Sunday, 19 September 2004

Partisanship moves

The left half of the blogosphere is in a tizzy over suggestions that Gallup is “oversampling” Republicans—allegedly deliberately, apparently since these folks think Frank Gallup thinks it’s a smart idea to destroy his business to help a particular party win the election.

The “oversampling” could have two, rather more innocent, explanations:

  1. By random chance, Gallup may have gotten a sample that is more Republican than usual; the 95% margin of error for the poll given the sample size of 767 (for “likely voters”) is around ±3.5%—for “registered voters,” it’s around ±3.1%.
  2. Partisanship may have “moved” as a result of the campaign. While early empirical studies such as The American Voter posited that partisanship was causally prior to vote choice, more recent research suggests that citizens’ partisanship changes over the course of a political campaign—people who are inclined to vote for Bush tend to become more Republican, while people who are inclined to vote for Kerry tend to become more Democratic. Thus the incidence of partisanship in the electorate may have actually moved in a Republican direction.

I’d also suggest that the incidence of “independent” voters appears to be relatively inflated, and probably includes a large number of voters with fairly strong partisan leanings; it is socially desirable to self-identify as an “independent,” and thus the polls (not just Gallup—all of them) tend to show more independent voters than truly exist, as “true” independents make up less than 10% of the contemporary elected. The NES-style “branching” partisanship measure appears to conform more reliably to the actual incidence of partisanship and partisan behavior in the electorate.

1 comment:

Any views expressed in these comments are solely those of their authors; they do not reflect the views of the authors of Signifying Nothing, unless attributed to one of us.

This shows why it’s dangerous to get a political methodologist riled up… :)

Comments are now closed on this post.