Many articles use regression discontinuity designs (RDDs) that exploit thediscontinuity in “close” election outcomes to identify various political andeconomic outcomes of interest. One of the most important types of diagnostictests in an RDD is checking for balance in observable variables within thewindow on either side of the threshold. Finding an imbalance raises concernsthat an unobservable variable may exist that affects whether a case ends upabove or below the threshold and also directly affects the dependentvariable of interest. This article shows that imbalance in RDDs exploitingclose elections are likely to arise even in the absence of any type ofstrategic sorting. Imbalance may arise simply due to variation in theunderlying distribution of partisanship in the electorate acrossconstituencies. Using both simulated and actual election data, the studydemonstrates that the imbalances driven by partisanship can be large inpractice. It then shows that although this causes a bias for the most naiveRDDs, the problem can be corrected with commonly used RDDs such as theinclusion of a local linear control function.