TNR BLOGS

January 08, 2009 | 6:03 PM
January 08, 2009 | 5:59 PM
January 08, 2009 | 5:46 PM

January 07, 2009 | 12:20 PM
January 07, 2009 | 12:13 PM
January 07, 2009 | 9:41 AM

January 08, 2009 | 6:31 PM
January 08, 2009 | 4:13 PM
January 08, 2009 | 2:50 PM

July 26, 2008 | 2:24 PM
July 23, 2008 | 1:55 PM
July 17, 2008 | 3:56 PM

January 08, 2009 | 5:12 PM
January 08, 2009 | 3:25 PM
January 08, 2009 | 1:16 PM
COLUMNISTS
TODAY'S STORIES
29.05.2008
Our "Rubbish" Polling Industry

Yglesias Yglesias guest-blogger Alyssa Rosenberg flagged this really interesting London Review of Books essay. Here's my favorite excerpt:

Yet if the voting patterns have been so predictable, why have the polls been so volatile? One of the amazing things about the business of American politics is that its polling industry is so primitive. Each primary has been preceded by a few wildly varying polls, some picking up big movement for Clinton, some for Obama, each able to feed the narrative of a contest that could swing decisively at any moment. All of these polls come with warnings about their margins of error (usually +/–4 per cent), but often they have been so far outside their own margins as to make the phrase ridiculous. A day before the California primary in February, the Zogby organisation had Obama ahead by 6 per cent – he ended up losing by 9 per cent. In Ohio, the same firm put Obama ahead by 2 per cent just before the actual vote – this time he lost by 10 per cent. The sampling of national opinion is even worse. Before the Indiana primary, two national polls released at the same time claimed to track the fallout from the appearance of Obama’s former pastor Jeremiah Wright on the political stage. One, for the New York Times, had Obama up by 14 per cent, and enabled the Times to run a story saying that the candidate had been undamaged. The other, for USA Today, had Clinton up by 7 per cent, leading the paper to conclude that Obama was paying a heavy price.

The reason for the differences is not hard to find. American polling organisations tend to rely on relatively small samples (certainly judged by British standards) for their results, often somewhere between 500 and 700 likely voters, compared to the more usual 1000-2000-plus for British national polls. The recent New York Times poll that gave Obama a 12 per cent lead was based on interviews with just 283 people. For a country the size of the United States, this is the equivalent to stopping a few people at random in the street, or throwing darts at a board. Given that American political life is generally so cut-throat, you might think there was room for a polling organisation that sought a competitive advantage by using the sort of sample sizes that produce relatively accurate results. Why on earth does anyone pay for this rubbish?

The answer is that in an election like this one, the polls aren’t there to tell the real story; they are there to support the various different stories that the commentators want to tell. The market is not for the hard truth, because the hard truth this time round is that most people are voting with the predictability of prodded animals. What the news organisations and blogs and roving pundits want are polls that suggest the voters are thinking hard about this election, arguing about it, making up their minds, talking it through, because that’s what all the commentators like to think they are doing themselves. This endless raft of educated opinion needs to be kept afloat on some data indicating that it matters what informed people say about politics, because it helps the voters to decide which way to jump. If you keep the polling sample sizes small enough, you can create the impression of a public willing to be moved by what other people are saying. That’s why the comment industry pays for this rubbish.

Update: See commenter AlanSP below for some significant caveats about this take on polls. 

--Michael Crowley

Posted: Thursday, May 29, 2008 6:00 PM with 7 comment(s)

Comments

You must be logged-in to comment.

Not a subscriber? Click here to get a digital or print and digital subscription to The New Republic!

AlanSP said:

Several problems with this analysis:

First, a minor factual error.  The last Zogby Ohio poll showed Obama and Clinton tied at 44% each.  I guess the bigger problem with the examples is cherry picking a couple of polls that were significantly off (and from one of the worst polling firms this cycle) as indicative of polling in general.

More importantly, the article displays some major misunderstandings about the statistics involved here.  The size of the country is irrelevant to the statistical error, provided that there the size of the population is much bigger than the size of the sample (this is the case for any national poll in either the US or Britain).  Also, there are diminishing returns on the reduction in error from increasing the sample size.  A sample of 600 has a MoE of 4%.  Increasing the sample size to 1000 gets you to 3.1%, and 1500 gets you down to 2.5%.  Polls with larger samples are more accurate, but they aren't *that* much more accurate.

Also, the margin of error reported for a particular poll is the margin of error for a single measurement, not for a lead of one candidate over the other, the variability in the lead is substantially larger.  ARG has a good explanation of this as well as a ballot lead calculator here americanresearchgroup.com/moe2.html

Finally, it's important to remember that, unlike polls, actual votes end with "undecided" at 0% (well, unless you're in Michigan).  The fact that many people tend to make up their minds after being polled obviously tends to increase the amount by which polls differ from the final result.

This isn't to say that there are no problems with the polling industry.  I just dislike it when people throw up their hands and say that polls are meaningless when what they really mean is that polls aren't as accurate as we'd like them to be.

May 29, 2008 7:11 PM

liberal reformer said:

By posting this quote, Michael, you are going to depress many on this website who love to parse polls, discuss crosstabs and compare poll against poll. I have not been impressed by the bulk of polling this electoral season. The US pollsters indeed should draw from larger samples.

May 29, 2008 7:22 PM

apfrankel said:

AlanSP is exactly right.  The author first states that the problem is not in the smallish 4% "margins of error" in sampling, but reflects something deeper.  He proposes to solve this by increasing sample size -- which would only serve to reduce margins of error to, say, 2 or 3%.  He and his editors lack a basic understanding of the issue and probably shouldn't be telling pollsters and statisticians how to do their jobs.

May 29, 2008 8:54 PM

bcbaird said:

AlanSP beat me to it, but yeah - the polling size isn't so much the issue as which questions are asked, the manner they're asked and, of course, the general flakiness of voters.

May 29, 2008 11:49 PM

tomeg said:

I find a graph of poll results taken over a longer period of time (e.g., 1st of April through end of May) a good deal more helpful than the difference between last week's results compared to this week's (or one week's Rasmussen vs. Zogby). (Brendan Nyhan prepares some pretty neat graphs, btw, whether you understand them or not. ;-)

May 30, 2008 12:51 AM

roidubouloi said:

The fact that most of the national polls consistently show biases in a particular direction, e.g., favoring Hillary or Obama or McCain or one's comparative performance compared to another, suggests that there is a problem other than sample size.  Also, in many instances the samples do run in the thousands which should make the sampling error small.  The problems seem to be:  the questions asked, the way the samples are drawn (this seems to me likely to be the source of the biases), the way tracking polls are adjusted for likely voters when they are.

It is helpful to be aware of the trending of the polls.  When all of them, with their biases, are trending in the same direction, the chance is that something is happening.  When all of them, with their biases, are on the same side, it is considerably more likely that at least that side is up.  When there is an extreme outlier, I ignore it.

It is actually possible to read a collection of poll results over time and have a decent idea of what is happening.  Right now, one can say with confidence that Obama is clearly preferred for the Democratic nomination and that a Democrat, either one, is preferred for the General Election but that it is close and  still volatile.  How much more can one expect to know?

May 30, 2008 8:34 AM

roidubouloi said:

By the way, the standard recommended sample size is 300.  I typically use 400-500 to be "on the safe side."  The reason for using a bigger sample than that is so that the cross-tabs are meaningful (since the involve sub-samples) rather than to improve the stability of the overall result.  If the sampling method is biased, sample size is not going to improve the outcome at all.  

May 30, 2008 8:37 AM