Nonlinear Binscatter Methods

RePEc Short ID: pcr107
ORCID iD: 0000-0002-7958-1818

Cattaneo, Matias D.; Crump, Richard K.; Farrell, Max H.; Feng, Yingjie

doi:10.59576/sr.1110

Report

Nonlinear Binscatter Methods

Abstract: Binscatters are a powerful tool for empirical work in the social, behavioral, and biomedical sciences. Available tools rely on least squares estimation of the conditional mean. We introduce novel binscatter methods based on nonlinear, possibly nonsmooth M-estimation, covering generalized linear, robust, and quantile regression models. We provide theoretical results and practical tools, including optimal bin selection, confidence bands, and statistical tests regarding functional form or shape restrictions. We demonstrate our methods by studying the relationship of income and (lack of) health insurance. We provide software for Python, R, and Stata. Our technical results may be of independent interest.

JEL Classification: C14; C18; C21;

https://doi.org/10.59576/sr.1110