Fedor Iskhakov, Australian National University

Nested Recursive Lexicographical Search (NRLS): Structural Estimation of Dynamic Directional Games with Multiple Equilibria. 

Abstract

We develop a fully robust full solution method for solving and estimating dynamic stochastic games models with multiple equilibria. The method is analogous to the Nested Fixed Point (NFXP) estimator by Rust, where the inner loop fixed point algorithm is replaced by Recursive Lexicographical Search algorithm by Iskhakov, Rust, Schjerning (2016). Until recently, NFXP have generally not been possible to implement since no algorithm is guaranteed to find all MPE. RLS algorithm fills this gap for directional dynamic games.

In models with billions of MPE's, our full solution method may be theoretically ideal, but computationally impractical due to the vast number of equilibria that needs to be computed. We propose a several numerical approaches to alleviate this problem. We show that our recursive MLE is computationally efficient and asymptotically equivalent to MLE. We compare it's performance to a variety of estimators and find our estimator to be remarkably robust, computational fast and is able to both obtain efficient MLE of the structural parameters and at the same time identify the equilibrium selection played in the data.

CCE organizes a weekly seminar, usually on Mondays from 13:00 to 14:15. These seminars are open to everyone.   

The seminar covers topics from all research fields in structural econometrics and computational economics including theoretical, empirical, methodological issues.