Time-Varying Risk Premia and Profits from Portfolio Trading Strategies in the U.S. Stock Markets

Abstract

This paper re-examines the profitability of two portfolio trading

strategies that are currently the most controversial in financial

research: the relative-strength strategy based on medium-term return

continuation (3 to 12 months) and the contrarian strategy based on

long-term return reversals (2 to 5 years). Using a sample of 1,500

stocks listed on the NYSE and AMEX from 1963 to 1989, the bootstrap

test result shows that large parts of the profits to the relative-strength

strategy can be explained by time-varying expected returns estimated

from a bivariate GARCH model for the conditional CAPM. This result

generally holds, even within subsamples classified by other measurees

of risk such as firm sizes and market model betas, except for the

medium- and large-size groups. However, profits to the contrarian

strategy are shown to be the most difficult to reconcile with existing

asset pricing models. The bootstrap distributions for the contrarian

profits under any null models average significantly lower profits than

the actual distributions at the 10% significance level. This indicates

that the asset pricing models are incapable of explaining long-term

mean reverting behavior of stock returns.