In Germany the CAPM is Alive and Well

Abstract

Using data on all firms listed in the top segment of the Frankfurt Stock Exchange during the years 1960 to 2007, we investigate how the (Sharpe-Lintner) CAPM performs under the assumption that the German capital market is totally segmented from other capital markets. We also check whether this model should be extended by the firm characteristics size and book-to-market. We can identify strong size and book-to-market effects in the German stock market. However, their direction, strength, and interaction are different in the two subperiods 1960-1990 and 1990-2007. We use the standard test procedures (BJS, GRS, Fama/MacBeth) to test the CAPM and do a large number of tests which differ by the length of the test period, the length of the return interval, beta calculations, firm level and portfolio data, sorting, and weighting. The total number of CAPM rejections is somewhat higher than what we would expect based on the statistical significance level. Long-term GRS tests often lead to rejections of the CAPM, especially in the second subperiod and in sorts on anomalies. Short term GRS-tests always reject the CAPM during the years 2000 to 2005. The results of Fama/MacBeth cross-sectional regressions depend on sorting, weighting and beta calculation. When we sort on beta and use value-weight portfolios the results for the full period, 1960 to 2007, are fully in line with the CAPM. Our interpretation of the results is that in Germany the pure domestic version of the CAPM works better than an extended model. It also works better for large firms than for small firms.