The Monkey Cage

I think you and Gelman missed the single most important conclusion that can be drawn from the data on the election. Based on the exit polls and the Pew data that Gelman has analysed, Obama did the same or slightly worse than Kerry among white, non-Hispanic voters aged 45 and up (or as I would prefer it, born 1963 or earlier). This group represents over 40% of the electorate. If they didn’t participate in the 6 point increase the Dems had overall, that seems pretty significant to me.
On the other hand, all voters born since 1978 have consistently voted for Democrats (at both the Presidential and Congressional levels) at about a 2-1 margin.

I think the data shows that the difference in the outcome between 2004 and 2008 is explained more by demographic changes in the electorate than by the performance of the economy. I’m not saying the economy has no impact, but why would changes in economic performance have no impact on the behavior the over-45 crowd (except to make more non-whites come out to vote) and such a disproportionate impact on younger voters?

My quick answer is to start with an additive model. John McCain had particular appeal to elderly whites. Even so, he did about the same as George W. Bush among that group (assuming your numbers are correct). There was a national swing toward the Democrats, which was stronger among some groups and weaker among others, sure. Our story is that the national swing is explainable by the economy, and that the deviations are candidate-specific. For example, Hillary Clinton probably would’ve done better among elderly whites and less well among young minorities, compared to Obama. And Mitt Romney probably would’ve done better among high-income voters and less well among low-income voters, compared to McCain. The maps of who won which states would’ve changed a bit too. The national outcome would’ve been similar though, I expect.

One thing that really is unprecedented, though, at least in the polling era, is the big swing of the young toward the Dems in 2006 and 2008.

I checked the Census Bureau (CPS) data. They show non-Hispanic whites aged 45 and up making up 47% of the electorate in 2008. (You can find the data for non-Hispanic whites at http://www.census.gov/population/socdemo/voting/cps2008/Table%2002-3.xls ). 61.4 million out of 131 million.

My reading of the Pew data (that you’ve used for your other analyses) is that 2004 and 2008 D/R splits for this cohort of white non-Hispanics voters (who were 41 and up in 2004) were essentially identical. If that’s correct, all the change from 2004 to 2008 was concentrated in 53% of the electorate. I don’t see how an additive model can account for this.

I don’t see any polling data that suggests that McCain was substantially stronger in this group than other potential Republican candidates. As I understand it, you’re claiming that McCain was about 5% stronger than Bush was for nearly 50% of the electorate. Where’s the evidence of that? It certainly didn’t show up in polling in 2000 when they ran against each other. Also, the turnout rate for this group was down in 2008 when it was up for the rest of the electorate. Doesn’t that make it harder to argue that McCain was particularly strong in this group? In comparison, Obama clearly increased both the turnout rate and the Dem advantage for blacks. Considering their already substantial lead in that group, I think we can safely say that was a candidate-specific effect.

The only evidence we have from the 2006 mid-terms is exit polling, but they show this group voting Republican at pretty much the same rate. Exit polling in the 2004 and 2008 Presidential elections largely agrees with the Pew data in terms of the D/R split. You’ve already noted that Congressional elections are tracking more and more closely with Presidential elections. If the economic forecasting model for presidential elections was valid, I wouldn’t expect that to happen. Taken together, all this suggests to me that the economic forecasting model has a significant weakness. I can’t imagine any mechanism that would explain the available data without considering the very real demographic changes in the electorate. Given that all the economic forecasting models completely blew the 2000 election (Gore should have won the popular vote by a significant margin instead of what was essentially a tie), I’m inclined to look for other explanations.

I agree that the swing looks much more uniform if you look geographically than if you break things down by demographics. Obama did about 5 percentage points better than Kerry: roughly 10 percentage points better than Kerry among minorities and young voters, and roughly 0 percentage points better than Kerry among older whites. Presumably, the increased difference between whites and minorities is largely an Obama effect, but the age pattern was happening in 2006, so I wouldn’t attribute it entirely to Obama. The key is that elections are, by and large, determined by the national vote swing.

I agree that the swing looks much more uniform if you look geographically than if you break things down by demographics. Obama did about 5 percentage points better than Kerry: roughly 10 percentage points better than Kerry among minorities and young voters, and roughly 0 percentage points better than Kerry among older whites. Presumably, the increased difference between whites and minorities is largely an Obama effect, but the age pattern was happening in 2006, so I wouldn’t attribute it entirely to Obama. The key is that elections are, by and large, determined by the national vote swing.