Here’s a thought experiment. You’re a Coakley supporter in MA, and you feel like neither the candidate nor your fellow Democrats are taking the election seriously because everyone believes she’ll win by a large margin. A pollster calls you and asks you for whom you are planning on voting. What should you say? You could tell the truth and say Coakley, and risk yet another poll that shows her with a big lead and risks lulling Democrats even further into complacency. Or you could say Brown, in the hope that a closer than expected poll would light a fire under your co-partisans and/or convince some people who are planning on voting for Brown to “make a statement” because they think he has no choice of winning to avoid doing so.
This got me thinking: do we, as political scientists, really have a good handle on why and when people tend to tell the truth to pollsters? Much has been made of building up and tearing down the so-called Bradley effect, whereby whites voters might be less likely to admit to voting against black candidates. We also know from Adam Berinsky’s work in particular that sometimes certain voters are unable to articulate positions that they probably hold (see for example here or here). But what about situations like the one above, where it would serve the partisan interest of the respondent to lie about her intended vote choice? Or, to give another example, Adam Meirowitz and I have a formalized argument about why certain poll respondents ought to exaggerate their positions on policy positions (see the section entitled Intuition “here”:http://homepages.nyu.edu/~jat7/Meirowitz_Tucker_2007.pdf; for more generalizable formal work on this topic by Meirowitz see here.)
Yet at the end of the day, we usually do seem to be able to predict most election results from the final set of polls before that election, which suggests people aren’t systematically lying in election polls. I’m not sure if we have any similar findings in terms of policy questions, but I certainly have never seen systematic evidence suggesting they do. So this seems to be an interesting puzzle: if people ought to lie to pollsters to advance their own political agenda, why don’t they? I posed this question to my colleague Neal Beck, and his guess was essentially that people who answer surveys want to get off the phone as quickly as possible, and putting any thought into answering a survey question would slow down that process. I’m curious as to what other people think. Why don’t people lie more often to pollsters? In my piece with Meirowitz, we raise the idea of a “small psychic cost” to lying, but why should that kind of psychic cost be any more troubling than, for example, not voting for your preferred candidate in an election for strategic reasons? Could it be that people don’t think politicians look at poll results? That seems overly naive. We could of course fall back on the “what would one different response matter?” argument, but we know that people vote strategically; certainly one strategic poll respondent out of 1000 can have a larger impact than one strategic voter out of a million.
Anyone aware of empirical research on these or related topics? Perhaps from the political pysch literature?
{ 10 comments }
I don’t talk to pollsters but I would hesitate to lie because I think people are swayed by polls, that is, they tend to vote for a winner (independents) or not vote if their candidate is winning (pro-Coakley) or go out to vote if their candidate is losing (anti-C).
Mary: Unless I misunderstood, you’ve just given a great reason to lie to pollsters. If, as you say, “people are swayed by polls,” and people are less likely to turn out if they think their candidate will win, then shouldn’t you always lie to pollsters and tell them that you plan to vote for the OTHER candidate in hopes of reducing that candidate’s turnout?
I’m a PhD candidate in PoliSci at Boston University, and I can’t tell you how many times I’ve wondered this as I’ve read the literature on polling. I actually spent a few months at one point looking for research on this, but I came up with absolutely nothing. My guess is that it happens often enough to matter, but that most elections are so close (i.e. within or close to the margin of error) that we don’t notice.
More importantly, I think its related to the idea of self-selection in polling response. Generally speaking, most people probably do want to get off the phone quickly, but my hunch is that this is less true with the people who agree to answer questions than with those that do not. In fact, almost by definition it would have to be, right?
Finally, giving the recent explosion in daily tracking polls and pollsters, and given the increase in polarization among those most interested and engaged in politics (a la Pryor’s Post-Proadcast Democracy) it seems to me that this sort of behavior is much more likely to take place now than even a few years ago.
I’ve even given thought to doing some work on this myself, but its so far outside the focus of my research that I haven’t been able to figure out a way to make it work. But yes, someone really should look into this!
Two possible answers:
(1) The marginal effect of a single poll response is essentially zero, since it would appear in decimal places not even reported. Hence, if there is the slightest cognitive cost to lying, you won’t do it.
(2) 99 percent of poll respondents don’t think that hard about politics to consider the strategic implications of their response. The remaining 1 percent understand (1).
If I get polled, I’m prepared to stop and think, for each question, about the best strategic answer to give to support my party and my candidates. But I’m also in the top .1% or whatever for political awareness. I doubt that very many people are prepared to think strategically in such situations (which typically show up without any warning or context). Remember that even if one intends to give a strategic answer, it’s not always clear what that answer might be (for liberals: Brown in order to scare Coakley supporters into action? Or Coakley, to discourage Brown supporters and stop the bandwagon?).
I’d expect a little more strategic answering on basic presidential approval questions (for Dems, saying one approves of Obama even if one doesn’t), but it would probably be difficult to separate from other effects driving partisans to give partisan answers.
@KenS: So then why should anyone ever vote strategically? The marginal effects are much, much smaller, yet we are pretty confident they do, right? To me, that’s the puzzle. Shouldn’t the “psychic cost” to lying be equivalent to “psychic cost” to betraying the responsibility the founding fathers gave you to cast your vote for the best candidate in the election?
I think Jonathan’s comments help make the response. There are two things that are different about a poll response when compared to a voting decision:
(1) Theories of poll response (e.g., Zaller) assume that people are merely grabbing the most accessible answer, which in this case is the truth. By contrast, conditional on someone having made it to a voting booth, they have probably given the matter a modicum of thought.
(2) The strategic logic here is less straightforward. Voting for the second best in order to prevent the third best from winning is not too hard (even so, there is much less strategic voting than one would expect). But in the polling case, the respondent would have to make a bet about how the slight change he/she could induce in the reported margins would influence supporters on each side. It is not always clear what the right choice is.
Put another way: in the state of the world in which my one vote is pivotal, I can predict with certainty what the outcome of my vote will be. In the state of the world in which my poll response nudges the margins through rounding, it is hard to predict what the net effect will be in terms of mobilizing or demoralizing voters on each side. The calculation is way too hard, even if poll respondents were inclined to think about it, which they are not.
I was being facetious, trying to show that you can’t win by lying to pollsters because you send the wrong message to someone anyway.
I suppose you could do your side some good if all your buddies were doing the same thing and the other guys weren’t.
But what if everyone lies?
Maybe that’s what happens since polls do a pretty good job predicting outcome.
When Ways of Life Collide by Sniderman and Hagerdoorn is an interesting book about the origins of anti-immigrant politics in Holland.
They do some interesting things to get around the tendency of people to give socially acceptable answers. For instance, in one of the surveys they took, they had the interviewer attempt to imply that he/she held a particular view.
KenS’s answer is awesome. I’d like to offer another answer:
People like their team, i.e. the team they root for, to win races, sports’ matches, any competition whatsoever. Polls are a form of competition. We all know that polls have little real-world implications, just like friendship sports’ matches, or little-league games, but we root for all of these.
Thus, when one is answering a poll, they want “their” candidate to win. This motivates them to answer according to their real preference.
A secondary point is that a single person actually has *more* weight in a poll than in the actual elections since the poll is conducted on a relatively-small sample. So, the contribution of a single respondents’ to his candidate’s victory is actually *much larger* in a poll than in the real election. This gives even more importance to give the true answer.
Comments on this entry are closed.