New Methods for the Arbitrage Pricing Theory and the Present Value Model.

This book consists of two essays on new approaches for the Arbitrage Pricing Theory and the Present Value Model, and one essay on cross-sectional correlations in panel data. The new approaches are designed to study a large number of securities over time. They can be employed by security analysts to discover market anomalies without assuming observable factors or constant risk premium. The book shows how these two approaches can be used to determine how many systematic factors affect the U.S. stock market.

This dissertation consists of two essays on developing new methods for testing the Arbitrage Pricing Theory (APT) and the Present Value Model (PVM), and one essay on correcting heteroskedasticity and cross-sectional correlation in panel study by using the Newey-West Adjustment Matrix. In the first essay, I develop an autoregressive method for testing the APT. Unlike methods currently being used in the literature, this method does not require prior estimation of factor loadings and risk premia. The new methodology is based on the observation that past returns of an asset carry information about its exposure to systematic risks and thus can be used to construct ex post risk adjustments for the asset via a cross-sectional autoregressive model. I derive several testable implications of the APT and drop a crucial assumption that factor risk premia are constant. The approach is robust to changes in factor loadings in some cases. I find little evidence that firm size contribute additional explanatory power to that of factor loadings in the APT model. The second essay studies the rational expectations present value model with variable expected returns. I develop an econometric method with which (i) to test a general model of expected returns and (ii) to test a linearalized version of the present value model. I find that share dividend-price ratios carry information about the structure of future dividend growth. I also find that the rejection of the present value model is dependent upon the variability of expected returns. The third essay is an outcome of joint work with Whitney Newey. We show that the Newey-West adjustment matrix can be very useful for correcting heteroskedasticity andcross-sectional correlation in panel studies. We apply that adjustment procedure to the vector autoregression model of Holtz-Eakin, Newey and Rosen and develop a chi-square test to determine the number of pervasive economic factors in an approximate factor model. Our empirical results suggest there are at least seven factors in the economy that affect returns of securities listed on the New York Stock Exchange from 1987-1988.