Abstract
1- Introduction
2- Data
3- The time-varying relation between inflation and consumption
4- The time-varying inflation risk premium
5- Origins of the time-varying inflation risk premium
6- Model
7- Calibration
8- Robustness checks
9- Conclusion
References
Abstract
We show that inflation risk is priced in stock returns and that inflation risk premia in the cross-section and the aggregate market vary over time, even changing sign as in the early 2000s. This time variation is due to both price and quantities of inflation risk changing over time. Using a consumption-based asset pricing model, we argue that inflation risk is priced because inflation predicts real consumption growth. The historical changes in this predictability and in stocks’ inflation betas can account for the size, variability, predictability, and sign reversals in inflation risk premia.
Introduction
We show that inflation risk is priced in stock returns and that both the price and quantities of inflation risk are strongly time-varying, even changing sign over our sample from 1962 to 2014. We argue that inflation is risky because it predicts real consumption growth in a time-varying way. Positive shocks to inflation sometimes contain bad news for future consumption, whereas at other times they contain good news. We provide new empirical evidence, for both the cross-section of stock returns and the aggregate stock market, and develop an equilibrium model to argue that it is the time-variation in the predictive content of inflation that drives the observed time-variation in the price and quantities of inflation risk. To quantify the predictive content of inflation at a given point in time, we define the nominal-real covariance as the slope coefficient in a conditional (that is, rolling) regression of real consumption growth on lagged inflation. We find that this nominal-real covariance is an economically strong and statistically significant predictor of the inflation risk premium in the cross-section of stocks. We measure the quantity of inflation risk of stocks by their conditional inflation beta, that is, the slope coefficient in a rolling regression of returns on inflation shocks. A one standard deviation increase in the nominal-real covariance predicts an increase in annualized expected return of 5.3% for a high-minus-low portfolio constructed by sorting stocks in deciles according to their inflation beta. Given an unconditional average return for the high-minus-low inflation beta portfolio of -4.2%, this increase implies that the inflation risk premium can switch sign.