Primaries might also contribute to party transformation by incentivizing candidates to move position within an election cycle. Candidates might face a “strategic positioning dilemma” if they must first satisfy an extreme selectorate to earn the nomination before facing a comparatively moderate general electorate. This chapter therefore tests whether all candidates in a primary adapt their positions away from the center during the nomination phase of a single election cycle, presenting general election voters with polarized choices. To scale positions both during and after a primary it uses a text-as-data approach based on candidates’ communication on Twitter during the 2020 election cycle. It finds that Democratic candidates who lost primaries became significantly more moderate immediately after their defeat, especially if they lost in ideological or factional primaries. It does not observe this pattern among Republican losers. This chapter demonstrates a further way in which primaries may contribute to polarization, incentivizing candidates to adopt positions further from the ideological center during the nomination phase of the election cycle.

Having shown that primaries can reorient parties in Chapter 4, this chapter tests the first mechanism through which primaries are said to contribute to partisan polarization: the selective effect of voter preferences. It therefore tests whether primary voters prefer noncentrist candidates, all else being equal. Through a set of four analyses it tests whether primary voters prefer candidates further from the center when they are presented with a comparatively moderate and polarized alternative, whether moderate incumbents are more threatened, whether candidates who emerge from (ideological and factional) primaries are more “extreme” than other candidates, and whether there is any relationship between turnout and nominee position. Taken together, the findings in this chapter demonstrate the absence of a select effect from primary voters in nonincumbent primaries and only a weak and substantively small effect when an incumbent is present, suggesting that the polarizing effect identified in Chapter 4 is largely independent from the preferences expressed by primary voters.

Chapter 5 tests whether legislators are accurate in their belief that primary voters are likely to punish them for compromising. Results from a survey experiment suggest that most voters, even most primary voters, reward legislators for making compromises. However, co-partisan primary voters who oppose compromise on a specific issue are willing to punish legislators who vote for the compromise. Although legislators may benefit electorally from supporting compromise, especially in the general election, they have reason to be cautious on compromise bills to avoid voter backlash from subsets of the politically active primary electorate. Just because the subset of voters who punish legislators for compromising is small does not mean it cannot be consequential – a small subset can mobilize a strong challenger, paint a legislators’ behavior as problematic in the eyes of less informed voters, or vote on the basis of a single important vote. Moreover, across many compromise votes a legislator may face, the small groups of voters who oppose each compromise might, when added together, represent a decisive portion of the primary electorate.