Are Two Factors Enough? The U.K. Evidence

Some studies in the 1990s documented that book value of equity to market value of equity (BV/MV) and the market value of equity (MVE) capture the cross-sectional variation of stock returns in the U.S. market in the 1963–90 period. Other researchers argued, however, that two other variables—the sales-to-price ratio (S/P) and the debt-to-equity ratio (D/E)—have more explanatory power for stock returns than BV/MV and MVE. The evidence in this article, from London Stock Exchange data, indicates that S/P and D/E do not entirely absorb the roles of BV/MV and MVE in explaining the cross-section of average stock returns in the U.K. market. We did find that S/P has significant explanatory power beyond the contribution of BV/MV and MVE, but the explanatory power of D/E is captured by S/P. The relationship between company-specific variables and subsequent stock returns has important practical applications—for portfolio selection and for other screening of individual stocks. The research into this topic has attracted considerable attention in the United States, but the extent to which the results obtained from U.S. stock data can be generalized to markets in other countries is a crucial question in asset pricing that has not received much attention. Resolving this issue is important not only for any person or organization diversifying investments internationally but also for the composition of domestic portfolios in the countries concerned.Our study is based on U.K. companies listed on the London Stock Exchange. The major U.S. studies on factor analysis provide the background to our analysis. In these studies, factors found to be influential in determining future stock returns are book value of equity to market value of equity (BV/MV), a size factor represented by market value of equity (MVE), the sales-to-price ratio (S/P), and the debt-to-equity ratio (D/E). We used data from the London Share Price Database and Datastream International to conduct an empirical analysis of the relationship of these factors to future returns for the period July 1980 through June 1996. Our final sample consisted of data for 1,420 nonfinancial U.K. companies.Accounting data for all fiscal year-ends in calendar year t − 1 were matched with market returns for July of year t to June of year t + 1. We used a company's accounting variables at the fiscal year-end t − 1 and formed portfolios as of the end of June of year t (for each year), which implies that we used only information that was available to the investor at the time of portfolio formation. We relied primarily on two methodologies in our empirical analysis—portfolio grouping and cross-sectional regression.In the portfolio-grouping approach, we formed portfolios by sorting stocks on this year's observable company-specific variables and comparing the portfolios' returns the following year. We formed the portfolios by (1) ranking the stocks on the basis of the chosen company-specific variable, (2) splitting the ranked list into 10 portfolios containing equal numbers of stocks, (3) calculating the average of the company-specific variable and the average of the following year's market return for the 10 portfolios, and (4) investigating the relationship between the average of the company-specific variable and the average of the following year's market return for each portfolio. We repeated this grouping procedure for the end of June of every year in the sample period; thus, the portfolios were rebalanced 16 times.Our findings for the London market were that BV/MV, MVE, S/P, and D/E are strongly related to future average stock returns. Some of the differences between the average returns of the first and tenth portfolios were striking. For example, when the stocks were sorted by BV/MV, the difference in returns between the top and bottom deciles turned out to be +18.8 percentage points (pps) annually. When MVE was used for the ranking, this return difference was −21.6 pps a year (i.e., smaller companies produced higher return). A similar exercise using S/P gave rise to a +18.6 pps a year difference, and using D/E produced a +15.2 pps a year difference.For the cross-sectional regressions, for each month of the sample period, we ran cross-sectional regressions of individual stock returns (not portfolio returns) on combinations of the company-specific variables. When we used data from the whole sample period, we found that, in contrast to evidence found in the U.S. stock market, the S/P and D/E variables did not absorb the roles of BV/MV and MVE in explaining the cross-section of average stock returns but that S/P has significant explanatory power beyond the contribution of BV/MV and MVE. We also found that the explanatory power of D/E was captured by S/P. To investigate the robustness of the cross-sectional regression results, we split the data into two subperiods and repeated the analysis. This time, S/P was significant for both subperiods but the influence of the other variables was less persistent; BV/MV was insignificant in Subperiod 1 but highly significant in Subperiod 2, and vice versa for MVE. The D/E factor was again dominated by S/P.Finally, we also ran regressions to test for an abnormal January influence. We did not find that a January effect was driving the results.