Comments on: 'The Trouble with Instruments: Re-Examining Shock-Based IV Designs' by Atanasov and Black

Abstract

Atanasov and Black (2015) (AB) analyzes potential limitations of empirical studies that use shock-based IV designs, focusing specifically on our article that studies the effect of board independence on firm value (Duchin et al., 2010). With regard to our study, AB raises three concerns with our analysis. This note presents our reaction to AB’s analysis. We agree with two of the concerns in the abstract; it turns out they do not matter for the substance of our analysis. We disagree on the critical issue concerning selection of covariates. As a guide to future research, we highlight the nature of the disagreement, and explain why we believe covariates should be motivated by theory, and why an a theoretical approach to selecting covariates can result in failure to identify effects that actually exist. An important lesson from the analysis is that researchers should exercise caution when including ad-hoc covariates in empirical specifications. We offer concluding thoughts about empirical research and causal inference.