Kaufmann, Gimpel, and Hoffman: A promise fulfilled

Kaufmann, Gimpel, and Hoffman. 2003. A promise fulfilled? Open primaries and representation. JOP 65:457-476.

RESEARCH QUESTION: Do open primaries produce more moderate (i.e. representative of the general population) candidates than closed primaries?

SAMPLE AND DATA: Uses exit poll from 113 presidential primaries from 1988 through 2000. Compares each party's primary voters to the same party's general election supporters to check for ideological and demographic representativeness (i.e. how well do primary voters represent those who will end up supporting the party in the general election?).

FINDINGS (see TABLE 2): Though the results are mixed, open primaries do (1) attract more demographically representative voters; (2) attract more ideological centrist voters; and (3) produce more moderate nominees than closed primaries.

Strangely, however, moderating one primary can radicalize the other. This can happen in two ways.

1. If one party's primary is open but the other's is closed, then the open party's primary will attract weak partisans and crossover voters--but the closed party's primary will attract only strong partisans (i.e. actual party members). Thus, the open primary produces a nominee better able to win in the general election, and the closed primary's nominee ends up being more radical than if both parties had closed primaries.
2. Even if both parties have open primaries, candidates matter. For example, in 2000 McCain was a moderate with strong crossover appeal. Therefore, many weak Democrats chose to participate in the Republican primaries. Thus, the pool selecting a Democratic nominee was more liberal than it might have been if McCain hadn't been running.

OTHER FINDINGS:

• Some of their "major" findings are: Modified open primaries are not markedly different, open primaries mean younger electorates, open Democratic primaries tend to be wealthier, primaries in the South are more conservative, early competitive Democratic primaries are more liberal, Republican crossovers are wealthy, and modified open primaries do not significantly alter the demographic composition.

CONCERNS:

• Some of the variable operationalizations are weird (especially measurements of age and income). See p 463. (of course, they may have been forced into this due to limitations in the exit poll data).
• Why measure ideology on a 3 point scale (footnote, Table 2)? (Possible that that's all they had from the exit polls).
• Strange logic on the "ideological representation" bit. Wants to compare the ideology of party voter's to the party's general membership, but measures ideology of party's general membership by looking at anyone who voted for the Democratic presidential candidate in the general election. But this is endogenous. If primary voters pick a more moderate candidate, then more people will vote for the Democratic candidate.