Introduction to Instrumental Variables
Instrumental variables (IV) analysis is the leading strategy for dealing with endogenous treatments in observational causal inference: IV can recover causal effects where OLS fails. The price, as always, is untestable assumptions. This session reviews the logic of IV with a focus on the substantive interpretation of the assumptions. We will develop visual intuition for different types of exclusion violations and understand why deviations from linearity greatly infringe upon the magic of IV. If there is time, we will discuss the use of genetic information as IV for the causal transmission of obesity in social networks.
Angrist, Joshua D., and Jörn-‐Steffen Pischke. 2009. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton: Princeton University Press. Pp. 113-138, 150-173.
O'Malley, A. J., Elwert, F., Rosenquist, J. N., Zaslavsky, A. M. and Christakis, N. A. (2014), Estimating peer effects in longitudinal dyadic data using instrumental variables. Biometrics, 70: 506–515.