Migration between Value- and Growth portfolios; The influence on the value premium

Migration between Value- and Growth

portfolios; The influence on the value

Migration between Value- and Growth

portfolios; The influence on the value

premium

Stefan Linskens

Rijksuniversiteit Groningen

Abstract

In this thesis, the value premium is examinated in the Dutch Stock Market between June 1st, 1981 and May 31st, 2007. Based on the Cash Flow to Price ratio the value premium and the migration

of value- and growth stocks are investigated.

However, on average, value stocks earn higher returns than growth stocks this is not backed by convincing statistical significance.This thesis further concludes that switching style value stocks

do not earn, on average, higher returns than switching style growth stocks. Although fixed style value stocks earn, on average, higher returns than fixed style growth stocks two until five years

Preface

During the course Behavioral Finance, the psychology of investors gained my attention. Because the value premium debate is a clash between rationalists and behaviorists this is the subject of this thesis. Writing it was a big challenge but writing it besides a full time job was even a bigger challenge!

In this preface I will use the opportunity to thank a couple of people for their support. In the first place I will thank my supervisor prof. dr. F.M. Tempelaar who was always willing to advice me when necessary on short notice. His feedback and his monitoring of the progress of the thesis helped me a lot.

Furthermore I will thank my girlfriend Marjolein for here practical help during busy periods. I also want to thank my parents for listening to me in times that my research stagnated. In the last place I will thank my friends for their understanding of my limited time and their support in finishing this thesis.

Table of

I. Introduction ... 4

II. Literature review and theoretical background... 7

IIa. Defining value- and growth stocks... 7

IIb. Returns on value- and growth investing... 8

IIc. Explanations for value stocks superior returns... 11

IId. Migration of value- and growth stocks ... 14

III. Hypotheses and methodology ... 16

IV. Data... 20

V. Results... 22

Va. The value premium... 22

Vb. Size adjusted returns ... 24

Vc. Migration... 25

VI. Summary and conclusion... 29

I. Introduction

Since the early nineties, the value premium is a widely discussed phenomenon in academic literature. Until today, a lot of questions remain unanswered and the outperformance of value stocks over growth stocks is still subject of an extensive debate. The value premium is based on the classifying of stocks in growth- and value stocks and, as its name implies, the premium that investors gain to hold value- instead of growth stocks. The problem goes back to the early 1930s. In their influencing book ‘Security Analysis’ Graham and Dodd (1934) argue that there can be a discrepancy between price and value. In their opinion investors have to be very careful with stocks with rosy prospects. They state that: ‘when growth is generally expected the price rarely is reasonable’. Conversely they state that: ‘securities of companies with unpromising outlooks may sell so low that they offer excellent investment opportunities’.

Nowadays these stocks with excellent growth prospect are labeled growth stocks and the stocks of companies with unpromising outlooks value stocks. The classifying is based on certain stock fundamentals relative to the stocks price. Examples of these ratios are the Book Equity to Market Equity ratio (BE/ME) (Fama and French, 1992), Earnings to Price ratio (E/P) (Basu, 1977) and the Cash Flow to Price ratio (CF/P) (Lakonishok, Shleifer and Vishny, 1994). A stock with a low market price relative to the specific fundamental is called a value stock and a stock with a high market price relative to the specific fundamental is called a growth stock (Wouters, 2006).

Investing in stocks based on value and growth characteristics is a form of style investing. When stocks are classified on similar performance characteristics this is called a style and when investors base their portfolio allocation on this style it is called style investing (Barberis and Shleifer, 2003). According to Barberis and Shleifer (2003) there are two reasons for the appearance of new styles. The first one is financial innovation and the second one is superior performance of a particular style. Value investing is an example of the second reason.

Many studies show that value stocks outperform growth stocks, so there is ample evidence on the existence of a positive value premium1. On the explanation of this positive premium however, is still an ongoing discussion. The discussion is mainly between the rationalists on the one hand and the behaviorists on the other hand. Rationalists believe in market efficiency. A market is efficient when prices always fully reflect available information (Fama, 1970). If prices reflect all available information this implies that investors cannot systematically outperform the market. Market outperformance of certain classes of stocks can, according to this theory, only be explained in

1

terms of risk. From this it follows that the outperformance of value- over growth stocks must be attributable to the higher risk of holding value stocks.

Behaviorists do not believe in efficient markets. According to this group, people do not make decisions in a rational way. Instead of following for example Bayes’ rule, people use heuristics to make decisions. According to this rule, people change expectations in the light of new evidence with the appropriate probabilities of certain outcomes of their decisions in mind. Kahneman and Tversky (1973) provide evidence in contrast with Bayes’ rule. They use experiments to show that people suffer from the representative heuristic in making decisions. Representativeness means that probabilities are judged by the degree that A is representative of B. It can lead to estimation errors because in the real world A is not as depended on B as assumed. An example of the representative heuristic is that investors use past stock performance (B) to evaluate future stock performance (A). In this case they overstate the importance of past stock performance. Because value stocks performed worse than growth stocks in the past, their prediction of the performance of growth stocks will be too optimistic. On the other hand their prediction about value stocks will be too pessimistic. Because future performance is an important part of a stocks price, value stocks become undervalued and growth stocks overvalued. When the actual future performance becomes known investors will be, on average, disappointed in the performance of the growth stocks and positively surprised about the performance of value stocks. This will lead to higher returns of value stocks than for growth stocks. De Bondt and Thaler (1985) were among the first that provided empirical evidence for this investor behavior. In their paper about overreaction in the stock market they compared the returns from companies, listed on the New York Stock Exchange, with low historical returns with the returns from companies with high historical returns. They created a ‘winner’ portfolio and a ‘loser’ portfolio based on the stock returns in a 36 month period. The ‘loser’ portfolio outperformed the ‘winner’ portfolio by 25% in the subsequent 36 months of portfolio formation.

The line of reasoning explained above is known as the extrapolation hypothesis or the expectational error hypothesis.

Recently, some papers on the anatomy of the returns from value- and growth portfolios appeared2. In these papers the migration of value- and growth stocks to respectively the growth- and value portfolio or an intermediary portfolio was discussed. It seems that a few value stocks that become growth stocks and a few growth stocks that become value stocks were largely responsible for the value premium. In these papers, portfolios were created based on (BE/ME). It appeared that there were no large differences in average performance between value- and growth

stocks when only stocks that stay at least two consecutive years in the same portfolio were taken into account.

This thesis will describe the main findings on the value premium until now and will provide some evidence on the value premium in the Netherlands. Furthermore it will provide evidence on a new phenomenon; the migration of value- and growth stocks and the effects on the value premium. This will be done with portfolios based on (CF/P) which is not done yet, as far as I know.

Central question in this thesis is:

Does the value premium exist, in the Dutch Stock Market between June 1st, 1981 and May 31st,

2007, and what is the influence of migrating growth- and value stocks, in this period, on the value premium?

II. Literature review and theoretical background

This section will discuss the main findings until now in the value premium debate. Returns from value investing will be discussed as well as explanations for the superior returns of value- over growth stocks. Lastly, the outcomes on studies about the migration of growth- and value stocks will be given. This section starts with a short explanation of the terms value- and growth stock.

IIa. Defining value- and growth stocks

As described in the introduction, the classifying of value- and growth stocks is based on certain stock fundamentals relative to the stock’s price. Value stocks have a low price relative to, for example, earnings, dividends, book assets and cash flow. On the other hand growth stocks have a high price relative to these fundamentals.

Embedded in these relative price differences is expected future growth. The rationale behind this is that expected future growth is an important component of a stocks price. According to traditional finance the value of a company is based on the discounted future cash flows of that company. The growth factor of these cash flows has therefore a huge impact on the value of the company and as a result the stock price of that company.

Graham and Dodd (1934) define growth stocks as stocks with rosy growth prospects. Growth stocks are therefore stocks with implied high future growth of, for example, earnings per share. Implied future growth is the reason that low BE/ME, CF/P and E/P stocks are growth stocks. This is because there is a positive relation between future growth and the denominator of these ratios. When the growth rate is increasing, price is also increasing and the ratios are, as a result, decreasing. Growth stocks can therefore be defined as stocks with high implied future growth and BE/ME, CF/P and E/P are ratios to capture this future growth. With value stocks it is the other way around. Value stocks can be defined as stocks with a low implied future growth.

IIb. Returns on value- and growth investing

A large amount of empirical evidence on the superior returns of value stocks over growth stocks appeared in the early 1990s. These papers extended the work of earlier research on stock market anomalies such as Basu (1977) and Reinganum (1981) on E/P, Rosenberg Reid and Lanstein (1984) on BE/ME and Bernard and Stober (1989) on CF/P.

One of the first influential articles was written by Fama and French (1992). In this paper the cross section of expected stocks returns was examined. Investigating the cross section of expected stocks returns means: studying the expected returns across different stocks or assets, for one time period. The opposite of cross section analysis is time series analysis which studies stocks or assets over time. This research, on non-financial firms in the US stock market, showed that firms with a high BE/ME ratio earn significantly higher returns than firms with a low BE/ME ratio. Fama and French (1992) created 12 portfolios on BE/ME and the average returns rose from 0,30% per month in the lowest to 1,83% in the highest BE/ME portfolio. High BE/ME stocks later were labeled value stocks and low BE/ME stocks growth stocks. This paper therefore provided evidence for the existence of a premium on value stocks. The low BE/ME portfolio existed of on average smaller stocks than the portfolio of growth stocks, so it was still conceivable that the results were driven by the small firm effect (Banz, 1981). The small firm effect is the effect that the expected stock returns on firms with a small market capitalization are on average higher than the expected stock returns on firms with a large market capitalization.

with future growth opportunities. A company with low risk and a low discount rate for future cash flows can also have a low BE/ME because there is a negative relation between Market Equity and the discount rate. Therefore BE/ME can have more than one explanation with respect to the characteristics of a firm. Lakonishok, Shleifer and Vishny (1994) conclude that the BE/ME is not a clean variable uniquely associated with economically interpretable characteristics and they prefer the use of CF/P. The difference between the lowest CF/P portfolio (glamour stocks) and highest CF/P portfolio (value stocks) was 11% per annum and on a size adjusted basis 8,8%, so the value premium was even a bit larger in comparison with the BE/ME portfolios.

La Porta (1996) used a different methodology to create value- and growth portfolios. Instead of using valuation measures, like other authors, he used a valuation related measure as a classifying variable. The value portfolio in this paper existed of stocks with low analyst earnings forecasts and the growth portfolio of stocks with high analyst earnings forecasts. Because of this, Doukas, Kim and Pantzalis (2002) doubt his conclusions. In their opinion it was not a direct test of the value-growth problem and the outcomes cannot be interpreted as evidence in favor of the value premium. In spite of this criticism, a lot of papers refer to this article so its results are respected. As Fama and French (1992) and Lakonishok, Shleifer and Vishny (1994), La Porta (1996) found significant higher returns for value stocks than growth stocks. The value portfolio outperformed the growth portfolio by 20% per annum. La Porta (1996) used data from firms listed on the New York Stock Exchange (NYSE) and the American Stock Exchange (AMEX).

Although most research on the value premium is performed with US data, there also exists evidence outside the US. Chan, Hamao and Lakonishok (1991) researched the influence of BE/ME, E/P, size and CF/P on stock returns in Japan. The BE/ME had the highest significant positive relation with expected stock returns from the four fundamentals. The CF/P ratio also had a significant positive relation with expected stocks returns. For the E/P ratio, the outcomes were mixed. In isolation, the portfolio of high E/P stocks significantly outperformed the low E/P portfolio, but in a multifactor regression the E/P coefficient was not significantly different from 0 anymore.

therefore concluded that the value premium was a global phenomenon that not only was present in the US.

Table 1 will present the summary of the most important papers on the value premium. Research period, market, fundamental and outcomes will be given.

Table 1. Summary of most important research on the value premium.

Authors Period Market Fundamental Results Fama and

French (1992)

July 1963- December 1990

US firms from the NYSE, AMEX and the Nasdaq

BE/ME High BE/ME portfolio outperformed the low

BE/ME by 1,53% per month Lakonishok, Shleifer and Vishny (1994) April 1968- April 1989

US firms from the NYSE and the AMEX.

BE/ME, E/P and CF/P

Outperformance of high BE/ME, E/P and CF/P portfolio with respectively 10,5%, 11% and 7,6% on average on a yearly basis La Porta

(1996)

July 1981- June 1991

US firms from the NYSE and the AMEX

Analyst forecasts on earnings growth

Low expected earnings growth portfolio outperforms high expected earnings growth portfolio by 20% a year Chan, Hamao and Lakonishok (1991) January 1971-December 1988

Japanese firms listed on the Tokyo Stock Exchange.

BE/ME, E/P and CF/P

BE/ME and CF/P had a positive significant relation with expected stock returns

Fama and French (1998)

1975-1995 US (NYSE, AMEX and Nasdaq) and MSCI firms from Japan, UK, France, Germany, Italy,

Netherlands, Belgium, Switzerland, Sweden, Australia, Hong Kong and Singapore

BE/ME, E/P, CF/P and D/P

12 out of 13 countries had a positive value premium on BE/ME, E/P and CF/P and 10 out of 13 on Dividend to Price (D/P)

IIc. Explanations for value stocks superior returns

The explanation of the value premium is, in contrast to its existence, subject of an extensive debate. One of the mentioned explanations is that it is the result of collective data snooping. By data snooping is meant that significance is the consequence of coincidence and not of a real relation between variables. When for example a randomly chosen hypothesis on a 5% level is tested, it will be on average significant in 5 out of 100 times. Because most evidence on the value premium comes from the US market, and as a consequence researchers use the same data, some people argue that the value premium is just a coincidental finding. There are some arguments against the data snooping explanation. In the first place, value premiums are also found for markets outside the US3. In the second place, there is a logical basis behind the value premium, because it can be tied to patterns of investor behavior and incentives of investment managers (Chan and Lakonishok, 2004). Because of this I believe it is defendable to reject data snooping as explanation for value stocks superior returns over growth stocks.

A more convincing story about the explanation of the value premium was introduced by Fama and French (1992). Besides that high BE/ME stocks earn higher returns than low BE/ME stocks they find something else which was striking. The Betas of high BE/ME stocks were not significantly higher than the Betas of the low BE/ME stocks. This raised questions on the existence of Beta and therefore the validity of the Capital Asset Pricing Model (CAPM) of Sharpe (1964) and Lintner (1965). Fama and French (1992) argued that when asset pricing is rational, BE/ME must proxy for risk because the higher returns for high BE/ME stocks cannot be explained otherwise. Their paper did not provide evidence for this.

This evidence was given a short time later (Fama and French, 1993). In this paper it was argued that to capture risk, the traditional CAPM was not sufficient anymore and the Three Factor Model was introduced. This model is based on three risk factors namely Size, BE/ME and Beta and explains the expected excess returns on a stock or portfolio with the sensitivity of its returns to these three factors. The model is specified as:

),

(

)

(

]

)

[E(

)

(

R

R

b

R

R

s

E

SMB

h

E

HML

E

−

=

−

+

+

where

E

(

R

)

−

R

,

E(

R

)

−

R

the market portfolio, E(SMB) the excess return on stocks of small firms over stocks of large firms and E(HML) the excess return on high BE/ME stocks over low BE/ME stocks.

b

s

h

the factor sensitivities slopes obtained from a time-series regression of excess stock returns on the excess factor returns,

i i i f m i i f i

R

a

b

R

R

s

SMB

h

HML

e

R

−

=

+

(

−

)

+

(

)

+

(

)

+

In this model BE/ME, Size and Beta are sources of systematic risk. Because of this, investors have to be compensated for holding high Beta, Size and BE/ME stocks. BE/ME is an indicator of systematic risk because it is interpreted as a proxy for financial distress by Fama and French (1993, 1995). Low BE/ME firms are companies with a low profitability, where high BE/ME firms are associated with sustained high profitability. Because the earnings of high BE/ME firms are under pressure, these firms often have a lower accessibility to the external capital markets and are forced to accept higher interest rates on loans. In unfavorable market conditions distressed firms are therefore more likely to go bankrupt. Investors have to be compensated for this risk so expected returns on high BE/ME stocks have to be higher than on low BE/ME stocks. The value premium is, following this line of reasoning, just the consequence of rational asset pricing and the linear relation between risk and return.

Fama and French’ (1993) findings are in line with the evidence of Chan and Chen (1991). They showed that there is covariation in returns related to relative distress that is not captured by the market return and is compensated in average returns.

risk explanation is not correct because the value portfolio is not associated with higher betas and higher volatility.

As explained earlier Lakonishok, Shleifer and Vishny (1994) criticized BE/ME as fundamental for the grouping of value- and growth stocks. Because Fama and French (1993) only used portfolios based on BE/ME to test their three factor model, the results were still debatable. When Fama and French (1996) repeat their research for E/P, CF/P and sales growth, however, the three factor model still holded.

Lakonishok, Shleifer and Vishny (1994) get support for their reasoning from La Porta (1996). He argued that if the extrapolation argument is true, investors have to react on average more positively on earnings announcement of value stocks than of growth stocks. He did find a significant difference between the announcement effects. La Porta, Lakonishok, Shleifer and Vishny (1997) provided some extra evidence on the extrapolation hypothesis. They showed that a significant proportion of the differences in returns between value- and growth stock is attributable to earnings surprises which are systematically more positive for value stocks. Their findings provided no evidence for the risk argument. They argued that the risk premium implies that event returns should be higher than nonevent returns for both growth- and value stocks. Earnings announcement days contain therefore a disproportionately fraction of the stock’s risk premium. Because the event returns for glamour stocks are lower than the nonevent returns, they reject the risk argument.

The conclusion is that the value premium puzzle is not solved yet. Its existence is not debatable anymore, however its nature is. It is very hard to choose between the two argumentations, because it is never possible to reject the metaphysical approach of risk. This approach means that all outperformance of specific classes of stocks has to be explained by risk. If a risk model cannot explain the cross section of returns this does not mean that the relation between risk and return is rejected but just that the model is not specified in the right way. The answer to the question which explanation of the value premium is the correct one, is not the topic of the rest of this study. The next subsection will give some recent findings on the anatomy of value- and growth portfolios.

IId. Migration of value- and growth stocks

Recently the dynamics of the returns on value- and growth portfolios has become subject of research. In a recent paper, Fama and French (2007) argued that the value premium can be separated in three factors. First, value stocks migrate to a neutral or growth portfolio because they earn high returns or become a target for acquisition. Second, growth stocks move from a growth- to a neutral portfolio or a value portfolio. Third, value stocks that do not migrate earn a slightly higher return than growth stocks that do not migrate. In this paper the authors constructed 6 portfolios based on size and price to book value (P/B)4 resulting in a Small Growth (SG), Small Neutral (SN), Small Value (SV), Big Growth (BG), Big Neutral (BN) and a Big Value (BV) portfolio. Migration to a lower P/B portfolio (from growth to value) was associated with large negative returns with a maximum of -36% for small stocks and -13,6% for large stocks per annum. Migration to a higher P/B portfolio on the other hand was associated with large positive returns with a maximum of 50,3% for small stocks and 29,7% for large stocks. The separation between large and small stocks was based on the NYSE median and the distinction between value and growth on the top and bottom 30% of the P/B spectrum. Fama and French (2007) argued that this can be in line with the risk argument from early papers published. The value premium then exists because investors do not know what is happening with the value- and growth stocks in the period after formation, and the value premium is the extra return for this uncertainty. On the other hand one can argue from a behavioral point of view that investors underestimate that value stocks will improve to growth stocks and growth stocks will deteriorate.

This paper extended the Fama and French (2005) paper. In this paper it was already concluded that migration of value- to growth stocks and the other way around had an important role in the

4

value premium. Fama and French (2005) argued that when value- and growth portfolios are created, the firms are on the opposite site of the profitability spectrum. The profitability, of a part of the growth stock companies, attracts competition which declines profitability. Some value stock firms on the other hand restructure and improve their profitability. This change in profitability goes in hand with the migration of the companies stocks from the value portfolio to the neutral or growth portfolio, or from the growth portfolio to the neutral or value portfolio. This can be an important implication for the cross section of average stock returns.

III. Hypotheses and methodology

The previous section of this thesis discussed the empirical evidence on the existence and explanation of the value premium until today. In this section the hypotheses and methodology of this thesis will be explained.

As a start, the existence of the value premium in the Dutch Stock Market will be researched for the period June 1st, 1981 and May 31st, 2007. In order to do this, portfolios based on the CF/P ratio will be created. I will use the CF/P ratio because earlier in this thesis I argued that this is a better fundamental to classify stocks into value- and growth portfolios. The CF/P ratios are determined for 1980 until 2001 at the end of each year. In each of these years, value-, neutral- and growth portfolios will be created, leading to 22 portfolio formation years. The year of portfolio formation will be defined as period t-1.

The top 30% stocks on the CF/P spectrum are labeled value stocks and the lowest 30% growth stocks. The middle 40% is the neutral portfolio. If the percentages of the total amount of stocks do not lead to a round number of stocks, the ‘rest stocks’ are labeled neutral so the value- and growth portfolio are always equal.

For each portfolio, the returns from one until five years after formation will be analyzed. Because I want to be sure that the accounting data is known before portfolio formation the accounting data from year t-1 is matched with the returns in June of year t. This 5 month gap is because companies listed on the Dutch Stock Market have 5 months to present their results after closing their fiscal year. In approximately 9% of the sample the end of the fiscal year is not December the 31st. Because in almost half of this group the fiscal year ends one month later or one month earlier the effect on the outcomes is small. When the fiscal year ends within six months from December the 31st the stock is included in the portfolio from that period. When the fiscal year ends more than 6 months later or earlier it is included in the portfolio from respectively the next or the previous year.

The annual return from June 1st, in year t until May 31st, in year t+1 will be defined as ‘one year after formation’. The returns from June 1st, in year t+1 until May 31st, in year t+2 are defined as ‘two years after formation’. This concept will continue until May 31st of year t+5. In each year, the portfolio return is the equally weighted average of all the stock returns in that year. The total return on value- and growth investing is the equally weighted average of all years, this holds from one until five years after formation. In figure 1 the analysis is summarized in a time line.

Figure 1. Time line of portfolio return analysis

Because portfolios are created in each year, and returns are tested until five years after portfolio formation, this methodology will lead to overlapping periods. The returns, for example, four and three years after formation from the value- and growth portfolios created in respectively 1982 and 1983 are both between June 1st, 1986 and May 31st, 1987.

The significance of the differences in value- and growth investing is tested with a one sided t-test. This is in accordance with important studies like Fama and French (1998) and Lakonishok, Shleifer and Vishny (1994).

In order to test the existence of the value premium in the Dutch Stock Market the following hypothesis is defined:

1a. H0: Value stocks do not earn, on average, higher returns than growth stocks in the Dutch

Stock Market between June 1st, 1981 and May 31st, 2007

Because of the outcomes of prior research it is expected that value stocks earn higher returns than growth stocks. The hypothesis is therefore tested against the one-tailed alternative hypothesis:

H1: Value stocks earn, on average, higher returns than growth stock in the Dutch Stock

Market between June 1st, 1981 and May 31st, 2007

Because this research is performed on the whole Dutch Stock Market there are large differences in size between the companies. Prior research provided evidence for the hypothesis that firms with a small market capitalization earn higher returns than firms with a large market capitalization. Because of this small firm effect (Banz, 1981), the portfolios will be corrected for size. In this way I can be sure that the results are attributable to the value premium and not to the small firm effect.

value- or growth stocks. This is in accordance with Wouters (2006). The size benchmark portfolios are constructed based on the relative size of the firms to the other firms in that year and the stocks of these firms are also assigned to a large, medium and small category. The average return of all the stocks in a size benchmark category will be the size benchmark return. Each stock return is subtracted by the size benchmark return from its corresponding category in that period to get the size adjusted returns.

The value premium with size adjusted returns is tested with the following hypothesis:

1b. H0: Value stocks do not earn, on average, higher size adjusted returns than growth

stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2007

The hypothesis is tested against the one-tailed alternative hypothesis:

H1: Value stocks earn, on average, higher size adjusted returns than growth stock in the

Dutch Stock Market between June 1st, 1981 and May 31st, 2007

After presenting the returns on value and growth investing the migration of value- and growth stocks will be investigated.

In the first place I will test if the migration of value- and growth stocks is largely responsible for the value premium. From prior studies it appeared that switching style value stocks earn much higher returns than switching style growth stocks. In order to test this for the Dutch Stock Market the following hypothesis is defined:

2a. H0: Switching style value stocks do not earn, on average, higher returns than switching

style growth stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2006

This hypothesis is tested against the one-tailed alternative hypothesis:

H1: Switching style value stocks earn, on average, higher returns than switching style

growth stock in the Dutch Stock Market between June 1st, 1981 and May 31st, 2006

Prior studies show that the switching behavior of value- and growth stocks is largely responsible for the value premium and further that the value premium is quite smaller when only fixed style stocks are included in the research. Therefore, I will compare the returns of fixed style value stocks and fixed style growth stocks. The following hypothesis is defined:

2b. H0: Fixed style value stocks do not earn, on average, higher returns than fixed style

growth stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2006

Because from prior research we expect that fixed style value stocks earn higher returns than fixed style growth stocks, the hypothesis is tested against the one-tailed alternative hypothesis:

H1: Fixed style value stocks earn, on average, higher returns than fixed style growth

IV. Data

All data in this thesis have been obtained from DataStream. In order to compare the returns on value- and growth investing in the Dutch Stock Market, based on the methodology explained in the last section, the following data is required:

• Year-end CF/P ratios from 1980 until 2001. • Annual returns from June 1st

, 1981 until May 31st 2007. • May 31st

company market values from 1981 until 2007.

Cash Flow is defined as earnings plus depreciation. This is in line with Chan and Lakonishok (2004) and Fama and French (1998). For earnings, DataStream company accounts item 625 (earned for ordinary) is used. Depreciation is company accounts item 136 (Depreciation). Price is the year-end market value.

Because this thesis will compare the returns of value- and growth portfolios, stock returns between June 1st, 1981 and May 31st ,2007 are required. For stock returns, the stocks’ total returns will be used. Total returns are returns with reinvested dividends, which make it possible to compare the total stock performance. The rationale behind this, is that the stock price will decline when companies pay dividends. As a consequence, a part of the stock performance is excluded from the results when returns without reinvested dividends are compared.

In total, the sample period covers returns from June 1st, 1981 until May 31st, 2007, which is 27 years. This is a sufficient time frame compared with the studies of Lakonishok, Shleifer and Vishny (1994), Fama and French (1998) and Wouters (2006) with respectively 22, 21 and 27 years.

Stocks which have no data for respectively earnings, depreciation and year-end market value are omitted from the research in that year. Companies with less than 24 months of return data are also omitted from the portfolio in that year. In total the sample exists of 280 firms and 2533 CF/P ratios between 1980 and 2001.

Table 2 summarizes the descriptive statistics of the sample categories

Table 2. Descriptive statistics of sample categories.

Total Average per year Value stocks 761 35

Neutral stocks5 1011 46 Growth stocks 761 35

Large size benchmark portfolio 333 15

Medium size benchmark portfolio 345 16 Small size benchmark portfolio 333 15

5

V. Results

In this section the results of the thesis are given and discussed. As a starting point, I will present the outcomes on value- and growth investing in the Dutch Stock Market from June 1st, 1980 until May 31st, 2007. The differences in returns will be discussed in comparison with prior research. Because of the possible influence of the small firm effect, size adjusted returns will be presented and discussed. After this, migration frequencies are given and the influence on the value premium is reviewed. In the end of this section I can conclude if the value premium exists for fixed style value stocks and fixed style growth stocks in the Dutch Stock Market.

Va. The value premium

Table 3 present the returns on value- and growth investing in the Dutch Stock Market from June 1st, 1981 until May 31st, 2007.

Table 3. Average annual returns on value- and growth investing in the Dutch Stock Market from June 1st, 1981 until May 31st, 2007.

Portfolio/Years after formation +1 +2 +3 +4 +5

Value 0,18 0,20 0,22 0,23 0,22 Growth 0,13 0,16 0,14 0,17 0,16 Difference Value-Growth 0,05 0,04 0,08 0,06 0,07

T-statistic 1,28 1,05 1,99** 1,57* 1,65*

*Significant on 10% level with critical value 1,323 **Significant on 5% level with critical value 1,721

From table 3 it appears that the value portfolio gains higher returns in all of the researched time periods. One year after portfolio formation the returns are 5% higher and from three until five years after formation the differences are even larger. In comparison with other studies the differences between the value- and growth portfolio are smaller. Lakonishok, Shleifer and Vishny (1994) for example found an outperformance of the value- over the growth portfolio of 10,5% in the first year. La Porta (1996) found an even larger difference with 20% in the first year. Wouters (2006) found outperformance of 10,5%, 10% and 8% in respectively 1, 2 and 3 years after the formation of value- and growth portfolios.

observations in the portfolios. Because the limited number of companies which are listed on the Dutch Stock Market there are on average 35 stocks in the value- and growth portfolio, and a few outliers in the returns can have a high impact on the volatility of the returns and therefore on the standard deviation.

Another possible explanation for the high standard deviation is the structure of the sample period. The sample period consists of periods with increasing and decreasing stock returns. For example, in the period 2000-2002 stock returns decreased sharply and from 2002-2007 stock returns increased steady. It is not inconceivable that the value premium is different for periods with increasing and decreasing stock returns. In order to test if the value premium is different for subperiods of the sample I will check the robustness of the outcomes for two subperiods. The sample will be separated in returns from June 1st, 1981 until May 31st, 1994 and June 1st, 1994 until May 31st, 2007. Table 4 and table 5 provide the returns on value- and growth investing in the subperiods.

Table 4. Average annual returns on value- and growth investing in the Dutch Stock Market from June 1st, 1981 until May 31st, 1994.

Portfolio/Years after formation +1 +2 +3 +4 +5

Value 0,24 0,25 0,26 0,29 0,28 Growth 0,19 0,20 0,15 0,17 0,16 Difference Value Growth 0,06 0,05 0,11 0,13 0,12

T-statistic 1,25 0,81 1,69* 2,01** 1,78*

*Significant on 10% level **Significant on 5% level

Table 5. Average annual returns on Value and Growth investing in the Dutch Stock Market from June 1st, 1994 until May 31st, 2007.

Portfolio/Years after formation +1 +2 +3 +4 +5

Value 0,08 0,14 0,17 0,18 0,18 Growth 0,05 0,10 0,13 0,17 0,15 Difference Value Growth 0,04 0,04 0,04 0,01 0,03

From table 4 and 5 can be concluded that there are some differences in the value premium between the two periods. The value premium from June 1st, 1981 until May 31st, 1994 is in all years after portfolio formation larger than the value premium from June 1st, 1994 until May 31st, 2007. In the first subperiod, the differences are significant on the 5% level four years after portfolio formation and on the 10% level for three and five years after portfolio formation. This is comparable with the results on the whole sample period. In the second subperiod the results are in all years after portfolio formation not significant. It is possible that this is the consequence of the volatile stocks market between 2000 and 2007. However, to draw conclusions further research is required. From the subperiod analysis can therefore be concluded that the significance of the value premium is not improved in the tested subperiods.

Summarizing can be concluded that, based on the analysis in section Va, hypothesis 1a that value stocks do not earn, on average, higher returns than growth stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2007 cannot be rejected.

Vb. Size adjusted returns

As described earlier it is possible that the results are influenced by the small firm effect. In order to test this table 6 provides the size adjusted returns.

Table 6. Average annual size adjusted returns on value- and growth investing in the Dutch Stock Market from June 1st, 1981 until May 31st, 2007.

Portfolio/Years after formation +1 +2 +3 +4 +5

Value -0,02 -0,03 0,05 0,09 0,08 Growth -0,03 -0,03 -0,02 0,03 0,03

Difference Value Growth 0,01 0,00 0,06 0,05 0,05

T-statistic 0,27 0,06 1,62* 1,30 1,32

*Significant on 10% level with critical value 1,323

size adjusted and not size adjusted. This result was not surprising because there were no big difference in size between the value and the growth portfolio in his study.

Concluding can be stated that the evidence on the value premium is even weaker as with size unadjusted returns. Hypothesis 1b that value stocks do not earn, on average, higher size adjusted returns than growth stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2007 cannot be rejected.

Vc. Migration

This subsection will provide information on the migration process and ultimately answer hypotheses 2a and 2b which are stated in section III. Firstly, table 7 provides some information on the switching behavior of stocks in the Dutch Stock Market between 1980 and 2001. It gives the relative frequencies of stocks moving from category i in year t to category j in year t+1.

Table 7. Migration frequencies between value- and growth portfolios in the Dutch Stock Market between 1980 and 2001.

Category j in year t+1

Category i in year t Growth Neutral Value

Growth 0,715 0,244 0,041

Neutral 0,142 0,695 0,162 Value 0,048 0,169 0,783

From table 7 it appears that there are some changes year-to-year in the composition of the different portfolios. In the year after formation approximately 29% of the growth stocks disappear from the growth portfolio and 22% of the value stocks disappear from the value portfolio. Most of these stocks migrate to the neutral portfolio but a small percentage of 4,8% and 4,1% even migrate from value- to growth stock or from growth- to value stock.

Table 8. Average annual returns of stocks moving from category i in year t to category j in year t+1 in the Dutch Stock Market between 1980 and 2001.

Category j in year t+1

Category i in year t Growth Neutral Value

Growth 0,115 0,033 0,435

Neutral 0,258 0,179 0,016

Value -0,156 0,217 0,176

As appears from table 8, growth stocks that migrate to value stocks earn 59,1% higher returns during the transition than value stocks which migrate to growth stocks. However the significance of this result is not tested, this is conspicuous because prior research concluded that a large part of the value premium is the result of value stocks migrating to growth stocks and the other way around (Fama and French, 2007 and Wouters, 2006). The returns in table 8 are contrarian with this.

Why these stocks, which migrate from growth- to value stock, earn such large returns is not straightforward. An explanation is that these are distressed stocks because the negative earnings in the formation year. A result of this is that investors can be afraid of bankruptcy which is reflected in the stock price. When the companies earnings are strongly improved in the next year (causing the transition to the value portfolio) investors change expectations and the stock price is recovering, causing large returns. Another remarkable finding is that neutral stocks which stay neutral stocks earn higher returns than value stocks which stay value stocks. It is expected that value stocks which stay value stocks earn the highest returns. However, in the paper of Wouters (2006) the same outcome was found.

In order to analyze switching behavior more extensively I make a distinction between switching style stocks and fixed style stocks. Stocks which stay at least two consecutive years in the same portfolio are labeled fixed style stocks, and stocks which are migrating after one year are labeled switching style stocks.

Table 9 provides the returns of switching style- and fixed style stocks for one until five years after portfolio formation. The returns are calculated in the same way as the value- and growth portfolios from section Va. From other studies it is expected that switching style value stocks earn higher returns than switching style growth stocks and that fixed style value stocks earn slightly higher returns than fixed style growth stocks.

Table 9. Returns for switching style and fixed style stocks in the Dutch Stock market between June 1st, 1981 and May 31st, 2006

Value Growth Difference Years

after formation

Switching Fixed Switching Fixed Switching T-statistic Fixed T-statistic

1 0,211 0,202 0,161 0,149 0,051 0,762 0,053 1,035 2 0,073 0,199 0,167 0,133 -0,094 -1,526 0,067 1,429* 3 0,241 0,189 0,194 0,114 0,047 0,796 0,075 1,444* 4 0,171 0,213 0,228 0,138 -0,057 -0,743 0,075 1,513* 5 0,174 0,233 0,152 0,157 0,022 0,314 0,076 1,710* *Significant on 10% level

The second year after formation, switching style growth stocks earn even higher return than switching style value stocks. In year three until year five after formation the results are mixed. In all years after formation the results are not significant. It is clear that these outcomes are no evidence that switching style value stocks earn higher returns than switching style growth stocks. The hypothesis that switching style value stocks do not earn, on average, higher returns than switching style growth stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2006 can therefore not be rejected.

A possible improvement is omitting negative earnings from the sample. Because these are largely responsible for the returns accompanied with the switching behaviour this can lead to different results. Lakonishok, Shleifer and Vishny (1994) for example omit firms with negative earnings. They argued that this is justified because including firms with negative earnings is not a test of the original value- and growth problem. This is because the initial value premium problem implies future growth for growth stocks but this is not the case with stocks of firms with negative earnings. In my sample there are 29 stocks which migrate from growth stock to value stocks and 34 which migrate from value stock to growth stock. Because the vast majority of the stocks which migrate form the growth- to the value portfolio are from firms with negative earnings in the year before migration, it is clear that omitting firms with negative earnings is not possible in this thesis.

Another possibility to improve the results is including amortization in the definition of Cash Flow because this will have a positive impact on the CF/P ratio. It is possible that these negative earnings firms have negative earnings because of high amortization levels.

The returns differences between fixed style value stocks and fixed style growth stocks are more convincing. In every year after formation, fixed style value stocks perform at least 5,3% better than fixed style growth stocks. In all years, expect year one, after formation these results are significant on the 10% level. Because this is not a very high significance level I have to be careful with conclusions. Hypothesis 2b that fixed style value stocks do not earn, on average, higher returns than fixed style growth stocks in the Dutch Stock Market between June 1st, 1981 and May 31st, 2006 can therefore not be rejected.

VI. Summary and conclusion

This thesis researched the value anomaly. The current situation in the value premium debate has been discussed in the light of prior research. Furthermore, the influence of the value premium on the Dutch Stock Market between June 1st, 1981 and May 31st, 2007 was discussed. Based on the CF/P ratio, growth- and value portfolios were constructed. From one until five years after portfolio formation the value premium was given with size unadjusted as well as size adjusted returns. After presenting and discussing the value premium, the influence of migration of stocks between different portfolios was tested. There has been made a distinction between fixed style stocks, which are stocks that stay at least two consecutive years in the same portfolio, and switching style stocks, which migrate to a different portfolio after one year.

From this thesis, it cannot be concluded that there is a value premium in the Dutch Stock Market between June 1st, 1981 until May 31st, 2007. In all tested periods after formation, the difference between the value- and growth portfolio is at least 4% but this is not backed by convincing statistical significance. When correcting for size the value premium decreases but is still visible. In year one and three after formation the premium of holding value- over growth stocks is 4%. Only in the second year after formation there is no difference in returns between the value- and the growth portfolio. These results are however not significant in all years after formation. The premium on holding value stocks instead of growth stocks is, although the value premium is lower, in line with prior research from for example Lakonishok, Shleifer and Vishny (1994), Fama and French (1998) and La Porta (1996).

The answer on the central question is clear. About the Dutch Stock Market between June 1st, 1981 and May 31st, 2007 cannot be stated that the value premium exists and further no conclusion can be drawn on the effect of the migration of stocks on the differences in returns between value- and growth stocks.

Limitations of this thesis are in the first place, the limited number of stocks in the portfolio samples. This is the consequence of performing a research on the Dutch Stock Market at which a limited number of companies are listed on. Because of this, there are on average 35 stocks in the value- and growth portfolio. This is a possible explanation for the very high standard deviations which lowers the significance of the results. A possible solution for this high standard deviation is performing the research with monthly returns instead of yearly returns. In this way there will be 12 times more data points in the research which has a positive effect on the standard deviation. On the other hand the effect on the volatility of the returns is uncertain. Therefore it is not possible to state, without research, that the significance of the results will be improved with monthly returns. Performing this research with monthly returns is however not possible due to time constraints. Another possible improvement of the significance of the results is changing the research period. From the subperiod analysis appeared that the value premium was a lot lower between June 1st, 1994 until May 31st, 2007 than between June 1st, 1981 until May 31st, 1994. This is possibly the consequence of the more volatile stock market in the first period. More research on the structure of the sample period can possibly improve the significance of the results. More a possible improvement than a limitation is changing the definition of the CF/P ratio. Because Cash Flow is defined, as in other papers (e.g. Chan and Lakonishok (2004), Fama and French (1998)), as earnings plus depreciation, amortization is omitted from the definition. The high amortization levels in especially the early 2000s could have a significant impact on the CF/P ratio if included. It is possible that a CF/P definition including amortization partly solved the negative earnings problem in the section on migration of stocks.

VII. References

Basu, S., 1977, Investment performance of common stocks in relation to their price-earnings ratios: a test of the efficient market hypothesis, Journal of Finance 32, 663-682.

Banz, R.W., 1981, the relationship of return and market value of common stocks, Journal of

Financial Economics 9, 3-18.

Bernard, V., and T. Stober, 1989, The nature and amount of information in cash flows and accruals, Accounting Review 64, 624-652.

Chan, L.K.C., and N. Chen (1991), Structural and return characteristics of small and large firms,

Journal of Finance 46, 1467-1484.

Chan, L.K.C., Y. Hamao and J. Lakonishok, 1991, Fundamentals and stock returns in Japan,

Journal of Finance, 46, 1739-1764.

Chan, L.K.C., and J. Lakonishok, 2004, Value and Growth Investing: Review and Update,

Financial Analyst Journal 60, 71-86.

De Bondt, W.F.M., and R.M. Thaler 1985, Does the stock market overreact? Journal of Finance 40, 793-805.

Doukas, J.A., C. Kim, and C. Pantzalis, 2002, A test of the errors-in-expectations explanation of the value/glamour stock returns performance: evidence from analysts’ forecasts, Journal of

Finance, 57, 2143-2165.

Fama, E.F., 1970, Efficient capital markets: a review of theory and empirical work, Journal of

Finance 25, 383-417.

Fama, E.F., and K.R. French, 1992, The cross-section of expected stock returns, Journal of

Finance 57, 427-465.

Fama, E.F., and K.R. French, 1993, Common risk factors in the returns of stocks and bonds,

Journal of Financial Economics 33, 3-56.

Fama, E.F., and K.R. French, 1995, Size and Book-to-Market Factors in Earnings and Returns,

Journal of Finance 50, 131-155

Fama, E.F., and K.R. French, 1996, Multifactor explanations of asset pricing anomalies, Journal

of Finance 51, 55-84.

Fama, E.F., and K.R. French, 1998, Value versus growth: The international evidence, Journal of

Finance 53, 1975-1999.

Fama, E.F., and K.R. French, 2005, The Anatomy of Value and Growth Stock Returns,

University of Chicago, Working Paper.

Gordon, M., and E. Shapiro, 1956, Capital equipment analysis: the required rate of profit,

Management Science 3, 102-110.

Graham, B., and D. Dodd, 1934, Security analysis, McGraw Hill Book Company Inc., New York. Kahneman, D., and A. Tversky, 1973, On the psychology of prediction, Psychological Review 80, 237-251.

Lakonishok, J., A. Shleifer, and R.W. Vishny, 1994, Contrarian investment, extrapolation, and risk, Journal of Finance 49, 1541-1578.

La Porta, R., 1996, Expectations and cross-section of stock returns, Journal of Finance 51, 1715-1742.

La Porta, R., J. Lakonishok, A. Shleifer, and R.W. Vishny, 1997, Good news for value stocks: further evidence on market efficiency, Journal of Finance 52, 859-87.

Lintner, J., 1965, The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets, Review of Economics and Statistics 17, 13-37.

Reinganum, M., 1981, Mis-specification of capital asset pricing: Empirical anomalies based on earnings yields and market values, Journal of Financial Economics 9, 19-46.

Rosenberg, B., K. Reid and R. Lanstein, 1985, Persuasive evidence of market inefficiency,

Journal of Portfolio Management 11, 9-17.

Sharpe, W.F., 1964, Capital asset prices: A theory of market equilibrium under conditions of risk,

Journal of Finance 19, 425-442.

Wouters, T.I.M., 2006, Style Investing: Behavioral Explanations of Stock Market Anomalies,