Articles and Analysis


AP/Pew: Why a 19 Days of Interviewing?

Topics: 2008 , The 2008 Race

The big poll release of the day is the new IA/NH/SC/national survey from the Associated Press and the Pew Research Center. Most that have posted results so far have noted an unusually long field period for the statewide surveys: 19 days from November 7 to November 25.

One reason for the longer field period is that the Pew Center, like the ABC/Washington Post surveys, but unlike the Register/Selzer and CBS News surveys, sample with a random digit dial methodology that begins with all adults in Iowa rather than all registered voters. As such, it requires more dialing.

However, the long field period and delayed release make it less than comparable to the new Des Moines Register survey that was released over the weekend and fielded from November 25-28. I asked Scott Keeter, Pew's director of survey research, about the rationale for their longer field period, and he responded via email:

The longer field period was a result of the fact that the purpose of the study was an in-depth look at the issues and voter considerations across three states, and not just the horse race. Long interview, two races, three states - the practical impact was a long field period.

As it happens, Gary Langer, polling director at ABC News, concluded his blog this morning about what constitutes a statistically significant "lead" in Iowa with a very similar point:

All this underscores one of the fundamental points about pre-election polls: They are estimates. Even with good-quality methodology, the notion of pinpoint accuracy is a myth. And the reason we do them is not simply to try to puzzle out who's ahead – but to understand how and why the voters are coming to their choices.

I'm curious to hear from our readers: Which is more important to you in the pre-election polls you follow: Their ability to accurately estimate who is ahead or how well they help you understand how and why the voters are coming to their choices?



What's most important to me is trend lines, and general comparability with other professional telephone polls.

The broader issues questions are important too, but they could have been conducted separately from a simple horse race poll, so as not to make the results so stale. The elapsed time between the first calls and the release of the results to me makes them fairly worthless.



I suspect I am unusual in that I tend to pay more attention to the in-depth material than the headline results, at least when we are more than a couple weeks away from the relevant election.

That said, it seems to me this is a bit of a false dichotomy. I assume that the in-depth material can change as well over time, and so to the extent we are looking at a dynamic picture even with respect to those matters, a longer field period is still a handicap.


Ben Martin:

I'm always more interested to see how and why people arrive at their decisions rather than who is "ahead" at that particular moment. The less pubicized questions about particular candidate attributes and positions are my favorite part of any poll. However, I have to admit that as the actual dates of the caucuses and primaries draw nearer, the horse race becomes increasingly interesting (especially if my candidate of choice takes a lead.)



Talk about a non-random sample. Asking your readers whether they are interested in the horse race polls (ie how accurately they represent who is ahead,) is asking them to admit that they are interested in polling for entertainment rather than science.
The truth is that most people want to see who is rising and who is falling, and a short fresh sample provides that information most quickly and accurately.



I'd like to see short-term results, because I am interested in how people react to recent events and changes in campain strategies. So may things happen in 19 days, the average is not very meaningful.



The Pew poll is great -- lots of stuff on how voters are thinking, and the comparisons to 2003 are really useful. Among the nuggets are that Dem. voters like the candidates more this time and are less focused on electability. Obama's 2nd-choice strength may explain the horserace developments we see now; OTOH HRC owns health care. You can pretty much map the strategies of the two major campaigns from these results. I see no daylight for Edwards in this poll.



I don't know how to interpret a poll taken over a 19 day period in light of the rapidly changing dynamics in IA. Potential voters are receiving information from candidates on a daily basis through new ads, candidate forums and debates. What was true 19 days ago may be quite different now and there appears to be no way of differentiating this from the effect of events two days ago. Your thoughts on interpretation Mr. Blumenthal.



Pew found 460 of the 2,111 registered voters it interviewed to "say they will definitely or probably attend a Democratic caucus." This would seem to correspond to a turnout for the Democratic caucuses of roughly 450,000 voters, which would be more than 3.5 times larger than the estimated turnout in 2004 and somewhere in the range of 2.5-3 times larger than the high end of predicted turnout for the Democratic caucuses next month. In short, one could question whether Pew's likely voter screen here was sufficiently tight.



" Which is more important to you in the pre-election polls you follow: Their ability to accurately estimate who is ahead or how well they help you understand how and why the voters are coming to their choices? "

That depends on who the "you" is. As a social scientist I am interested in why people vote the way they do. If I were a New Hampshire voter, I imagine that I would have different considerations. Voting strategically, I would want to avoid wasting my ballot. Therefore, the candidates' position in the horse race would matter a great deal. (I say New Hampshire rather than Iowa as caucus night rules make strategic voting much easier than it is in a ballot primary)



I think the PEW/AP poll is just fine (though i am not a Hillary supporter) from a statistical viewpoint. Their polls are the most accurate and most reliable. Yes, it was done over a longer period of time (November 7 to November 25) than the other polls, which gives this poll more validity and a better longitudinal analysis. A longer field period allows to capture changing trends over a longer time. And as we see here, at least on a national level, Hillary is still in a very strong position.



To the previous poster. You failed to explain how this poll exactly shows changing trends. It's not like the poll was broken down in three groups showing us how things varied from say, Nov. 7th-12th to the next 5-day polling period, etc.

This poll is useless. It tells us that Hillary is leading Obama, but the bulk of recently conducted polls suggests she isn't. Obama has taken a slight lead. We are living in December 3rd, 2007, and we want to know the political reality, and what most approximates this reality is a short period of time not far from December 3rd, which is what the Des Moines poll has that Pew hasn't.

Debates have been broadcast, media hit jobs have been published, rumors have been spread, candidates have screwed up, since November 7th to this day.

Some people who voted in this poll during the early days of November probably changed their mind by the time the poll ended on the 27th.

This method is ludicrous.


Nick Panagakis:

What's more important, their ability to accurately estimate who is ahead or how well they help you understand how and why the voters are coming to their choices?

Both are important.

These are four separate polls. They should have been done sequentially, IA then NH then SC then National and results released as avaiable so they could be both timely and useful in terms of depth of coverage.

Comparisons between states would not be exact in terms of field periods but close enough. This is the "least/worse" option.


Chris G:

both are important, for very different reasons.

as far as the horserace, I'm not just interested in the most recent estimate, but also *real* variability in voter preference as a function of time. I think many of us here would agree that trends aren't always smooth.

so totally agreed with above posters that a 19-day interval is way too long, even for a deep understanding of why voters choose a certain candidate. we should at least see a distribution of respondents by demographic as a function of time.


Horse race polling seems to give people even less reason to pay attention--the narrative only reports who is ahead.

Or maybe I'm just bitter that my candidate isn't winning.



I watch the polls daily, and whether you like it or not Hillary Clinton will win the Democratic Nomination, period. Barak Obama will be ready in another 8 years. I know some of you would like it to be otherwise. But he is just not ready to be, President. Sorry, Mr Blumenthal. Hillary is ahead. Good wishful thinking, though.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.