Dimension and Book-to-Market Ratio Again The English Case

Pedro Rino Vieira ISEG Technical University of Lisbon rinovieira@iseg.utl.pt

Jos Azevedo Pereira ISEG Technical University of Lisbon jpereira@iseg.utl.pt

WORKING PAPER N. 5/2006 September 2006 Abstract Using the Fama-French Model (1993) Daniel and Titman (1997) show that size and bookto-market effects cannot be understood as distress factor proxies, but as characteristics that explain the cross section variation in stock returns. Davis et al. (2000) refute these results using a different set of data. While addressing this question, we have found unexpected evidence against the Fama-French Model in the UK market and challenging results regarding the size and book-to-market effects in both the UK and USA. Our findings, at the very least, suggest a bad CAPM specification and, at most, suggest that financial markets are not efficient. Key words: Behavioural Finance, Size Effect, Book-to-Market Effect, CAPM, Efficient Market Hypothesis, Financial Investments JEL Classification: G12, G14

Department of Management Working paper Series ISSN 0874-8470

Mainstream finance advocates that financial markets are efficient and their investors fully rational in the Efficient Markets Hypothesis (EMH) sense. Being fully rational means two things. In the first place, investors accurately and continuously update in their decision decision-making framework on the basis of new data made available in the market. Secondly, given this updated setting, investors try to take the best decisions in order to maximize their utility. This is a simple and appealing vision of the world that is particularly useful in theoretical modelling exercises. Unfortunately it is only half the story, as recent work from neurobiology tells us (Damsio, 2003). Actually, Damsio explains that human beings have two complementary ways of making decisions. Way A, the fully rational one, leads to some mental images being projected in our minds, such as options for action and their future consequences. Based on these images we come out with a decision through different reasoning strategies. However, at the same time, we recall ancient emotional memories lived in similar situations, which is way B. This recall, conscious or not, influences our decision making process conditioning our analysis of the situation. If a decision made in a previous situation has provoked some pleasant outcomes and feelings, in a similar current situation we will tend to make a decision without comprehensive rational reasoning that takes into account the differences of the new context. If the framework has changed, most likely the decision taken will not be the best or the same as that we would reach using the rational way alone. Most likely, phenomena such as representativeness, anchoring or availability biases found by Kahneman and Tversky (1974) are the result of this way B (Damsio, 1974). This, and further research on psychology1 raises, at least, profound doubts about the hypothesis that investors are fully rational and the market is efficient in any of the three forms considered by Fama (1970). So, it can be expected that EMH models, namely the Sharpe-Lintner-Black (SLB) or Capital

For example Alpert and Raiffa (1982), Fischhoff et al. (1977), Buehler et al. (1994), Kahneman and

Tversky (1974 and 1982), Edwards (1968), Lord et al. (1979), Camerer and Hogarth (1999)

Asset Pricing Model (CAPM) [Sharpe (1964), Lintner (1965) and Black (1972)], do not apply properly to real world situations. Not surprisingly several Behavioural Finance researchers have found several anomalies in the CAPM. Banz (1981) found that shares issued by big firms tend to produce lower returns than shares issued by small companies. Keim (1983) confirmed this result. This is the size effect. Basu (1983) analyzed the relationship between earnings yield, market value and return for NYSE common stocks. He confirmed the size effect and found that shares issued by firms with higher earnings yield have had higher returns. De Bondt and Thaler (1985, 1987), in a more behavioural study, discovered that most people overreact to unexpected and dramatic news events in such a way that, in the long term, portfolios of prior losers are found to outperform prior winners. Jegadeesh and Titman (1993) analyzed this same question, finding that in the short term, prior winners are still winners and prior losers are still losers. Rosenberg et al. (1985) discovered that average returns on U.S. stocks are positively related to the book-to-market ratio. Bhandari (1988) documented a positive relation between leverage and average return. All these results that have challenged the SLB model might have two meanings. One possibility is a misspecification of the SLB model, which can be addressed with a more sophisticated one, perhaps with a multi-factor model based on the work of Merton (1973) and Ross (1976). Another and more radical possibility is to assume that those results imply the inefficiency of the financial markets, which sounds appealing under the Behavioural Finance approach. To address this issue, Fama and French (1992, 1996) examined all these variables and concluded that the cross-sectional variation in expected returns can be explained by size and book-to-market, while , the traditional SLB model measure of risk fails to explain the return dispersion. Fama and French (1993) defend that if higher returns are compensation for higher systematic risk then the book-to-market and size are proxies for distress and distressed firms are riskier because they are more sensitive to certain business cycle factors. So, the EMH still applies, even if the SLB model suffers from misspecification. Fama and French (1993) run several tests that support their claims. First they show that prices of high book-to-market and small size stocks tend to move in a way that supports the idea of a common risk factor. Secondly, they developed a multi-factor model in the sense of Merton (1973) and Ross (1976) with three factors that explain the excess returns of a full set of book-to-market and size sorted portfolios. These factors are 3

(1) size (a small capitalization portfolio minus big capitalization portfolio they called SMB); (2) bookto-market ratio (a high book-to-market portfolio minus a low book-to-market portfolio they called HML); (3) value-weighted market portfolio (Rm). The Fama-French model is:

Ri R f = bi (Rm R f ) + s i SMB + hi HML

(1)

Fama and French (1993) support the EMH, but cannot explain the economic role of size and book-tomarket. They state that size and book-to-market are proxies of risk, but they cannot detail this relationship or describe the specific kind of risk that is measured. Under the Behavioural Finance paradigm and taking into account the teachings of neurobiology and psychology we might want to consider different explanations. Lakonishok et al. (1994) suggest some investors tend to get overly excited about stocks that have done very well in the past and buy them up, so that these glamour stocks become overpriced. At the same time, they overreact to stocks that have done very badly, overselling these value or high book-to-market stocks and they become underpriced. So, strategies that invest disproportionately in these underpriced stocks and underinvest in stocks that are overpriced outperform the market, even when those are not fundamentally riskier. This explanation challenges the EMH, but does not deny a relation between systematic risk and return or the supposition that the return premium of high book-to-market and small size stocks can be explained by a factor model. However, we can ask a more fundamental question: are the return patterns of characteristic-sorted portfolios really consistent with a factor model at all? This question was raised by Daniel and Titman (1997), who wanted to know whether the high returns of high bookto-market and small size stocks can be attributed to their factor loadings. In short, Daniel and Titman (1997) question if relative distress drives stock returns and if book-to-market is a proxy for relative distress. The underlying idea is that low book-to-market (strong firms) produce low stock returns and high book-to-market stocks (distressed firms) have high returns, regardless of risk loading. Their results indicate that (1) there is no discernible separate risk factor associated with high or low book-to-market firms, and (2) there is no return premium associated with any of the three factors identified by Fama and French (1993), suggesting that the high returns related to these portfolios cannot be viewed as compensation for factor risk. Actually, they found that although high book-to-

market stocks do covary strongly with other high book-to-market stocks, the covariances do not result from there being particular risks associated with distress, but rather reflect the fact that high book-tomarket firms tend to have similar properties. Responding to this study, Davis et al. (2000) use a different approach for a much longer time period. While Daniel and Titman study returns from July 1973 to December 1993, Fama and French use data from July 1929 to June 1997. They confirm Daniel and Titmans results for the 1973-1993 period of time. However, when they use the entire sample the empirical evidence supports the Fama-French Model. Once again we have contradictory results. In this case, we have the same phenomenon but studied with different methodologies and samples, which might weaken the findings, since a good model should apply in all situations (Kothari et al., 1995). In summary, there is an open debate centered on whether these factors can possibly represent relevant aggregate risk. In this article, while readdressing this question we have found unexpected and surprising results. We found evidence that (1) the book-to-market effect is different between the British and American samples, (2) in the USA sample, the relationship between returns and volatility is negative (higher returns are associated to lower volatility and thus lower risk), (3) the market factor of the Fama-French model is the only one able to explain the UK firms returns and (4) size effect and book-to-market are strongly correlated. In Section I we present the data. In Section II we characterize and re-analyze the size and book-tomarket effects in the UK and USA, comparing the results and calculating the correlation between size and book-to-market. In Section III we analyze the Fama-French model and the variables used by them, including their excess return explanatory power. We also calculate the correlation between SMB and HML. In Section IV we present the main conclusions of our work.

Data

The London Stock Exchange sub-sample data used in this study was retrieved from DataStream and covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. Overall, we considered 2901 firms. Because not every firm is quoted every 5

year, the number of stocks analyzed in each year varies along the period. We considered 887 firms in the first year and 1892 in the last one. The average number of firms listed is each year is 1478. The book-to-market ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. According to Fama and French (1993), the end of June is used as the portfolio formation date to guarantee that the book-equity value for year t-1 is public information. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were reranked in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and a ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. For each portfolio we computed the weighted-average return. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we computed the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return. For the Fama-French model study, we followed the Fama and French work (1993) and formed six new portfolios based on the intersection of the three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%) and two size categories (Small and Big with breakpoint at 50%). These portfolios are designated LS, MS, HS, LB, MB and HB. With these we formed two new portfolios to capture the book-to-market and size effect, as proposed by Fama and French (1993). The first portfolio is the SMB (small minus big) that captures the size effect and is the difference between the return of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. The second portfolio is the HML (high minus low) and captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios. 6

II The size and book-to-market effects Following the work of Daniel and Titman (1997) we first analyzed the size and book-to-market effects in the English market. The results summarized in Table 1 confirm the size effect found in the American sample. Actually, smaller firms show a monthly return 1.3% higher than bigger firms, showing also higher risk measured by the standard deviation (differential of 5.37%). The return premium does not seem too high for this risk level, although this question is not directly addressed here. The first unexpected result is related to book-to-market. The results suggest that there is no BE/ME effect in the UK, since no linear relationship was found between return and book-to-market. These results are statistically significant (t-statistics above 3 to size effect and 5 to BE/ME effect). Table 1 Descriptive statistics of the 25 portfolios - UK Taking into account that the size and book-to-market effects were first reported in the eighties and that people are not totally rational and use short-cuts in their decision-making processes, we might expect to find a change in investor behaviour related to a different understanding and usage of book-tomarket ratios. Therefore, the period under study was divided into two smaller periods of ten and nine years respectively. The first goes from July 1983 to June 1993 (first decade) and the second from July 1993 to June 2002 (second decade). Panel B of Table 1 shows that, in the first decade, the results are still surprising. Comparing the extreme portfolios we verify a book-to-market effect, but when we consider the intermediate portfolios, no relationship could be established. Apparently there is no relationship between risk and return. The size effect is verified in the first decade, although without a full linear relation between return and risk. In the second decade, the book-to-market effect has reversed with firms with low BE/ME showing higher returns than firms with high BE/ME (1.58% higher). Once again, the relationship between risk and return is not totally linear, but tends towards the expected behaviour from a rational perspective. So, the risk profile of these companies has changed and then the return profile has also changed, but not the book-to-market, which is really striking. What phenomenon might explain this modification? Why are firms with high book-to-market now more risky?

Quintiles allow the analysis to be fine tuned, but may hide some general trends. Consequently, the same analysis was performed on the Fama-French portfolios. Table 2 shows the results, which are similar but not the same. Size effect is confirmed. Across the period the reversal in the book-to-market effect seems to be confirmed (lower BE/ME firms earn a monthly higher return of 0.25%). Table 2 Descriptive statistics of the Fama-French portfolios - UK However, in the first decade, the BE/ME shows the expected behaviour with a monthly return premium of 0.45% associated with higher BE/ME firms. However, in the second one, the results have reversed. Firms with lower BE/ME earn a monthly return premium of 0.9%. These results are statistically significant, but not as strong as those obtained with the use of 25 portfolios. The usage of these two different sets of portfolios does not alter the main conclusions in the period considered, but it seems that some kinds of portfolio are more appropriate to illustrate certain results than others. These results apply to the UK market and contradict the findings reported for the United States. Nevertheless, these different studies use samples that cover different time periods. We therefore extended our analysis to the US market with a sample that covers the Amex, Nasdaq and NYSE (Table 3). During the overall period, as well as in each decade, the BE/ME effect is as expected and the results are statistically significant. What is odd is the size effect. The biggest firms with lower book-tomarket have higher returns than small firms with lower BE/ME, but small firms with high book-tomarket have higher returns than big firms with high BE/ME. Thus, the evidence from the complete sample does not support size effect. Also interesting is that in the first decade we can observe a size effect that is the opposite of what could be expected, and, in the second decade, the size effect corresponds to expectations. The major surprise is, however, the relationship between return and volatility. In the BE/ME portfolios, the stocks with lower returns show consistently higher volatility in all time periods considered. In the size portfolios this relation is not clear. During the first decade, a negative relationship was found and by contrast, in the second decade, we found a positive relationship, as would be expected by the EMH. If volatility measures risk and the data are good, these results clearly harm the EMH.

Table 3 Descriptive statistics of the 25 portfolios - USA As done for the LSE, we also studied the Fama-French portfolios in the USA. Table 4 summarizes the results, which are similar to those observed in the 25 portfolio analysis. Firms with higher BE/ME show consistently higher returns and lower volatility than firms with lower BE/ME. Size effects still exhibit contrary behaviour from one decade to the next. Big firms show higher returns in the first decade and small firms have higher returns in decade two. In all periods of time, small companies with lower BE/ME presented lower volatility, independently of their returns. Table 4 Descriptive statistics of the Fama-French portfolios - USA This relation between return and volatility is a puzzle and challenges the EMH. It just seems that risk is not relevant in investment decisions any more, although these results could be spurious. Size effect and book-to-market measure different things, but have one common variable: market equity, which means that these two effects are correlated. This correlation might provoke some unexpected results. To study correlation the natural logarithm of ME and BE/ME is used following Fama and French (1992), since in this way we can better capture the underlying effects. Table 5 shows the existence of a negative correlation between size and book-to-market (-0.299 across the period). This correlation is not extreme, but it might be enough to destroy their ability to together explain the cross section returns. When we analyze the 25 portfolios, the correlation seems to be stronger, about 0.5, meaning that these two variables are highly correlated. Multi-factor models like the Fama-French might loose explanatory power due to this high correlation. Table 5 Annual correlations between size and BE/ME Table 6 Annual correlations between size and BE/ME for the 25 portfolios III The Fama-French Model From the previous discussion we know that the BE/ME effect in UK exhibits strange behaviour when compared with the expected results. Its relation with volatility is, at the very least, peculiar, and the correlation between market value and book-to-market ratio is considerable, although not extreme. These results have obvious consequences for the Fama-French model. If the variables are correlated the model will suffer from autocorrelation. If the non-linear relation between the variables and 9

volatility discovered in this sample were verified in n different samples of the same universe, then size and book-to-market could not be risk proxies. Thus, the Fama-French model seems to be in trouble. Therefore, instead of following our intended path, it was necessary to test the model for the UK. The results are as follows: To test the model we ran the regression

Ri R f = ai + bi ( Rm R f ) + si SMB + hi HML + i

(2)

However, first we described the model variables shown in Table 7. For the portfolios SMB and HML the data confirms the previous results obtained in relation to dimension and book-to-market, in each period considered. As expected, the factor risk of market has a positive premium. Each factor average has a strong statistical significance as measured by one sample t-test. However, there is one worrying result: the correlation between dimension and book-to-market is stronger when the portfolios SMB and HML are used. Across the period the correlation is -0.596 with statistical significance at the 0.01 level. In the second decade, the correlation is -0.750 with the same statistical significance, which is high and unexpected. However, no correlation was detected in the first decade. The correlation between SMB and HML on the one hand and Rm-Rf on the other is not significant in any periods. However, the correlation between SMB and Rm-Rf in the first decade is relevant. This result confirms the findings of Fama and French (1992, 1993). Table 7 Description of Fama-French model variables As in the previous section, here we also analyze the correlation for each of the 25 portfolios. The results are in Table 8 and show that during the entire period the correlation is usually above 0.5, which is an important correlation level. However, analyzing each decade, we found some interesting results. Just as Fama and French (1993) argued, the usage of HML and SMB substantially reduces the correlation between ME and BE/ME in the first decade. In the second decade, HML and SMB are strongly correlated. Once again, we have to ask what might have provoked these unexpected changes between the two decades.

10

Table 8 Correlation between HML and SMB for the 25 portfolios These results promise to raise autocorrelation problems in the OLS regression with HML and SMB as independent variables. This is so, as shown in Table 9. The regressions have statistical significance (test F, not reported here) and each independent variable can explain the returns reported, even if only to a very low degree. Actually the adjusted R-square assumes a considerably low level, failing to confirm previous evidence (Fama and French, 1993, 1996). However, this low adjusted R-square is not totally unexpected if one considers that this model has a very strong sectional component. What is problematic is the positive slope of both factors, which associated with the high correlation level between SMB and HML, suggests notorious problems of autocorrelation. Note that the HML slope should be negative in the second decade and when the entire sample is used. Table 9 - Regression Ri-Rf=ai+hiHML+siSMB+i Associated as independent variables HML and SMB cause technical problems, but what happens when they are considered alone? Table 10 summarizes the results of a regression with HML as a unique independent variable. In each and every sample used, the regressions are statistically strong and the slope is as expected: erratic over the period; positive in the first decade, reflecting the positive relation between returns and BE/ME; and negative in the second decade as the return-BE/ME relation became negative in this last period. The relation with risk is linear and as expected, but may not be enough to explain the earnings difference, although this question is not directly addressed here. Once again, the adjusted R-square is quite low, making the BE/ME almost irrelevant as an explanatory variable of return. Table 11 reports the results of the factor SMB regression. When we considered the SMB alone it presented the expected behaviour (positive slope). Smaller firms had a higher risk premium when compared with bigger firms. This result is statistically significant. The relationship between risk and return tends towards linearity. Again, we cannot say if the return premium is or is not appropriate to compensate for the different risk levels. The adjusted R-square is very low, but higher than the HML adjusted R-square, suggesting that dimension is a more important explanatory factor of stock returns.

11

Table 10 Regression Ri-Rf=ai+hiHML+i Table 11 Regression Ri-Rf=ai+siSMB+i If the HML and SMB are of lesser importance in terms of their capacity for explaining returns, what about the market factor? Will it also be irrelevant, or perhaps it will be relevant but not as important as predicted in the SLB model? To answer this question we ran a regression of excess firm return on the market risk premium, using a model similar to the Fama-MacBeth Model (1983). The results are in Table 12 and are the more significant in statistical terms (highest t and F statistics). The factor has a positive slope and is usually higher in the portfolios with highest returns. This effect is consistent across the entire period of time. Although erratic, the adjusted R-square is considerably higher than in the other factors, thus the market factor has, indeed, more explanatory potential, particularly in the portfolios with bigger and higher book-to-market firms. Nevertheless, it is still insufficient to fully explain the excess return of common stock. At most, it explains 43.2% of the risk premium (portfolio with size Big and BE/ME quintile 4 in decade 1. Another interesting point is that the has stronger explanatory ability in decade 1 than in decade 2. In certain cases the capacity has decreased by 20% (portfolios with big size and BE/ME less than 5). What can explain these changes? These results challenge the Fama-French model, since its explanatory capacity in the UK seems to be quite inferior to that reported by Fama and French (1993, 1996) and SMB and HML seem to be strongly correlated. Nevertheless, we tested the model, and the results are presented in Table 13. As expected, the results provide evidence on the autocorrelation already reported. The HML factor when analyzed together with SMB presents positive slopes, what does not correspond to the relation between HML and excess return. Thus, the Fama-French model seems to be affected by an autocorrelation problem in the UK market. If so, we should rethink the meaning of dimension and BE/ME and we cannot test if they are risk proxies or characteristics that cannot be associated with a specific risk profile. Table 12 Regression Ri-Rf=ai+bi(Rm-Rf)+i Table 13 Regression Ri-Rf=ai+bi(Rm-Rf)+hiHML+siSMB+i This model appears not to work due to the strong correlation between SMB and HML. If this is the case, an alternative model without one of these two factors might be successful. Because SMB has 12

more consistent behaviour through the time period and reveals higher capability to explain excess returns than HML (the adjusted R-square tends to be higher), we chose to keep the SMB factor in the model. Therefore, we ran a regression on:

Ri R f = ai + bi ( Rm R f ) + si SMB + i

(3)

The results of the new regression are summarized in Table 14. Without the perverse effects of autocorrelation, we can conduct a more comprehensive analysis of the tested model. In each studied period we find similar results with an important exception: the model has a powerful explanatory capacity in the first decade, which is not that surprising if we take into account the previous evidence. Actually, this first period shows results close to those expected by the literature. It is consistently in the second decade that the results are unexpected and unforeseen. Table 14 Regression Ri-Rf=ai+bi(Rm-Rf)+siSMB+i In all portfolios and periods of time the market factor presents similar values statistically no different than 0.85, which supports the finding reported by Fama and French (1993). Thus, the market premium risk seems to compensate for a systematic risk factor common to all firms. This interpretation is consistent with the EMH and the multi-factor models of Merton (1973) and Ross (1976). The factor SMB shows very regular behaviour, with higher loadings in smaller firms and lower loadings in bigger firms, in an almost linear relation across different size quintiles for each BE/ME quintile. The adjusted R-square suggests (1) that model 3 is as able to explain the excess returns as model 2 and (2) these two variables are still insufficient to provide a comprehensive and full explanation of excess returns. The unexplained and unconsidered factors should have been captured by the intersections of Fama-French Model and the SMB model tested here. Table 15 shows the intersections and correspondent t statistic for the 25 portfolios in each period of time. As expected, the intersections are significantly different from zero in both portfolios. Also as expected, the intersections are lower and less statistically significant in the first decade than in the second one. Our findings show that the Fama-French model does not apply in UK, meaning that our initial project cannot be pursued. However, interesting evidence has been found that requires some reflection. It is presented in the final section.

13

Table 15 Regression Interception IV Conclusions The results are quite unexpected. We have found that: In the UK, the book-to-market effect is the opposite of the one verified in the USA; In the American sample, there is a negative correlation between return and volatility (higher returns associated to lower volatility); In the English sample, the Fama-French model does not apply. The market factor is unique and is the only one with a good ability to explain excess returns; and Size effect and book-to-market are strongly correlated in the UK sample.

Thus, the results suggest there are important differences between the English and the American markets. One possible cause is a biased sample (Banz and Breen, 1986 and Kothari et al. 1995). In this study, there are some structural differences in the portfolio composition of both sets of portfolios between the UK sample and the USA sample (see Table 16 and Table 17). The market value and book value computation are similar to the Fama-French (1992, 1996, 2000) and Daniel and Titman (1997) one, but not exactly the same. These differences in data and methodology might explain the findings. In the UK, the firm distribution by quintiles is approximately uniform, which does not happen in the USA sample where 60% of the firms are in the first size quintile. Similarly, the first and last BE/ME quintiles are those with higher number of firms (28% in the first quintile and 21% in the fifth quintile). These differences might explain, to a certain extent, the different results. However, our conclusions may not be so surprising considering that Kothari et al. (1995), with a different sample, report there is a weak relationship between BE/ME and excess return, and that the properly measures the relation between return and risk premium, as long as the is calculated on an annual basis. These findings, similar to ours, raise the question of possible data mining. This question is also addressed by Conrad et al. (2003, pp. 1969). These authors suggest that the methodologies developed to explore CAPM weaknesses and our familiarity with the data might explain up to 50% of the in-sample relation between firm characteristics and returns uncovered using single (one-way) sorts. The biases can be much larger if we simultaneously condition returns on two or more characteristics. 14

In this case, we are so extremely familiarized with the USA data that we might have used data in such a way as to help us support certain desired conclusions. In the UK, this does not happen, once the data are less well known. We believe that our work does not suffer from this bias because we have used American methodologies with an English sample. Thus, our results should be more impartial. Table 16 Structure of the 25 portfolios Table 17 Fama-French portfolios Moreover, there is one set of results that cannot be explained this way: the difference between the two decades, which is the most striking finding. In the first decade, the results correspond to what would be expected, in line with other evidence reported throughout the paper. However, in the second decade something has changed in an unanticipated way, allowing an opposite book-to-market effect. However, if BE/ME is a risk proxy, when we study this effect we are, actually, analyzing the relation between risk and return, using book-to-market as an intermediate that represents risk. So, a real change in the BE/ME effect implies a different relation between BE/ME and risk. If volatility is a measure of risk, than what really changed in the second decade was the relation between return and risk/volatility, not the one between BE/ME and volatility. So, the BE/ME effect is the same, but the risk profile has changed. Thus, this seems to confirm BE/ME as a risk proxy. But this is not the complete story. We still have a question to answer: what may have caused the change in the relation between risk and returns? The obvious explanation brings us back to neurobiology, behavioural finance and the main idea that investors behaviour is full of irrationality, emotions, feelings, and memories that are taken into account in the decision process that can lead us to irrational decisions. Our hypothesis is: something has induced a change in the way investors evaluate firms with lower BE/ME, creating higher demand and turnover for this stock. Because investors usually buy and sell intraday or overnight, they force greater price floating and, thus, higher volatility. In this sense, they are riskier and have a higher return level associated. This possibility should be followed up in future studies. One can argue that this explanation is not good enough because it does not consider the size effect. It is true, but we would not expect that. Literature refers to the BE/ME as the main explanatory factor regarding excess returns, with a much stronger explanatory capacity than size effect. In addition, the findings suggest that the size effect appeared in the eighties (Fama and French, 1992). So, the 15

preference for the BE/ME factor in investment decisions is normal under this assumption. A similar phenomenon may have affected . In the early nineties several scholars such as Fama and French (1992) announced the death of . So, if is not a relevant factor, it will be less and less used. The relationship between and excess returns tends to decrease, as our results suggest. These results have obvious implications for the question raised initially: what is the real role of the size and book-to-market effects? Considering the contradictions between the USA and the UK samples and the reduced explanatory power when regressed against excess return, they may not be risk proxies or relevant characteristics for explaining return or excess return. If we considered only the UK sample, then HML and SMB could be understood as proxies of risk. The overall analysis, specifically the UK-USA contradictions, leads us to a conclusion: either we have a data bias or financial markets are inefficient and reflect serious irrationality.

V References Alper, M. and H. Raifa (1982), A Progress Report on the Training of Probability Assessors, in D. Kahneman, P. Slovic and A. Tversky (eds.), Judgment Under Uncertainty: Heuristics and Biases (Cambridge University Press) Banz, R. (1981), The Relationship Between Return and the Market Value of Common Stocks, Journal of financial Economics, Vol. 9, No. 1 (March), pp. 3-18 Banz, R. and W. Breen (1986), Sample-Dependent Results Using Accounting and Market Data: Some Evidence, Journal of Finance, Vol. 41, No. 41 (September), pp. 779-793 Basu, S. (1983), The Relationship Between Earnings Yield, Market Value, and Return for NYSE Common Stocks, Journal of Financial Economics Vol. 12, No. 1 (June), pp. 129-156 Bhandari, Laxmi C. (1988), Debt/equity Ratios and Expected Common Stock Returns: Empirical Evidence, Journal of Finance, Volume 43, No. 2 (June), pp. 507-528 Black, F. (1972), Capital Market Equilibrium with Restricted Borrowing, Journal of Business, Vol. 45, No. 3 (July), pp. 444-455 Buehler R., D. Griffin, and M. Ross (1994), Exploring the Planning Fallacy: Why People Underestimate their Task Completion Times, Journal of Personality and Social Psychology, Vol. 67, No. 3 (March), pp. 366-381 Camerer, C. and R. Hogarth (1999), The Effects of Financial Incentives in Experiments: A Review and Capital-Labor Production Framework, Journal of Risk and Uncertainty, Vol. 19, No. 1-3 (December), pp. 7-42 Conrad J., M. Cooper and G. Kaul (2003), Value versus Glamour, Journal of Finance, Vol. 58, No. 5 (October), pp. 1969-1996 16

Damsio, A (2003), Looking for Spinoza. Joy, Sorrow and the Feeling Brain (Orlando: Harcourt Books) Daniel, K. and S. Titman (1997), Evidence on the Characteristics of Cross-Sectional Variation in Stock Returns, Journal of Finance, Vol. 52, No. 1 (March), pp. 1-33 Davis, J., E. Fama, and K. French (2000), Characteristics, Covariances, and Average Returns: 1929 to 1997, Journal of Finance, 55, No. 1 (February), pp. 389-406 De Bondt, W. and R. Thaler (1985), Does the Stock Market Overreact?, Journal of Finance, Vol. 40, No. 3 (July), 793-808 ------ ------ (1987), Further Evidence on Investor Overreaction and Stock Market Seasonality, Journal of Finance, Vol. 42 No. 3 (July), pp. 557-581 Edwards, W. (1968), Conservatism in Human Information Processing, in B. Kleinmutz (ed.), Formal Representation of Human Judgment (New York: John Wiley and Sons) Fama, E. and J. MacBeth (1973), Risk, Return and Equilibrium: Empirical Tests Journal of Political Economy, Vol. 81, No. 3 (May-June), pp. 607-636 Fama, E. (1970), Efficient Capital Markets: A Review of Theory and Empirical Work, Journal of Finance, Vol. 25, No. 2 (May), pp. 383-417 Fama, E. and K. French (1992), The Cross-Section of Expected Stock Returns, Journal of Finance, Vol. 47, No. 2 (June), pp. 427-465 ------ ------ (1993), Common Risk Factors in Returns on Stocks and Bonds, Journal of Financial Economics Vol. 33, No. 1 (February), pp. 3-56 ------ ------ (1996), Multifactor Explanations of Asset Pricing Anomalies, Journal of Finance, Vol. 51, No. 1 (March), pp. 55-84 Fischhoff, B., Slovic, P. and S. Lichtenstein (1977), Knowing With Certainty: The Appropriateness of Extreme confidence, Journal of Experimental Psychology: Human Perception and Performance, Vol. 3, No. 4 (November), pp. 552-564 Jegadeesh, N. and S. Titman (1993), Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency, Journal of Finance, Vol. 48, No. 1 (March), pp. 65-91 Kahneman, D, and A. Tversky (1982), Intuitive Prediction: Biases and Corrective Procedures in D. Kahneman, P. Slovic and A. Tversky (eds.), Judgment Under Uncertainty: Heuristics and Biases (Cambridge: Cambridge University Press) ------ ------ (1974), Judgement Under Uncertainty: Heuristics and Biases, Science, Vol. 185, No. 4157 (September), pp. 1124-31 Keim, Donald B. (1983), Size Related Anomalies and Stock Return Seasonality: Further Evidence, Journal of Financial Economics Vol. 12, No. 1 (June) pp. 13-32 Kothari, S. P., J. Shanken and R. Sloan (1995), Another Look at the Cross-Section of Expected Returns, Journal of Finance, Vol. 50, No. 1 (March), pp. 185-224 Lakonishok, J., A. Shleifer and R. Visnhy (1994), Contrarian Investment, Extrapolation and Risk, Journal of Finance, Vol. 49, No. 5 (December), pp. 1541-1578 17

Lintner, J. (1965), The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets, Review of Economics and Statistics, Vol. 47, No. 1, (February), pp. 13-37 Lord, C., Ross and M. Lepper (1979), Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence, Journal of Personality and Social Psychology, Vol. 37, No. 11 (November), pp. 2098-2109 Merton, R. (1973), An Intertemporal Capital Asset Pricing Model, Econometrica, Vol. 41, No. 5 (September), pp. 867-887 Rosenberg, B., K. Reid and R. Lanstein (1985), Persuasive Evidence of Market Inefficiency, Journal of Portfolio Management Vol. 11, No. 1 (Spring), pp. 9-17 Ross, S. (1976), The Arbitrage Theory of Capital Asset Pricing, Journal of Economic Theory, Vol. 13, No. 3 (December), pp. 341-360 Sharpe, W., 1964, Capital Asset Prices: a Theory of Market Equilibrium Under Conditions of Risk, Journal of Finance, Vol. 19, No. 3 (September), pp. 425-442

18

Table 1 Descriptive statistics of the 25 portfolios - UK The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were reranked in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. For each portfolio we computed the weighted-average return. To calculate the return rate, we considered the adjusted price of the last day of the month.
Size Panel A - 1983-2002 Book-to-Market 4 High 0.97% 2.02% 0.99% 0.71% 0.52% 0.59% 1.34% 2.58% 1.45% 1.22% 0.94% 0.52%

Low 1.82% 3.58% 2.26% 1.30% 1.51% 0.43%

2 1.38% 2.73% 1.86% 1.08% 0.77% 0.47%

Small 2 3 4 Big

2.80% 1.48% 1.13% 0.92% 0.50%

3 Return 1.33% 3.11% 0.85% 1.31% 0.85% 0.51%

Low

9.69% 10.43% 12.01% 7.53% 12.37% 5.74% 6.98% 6.29% 9.96% 5.06% 4.73%

2 3 4 Standard Deviation 6.80% 8.31% 5.25% 9.55% 15.35% 6.70% 7.69% 5.12% 4.56% 5.68% 5.68% 4.85% 4.85% 5.04% 4.82% 4.88% 4.95% 5.03%

High 5.43% 5.58% 5.07% 5.42% 5.27% 5.68%

Small 2 3 4 Big

9.17 6.72 6.69 4.98 3.40

6.40 4.56 2.80 2.85 2.31 1.39

t-statistics (return) 6.93 5.44 4.37 3.10 3.70 2.54 2.92 3.53 2.42 2.59 1.47 1.57

6.28 4.62 3.33 2.25 1.66 1.78

8.43 7.06 4.38 3.45 2.73 1.40

57 57 58 60 59

Average number of firms 56 59 59 60 35 39 53 62 38 50 58 67 56 58 58 63 71 69 62 59 80 78 63 47 Average Book-to-Market 0.29 0.52 0.77 1.14 0.40 0.63 0.89 1.27 0.32 0.56 0.79 1.13 0.28 0.50 0.74 1.14 0.25 0.46 0.73 1.08 0.22 0.44 0.69 1.07

59 96 73 58 40 30

Small 2 3 4 Big

6,025 17,759 44,848 135,012 1,840,312

Average size 531,854 564,691 441,181 248,161 258,070 6,576 6,284 5,864 5,934 5,466 18,180 17,800 17,145 17,777 17,894 46,280 45,030 45,233 44,472 43,225 138,868 139,448 130,154 131,944 134,643 2,449,363 2,614,892 2,007,506 1,040,678 1,089,121

1.05 0.89 0.85 0.79 0.79

1.65 2.04 1.68 1.59 1.44 1.52

19

Size

Low 1.13% 2.46% 1.20% 0.92% 0.62% 0.43%

2 1.29% 3.02% 1.03% 0.89% 0.86% 0.65%

Small 2 3 4 Big

2.61% 1.28% 1.04% 0.86% 0.64%

3 Return 1.22% 2.29% 1.10% 1.25% 0.91% 0.57%

Panel B - 1983-1993 Book-to-Market 4 High Low 1.07% 2.09% 1.29% 0.79% 0.66% 0.52% 1.71% 3.17% 1.79% 1.34% 1.21% 1.03% 5.90% 6.83% 6.09% 5.62% 5.46% 5.50%

6.96% 5.79% 5.61% 5.59% 5.61%

3 4 Standard Deviation 6.25% 5.90% 5.69% 8.45% 6.50% 6.54% 5.79% 5.96% 5.43% 5.51% 5.86% 5.43% 5.31% 5.91% 5.38% 5.60% 5.42% 5.74% Average number of firms 48 49 48 35 42 53 44 46 52 50 48 49 53 53 49 58 53 37 Average Book-to-Market 0.56 0.84 1.16 0.62 0.93 1.26 0.63 0.88 1.17 0.55 0.82 1.15 0.51 0.80 1.12 0.50 0.74 1.09

High 6.11% 6.64% 6.00% 5.92% 6.08% 5.93%

Small 2 3 4 Big

9.09 5.37 4.50 3.73 2.79

4.64 3.95 2.15 1.80 1.25 0.85

t-statistics (return) 5.04 5.04 3.91 3.87 1.95 2.02 1.78 2.34 1.78 1.68 1.27 1.15

4.58 3.50 2.61 1.60 1.35 1.00

6.81 5.22 3.27 2.48 2.19 1.91

47 46 48 49 49

47 30 35 47 55 65

48 74 53 43 37 33

Average size 313,437 285,628 329,961 236,491 Small 4,854 5,551 4,794 4,554 4,747 2 13,674 13,440 13,682 13,777 13,593 3 35,586 35,655 36,072 35,994 36,016 4 108,370 110,438 110,314 108,642 105,135 Big 1,179,865 1,402,101 1,263,278 1,486,838 1,022,966

176,831 4,625 13,878 34,193 107,318 724,141

1.07 0.94 0.90 0.84 0.83

0.33 0.42 0.35 0.31 0.29 0.26

1.70 2.11 1.67 1.66 1.50 1.55

Size

Low 2.54% 4.76% 3.39% 1.70% 2.43% 0.43%

Small 2 3 4 Big

3.01% 1.70% 1.22% 0.98% 0.36%

3 Return 1.48% 1.44% 2.43% 3.96% 2.73% 0.59% 1.28% 1.38% 0.66% 0.80% 0.28% 0.45% t-statistics (return) 4.80 3.34 2.44 2.02 3.16 1.54 2.34 2.67 1.64 2.16 0.73 1.08

Panel C - 1993-2002 Book-to-Market 4 High Low 0.86% 1.95% 0.68% 0.63% 0.37% 0.65% 0.96% 1.95% 1.10% 1.10% 0.65% -0.02% 12.47% 13.14% 15.67% 9.01% 16.55% 5.87% 8.18% 6.96% 13.09% 4.39% 3.79%

3 4 Standard Deviation 7.35% 10.26% 4.74% 10.62% 20.97% 6.90% 9.23% 4.05% 3.42% 5.87% 5.52% 4.18% 4.32% 3.95% 4.17% 4.01% 4.42% 4.17% Average number of firms 70 69 71 43 64 71 56 70 81 66 67 77 85 72 69 98 73 57

High 4.55% 4.12% 3.86% 4.85% 4.27% 5.37%

Small 2 3 4 Big

5.47 4.49 4.96 3.37 1.95

4.87 3.25 2.18 2.22 1.98 1.21

4.32 3.02 2.12 1.61 0.96 1.67

5.01 5.07 3.04 2.42 1.63 -0.34

67 68 69 71 70

65 40 41 64 86 95

70 118 93 73 43 26

Average size 761,766 858,441 558,254 260,444 343,585 Small 7,257 7,655 7,851 7,244 7,183 6,355 2 22,059 23,170 22,134 20,690 22,180 22,122 3 54,598 57,464 54,460 54,959 53,373 52,732 4 163,056 168,795 170,116 152,799 160,163 163,406 Big 2,535,520 3,551,745 4,037,643 2,555,578 1,059,324 1,473,312

1.12 0.93 0.84 0.75 0.73

0.28 0.43 0.35 0.24 0.20 0.16

Average Book-to-Market 0.46 0.76 1.12 0.58 0.99 1.26 0.49 0.79 1.10 0.46 0.73 1.13 0.40 0.66 1.09 0.37 0.63 1.03

1.75 2.32 1.94 1.62 1.39 1.48

20

Table 2 Descriptive statistics of the Fama-French portfolios - UK The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. For each portfolio we computed the weighted-average return. To calculate the return rate, we considered the adjusted price of the last day of the month.
Size Panel A - 1983 - 2002 Book-to-Market Low Medium High Low Medium High Return Standard Deviation 1.23% 1.04% 0.98% 6.40% 5.16% 4.99% 2.03% 1.47% 1.44% 6.16% 7.73% 5.59% 4.80% 0.43% 0.61% 0.52% 4.80% 4.59% 4.66% 5.15% t-statistics (return) 4.15 4.37 4.01 4.02 1.43 2.01 Average number of firms 213 294 223 239 155 281 281 248 272 308 166 Average Book-to-Market 0.35 0.79 1.52 0.98 0.40 0.86 1.67 0.80 0.30 0.72 1.37

Small Big

1.65% 0.52%

Small Big

7.07 2.87

4.23 4.58 1.54

Small Big

16,726 820,083

Average size 628,725 426,804 199,685 18,628 16,436 15,115 1,238,823 837,171 384,256

21

Size

Small Big

1.50% 0.66%

Panel B - 1983 - 1993 Book-to-Market Low Medium High Return 0.88% 1.05% 1.33% 1.29% 1.46% 1.76% 5.63% 0.46% 0.63% 0.90% 5.40% t-statistics (return) 2.48 2.90 2.52 2.77 0.94 1.28

Low Medium High Standard Deviation 5.47% 5.57% 5.56% 5.63% 5.77% 5.51% 5.30% 5.34% 5.59%

Small Big

5.07 2.33

3.71 3.50 1.77

Average number of firms 176 241 180 194 140 227 217 204 213 256 144 Average Book-to-Market 0.40 0.85 1.51 0.99 0.45 0.90 1.61 0.85 0.34 0.79 1.40

Small Big

13,018 528,756

Average size 333,816 293,132 185,713 14,379 12,979 11,695 653,252 573,286 359,730

Size

Small Big

1.79% 0.37%

Panel C - 1993 - 2002 Book-to-Market Low Medium High Return 1.60% 1.04% 0.61% 2.79% 1.48% 1.10% 6.69% 0.40% 0.60% 0.11% 4.08% t-statistics (return) 3.33 3.35 3.17 2.93 1.14 1.67

Low Medium High Standard Deviation 7.25% 4.70% 4.30% 9.42% 5.42% 3.91% 3.73% 3.84% 4.63%

Small Big

4.96 1.68

2.12 3.00 0.26

Average number of firms 251 347 266 283 171 334 345 293 330 360 188 Average Book-to-Market 0.29 0.73 1.53 0.96 0.34 0.81 1.74 0.74 0.24 0.66 1.33

Small Big

Average size 939,156 567,510 214,393 20,630 23,100 20,075 18,714 1,126,743 1,855,213 1,114,945 410,072

22

Table 3 Descriptive statistics of the 25 portfolios The NYSE, AMEX and Nasdaq sub-sample data used covers the period from December 1982 to June 2002 and were arranged by Fama and French from Compustat and CRSP data. BE is the COMPUSTAT book value of stockholders equity, plus balance sheet deferred taxes and investment tax credit (if available), minus the book value of preferred stock. The book-to-market ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. Within each year, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were reranked in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. For each portfolio the weighted-average returns were computed. To calculate the return rate, we considered the adjusted price of the .
Size Panel A - 1983-2002 Book-to-Market 2 3 4 High Return 1.00 1.09 1.22 1.24 0.93 1.11 1.33 1.27 6.34 0.86 1.20 1.26 1.11 5.77 1.02 1.04 1.17 1.41 5.38 1.05 1.04 1.20 1.26 5.16 1.13 1.07 1.11 1.15 4.73 t-statistics (return) 6.01 7.65 9.10 2.02 3.06 4.08 2.24 3.81 4.12 2.89 3.46 4.02 3.19 3.24 4.21 3.59 3.64 3.91

Low 0.64 0.00 0.44 0.67 1.00 1.07

Low

Small 2 3 4 Big

0.93 0.98 1.06 1.11 1.11

2 3 4 High Standard Deviation 7.03 5.67 4.89 4.57 4.98 8.36 7.02 5.53 4.99 5.11 7.69 5.87 4.83 4.69 5.26 7.12 5.41 4.63 4.46 4.85 6.45 5.04 4.93 4.36 4.81 5.10 4.80 4.51 4.35 4.89 Average number of firms 283 172 164 166 210 786 451 449 518 829 231 153 151 125 94 162 105 93 76 54 119 82 71 61 43 117 71 57 49 32

Small 2 3 4 Big

5.01 5.79 6.77 7.37 8.00

3.10 0.00 0.89 1.45 2.38 3.20

8.52 3.81 3.23 4.44 4.00 3.60

607 151 98 75 65

Small 2 3 4 Big

47 261 648 1,703 10,433

3,648 49 256 569 1,605 15,764

Average size 2,796 2,560 2,131 1,956 53 52 47 34 262 267 266 256 655 661 667 685 1,710 1,718 1,755 1,726 11,300 10,101 7,918 7,080

0.84 0.80 0.78 0.78 0.75

Average Book-to-Market 0.23 0.48 0.70 0.95 1.59 0.21 0.48 0.70 0.96 1.85 0.22 0.48 0.70 0.95 1.66 0.23 0.48 0.70 0.95 1.54 0.24 0.48 0.70 0.96 1.51 0.23 0.48 0.69 0.95 1.40

23

Size

Low 0.69 -0.16 0.44 0.85 1.13 1.17

2 1.04 0.73 0.91 1.23 1.06 1.28

Small 2 3 4 Big

0.71 1.07 1.24 1.25 1.28

Panel B - 1983-1993 Book-to-Market 3 4 High Low 2 3 4 Return Standard Deviation 1.11 1.27 1.44 6.06 5.47 4.81 4.42 0.82 1.05 1.13 5.47 6.43 5.77 5.05 4.83 1.36 1.33 1.29 5.49 6.71 5.81 4.88 4.54 1.12 1.43 1.59 5.19 6.29 5.53 4.64 4.39 1.04 1.38 1.61 5.01 5.63 5.25 5.11 4.29 1.19 1.18 1.58 4.67 5.07 5.00 4.41 4.10

High 4.96 5.14 5.27 4.92 4.72 4.77

Small 2 3 4 Big

3.19 4.76 5.88 6.09 6.72

2.78 -0.28 0.72 1.48 2.21 2.53

t-statistics (return) 4.67 5.64 1.39 1.77 1.72 3.06 2.44 2.65 2.20 2.23 2.81 2.96 Average size 1,621 1,559 32 31 173 180 453 459 1,193 1,198 6,253 5,927

7.06 2.38 3.22 3.57 3.53 3.16

7.12 2.40 2.67 3.54 3.75 3.64

540 131 86 64 55

Average number of firms 269 157 141 130 179 837 435 384 370 676 202 136 128 100 88 136 86 82 72 54 93 66 61 58 42 79 60 52 50 34 Average Book-to-Market 0.28 0.57 0.81 1.08 1.71 0.24 0.57 0.81 1.08 2.00 0.27 0.57 0.82 1.07 1.79 0.28 0.57 0.81 1.07 1.64 0.29 0.57 0.81 1.08 1.61 0.30 0.57 0.81 1.08 1.52

Small 2 3 4 Big

29 175 455 1,193 6,195

1,844 29 170 394 1,093 7,533

1,529 30 179 483 1,258 5,696

1,493 22 171 483 1,221 5,566

0.94 0.90 0.88 0.87 0.85

Size

Low 0.59 0.17 0.45 0.49 0.86 0.96

2 0.95 1.13 0.81 0.80 1.05 0.96

Small 2 3 4 Big

1.16 0.88 0.87 0.97 0.92

Panel C - 1993-2002 Book-to-Market 3 4 High Low 2 3 4 Return Standard Deviation 1.08 1.16 1.03 7.93 5.89 4.97 4.73 1.41 1.63 1.43 7.14 10.02 8.15 6.01 5.17 1.03 1.19 0.93 6.06 8.63 5.96 4.79 4.86 0.96 0.90 1.21 5.57 7.92 5.29 4.63 4.54 1.05 1.01 0.89 5.31 7.23 4.84 4.74 4.45 0.95 1.04 0.69 4.79 5.14 4.61 4.63 4.61

High 5.00 5.10 5.27 4.78 4.90 5.00

Small 2 3 4 Big

3.86 3.47 3.75 4.36 4.59

1.76 0.18 0.56 0.66 1.27 1.99

t-statistics (return) 3.86 5.18 1.49 2.51 1.45 2.30 1.62 2.22 2.31 2.36 2.23 2.18

5.84 3.37 2.61 2.13 2.43 2.41

4.91 3.00 1.88 2.70 1.93 1.48

224 676 172 111 86

Average number of firms 297 189 188 203 243 733 468 517 673 987 262 171 176 151 99 188 125 105 81 55 146 98 81 64 44 155 82 61 48 29 Average Book-to-Market 0.18 0.40 0.58 0.83 1.47 0.18 0.40 0.59 0.84 1.69 0.18 0.40 0.59 0.82 1.54 0.18 0.40 0.58 0.82 1.43 0.18 0.40 0.58 0.83 1.42 0.17 0.39 0.58 0.82 1.28

Small 2 3 4 Big

66 353 851 2,239 14,893

5,548 71 346 752 2,143 24,427

Average size 4,033 3,613 2,764 75 73 64 356 359 358 868 874 862 2,254 2,265 2,278 16,611 14,496 10,257

2,444 46 345 897 2,256 8,674

0.74 0.71 0.68 0.68 0.65

24

Table 4 Descriptive statistics of the Fama-French portfolios The NYSE, AMEX and Nasdaq sub-sample data used covers the period from December 1982 to June 2002 and were arranged by Fama and French from Compustat and CRSP data. BE is the COMPUSTAT book value of stockholders equity, plus balance sheet deferred taxes and investment tax credit (if available), minus the book value of preferred stock. The book-to-market ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. Within each year, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were reranked in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. For each portfolio the weighted-average return were computed. To calculate the return rate, we considered the adjusted price of the last day of the month.
Size Panel A - 1983 - 2002 Book-to-Market Low Medium High Low Medium High Return Standard Deviation 0.76 1.11 1.20 6.17 4.65 4.51 0.98 0.49 1.15 1.28 5.70 7.12 4.92 4.74 1.08 1.03 1.08 1.13 4.57 5.05 4.37 4.28 t-statistics (return) 19.09 19.81 20.53 51.76 1.06 3.59 4.13 60.94 3.13 3.77 4.03 Average size 2,131 2,174 1,642 119 139 136 81 3,846 4,123 4,211 3,203 Average number of firms 933 821 737 1352 1451 1301 1304 308 415 340 169 Average Book-to-Market 0.29 0.71 1.39 0.83 0.28 0.72 1.49 0.76 0.29 0.71 1.28

Small Big

Small Big

Small Big

Size

Small Big

Panel B - 1983 - 1993 Book-to-Market Low Medium High Low Medium High Return Standard Deviation 0.84 1.16 1.36 5.69 4.61 4.52 0.99 0.51 1.16 1.29 4.93 6.22 4.90 4.82 1.25 1.17 1.17 1.42 4.26 5.11 4.32 4.22 t-statistics (return) 14.71 15.16 15.86 70.06 0.90 2.59 2.94 63.41 2.51 2.96 3.70 Average size 1,608 1,429 1,259 77 81 91 59 2,787 3,135 2,767 2,459 Average number of firms 886 695 612 1199 1454 1090 1053 263 318 300 172 Average Book-to-Market 0.35 0.84 1.53 0.94 0.34 0.84 1.64 0.87 0.36 0.84 1.41

Small Big

Small Big

25

Size

Low 0.68 0.48 0.89

Small Big

0.96 0.89

Panel C - 1993 - 2002 Book-to-Market Medium High Low Medium High Return Standard Deviation 1.06 1.04 6.66 4.69 4.51 1.14 1.26 4.70 8.00 4.96 4.68 0.98 0.81 4.34 5.00 4.44 4.34

Small Big

t-statistics (return) Average number of firms 13.98 15.62 14.98 983 953 868 51.45 0.64 2.47 2.88 1514 1449 1524 1568 56.87 1.89 2.36 2.00 355 516 382 167 Average size 2,682 2,964 2,045 162 200 183 104 4,965 5,164 5,745 3,985 Average Book-to-Market 0.22 0.58 1.23 0.71 0.23 0.59 1.33 0.64 0.21 0.56 1.14

Small Big

26

Table 5 Annual correlations between size and BE/ME The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. To compute correlation is used the natural logarithm of ME and BE/ME is used to better capture the underlying effects.
Variables 1983-2002 ln(BE/ME) ln(MV) 1.000 -0.299 Time period 1983-1993 ln(BE/ME) ln(MV) 1.000 -0.275 1993-2002 ln(BE/ME) ln(MV) 1.000 -0.304

ln(mv) ln(beme)

1.000

1.000

1.000

27

Table 6 Annual correlations between size and BE/ME for the 25 portfolios The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year.
Size Low Book-to-Market 2 3 4 High

Small 2 3 4 Big

Panel A - 1983-2002 -0.247 -0.330 -0.379 -0.319 -0.505 -0.413 -0.360 -0.522 -0.492 -0.326 -0.431 -0.380 -0.151 -0.177 -0.270 Panel B - 1983-1993 -0.174 -0.334 -0.378 -0.247 -0.463 -0.512 -0.332 -0.513 -0.511 -0.228 -0.464 -0.503 -0.128 -0.158 -0.234 Panel C - 1993-2002 -0.223 -0.286 -0.340 -0.261 -0.440 -0.305 -0.266 -0.452 -0.431 -0.266 -0.325 -0.198 -0.091 -0.086 -0.231

-0.312 -0.406 -0.375 -0.216 -0.285

-0.310 -0.273 -0.353 -0.344 -0.153

Small 2 3 4 Big

-0.373 -0.521 -0.444 -0.293 -0.356

-0.423 -0.363 -0.467 -0.367 -0.146

Small 2 3 4 Big

-0.243 -0.337 -0.374 -0.129 -0.208

-0.289 -0.243 -0.348 -0.311 -0.153

28

Table 7 Description of Fama-French model variables The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we compute the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. With these we form two new portfolios to capture the book-to-market and size effect. The first portfolio is the SMB (small minus big) that captures the size effect and is the difference between the return of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. The second portfolio is the HML (high minus low) and captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios.
Panel A - 1983 - 2002 Independent variables Correlations Average Std Dev T-statistics Rm-Rf SMB HML 1.22% 4.22% 165.36 Rm-Rf 1 0.57% 4.22% 77.14 SMB -0.142* 1 1.16% 4.55% 146.73 HML -0.048* -0.596* 1 -0.46% 3.39% -78.73 *Correlation is significant at the 0.01 level (2-tailed)

Name Rm Rm-Rf SMB HML

Name Rm Rm-Rf SMB HML

Panel B - 1983 - 1993 Independent variables Correlations Average Std Dev T-statistics Rm-Rf SMB HML 1.40% 5.09% 104.10 Rm-Rf 1 0.53% 5.11% 38.95 SMB -0.244* 1 0.74% 3.61% 77.23 HML 0.095* -0.002 1 0.38% 1.84% 77.93 *Correlation is significant at the 0.01 level (2-tailed)

Name Rm Rm-Rf SMB HML

Panel C - 1993 - 2002 Independent variables Correlations Average Std Dev T-statistics Rm-Rf SMB HML 1.07% 3.39% 136.58 Rm-Rf 1 0.60% 3.38% 76.99 SMB -0.080* 1 1.49% 5.13% 125.35 HML -0.134* -0.750* 1 -1.11% 4.91% -116.95 *Correlation is significant at the 0.01 level (2-tailed)

29

Table 8 Correlation between HML and SMB for the 25 portfolios The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we computed the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. With these we formed two new portfolios to capture the book-to-market and size effect. The first portfolio is the SMB (small minus big) that captures the size effect and is the difference between the return of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. The second portfolio is the HML (high minus low) and captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The correlation between HML and SMB was computed for each portfolio.
Size Low Book-to-Market 2 3 4 High

Small 2 3 4 Big

Panel A - 1983-2002 -0.604 -0.588 -0.546 -0.570 -0.612 -0.601 -0.562 -0.604 -0.607 -0.581 -0.590 -0.602 -0.623 -0.594 -0.595 -0.583 -0.570 -0.620 -0.663 -0.618

-0.608 -0.609 -0.565 -0.621 -0.471

Panel B - 1983-1993 Small -0.039 -0.015 -0.024 -0.014 0.036 2 -0.026 -0.011 0.015 -0.018 0.014 3 0.005 -0.021 0.002 0.018 -0.011 4 0.016 0.021 -0.013 0.006 -0.064 Big 0.005 0.008 0.008 0.007 -0.754 Panel C - 1993-2002 -0.754 -0.779 -0.751 -0.776 -0.756 -0.760 -0.740 -0.738 -0.724 -0.724 -0.749 -0.747 -0.729 -0.753 -0.760 -0.745 -0.761 -0.750 -0.730 -0.766

Small 2 3 4 Big

-0.730 -0.758 -0.785 -0.767 -0.694

30

Table 9 - Regression Ri-Rf=ai+hiHML+siSMB+i

Ri R f = ai + hi HML + si SMB + i
The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. With these we form two new portfolios to capture the book-to-market and size effect. The first portfolio is the SMB (small minus big) that captures the size effect and is the difference between the return of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. The second portfolio is the HML (high minus low) and captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Size Small 2 3 4 Big Panel A - 1983-2002 Book-to-Market 4 High Low 2 0.369 0.388 0.324 0.159 0.013 0.519 0.462 0.346 0.198 0.102 0.980 -4.731 -3.372 -2.467 -9.898

Low

0.104 0.312 -0.380 0.109 -0.159 0.179 -0.101 0.176 -0.299 -0.046

3 h 0.278 0.414 0.349 0.316 0.051

3 4 High t-statistics (h) 4.029 4.149 7.355 11.478 2.110 10.309 12.384 13.817 4.627 9.724 10.764 10.531 5.629 9.578 5.854 5.461 -1.840 1.785 0.443 2.453

Small 2 3 4 Big

s 1.058 0.842 0.842 0.736 0.656 0.718 0.598 0.535 0.473 0.504 0.609 0.568 0.529 0.406 0.413 0.323 0.314 0.227 0.148 0.200 -0.005 -0.159 -0.138 -0.073 -0.179 Adjusted R-Square 0.033 0.027 0.032 0.028 0.023 0.026 0.034 0.029 0.022 0.012 0.008 0.004 0.005 0.006 0.002

13.500 12.366 17.174 10.365 -0.225

t-statistics (s) 15.877 16.778 15.569 1.783.000 19.797 19.632 13.555 9.448 -8.567 -6.463 s(e) 0.200 0.126 0.113 0.103 0.093

20.389 19.073 20.241 19.979 17.941 16.960 7.099 7.663 -3.451 -5.959

Small 2 3 4 Big

0.033 0.042 0.040 0.013 0.008

0.016 0.024 0.021 0.006 0.010

0.259 0.199 0.141 0.141 0.111

0.200 0.149 0.012 0.107 0.090

0.159 0.105 0.099 0.087 0.080

0.184 0.117 0.103 0.096 0.088

31

Size Small 2 3 4 Big

Low 0.069 -0.062 -0.142 0.240 0.092

2 0.171 0.035 0.243 0.407 0.305

3 h 0.512 0.456 0.463 0.538 0.544

Panel B - 1983-1993 Book-to-Market 4 High Low 0.325 0.540 0.642 0.563 0.634 0.807 0.814 0.887 0.899 0.689 -0.789 -4.644 -3.146 -3.186 -0.003

3 4 t-statistics (h) 1.063 0.424 0.229 -4.215 -5.169 -3.725 -5.233 -2.284 -3.844 -1.179 -2.145 0.630 3.256 2.154 1.291 t-statistics (s) 1.073 3.930 0.370 5.004 3.286 6.136 6.209 7.551 5.180 8.623 s(e) 0.170 0.125 0.108 0.104 0.093

High 3.365 -0.028 -1.864 -0.984 1.540

Small 2 3 4 Big

s 0.747 0.905 0.806 0.825 0.788 0.685 0.580 0.665 0.556 0.504 0.552 0.565 0.583 0.416 0.415 0.177 0.201 0.181 0.028 0.069 -0.228 -0.328 -0.284 -0.242 -0.325 Adjusted R-Square 0.029 0.031 0.048 0.025 0.041 0.041 0.038 0.042 0.035 0.012 0.012 0.014 0.020 0.023 0.029

0.435 -0.542 -1.663 3.374 1.663

3.369 7.282 8.934 9.139 9.351

5.977 10.170 10.550 10.536 7.760

Small 2 3 4 Big

0.024 0.033 0.028 0.005 0.008

0.017 0.038 0.038 0.024 0.030

0.173 0.134 0.118 0.109 0.091

0.188 0.127 0.104 0.095 0.091

0.138 0.108 0.101 0.089 0.084

0.244 0.119 0.111 0.099 0.097

Size Small 2 3 4 Big

Low

0.363 0.265 -0.524 0.154 -0.159 0.139 -0.144 0.168 -0.302 -0.051

3 h 0.203 0.290 0.287 0.287 0.250 0.209 0.278 0.086 -0.012 -0.066

Panel C - 1993-2002 Book-to-Market 4 High Low 0.256 0.359 0.189 0.092 0.015 2.095 -3.811 -2.229 -2.342 -6.681

3 4 t-statistics (h) 2.375 2.090 3.852 1.966 5.354 7.190 2.466 4.987 5.352 3.807 6.221 2.387 -1.617 -0.320 -1.851 t-statistics (s) 10.165 10.503 11.324 10.176 9.519 11.980 12.137 11.203 10.552 10.073 6.345 5.183 -4.180 -3.690 -2.274 s(e) 0.219 0.126 0.117 0.103 0.093

High 6.249 8.313 4.660 2.025 0.310

Small 2 3 4 Big

s 1.345 0.901 0.817 0.658 0.460 0.662 0.633 0.410 0.382 0.439 0.643 0.552 0.451 0.332 0.316 0.358 0.353 0.221 0.153 0.198 0.082 -0.106 -0.114 -0.065 -0.093 Adjusted R-Square 0.038 0.026 0.024 0.029 0.012 0.016 0.032 0.021 0.015 0.014 0.006 0.004 0.002 0.003 0.001

9.728 6.121 10.981 7.147 2.316

13.959 12.745 10.038 5.557 -2.266

Small 2 3 4 Big

0.037 0.045 0.046 0.018 0.017

0.017 0.016 0.016 0.009 0.004

0.315 0.248 0.158 0.160 0.123

0.201 0.166 0.130 0.114 0.088

0.175 0.102 0.097 0.085 0.076

0.125 0.116 0.097 0.086 0.071

32

Table 10 Regression Ri-Rf=ai+hiHML+i

Ri R f = ai + hi HML + i
The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. From this was calculated the HML (high minus low) that captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Size Small 2 3 4 Big Panel A - 1983-2002 Book-to-Market 4 High Low -8.908 -14.700 -15.031 -10.162 -12.299

Low -0.761 -0.964 -0.599 -0.343 -0.295

2 -0.440 -0.375 -0.252 -0.080 -0.083

3 h -0.404 -0.269 0.027 -0.002 0.011 0.048 -0.067 0.047 0.023 0.128 0.073 0.026 0.155 0.027 0.220

3 4 High t-statistics (h) -7.092 -7.949 -6.753 0.733 -8.895 -0.054 0.443 1.806 -7.747 -2.277 2.107 -0.939 -3.183 4.858 3.294 0.921 4.233 6.579 0.733 5.948 s(e) 0.202 0.127 0.115 0.104 0.093

Small 2 3 4 Big

0.010 0.025 0.017 0.006 0.008

Adjusted R-Square 0.006 0.005 0.003 0.007 0.000 0.000 0.005 0.000 0.000 0.001 0.002 0.001 0.001 0.003 0.000

0.000 0.000 0.000 0.000 0.005

0.262 0.201 0.143 0.142 0.111

0.198 0.151 0.121 0.108 0.090

0.162 0.106 0.100 0.087 0.080

0.185 0.119 0.104 0.093 0.088

33

Size Small 2 3 4 Big

Low 0.010 -0.099 -0.136 0.246 0.090

2 h 0.143 0.023 0.220 0.416 0.300

3 0.475 0.476 0.465 0.534 0.540

Panel B - 1983-1993 Book-to-Market 4 High Low 0.301 0.521 0.657 0.564 0.630 0.858 0.828 0.878 0.890 0.727 0.062 -0.845 -1.574 3.446 1.617

3 4 High t-statistics (h) 0.888 3.590 3.046 6.324 0.244 5.120 6.905 10.224 2.923 6.051 9.045 10.350 6.326 7.476 9.145 10.445 5.052 8.497 9.253 8.139 s(e) 0.172 0.127 0.110 0.104 0.093

Small 2 3 4 Big

0.000 0.000 0.000 0.002 0.000

Adjusted R-Square 0.000 0.002 0.001 0.000 0.005 0.007 0.001 0.006 0.014 0.006 0.009 0.014 0.004 0.011 0.019

0.004 0.016 0.020 0.024 0.016

0.175 0.136 0.119 0.109 0.092

0.191 0.129 0.106 0.095 0.092

0.141 0.110 0.102 0.088 0.084

0.246 0.120 0.112 0.099 0.098

Size Small 2 3 4 Big

Low -0.908 -1.160 -0.726 -0.464 -0.381

2 -0.619 -0.453 -0.357 -0.167 0.048

3 h -0.562 -0.090 -0.171 0.062 0.091

Panel C - 1993-2002 Book-to-Market 4 High Low -0.371 -0.066 -0.099 -0.053 -0.004 -0.162 -0.058 -0.131 -0.101 0.091 -7.890 -12.849 -14.641 -10.996 -13.016

3 4 t-statistics (h) -8.753 -8.701 -7.760 -8.816 -2.492 -2.422 -9.079 -5.091 -3.777 -5.709 2.145 -2.204 2.290 3.475 -0.171 s(e) 0.221 0.127 0.118 0.103 0.093

High -5.733 -2.049 -5.172 -3.475 2.616

Small 2 3 4 Big

0.015 0.037 0.029 0.013 0.016

Adjusted R-Square 0.016 0.011 0.008 0.012 0.001 0.001 0.011 0.003 0.002 0.003 0.000 0.001 0.000 0.001 0.000

0.002 0.000 0.003 0.002 0.002

0.319 0.249 0.159 0.161 0.123

0.203 0.167 0.131 0.115 0.088

0.176 0.103 0.097 0.085 0.076

0.126 0.117 0.098 0.086 0.071

34

Table 11 Regression Ri-Rf=ai+siSMB+i

Ri R f = ai + si SMB + i
The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. From this was computed the SMB (small minus big) that captures the size effect and is the difference between the return of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Size Small 2 3 4 Big Panel A - 1983-2002 Book-to-Market 4 High Low 16.202 18.697 22.678 14.326 7.283

Low 1.011 0.879 0.675 0.366 0.130

3 s 0.765 0.716 0.571 0.431 0.549 0.355 0.301 0.288 0.494 0.375 0.261 0.243 0.236 0.089 0.077 0.112 -0.138 -0.160 -0.079 -0.214

3 4 High t-statistics (h) 16.944 17.926 20.175 15.205 17.897 14.489 15.936 14.462 20.776 17.148 14.296 13.272 12.728 4.599 4.531 5.447 -9.379 -9.054 -4.751 -8.063 s(e) 0.200 0.126 0.114 0.104 0.093

Small 2 3 4 Big

Adjusted R-Square 0.033 0.032 0.026 0.028 0.039 0.027 0.016 0.017 0.039 0.032 0.022 0.014 0.013 0.010 0.001 0.001 0.003 0.005 0.006 0.002

0.010 0.013 0.013 0.003 0.009

0.259 0.200 0.141 0.141 0.111

0.195 0.149 0.119 0.107 0.090

0.159 0.105 0.099 0.087 0.080

0.184 0.118 0.104 0.093 0.088

35

3 s Small 0.746 0.903 0.800 0.823 0.804 2 0.686 0.580 0.668 0.552 0.510 3 0.552 0.563 0.584 0.422 0.410 4 0.180 0.206 0.178 0.030 0.042 Big -0.228 -0.326 -0.282 -0.240 -0.344 Adjusted R-Square 0.029 0.028 0.046 0.025 0.036 0.033 0.036 0.036 0.022 0.006 0.004 0.000 0.016 0.012 0.010

Size

Low

Panel B - 1983-1993 Book-to-Market 4 High Low

3 4 High t-statistics (h) 9.539 11.218 12.148 17.570 10.943 12.135 11.655 14.519 14.586 12.282 12.614 14.937 14.813 11.660 9.543 4.804 6.356 4.848 0.892 1.034 -7.960 -10.760 -8.685 -6.835 -7.957 s(e) 0.170 0.125 0.108 0.104 0.093

Small 2 3 4 Big

0.024 0.033 0.027 0.003 0.008

0.013 0.023 0.017 0.000 0.015

0.173 0.134 0.118 0.109 0.091

0.188 0.127 0.105 0.095 0.091

0.138 0.109 0.102 0.089 0.085

0.245 0.120 0.112 0.100 0.098

Size Small 2 3 4 Big

Low 1.127 0.974 0.738 0.443 0.264

3 s 0.737 0.695 0.484 0.309 0.540 0.239 0.213 0.222 0.471 0.301 0.206 0.201 0.252 0.056 0.100 0.143 -0.075 -0.107 -0.024 -0.102 Adjusted R-Square 0.037 0.026 0.022 0.028 0.009 0.011 0.031 0.017 0.012 0.013 0.001 0.003 0.002 0.003 0.000

Panel C - 1993-2002 Book-to-Market 4 High Low 12.401 13.742 18.260 12.926 11.388

3 4 High t-statistics (h) 13.261 13.519 13.202 13.733 13.365 8.245 9.860 9.845 15.007 11.251 9.841 10.288 10.943 2.482 5.101 6.237 -4.484 -5.061 -1.332 -3.449 s(e) 0.219 0.126 0.118 0.103 0.093

Small 2 3 4 Big

0.036 0.042 0.045 0.018 0.012

0.014 0.010 0.013 0.008 0.004

0.316 0.248 0.158 0.160 0.123

0.201 0.166 0.130 0.114 0.088

0.175 0.102 0.097 0.085 0.076

0.125 0.116 0.097 0.086 0.071

36

Table 12 Regression Ri-Rf=ai+bi(Rm-Rf)+i

Ri R f = ai + bi ( Rm R f ) + i
The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we computed the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Size Small 2 3 4 Big Panel A - 1983-2002 Book-to-Market 4 High Low 2 0.615 0.713 0.874 0.974 1.093 0.489 0.746 0.905 0.987 1.046 9.848 19.099 31.464 39.719 56.918

Low 0.663 0.918 0.930 1.029 1.041

2 0.776 0.781 0.823 0.860 0.998

3 b 0.743 0.772 0.796 0.955 1.093

3 4 t-statistics (b) 16.002 17.104 19.628 23.884 29.779 36.359 34.703 34.921 47.282 44.218 50.797 60.410 68.701 68.616 71.936 s(e) 0.200 0.123 0.110 0.095 0.081

High 16.140 33.962 44.148 50.775 53.497

Small 2 3 4 Big

0.012 0.041 0.072 0.090 0.153

Adjusted R-Square 0.028 0.024 0.027 0.048 0.063 0.081 0.085 0.086 0.137 0.112 0.156 0.216 0.212 0.248 0.327

0.012 0.066 0.130 0.219 0.297

0.261 0.199 0.139 0.135 0.102

0.196 0.148 0.116 0.101 0.080

0.160 0.102 0.093 0.077 0.066

0.184 0.115 0.097 0.083 0.074

Size Small 2 3 4 Big

Low 0.706 0.810 0.843 0.961 1.009

2 0.852 0.769 0.858 0.897 1.068

3 b 0.839 0.836 0.885 1.049 1.146

Panel B - 1983-1993 Book-to-Market 4 High Low 2 0.697 0.792 0.874 0.986 1.091 0.483 0.820 0.912 0.983 1.068 13.094 21.724 28.940 40.452 58.425

3 4 t-statistics (b) 15.359 18.362 21.151 22.908 25.712 31.541 34.834 33.843 37.280 43.514 48.669 50.353 61.342 64.552 58.246 s(e) 0.167 0.121 0.101 0.089 0.073

High 9.255 28.800 32.507 40.581 43.952

Small 2 3 4 Big

0.045 0.100 0.129 0.198 0.306

Adjusted R-Square 0.053 0.062 0.065 0.091 0.106 0.137 0.169 0.165 0.191 0.230 0.272 0.303 0.352 0.396 0.432

0.009 0.115 0.169 0.271 0.326

0.171 0.129 0.111 0.098 0.077

0.186 0.123 0.097 0.084 0.074

0.136 0.103 0.093 0.074 0.064

0.245 0.114 0.103 0.086 0.081

37

Size Small 2 3 4 Big

Low 0.572 1.163 1.102 1.139 1.094

2 0.607 0.804 0.758 0.804 0.893

3 b 0.580 0.673 0.641 0.768 0.997

Panel C - 1993-2002 Book-to-Market 4 High Low 2 0.458 0.584 0.876 0.951 1.097 0.495 0.640 0.895 0.998 0.975 3.964 10.647 19.095 23.464 31.716

3 4 t-statistics (b) 6.684 7.374 7.710 12.784 15.936 18.450 16.858 15.914 29.108 23.410 22.547 34.886 37.316 34.464 43.308 s(e) 0.221 0.125 0.117 0.100 0.087

High 15.224 18.638 28.819 29.464 27.893

Small 2 3 4 Big

0.004 0.025 0.049 0.056 0.089

Adjusted R-Square 0.010 0.008 0.008 0.026 0.032 0.037 0.039 0.034 0.093 0.056 0.062 0.141 0.116 0.130 0.233

0.018 0.034 0.095 0.155 0.219

0.321 0.250 0.157 0.157 0.118

0.204 0.166 0.129 0.112 0.083

0.176 0.101 0.093 0.079 0.067

0.125 0.115 0.093 0.080 0.063

38

Table 13 Regression Ri-Rf=ai+bi(Rm-Rf)+hiHML+siSMB+i

Ri R f = ai + bi ( Rm R f ) + hi HML + si SMB + i
The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we computed the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. With these we form two new portfolios to capture the book-to-market and size effect. The first portfolio is the SMB (small minus big) that captures the size effect and is the difference between the returns of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. The second portfolio is the HML (high minus low) and captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Size Small 2 3 4 Big Small 2 3 4 Big Small 2 3 4 Big Small 2 3 4 Big Panel A - 1983-2002 Book-to-Market 4 High Low 2 0.630 0.880 1.042 1.089 1.090 0.651 0.679 0.647 0.544 0.432 0.799 0.707 0.673 0.500 0.167 0.035 0.112 0.184 0.258 0.312 12.937 22.161 36.560 43.143 57.883 3.433 -0.598 2.519 4.126 -2.416 16.229 17.425 25.614 19.753 10.697 0.256 0.194 0.134 0.133 0.102

Low 0.878 1.066 1.073 1.127 1.071 0.367 -0.048 0.114 0.162 -0.068 1.292 1.010 0.886 0.595 0.224 0.053 0.093 0.131 0.116 0.164

3 b 0.972 0.925 0.782 0.915 0.910 0.838 0.960 0.926 0.996 0.961 1.048 1.056 1.023 1.136 1.135 h 0.575 0.526 0.583 0.326 0.627 0.597 0.411 0.564 0.580 0.422 0.587 0.427 0.227 0.359 0.306 s 1.152 1.072 0.931 0.816 0.747 0.678 0.806 0.743 0.646 0.544 0.488 0.411 0.094 0.167 0.195 Adjusted R-Square 0.076 0.063 0.072 0.090 0.107 0.133 0.144 0.141 0.191 0.145 0.187 0.245 0.216 0.259 0.339

3 4 t-statistics (b) 20.045 21.209 24.903 28.009 35.121 42.987 40.903 40.979 54.290 49.146 55.499 65.100 68.820 69.777 73.580 t-statistics (h) 7.493 7.872 11.684 6.446 16.118 19.912 11.145 16.508 20.848 14.267 19.402 17.768 10.182 14.387 13.095 t-statistics (s) 20.176 21.236 25.724 21.493 25.477 30.037 29.161 28.685 30.715 24.630 21.963 22.041 5.561 8.811 11.110 s(e) 0.191 0.196 0.156 0.145 0.120 0.099 0.112 0.106 0.090 0.100 0.093 0.076 0.079 0.080 0.065

High 20.605 40.135 51.067 55.839 54.514 14.389 21.005 21.154 17.033 12.224 22.969 28.760 29.533 21.574 6.455 0.182 0.112 0.094 0.080 0.073

39

Size Small 2 3 4 Big

Low 0.890 0.990 1.003 1.064 1.044

2 1.075 0.942 1.022 0.982 1.073

3 b 1.024 1.006 1.037 1.142 1.165

Panel B - 1983-1993 Book-to-Market 4 High Low 0.883 0.931 0.994 1.044 1.101 0.632 0.936 1.028 1.045 1.076 16.351 26.812 34.756 43.889 58.351

3 4 t-statistics (b) 19.278 22.330 27.167 27.894 31.195 37.546 42.382 40.383 42.926 47.308 52.627 52.379 59.590 63.578 57.277 t-statistics (h) -0.490 2.161 -3.032 1.394 -0.545 1.704 2.723 4.664 1.116 6.283

High 11.746 32.894 36.839 43.020 43.108

Small 2 3 4 Big

h -0.082 -0.075 0.270 0.152 0.578 -0.125 -0.268 0.118 0.252 0.517 -0.309 -0.035 0.115 0.367 0.618 -0.124 0.154 0.278 0.280 0.796 -0.306 0.054 0.311 0.355 0.614 s 1.148 0.998 0.929 0.590 0.150

-0.531 -1.179 -3.984 -1.958 -6.528

1.657 4.270 3.733 6.925 5.814 8.217 5.476 11.094 6.868 8.359

Small 2 3 4 Big

1.058 1.046 0.898 0.556 0.111

1.283 0.917 0.926 0.523 0.038

1.121 0.877 0.761 0.393 0.125

1.014 0.811 0.773 0.436 0.090

13.580 19.372 21.943 16.348 4.507

t-statistics (s) 16.104 17.757 24.555 19.117 22.841 24.999 27.162 25.993 23.481 18.325 18.799 14.033 1.489 5.731 4.617 s(e) 0.162 0.115 0.095 0.087 0.072

13.462 20.697 19.800 12.491 2.443

Small 2 3 4 Big

0.090 0.173 0.199 0.229 0.311

Adjusted R-Square 0.107 0.117 0.146 0.151 0.184 0.218 0.260 0.253 0.265 0.270 0.313 0.329 0.352 0.403 0.440

0.032 0.178 0.237 0.312 0.338

0.167 0.124 0.107 0.096 0.076

0.180 0.119 0.092 0.082 0.074

0.130 0.098 0.088 0.073 0.064

0.242 0.110 0.099 0.083 0.080

Size Small 2 3 4 Big

Low 0.899 1.256 1.278 1.276 1.149

2 0.838 0.989 0.920 0.960 0.962

3 b 0.786 0.843 0.811 0.911 1.086 h 0.497 0.592 0.547 0.599 0.388 s 1.034 0.638 0.670 0.459 0.192

Panel C - 1993-2002 Book-to-Market 4 High Low 0.647 0.743 1.038 1.106 1.188 0.640 0.827 1.076 1.157 1.090 6.016 11.212 21.751 25.429 32.240

3 4 t-statistics (b) 8.948 9.653 10.550 15.312 19.315 22.820 20.079 19.544 33.730 27.122 25.953 39.032 38.420 35.906 45.162

High 19.130 23.355 33.673 33.171 29.885

Small 2 3 4 Big

0.711 -0.046 0.300 0.312 0.100

0.594 0.514 0.456 0.514 0.304

0.525 0.550 0.582 0.519 0.370

0.476 0.661 0.607 0.535 0.445

t-statistics (h) 3.904 5.095 4.921 6.737 -0.327 6.387 10.823 13.603 4.150 7.983 10.691 15.241 5.023 11.582 13.444 15.029 2.225 9.727 10.464 11.358 t(s) 12.851 14.611 16.422 13.268 6.413 s(e) 0.218 0.123 0.114 0.098 0.086

11.322 15.028 15.184 12.468 10.059

Small 2 3 4 Big

1.599 1.001 0.996 0.704 0.374

1.139 0.895 0.789 0.612 0.161

0.833 0.581 0.606 0.475 0.254

0.626 0.663 0.619 0.521 0.240

11.104 9.028 16.892 14.014 10.640

12.405 14.125 17.239 17.468 6.505

13.880 18.059 19.831 16.921 9.891

18.619 19.004 20.101 15.530 6.417

Small 2 3 4 Big

0.045 0.072 0.106 0.082 0.107

Adjusted R-Square 0.055 0.039 0.037 0.065 0.058 0.071 0.084 0.070 0.135 0.087 0.086 0.174 0.123 0.142 0.249

0.044 0.068 0.140 0.195 0.246

0.314 0.244 0.153 0.155 0.117

0.199 0.163 0.126 0.110 0.083

0.174 0.099 0.091 0.077 0.066

0.123 0.113 0.091 0.078 0.062

40

Table 14 Regression Ri-Rf=ai+bi(Rm-Rf)+siSMB+i

Ri R f = ai + bi ( Rm R f ) + si SMB + i
The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we computed the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. From this was calculated the SMB (small minus big) that captures the size effect and is the difference between the returns of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Size Small 2 3 4 Big Panel A - 1983-2002 Book-to-Market 4 High Low 2 0.719 0.775 0.931 1.006 1.101 0.567 0.803 0.959 1.025 1.048 12.504 22.681 36.638 43.003 58.821

Low 0.834 1.071 1.061 1.110 1.077

2 0.909 0.883 0.920 0.914 0.996

3 b 0.865 0.845 0.869 0.989 1.095 s 0.825 0.467 0.486 0.223 0.008

3 4 t-statistics (b) 18.983 20.091 23.129 27.296 32.694 39.775 39.474 38.503 50.732 47.112 52.362 62.253 67.896 67.822 71.834 t-statistics (s) 19.793 20.801 23.600 22.165 19.579 22.297 27.683 23.237 22.374 19.982 12.559 14.077 -1.000 0.536 3.608 s(e) 0.197 0.121 0.108 0.095 0.081

High 18.670 36.648 47.091 52.727 52.633

Small 2 3 4 Big

1.122 1.031 0.837 0.523 0.255

0.884 0.666 0.629 0.349 -0.013

0.662 0.405 0.379 0.213 0.049

0.508 0.380 0.347 0.250 0.011

17.975 22.338 29.230 21.345 15.505

17.864 19.714 20.323 13.715 0.503

Small 2 3 4 Big

0.052 0.093 0.131 0.116 0.164

Adjusted R-Square 0.070 0.058 0.063 0.087 0.089 0.110 0.136 0.123 0.166 0.134 0.166 0.227 0.212 0.248 0.328

0.026 0.088 0.156 0.235 0.297

0.256 0.194 0.134 0.134 0.102

0.191 0.145 0.112 0.100 0.080

0.157 0.100 0.091 0.076 0.066

0.183 0.113 0.096 0.082 0.074

41

Size Small 2 3 4 Big

Low 0.888 0.989 0.996 1.058 1.027

2 1.073 0.929 1.021 0.987 1.075

3 b 1.032 1.012 1.043 1.150 1.174 s 1.147 1.001 0.931 0.592 0.154

Panel B - 1983-1993 Book-to-Market 4 High Low 0.887 0.942 1.008 1.055 1.113 0.665 0.960 1.050 1.054 1.081 16.350 26.791 34.530 44.004 57.864

3 4 t-statistics (b) 19.306 22.595 27.345 27.710 31.630 38.181 42.547 40.933 43.623 47.755 53.109 53.125 59.918 64.028 57.884 t-statistics (s) 16.103 17.746 24.556 19.043 22.925 25.014 27.161 26.058 23.686 18.451 18.808 14.183 1.522 5.889 4.795 s(e) 0.162 0.115 0.095 0.087 0.073

High 12.478 33.861 37.576 42.830 42.951

Small 2 3 4 Big

1.059 1.047 0.895 0.552 0.104

1.283 0.914 0.925 0.527 0.039

1.121 0.878 0.769 0.398 0.131

1.037 0.823 0.778 0.415 0.075

13.597 19.398 21.838 16.268 4.243

13.786 20.933 19.787 11.753 2.029

Small 2 3 4 Big

0.090 0.173 0.197 0.229 0.307

Adjusted R-Square 0.107 0.116 0.146 0.149 0.184 0.216 0.260 0.252 0.261 0.269 0.311 0.326 0.352 0.399 0.435

0.030 0.172 0.227 0.293 0.327

0.167 0.124 0.107 0.096 0.076

0.180 0.119 0.092 0.082 0.074

0.130 0.098 0.088 0.073 0.064

0.243 0.110 0.099 0.084 0.081

Size Small 2 3 4 Big

Low 0.713 1.267 1.207 1.203 1.126

2 0.688 0.869 0.819 0.842 0.890

3 b 0.665 0.707 0.679 0.780 0.991

Panel C - 1993-2002 Book-to-Market 4 High Low 0.529 0.617 0.902 0.969 1.100 0.536 0.670 0.926 1.021 0.971 5.029 11.857 21.451 25.013 32.912

3 4 t-statistics (b) 7.715 8.553 8.972 14.015 16.784 19.558 18.515 17.008 30.204 24.664 22.860 35.582 37.046 34.116 43.315

High 16.590 19.591 30.031 30.330 27.649

Small 2 3 4 Big

s 1.163 0.765 0.729 1.029 0.580 0.279 0.814 0.517 0.335 0.515 0.299 0.097 0.313 -0.024 -0.044

0.512 0.341 0.249 0.256 0.249 0.243 0.151 0.192 0.025 -0.032

12.793 14.714 20.701 15.468 14.179

t-statistics (s) 13.824 14.208 13.987 15.229 14.549 9.752 11.757 11.510 16.833 12.728 12.513 13.087 13.320 4.422 8.269 9.120 -1.544 -2.219 1.564 -1.230 s(e) 0.218 0.124 0.115 0.100 0.087

Small 2 3 4 Big

0.042 0.072 0.104 0.080 0.107

Adjusted R-Square 0.050 0.036 0.032 0.058 0.044 0.052 0.076 0.055 0.110 0.074 0.065 0.149 0.116 0.131 0.233

0.035 0.046 0.115 0.169 0.219

0.315 0.244 0.153 0.155 0.117

0.200 0.164 0.127 0.111 0.083

0.174 0.100 0.092 0.079 0.067

0.124 0.114 0.092 0.079 0.063

42

Table 15 Regression Interception The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. To calculate the return rate, we considered the adjusted price of the last day of the month. Then we computed the monthly market return as the weighted-average return of all firms listed in the considered month. The risk free return rate used is the English three-month T-Bill return Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. These portfolios are designated LS, MS, HS, LB, MB and HB. With these we form two new portfolios to capture the book-to-market and size effect. The first portfolio is the SMB (small minus big) that captures the size effect and is the difference between the returns of large and small firms with similar book-to-market. It was calculated as the simple average of LS, MS and HS returns minus the simple average of LB, MB and HB returns. The second portfolio is the HML (high minus low) and captures the book-to-market effect. It is the difference between the average return of the higher BE/ME portfolios and the lower BE/ME portfolios. After this, all stocks were ranked in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were re-ranked again in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and an ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The regression was run for each portfolio.
Panel A - 1983-2002 Book-to-Market Size Low Small 2 3 4 Big -0.012 -0.020 -0.017 -0.015 -0.012 2 -0.012 -0.014 -0.015 -0.010 -0.005 a 3 -0.007 -0.015 -0.010 -0.008 -0.005 t-statistics(a) High Low 2 3 4 High Ri-Rf=ai+bi(Rm-Rf)+siSMB+i -0.005 0.003 -3.965 -5.501 -3.858 -3.832 2.141 -0.009 -0.003 -9.250 -9.973 -13.350 -10.462 -3.573 -0.010 -0.007 -13.945 -14.891 -9.631 -12.934 -8.417 -0.008 -0.006 -13.926 -12.427 -9.603 -11.120 -7.157 -0.005 -0.002 -15.520 -8.635 -7.357 -6.950 -1.778 4 Ri-Rf=ai+bi(Rm-Rf)+hiHML+siSMB+i -0.006 0.001 -4.023 -5.945 -4.189 -4.235 1.133 -0.009 -0.004 -9.212 -10.352 -14.138 -11.148 -4.723 -0.011 -0.008 -14.057 -15.468 -10.540 -14.296 -9.225 -0.009 -0.007 -14.115 -13.142 -10.751 -12.351 -7.509 -0.005 -0.002 -15.367 -9.048 -8.063 -7.805 -2.072

Small 2 3 4 Big

-0.012 -0.020 -0.018 -0.016 -0.012

-0.013 -0.015 -0.016 -0.011 -0.006

-0.008 -0.016 -0.010 -0.009 -0.006

43

Panel B - 1983-1993 Book-to-Market Size Low Small 2 3 4 Big 2 a 3 t-statistics (a) High Low 2 3 Ri-Rf=ai+bi(Rm-Rf)+siSMB+i -0.004 0.006 -2.722 -1.422 -2.274 -0.010 -0.005 -8.613 -8.556 -9.743 -0.011 -0.008 -9.940 -11.222 -7.573 -0.007 -0.005 -9.949 -7.638 -7.637 -0.005 -0.001 -8.575 -4.789 -5.329 4 Ri-Rf=ai+bi(Rm-Rf)+hiHML+siSMB+i -0.005 0.004 -2.595 -1.309 -0.011 -0.007 -8.254 -7.820 -0.012 -0.010 -8.919 -10.942 -0.008 -0.007 -9.393 -8.004 -0.006 -0.003 -7.393 -4.912 4 -2.501 -7.924 -9.029 -7.034 -4.557 High 2.353 -3.497 -5.470 -3.864 -0.577

-0.008 -0.004 -0.005 -0.017 -0.014 -0.016 -0.015 -0.014 -0.010 -0.012 -0.008 -0.009 -0.008 -0.004 -0.005

Small 2 3 4 Big

-0.007 -0.004 -0.006 -0.016 -0.013 -0.016 -0.013 -0.014 -0.010 -0.012 -0.009 -0.010 -0.007 -0.005 -0.006

-2.608 -9.837 -7.755 -8.399 -6.474

-2.749 -8.439 -9.970 -7.973 -5.847

1.458 -4.675 -6.820 -5.513 -2.086

Panel C - 1993-2002 Book-to-Market Size Low Small 2 3 4 Big 2 a 3 t-statistics (a) High Low 2 3 Ri-Rf=ai+bi(Rm-Rf)+siSMB+i -0.005 -0.001 -2.972 -6.080 -3.061 -0.008 -0.003 -6.176 -6.363 -9.534 -0.010 -0.007 -10.507 -10.344 -6.531 -0.009 -0.008 -10.716 -9.920 -5.779 -0.005 -0.002 -12.965 -6.858 -5.036 4 Ri-Rf=ai+bi(Rm-Rf)+hiHML+siSMB+i -0.005 -0.001 -2.774 -6.040 -0.007 -0.002 -6.182 -6.283 -0.010 -0.007 -10.380 -10.284 -0.008 -0.007 -10.607 -9.788 -0.005 -0.002 -12.956 -6.561 4 -2.581 -6.767 -9.729 -9.048 -5.360 High -0.932 -2.086 -6.721 -6.151 -1.946

-0.016 -0.019 -0.009 -0.024 -0.014 -0.014 -0.020 -0.017 -0.009 -0.018 -0.012 -0.007 -0.016 -0.006 -0.005

Small 2 3 4 Big

-0.015 -0.019 -0.008 -0.024 -0.014 -0.014 -0.020 -0.016 -0.009 -0.018 -0.012 -0.007 -0.016 -0.006 -0.005

-2.956 -9.330 -6.366 -5.708 -4.710

-2.433 -6.478 -9.855 -8.767 -5.373

-0.767 -1.973 -6.365 -5.878 -1.416

44

Table 16 Structure of the 25 portfolios The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into five quintiles. Next, the stocks were reranked in accordance with their market-equity value at the end of June of year t and divided into five new quintiles. This way, each stock belongs simultaneously to a certain BE/ME quintile and a ME quintile. The matching of these two sets of quintiles allowed the creation of 25 portfolios per year. The NYSE, AMEX and Nasdaq sub-sample data used covers the same period of time and was arranged by Fama and French from Compustat and CRSP data. BE is the COMPUSTAT book value of stockholders equity, plus balance sheet deferred taxes and investment tax credit (if available), minus the book value of preferred stock. The book-to-market ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with nonpositive book-equity have been excluded. Within each year, all stocks were ranked as in the London Stock Exchange sub-sample.
Size Low 2 3 4 High Book-to-Market Total Low LSE Average number of firms 2 3 4 High 39 53 62 50 58 67 58 58 63 69 62 59 78 63 47 294 294 298 3 4 High NASDAQ/AMEX/NYSE Average number of firms 2 3 4 High 451 449 518 829 153 151 125 94 105 93 76 54 82 71 61 43 71 57 49 32 862 821 830 1051 2 Total

Low Small 2 3 4 Big Total 35 38 56 71 80 279

Low 96 73 58 40 30 297 286 286 292 300 297 1462 786 231 162 119 117 1414

3033 755 491 375 324 4977

Average number of firms (%) Low 2 3 4 High Small 2.42% 2.67% 3.63% 4.25% 6.56% 2 2.61% 3.42% 3.99% 4.56% 5.01% 3 3.80% 3.95% 3.94% 4.32% 3.97% 4 4.82% 4.70% 4.25% 4.04% 2.72% Big 5.46% 5.34% 4.31% 3.21% 2.02% Total 19.12% 20.09% 20.13% 20.38% 20.28%

19.54% 19.59% 19.99% 20.54% 20.35% 100.00%

Average number of firms (%) Low 2 3 4 High 15.78% 9.06% 9.02% 10.42% 16.65% 4.65% 3.08% 3.04% 2.52% 1.88% 3.25% 2.11% 1.88% 1.53% 1.09% 2.39% 1.64% 1.42% 1.22% 0.86% 2.34% 1.42% 1.14% 0.98% 0.63% 28.41% 17.31% 16.50% 16.67% 21.11%

60.94% 15.16% 9.86% 7.53% 6.52% 100.00%

45

Table 17 Fama-French portfolios The London Stock Exchange sub-sample data used covers the period from December 1982 to June 2002. It includes all firms that have been listed for at least two years in the Worldscope UK 2003 database and have prices available in both December of year t-1 and June of year t. The book-tomarket ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with non-positive book-equity have been excluded. The market value is defined as the number of shares outstanding times the share price. The number of shares outstanding is updated when new stock is issued or when it changes (e.g. stock split). The book-equity value is calculated as the fixed assets less intangible assets minus total debt, minority interests and preferred stock. Within each year, we ranked all stocks in the sample based on their book-to-market ratio at the end of year t-1. Given these rankings the sample was divided into three book-to-market categories (High, Medium and Low with breakpoints at 30% and 70%). Next, the stocks were re-ranked in accordance with their market-equity value at the end of June of year t and divided into two size categories (Small and Big with breakpoint at 50%). This way, each stock belongs simultaneously to a certain BE/ME group and an ME group. The matching of these two sets of groups allowed the creation of 6 portfolios per year. For each portfolio we computed the weighted-average return. To calculate the return rate, we considered the adjusted price of the last day of the month. The NYSE, AMEX and Nasdaq sub-sample data used covers the same period of time and was arranged by Fama and French from Compustat and CRSP data. BE is the COMPUSTAT book value of stockholders equity, plus balance sheet deferred taxes and investment tax credit (if available), minus the book value of preferred stock. The book-to-market ratio (BE/ME) and market equity (ME) were then used to form the portfolios from July of t through June of Year t+1. All firms with nonpositive book-equity have been excluded. Within each year, all stocks were ranked as in the London Stock Exchange sub-sample.
Size Low Medium High LSE Average number of firms 155 281 281 272 308 166 427 588 447 Book-to-Market Total Low Medium High NASDAQ/AMEX/NYSE Average number of firms 716 1451 1301 1304 745 415 340 169 1462 1866 1641 1474 Average number of firms (%) 29.14% 26.12% 26.18% 8.33% 6.83% 3.40% 37.47% 32.95% 29.58% Total

Small Big

4057 924 4981

Small Big

Average number of firms (%) 10.62% 19.19% 19.20% 18.58% 21.06% 11.36% 29.20% 40.25% 30.56%

49.01% 50.99% 100.00%

81.44% 18.56% 100.00%

46