Author: Hill, Steven C. ; Kreider, Brent
Working Paper: Partially Identifying Treatment Effects with an Application to Covering the Uninsured (PDF) ; April 2006
We extend the nonparametric literature on partially identified probability distributions and treatment e¤ects. We then use our analytical results to bound the impact of universal health insurance on provider visits and medical expenditures. Our approach accounts for uncertainty about the reliability of self-reported insurance status as well as uncertainty created by unknown counterfactuals. As part of the contribution, we provide sharp bounds on the conditional mean of a random variable for the case that a binary conditioning variable is subject to arbitrary endogenous measurement error.
Using data from the 1996 Medical Expenditure Panel Survey (MEPS), we construct health insurance validation data for a nonrandom portion of the sample based on insurance cards, policy booklets, and follow-back interviews with employers and insurance companies. Under relatively weak verification and monotonicity assumptions, we estimate that monthly per capita provider visits under universal coverage would rise by no more than four-tenths of a visit (a 9% change) across the nonelderly population and that per capita expenditures would rise by less than 15%.