Heterogeneity and Heteroskedasticity in Endogenous Switching Models: Estimating the Effects of Physician Advice on Calorie Consumption
提出一种两步控制函数估计方法，用于处理二元处理变量内生转换模型中的异质性和异方差性，并应用该方法评估医生建议对卡路里消费的因果效应，发现传统方法低估了其效果。
ABSTRACT We describe two‐step control function estimation procedures for estimating average treatment effects in constant coefficient endogenous switching models with a binary treatment. We allow for additional heterogeneity by allowing for heteroskedasticity in the latent error in the treatment assignment equation. We apply our estimation procedures to evaluate the causal effect of physician advice on calorie consumption using National Health and Nutritional Examination data (2007–2016) and find that the effect of physician advice to lose weight on average calorie consumption roughly doubles compared to the traditional methods. Our findings recognize physician advice as a relevant low‐cost intervention in the campaign against rising obesity.