A dynamic quantile model for distinguishing intertemporal substitution from risk aversion

Abstract

This paper uses a dynamic quantile model to estimate the elasticity of intertemporal substitution (EIS) and risk attitude using large disaggregated data from the NielsenIQ Consumer Panel. This data is transactional at the consumption purchase level, which minimizes measurement error, and the final sample contains more than two million observations. In the quantile model, the risk attitude is captured by the quantile and is, therefore, separable from the EIS. To estimate the parameters of interest we use the Euler equation along with instrumental variables quantile regression. First, we estimate the model across different levels of risk attitude. Empirical results document evidence of monotonically decreasing EIS along quantiles. For large risk aversion, the EIS is greater than one, whereas for small risk aversion it descends into negative values. Subsequently, we estimate the risk attitude and the EIS simultaneously. The results substantiate a risk attitude close to the median, with EIS consistently positive and smaller than one.