Return Predictability: Can correlation effectively predict returns?

Sammanfattning: Previous research shows that index variance can be decomposed into average constituent correlation and average constituent variance. These studies hold that the average correlation captures features of the aggregate market risk and under a risk-reward relationship is a predictor of future excess returns. Based on these findings, this paper looks at contemporaneous and forecasting features of the risk variables, with market data from Eurostoxx50 and Swedish OMXS30. This study contributes to previous research by specifically studying the predictability of average correlation during market downturns and the conditional nature of risk and return. In accordance with previous research, the results confirmed that risk can be decomposed contemporaneously into average constituent correlation and average variance. When examining if average correlation is a predictor of future excess returns, a forecasting relationship is only found during market downturns, indicative of a conditional risk-reward trade-off. This is explained partly by the (i) Roll Critique, as different market proxies cause deviating results depending on the proxy's constituents, (ii) distinguishing changes in idiosyncratic risk from systematic risk as the risk proxy´s components vary over time, and (iii) the conditionality of market efficiency.