Sign and Quantiles of the Realized Stock-Bond Correlation

We scrutinize the monthly realized stock-bond correlation based upon high frequency returns. In particular, we use a probit model to track the dynamics of the sign of the correlation relative to its various economic forces. The sign is predictable to a large extent with bond market liquidity being the most important variable. Moreover, stock market volatility, inflation uncertainty, short rate volatility, and bond volatility have significant effects upon the sign. In addition, we use quantile regressions to pin down the systematic variation of the extreme tails of the realized stock-bond correlation over its economic determinants. We document that the correlation behaves di¤erently when it is large negative (0.10 quantile) as opposed to when it is large positive (0.90 quantile). Nevertheless, the empirical findings are only partially robust to using other, possibly less precise, measures of the stock-bond correlation.