11/10/2016
More Info | Tim Christensen (NYU)
Nonparametric value functions and term structures
Many contemporary macro/finance models feature agents with recursive preferences or robust preferences. These models present a challenge for conventional semi/nonparametric estimation techniques such as GMM, as the pricing kernel is a function of the continuation value of the future consumption plan, which is unobservable when state dynamics are modeled flexibly. We describe sieve procedures for estimating the value function in models with recursive or robust preferences. The procedures solve a nonparametric fixed-point problem that is different from the usual Bellman equation obtained under time-separable preferences. To improve the finite-sample properties of the estimator, we introduce a new nonparametric regression procedure for estimating the conditional mean of a positive stochastic process. We apply the methodology to analyze the term structure of equity allowing for general nonlinear state dynamics. |