Work

A Semiparametric Machine Learning Estimator for Sample Selection Models

Year: 2020
Type: report
Abstract: We propose a semiparametric machine learning estimator for sample selection models that is asymptotically normal and √ n-consistent, and show its improvement over existing methods in an empirical Monte Carlo.The simulated data relates to prison sentencing and is closely aligned to real court case data from North Carolina.The improvement is particularly pronounced in the absence of an exclusion restriction and when the proportion of censored observations increases; in some cases, mean squared error is reduced by more than 99%.Finite-sample evidence of the accuracy of estimated standard errors is also provided.
Language: English
Open Access status: green
OpenAlex ID: W6987154109
Cites: 20
Cited by:
Related to: 10