What should we make of evidence that people give more factually accurate answers on political surveys when money is on the line?

Co-blogger Orin Kerr links to an excellent recent study by political scientist John Bullock and several coauthors, which finds that survey respondents give more correct factual answers on questions about political issues when they are paid to do so. The authors also find that partisan gaps in factual perceptions diminish greatly when money is on the line. Their results are consistent with a number of previous studies, which show that that survey respondent give more accurate answers to factual questions about political issues when offered incentives to do so.

Such results might lead us to be more optimistic about the problem of political ignorance. Perhaps the voters know more than traditional surveys suggest. But any such optimism must be tempered by the realization that voters do not get paid for accurate factual assessments when they go to the polls.

As a general rule, partisans are highly biased in their evaluation of facts and evidence, in part because they have little incentive to try to keep their biases under control. The chance that any one voter's ignorance or bias will change an electoral outcome is infinitesmally small. Financial incentives for correct answers change the payoff structure, and therefore lead survey respondents to consider what they know more carefully, and try to keep their biases in check. When money is on the line, we reason more carefully than when the only thing at stake is a potential error in casting a vote that is highly unlikely to change anything. Interestingly, things are very different in the rare case when a voter knows her choice is going to be decisive.

Although they acknowledge the possibility that "incentives may reduce partisan gaps by causing
respondents to think more carefully about correct answers," the authors instead emphasize the more optimistic possibility that many people are simply "insincere" when they give inaccurate, biased answers on conventional surveys. They know the truth, but deliberately lie in order to "cheerlead" for their favorite political "team."

But such deception makes little sense. There is little point to lying about factual issues on an anonymous survey in order to support your party. Indeed, doing so might well harm your party's interests rather than benefit it. The predictable effect of giving inaccurate, biased answers to survey questions is not to help your party, but to make you and its other supporters seem ignorant and stupid. Consider the reaction to the recent survey about bombing the fictional nation of Agrabah, and how it made Republicans look bad. People smart and sophisticated enough to deliberately lie for the purpose of supporting their party are unlikely to ignore this fairly obvious point. After all, if you know the correct answer, you will likely recognize that the people who sponsored the survey probably do so as well. You are unlikely to fool them. Thus, it is far more likely that financial incentives actually increase knowledge (e.g.—by leading respondents to recall information they had forgotten) and diminish bias, than that they eliminate "insincere" responses intended to "cheerlead" for a party.

The authors suggest that the "insincerity" of many respondents is shown by the fact that a higher percentage admit ignorance on surveys where they are paid more for a "don't know" response than for a wrong one (but less than for getting the correct answer). But this could easily be the result of thinking harder about what they know and don't know (and with what degree of certainty) when money is on the line. Moreover, some risk-averse respondents might prefer a certain (though smaller) payoff for picking the "don't know" option, than a larger, but less certain, one for picking what they think is the correct answer, but still have some doubts about.

All of this suggests that we can reduce political ignorance by giving voters financial incentives to increase their knowledge. Economist Bryan Caplan has proposed a "Voter Achievement Test" that would do just that, and I tentatively outlined a similar proposal in my book Democracy and Political Ignorance. Caplan's proposal and others like it deserve serious consideration. But I have stopped short of actually endorsing any such plan, because I am skeptical that any real-world government can be trusted to design and administer such a test fairly.

The effectiveness of incentives in increasing knowledge and reducing bias also suggests that we should consider making more decisions by "voting with our feet" and fewer at the ballot box. When we decide what products to buy in the market, or choose which jurisdiction to live in a federal system, we usually try much harder to to acquire relevant information and evaluate it wisely than when we vote in any election. It is no accident that most people devote a lot more time and effort to figuring out which car or TV to buy, than to deciding who to vote for in a presidential election. We know that the former decision will make a real difference, while the latter has only a tiny chance of doing so.