This Week in Political Science: Divisive Primary Edition

The “Niggerhead” incident and related mudslinging among the GOP presidential candidates prompts the question: how competitive is too competitive when it comes to primaries? Political scientists have long studied what is called the divisive primary hypothesis, and John wrote a post back in early 2008 (see also the comments) that presented some relevant research. John suggested that divisive primaries had little effect on the nominee’s performance in the general election:

The logic is this: a divisive primary is more likely to arise if an incumbent is unpopular or presiding over a weak economy, simply because this incumbent will attract more challengers. Leaving the economy and presidential popularity out of the equation risks overestimating the effects of divisiveness.

Paul-Henri Gurian, author of this working paper, disagreed and responded in the comments:
We find that a divided party will lose up to 5% nationally in the general election, as well as losing up to 2% in individual states that had divisive state primaries.

See also this blog post by Josh Putnam for more.

The 2008 election afforded a fresh opportunity to examine this question, especially because of the lengthy battle between Barack Obama and Hillary Clinton.  Several studies of 2008 suggest that the Democratic primary contest had little impact on Obama’s general election performance.

Using a survey that tracked individual voters from the primary to the general election, Michael Henderson, D. Sunshine Hillygus, and Trevor Thompson (gated, ungated) examine whether and why Clinton supporters did or did not support Obama in the general election.  They find that 71% of Clinton supporters ended up voting for Obama.  Moreover, supporters of Clinton and the other Democratic candidates were no more likely to stay home on Election Day.  The most important factors that predicted a vote for McCain among supporters of the other Democratic candidates were not frustration with the primary election’s outcome but ideology and political issues, especially the Iraq War.  The authors conclude:

Our analysis offers individual-level results that call into question the longstanding assumption that thwarted voters will necessarily stay home or defect to the opposing-party candidate because of hard feelings from a divisive nomination phase.

In a 2010 paper, Amber Wichowsky and Sarah E. Niebler (gated) argue that then Senator Obama was actually helped by the primary’s competitiveness. Wichowsky and Niebler seek to disentangle competitiveness  (the closeness of the election) from divisiveness (the negativity of the campaign) and then estimate their respective effects on Obama’s general election performance.  The authors find that more competitive primaries featured more political ads, but a smaller proportion of negative ads. Competitiveness did not breed divisiveness.  Moreover, the more competitive the state’s primary or caucus, the better Obama did in that state in the general election:
After controlling for Democratic performance in the previous presidential election, a one percentage point increase in the relative competitiveness of the Democratic contest over the Republican contest leads to three tenths of a percentage point increase in Democratic vote share, a relatively small effect, but opposite that suggested by earlier research.

Wichowsky and Niebler explain why the 2008 Democratic primary battle may have helped Obama:
For one, Obama was an unknown candidate with an issue position on Iraq that was popular among Democrats and many Independents. Second, he was able to flex his substantial organizational and financial muscle throughout the primary season. Indeed, the protracted nomination may have helped Obama increase his donor and volunteer base…Moreover, absent geographic, demographic, and ideological factions, such as those characterizing the contests of 1968, 1976, and 1980, a more unified Democratic party may have been in a better position to reap the benefits of a vigorously contested nomination.

In general, political commentators still exhibit more concern about the effect of divisive primaries than the evidence warrants.

3 Responses to This Week in Political Science: Divisive Primary Edition

  1. Paul g October 5, 2011 at 3:47 pm #

    I think one of the major contributions of this site is to help journalists understand what political science can–and can not–say about current politics. Here, I think you may not be as careful about the can not portion.
    Perhaps we read the scholarship differently, but my recollection is that divisive primaries may be a concern not primarily because they frustrate voters who supported the losing candidate, but because they expose weaknesses for the nominee, such as highlighting vulnerable issues, showing soft areas of support (groups, regions), or raising questions about viability.
    None of these, as far as I can tell, are tested above. Henderson et al. focus almost exclusively on the “sour grapes” hypothesis and find a 1.1% predicted drop in Obama’s vote as a result. The impact is small but statistically significant. Wichowsky and Niebler seems more applicable, but does not try to tell us what Obama’s vote share would have been in the counterfactual condition of a non divisive primary. They show (similar to Henderson et al in fact) that the potential negative impact of a divisive (not the same thing as competitive by the way!) primary can be ameliorated by party unity.

    • John Sides October 5, 2011 at 4:58 pm #

      Paul: The conventional wisdom about a divisive primary, as I hear it expressed, is that it costs a candidate votes in the general election. What you discussing — exposing weaknesses in the nominee etc. — are mechanisms by which votes are lost. What I think the literature has not done is show conclusively that votes are lost.

      I think the Wichowsky and Niebler paper is interesting in this regard. Yes, we can’t observe the counterfactual, but we can leverage the variation across states: that variation suggests that opposite of what conventional wisdom suggests. And they do distinguish competitiveness from divisiveness, so they are certainly aware of the difference.

      Your last sentence actually suggests why the conventional wisdom is likely wrong: it overestimates the party disunity created by a divisive primary.

  2. Phill Snow October 6, 2011 at 1:27 pm #

    I wonder, though, how this might differ when the primary contest is between representatives of entirely different factions of the the same party. While you could argue that Obama was a relative outsider and Clinton was an insider, the distinction between them in terms of policy issues and priorities was fairly close. Both candidates attracted the same kind of voter, so the Clinton voter was not nearly as thwarted as, say, a Bachman or Perry voter would be by the nomination of Romney.