Estimating Parameters of Structural Models Using Neural Networks
49 Pages Posted: 9 Dec 2019 Last revised: 7 Dec 2023
Date Written: December 1, 2023
Abstract
We explore an alternative use of machine learning. We train neural nets to provide the estimate for the parameter of a given (structural) econometric model, e.g., discrete choice, consumer search. The training examples consist of datasets generated by the econometric model under a range of parameter values. The neural net takes the moments of a dataset as input and tries to recognize the parameter value underlying that dataset. In addition to point estimate, the neural net can also be trained to provide statistical accuracy. We establish that this neural net estimator (NNE) converges to limited-information Bayesian posterior when the number of training datasets is sufficiently large. We compare NNE to the prevailing estimation approach in a consumer sequential search application. NNE gives accurate and robust estimates at light computational costs. We discuss more broadly what types of applications are suitable (and unsuitable) for NNE.
Keywords: neural networks, machine learning, structural estimation, redundant moments, simulation burden, sequential search
Suggested Citation: Suggested Citation