An empirical examination of the implications of arbitrage pricing theory

This paper presents a comprehensive set of tests of the implications of the Arbitrage Pricing Theory. We find, unlike previously reported results, a very limited relationship between the expected returns and the covariance (factor loadings) measures of risk. Furthermore, unique variance measures of risk, while generally making only small contributions to the explanation of asset returns, turn out to be significant about as frequently as the coveriance measures of risk — which is inconsistent with the Arbitrage Pricing Theory model. The intercept tests are more mixed but provide only limited support to the model.