Showing 1–1 of 1 results for author: Wei, Y '
-
Pre-Training Estimators for Structural Models: Application to Consumer Search
Authors:
Yanhao 'Max' Wei,
Zhenling Jiang
Abstract:
We explore pretraining estimators for structural econometric models. The estimator is "pretrained" in the sense that the bulk of the computational cost and researcher effort occur during the construction of the estimator. Subsequent applications of the estimator to different datasets require little computational cost or researcher effort. The estimation leverages a neural net to recognize the stru…
▽ More
We explore pretraining estimators for structural econometric models. The estimator is "pretrained" in the sense that the bulk of the computational cost and researcher effort occur during the construction of the estimator. Subsequent applications of the estimator to different datasets require little computational cost or researcher effort. The estimation leverages a neural net to recognize the structural model's parameter from data patterns. As an initial trial, this paper builds a pretrained estimator for a sequential search model that is known to be difficult to estimate. We evaluate the pretrained estimator on 14 real datasets. The estimation takes seconds to run and shows high accuracy. We provide the estimator at pnnehome.github.io. More generally, pretrained, off-the-shelf estimators can make structural models more accessible to researchers and practitioners.
△ Less
Submitted 1 May, 2025;
originally announced May 2025.
