I disbelieve what I call the Self-Interested Voter Hypothesis (SIVH for short). Political scientists like David Sears have amassed a mountain of empirical evidence against the SIVH, and the evidence on the other side (mostly economists, I’m afraid) is awfully weak.

When I say that, I don’t mean that that the SIVH occasionally fails to work. I mean that it virtually never works. Knowing someone’s objective self-interest tells you essentially nothing about his political beliefs. To take the easiest example, it is simply not true that Democrats are poor and Republicans are rich. The correlation between income and party is very close to zero. I’ve run lots of regressions myself on two unrelated data sets. It checks out.

However, Alex Tabarrok recently brought a little counter-evidence to my attention. According to the Statistical Modeling, Causal Inference, and Social Science blog, here is the individual-level income/vote breakdown from exit polls for the 2004 presidential election:

Income    %R    %D
$200K    62%   37%

Is it time to reverse my position? Not given all the other evidence I’ve seen. But it IS time to get a little less sure, and wonder about the source of the disparity. My main thoughts:

1. The vote share is very flat from $50K-$200K. And since perhaps 1% of families make over $200K, they wouldn’t have any discernable effect on the income-party correlation. So maybe the individual-level correlation from exit polls is not that far from what I’ve seen before (around .05).

2. The results I know well are for party identification, not vote. Could presidential vote be more selfish than your choice of party?

3. Presumably personality matters more in presidential races than in any other electoral contest. That adds a lot of noise, so perhaps it is not so surprising that the results are different than for party ID. Still, the implication is that Republicans making less than $50K were unusually likely to vote against their party. Is that credible?

4. The results I’ve seen are for the whole population, not just voters. Could voters fit the SIVH better than non-voters? I’ve looked at this issue before, however, and found no evidence of it.

Many people dislike empirical work because it never gives you a definite answer. That doesn’t stop you, however, from being nearly definite. I get less confident when a little counter-evidence comes in, but it would take a lot more to drastically change my mind. As one of my favorite sayings goes: “Keep an open mind – but not so open that your brain falls out of your head.”