Approximate factor structures, macroeconomic and financial factors, unique and stable return generating processes and market anomalies: An empirical investigation of the robustness of the arbitrage pricing theory

Abstract

This thesis presents an empirical investigation into the Arbitrage Pricing Theory (APT). At the onset of the thesis it is recognised that tests of the APT are conditional on a number of preconditions and assumptions. The first line of investigation examines the effect of the assumed nature of the form of the return generating process of stocks. It is found that stocks follow an approximate factor structure and tests of the APT are sensitive to the specified form of the return generating process. We provide an efficient estimation methodology for the case when stocks follow an approximate factor structure. The second issue we raise is that of the appropriate factors, the role of the market portfolio and the performance of the APT against the Capital Asset Pricing Model (CAPM). The conclusions that we draw are that the APT is robust to a number of specified alternatives and furthermore, the APT outperforms the CAPM in comparative tests. In addition, within the APT specification there is a role for the market portfolio. Through a comparison of the results in chapters 2 and 3 it is evident that the APT is not robust to the specification of unexpected components. We evaluate the validity of extant techniques in this respect and find that they are unlikely to be representative of agents actual unexpected components. Consequently we put forth an alternative methodology based upon estimating expectations from a learning scheme. This technique is valid in respect to our prior assumptions. Having addressed these preconditions and assumptions that arise in tests of the APT a thorough investigation into the empirical content of the APT is then undertaken. Concentrating on the issues that the return generating process must be unique and that the estimated risk premia should be stable overtime the results indicate that the APT does have empirical content. Finally, armed with the empirically valid APT we proceed to analyse the issue of seasonalities in stock returns. The results confirm previous findings that there are seasonal patterns in the UK stock market, however, unlike previous findings we show that these seasonal patterns are part of the risk return structure and can be explained by the yearly business cycle. Furthermore, the APT retains empirical content when these seasonal patterns are removed from the data. The overall finding of this thesis is that the APT does have empirical content and provides a good description of the return generating process of UK stocks.