Electoral Composition and the Midterm Loss Cycle

Abstract

The president’s party has lost congressional seats in all but three midterm elections since the Civil War, a phenomenon known as the “midterm loss cycle.” This paper uses individual-level administrative voter files covering the vast majority of the U.S. electorate from 2008-2018 to study how compositional changes in the electorate influence the midterm loss cycle. My approach overcomes serious data limitations in prior research, which uses unreliable voting surveys or aggregate election results. Three main findings emerge. First, I show that the midterm electorate is substantially older, less sociodemographically diverse, and more educated than the presidential electorate. This runs contrary to results from survey data. Second, I identify a persistent “alignment gap,” in which voters are more likely to drop off in midterm elections when they share the president’s party than when they belong to the opposite party. I provide suggestive evidence that this alignment gap results from irregular, peripheral voters of the president’s party being differentially mobilized in the presidential election and then abstaining in the following midterm. Third, I estimate that the alignment gap cost the president’s party 14, 7, and 10 seats in the 2010, 2014, and 2018 midterm elections, respectively, or about one-third of the total seats lost. These results imply the necessity of accounting for compositional changes when studying the midterm loss cycle and illuminate substantial differences in the determinants of voting between presidential and midterm elections.