Abstract

This study addresses the problem of accurately forecasting and attributing risk in equity portfolios. We develop a hybrid methodology which takes advantage of the superior forecasting power of implicit factor models while also attributing portfolio risk to economic factors and firm-specific characteristics. We then compare the relative accuracy of risk attribution using our hybrid approach versus an explicit cross-sectional factor model. We present simulation results which suggest, given realistic parameter values, that the estimation efficiency gained by using the hybrid approach yields substantial improvements over explicit models.

13.1 Introduction

It is well known that the tendency of stock prices to move together is the primary source of return risk for equity portfolios containing more than just a few stocks. Factor models are used to describe and predict these price co-movements across stocks. The factor models can, roughly, be categorized as either implicit or structural; implicit models infer the common factors driving stock returns by looking at the factors footprints in observed returns, while the structural models specify the factors a priori in terms of observable characteristics of stocks or macroeconomic variables.

It seems generally agreed that the implicit models afford superior prediction of stock return risk, e.g. King et al. (1994) who show that changes in stock price volatilities can generally be...