Master Thesis Finance. Value Premium and Business Cycles:

Transcription

1 Master Thesis Finance Value Premium and Business Cycles: a perspective on major European markets. Author: Anar Bayramov ANR: Program: MSc. Finance Graduation department: Supervisor: Chairperson: Finance Alberto Manconi Assistant Professor of Finance, Tilburg University Fabio Castiglionesi Associate Professor in Finance, Tilburg University Faculty: Finance, Tilburg University School of Economics and Management Date: January 29, 2013

2 Acknowledgements Firstly I would like to thank my supervisor, Alberto Manconi, Assistant Professor of Finance at Tilburg University, for his most valuable guidance. With no prior experience in this kind of work, I believe I have extensively deepened my knowledge while working on this challenging and interesting project. Also I would like to thank my family and friends for their support during my studies. Master Thesis - Anar Bayramov Page 1

3 Abstract This paper investigates the relation between value premiums and business cycles in France, Germany, the UK and the aggregate European market. We investigate whether it is possible to form profitable strategies from a universe of blue chip stocks. The study analyzes cyclicality of value premium, mean reversion in profitability of value and growth companies during various holding period returns in good and bad times. Further we evaluate risks and rewards of style strategies. We find that value strategies from a range of blue chip stocks are pursuable; the value premium is positively related to the state of economy and the largest returns are obtained by maintaining relatively short and mid horizon holding periods. While value stocks are riskier, the returns more than compensate for this risk. Master Thesis - Anar Bayramov Page 2

4 Table of Contents Acknowledgements 1 Abstract 2 Table of Contents 3 I. Introduction 4 II. Literature Review and Hypothesis Development Evolution of Value and Growth Emergence of Value Premium Value Premium during Expansions and Recessions Mean Reversion during Business Cycles Holding Period Returns Risk and Reward of Value and Growth Strategies 12 III. Data and Methodology Sample Selection Methodology Filtering and Sorting Significance Tests and Mean Reversion OLS Regressions and Risk-Reward Measures 18 IV. Empirical Results The Value Premium Cyclicality of Value Premium Post-formation Profitability Holding Period Returns Risk and Reward 24 V. Conclusions 26 VI. References 28 VII. Appendices 31 Master Thesis - Anar Bayramov Page 3

5 I. Introduction. A seminal contribution of Fama and French (1992) exposed academicians and practitioners to factors explaining the cross section of stock returns other than beta. Since then hedge funds, pension funds and other investment houses have formed technical portfolio management teams in an attempt to profit from these factors. Today, we observe emergence of various style indices, mutual funds and trackers following factor based strategies in order to serve the increasing demand for such products from investors. It has become a common practice to maintain factor portfolios in large funds. However, not all strategies perform well when markets are down, and it has become extremely difficult to beat the market in the wake of drastic fluctuations in economies worldwide: financial crisis, high market volatility, prolonged periods of stagnation and even double dip recessions for large economies like Spain and the Netherlands. In this paper, therefore, we investigate the performance of factor strategies involving the Book-to-Market ratio, which was branded by Fama and French (1992) as the most persistent factor in explaining the cross section of stock returns. We are specifically interested in the performance of value strategies during business cycles. Furthermore, along with the Book-to-Market ratio we analyze Price-to-Earnings and Price-to-Cash-Flows ratios, which are commonly considered alternatives. We wonder whether employment of strategies associated with the value premium can be a flight to safety during bad times. Thus, can we identify profitable and relatively low risk patterns in the behaviour of the value premium in expansions? What are the risks and rewards of inclusion of growth and value stocks in a portfolio during contractions? Based on these queries, we formulate our main research question: How do business cycles affect the value premium in Europe? Thereby, this paper addresses the relationship between value premiums in blue chip stocks and business cycles in German, French, British and aggregate European markets. We contribute to the previous research by performing in-depth analysis of the relationship between Master Thesis - Anar Bayramov Page 4

6 the value and growth during expansion and recession stages of the business cycles. We investigate mean reversion of style strategies and its effect on the holding period returns, and performance of these strategies in terms of risk-reward measures during economic ups and downs. We consider the aggregate European market as a one investment universe of blue chip stocks and examine how it confines with modern financial theories. We do this with the hope of providing additional insight into the stock investment universe from a different perspective so as to assist investment professionals in their evaluation of appropriate styles. The main findings of this paper are as follows. It is possible to pursue growth and value strategies in the universe of blue chip stocks in major European markets within a business cycle framework. The value premium is pro-cyclical, and we observe mean reversion in profitability of companies in good as well as bad times. A faster mean reversion results in shorter holding periods being more profitable than longer ones. The value stocks are riskier in three out of four markets investigated, but rewards associated with the extra amount of risk are more than expected. For the UK we observe consistently negative value premiums, especially during contractions. This makes the growth stock investments a valuable alternative during bad times. Moreover, investors can form profitable value strategies by treating Europe as one aggregate market during economic cycle stages. The remainder of this paper proceeds as follows. Section II documents the theory behind value premiums and develops testable hypotheses. Section III discusses the sample and methods used in analyzing it. The penultimate section presents the empirical findings. Section V concludes, and discusses limitations and suggestions for future research. II. Literature Review and Hypotheses Development. 2.1 Evolution of Value and Growth The classification of strategies by growth and value is based on fundamental analysis. Stocks classified as value are the ones on the lower end of Price-to-Earnings (), Market-to-Book- Value (P/B), Price-to-Cash-Flows (P/CF), Price-to-Dividend-Yield (P/D) ratios, or combinations of these. Growth stocks, however, are the ones with good past performance and growth opportunities (as viewed by the market), which are concentrated on the higher end of, P/B, Master Thesis - Anar Bayramov Page 5

7 P/CF, and P/D rankings. The value premium is referred to as the difference between returns of value and growth stocks. In recent years, in order to differentiate the approach and possibly add more value to the analysis, financial agencies started to employ various techniques. Some identified stocks as pure value and growth, or with mixed characteristics, others used combinations of several ratios to classify stocks into styles. The following are excerpts from Morgan Stanley Capital International, Financial Times and Stock Exchange, and Standard & Poor s style index guides. Morgan Stanley Capital International changed its classification over the years: ( ) From 1997 to May 2003 the value and growth indices have been constructed based on a single-dimensional framework that allocates securities in a MSCI Standard Country Index into either value or growth based on their Price to Book Value ratios. Effective as of the close of May 30, 2003, MSCI will apply a two-dimensional framework for style segmentation ( ) (p. 5) Financial Times and Stock Exchange defines styles as following: Value companies tend to show good value in relation to selling price and as such companies tend to have low price-to-earnings or price-to-book (assets minus liabilities) ratios. If you are investing in a Value company you are investing on the basis of the ratio between the share price and the underlying value of the company. Growth companies are expected to exceed the average return of the market. Growth stocks are often characterized by favorable fundamental ratios such as steadily increasing revenue growth, as well as steadily increasing earnings. If you are investing in a Growth company you are investing on the basis of the anticipated growth of the company. (p. 1) Standard & Poor s uses pure style classification: The first generation of style indices from S&P Indices and Russell Investments was constructed using all the stocks in the corresponding parent index universe, grouping them into growth and value classifications. In 2005, S&P Indices introduced a unique set of pure style indices. These pure style indices employ a more robust definition of style and have improved discriminatory power to differentiate between growth and value. The Master Thesis - Anar Bayramov Page 6

8 pure style indices include only stocks with pure growth and pure value characteristics. There are no overlapping stocks ( ). (p. 1) As we can see, while the classification and classification methods differ to a certain extent, the basic principles of identification of value and growth stocks stay the same. Growing sophistication might seem confusing sometimes, but still it all depends on the fundamental analysis of the abovementioned ratios Emergence of Value Premiums Since the publication of Graham and Dodd s Security Analysis in 1934, investment professionals and scholars have debated on the superior performance of value strategies and usefulness of the fundamental analysis. Later, Basu (1977) documented that by investing in low stocks investors can obtain above average market returns. He found that the low stocks perform significantly better in terms of absolute returns by approximately 7%. Further works of Stattman (1980), and Rosenberg, Reid, and Lanstein (1984) illustrate that the US stocks with high Book-to-Market- Value (B/M) also beat the market. Chan, Hamao, and Lakonishok (1991) discover a positive relationship between the Cash-Flow-to-Price ratio and excess returns. Cross section of expected stock returns by Fama and French in 1992 uses a different approach and extends the results for the US market, once more showing the existence of and offering a rational explanation for the value premium. One of the opponents, Black (1993), explains the results in Fama and French (1992) as being interpreted incorrectly and simply a product of data mining, which cannot warrant the results in the presence of estimation error and changing risk premiums. Supporting Black s position, MacKinlay (1995) argues that on the ex post basis, by grouping assets with common disturbance terms it is possible to obtain statistically significant deviations. However, he considers such results merely a product of data snooping, and shares the view that with direct adjustment to prevent data snooping the real deviations are hardly obtainable. To test Black (1993) and MacKinlay s (1995) assertions that the value premium is sample-specific, Fama and French (1998) look at the stock returns of companies in Japan, the Master Thesis - Anar Bayramov Page 7

9 US, the UK, France, Germany, Italy, the Netherlands, Belgium, Switzerland, Sweden, Australia, Hong Kong, and Singapore from 1975 till They form portfolios at the end of each calendar year and show that the value premium exists in twelve of thirteen major markets. Furthermore, Liew and Vassalou (2000) look at security returns from international markets - Australia, Canada, France, Germany, Italy, Japan, the Netherlands, Switzerland, the UK, and the US, - and confirm previous findings of Fama and French (1998): returns on zero investment portfolios are both statistically and economically significant in most markets. Yearly holding portfolios were constructed from 1978 until and rebalanced in June. Regardless of portfolio construction and rebalancing dates, these studies confirm the existence of the value premium in most major markets. So, firstly, we want to perform a standard check on the existence of value premiums among the variety of blue chip stocks in the three largest European economies and the aggregate European market during expansions and recessions. As suggested earlier, in addition to P/B we will also run tests with two other popular measures employed by scholars and practitioners in determining value and growth stocks: and P/CF. Hypothesis 1 a. The average monthly return of low, P/B and P/CF stocks is higher than that of high, P/B and P/CF stocks for business cycle stages. Hypothesis 1 b. The average monthly return of low, P/B and P/CF stocks is equal to that of high, P/B and P/CF stocks for business cycle stages. Hypothesis 1 c. The average monthly return of low, P/B and P/CF stocks is lower than that of high, P/B and P/CF stocks for business cycle stages. 2.3 Value Premium during Expansions and Recessions One of the popular explanations of value premiums is rational variation of expected returns, as initially suggested by Fama and French (1992). According to this position, value stocks are 1 Due to the lack of observations and data for countries, other than Netherlands, UK, US and Australia, Liew and Vassalou (2000) constructed portfolios at later points that They also note that for Italy, Netherlands and Switzerland the number of securities is small for some years and results should be interpreted with caution. Master Thesis - Anar Bayramov Page 8

10 fundamentally riskier. They argue that risks associated with stocks are multidimensional after testing whether size, leverage, B/M, and Earnings-to-Price ratio (E/P) add any additional information about expected average returns. Fama and French discover that size and B/M seem to absorb the explanatory power of leverage and E/P. They also note that B/M is the strongest factor in terms of significance and persistence. They suggest the risk captured in the B/M could be the relative distress factor of Chan and Chen (1991). Risk factors are always present in the earnings prospects of a firm, and the ones considered to have the weakest prospects are identified by the market as those with a high B/M ratio and low stock prices. Liew and Vassalou (2000) investigate how strongly future economic growth and profitability of High-Minus-Low (HML) are linked: A positive relation would exist, if high returns in HML ( ) are associated with future good states of the economy. That would mean that high B/M ( ) stocks are better able to prosper than low B/M ( ) stocks when periods of high economic growth are expected. The observed positive relation between HML ( ) and future GDP growth makes sense. Presumably, investors would rather hold stocks whose returns are relatively high when they discover that the economy is in a bad state. They therefore hold low B/M ( ) stocks with good growth opportunities ( ). (p. 233) They provide evidence of a positive relationship between the two and suggest that the explanation offered by Fama and French (1993) is plausible and likely. If low P/B stocks are less able to prosper during down markets as suggested by Liew and Vassalou (2000), then value premiums should be smaller during economic recessions and stagnations than during expansions. Therefore, based on rational expectation theory we can expect that in contrast with good times, the value premium is smaller during bad times when the price of risk is even higher for risk-averse investors. In other words, we can predict that the value premium is pro-cyclical. Master Thesis - Anar Bayramov Page 9

11 Hypothesis 2 a : Value premium is negatively related to contemporaneous state of economy. Hypothesis 2 b : Value premium is positively related to contemporaneous state of economy. 2.4 Mean Reversion during Business Cycles The rational explanation for the existence of value premium is based on the theory of mean reversion in profitability. It is a process during which returns of underperforming and outperforming firms revert to a certain industry mean. The forces deemed responsible for it are competition, supply and demand. As new industries are born some of the first comers are very profitable. Such success attracts more and more new entrants who mimic and reproduce initial products, thus increasing the competition and lowering profitability of the industry. Thus profitability of all the companies in the industry tends to converge. Fama and French (2000) construct a model, which estimates mean reversion of about 38% per year. They note that ( ) mean reversion is faster when profitability is below its mean and when it is further from its mean in either direction. (p. 161). Lakonishok, Shleifer, and Vishny (1994), and Jegadeesh and Titman (1993) provide empirical evidence of mean reversion in profitability of the US firms. Further, Fama and French (2007) elaborate on the causes of the value premium: The capital gains of value stocks trace mostly to convergence: P/B rises as some value companies become more profitable and their stocks move to lower-expected-return groups. Growth in book equity is trivial to negative for value portfolios but is a large positive factor in the capital gains of growth stocks. For growth stocks, convergence is negative: P/B falls because growth companies do not always remain highly profitable with low expected stock returns. Relative to convergence, drift is a minor factor in average returns. (p. 96) Based on the abovementioned findings, we should observe mean reversion in profitability in our sample too. Moreover, if value stocks are performing better in the up markets (Hypothesis 2 b ), and are riskier than growth stocks in the down markets (Fama and French (1993)), then we can expect slower mean reversion in profitability in bad times, if any at all, and faster mean Master Thesis - Anar Bayramov Page 10

12 reversion in good times. As a measure of profitability of firms, we employ the Return on Equity (ROE) ratio. Hypothesis 3 a : The returns of the value and growth companies do not exhibit mean reversion in profitability. Hypothesis 3 b : The returns of the value and growth companies exhibit slower or no mean reversion in profitability during expansions, and faster mean reversion during recessions. Hypothesis 3 c : The returns of the value and growth companies exhibit faster mean reversion in profitability during expansions, and slower or no mean reversion during recessions. Hypothesis 3 d : The returns of the value and growth companies exhibit mean reversion in profitability of a similar intensity during both expansions and recessions. 2.5 Holding Period Returns The fact that investors might fail to recognize the mean reversion in profitability can result in misestimation of future earnings. This potentially makes value strategies more or less attractive for different time horizons. If the returns of firms are mean reverting in profitability (Hypothesis 3 b, 3 c, 3 d ) and converge to a market average at some point, then we can further hypothesize that the return difference of value and growth portfolios should be less for relatively longer holding periods in both good and bad times. Hypothesis 4 a : The value premium is economically more significant for relatively longer holding periods than shorter periods during both expansions and recessions. Hypothesis 4 b : The value premium is economically more significant for relatively shorter holding periods than longer periods during both expansions and recessions. Master Thesis - Anar Bayramov Page 11

13 2.6 Risk and Reward of Value and Growth Strategies Another popular explanation for the causes of value premiums is an irrational investor behavior. Adherents of this view suggest that investors are overly pessimistic about the future prospects of value stocks and overly optimistic about growth stocks. This is due to extrapolation of past performance into the future. As the information about performance arrives, investors start making adjustments to the views. Therefore, the prices start reverting to their actual values prices of stocks in the high spectrum decrease and prices in the lower spectrum increase. Supporting this view, Lakonishok et al (1994) argue that the irrational behavior is not associated with an extra amount of risk. This result undermines the rational explanation offered by Fama and French (1993). They report that the betas of value stocks are higher than those of growth stocks in good times and lower in bad times in the sample for the US market. While Petkova and Zhang (2005) refine the earlier findings in previous research by showing that the value betas tend to co-vary positively with the expected market premiums, while the growth betas co-vary negatively. The findings in the previous research discussed are contradictory. We will also look at the risk and reward measures of value and growth portfolios. This will help us to determine the relative attractiveness of the two styles. If the rational explanation holds, then we can expect that during bad times value stocks have higher volatility and beta than growth stocks. Hypothesis 5 a : The low, P/B and P/CF stocks markets have lower volatility and beta than the high, P/B and P/CF stocks during recessions. Hypothesis 5 b : The low, P/B and P/CF stocks markets have higher volatility and beta than the high, P/B and P/CF stocks during recessions. Furthermore, if the value premium is indeed pro-cyclical (Hypothesis 2 b ) and growth stocks tend to co-vary negatively with the market (Petkova and Zhang (2005)), we can further hypothesize that Treynor s and Sharpe risk-reward measures should be higher for growth stocks than value stocks during down markets. Master Thesis - Anar Bayramov Page 12

14 Hypothesis 6 a : Low, P/B and P/CF stocks perform better than high, P/B and P/CF stocks in terms of Treynor s measure and Sharpe ratios during recessions. Hypothesis 6 b : High, P/B and P/CF stocks perform better than low, P/B and P/CF stocks in terms of Treynor s measure and Sharpe ratios during recessions. III. Data and Methodology 3.1 Sample Selection The initial set of stocks in the dataset consists of the securities included in the FTSE Developed Europe ex UK Index and FTSE 100 Index for UK. Table 1 presents information about the number of companies per country and source indices. The FTSE Developed Europe ex UK Index consists of Large (45%) and Mid (55%) capitalization blue chip stocks of 15 developed markets in Europe excluding the UK. To include the UK market the largest in Europe we also add the FTSE 100 index stocks. The latter is a market capitalization weighted index of blue chip stocks from the UK stock market. Such sample selection has several advantages. The FTSE Developed Europe ex UK index and FTSE 100 are representative benchmark indices of corresponding economic regions. They represent the investment grade securities that are of primary interest to global investors. Moreover, to ensure attractiveness, constructing agencies screen constituent equities for the following features: - Investability securities are free-float weighted to ensure investment freedom by potential investors; - Liquidity stocks are checked to ensure index tradability - Transparency Large and international investors usually do not invest heavily in securities that are illiquid and not closely watched by the market analysts. Securities of blue chip companies are especially appreciated during the times of economic depressions because they are considered to have the lowest bankruptcy risks among all classes of securities. Master Thesis - Anar Bayramov Page 13

15 Taking into account developments in the field, we perform the classification to value and growth within the index universe. Currently such practice is very common. S&P, MSCI and other financial service agencies construct value and growth indices using the stocks in the parent index universe, as discussed in Section I. The end product of this practice is a sub-index like the S&P 500 Growth Index and S&P 500 Value Index, MSCI EAFA Growth Index and MSCI EAFA Value Index, etc. We obtained all the information on stocks from December, 1981 till June, 2012, from Reuters Thomson Datastream Database. Information on the index constituents was kindly provided by FTSE. FTSE indices are rebalanced quarterly. The last rebalancing for both indices before the start of our research took place in March, So the FTSE Developed Europe ex UK as of April, 30, 2012, consisted of 396 ordinary and preferred stocks. 14 companies (Volvo, Henkel, Exor, Volkswagen, etc.) were represented by two classes of stock. Due to a high correlation and similarity of variables in different classes of the same company equity, only one class, the major security according to Datastream classification, was selected. This approach resulted in dropping 14 securities. On the other hand, the FTSE 100 consisted of 100 stocks, and in the presence of no similarity issue all stocks were included in the final sample. Sometimes financial companies are dropped in the analysis of value premiums due to debt and equity structures, which may differ significantly from non-financials. We do not exclude financial companies, as Barber and Lyon (1997) have shown it is possible to pursue value strategies in the universe of financial companies. The following are descriptions of monthly variables, also obtained from Datastream; all variables are in EURO currency or percentage terms: - Price represents the official closing share price - P/CF share price divided by the cash earnings per share for the appropriate financial period - share price divided by the earnings rate per share at the required date - P/B share price divided by the book value per share - Return on Equity based on a trailing twelve month period if applicable and represents the sum of the relevant item reported in the last twelve months Master Thesis - Anar Bayramov Page 14

16 - Market Capitalization share price at the year end multiplied by the number of common shares outstanding Extreme ROE outliers, constituting observations below the 0.5 th percentile and above the 99.5 th percentile, were dropped. For the calculation of market betas, alphas, and risk-reward measures, information on the risk free rate and the market return proxy were also obtained from Datastream. As a proxy for the risk free rate we used the Money market rate 3-month Frankfurt Banks Middle rate, and for the market return Total Return index on MSCI Europe index was taken. Datastream defines Middle rate and Total Return index as follows: - Middle rate midpoint between the bid and offered rates - Return Index return indices for interest rates show a theoretical growth in the value of an interest rate deposit. The index is based on an initial value of 100 at the start date. Thereafter it is assumed that, each calendar day, the value of the deposit is increased by a percentage equal to 1/365 times the value of the interest rate. France, Germany, and the UK are the largest stock markets in Europe and the historical information available on the company securities of these countries allows us to test several business cycles and identify patterns. For other large economies of the region, such as Switzerland, the Netherlands and Sweden, it was impossible to conduct the tests because of a lack of historical information on securities. With the available information we are able to look at only a few cycles, and cannot arrive at solid conclusions. From Table 2 we can see that even for large economies the business cycles do not exactly coincide. However, the growing economic convergence in Europe allows us to consider Europe as one regional stock market in a business cycle framework. Mink et al. (2008) argue that the euro area business cycles exhibit more similarities than the regional business cycles in the US. They also suggest that though not all the countries exhibit same patterns in business cycles, most do, and differences between them have declined from the 1990s onward. And since we are able to obtain information on the business cycle dating for the 16-country European region, we also tested all abovementioned hypotheses from a single European market standpoint. We believe this Master Thesis - Anar Bayramov Page 15

17 is an interesting perspective from an international investor s point of view and may unveil points for future research. Table 2 shows the business cycle peak and trough dates for the markets tested, and corresponding sources of the information. The Economic Cycle Research Institute (ECRI) and the Center for Economic Policy Research Business Cycle Dating Committee (CEPR), two highly regarded institutions among practitioners and scholars, provide business cycle dating. Both ECRI and CEPR use leading macroeconomic indicators for determining business cycle turning points. CERP (Dating Committee Findings 22 September 2003) determines a recession as: ( ) a significant decline in the level of economic activity, spread across the economy of the euro area, usually visible in two or more consecutive quarters of negative growth in GDP, employment and other measures of aggregate economic activity for the ( ) area as a whole, and reflecting similar developments in most countries. A recession begins just after the economy reaches a peak of activity and ends when the economy reaches its trough. Between trough and peak, the economy is formally in an expansion; between peak and trough it is in a recession. A recession begins just after the economy reaches a peak of activity and ends when the economy reaches its trough. Between trough and peak, the economy is formally in an expansion; between peak and trough it is in a recession. In both cases, growth rates may be very low. (p. 1) 3.2 Methodology This section provides information on primary techniques used in the empirical analysis. We start with formation of the portfolios for each of the three ratios, which lays a foundation for our further analysis. Then we perform the significance and mean reversion tests and calculate the risk and reward measures for value and growth portfolios Filtering and Sorting Depending on whether the dating is in months or quarters, we form 16 portfolios at the start of every month or quarter following the peak and trough dates. We do this for each of the regions presented in Table 2. Firstly, we sort the stocks from highest to lowest according to the market capitalization and assign them into size quartiles. Next, in each size quartile we sort stocks by Master Thesis - Anar Bayramov Page 16

18 ratio and group them into quartiles. These steps are done for each of the three ratios. In each size quartile, companies in the top quartile of the P/B, and P/CF rankings are classified as growth, and those in the bottom quartile are classified as value. The reason for two-dimensional sorting is that the value premium might be due to the size of firms, as smaller firms are producing higher returns than large cap firms (Fama and French (1992)). Bauman et al. (1998) also show that small capitalization stocks in different markets tend to yield higher returns than large cap stocks. Therefore, such sorting is employed in order to control for size effect. Portfolios contain roughly the same number of stocks. When the number of stocks in a size quartile is odd, some portfolios are assigned a larger number of stocks depending on the value of the ratio. Also, due to the lack of observations in early business stages some portfolios may contain as few as two stocks. To be included in the sample for the business cycle stage, the company should have market capitalization information, a positive ratio at the time of formation of the portfolio, and be traded on the last day of the month. Exclusion of companies with negative ratios should have a minimal effect on returns, as shown by Basu (1977). In the sample we have companies with different fiscal years. Fama and French (1992) find that combining companies with various fiscal years does not affect the results. Post formation average monthly returns are calculated for each of the portfolios for the whole length of the cycle - six month, one, two, three, four and five year holding periods. This approach assumes a simple buy and hold strategy Significance Tests and Mean Reversion For each of the portfolios, one sample and individual sample t-tests are conducted to examine the significance of individual portfolio returns and value premiums for the abovementioned holding periods during each business cycle stage. Holding period returns for some cycle stages, the duration of which exceed the lengths of the tested period, are not calculated. Master Thesis - Anar Bayramov Page 17

19 Firstly, we perform Levene s test to identify whether samples have equal variance. Based on the results of the Levene s test, individual sample t-tests, either with equal or unequal variances, are run to estimate statistical significance of the value premium. As a proxy of profitability we use ROE and calculate average ROE for growth and value portfolios for each of the years following portfolio formation. Then, we perform individual sample t-tests of ROE during each of the subsequent years to estimate the difference in profitability after portfolio formation. We perform this for each of the four regions. For expansion we average ROE for each of the seven following years, for recessions and stagnations stages for three years (due to shorter duration of the contraction stages) OLS Regressions and Risk-Reward Measures Performance should not be evaluated only in terms of risk, but also reward. We calculate riskreward measures with respect to every ratio for expansions, recessions and whole sample periods. Necessary variables for calculations of risk-reward measures are obtained by running the following CAPM based regression separately for expansions, recessions, and the whole sample period: (1) where is the return of value or growth portfolio in month t; is the risk-free rate in month t; is the return of market in month t; is the estimated intercept (which is also Jensen s alpha); and is the estimated slope. Treynor s measure shows the excess return obtained for every unit of systematic nondiversifiable risk. By using beta from regression (1), it is calculated as follows: (2) where is the average return of style portfolios, is the average risk free rate. total risk: In contrast with Treynor s measure, Sharpe ratio relates excess return per each unit of Master Thesis - Anar Bayramov Page 18

20 (3) where is the standard deviation of excess returns. IV. Empirical Results 4.1 The Value Premium Tables 3, 4, 5 and 6 present the returns on value and growth portfolios for business cycle stages for France, Germany, the UK and Europe, respectively. Regardless of length, each stage is treated as one holding period. Results in Table 3 confirm the presence of a value premium in the French market. Returns on most value and growth portfolios are statistically significant at conventional levels. From the results in Table 3 we can also see a decrease in the historical value premium in economic terms. If the average value premium for one stage was 0.72%, for any ratio in the first complete business cycle in our sample (consisting of expansion followed by recession), it equals -0.19% for the last business cycle. For all ratios, except three expansions and one recession, the returns are economically significant. We also must note that, in general, the returns for low portfolios have higher standard errors than those for high portfolios. P/B seems to be a better predictor of the value premium for French blue chip stocks. In Germany we observe the same tendency regarding test statistics higher for low portfolios. Economic significance over time is relatively constant. Though the ratio predicts positive value premiums on more occasions than the P/B ratio, the average value premium for based portfolios is 0.51% and 0.98% for P/B 2. For the UK market the difference in returns for most cycle stages is either insignificant or significantly negative in economic terms. High mean reversion could explain economic insignificance. Later we will look at the holding period returns, and see if shorter holding periods could help us obtain a positive value premium. These results suggest that among a range of blue chip stocks, growth stocks are better performers than value stocks at any time in the UK. In 2 for portfolios average with difference higher than one basis point Master Thesis - Anar Bayramov Page 19

21 contrast with what we observed for Germany and France, test statistics for high portfolios are at least as high as those for low portfolios. For Europe, possibly due to increased number of observations, we observe that the portfolio returns are statistically as well as economically significant on more occasions. is a better predictor of the value premium. Until the recession of 2008 the value premium is positive and pervasive. After the start of the financial crisis, we do not observe any positive difference between the average returns of value and growth portfolios. The same could be said for France and the UK. One of the possible explanations from an irrational behaviour point of view is that when the financial crisis started, investors trying to maintain positive returns and assumingly extrapolating past positive premiums into the future, fled to value stocks, thus increasing the demand and causing an increase in prices. So later when the market converged to equilibrium, the decrease in price of value stocks was larger due to the preceding spike. Alternatively, a rational explanation could be that since value stocks were always riskier, smaller returns for value stocks are just a result of the higher risk they bore and the more severe the economic recession, the lower the returns on low ratio stocks. In three of four markets tested, the value premium is present during the whole business cycle stage, whilst somewhat weaker for the French market: the low ratio stock based portfolios perform better than the high ratio based ones during business cycle stages. Thus, we fail to reject Hypothesis 1 a. This evidence supports the earlier findings of Fama and French (1992, 1998) and refutes the stance of Black (1993) and MacKinlay (1995), who considered value premiums to be sample specific and a result of data snooping. 4.2 Cyclicality of Value Premium Next, we look at the relationship between the state of the economy and value premiums. The results are presented in Tables 7, 8, 9, 10, and plotted in Figures 1, 2, 3, and 4 for France, Germany, the UK and Europe, respectively. For each ratio we performed individual sample t- tests of style portfolios returns during each cycle stage. Master Thesis - Anar Bayramov Page 20

22 Findings for France confirm our prediction for P/B based portfolios. Value premium during each recession is strictly lower than that of a successive expansion. and P/CF based value premium seems to follow the same trend in a relatively smoother pattern. For Germany, as noted earlier, the difference of the low and high P/B based portfolios is higher in absolute terms, although the ratios show a higher relation to business cycles. In the UK and Europe, both and P/B ratios are almost equally strongly related to business cycles. With the exception of a few expansions and recessions, we see that during recessions value premiums usually go down and then increase during expansion. This trend is common for the UK as well, although in the UK the value premium is mainly negative. From a statistical point of view, the difference between the returns of low and high portfolios is not significant for the most part; however they are significant from an economic point of view. Fama and French (1998) also get a low standard error and as they put it This is testimony to the high volatility of the country returns. The market returns of many countries have standard deviations of approximately 30 percent per year, about twice that of the global market portfolio ( ). (p. 1981) In all four markets we observe a positive correlation between the value premiums and economic cycles for and P/B based portfolios. The value premium suffers during economic downturns and goes up during recoveries. These results support our prediction of value premiums being pro-cyclical, and on the basis of these we reject hypothesis 2 a. 4.3 Post-formation Profitability Tables 11, 12, 13 and 14 present post-formation fundamental performance of companies in France, Germany, the UK and Europe. During expansions in France we observe faster mean reversion in profitability for the P/CF based portfolios. Up to year 5 high P/CF stocks outperform low stocks. In year 5 the difference between low and high stocks average ROE is already positive 0.17 and in the subsequent year increases to In terms of speed of mean reversion P/CF based stocks are followed by based. The P/B ranked stocks demonstrate the slowest mean reversion rates Master Thesis - Anar Bayramov Page 21

23 from all three. For expansions overall we observe convergence around year 5. This is true for all three ratios. During recessions the gap in performance is actually widening or at least staying the same. For P/B stocks during the first year after portfolio formation, value underperforms growth by 13.15%, whereas in the third year the difference increases up to 22.09%. The results are statistically significant at a 1% level. As with expansions, during recessions we observe similar performance for with an increase in the profitability gap during the second year and a decrease during the third. For P/CF the difference remains almost the same only a 0.19% relative improvement for value. The results for France are in line with our prediction we observe a rapid mean reversion in the profitability during good times and slow or almost no reversion during bad times. Expansions in the UK are associated with the smallest performance difference around the 2 nd and 3 rd years, and increase thereafter. However it never quite converges or flips the sign. For recessions as it is the case in French market, Value minus Growth in ROE increases monotonically for all three ratios: for the increase in difference from year 1 to year 2 is 5.93%, for P/B 2.14%, and for P/CF 4.45%. Tests for Germany and Europe provide evidence of a mean reverting process for both states of the economy. For expansions, convergence in all three ratios takes place around year 6 in Germany and year 7 in Europe. In year 3 in Germany the difference falls from 2.31% to 0.68% for the stocks, for P/B from -8.34% to -6.64%, and for P/CF from -3.30% to -1.84%. The European market shows a similar pattern. Hence, in the UK and France we observe a faster convergence during expansions, and a small or almost no convergence during recessions. In addition, findings for Germany and Europe show a rather swift mean reversion during recessions for the and P/CF ranked stocks. The results for all the ratios support our prediction except the two occasions for and P/CF stocks. Based on this, we reject Hypothesis 3 a, 3 b, and 3 d, and fail to reject Hypothesis 3 c : The returns of the value and growth companies exhibit faster mean reversion in profitability in expansions, and slower or no mean reversion in recessions. Master Thesis - Anar Bayramov Page 22

24 4.4 Holding Period Returns. Since our results have shown that, indeed, stocks exhibit a faster mean reversion during expansions than recessions, it is further possible to test our predictions on holding period returns. In Section 2.5 we have predicted that the value premiums should decrease over long periods because stocks tend to revert to a market average. Findings on this are presented in Table 15 for France, Germany Table 16, UK Table 17, and Europe Table 18. During expansions in the French market, when the value premium is positive, we observe the largest positive difference at different points of the first two years of portfolio life. From 1/1/1985 to 2/28/1992 the value premium increases from a six month holding period onwards, reaching its peak at twelve months, and then starts to decrease. The pattern is similar for all three ratios and statistically significant at a 1% level. For P/B during 6/1/2003 to 2/28/2008 expansion the difference is most economically significant in two year holding period at a 1.76% monthly, statistically significant at 1% level and P/CF during twelve month with a 1.36%; statistically significant at 10% level. Only the stocks see the maximum premium during the three year holding period with a 0.58%. For the last expansion the value premium is highly significant during the first six months a rocking 8.64% monthly for P/B, 1.11% for and 3.81% monthly for P/CF; P/B and P/CF value minus growth difference is statistically significant at 1% and 10% levels. The findings for recessions are somewhat vague. If the value premium is positive during the first six months, it tends to turn negative as time passes. However, when the difference is negative, it tends to increase at a slow speed. This is true for the and P/CF ratios for all contraction periods and in half of the cases for P/B. These results can be explained by the mean reverting process. In France we observed slow or no mean reversion during recessions. So as time passes, value stocks tend to perform worse, or at least proportionately as bad as growth stocks. In the first case this results in a decrease of a positive value premium, in the second it leads to an increase when the difference is negative. Both values move towards the equilibrium with similar returns. Master Thesis - Anar Bayramov Page 23

25 In Germany and Europe we see similar patterns. The value premium has the maximum value during the first six or twelve month holding periods. If differences are negative, with time they tend to move towards the positive values. Negative values during recessions are more likely to turn positive, unlike in France. Possibly, this is due to the fact German and European stocks exhibited at least as fast or faster mean reversion in profitability during recessions. In the UK the value premium is almost always negative at any holding period. This rejects our earlier assumption that negative premiums during business cycles in the UK are due to faster mean reversion, and the differences between low and high stocks should be positive on shorter horizons. Indeed stocks exhibited reversion up until the Year 3. None of the holding periods yield consistently positive value premiums. Only a few two and three year holding periods do so, but are statistically insignificant. When the value premium is economically significant and remains so for several consecutive holding periods, we observe higher standard errors. In France, for example, in two out of tree expansions the differences are statistically significant on a conventional level up to five years. This means that characteristics of the business cycle stages affect the significance and persistence of value premiums. The stronger the recovery, the more likely we will observe a persistently positive value premium. On the basis of the discussed results we are able to reject Hypothesis 4 a and fail to do so with Hypothesis 4 b : The value premium is economically more significant for relatively shorter holding periods than longer periods during both expansions and recessions. 4.5 Risk and Reward. In this section we look at the risks and rewards of value and growth styles during business cycles. Tables 19, 20, 21 and 22 illustrate findings for France, Germany, the UK and Europe, respectively. of growth portfolios is always lower than of value portfolios for French business cycles. Difference between the two is the largest in the case of P/B stocks during recessions high P/B is 0.88, while for low stocks it is The values are significant at a 1% level. The Master Thesis - Anar Bayramov Page 24

26 difference of 0.45 shows value stocks are way more sensitive to times when the price of risk is high. For and P/CF the difference is about 0.2. Volatilities are higher for value stocks during recession, and the largest difference is 1.29% for the P/B ranked stocks. The UK stocks demonstrate similar findings: higher betas and volatilities for the value. The difference in volatilities of monthly returns for an average ratio is higher for the UK than for France. The low stocks demonstrate a standard deviation of 10.08%, low P/B 9.27%, and low P/CF 10.18%, while high, P/B and P/CF stocks have 7.48, 7.23, and 7.80%, respectively. Surprisingly, we find opposite results for Germany. Volatility and beta of growth portfolios is higher during recessions for any of the three ratios. The standard deviation of high P/B stocks stands at 10.48%, whereas for low P/B it is In our sample, returns on German blue chip stocks are more volatile than on French ones. For P/B and beta estimates are higher for value stocks during recessions, and approximately equal for P/CF. The estimate of slope from the regression (1) is statistically significant at all times in every market. The beta estimates and standard deviations of returns in French, British, and European markets support our prediction. Opposite results for the German stocks could be due to country specific characteristics of growth companies in the country. Based on supportive findings in three out of four markets, we conclude that the low ratio stocks in Europe are generally more volatile and have higher betas than high ratio stocks. Therefore, we reject Hypothesis 5 a : Low, P/B and P/CF stocks markets have lower volatility and beta than high, P/B and P/CF stocks during recessions, and thus, we fail to reject alternative Hypothesis 5 b. These results are in line with the risk based explanation offered by Fama and French (1996). We also analyze risk-reward measures Sharpe and Treynor. Based on cyclicality, we predicted that growth portfolios should have better risk-reward characteristics, due to the higher riskiness of the value stocks in states of economic disturbance. Master Thesis - Anar Bayramov Page 25