Associate Professor David Slichter, State University of New York at Binghamton
MRC INTEGRATIVE EPIDEMIOLOGY UNIT (IEU)
Tuesday, 10th January, 2017
13.00 – 14.00 - Room OS6 – Oakfield House
Associate Professor of Economics
State University of New York at Binghamton
"Testing Instrument Validity and Identification with Invalid Instruments"
Instrumental variables are a widespread method for measuring causal relationships. Their use is limited in many contexts, though, by the difficulty of arguing for the required validity assumption. I describe a framework for testing the validity of instruments using covariates. The motivating example comes from the literature on measuring the effects of education on earnings, in which Card (1995) proposes that growing up near a university can be used as an instrument for the wage returns to attending college. I focus on the case of a binary instrument for a binary treatment, but the approach can be extended to continuous variables. The approach allows quantification of the extent of the violation of the validity assumption, and therefore can provide point or set identification of the treatment effect of interest even when the proposed instrument is invalid. Applying the approach, I find that the proximity to college instrument is invalid and likely overstates the returns to college.
David Slichter is Assistant Professor of Economics at the State University of New York at Binghamton. He completed his PhD in Economics from the University of Rochester in the spring of 2016. His main research interest is in understanding the role of early life inputs, especially school and teacher quality, in skill development. He is also involved in developing new methods for measuring treatment effects in observational data, and in developing tests for common assumptions in empirical work.