Tracing Polarization's Roots: A Panel Study of Voter Choice in Congressional Primary and General Elections

Competing theories alternatively blame Congressional polarization on interest group control over nominations, extreme and highly-knowledgeable primary voters, or uninformed general electorates. We argue that these actors all prefer policy-aligned candidates, but that *primary* voters face unique challenges electing them because of the absence of distinguishing party cues. We report an original four-wave panel survey (N=31,254) spanning primary and general elections in 27 congressional districts in 2024, supplemented with candidate position and endorsement information. The findings support our argument. Both primary and general election voters are ~14 percentage points more (less) likely to vote for a candidate after learning they (dis)agree with them on an issue. However, primary voters know less than general election voters; difficulty distinguishing candidates from partisan reputations is a likely mechanism. Consistent with group influence, learning endorsements meaningfully affects voter choice. Our findings shed new light on why Congressional candidates fail to converge to the median voter.