Q&A on Minority Turnout Model

There were several good questions on today's analysis of Obama's minority turnout scenarios, so let me extract a couple of those here.

Q. I am curious why you use census estimates of turnout (which are self-reports) instead of exit poll estimates of turnout (which vary substantially across states). Might not matter much.

We played around with both, and we made a decision to use the Census Bureau numbers as those results have somewhat larger sample sizes in all but a couple of states. The self-reporting might be a bit of an issue, but then again, we've seen plenty of problems in the past couple of cycles with the way that exit polls determine their samples as well.

Q. The main problems I see with your analysis that should be considered is (1) white/Anglo counter-mobilization; and (2) diminishing returns. While the Republicans are unpopular now, the election will not occur in a vacuum, and there likely will be some countermobilization among whites and Anglos (everyone is energized now, and you are assuming the black + latino voters increase but not that white/Anglo turnout is higher). I could see this mattering in a state like Texas, which has a history of white countermobilization to increasing black registration.

I've actually seen some academic work that the presence of a black candidate on the ticket tends to increase black turnout -- but also tends to increase white turnout. So you may be on to something there. At the same time, the white conservative vote was pretty darn mobilized in 2004, and I'm not sure I see the same thing happening this year at a time when the incumbent, Republican president is extremely unpopular, and when John McCain is not particularly well liked by elements of that conservative base.

I still think the most instructive piece of evidence are the turnout figures from the Democratic primaries. Turnout among these groups has not just increased in absolute terms -- everyone's turnout has increased in absolute terms. It has also increased in relative terms; slightly so for black voters, and very substantially so for Latinos and young voters.

Q. Are there any newer estimates of population by race & ethnicity (w/ a voter screen) than 2004 CPS that would be helpful to estimate 08 turnout rates?

Yes, the Census Bureau puts out estimates of the racial composition of each state every year, but the differences only amount to a percentage point here and there. I think you guys may be overestimating the level of precision that I'm aiming for with this analysis. Unlike some of the other stuff I do here, this is not necessarily meant to be predictive. I'm not necessarily saying there will be big increases in turnout among these groups. Instead, it's meant to be illustrative of how the map would change if Obama did get higher turnout from these groups. The whole thing is kind of in the conditional tense.

Q. What accounts for black turnout decreasing in NY, MS, FL, CA, & OK between '04 and '08? Is it just population change?

It might just be random noise. However, in three of these states (NY, OK, FL), Obama made only a half-hearted effort to compete. In California, there was a huge surge in Latino turnout that swamped everything else. In Mississippi -- I'm not sure. Obama actually spent very little time on the ground there -- just a day of campaigning, I think -- and there was a pretty large Republican crossover vote.

Q. Do these estimates account for cases in which the young voter might also be African-American or Latino? In other words is there some double counting going on?

The double counting thing is definitely something I was aware of, but I'm not sure that there's an elegant solution. Certainly with the Latino vote, for instance, there is a fairly big generational divide. If Obama turns out younger Latinos, that vote might go for him 70:30 or 75:25, whereas for older Latinos, the vote will probably be closer to 50:50. Nevertheless, some sort of adjustment is probably warranted. But I stated above, this analysis is intended more to be descriptive than predictive.

Q: Poblano - your analysis is getting to be way way too biased and losing its usefulness - you have to realize that it is very difficult for a democrat to get more than 50% of the vote, it has not happened since 1976.

Sure, but look how the electorate has changed since then. In 1976, just 2 percent of the electorate was Latino. That number was 8 percent in 2004 and will probably be at least 10 percent in 2008. Just 27 percent were college graduates in 1976, versus 42 percent in 2004. There are some trends that tend to favor the Republicans too, like the population getting older.

Q. If this could be real, it has huge implications for Obama's strategy. Should he tack right or tack left for the general?

As I've written before, Obama has two fundamental ways he can win. First, he can attempt to unifty and turn out the Democratic base. If he does that, he'll win based on the Democrats' party identification advantage, unless he gets absolutely killed with independents. Secondly, he could write off certain of those Reagan/Clinton Democrats, and instead make more of an appeal to the sort of the libertarian-leaning investor class, which makes up a pretty high fraction of the independent voting pool. The former strategy -- consolidating Democrats -- is certainly less of an uphill climb. The more confident the Obama campaign is that it *can* achieve increases in turnout among traditionally Democratic-leaning groups, the more that looks like the right way to go.