Recursive Utility and the Equity Premium Puzzle: A Discrete-Time Approach

Abstract

We study the Epstein-Zin model with recursive utility. Recognizing that recursive preferences implies that the underlying model is not Markovian, we use methods not depending upon the Markov property to solve the model. We work with the returns directly, which we approximate by Taylor series expansions, in log terms. Leaving out moments of order three and higher, we calibrate the resulting model to the data of Mehra and Prescott (1985) under various assumptions about the wealth portfolio. The results are very reasonable for the US-data. We also calibrate to a newer Norwegian data set, where we also have the relevant estimates for the national wealth portfolio. Again, the results are consistent with plausible values for the preference parameters.

Keywords: Recursive utility, the Epstein-Zin model, utility gradients, calibrations, the Markov property

JEL Classification: G10, G12, D9, D51, D53, D90, E21