December 2014
University of Groningen
Faculty of Economics and Business
Have a free lunch in Budapest or Bratislava:
The value premium in the post-communist countries
of the European Union
Master Thesis
Student Nr: S2625822
Name: Roman Uhlir
Have a free lunch in Budapest or Bratislava:
The value premium in the post-communist countries
of the European Union
Abstract: This paper finds a value premium ranging between ten and twenty
percentage points of additional annual investment returns in the stock markets of post-communist Europe in 2005-2013. Its overall magnitude in this period is higher than in other European markets. However, this result is driven by abnormally high performance of Eastern European stocks before the financial crisis. Further, this article challenges the risk-based explanation of the value premium both on argumentative and empirical level. This paper explains the excess returns with behavioral patterns and provides evidence for the behavioral explanation on the analysis down-market returns, standard deviation and CAPM alphas. Thus, value investments are apparently close to a ‘free lunch’ in the stock market.
1. Introduction
This paper examines the magnitude of the value premium in stock markets of the post-communist countries of the European Union (also labeled “the East of the EU”, “new EU countries”, “post-communist Europe” or simply “Eastern Europe”) both in absolute terms as well as in comparison with other European markets. Further, I bring an empirical evidence to support the hypothesis that the value premium can be explained by behavior rather than by risk and thus comes close to a “free lunch”. The value premium, in the sense of e.g. Fama and French (1998), stands for a phenomenon of so called value stocks tending to have higher average returns compared to so called growth stocks. Value stocks are considered those ones with “low prices relative to earnings, dividends, historical prices, book assets or other measure of value” (Lakonishok et al. 1994). On the other hand, to relatively expensive stocks with, often with high future growth expectations one refers as to growth stocks (sometimes also called “glamour stocks”). The most popular criterion of distinguishing between the two groups of equities in the literature to date is the Book-to-Market ratio, followed by Earnings-to-Price ratio and Cash-Flow-to-Price ratio.
Warren Buffett’s Berkshire Hathaway (Buffett and Clark, 2007). The value premium research brings possible scientific explanation for the success of these strategies. The value indicators such as Book-to-Market and Earnings-to-Price for any exchange-listed firm are easily available to the broad public and identifying value stocks based on these indicators does not require any sophisticated skills. Thus, the conclusions of the value premium research seem to show a direction for simple and evidently market-outperforming strategies for any investor.1
Most of the research to date focuses on the most developed markets of North America, Western Europe or Japan. In all these geographies, the existence of the value premium has been established. However, there are differences in the magnitude and persistence of the value premium. This paper aims to broaden the geographic scope of the value premium research to the Eastern European markets. This article shows that the value premium is also present in post-communist Europe and compares the magnitude and persistence of the value premium in Eastern Europe to other markets in the rest of the EU. The most debated issue in the literature to the topic is whether the value premium is driven by behavioral patterns, i.e. irrationality of market participants or risk. I understand risk as the “possibility of loss”. The major focus of this article is the explanation of the value premium in post-communist Europe. My paper builds on the foundations by Lakonishok et al. (1994) in two ways. First, I argue for the theoretical behavioral argumentation of Lakonishok et al. (1994) clashing with the critique of Fama and French (1996 and 1998). Second, I bring an empirical evidence that confirms the behavioral value premium explanation of Lakonishok et al. (1994) and rejects the risk-only-based explanation of Fama and French (1996 and 1998). Similarly to e.g. van der Hart et al. (2003 and 2005), this article belongs to the body of more recent literature that further develops the research of Lakonishok et al. (1994). This body of literature represents an alternative to newer articles of Fama and French (2006a and 2012). This paper uses methods that are in line with both streams of the literature.
The research aim can be expressed as a set of three hypotheses: 1. H0: Value premium in post-communist Europe = 0;
H1: Value premium in post-communist Europe ≠ 0
The first hypothesis tests whether value stocks in post-communist Europe bring higher returns than other stocks. At this stage, the riskiness is not analyzed as the main focus is to identify and quantify the value premium.
2. H0: Value premium in post-communist Europe ≤ Value premium in the rest of the EU;
H1: Value premium in post-communist Europe > Value premium in the rest of the EU Tests for the second hypothesis compare the value premium in post-communist Europe with other EU countries. Both magnitude and the persistence of the value premium are compared.
3. H0: Value premium in post-communist Europe can be explained by behavioral patterns;
H1: Value premium in post-communist Europe can be explained by risk.
The third hypothesis attempts to explain the value premium in post-communist Europe. Analyses of the down-markets, CAPM betas and standard deviations of returns are used to measure the riskiness of stocks. The riskiness of value stock portfolios is then compared with other portfolios.
The paper is structured as follows: the next section provides a review of the literature to date, the third section describes the data, the fourth section explains the methodology and presents the results. Finally, the fifth section concludes.
2. Literature review
The literature dealing with value premium can be distinguished in three major groups: First, papers that only aim to find a value premium without attempting to explain it (e.g. Chan et al. 1991, Barry et al., 2002), second literature that explains the value premium with higher risk of the value investments (e.g. Fama and French 1996 and 1998) and third research claiming that behavioral aspects such as expectational errors based on over-extrapolation of past returns (started by Lakonishok et al. 1994). The paper of Chan and Lakonishok (2004) presents a broader review of literature regarding the value premium for further reference.
sub-section 2.3, I present the baseline paper of Lakonishok et al. (1994) and its behavior-based value premium explanation. Sub-section 2.4 approaches the clash in the literature about whether the value premium can be explained by risk or behavior. In sub-section 2.5 I introduce the CAPM and the three-factor model in the value premium context and in 2.6, I examine critically the arguments of Fama and French (1996 and 1998) used in the debate with Lakonishok et al. (1994) as well as more recent work of Fama and French.
2.1 The computation of the value premium
Table 1.1 presents compactly 11 relevant articles, providing information about the methodologies (columns ‘Value-Growth distinguisher’, ‘Basis for portfolios’ and ‘Structure of portfolios’), the data (columns ‘Time period’, ‘Geography’ and ‘Number of firms’) as well as the results and conclusions (columns ‘Val. prem. magnitude’ and ‘Val. prem. explanation’). In the value premium literature some methods have been established as standards used by every researcher.
Notably, portfolios are used instead of single stocks. The universal indicator to distinguish the value and growth stocks is the Book-to-Market value (or its reverse value Price-to-Book ratio). This measure is usually used parallelly with the E/P (or P/E) ratio and/or with other indicators capturing dividends, cash flow or past growth in sales. In addition, since a size premium has been established as a robust pattern favoring rather smaller stocks (Chan and Lakonishok, 2004), it is necessary to control for its effects to capture the value premium. This control takes place either in form of including market capitalization in the regression analysis directly or splitting the stocks into size-based portfolios. The portfolios are always based on quantiles or certain percentiles in one or two dimensions: value and mostly also size. Often authors use univariate analysis: they compare the performance of the value portfolios with the growth portfolios and/or the benchmark such as the stock market index in given country. Regression analysis is used to estimate the parameters of capital asset pricing models. Transaction costs resulting from portfolio rebalancing are not considered in any of the articles.
multi-year average. However, as e.g. Fama and French (1998, p. 1980, Table III) and Chan and Lakonishok (2004, p. 82-83, Table 8) show, value premia fluctuate and do not appear every year in every market.
2.2 Research on European post-communist countries
To my knowledge, there is so far no paper that would analyze the value premium exclusively in the Eastern European markets. Lischewski and Voronkova (2011) research the value premium in the Polish market which is by far the largest one in the region of interest. Van der Hart (2003 and 2005) and Barry et al. (2002) examine over 30 emerging markets worldwide. In their studies, the Eastern European countries Czech Republic, Hungary, Poland and Slovakia are covered too. Notably, the region has not been covered in the internationally oriented paper of Fama and French (1998) focusing on emerging markets outside of Europe neither in Fama and French (2012) focusing rather on developed markets.
2.3 Baseline paper: Lakonishok, Schleifer and Vishny (1994)
Lakonishok et al. (1994) define value stocks as stocks with low prices relative to a value measure such as earnings, dividends, historical prices or book assets. They find a consensus in the literature that the value premium is existent and well established across the globe. What differs are the explanations why it exists.
Value strategies i.e. strategies that exploit the value premium are contrarian to so called ‘naive strategies’ or ‘popular models’. Naive strategies are to smaller or larger extent irrational: naive investors overextrapolate past performance assuming a trend or overreact to good or bad news. Lakonishok et al. (1994) mention also other examples of naive irrational behavior: Investors are equating a well run company with a good stock purchase ignoring the price paid. As naive investors get excited about glamour companies, glamour stocks get overpriced because of excess buying. Contrary to the so called popular models, the value strategies invest in
underpriced stocks. This contrarian model is based on the assumption that the past growth of
glamour stocks is overextrapolated to the future and thus the expected returns are too high. The value strategy exploits these expectational errors.
Lakonishok et al. (1994) ignite the debate in the literature with their reference to Fama and French (1992) who state that value stocks are fundamentally riskier which explains the higher returns.
Lakonishok et al. (1994) use the Earnings-to-Price, Book-to-Market and the Cash-Flow-to-Price ratio as measure of future growth expectations. They delete stocks which have negative values for these indicators from their sample. Besides the future growth expectations, the authors also consider past growth to distinguish between glamour (growth) and value stocks. Only stocks traded in the USA are considered, the time frame is 1963 - 1990.
others are expected to have a rationally justifiable low Book-to-Market ratio as far as the growth prospects are not overrated.
The portfolios are built based on market capitalization deciles. All stocks in these portfolios are equally weighted. The portfolios are built based on deciles according to the three ratio measures of stock value. In addition, they use the growth in sales as a further dimension of distinguishing between value and growth stocks. The assumption behind it is that naive investors will overextrapolate the trends in sales to the future after the portfolio formation. Thus stocks with low past growth in sales are considered value stocks according to this indicator. The portfolio returns are size adjusted, i.e. adjusted for the returns of the size decile the given stock belongs to.
They prove that the value strategies beat the market in the long run. The two-dimensional strategy investing in stocks that are identified as value stocks both by the past growth in sales and by one of the three ratios perform even better than the one-dimensional strategies based on only one indicator. The analysis is performed on comparison of nine portfolios: the stocks are split into top 30%, middle 40% and bottom 30% in each dimension.
The results confirm the hypothesis that investors are too optimistic about the growth stocks and too pessimistic about the value stocks. The sales growth momentum was proven in a short term horizon of 1 to 2 years only, though not in the long term horizon of 3 to 5 years.
The research shows a significant value premium resulting from various types of portfolio performance analysis. In a 5 year horizon, the value stocks outperform glamour stocks by about 90 percent. Besides the value effect, the size premium is confirmed as well.
Lakonishok et al. (1994) also look for the explanation of the excessive returns of value stocks and test whether these investments are riskier. The authors bring following view on risk: “To be fundamentally riskier, value stocks must underperform glamour stocks with some frequency, and particularly in the states of the world when the marginal utility of wealth is high” (Lakonishok et al., 1994, p.1543).
risk, the difference between value and growth portfolios in the CAPM betas is neglectable. The standard deviation of the returns was higher for value stocks, though Lakonishok et al. (1994) do not see the difference in this criterion to be large enough to fully capture the superior returns of the value strategies. In addition, as the authors state, larger standard deviation does not necessarily mean higher downside risk.
2.4 The value premium explanation: Risk or Behavior?
In the debate about whether value premium can be explained by risk or irrational behavior of the investors in the markets, van der Hart et al. (2003 and 2005), Athanassakos (2009), Chan and Lakonishok (2004) and Lakonishok et al. (1994) represent the side of the behavioral explanation.
Fama and French (1992, 1996, 1998 and 2006a) also find the value premium. However, they explain it with higher risk as calculated in various pricing models. Fama and French (1998 and 2006a) find nevertheless that simple CAPM cannot explain the value premium in the time period 1963 - 2004. Fama and French (1998) conclude that value premium can be captured by the Arbitrage Pricing Theory (APT) of Ross (1976) and ICAPM of Merton (1973). Remarkably, from the group of authors researching this topic it is only Fama and French who defend the risk-only explanation of the value premium. The next two subsections shed some more light on Fama and French’s reasoning.
2.5 Asset Pricing models
Fama and French (1996 and 1998) label the value premium as one of “anomalies” in the stock market that cannot be explained by the traditional CAPM of Fama and MacBeth (1973). In the paper of 1998 (p.1984) it is the two factor pricing model in line with ICAPM and APT that explains the portfolio returns in excess of the risk-free rate with global markets return in excess of the risk free rate and the global value premium:
R - F = α + b[M - F] + c[H - LB/M] + e (1)
In the equation (1), R is the portfolio return, F is the risk-free rate, M is the return of the market portfolio, H - LB/M is the global value premium (standing for High minus Low Book-to-Market), and e is the regression error term.
E(Ri) - Rf = bi[E(RM) - Rf] + siE(SMB) + hiE(HML) (2) The three factor model is in fact the two factor ICAPM (1) extended by the size premium. The notation in the two papers of Fama and French differs slightly, so that the value premium is expressed as HML (high-minus-low). The size premium is noted as SMB (small-minus-big). HML are calculated by subtracting the average returns for the lowest value portfolios from the returns of the highest value portfolio. Analogously, the average returns of the highest market capitalization portfolio are subtracted from the returns of the lowest market capitalization portfolio to calculate SMB. Both are computed for every year considered in the analysis. The coefficients si and hi explain to what extent the returns of a given portfolio are correlated with the size and value premium for that given year.
Fama and French find empirically that these models can successfully explain large portion of excess returns in the stock markets. Size and value are presented as another risk dimensions.
2.6 Critique of Fama and French’s (1996 and 1998) approach
Lakonishok et al. (1994) challenge the argumentation and the risk-based value premium explanation of Fama and French (1992). Fama and French (1996 and 1998) approach the argumentation of Lakonishok et al. (1994) and attempt to refute it. Further on, most of the literature used except Fama and French confirm empirically the explanation of Lakonishok et al. (1994) (such as van der Hart et al., 2003 and 2005 or Athanassakos, 2009). I analyze and judge the debate to capture the essence of the issue. Closer look at the arguments used in the literature helps to explain the value premium also in post-communist Europe and thus is relevant for this paper’s third hypothesis. Following subsections approach the third hypothesis from a theoretical and argumentative angle and thus are complementary to the subsection 4.4 that attempts to find the explanation of the value premium in post-communist Europe empirically.
2.6.1 The Argumentation of Fama and French (1998)
which are used to compute the High-Minus-Low Value portfolio returns for the four indicators for every year. As reported in Table II of Fama and French (1998), for High-minus-Low Book-to-Market, they obtain a mean of 7.68% and a standard deviation of 9.94 percentage points, resulting in a t-statistics of 3.45. For the remaining three indicators, the t-statistics ranges between 2.38 and 3.45. The calculated market premium equals 9.60% on average with a standard deviation of 15.67 percentage points and a t-statistics of 2.74.
Fama and French (1998, p.1986) compare the value premium with the market premium (based on the empirical results in Fama and French 1998, Table XI). Based on the fact that the magnitude and t-statistics for the value and market premium are comparable, Fama and French close the discussion as follows “We conclude that value-growth premiums are no more suspicious than the market premium. At a minimum, the large standard deviations of the value-growth premiums say that they are not arbitrage opportunities.”
of the value premium, but to find a mutation of the Capital Asset Pricing Model that would explain well the international stock returns.
Further deficit of Fama and French (1998) is that they do not deal with the argumentation nor the methodology of Lakonishok et al. (1994). Lakonishok et al. (1994) bring the point that standard deviation of the excess returns itself is a limited measure of risk, since what really matters for an investor is the downside risk, especially in crises or recessions when the marginal utility of wealth is high. Fama and French (1998) use the high standard deviation of the value premium as their only argument for the risk-based explanation. However, this approach was already refuted by Lakonishok et al. (1994) and not defended by any of the Fama and French papers to the topic.
2.6.2 Main arguments of Fama and French (1996)
It is rather the paper of Fama and French (1996, especially p.79) that deals with the argumentation of Lakonishok et al. (1994). Fama and French (1996) research the value premium on American stocks in the time frame from 1964 to 1993. They bring four arguments against the approach of Lakonishok et al. (1994).
First, in a similar manner as in Fama and French (1996), they compare the value to the market premium, this time called relative-distress premium. They state that the value premium is as “suspicious” as the market premium.
Second, they claim that investors’ overreaction “cannot be the whole story”. Fama and French (1998, p.56) understand firms with high Book-to-Market ratio as financially distressed firms with persistently low earnings and firms with low Book-to-Market ratio as financially strong firms with persistently high earnings. There is an evidence of mean reversion in the earnings growth. The value premium persists for at least five years after the portfolio formation, whereas the earnings reversion is expected to occur much sooner. From this information, Fama and French (1998, p.79) conclude that investors’ overreaction and absent understanding of the mean reversion trend in earnings growth cannot fully explain the value premium.
variables like GNP. Thus, the value premium is not correlated with GNP or the whole market, even though value stocks are still riskier than other stocks.
Fourth, the variance analysis of returns as performed by Lakonishok et al. (1994) is according to Fama and French (1996) not sufficient for portfolio risk. As results from the three-factor model equation, the positive HML slopes of value firms raise their return variances and imply higher average returns, whereas negative HML coefficients for growth stocks also raise their return variances but imply lower average returns.
2.6.3 Critique of Fama and French (1996)
A closer look at the arguments of Fama and French (1996) shows that the approach of Lakonishok et al. (1994) withstands the critique from Fama and French.
The first argument of value premium being suspicious as the market premium was addressed in sub-section 2.6.1.
The second argument shows the limits of the explanatory power of the overreaction model. Though even if the overreaction of market participants explained only a small or even a neglectable part of the value premium, it would not necessarily imply that risk actually is the explanation. Overreaction of market participants means in the context of Lakonishok et al. (1994) over-extrapolation of past returns too far to the future which they report as the main driver behind the value premium based on their econometric analysis. However, as Lakonishok et al. (1994) acknowledge, there can be also other market irrationality-based explanations of the value premium such as incentives for institutional investors to overinvest in growth stock. Another of several possible explanations brought by Lakonishok et al. (1994) could be e.g. private investors confusing well-run companies with good investments. The conclusion of Lakonishok et al. (1994) is not that value premium can be explained with overreaction (i.e. over-extrapolation of past returns) only. Rather, Lakonishok et al. (1994) find that the value premium cannot be fully explained by risk and consider several behavioral explanations, backed-up by the evidence for the over-extrapolation model. Thus, the second argument of Fama and French does not support their statement of value premium being a risk factor rather than an arbitrage opportunity.
in the long run. If risk was the explanation, an investor would intuitively expect more severe losses in crises and/or recessions. If risk was the explanation and there are no excess losses of value stocks in crises or recessions, Fama and French should communicate in what states of the world excess losses of the value stocks do occur to illustrate the existence and nature of the risk. Though this illustration is missing in the article. The point of industry-specific risks certainly enriches the discussion about the anomalies in the stock market. Possibly industry cycles are another important factor in explaining stock returns and could be successfully used as further extension of the three-factor model. Nevertheless, this is not what Fama and French (1996) do. In Fama and French (1996) there is neither an empirical nor a logical proof why value stocks should be exposed to higher industry-specific risks than other stocks. Intuitively, investors can expect lower book-to-market ratios for instance for technology-oriented companies as they can be expected to have larger share of intangible assets that do not enter the book value. Therefore, it is conceivable to expect certain industry-tilt differences between growth and value stocks. Though in the article, there is not even an analysis of industry representation among the value stocks, let alone return and risk analysis of value and growth stocks based on their industry and the industry cycle stage or the like. Therefore, I conclude that the third argument of Fama and French (1996) does not tackle the validity of the down-market analysis as the argument is not backed up by any evidence related to the value premium.
2.6.4 The inherent problem of CAPM and the three-factor model in the context of the value premium explanation
Even the structure of Fama and French’s multifactor pricing models (equations (1) and (2)) actually indicates that risk is not the only possible explanation of the value premium. The global value premium is put as an explanatory variable for excess portfolio returns. However, the results of these and other pricing models do not give the answers why the global value premium is there. Rather, the results of the pricing models can be interpreted as an evidence of a global value premium presence. The internationally established value premium pattern can, besides risk and company size, explain excessive portfolio returns. In other words, having a closer look on the pricing model equations, the results of Fama and French explain excess returns with risk, value premium and size and not the value premium with risk.
The origin of the problems of the argumentation of Fama and French (1996 and 1998) inhere in the assumptions of the traditional one-factor CAPM: “Investors are assumed to be risk averse
and to behave as if they choose among portfolios on the basis of maximum expected utility. A perfect capital market, investor risk aversion, and two-parameter return distributions imply the important “efficient set theorem”: The optimal portfolio for any investor must be efficient in the sense that no other portfolio with the same or higher expected return has lower dispersion of return” Fama and MacBeth (1973, p.607).
If there are behavioral patterns in the market, i.e. irrationality of the investors driving the value premium, the set of assumptions labeled “efficient set theorem” by Fama and MacBeth (1973) is violated. In that case, also e.g. the three factor model would hit its limitations as it builds on the traditional CAPM. Using the mutations of the CAPM, Fama and French assume rational markets that are not driven by behavioral aspects. Though reality and assumptions of theoretical models should be distinguished (Sedlacek, 2009, p.299 ff). Investors’ rationality in the context of the CAPM and the three-factor model should be further considered just a model assumption as it was meant to be originally.
Nevertheless, having closer look at their argumentation, model set-up and results as discussed above, Lakonishok et al. (1994) bring a behavioral value premium explanation backed up by argumentation and empirical evidence that is stronger than the risk-based explanation of Fama and French (1996 and 1998).
2.6.5 More recent work of Fama and French
Fama and French (2012) apply the four-factor model (i.e. covering value, size and momentum) to explain the stock returns in four macro-regions: North America, Western Europe, Asia Pacific and Japan (23 countries in total) in the timeframe 1989 - 2011. The aim of the paper is to measure the performance of global and local four-factor models. As expected, the value premium is found. Without going deeper into detail, the main point of critique of this paper is that it ignores the by then already well established body of literature that offers the behavioral patterns as another conceivable explanation of the value premium. Instead, Fama and French (2012) look for the combination of global and local factors of the model that would explain the stocks returns the best. They see high CAPM alphas, i.e. excess returns that cannot be explained by risk as a measure of a poor performance of the model which they take as an impulse to adjust the model. However, they do not admit that alphas could be possibly explained by behavioral patterns. They observe “a mixed performance” of the four factor models.
Fama and French (2006a) analyze both American and international data in an extensive time period from 1926 to 2004. The aim of the paper is to test the performance of the CAPM. They judge the performance of the CAPM based on the intercepts α under the assumption that in the CAPM world, the intercepts α must equal zero. Again, the behavioral explanation is not considered as a possible explanation of the alphas. As expected, they come to the conclusion that the CAPM cannot fully explain the value premium.
The value premium debate between 1992 and 1998 between Fama and French and Lakonishok et al. (1994) was the most interesting clash in the literature to date. The newer literature can be split into two main streams: first, empirical literature confirming the behavioral explanation of Lakonishok et al. (1994) on various datasets (see Table 1.1 for details) and second, Fama and French (2006a, 2006b and 2012) who do not consider investors’ irrationality as one of the factors behind stock returns nor challenge the arguments of the behavioral explanation. Instead they search for a model explaining stock returns with non-behavioral drivers. Thus, the newer behavior-based literature stream has a higher persuasive power as they do control for the risk in their analysis whereas Fama and French do not control for behavior.
In context of subsection 2.6, behavioral aspects can be expected to drive the value premium also in post-communist Europe. Sub-section 4.4 confirms this later empirically while testing the third hypothesis.
3. Data
This paper’s analysis is based on the data on listed companies in the 28 member countries of the European Union. I distinguish between 11 post-communist countries (Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, Slovenia) and the Rest of the EU (also labeled “non-post-communist Europe” or simply “Western Europe” [even though the term also includes Southern and Northern European countries]). The list of exchange-traded companies in the geography of interest was obtained from the Orbis database resulting in more than 11,000 ISINs. Those were used to extract the variables of interest from Datastream: the annual prices between 2005 and 2013, the market-to-book ratio and market capitalization2. Since the information on price or one of the two indicators is missing for a large
portion of the sample, there are eventually about six thousand companies that can be used for my analysis.
As expected, the companies from non-post-communist countries dominate the sample being represented with more than 3,000 to about 5,000 stocks. From these 18 countries, about 60% of the companies are listed in the United Kingdom, Germany or France.
countries. In addition, with notable exception of Poland, Eastern European countries are smaller economies. Therefore the sample size is one of the limitations of this paper’s research compared to earlier studies performed on data from large and well developed markets such as Japan or the United States (see Table 1.1 for further reference). Over time, the number of companies listed in Eastern Europe increases. This paper’s analysis begins in the year 2005 and ends in 2013 so that the number of listed companies on Eastern European stock exchanges with sufficient data ranges between 389 and 1041. The largest share of the Eastern European stocks is listed in Poland, the largest economy and market in the region, followed by Bulgaria and Romania. Further information about the geographical representation in the sample as well as number of stocks in the portfolios each year can be found in tables A1 - A4 in the Appendix.
In line with Lakonishok et al. (2004), I use annual stock price data that correspond to prices as of January 01 of the given year. The annual performance in percent is computed based on this stock price time series. The stock prices and therewith also the annual performance are in local currencies to capture investors’ returns that are unaffected by exchange rate changes, in line with Fama and French (2012). I acknowledge the disadvantage of this approach: the returns cannot be directly interpreted as returns of an investor investing from the dollar- or euro-area into post-communist Europe. I use market capitalization data in Euro to enable comparison between company sizes across currency areas. This way, exchange rate fluctuations affect the results of all analyses where size-weighting is used: Stocks from currency areas with relatively strong currencies in a given year have greater weighting. This is acceptable under the assumption that value stocks are not over- or under-represented in a currency area with an extreme foreign exchange rate spike in one of the years considered. All analyses using equal weighting of returns are not affected by foreign exchange rate changes.
4. Establishing the Value Premium in Post-Communist Europe
4.1 MethodologyAthanassakos (2009) quartiles for both dimensions are used. Based on the Market-To-Book (M/B) value and Market Capitalization in Euro, the stocks are assigned to one of 16 portfolios. The explanation and quantification of size effects are out of this paper’s scope. Nevertheless, the size effects are also included in the analysis as control variables which is a standard in all literature on the value premium topic. The consideration of the size effect is based on the finding of Banz (1981) of smaller companies tending to bring higher returns.
As in Lakonishok et al. (2004), the portfolios are rebalanced every year according to size and value quartiles that are re-calculated for every year. Thus, stocks move between the portfolios as their market capitalization and book value changes across years relative to other stocks. As companies get listed and delisted, the number of stocks in the portfolios changes (as can be seen in Tables A1 and A3). As in Fama and French (1992), portfolios used for the mean equality analysis are Market Value weighted, larger capitalization companies being more strongly represented. As can be seen on the cut-off values in the Table 4.1, the market capitalization differences within the size quartiles are relatively large so that the smaller companies within the quartile have almost neglectable effects on the analysis results whereas larger companies drive the results. There is also literature using equal weighting (e.g. van der Hart et al. 2005 or Lakonishok et al. 1994). In this paper, equal weighting is used for regression analysis (including CAPM and the three-factor model in sub-section 4.4). As will be shown later, both approaches lead to similar results.
The quartile based portfolios are built separately for Eastern European countries and the rest of the EU, enabling to compare the stock performance both between the two regions and value and size quartiles. The annual stock returns are capped by 400% to reduce outlier bias in a similar manner as van der Hart et al. (2005) who cap monthly returns.
Capital Asset Pricing Models are used to test the third hypothesis.
2005 2006 2007 2008 2009 2010 2011 2012 2013 Value Market-to-Book 25th percentile 0.515 0.650 0.660 0.930 0.390 0.450 0.440 0.340 0.320 median 0.990 1.155 1.310 1.820 0.680 0.840 0.940 0.650 0.660 75th percentile 1.815 1.973 2.350 3.238 1.215 1.440 1.640 1.220 1.260 Market Cap. in Millions of € 25th percentile 6.1 8.3 6.8 11.0 4.6 4.5 4.6 3.1 2.6 median 20.0 29.1 28.3 42.4 14.2 15.2 16.3 10.7 10.4 75th percentile 97.2 119.1 117.8 152.4 47.9 54.8 63.2 46.4 46.7
4.2 Results: establishing value premium in Eastern Europe
This section shows that the value premium is existent in the post-communist countries of the EU and quantifies its magnitude. In line with Lakonishok et al. (1994: Table I) and van der Hart et al. (2005: Table 1), I perform simple comparison of the average portfolio as the first step of quantitative analysis. Table 4.1 presents the cut-off values for the quartile-based portfolios. Notably the cutoff Market-to-Book ratio for all companies in the top value quartile in all years is below 1, i.e. the market price of the companies is even less than their book value. In this quartile, also companies with negative Market-to-Book ratios are included.
The annual performances of all size-weighted portfolios for every year in the time period 2005 - 2003 can be found in Table 4.2. Van der Hart et al. (2005) analyze the success of the value-focused stock selection strategy considering 15% top and bottom Book-to-Market stocks from the sample. Similarly, in this paper the value premium is established based on the comparison of the first B/M quartile (value stocks) and the fourth B/M quartile (growth stocks). The size quartiles enable to separate the value effects from the size effects.
It can be observed that the performance of the top-value-smallest-size quartile as well as the average performance of all four size quartiles for the top-value quartile exceeds the weighted
Value
Quartile QuartileSize 2005 2006 2007 2008 2009 2010 2011 2012 2013
1 1 51.2% 127.8% 117.4% -30.3% 48.0% 34.2% -2.4% 2.5% 9.7% 1 2 55.2% 138.5% 88.5% -47.0% 53.6% 0.9% -20.4% -0.1% 9.3% 1 3 34.0% 88.2% 64.2% -45.4% 32.7% -3.9% -20.3% -3.6% 40.1% 1 4 144.9% 24.6% 96.6% -51.1% 91.2% 26.4% -8.4% 3.5% -6.1% 2 1 9.4% 120.5% 101.7% -43.4% 39.4% 6.4% -8.3% -5.7% 7.9% 2 2 63.2% 75.7% 111.8% -57.3% 20.1% 20.7% -18.1% 1.6% 25.8% 2 3 39.7% 58.3% 42.0% -59.2% 45.8% 17.4% -20.2% -9.6% 30.8% 2 4 58.2% 22.6% 30.2% -46.6% 76.0% 27.9% -14.7% 21.9% 14.1% 3 1 29.1% 74.5% 109.4% -54.1% 42.2% 54.6% -23.9% -11.3% 15.7% 3 2 25.3% 84.8% 41.5% -60.2% 19.1% 21.4% -30.3% -10.4% 10.9% 3 3 36.9% 41.5% 41.2% -68.2% 15.6% 16.2% -24.3% 1.9% 22.3% 3 4 44.0% 19.9% 18.6% -53.1% 34.2% 6.6% -22.3% 11.5% 3.9% 4 1 80.6% 63.0% 42.5% -47.1% 43.2% -7.9% -34.5% -9.4% 1.9% 4 2 -16.1% 28.2% 21.1% -58.2% -2.7% -1.5% -29.4% -21.7% -3.2% 4 3 10.2% 58.9% 17.6% -68.8% 4.2% -3.3% -37.3% -7.2% 6.6% 4 4 35.3% 34.8% 12.7% -53.9% 14.6% 7.1% -13.7% 3.7% 3.1% Weighted Average 40.5% 31.3% 18.2% -53.1% 27.2% 9.3% -17.1% 7.8% 4.7%
Table 4.2: Portfolio Performance: Post-Communist EU countries; There are 16 portfolios: for each value quartile
average of Eastern European stocks every year except 2012. The value premium is statistically significant with a t-test probability of 2.3% testing for mean equality between the average returns of the first value quartile portfolios and all remaining portfolios. The t-test for the mean equality of portfolio returns between the first and the fourth value quartile is significant having a p-value of 0.6%. Performing the t-test for mean equality for every year, statistically significant higher returns of the top value quartile occur in 6 out of the 9 years (2005-2009 and 2011). Table 4.3 summarizes the performance over the years and demonstrates the superior returns of value stocks relative to the growth stocks as well as to the size-weighted average return in that period of 7.6% per annum. The value stocks outperform the size-weighted return in that period by 24.2 percentage points of additional returns per annum on average. Using equally weighted portfolio returns, a value premium of 19.2 percentage points is obtained. However, these figures are strongly influenced by abnormally high stock returns up till and including 2007. Considering only the time from 2008 on, the magnitude of the value premium equals 9.1 percentage points (computed as average value portfolio returns minus average returns of all other portfolios, using size-weighted returns). This magnitude is within the range of 7.8 - 13.5 percentage points found by Chan and Lakonishok (2004).
Value Growth Quartile 1 2 3 4 Small 1 39.8% 25.3% 26.2% 14.7% 2 30.9% 27.1% 11.3% -9.3% Big 3 20.7% 16.1% 9.2% -2.1% 4 35.7% 21.1% 7.0% 4.9%
Table 4.3: Average portfolio performance of Eastern European stocks in the period 2005-2013; Stocks are annually
assigned to their size and value quartile, 16 portfolios are built based on the quartiles in the two dimensions
its essence similar to Capital Asset Pricing Model as performed later for testing the third hypothesis. In a CAPM, the yearly returns of the portfolios are regressed with the returns of the market (both risk free rate adjusted). This way alpha (the intercept standing for excess returns that cannot be explained by higher riskiness of the investment) and beta (the slope standing for the riskiness of the investment) are calculated. Comparing the alphas and betas of size and value portfolios with each other, distilling the size and value effects, conclusions can be drawn on existence and possible rationales of the value premium. In the regression analysis, the stock returns are explained also with the size and value effects. Since all traded stocks with available data from post-communist Europe are included in the regression, the constant reflects the size- and value-adjusted portfolio returns in the Eastern European region for the given year. The constant is estimated for every year, thus its magnitude reflects the market trends. The CAPM approach has the advantage of distinguishing between alpha and beta which makes it suitable for testing whether the value premium is driven by risk or behavior. On the other hand, the regression analysis enables to distinguish between the years and thus it is better suitable for testing whether, when and with which economic and statistical significance the value premium occurs.
The results of the regression analysis are presented in Table 4.4. All stocks are weighted equally. Significant market-to-book effects can be observed in the sub-period 2009 - 2011. Compared to the mean equality analysis of annual returns, the value premium comes out as less persistent.
Stock Returns: Post-Communist Countries
Dependent Variable: Annual portfolio Return in %
2005 2006 2007 2008 2009 2010 2011 2012 2013 Constant 0.000***0.546 0.000***1.355 0.000***1.017 0.000***-0.252 0.000***0.266 0.001***0.112 0.000***-0.111 -0.0100.672 0.000***0.202 Market-to-Book 0.0020.535 0.0030.631 -0.0030.459 -0.0030.362 0.010***-0.006 0.003***-0.012 0.001***-0.010 0.5720.000 -0.0020.170 Ln(MarketCap) -0.043 -0.159 -0.119 -0.067 0.011 0.006 -0.024 0.007 -0.003 0.037** 0.000*** 0.000*** 0.000*** 0.378 0.538 0.000*** 0.326 0.763 Adjusted R2 0.0064 0.0803 0.0684 0.1447 0.0062 0.0082 0.0271 -0.0007 -0.0001 F-Test P-Value 0.111 0.000 0.000 0.000 0.026 0.009 0.000 0.516 0.378 Sample Size 375 475 664 771 852 889 958 994 999
Table 4.4: Regression results, Coefficients and their p-values, *** for 1%, ** for 5% and * for 10% confidence levels; Market-to-Book and Ln(MarketCap) are included as continuous variables, all stocks are
More significant results are obtained if the top quartile dummy is considered as a distinguisher of value stocks (results in Table 4.5). The dummy equals 1 if given stock belongs to the top value quartile and zero if it belongs to any other value quartile. This approach builds on the literature that evaluates and quantifies the value premium based on the performance of certain top value percentile of the stock sample, such as van der Hart (2003) or Chan and Lakonishok (2004). This method enables to judge whether an investment strategy betting on top value stocks would beat the market.
This way, statistically significant value premium appears in six years out of the nine years period. If statistically significant, the value premium is always positive and economically significant: in the years with the value premium presence, the top quartile stocks bring at least 10.6 percentage points higher returns than other stocks in the sample.
Stock Returns: Post-Communist Countries
Dependent Variable: Annual portfolio Return in %
2005 2006 2007 2008 2009 2010 2011 2012 2013 Constant 0.000***0.298 0.000***1.040 0.000***0.867 0.000***-0.434 0.000***0.223 0.000***0.138 0.000***-0.240 0.008***-0.065 0.000***0.128 Top Value Quartile 0.297 0.387 0.310 0.129 0.252 -0.017 0.106 0.026 0.019 0.003** 0.002*** 0.005*** 0.000*** 0.000*** 0.738 0.002*** 0.485 0.727 Ln(MarketCap) 0.0100.854 0.000***-0.323 0.000***-0.316 0.000***-0.112 0.7210.012 -0.0210.411 -0.0070.692 0.010***0.048 0.022***0.053 Adjusted R2 0.0211 0.0922 0.0803 0.1064 0.0184 -0.0016 0.0102 0.0051 0.0038 F-Test P-Value 0.007 0.000 0.000 0.000 0.000 0.711 0.004 0.037 0.065 Sample Size 376 454 609 729 801 830 888 902 907
Table 4.5: Regression results, coefficients and their p-values, *** for 1%, ** for 5% and * for 10% confidence levels; Book-to-Market included as a dummy for the top value quartile
Table A5 in the Appendix shows the alphas and betas of a standard CAPM for the Eastern European dataset based on annual returns3, all stocks weighted equally. Using the average
alphas of the value portfolios over the years, the value premium cannot be quantified as 3 of 4 alphas are not statistically significant.
4.3 Results: Value premium in post-communist EU countries compared to the value premium in the rest of the EU
This section identifies the differences between the value premiums in the post-communist countries and the rest of Europe. Some of the differences between the universe of stocks from Eastern and Western Europe can be deducted from the Table 4.6. Listed companies in Western Europe tend to have much larger market capitalization. The Market-to-Book ratio tends to be lower in Eastern Europe making stocks from that region “cheaper”. To some extent this Market-to-Book difference can be expected due to higher Research and Development intensity in Western Europe: e.g. in 2011 in Eastern Europe, the R&D expenditures per capita in Euro in the private sector were less than a fifth compared to the rest of the EU4. As the R&D costs do not
enter the book value of the firm, companies with higher R&D expenditures should have ceteris paribus higher Market-to-Book ratios as pointed out by Lakonishok et al. (1994).
The rest of the EU in the context of this paper is geo-economically more diverse as it aggregates countries in Southern and Northwestern Europe. Compared to Eastern Europe, the value premium in the Western European countries has received more attention from the research community (at least the most important countries) e.g. in Fama and French (1998 and 2012). Both papers find the value premium in Western European countries.
2005 2006 2007 2008 2009 2010 2011 2012 2013 Value Market-to-Book 25th percentile 0.820 0.910 1.010 1.050 0.520 0.690 0.750 0.630 0.630 median 1.350 1.600 1.750 1.710 0.880 1.070 1.230 1.000 1.060 75th percentile 2.378 2.790 3.080 3.000 1.600 2.020 2.300 1.930 2.060 Market Cap. in Millions of € 25th percentile 16.7 20.7 23.6 22.4 9.5 12.8 14.3 11.6 11.6 median 73.2 89.0 103.8 91.0 39.4 54.3 61.6 51.8 55.0 75th percentile 410.3 491.8 597.3 512.4 214.4 323.1 383.2 304.5 341.2
Table 4.6: Cut-off values for value and size quartiles for Western European stocks
Value
Quartile QuartileSize 2005 2006 2007 2008 2009 2010 2011 2012 2013
1 1 39.4% 34.7% -2.6% -44.9% 63.0% 14.3% -18.3% 10.1% 25.6% 1 2 28.4% 43.2% 1.5% -48.2% 51.5% 7.4% -21.2% 18.7% 34.8% 1 3 29.8% 34.9% -10.0% -45.2% 48.2% 8.8% -20.1% 17.6% 32.6% 1 4 15.5% 5.1% 3.9% -48.2% 45.1% 2.2% -26.0% 18.2% 26.1% 2 1 29.7% 27.2% -2.4% -50.8% 31.3% 21.9% -22.3% 7.4% 21.7% 2 2 31.0% 32.8% -4.6% -45.4% 51.2% 30.4% -18.6% 12.4% 24.8% 2 3 24.1% 28.7% 7.4% -47.5% 35.8% 6.4% -18.5% 14.2% 27.8% 2 4 23.6% 16.2% -0.7% -39.9% 21.0% -3.2% -12.1% 4.6% 20.0% 3 1 34.1% 27.4% -1.6% -44.4% 37.8% 20.0% -19.2% 6.8% 27.9% 3 2 28.3% 29.5% 2.4% -40.4% 41.0% 30.1% -13.6% 18.0% 33.4% 3 3 31.3% 29.1% 2.6% -46.6% 33.4% 15.2% -14.3% 15.6% 25.2% 3 4 17.2% 14.1% 14.1% -39.3% 17.9% 8.5% -11.9% 14.3% 19.4% 4 1 24.7% 23.2% -9.4% -51.9% 33.6% 14.4% -14.8% -8.4% 8.5% 4 2 23.6% 21.7% -4.5% -45.8% 30.3% 26.7% -20.1% 18.2% 24.4% 4 3 17.4% 18.8% -2.5% -36.5% 29.0% 15.8% -8.8% 19.3% 27.3% 4 4 19.8% 9.8% 13.2% -37.8% 13.3% 8.7% -3.6% 9.7% 14.6% Weighted Average 22.2% 17.4% 3.6% -42.3% 27.0% 7.6% -14.4% 12.5% 21.3%
Table 4.7: Portfolio Performance: Western European countries; There are 16 portfolios: for each value quartile
there are 4 size portfolios. To calculate the average return in a given year, stocks are weighted by their market capitalization.
As can be seen in the Table 4.8, the stocks of the top value quartile in Western Europe bring rather higher returns than stocks in the remaining portfolios. Though this difference is not statistically significant as follows from the mean equality t-tests comparing the annual returns of the top value quartile portfolios with first all remaining portfolios and second the portfolios in the bottom quartile. Value Growth Quartile 1 2 3 4 Small 1 13.5% 7.1% 9.9% 2.2% 2 12.9% 12.7% 14.3% 8.3% Big 3 10.7% 8.7% 10.2% 8.9% 4 4.6% 3.3% 6.0% 5.3%
Table 4.8: Average portfolio performance of Western European stocks in the period 2005-2013
it is listed in post-communist Europe. The “Rest of the EU Top Quartile” dummy has analogously the value 1 only for top value quartile stocks from Western Europe. It is controlled for size effects. Another control variable is the “Eastern EU dummy” capturing the differences in the overall market returns in the two regions. It is assumed that Western Europe can be considered one market.
Stock Returns: All EU Countries
Dependent Variable: Annual portfolio Return in %
2005 2006 2007 2008 2009 2010 2011 2012 2013 Constant 0.368 0.258 0.088 -0.391 0.282 0.154 -0.083 -0.038 0.185 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.013*** 0.000*** Eastern EU Top Value Quartile 0.105 0.597 0.501 0.277 0.051 0.048 0.096 0.124 -0.013 0.165 0.000*** 0.000*** 0.000*** 0.393 0.325 0.003*** 0.001*** 0.787 Rest of the EU Top Value Quartile 0.156 0.152 0.117 0.058 0.375 0.083 0.044 0.074 0.106 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.003*** 0.000*** 0.000*** Ln(MarketCap) -0.014 -0.009 -0.019 -0.014 0.007 -0.002 -0.017 0.023 0.007 0.004*** 0.039*** 0.000*** 0.000*** 0.157 0.536 0.000*** 0.000*** 0.048*** Eastern EU dummy 0.064 0.431 0.461 -0.131 -0.040 -0.053 -0.094 -0.053 -0.018 0.101 0.000*** 0.000*** 0.000*** 0.211 0.039*** 0.000*** 0.007*** 0.492 Adjusted R2 0.0137 0.0971 0.1404 0.0489 0.0431 0.0040 0.0184 0.0152 0.0045 F-Test P-Value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Sample Size 3765 4154 4743 5247 5503 5618 5862 6007 6005
Table 4.9: Regression results, Coefficients and their p-values, *** for 1%, ** for 5% and * for 10% confidence levels; Both Eastern and Western European stocks are considered. Value quartiles are built separately for Eastern and
Western Europe. Thus, there are two top value dummies for the two geographies. The Eastern EU dummy captures the return differences between Eastern and Western European stocks markets. All stocks are weighted equally.
As the Table 4.9 shows, the regression analysis confirms a positive and both statistically and economically significant value premium in Western Europe. It occurs every year in the period and it brings 4.4 - 15.6 percentage points p.a. higher returns relative to other portfolios. In comparison, statistically significant value premium in the East of the European Union only occurs in five out of the nine years. Compared to the rest of the EU, Eastern European markets show high economic significance of the value premium, ranging from 9.6 to 59.7 percentage points of additional annual returns.
persistent, though stronger up to and including 2008. After this crisis year, the value premium in post-communist countries becomes comparably strong as in the rest of the EU. Interestingly, also the overall market in the East of the EU performs worse in and after the crisis as can be seen on the “Eastern EU” dummy trend. As the regression results show, before the crisis there was even a double premium in post-communist EU countries relative to the rest of the EU: first, the stock markets in Eastern Europe outperformed those in the rest of the EU and second, the value premium was higher in the Eastern Europe than in Western Europe. However, after the crisis year 2008 the Eastern European “premium on the value premium” disappeared and also from 2010 on, the Eastern European markets underperformed the rest of the EU.
From the regression analysis in Table 4.9, I conclude that the value premium in post-communist Europe was stronger before the global financial crisis whereas in the recent years, its magnitude declined to a level of 9.6 and 12.4 percentage points of additional returns which is comparable to the rest of the EU. The value premium in Eastern Europe is also less persistent than in Western Europe.
4.4 Results: Explaining value premium in post-communist Europe: risk or behavior?
Similarly to Lakonishok et al. (1994, p.1569) and van der Hart et al. (2003 and 2005)5, this
section attempts to capture the value premium in post-communist Europe with versions of the Capital Asset Pricing Model. In addition, I analyze the standard deviation of value portfolio returns in relation to other portfolios.
Before that, I perform the downside risk analysis as in the baseline paper, motivated by the intuition that in the “bad states of the world” the marginal utility of wealth is especially high. The “bad states of world” of interest are down-markets or recessions, which I examine on a yearly basis. I define down-market year as a year when the relevant pan-European index, Euro Stoxx 50 as well as the weighted annual performance averages of the stocks considered dropped by more than 10%. Recession is commonly understood as a period of general decline in economic activity for two or more consecutive quarters (as e.g. in Blanchard, 2003). Based on the Euro Stoxx 50 index and GDP growth data for EU28, I obtain 2008 and 2011 as down-market years and 2008, 2009, 2012 and 2013 as recession years.6 The performance of the
value portfolios in these years can be judged based on the Tables 4.2 and 4.5.
Table 4.2 dealing with the size-weighted portfolio returns shows value portfolios losses in the range of -51.1% to -30.3% in 2008 and gains in the range of 32.7 to 91.2% in 2009. The size-5 Van der Hart et al. (2003 and 2005) work with the four-factor model that also includes momentum.
weighted performance averages were -53.1% and 27.2%. The value portfolios outperformed the weighted average of the Eastern European stock universe in the crisis of 2008 and in the aftermath recession on 2009. Considering the regression analysis in the Table 4.5 (based on equal weighting of all stocks), the results are confirmed as we observe a positive and statistically and economically significant value premium in 2008 and 2009.
As can be seen in Table 4.2, all four top value portfolios underperformed the market average in the recession year 2012. The size-weighted top value portfolio returns for the 2011 downmarket and the recession year 2013 vary heavily depending on the size quartile.
Table 4.5 shows a positive and significant value premium in 2011 whereas the top value quartile dummy is not statistically significant in 2012 and 2013. The evidence for the 2011 down-market and the recession years 2012 and 2013 is therefore less clear. In the down-market year 2011, value stocks outperformed other stocks as results from the regression analysis in Table 4.5. However, the investor would be tendentially better off keeping rather other than value stocks in her size-weighted portfolio in 2012 (Table 4.2). This trend cannot be observed in 2013 anymore though.
I conclude from this analysis that Eastern European value stocks were still a less risky investment in the most recent “bad states of the world” since they proved themselves to be more stable in the down-markets of 2008 and 2011 as well as in the recession year 2009. There is a mixed evidence on the performance in the recessions of 2012 and 2013. Certainly value stocks cannot be labeled as riskier than other stocks from the point of view of the downside risk analysis.
In the next step, I analyze the value premium based on the traditional CAPM (Fama and MacBeth, 1973) as in equation (3)
R - F = α + β (M - F) + e (3)
where R is the portfolio return, F is the risk-free rate, M is the market return and e is the regression error term. Thus the adjusted portfolio returns are explained with the risk-adjusted market returns. Beta is interpreted as a riskiness of given portfolio or the reagibility of the portfolio to the (risk-adjusted) market movements. Alpha are the excess returns of given portfolio that cannot be explained by higher riskiness of that portfolio. The Euribor7 is used as
accuracy of the results. The CAPM coefficients are based on monthly returns that were capped by 150% as in van der Hart (2005). All stocks are equally weighted.
The Euribor7 in the given year was subtracted from those. The risk adjusted returns over each
portfolio are the dependent variable. As the market return I use the market- capitalization-weighted average of all stocks with available data in post-communist EU countries. OLS regressions are performed to obtain the alpha and beta coefficients. The results are presented in Table 4.10.
Value
Quartile QuartileSize Betas Alphas
In Table 4.10, trends can be observed: the alphas of value stocks tend to be higher than alphas of the rest of the sample and they are also always positive whereas growth stocks rather have negative intercepts. Betas tend to be higher for value stocks with larger market capitalization. Overall, the magnitude of the top value quartile betas is comparable to other stocks in the sample, though it is significantly higher for the companies in the largest capitalization quartile. Betas should not be considered in isolation if the alphas are non-zero. Understanding risk as the “possibility of loss” and alpha as excess return, then two portfolios with equal betas and different alphas bear different risk levels: The portfolio with higher alpha is less risky as it is less likely to show losses in a given month. In that context, value stocks appear to be even a safer investment as other stocks because of their high alphas combined with betas at a normal level. As can be seen in the Table A6 in the Appendix, slightly different patterns can be recognized for the stocks from Western Europe, as the magnitude of their value stocks quartile alphas is lower (still positive though) whereas the beta magnitudes for value stocks are higher.
In line with Lischewski and Voronkova (2011), as the next step I explain the portfolio returns with the three factor CAPM as used by Fama and French (1996 and 1998). In the three factor CAPM, the SMB and HML factors are added as further explanatory variables. SMB (Small-Minus-Big) stands for the performance difference between the bottom market capitalization quartile and the top one. Analogously HML is the performance difference of the top and bottom value quartiles. The regression can be rewritten as equation 4. Results are presented in Table 4.11.
R - F = α + β (M - F) + γ SMB + δ HML+ e (4)
Value
Quartile QuartileSize HML SMB Market Beta Alpha
1 1 0.31 0.53 0.88 0.68% 0.000*** 0.000*** 0.000*** 0.015 1 2 0.50 0.06 1.04 -0.35% 0.000*** 0.359 0.000*** 0.297 1 3 0.57 0.00 1.03 -1.09% 0.000*** 0.957 0.000*** 0.013 1 4 0.76 -0.57 1.04 -0.11% 0.000*** 0.000*** 0.000*** 0.836 2 1 0.07 0.44 0.92 -0.29% 0.378 0.000*** 0.000*** 0.491 2 2 0.02 0.03 1.09 0.44% 0.688 0.677 0.000*** 0.147 2 3 0.05 -0.26 0.99 -0.09% 0.320 0.000*** 0.000*** 0.746 2 4 -0.08 -0.40 0.98 0.81% 0.052* 0.000*** 0.000*** 0.000*** 3 1 0.20 0.70 1.22 0.28% 0.049*** 0.000*** 0.000*** 0.593 3 2 -0.10 0.03 1.11 -0.27% 0.121 0.660 0.000*** 0.408 3 3 -0.18 -0.28 1.04 -0.01% 0.000*** 0.000*** 0.000*** 0.956 3 4 -0.19 -0.43 0.95 0.19% 0.000*** 0.000*** 0.000*** 0.347*** 4 1 -0.48 0.54 0.95 0.27% 0.000*** 0.000*** 0.000*** 0.551*** 4 2 -0.58 0.04 1.01 -0.16% 0.000*** 0.653 0.000*** 0.726 4 3 -0.43 -0.15 1.04 -1.03% 0.000*** 0.005 0.000*** 0.000*** 4 4 -0.38 -0.40 1.00 0.05% 0.000*** 0.000*** 0.000*** 0.806
Table 4.11: Regression coefficients of the three-factor CAPM: value and size portfolios in post-communist Europe for the period 2005 – 2003 based on monthly returns, *** for 1%, ** for 5% and * for 10% confidence
The results of standard CAPM show that there are no big differences in the riskiness of value and all other stocks since the betas of the two portfolio groups are not significantly different from each other. However, we observe higher alphas for value stocks which is a further evidence for the statement that the value premium cannot be explained solely by higher risk.
attributed to the HML factor. The HML factor reflects the value premium since it is based on the performance difference between value and growth stocks which was 2.1% per month on average.
As the next step, I analyze the volatility of portfolio returns. Standard deviation in monthly returns capped by 150% is used as the volatility measure. As one of the founders of the traditional CAPM states, “in equilibrium there will be a simple linear relationship between the expected return and standard deviation of return for efficient for efficient combinations of risky assets” (Sharpe 1964: p.36). If the value stocks are riskier, then according to the CAPM principle, the volatility of their portfolio returns should be higher than the volatility of the returns of other stocks. The standard deviation of monthly returns for each portfolio are presented in Table 4.12. The volatilities for the value portfolios are not statistically significantly different from returns of the remaining portfolios. The t-test for mean equality brings a p-value of about 35% for volatility equality of top and bottom value portfolios. Very similar p-value is obtained when testing for the volatility equality between the value portfolios and all remaining portfolios. From the volatility point of view the value stocks are not riskier than other stocks. The results again raise the doubt whether various versions of the Capital Asset Pricing Model can explain the nature of the value premium.
Value Growth Quartil e 1 2 3 4 Small 1 0.057 0.062 0.082 0.068 2 0.070 0.067 0.066 0.070 Big 3 0.075 0.063 0.063 0.063 4 0.095 0.062 0.061 0.062
t-test p-value Value quartile 1 vs 2, 3, 4 0.344 Value quartile 1 vs 4 0.347 Table 4.12: Standard Deviation of monthly portfolio returns of Eastern European stocks in the period 2005 - 2013
5. Conclusions
in the East of the EU changed its character in the relation to the value premium observed rest of the EU in the time period examined. Before the crisis of 2008 its magnitude was significantly higher, whereas in the years after the crisis its magnitude was at a similar level as in more developed markets of the EU. At the same time, the value premium in post-communist Europe appeared less persistently than in the rest of the EU.
The main focus of this article was the explanation of the value premium in Eastern European markets: Is it risk or behavior? In the literature section, I judge the debate between Fama and French (1992, 1996 and 1998) and Lakonishok et al. (1994) on argumentative and theoretical level, find deficits in the arguments of Fama and French (1996 and 1998) and recognize the behavioral explanation as overall stronger. I provide a rationale why versions of the Capital Asset Pricing Model can identify, but not explain the value premium: These models assume rational pricing. Thus if the value premium indeed is driven by behavioral patterns, the model cannot explain it as it violates the assumptions of the model.
I confirm the behavioral explanation of the value premium in post-communist Europe empirically. Value stocks in Eastern Europe in the time frame from 2005 to 2013 are not riskier than other stocks according to any risk measure used. In frame of the risk analysis, I covered the riskiness of value stocks in recession and down-markets as well as traditional risk measures as CAPM betas and standard deviation of portfolio returns. At the same time, value stocks bring higher returns than other stocks in the long run.
This paper confirmed that value stocks (not only) in post-communist Europe are an attractive investment opportunity that offers a risk-return balance that is close to a ‘free lunch’. Its economic significance and thus the return gain potential for an investor was impressive in the period from 2005 to 2013: from 10.7 up to 24.2 percentage points of additional return, depending on the estimation method. However, the magnitude of the value premium in post-communist Europe declined after the 2008 crisis.
References
Athanassakos, G. (2009), Value versus growth stock returns and the value premium: The Canadian experience 1985–2005. Canadian Journal of Administrative Sciences vol. 26: pp. 109–121
Banz, R.W. (1981), The relationship between return and market value of common stocks, Journal of Financial Economics, Volume 9, Issue 1, March 1981, pp. 3-18,
Barry, C.B., Goldreyer E., Lockwood, L., Rodriguez M. (2002), Robustness of size and value effects in emerging equity markets, 1985–2000, Emerging Markets Review, Volume 3, Issue 1, 1 March 2002, pp. 1-30
Blanchard, O. (2003), Macroeconomics, Third Edition, International Edition, Prentice Hall Buffett, M., Clark, D. (2009) - The Tao of Warren Buffett, Pocket Books
Chan, L. K. C, Lakonishok, J. (2004), Value and Growth Investing: Review and Update, Financial Analysts Journal, Vol. 60, No. 1 (Jan. - Feb., 2004), pp. 71-86
Chan, L. K. C., Hamao, Y. and Lakonishok, J. (1991), Fundamentals and Stock Returns in Japan. The Journal of Finance, 46, pp. 1739–1764
Fama, E. F. and French, K. R. (1992), The Cross-Section of Expected Stock Returns. The Journal of Finance, 47: 427–465.
Fama, E. F. and French, K. R. (1996), Multifactor Explanations of Asset Pricing Anomalies. The Journal of Finance, 51: pp. 55–84
Fama, E. F. and French, K. R. (1998), Value versus Growth: The International Evidence. The Journal of Finance, 53: pp. 1975–1999.
Fama, E. F. and French, K. R. (2012), Size, value, and momentum in international stock returns, Journal of Financial Economics, Volume 105, Issue 3, September 2012, pp. 457-472
Fama, E. F. and French, K. R. (2006a), The Value Premium and the CAPM. The Journal of Finance, 61: pp. 2163–2185.
