Beschreibung:
Investors rely on the stock-bond correlation for a variety of tasks, such as forming optimal portfolios, designing hedging strategies, and assessing risk. Most investors estimate the stock-bond correlation simply by extrapolating the historical correlation of monthly returns and assume that this correlation best characterizes the correlation of future, annual or multi-year returns, but this approach is decidedly unreliable. The authors introduce four innovations for generating a reliable prediction of the stock-bond correlation. First, they show how to represent the correlation of single period cumulative stock and bond returns in a way that captures how the returns drift during the period. Second, they identify fundamental predictors of the stock-bond correlation. Third, they model the stock-bond correlation as a function of the path of some fundamental predictors rather than single observations. And fourth, they censor their sample to include only relevant observations, in which relevance has a precise mathematical definition