In 2008, Democrats enjoyed a solid advantage in partisan identification. By 2010, that advantage had largely evaporated. As I detailed in a previous post, in every state, the Democratic partisan ID advantage has declined, and by an average of nine percentage points.
But the decline has not been equal across the nation. In fact, there is a good deal of variation in the change in Democratic identification across states, ranging from a ranging from a drop of 22.2 percent in New Hampshire (from +13.2% to -9.0%) to a drop of just 1.6 percent in Mississippi (see this table for state-by-state numbers).
Why should these changes vary so much from state to state? Are there demographics that might explain this?
As it turns out, the only statistically significant predictor of the decline in democratic partisan affiliation advantage is the percentage of white people in the state. Surprisingly, the state economy (at least as measured by unemployment rate or change in unemployment rate) doesn’t seem to matter.
Unemployment
Let’s begin with the unemployment rate, since a good deal of the analysis around the 2010 election was an “it’s the economy stupid” story: voters blamed Democrats for high unemployment, and voted Republican to express their anger and frustration.
Yet, what’s remarkable about this scatterplot (above) is that the story doesn’t hold up. If anything, the relationship seems to be slightly opposite what the conventional wisdom would lead us to expect: the Democrats appear to have lost more support in states that have relatively lower unemployment rates. However, it is not statistically significant.
Still, it’s possible that what matters is not the absolute unemployment rate, but rather the change. Yet, once again, the scatterplot (below) shows that this is not the case. The more unemployment dropped between November 2008 and November 2010, the less the average decline in Dems’ partisan ID advantage. Though the relationship is actually stronger than above, it is still not a statistically significant one.
These numbers just don’t fit with the story of voters turning against Democrats for a failing economy. Take Nevada: Unemployment jumped from 8.0 percent to 14.3 percent, yet Democrats partisan ID declined by only; Similarly, California: Unemployment goes up from 8.4 percent to 12.4 percent.
On the other side, consider New Hampshire: Unemployment goes up from 4.3 percent to 5.4 percent (both among the lowest in the nation), but Democrats lose 22.2 percentage points in partisan ID advantage; Or South Dakota: Enemployment up from just 3.4 percent to just 4.5 percent, but the Dem partisan ID advantage falls up 10.4 percent.
Manufacturing
Another possibility is that what matters is the economic make-up of the economy, and in particular, perhaps states that rely disproportionately on manufacturing are more likely to have a lot of anxious voters, since manufacturing is a dying industry. But if we plot the decline in Democratic partisan ID and the manufacturing as share of the state GDP, there is no relationship.
Seniors
Another possibility is that Democrats are losing out in states with more seniors, since senior citizens are reportedly turning against Democrats. A scatter-plot shows a clear relationship, though not quite a statistically significant one (but close!). Generally, the more seniors in a state, the more Democrats have lost in their partisan ID advantage. However, the number of seniors explains only three percent of the variation in the Democratic vote share decline.
Whites
Finally, we come to the share of white voters. Here we have a consistent pattern, and one that is statistically significant (and explains 13 percent of the state-level variation). For every ten percent increase in white voters as a share of the electorate, the predicted decline in Democratic ID advantage is almost one full percentage point (the one outlier in the lower left is Hawaii, which is highly Asian. Without that outlier, the relationship would be even stronger).
This re-emphasizes the problems that Democrats seem to be having with white voters. (Democrats have not enjoyed parity with Republicans among white voters in 20 years, but 2010 was especially bad, with white voters breaking 62-to-38 for Republicans in the mid-term elections.)
This explains why the Democratic decline in diverse states like California (47 percent white) and Nevada (66 percent white) is less than in lily-white states like South Dakota (90 percent white) and New Hampshire (95 percent white), even though California and Nevada have much higher levels of unemployment.
These results exist regardless of economic circumstances (these findings are robust even in a statistical model that controls for all the other possible factors discussed).
Conclusions
The brief summary of this analysis is that race may matter more than the economy for why voters have been identifying more and more as Republicans for the last two years.
Of course, there are obvious caveats to this interpretation, most significantly the fact that I am playing around with state-level data, as opposed to individual-level data.
But the patterns are discouraging for Obama and the Democrats. Much prognostication has argued that the number one factor for 2012 will be the unemployment rate, because historically, the unemployment rate has been a very strong predictor of whether the incumbent party wins or not. This analysis suggests that something else is going on as well. Democrats are having a hard time with seniors and particularly white voters, and it’s not just a story about the state of the economy. Democrats ignore these scatterplots at their peril.
Update: I’ve written a response to some of the comments entitled “Am I a Race-Baiter?”
