Estimation of Marginal Treatment Effects using Local
Instrumental Variables
Description
In the generalized Roy model, the marginal treatment effect (MTE) can be used as
a building block for constructing conventional causal parameters such as the average treatment
effect (ATE) and the average treatment effect on the treated (ATT) (Heckman, Urzua, and Vytlacil 2006
). Given a treatment selection model and an outcome model, the function mte()
estimates the MTE via local instrumental variables (or via a normal selection model) and also the
projection of MTE onto the 2-dimensional space of the propensity score and a latent variable representing
unobserved resistance to treatment (Zhou and Xie 2018 ).
The object returned by mte() can be used to estimate conventional parameters such as ATE and ATT
(via average()) or marginal policy-relevant treatment effects (via mprte()).