Consistent Biases in Electoral Environments: Evidence from Entry and Exit of Senators

Abstract

In this paper, we compare senators first elected in midterms with those first elected in presidential elections and find them strikingly different: The cohort of senators first elected in presidential elections is consistently more ideologically extreme and party disciplined than the cohort first elected in midterms. This result is surprising in light of empirical evidence suggesting that the electorate in presidential elections is more ideologically moderate and less partisan than the electorate in midterm elections. Furthermore, we find that senators who are ousted or retire from office during the time period around presidential elections are significantly more ideologically moderate and vote more independently than those who exit around midterms. Together, these two empirical regularities suggest that the relatively more moderate electorate in presidential elections generates a more extreme and polarized Senate. These findings suggest that holding concurrent races for office is not outcome neutral and raise policy questions about the timing of elections and ballot initiatives. Our empirical approach is robust to econometric specification and outliers and can be extended to examining models of electoral competition and voting behavior.