Handbook of Exchange Rates

15.1 Introduction

The expected volatilities and correlations of asset returns are a critical input in the optimal portfolio choice of a risk-averse investor. Extensive empirical evidence indicates that both the volatility of asset returns and their correlations change over time.¹ However, forecasting the dynamics of volatility and correlation requires estimation of suitable multivariate models, which are notoriously complicated and difficult to handle. This has spurred a large body of empirical research exploring tractable multivariate models of time-varying volatility.² Among them, the dynamic conditional correlation (DCC) model (Engle, 2002) has emerged as a benchmark, as it provides a parsimonious and flexible framework for modeling the dynamics of asset return volatilities and correlations. Hence, it can be readily used in realistic applications of dynamic asset allocation.

This chapter addresses an essential question that lies at the core of a long line of research in empirical finance: does volatility and correlation timing matter for the optimal asset allocation of a risk-averse investor and, if so, how? We contribute to the literature on the economic value of volatility timing, which focuses primarily on the dynamics of volatilities, while, however, for the most part treats the impact of dynamic correlations as an afterthought; in some cases correlations are assumed constant (e.g., Della Corte et al. 2009), and in other cases they are modeled using rolling estimators (e.g., ...