The average treatment effect and average partial effect in nonlinear models

In the literature on program impact evaluation, the popular impact parameters can the average treatment effect, the average treatment effect on the treated, the average partial effect, and the average partial effect on the treated. In empirical studies, these parameters are not always presented and estimated clearly. In addition, when outcome functions are nonlinear, the estimation of these parameters is not straightforward. This paper discusses the estimation of these parameters in nonlinear models of outcomes and illustrates the estimation in an example of a micro-credit program in Vietnam.