A Structural Approach to Measuring Time-varying Risk Aversion

Abstract

Non-homothetic preferences have the potential to rationalize important asset pricing facts including time-varying risk premia and business cycle movements in asset prices (e.g., Campbell and Cochrane (1999)). This paper offers a structural approach to measuring time-varying risk aversion. Motivated by the literature on consumption commitments (e.g., Flavin and Nakagawa (2008), Chetty and Szeidl (2016), Chetty, Sandor, and Szeidl (2017)), I develop a model in which investors have nonseparable preferences over housing and nonhousing consumption, and investors must consume a minimum amount of housing each period. Non-housing consumption is assumed to be flexibly chosen. The key insight is that the intratemporal optimality condition between the two goods reveals information about the surplus consumption ratio, a key variable driving risk aversion. A cointegrating relationship between relative quantities and prices allow us to identify the elasticity of intratemporal substitution and measure surplus housing consumption. Using aggregate U.S. consumption data from 1959 to the present, the measured surplus consumption ratio demonstrates clear business cycle fluctuations, rising during expansions and falling during recessions. Consistent with the theory, this measure also predicts future excess returns.