Files
Abstract
I study the term structure of stock-bond covariances after the Global Financial Crisis. The term structure of covariances refers to the covariance between the aggregate market and government bonds of different maturities. I establish five new facts about the term structure of covariances: (1) The average term structure of covariances is downward sloping: long-maturity covariances are more negative than their short-maturity counterparts, whereas the term structure is flat across maturities before the crisis. (2) After the crisis, short-maturity covariances are acyclical. Short-maturity covariances are of the same magnitude in good and bad times, and so are acyclical. In contrast, long-maturity covariances are more negative in bad times, and so are countercyclical. Taken together, the first two facts imply that short-maturity bonds have weaker hedging properties relative to long-maturity bonds. (3) To generalize the intuition from the first two facts, I run regressions of long-maturity covariances on short-maturity covariances in changes. Before 2007, the sensitivity of long to short, the regression slope, is about one and stable. After 2010, the sensitivity roughly doubles. (4) The sensitivity of long to short is related to unconventional monetary policy. First, in the United States, the sensitivity rises only after the announcement of large-scale asset purchases and varies with Post-GFC programs implemented by the Federal Reserve. Second, around the world, the rise in the sensitivity is larger in countries with larger quantitative easing programs. (5) The break in the term structure has implications for bond risk premia. I construct a stock-bond factor, analogous to the Cochrane-Piazzesi factor, from the term structure of covariances. After the crisis but not before, the stock-bond factor predicts bond excess returns, even after controlling for typical predictors of bond returns. These five facts provide new directions for the literature on stock-bond comovement.