Abstract: The weak instrument literature focuses heavily on the problem of size inflation in 2SLS/GMM t-tests that arises if instruments are weak. A common standard for acceptable instrument strength is a first-stage F of 10, which renders this size inflation modest. However, Keane and Neal (2023, 2024) showed that IV estimators suffer from another problem even at much higher levels of instrument strength: standard errors tend to be artificially small in samples where the coefficient estimate is close to the OLS bias. This ‘power asymmetry’ means the t-test has terrible power to detect true negative effects when the OLS bias is positive. This presentation will quickly review these findings before considering two additional consequences for applied work: (i) how to do inference in fuzzy regression discontinuity designs, and (ii) how this feature of IV makes it particularly vulnerable to p-hacking through instrument selection.”

Contact the host of The Power Asymmetry of IV Estimators: Consequences for Applied Work

There were 0 errors found in the information you submitted: