Forecasting Consumption Growth with the Real Term Structure

From the log-linearized consumption Euler equation, consumption growth of any horizon m is a function of the expected real return of maturity m, and they are linked through the elasticity of intertemporal substitution (EIS). Instead of using only the 1- period return and consumption growth, this result allows us to use the term structure of interest rates to identify the EIS. Using quarterly US data from 1954Q1 to 2007Q4, GMM results show that the real term structure is unrelated to future consumption growth: after controlling for small sample bias, we cannot reject the hypothesis that the EIS is zero. However, allowing a break in 1979 changes the results dramatically: the EIS is around 0.4 in the first period and it drops to around 0.2 in the second period. Not only is the EIS smaller, the out-sample forecasting power of ex post real return also drops in the second subsample compared to a simple AR(1) model for consumption growth. I find a lower EIS also for annual data.