Red State Blue State in 2012

Avi Feller, Boris Shor, and I write:

The so-called “red/blue paradox” is that rich individuals are more likely to vote Republican but rich states are more likely to support the Democrats. Previous research argued that this seeming paradox could be explained by comparing rich and poor voters within each state – the difference in the Republican vote share between rich and poor voters was much larger in low-income, conservative, middle-American states like Mississippi than in high-income, liberal, coastal states like Connecticut. We use exit poll and other survey data to assess whether this was still the case for the 2012 Presidential election. Based on this preliminary analysis, we find that, while the red/ blue paradox is still strong, the explanation offered by Gelman et al. no longer appears to hold. We explore several empirical patterns from this election and suggest possible avenues for resolving the questions posed by the new data.

I love that—”the explanation offered by Gelman et al. no longer appears to hold.” As I said to Boris, if someone’s going to shoot us down, I’d like that someone to be us.

Click through to the paper—we have lots of graphs. Not too many answers, but lots of questions.

12 Responses to Red State Blue State in 2012

  1. Mark July 22, 2013 at 9:54 am #

    Is there data available to do a breakdown of voting patterns for those with family incomes exceeding $150,00 ($50o,000, $1 million, $10 million for instance)?

    • Andrew Gelman July 22, 2013 at 11:20 am #


      There are no conclusive data yet, but Page, Bartels, and Seawright are conducting a survey of rich Americans.

  2. Daniel July 22, 2013 at 10:26 am #

    This is consistent with the anecdotal stories about rich hedge fund and private equity managers in blue states turning against Obama after supporting him in 2008. After all, Romney is a blue-state rich guy who “everyone knows” to be only pretending to be socially conservative.

  3. Mark July 22, 2013 at 12:49 pm #

    Your statement that “much of the differences between the two parties can
    be identified as differences between rich liberals and rich conservatives” seems to be supported by your data and analysis. However, I have a problem with your nomenclature, particularly the use of the term “rich”. First, you refer in the post and article “rich individuals” but the data you have reflects family income. Second, and more importantly, it is odd to consider someone with a family income of $150,000 as “rich”. In my neck of the woods, husband and wife public school teachers can have that income.

    It also obscures the difference between income and wealth. For instance, there is a big difference between a married couple with new jobs and salary increases giving them an income of 200K, a 400K house with a mortgage, two children and very little in savings versus someone with an income of 200K derived from 10M in tax-free munis, who owns a 1M home without a mortgage and is single.

    To be clear, your data is very interesting; it is the use of a politically loaded term (and one that makes your results easily subject to misunderstanding) that I suggest you reconsider. It’s also why I would be very interested in more granular data as incomes escalate above 150K (though that would still not address the income v wealth aspect).

    • eric July 23, 2013 at 1:51 am #

      The median household income in the US is $50,000 with households $150,000 and up accounting for fewer than 7% of those households. Rich is always going to be a relative term. Even if teachers in your neck of the woods make that kind of money; nationwide, it’s an awful lot of money, closer–percentage-wise–to the guy with the 1M home than the vast majority of the country.

      Sure, in your neck of the woods it may not seem like much. In my neck of the woods it would be an enormous sum.

  4. Nameless July 22, 2013 at 5:07 pm #

    Given the estimated state-level vote vs. income slope, the difference between incomes between rich states and poor states corresponds to a 5-7% increase in the conservative vote in the rich group. From this point of view, the red/blue paradox is not really a paradox – if white voters in Massachusetts are uniformly more liberal than white voters in Missouri (so that Romney gets 25% more votes in MA than in MO in every income group), but voters in MA are relatively wealthier overall and that subtracts 7% from Romney’s total vote, Romney still does better in MO than in MA by a double-digit percentage amount.

    As to the reasons why voters in MA are uniformly more liberal than in MO, that’s a whole different question. (In addition, they are also uniformly less religious, more tolerant, etc. – there’s a full package of cultural differences.)

  5. David Karger July 23, 2013 at 3:01 am #

    You use two fixed thresholds of $30K and $150K, as well as graphs that plot as a function of absolute income. This seemingly ignores the fact that the same income can mean different things in different places. $150K in the rural south is probably quite well off; in NYC it’s poor. Have you looked at ways to normalize income to the environment of the recipient? E.g., looked at how voting outcomes differ among the upper quintiles in different states as opposed to some absolute income bound? Although it might be even more revealing to look at city-level data. Perhaps the reason blue states leaned Democrat is that even the relatively high-income earners in those states feel poor?

    • Nameless July 23, 2013 at 11:37 am #

      That’s a good point. In addition to the fact that $150K “feels” poorer in NYC than in Alabama, people themselves are different – the same amount of schooling that buys you $150K in NYC would buy you only, say, $80K in Alabama.

      And here’s an interesting observation. As the article says, if you plot per capita state income vs. Romney vote, there’s a negative correlation but it isn’t very strong. (Using data from Wikipedia to recreate the plot, I get R^2 in the neighborhood of 0.41 to 0.43 depending on the type of trend line.) If, instead of income, I use cost of living (first table I could google –, R^2 jumps to 0.56 for the full set, and as high as 0.65 if I exclude Alaska. (The table says that Alaska is more expensive overall than all states except Hawaii. I suspect that it underweights housing and overweights everything else.) If you look at the original table in the article, biggest outliers are, on one end, places like Wyoming (high income, low cost of living, too conservative for trend); on the other, Hawaii (similar income to Wyoming, significantly higher cost of living, too liberal for trend.)

      • eric July 24, 2013 at 1:07 am #

        Heck, why not just make an index of cost of living/median income and run a regression that might not be a bad way to account for these issues of relative wealth.

        Those breakdowns by category at might be interesting too.

        • Nameless July 25, 2013 at 12:38 am #

          Straightforward regression is problematic because relationships are nonlinear. For example, Romney vote steadily decreases with increasing cost of housing from TX to MA, but barely changes as the cost of housing continues to increase from MA to MD and NY. And do we want absolute cost of housing or the ratio of housing to income?
          As usual with nonlinear relationships, we lose some predictive power if our input variables are averages. (If Manhattan is significantly different from upstate NY and San Francisco is significantly different from Bakersfield, we won’t predict the vote in NY or CA correctly from weighted averages of income & cost of living in those.)
          In breakdowns by category, the strongest correlation is between Romney vote and the cost of transportation. But I suspect that, in this case, correlation runs at least partly backwards (higher liberal vote – higher gas taxes).
          Nevertheless, a pretty decent correlation can be achieved by regressing the vote against just two variables – per capita income and income-adjusted cost of housing.

      • Matt July 24, 2013 at 1:32 am #

        Things “cost more” because they’re “worth more”. $150 grand in New York is the same as $150 grand in Alabama, you just get paid a discount for living in Alabama.

  6. Mark July 23, 2013 at 2:05 pm #

    Would overlaying measurements of income inequality provide any additional insight?