Stock-Bond Correlation and Out-of-Sample Portfolio Performance Using Analyst Forecasts

Abstract

In this paper I examine whether one can use analyst forecasts of macroeconomic variables to improve investors ex-ante allocation of wealth between stocks and bonds. Such forecasts provide a forward-looking approach which I find improves investor's information set for the myopic stock-bond portfolio. In-sample I find that the best forecast of the volatility and correlation is simply the previous realised estimate. However, out-of-sample, analyst forecasts provide information not contained in historical/realised data which improves investors' diversification benefits. Out-of-sample performance is evaluated using five metrics: volatility, Sharpe ratio, certainty-equivalent return, turnover and opportunity cost. For minimum-variance portfolios formed using analyst forecasts, although the volatility of the portfolios increase, the Sharpe ratios substantially increase compared with an equally-weighted portfolio. Whilst the portfolio turnover is higher, the certainty equivalent (riskless return investor would accept to forego risky strategy) is also larger. I find that the opportunity cost, which measures the economic value of timing stock-bond correlation using analyst forecasts is 1.56% annually. Robustness checks highlight the use of analyst forecasts during the financial crisis, substantially improving the stock-bond portfolio performance versus other methods. Also, the benefits of these forecasts seem to persist across time horizon: for different holding periods, those portfolios formed using analyst forecasts consistently outperform all other methods.