Value Portfolio Theory

Abstract

This paper considers the role of valuation in portfolio theory. In particular, we examine value-price ratios and their dynamic properties. Five valuation principles are proposed, namely the separability of prices and valuations, the asymptotic convergence of value-price ratios, the finite horizon convergence of value-price ratios, the predictability of asset returns using value-price ratios, and the irrelevance of terminal values for return predictability. We then consider the use of value-price ratios in optimising portfolio weights. First, we estimate three systems of value-price return predictability using GMM. The first system is based on own-asset predictability, the second uses cross-asset return predictability, and the third uses a mean-reverting specification. The coefficients of return predictability and the covariance matrix of the residuals from the return predictability regressions are used to construct portfolio weights and form the value portfolios. For a sample of 60 US stocks from December 1979 to November 1998, the value portfolios are compared to standard Markowitz portfolios. The comparisons are invoked for portfolios formed across sectors, industries and sub-industries. The results show that value portfolios exhibit superior performance for portfolios where the assets are relatively homogeneous. The more refined the classification, the better the performance. Thus, value portfolios perform better for sub-industries than for industries, and better for industries than for sectors. In general, value-price predictability confers benefits for portfolio performance when portfolios are small and homogeneous.