Partisan Imbalance in Regression Discontinuity Studies Based on Electoral Thresholds

Abstract

Many articles use regression discontinuity designs (RDDs) that exploit the discontinuity in “close” election outcomes to identify various political and economic outcomes of interest. One of the most important types of diagnostic tests in an RDD is checking for balance in observable variables within the window on either side of the threshold. Finding an imbalance raises concerns that an unobservable variable may exist that affects whether a case ends up above or below the threshold and also directly affects the dependent variable of interest. This article shows that imbalance in RDDs exploiting close elections are likely to arise even in the absence of any type of strategic sorting. Imbalance may arise simply due to variation in the underlying distribution of partisanship in the electorate across constituencies. Using both simulated and actual election data, the study demonstrates that the imbalances driven by partisanship can be large in practice. It then shows that although this causes a bias for the most naive RDDs, the problem can be corrected with commonly used RDDs such as the inclusion of a local linear control function.