The book-to-market effect before and after the market decline of 2000.
Fama and French (1992) and others document and describe an extensive book-to-market (BEME) ratio effect in stock returns. Trecartin's (2000) results, however, question the pervasiveness of this finding across time periods. Thus, we reexamine the inter-temporal stability of the BEME effect by examining the months prior to and following the market decline of 2000. In general, we find that the BEME effect is much stronger during the bear market period, but it appears that much of the BEME effect during this period is driven by firms in the technology sectors.
It is now commonly accepted that high book value of equity to market value of equity (BEME) portfolios outperform low BEME portfolios in stock-price return. Fama and French (1992) were the first to document this relationship, and since that time many articles have examined the persistence of this relationship across time (e.g., Davis 1994 and Ciccone 2003) and across international boundaries (Lakonishok 1991).
The finance literature is also concerned with explaining the dominance of high BEME portfolios since this result is contrary to the risk-return relationship forecasted by the traditional capital asset pricing model (CAPM). Fama and French (1992, 1993) attribute the BEME effect to a risk factor not captured by the CAPM. According to Tai (2003), BEME is a risk factor that varies across time. However, the literature also indicates that the BEME effect may be the result of other factors. Loughran (1997) suggests that the dominance of high BEME portfolios is due to the January effect, but Best, Best, and Yoder (2000) use tests of stochastic dominance to show that the BEME effect is not driven by the January effect. The BEME effect is also shown by Griffin and Lemmon (2002) to result from the mispricing of distressed firms.
Lakonishok, Shleifer, and Vishny (1994) and Haugen and Baker (1996) attribute the BEME effect to irrational investors and inefficient markets. If the BEME effect is indeed the result of irrationality, it is likely to change as market perceptions change. In this paper, we investigate further the inter-temporal persistence of the BEME effect. Trecartin (2000) shows that BEME is positively and significantly related to stock-price return in only 43% of monthly regressions from July 1963 to December 1997. Trecartin's (2000) results suggest that the use of BEME in forming short-term portfolios is not necessarily appropriate. We continue Tracartin's (2000) work by examining the use of the BEME in forming short-term portfolios leading up to and following the market downturn in the late spring and summer of 2000. Because aggregate stock returns during this period are heavily influenced by internet and technology oriented firms, we also investigate the use of a BEME strategy for firms outside of these market sectors during the bull and bear markets covered in our analysis.
DATA AND METHODOLOGY
We collect data for the firms used in this study using Research Insight and the Center for Research in Security Prices (CRSP) databases. Our analysis period includes July 1997 through December 2002. As in Fama and French (1992), the book value of equity is taken from the fiscal year-end of the year prior to portfolio formation. Using Trecartin's (2000) approach, we collect the market value of equity at the end of June in the year of portfolio formation and compound daily returns to compute monthly returns for each month from July of one year through June of the following year. In all cases in which we cannot match Research Insight data firms to CRSP data firms, we delete those observations. We then examine the monthly returns for 10 equally-sized (i.e., decile) portfolios ranging from a high to low BEME portfolios. Because we cannot match all firms across Research Insight and CRSP, our sample has various-sized portfolios across the sample period (ranging from a low of 181 firms per decile portfolio to a high of 257 firms per decile portfolio when all available firms are included in the sample).
To ascertain the impact of the recent bull and bear market on the BEME strategy, we break the sample into two distinct time periods--July 1997 through June 2000, and July 2000 through December 2002. Over these time periods (and in aggregate), we calculate the compounded monthly return on each of the decile BEME portfolios. We then compare the return on the lowest decile rank portfolio (labeled "High BEME" and which includes the firms with the highest BEME ratio) to the highest decile rank portfolio (labeled "Low BEME"). This comparison also facilitates the use of a difference in means t-test to determine whether any differences in portfolio returns are statistically significant.
Finally, the BEME effect as documented in previous studies implies that returns should increase monotonically from low BEME portfolios to high BEME portfolios. Because our statistical tests revolve around the "extreme" decile portfolios (high BEME versus low BEME), which may mask underlying variations in returns on the decile portfolios, we also ascertain which portfolio (among the ten deciles) has the highest return for the given month. Although we do not perform statistical tests on these returns, this examination allows us to determine the stability of the BEME effect across rank portfolios in addition to time-specific variations in the BEME effect.
For each of our statistical tests and time period comparisons, we report results in which all available firms are included and the results once technology-oriented firms are excluded. We consider firms with 3-digit SIC codes of 2830, 3570, 3660, 3670, 3690, 7370 and 4800-4890 to be high-tech firms. Thus, we can determine the impact of the "technology bubble" on the BEME effect. Our decile portfolio sizes range from 147 firms per portfolio to 198 firms per portfolio across time periods when we exclude technology firms.
Although we calculate returns on all decile portfolios and conduct statistical tests for each of the 66 months in our sample, we report only summarized data in the interest of space (month by month results can be obtained from the authors upon request). Our first analysis is included in Table 1, which lists the number of months in which the High BEME portfolio return is greater than the Low BEME portfolio and vice-versa. Again, to help us determine the impact of technology-oriented firms on the BEME effect over our analysis period, we report two sample results--"All" firms in Panel A of Table 1 and "Non-tech" firms in Panel B.
For all firms over the entire sample period, we find that in 43 of the 66 months (65.2%), the High BEME portfolio has a higher return than the Low BEME portfolio. The frequency for the non-tech sample is exactly the same over the entire sample period. Although this is greater than the frequency predicted by chance (50%), in real terms the frequency is small enough for us to begin to question the economic significance of the BEME effect. When we divide the sample into the pre- and post-market decline periods (July 1997-June 2000 and July 2000-December 2002), we find something striking in the all firms sample. The High BEME portfolio return exceeds the Low BEME portfolio return in exactly 50% of the months (and vice-versa) during the pre-market decline period--exactly as predicted by chance. In the post market decline period, the High BEME portfolio return is higher than the Low BEME return 83.3% of the months. Thus, there appears to be a shift in the BEME effect that occurs around the time of the market decline of 2000. This strong BEME effect, however, appears to be driven by technology-oriented firms as indicated in the sub-sample periods of Panel B. The relative frequencies for the non-tech firms are similar across time periods and are close to the full sample period relative frequencies.
To put these results in better perspective, and to shed greater light on the economic significance of our results, we identify those return differences (High BEME portfolio return minus Low BEME portfolio return) that are statistically significant. In Table 2, we report the number of occurrences in which these return differences are statistically different (using a 10% significance level). As indicated in Table 1, there are a number of occurrences in which the Low BEME portfolio has a higher return than the High BEME portfolio (contrary to expectations derived from previous studies). Thus, we divide the results in Table 2 into whether the return difference (High BEME-Low BEME) is positive (i.e., the return on the High BEME portfolio is highest) or negative. We also report the results for the pre- and post-market-decline periods.
As shown for the entire sample period and all firms in Panel A, only 39.4% of all (66) months have a High BEME portfolio return that is statistically greater than the Low BEME portfolio return. Additionally, 12.1% of all months have a High BEME portfolio return that is statistically lower than the Low BEME portfolio return. Thus, our full sample findings appear to be consistent with Tracartin (2000). As the anecdotal evidence from Table 1 suggests, the BEME effect is stronger in the post-market decline period. During this period, 60% of all months have a High BEME return that is statistically greater than the Low BEME portfolio return, while only 1 month (3.3%) has a greater Low BEME return than High BEME return. Again, however, these results appear more pervasive for technology firms. As Panel B shows for the non-tech firms, only 43.3% of the post-market decline months have a High BEME portfolio return that is statistically greater than the Low BEME portfolio return. Thus, given the results in Tables 1 and 2, we cannot support the universality of the BEME effect. This effect appears to be concentrated in certain time periods and certain sectors.
Next, we report the number of times that a particular decile portfolio has the highest return in a given month. The frequencies (divided by all firms and non-technology firms only) for the entire sample time period appear in Table 3. Portfolio 1 represents the decile portfolio with the highest BEME firms and Portfolio 10 represents the portfolio with the lowest BEME firms. Although we provide no statistical tests on these frequencies, we should see a large frequency associated with Portfolio 1 and little or no frequencies among the other portfolios if the BEME effect described initially by Fama and French (1992) is pervasive.
Surprisingly, for all firms, the most frequent portfolio with the highest return is the lowest BEME portfolio. In 16 months (24.2%) out of the entire sample period, this portfolio has the highest return. Further, although the two portfolios with the highest BEME have the higher returns 33.3% of the time, the two portfolios with the lowest BEME have the higher returns 34.2% of the time. The relative frequencies for the non-tech firms only are similar. The two highest BEME portfolios have the higher return 31.8% of the months, while the two lowest BEME portfolios have the higher return 30.4% of the months during the full sample period. Thus, we must seriously question the extensiveness of the BEME effect as documented by previous studies.
Finally, to highlight the inter-temporal variation in the BEME effect, we divide the frequencies from Table 3 into our two sample periods--July 1997-June 2000 and July 2000 December 2002. These results are in Table 4. In Panel A, which includes frequencies for the pre-market-decline period, the relative frequencies are similar to, if not stronger than, the results from Table 3. The results from the post-market decline period in Panel B, however, reveal a pattern. As was evident from our previous evidence, it appears that the BEME effect becomes stronger in this latter period. The portfolios with the higher BEME (that is, portfolios 1-5), overwhelmingly have the higher returns during this period. Although this is true for all firms and for the non-tech firms, it is apparent that the results are stronger among the tech firms. Thus, we are left with our initial conclusion-the BEME effect is largely time and sector dependent.
Fama and French (1992) and others document and describe an extensive book-to-market (BEME) ratio effect in stock returns. Trecartin's (2000) results, however, question the pervasiveness of this finding across time periods. Thus, we reexamine the inter-temporal stability of the BEME effect by examining the months prior to and following the market decline of 2000. This time period allows us to contribute to the literature on the BEME effect in two ways.
First, we are able to determine the impact of the BEME effect in a bull and bear market. In general, we find that the BEME effect is much stronger during the bear market period. Second, because the bull market of the late 1990s was driven primarily by technology-oriented stocks, we are able to subdivide our sample to determine whether the tech sector has a large influence on the BEME effect. Here, we find that BEME effect is generally non-existent in the pre-market decline even among tech stocks. In the post-market decline, however, much of the BEME effect we document can be attributed to the tech firm sectors.
Thus, given our findings, we are left to question the extent of the BEME effect. Although the BEME ratio may be systematically related to the returns of some firms during certain time periods, it seems difficult to expect that a typical investor could consistently profit from forming portfolios on this basis.
We leave for future research two items unexplored in our analysis. First, are there systematic seasonalities in our findings? Although the results of Best, Best and Yoder (2000) would suggest otherwise, we find much different results than in that research. Second, although we do not report this finding previously in this paper, we note that there appears to be a "momentum" effect within the decile portfolios that we form. That is, when a particular decile portfolio has the highest return in a given month, that same rank portfolio tends to repeat as the highest return portfolio in subsequent months. Thus, there may be fruitful trading strategies based on a BEME and momentum effect.
Best, R. J., R. W. Best & J. A. Yoder. (2000). Value stocks and market efficiency. Journal of Economics and Finance, 24, 28-35.
Ciccone, S. J. (2003). Forecast dispersion and error versus size, book-to-market ratio and momentum: A comparison of anomalies from 1992 to 2001. Journal of Asset Management, 3, 333-344.
Davis, J. L. (1994). The cross-section of realized stock returns: the pre-COMPUSTAT evidence. Journal of Finance, 49, 1579-1593.
Fama, E. F. & K. R. French (1992). The cross-section of expected stock returns. Journal of Finance, 47, 427-465.
Fama, E. F. & K. R. French (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3-56.
Griffin, J. M. & M. L. Lemmon (2002). Book to market equity, distress risk, and stock returns. Journal of Finance, 57, 2317-2336.
Haugen, R. A. & N. L. Baker (1996). Commonality in the determinants of expected stock returns. Journal of Financial Economics, 41, 401-439.
Lakonishok, J., A. Shleifer, & R. Vishny (1994). Contrarian investment, extrapolation, and risk. Journal of Finance, 49, 1541-1578.
Loughran, T. (1997). Book-to-market across size, exchange, and seasonality: Is there an effect? Journal of Financial and Quantitative Analysis, 32(3), 249-268.
Tai, CS. (2003). Are Fama and French and momentum factors really priced? Journal of Multinational Financial Management, 13, 359-384.
Trecartin Jr., R. R. (2000). The reliability of the book-to-market ratio as a risk proxy. Financial Services Review, 9, 361-373.
Elaine Jones, Central Missouri State University
Roger J. Best, Central Missouri State University
Table 1: Frequency of Returns on High BEME Portfolio versus Low BEME Portfolio Panel A: All Firms High BEME> Low BEME Time Period Low BEME % of Months > High BEME % of Months 07/97-12/02 43 65.2% 23 34.8% 07/97-06/00 18 50.0% 18 50.0% 07/00-12/02 25 83.3% 5 16.7% Panel B: Non-tech Firms High BEME> Low BEME Time Period Low BEME % of Months > High BEME % of Months 07/97-12/02 43 65.2% 23 34.8% 07/97-06/00 20 55.6% 16 44.4% 07/00-12/02 19 63.3% 11 36.7% Table 2: Frequency High BEME Return--Low BEME Return Is Statistically Significant Panel A: All Firms Number of Months: High-Low is High-Low is Time Period Positive % of Months Negative % of Months 07/97-12/02 26 39.4% 8 12.1% 07/97-06/00 8 22.2% 7 19.4% 07/00-12/02 18 60.0% 1 3.3% Panel B: Non-Tech Firms Number of Months: High-Low is High-Low is Time Period Positive % of Months Negative % of Months 07/97-12/02 18 27.3% 3 4.5% 07/97-06/00 5 13.9% 2 5.6% 07/00-12/02 13 43.3% 1 3.3% Table 3: Number of Months in which Each Decile Portfolio Has Highest Return All Firms Non-Tech Firms Portfolio Frequency % of Months Frequency % of Months 1 7 10.6% 13 19.7% 2 15 22.7% 8 12.1% 3 8 12.1% 6 9.1% 4 3 4.5% 6 9.1% 5 6 9.1% 9 9.1% 6 3 4.5% 2 3.0% 7 0 0.0% 1 1.5% 8 1 1.5% 1 1.5% 9 7 10.6% 10 15.2% 10 16 24.2% 10 15.2% Table 4: Number of Months in which Each Decile Portfolio Has Highest Return by Time Periods Panel A: July 1997-June 2000 All Firms Non-Tech Firms Portfolio Frequency % of Months Frequency % of Months 1 4 11.1% 11 30.6% 2 8 22.2% 4 11.1% 3 3 8.3% 3 8.3% 4 0 0.0% 0 0.0% 5 0 0.0% 0 0.0% 6 2 5.6% 1 2.8% 7 0 0.0% 1 2.8% 8 1 2.8% 1 2.8% 9 6 16.7% 10 27.8% 10 12 33.3% 5 13.9% Panel B: July 2000-December 2002 All Firms Non-Tech Firms Portfolio Frequency % of Months Frequency % of Months 1 3 10.0% 2 6.7% 2 7 23.3% 4 13.3% 3 5 16.7% 3 10.0% 4 3 10.0% 6 20.0% 5 6 20.0% 9 30.0% 6 1 3.3% 1 3.3% 7 0 0.0% 0 0.0% 8 0 0.0% 0 0.0% 9 1 3.3% 0 0.0% 10 4 13.3% 5 16.7%
|Printer friendly Cite/link Email Feedback|
|Author:||Jones, Elaine; Best, Roger J.|
|Publication:||Academy of Accounting and Financial Studies Journal|
|Date:||Jan 1, 2005|
|Previous Article:||On corporate debt policy.|
|Next Article:||The integration of capital markets: correlation analyses of the market indexes in Greater China economies.|