The bivariate probit model, maximum likelihood estimation, pseudo true parameters and partial identification

Research output: Contribution to journalArticleResearchpeer-review

Abstract

This paper examines the notion of “identification by functional form” for two equation triangular systems for binary endogenous variables by providing a bridge between the literature on the recursive bivariate probit model and that on partial identification. We evaluate the impact of functional form on the performance of (quasi) maximum likelihood estimators, and investigate the practical importance of available instruments in both cases of correct and incorrect distributional specification. Finally, we calculate average treatment effect bounds and demonstrate how properties of the estimators are explicable via a link between the notion of pseudo-true parameters and the concepts of partial identification.

Original languageEnglish
Pages (from-to)91-113
Number of pages20
JournalJournal of Econometrics
Volume209
Issue number1
DOIs
Publication statusPublished - Mar 2019

Keywords

  • Average treatment effect
  • Binary outcome models
  • Copula
  • Identified set
  • Instrumental variables
  • Misspecification

Cite this

@article{b6eaf87662064761a42c4e5b60d0b3d7,
title = "The bivariate probit model, maximum likelihood estimation, pseudo true parameters and partial identification",
abstract = "This paper examines the notion of “identification by functional form” for two equation triangular systems for binary endogenous variables by providing a bridge between the literature on the recursive bivariate probit model and that on partial identification. We evaluate the impact of functional form on the performance of (quasi) maximum likelihood estimators, and investigate the practical importance of available instruments in both cases of correct and incorrect distributional specification. Finally, we calculate average treatment effect bounds and demonstrate how properties of the estimators are explicable via a link between the notion of pseudo-true parameters and the concepts of partial identification.",
keywords = "Average treatment effect, Binary outcome models, Copula, Identified set, Instrumental variables, Misspecification",
author = "Chuhui Li and D.S. Poskitt and Xueyan Zhao",
year = "2019",
month = "3",
doi = "10.1016/j.jeconom.2018.07.009",
language = "English",
volume = "209",
pages = "91--113",
journal = "Journal of Econometrics",
issn = "0304-4076",
publisher = "Elsevier",
number = "1",

}

The bivariate probit model, maximum likelihood estimation, pseudo true parameters and partial identification. / Li, Chuhui; Poskitt, D.S.; Zhao, Xueyan.

In: Journal of Econometrics, Vol. 209, No. 1, 03.2019, p. 91-113.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - The bivariate probit model, maximum likelihood estimation, pseudo true parameters and partial identification

AU - Li, Chuhui

AU - Poskitt, D.S.

AU - Zhao, Xueyan

PY - 2019/3

Y1 - 2019/3

N2 - This paper examines the notion of “identification by functional form” for two equation triangular systems for binary endogenous variables by providing a bridge between the literature on the recursive bivariate probit model and that on partial identification. We evaluate the impact of functional form on the performance of (quasi) maximum likelihood estimators, and investigate the practical importance of available instruments in both cases of correct and incorrect distributional specification. Finally, we calculate average treatment effect bounds and demonstrate how properties of the estimators are explicable via a link between the notion of pseudo-true parameters and the concepts of partial identification.

AB - This paper examines the notion of “identification by functional form” for two equation triangular systems for binary endogenous variables by providing a bridge between the literature on the recursive bivariate probit model and that on partial identification. We evaluate the impact of functional form on the performance of (quasi) maximum likelihood estimators, and investigate the practical importance of available instruments in both cases of correct and incorrect distributional specification. Finally, we calculate average treatment effect bounds and demonstrate how properties of the estimators are explicable via a link between the notion of pseudo-true parameters and the concepts of partial identification.

KW - Average treatment effect

KW - Binary outcome models

KW - Copula

KW - Identified set

KW - Instrumental variables

KW - Misspecification

UR - http://www.scopus.com/inward/record.url?scp=85058795974&partnerID=8YFLogxK

U2 - 10.1016/j.jeconom.2018.07.009

DO - 10.1016/j.jeconom.2018.07.009

M3 - Article

VL - 209

SP - 91

EP - 113

JO - Journal of Econometrics

JF - Journal of Econometrics

SN - 0304-4076

IS - 1

ER -