Conor Clarke has kindly replied to my earlier rebuttal to his proposal to “get rid of polls”:
bq. A few days ago John Sides published a reasonably snarky blog post about a piece I wrote last week: The case against polling. The basic argument of my piece was that polls are annoying because (1) we should want our democratic institutions to operate according to pre-established mechanisms, not random quasi-referenda; (2) Lots of polls are wrong or misleading (duh); and (3) Present opinion polls can affect future opinion polls, due to information cascades. My feeling is that one’s opinions should change after an exchange of reasons, not after he or she acquires the knowledge that one opinion is more popular than another.
bq. Anyway, John gave me an exchange of reasons. Since he studies this stuff for a living, he’s certainly got the moral authority to get snarky. And I think he makes some good points, some of which I find convincing and some of which I need to think about a bit more. But what I think is missing from John’s post is the affirmative case for polls. (And I mean political polls — how many people support Obama’s health care plan and so forth — not social science research polls.) Weakening the case against polling strikes me as a necessary but insufficient defense. What good reason do we have (besides morbid curiosity) to consume polls we see in the morning’s paper? What value is there in letting the public know what the public already thinks?
I thought my post was not really _that_ snarky. Okay, maybe the opening paragraph. Nevertheless, Clarke poses a good question and I want to respond.
But first, a detour into some research I should have cited in my first post. Clarke raises the possibility of herding or bandwagoning in response to polls — a possibility that I poo-poo’d. In addressing this possibility, I should have turned to Diana Mutz’s Impersonal Influence. The subtitle of the book explains why: “How Perceptions of Mass Collectives Affect Political Attitudes.” She identifies various ways in which these perceptions might matter. Here is one:
bq. Strategic use of information about others’ views is obviously indicative of a highly informed and extremely thoughtful public.
This gets at what Hans Noel and Henry were pointing out: polls enable people to vote strategically. But here’s the kicker about “highly informed and extremely thoughtful” citizens, according to Mutz:
bq. These people are unfortunately not numerous.
In part, as I noted in the first post, this is because so few people pay diligent attention to polls — leaving aside the question of how many of those few people do so to behave strategically.
Mutz also finds some evidence of a “consensus heuristic,” in which people follow majority opinion as a rule-of-thumb. But there are serious limits to how frequently this heuristic applies:
bq. …the consensus heuristic requires extremely low levels of information in order to be applicable to political candidates and controversies … Few political situations involve … extremely low levels of information … and in this uncommon situation where information on a candidate or issue is low, the likelihood that such a citizen will be sufficiently involved politically to bother voting or even expressing an issue opinion seems slight.
This is perhaps her punchline:
bq. The bulk of impersonal influence from perceptions of mass opinion involves the integration of consensus information with other candidate or issue knowledge…people make use of consensus information in ways that are far removed from their sheeplike reputation.
In short, even if the vast majority of people paid attention to, well, the vast majority of people, it’s not clear that they would do so by, in Clarke’s phrase, “following the flock.” As such, I am not able to conjure up an example of large and purely poll-driven changes in attitudes.
End of detour. Now to Clarke’s main point: what is the positive case for polls? I would make the case on three grounds:
1) Accountability. I raised this in my original post. I don’t think elections are a sufficient mechanism to ensure accountability, in part because many if not most voters do not make policy a central factor in their choice. A more robust mechanism of accountability also needs to be ongoing: every 2 years or 4 years is not necessarily enough. What I mean by accountability is this: we should know whether policymakers and policy are acting in accord with public opinion (to the best of our ability to define opinion). This does _not_ mean that policy must always agree with the majority’s position. Accountability is not plebiscitary democracy, and fidelity to the majority is not the only or best criterion for evaluating policy. But surely public opinion provides a relevant criterion. This is where polls come in — above and beyond periodic elections.
2) Evidence. Public opinion is a currency of democratic politics — invoked by leaders, commentators, etc. My frequent frustration is that this rhetoric about “the American people” has nothing to back it up. In the mostly harmless cases, commentators cloak their own opinions with a popular garb. A couple examples are in this post. In the worst cases, rhetoric about public opinion becomes the linchpin of demagoguery. With polls, we can at least subject such claims to scrutiny.
For example, yesterday Ben Nelson expressed opposition to the proposed tax surcharge to pay for health care by invoking public opinion. In his view, voters think “tax is a four-letter word” and, apropos of the surcharge, “don’t necessarily think it’s fair.” Jonathan Chait responds by noting an ABC/Washington Post poll in which 60% of respondents supported “raising income taxes on Americans with household incomes over 250 thousand dollars to help pay for health care reform.” I’m not at all claiming that Ben Nelson is a demagogue. He may even be right that a tax surcharge is bad policy. But it’s not a policy definitively opposed by voters.
A skeptic like Clarke might rejoin: do we need so much “evidence” — that is, so many polls? It’s worth keeping this in mind: if the media organizations stopped polling, and Gallup et al. went under, we would still have polls. In response to Clarke’s original post, Ed Kilgore debunked “the real problem with poll-haters”:
bq. …the idea that suppressing or delegitimizing one form of information (and that’s all polls are, after all) will somehow create a data-free political realm in which “pure” or “real” or “principled” decisions are made.
In fact, the political realm would be full of polls conducted by politicians and interest groups and others, who would trumpet their findings much as they already do. However, there’s no guarantee that these polls would constitute reliable evidence, given the incentive that politicians have to trumpet only congenial findings. With more polls, we can actually better pinpoint what Americans think. As a social scientist, I would rather have more data points than fewer, although I realize that the proliferation of polls does facilitate the proliferation of crappy interpretations and hence the proliferation of crabby blog posts by me.
3) Representativeness. If we can agree that evidence about public opinion is useful, then there is the question of what evidence. There are lots of potential measures of public opinion: elections, polls, letters to the editor, letters to members of Congress, protests and other forms of collective action, etc. Letters and protests and their kin are important, particularly in how they convey intensity of opinion. But it’s also important to know what the public, rather than a self-selected group of activists, thinks. This is the real value of polls. Of course, no poll reaches a perfectly representative sample, but good polls do well enough.
Lest I sound like a cheerleader: some polls bug me, much as they must bug Clarke. I hate the vastly premature 2012 presidential trial heats, for example. But when I think of politics without polls, I think again of Howard Kurtz essentially fabricating an answer to “how things are playing out in the country.” And that really bugs me.