Identification and Estimation of Triangular Models without Exclusion Restrictions

43 Pages Posted: 22 Nov 2022

Date Written: November 16, 2022

Abstract

This paper studies different identification strategies for triangular simultaneous semipara- metric equation models without exclusion restrictions. We start our study with an identification strategy based on a functional form assumption. In this setting, we derive an orthogonal score function and we provide a two-step estimator procedure using neural networks for the estimation of the nuisance parameter. Later, we relax the functional form assumption and impose restrictions on the unobservables to obtain additional moment conditions. The restrictions rely on asymmetrically distributed or heteroskedastic errors terms. For estimation, we propose two-step semiparametric estimators. In the first step, we use neural networks for estimation of the nuisance parameters and in the second step we estimate the causal parameter of interest.

Keywords: Endogeneity, Artificial Neural Network, Identification, Control Function, Orthogo- nal score function

JEL Classification: C3, C31

Suggested Citation

Avila Marquez, Monika and Krishnakumar, Jaya, Identification and Estimation of Triangular Models without Exclusion Restrictions (November 16, 2022). Available at SSRN: https://ssrn.com/abstract=4278407 or http://dx.doi.org/10.2139/ssrn.4278407

Monika Avila Marquez (Contact Author)

University of Geneva ( email )

102 Bd Carl-Vogt
Genève, CH - 1205
Switzerland

Jaya Krishnakumar

University of Geneva ( email )

40 Bd. du Pont d'Arve
Genève 4, CH - 1211
Switzerland

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
6
Abstract Views
61
PlumX Metrics