Poll results are ubiquitous in political discussions. This morning, liberal columnist Frank Rich quoted a number I’ve heard a lot lately, “41% of Republicans believe President Obama is not a citizen.” Well, hold on a second. Forty-one percent told a pollster that they believe that. Why would they tell the truth? If I happened to believe (as most Republicans presumably do) that the president’s policies are a dire threat to the country, wouldn’t I be morally as well as practically compelled to give whichever answer would make him look as bad as possible? I might think that higher poll numbers for the non-citizenship claim would lend it credibility, which I would like. Actually, the correct strategy for Republican respondents isn’t clear here – the contention that Obama is not a citizen may be so absurd that it discredits his opposition. The general point remains.

Of course, another phenomenon that is probably more important is psychological rather than strategic: once people join a side, they are likely to believe anything that is good for their side and bad for the other. Just using no-stakes anonymous polls, this is largely indistinguishable from the strategic poll responses I talk about above. You could try to distinguish the two by putting stakes on “correct” answers, but most poll questions are too subjective for that. Even in the objective example of Obama’s citizenship, deniers would be unlikely to pay up their bets based on birth certificates or newspaper clippings. One thing that might actually help distinguish strategic lying from partisan beliefs is using human pollsters vs. automated systems. People might be embarrassed to lie to a human pollster. Further comments on this are encouraged.

A topic of much-discussed frustration in these poll-happy times is that polls are, to say the least, not a great way to determine policy. My point here is that polls may not even be a great way to find out what people think.
————————————————-
A note to non-technical readers about the title: “Incentive compatible” is a term used in the theory of auctions, matching and voting to mean that it is in each person’s best interest to tell the truth. So if you wander into an economics talk and hear someone say “the IC constraint is binding,” they are just saying that we have to worry about whether people will tell the truth.