Citation:
Gruenbaum, Ben, and Chris Celaya. “Can Close Elections Ever Be "As-Good-As-Random"? A Matching Technique For Imbalanced Regression Discontinuity Designs,” Working Paper. Copy at https://tinyurl.com/y8t5fqh2
Abstract:
We adapt and apply recently developed matching methodologies to improve covariate balance across treatment groups for the purpose of estimating the causal effect of incumbency in the
US House of Representatives. In doing so we assess a recent finding by Caughey and Sekhon (2011)
that Regression Discontinuity (RD) design is invalid for estimating this quantity
because of unobserved differences in the ability of parties or candidates to affect the
outcome of extremely close elections. We also discuss more generally how covariate matching and
RD designs can be combined in future applied research.