A notion you hear a lot these days is that the 2010 election will depend primarily on who votes, and especially on whether Obama voters, or maybe first-time Obama voters from 2008, actually show up. Dan Hopkins sent me this post from Tom Jensen. Jensen writes:
bq. There continues to be no doubt this fall’s election will have more to do with whether Democrats can turn out Obama voters than keep them in the fold.
This raises a question that political science can address: is variation in election outcomes due more to changing minds — such that some voters who preferred a candidate of one party in one election then prefer a candidate of the other party in the next election — or to changes the electorate, such that the election outcome changes because different groups of voters either stayed home or turned out.
bq. I estimate the effect of changing electorate composition on vote share using millions of individual records of turnout behavior, hundreds of thousands of ballot images, and tens of thousands of precinct election returns from the state of California in 2004 and 2006.
Here is what Hill found:
bq. Consistent with survey findings, I show that the partisan composition of the electorate between the two elections does not change much despite a large decline in turnout. I find that the effect of changes in who votes depends upon the nature of the contest…Of Republican Governor Arnold Schwarzenegger’s 10.7 point improvement in vote share in 2006 over Republican President George W. Bush in 2004, I estimate that 9.2 points are due to preference change and 1.5 points due to change in the group of citizens who vote. In contrast, change in U.S. Senate vote share across the two elections is about equally due to preference and composition change.
The 2004 and 2006 Senate races featured Democratic incumbents, Boxer and Feinstein, who each won a similar proportion of the vote.
Much more research is needed to determine the prevalence of preference vs. composition change across elections. I would suggest caution, however, in assuming that mobilizing voters (i.e., composition) is the key in 2010. Narratives about enthusiastic vs. dispirited parties don’t always survive scrutiny, as I argued with regard to the 1994 election. And survey evidence of voter “enthusiasm” is not a reliable predictor of turnout.
Here’s another point worth considering. In a forthcoming paper (pdf), Eric McGhee and I measured the partisan composition of voters using state-level exit polls from 1988-2006 and looked to see how it responded to presidential approval, economic performance, and the balance of campaigning in the state by presidential, Senate, house, and gubernatorial candidates (as relevant).
We find little evidence that presidential approval and economic performance affect the partisan composition of voters. The things you think might enliven or dispirit partisans — a popular or unpopular incumbent, for example — don’t really seem to matter. Campaigning, by contrast, does matter. The more a party dominates the other in spending, ads, etc., the more its partisans dominate that year’s electorate. (The paper dwells at some length on the possibility of reverse causation and finds little evidence of this.) But here’s the irony: the effects of campaigning are largest when one party spends much more than the other, and those races are precisely the ones where the dominant party has little chance of losing.
None of this research obviates the prospect that successful mobilization might change the electorate just enough to tip a very competitive race. It does, however, insert some uncertainty into the many confident predictions that 2010 depends mostly on who goes to the polls, and on whether the party’s respective campaigns can mobilize their faithful.