Publication Type: Conference Paper/Unpublished Manuscript
Abstract: The "divisive primary" hypothesis is a now-elderly, dubiously-honorable, and one-of-the-least-well supported, albeit well-known, hypotheses in all of American political science. Many have assessed the hypothetical impact of the "divisive primaries" on general election outcomes, but, generally speaking, the analytic results have been weak to null.

We argue the case for an omitted "contestation" variable: One path to getting a handle on division and conflict is by including a measure of the ideological positioning of candidates in the primary contests. We hypothesize the following: The negative effect of a divisive primary on the general election "outcome" will manifest itself when the more ideologically extreme of the candidates is victorious in the primary contest. (1) In a primary contest, a contest ensues between two candidates and the more ideologically extreme candidate wins. In the spatial "presentation" of the general election contest that follows, moderate voters in the contested party may well split their ballots between their party's "extremist" nominee and defecting to the other party's, now more-adjacent candidate, and, thus, the opposing party wins. (2) The complement, of course, is that the moderate wins the nomination and the more extremist, albeit primary-losing, voters have little choice but to trudge to the polls and cast their ballots for their party's nominee. Thus, the perverse effects of a divisive primary occurs when the more extreme candidate is the party’s nominee.

We also point our analysis in two precise directions in measuring the "general election outcome." First, in a traditional "aggregate" fashion, we expect that contested primary elections under varied "ideological contestation scenarios" will affect the distribution of the candidates' vote support in percentages in the general election outcome - primary contests lower GE vote. Second, we also claim, that the impact of candidate ideological positioning should affect individual voters defection rates conditioned on the ideological position of voters. We expect to find, by examining exit polls for gubernatorial races from 1990 to 2006, that in these races with extreme candidates as the nominee, self-described "moderate" voters of the party represented by the extremist candidate will more likely defect to the opposing party. This, of course, should be most pronounced when only one party has this type of spatial configuration.

Our data set focuses on the outcomes of 210 gubernatorial primaries held in XX American states during the 1990 to 2006 period. We have over 450 survey responses from political scientists, journalists, and activists that we asked to “rate” the 1990-2006 gubernatorial candidates in their states. We also have collapsed the 1990-2006 Voter News Service exit polls for gubernatorial races over this time period with its question on gubernatorial general election choices. We are positioned to test this “spatial” understanding of the divisive primary hypothesis at the aggregate level (N=210 elections) and at the individual level (N=210 elections with 300,000 voters).