Econometrics Seminar
Speaker: Ertian Chen (UCL)
Title: Model-Adaptive Estimation of Dynamic Discrete Choice Models with Large State Spaces
Format: Hybrid
Homepage:
Organisers: Julien, Pietro
Abstract: We consider the estimation of Dynamic Discrete Choice (DDC) models with large state spaces using the Conditional Choice Probability (CCP) estimators. CCP estimation is a constrained optimization, where the constraint has to be solved numerically in models with large state spaces. Sieve approximation methods often are used to solve the constraint in practice. The performance of the approximation depends crucially on the choice of the sieve space chosen by the researcher. In this paper, we propose a model-adaptive sieve space, which is constructed by iteratively augmenting the space with the residual from the previous iteration. We show both theoretically and numerically that model-adaptive sieves can dramatically improve performance. In particular, the bias decays at a superlinear rate in the sieve dimension, unlike a linear rate achieved using conventional bases such as splines or polynomials. Moreover, our method also contributes to the full-solution method with policy iteration, as the same constraint needs to be resolved at each policy iteration step.