free Cash Flow and Abnormal Earnings Equity Value Estimates. Journal
of Accounting Research, 38, (1), 45-70.

Giner, B. and Iiguez, R. (2006). An empirical assessment of the Feltham-Ohlson models
considering the sign of abnormal earnings. Accounting and Business Research, 36, (3),
169-190.

Gode, D. and Mohanram, P. (2003). Inferring the Cost of Capital Using the Ohlson-Juettner
Model. Review of Accounting Studies, 8, 399-431.

Healy, P.M. (1985). The effect of bonus schemes on accounting decisions. Journal of
accounting and Economics, 7, (1-3), 86-107.

Isidro, H., OHanlon, J., and Young, S. (2006). Dirty Surplus Accounting Flows and
Valuation Errors. Abacus, 42, (3/4), 302-344.

Lee, C.M.C. and Swaminathan, B. (1999). Valuing the Dow: A Bottom-Up Approach.
Financial Analysts Journal, 4-23.

Lee, C.M.C., Myers, J., and Swaminathan, B. (1999). What is the Intrinsic Value of the Dow?
Journal of Finance, 54, (5), 1693-1741.

Ohlson, J.A. (1995). Earnings, book values, and dividends in equity valuation. Contemporary
Accounting Research, 11, (2), 661-687.


38

Ohlson, J.A. (2005). On Accounting-Based Valuation Formulae. Review of Accounting
Studies, 10, 323-347.

Ohlson, J.A. and Juettner-Nauroth, B.E. (2005). Expected EPS and EPS Growth as
Determinants of Value. Review of Accounting Studies, 10, 349-365.

Palepu, K.G., Healy, P.M., Bernard, V.L. and Peek, E. (2007). Business Analysis and
Valuation. London: Thomson Learning.

Penman, S. H. (1998). A Synthesis of Equity Valuation Techniques and the Terminal Value
Calculation for the Dividend Discount Model. Review of Accounting Studies, 2, 303-
323.

Penman, S. H. (2005). Discussion of On Accounting-Based Valuation Formulae and
Expected EPS and EPS Growth as Determinants of Value. Review of Accounting
Studies, 10, 367-378.