Estimation and Inference by Stochastic Optimization: Three Examples
AEA Papers and Proceedings
vol. 111,
May 2021
(pp. 626-30)
Abstract
This paper illustrates two algorithms designed in Forneron and Ng (2020): the resampled Newton-Raphson (rNR) and resampled quasi-Newton (rQN) algorithms, which speed up estimation and bootstrap inference for structural models. An empirical application to BLP shows that computation time decreases from nearly five hours with the standard bootstrap to just over one hour with rNR and to only 15 minutes using rQN. A first Monte Carlo exercise illustrates the accuracy of the method for estimation and inference in a probit IV regression. A second exercise additionally illustrates statistical efficiency gains relative to standard estimation for simulation-based estimation using a dynamic panel regression example.Citation
Forneron, Jean-Jacques, and Serena Ng. 2021. "Estimation and Inference by Stochastic Optimization: Three Examples." AEA Papers and Proceedings, 111: 626-30. DOI: 10.1257/pandp.20211038Additional Materials
JEL Classification
- C15 Statistical Simulation Methods: General
- C61 Optimization Techniques; Programming Models; Dynamic Analysis
- C63 Computational Techniques; Simulation Modeling
- C23 Single Equation Models; Single Variables: Panel Data Models; Spatio-temporal Models
- C25 Single Equation Models; Single Variables: Discrete Regression and Qualitative Choice Models; Discrete Regressors; Proportions; Probabilities
- C26 Single Equation Models: Single Variables: Instrumental Variables (IV) Estimation