| In the first essay, I examine the relationship between the selection model and instrumental variable, local average treatment effect (LATE) frameworks. I show that the assumption of an unobserved index crossing a threshold that defines the selection model is equivalent to the independence and monotonicity assumptions at the center of the LATE approach. In the second essay, I consider the semiparametric identification of the average treatment effect within a selection model framework when the treated and the untreated outcome equations contain the same error term but the error term is not additively separable from the observed covariates in the outcome equations. In the third essay, I examine the use of a nonparametric selection model to identify or bound treatment parameters while allowing the treated and the untreated outcome equations to contain distinct error terms. |
