This research is concerned with the empirical verification of the assumptions necessary to ensure an unambiguous test of the arbitrage pricing model for the London Stock Exchange. More specifically, the purpose of this study is threefold: First, to test the normality assumption regarding the distributions of security returns and the intertemporal stationarity assumption of the security mean returns and the covariance (correlation) matrix of security returns. Second, to verify whether the number of common factors determining the security returns is the same across various groups of securities having different sizes and across different security groups having the same size. Third, to test whether the number of common factors affecting the security returns remain unchanged across various time periods for the same group of securities and across various time periods for different groups of securities. The research findings indicate that the distributions of security monthly returns are approximately normal and they are not intertemporally stationary. The correlation matrix of security returns seems to be stationary through time and thus the correlation matrix has to be used for the arbitrage pricing model's tests. Furthermore the number of factors changes as the group size changes. Such results highlight that the methodology used for testing the arbitrage pricing model is not the appropriate one, and previous tests of the arbitrage pricing model are not necessarily tests of the model. The arbitrage pricing model may be held, but the existing statitistical methodology does not provide an unambiguous test of the model for the London Stock Exchange. Finally, the number of factors changes across various time periods for the same group of securities and for different security groups. These findings suggest that the security returns generating model of the arbitrage pricing theory cannot be used for making predictions. These results, however, do not constitute evidence against the arbitrage pricing model. The arbitrage pricing model may be held, but the present state of the statistical methodology cannot be utilized to provide an unambiguous test of the model for the London Stock Exchange.
|Date of Award||1983|