Simulated Data and Code for Adversarial Estimation of Structural Models

Zenodo

2023/09/01

This dataset accompanies the 2023 Econometrica paper by Tetsuya Kaji, Elena Manresa, and Guillaume Pouliot. The authors propose a novel adversarial estimation method for structural econometric models, framed as a minimax game between a data-generating model and a classifier. The dataset contains the simulated data and source code used to benchmark the estimator’s performance in several canonical structural model settings. The approach combines simulation-based estimation with neural network-based discriminators, enabling robust performance under both correct specification and model misspecification.