Abstract:
Regression discontinuity (RD) is a popular method to estimate the effect of endogenous treatment. In a “fuzzy” RD design, the probability of treatment jumps discontinuously when a “running variable” (R) passes a threshold (R0). Fuzzy RD estimates are obtained via a two-stage least squares (2SLS) regression, where an indicator I(R > R0) plays the role of the instrument. Recently, Keane and Neal (2023, 2024) showed that 2SLS t-tests suffer from a “power asymmetry” problem: 2SLS standard errors are spuriously precise (imprecise) when the 2SLS estimate is close to (far from) the OLS estimate. One consequence is that the t-test has little power to detect true negative effects when the OLS bias is positive. This problem exists even if the instrument is very strong. Here we show that the same problem arises in fuzzy RD designs. A simple way to avoid this problem is to instead rely on the reduced form (or intent to treat) regression, which in this case is a sharp RD of the outcome on I(R > R0), to assess significance of the treatment effect.
 
About the presenter:
Daniel Kaliski is a Lecturer in Economics at Birkbeck, University of London. His research investigates the causes of socioeconomic disparities in health. He also examines how individuals respond to the health risks they face. He received his DPhil in Economics from the University of Oxford