Poll gap, why must you narrow?

October 14, 2008 9:06 PM

Why does a political candidate's lead in opinion polls tend to shrink as the election approaches?

I read this in the New York Times today: "wide gaps in polls have historically tended to narrow in the closing weeks of the race". I read that ALL the time, but I've never seen an explanation for the phenomenon. Anyone know the answer? Wild guesses are fine, but a substantive answer is preferred...

posted by mcstayinskool to law & government (7 comments total)

My offhand guess would be our old friend regression to the mean.

Major political races tend me to be close; so large leads probably tend to be aberrations more often than not. Also candidates who are behind have nowhere to go but up, so they pull out all the stops in an attempt to come back. Also also, things that create bounces, like conventions and debates, are over by that point, so that might return the polls to a more balanced state.
posted by drjimmy11 at 9:21 PM on October 14, 2008

Finally, it's possible that the pollsters have improved their methodology in one way or another and are just getting more accurate as the race goes on.
posted by drjimmy11 at 9:23 PM on October 14, 2008

It may have something to do with the nature of the polling organizations.

I read that some polls don't consider people leaning one way or the other and just group them with the undecideds, even though they really may end up voting for the party they were inclined to vote for back in the primaries.

In short, my guess is that less people are marked down as undecided.
posted by abdulf at 9:24 PM on October 14, 2008

Some folks believe that the press's need for a tighter race to keep interest and ratings/circulation high affects coverage so that underdogs get more favorable coverage in the final weeks. Some journalists have acknowledged this to be the case; Media Matters just posted an excerpt from Howard Fineman saying just that about Bush/Gore in September 2000:

I don't think the media was going to allow just by its nature the next seven weeks and the last seven or eight weeks of the campaign to be all about Al Gore's relentless triumphant march to the presidency.

We want a race I suppose. If we have a bias of any kind, it's that we like to see a contest, and we like to see it down the end if we can. And I think that's partly the psychology at play here.

His implication being that "the psychology at play" in newsrooms affects coverage to make races seem closer. The update at the bottom of the post links an NYT reporter saying the same thing on Monday:

Campaigns have rhythms, and inevitably swing back and forth for all kinds of reasons, including mistakes by candidates...and the news media’s desire for a competitive race and tendency to find the “underdog is surging” story line irresistible.
posted by mediareport at 9:53 PM on October 14, 2008 [1 favorite has favorites]

It's as simple as this: Undecideds, the closer to election date, tend to break to the underdog. The thinking among pollsters is that if they were going to go with the frontrunner, they would have already jumped on the bandwagon. This especially holds true the last 10 days of an election.
posted by Gerard Sorme at 10:01 PM on October 14, 2008 [2 favorites has favorites]

There are many reasons, some practical and some technical. As the election approaches, one thing that happens is that pollsters tighten, ratchet by ratchet, their "likely voter" model and weight their results accordingly.

The media certainly may play an underacknowledged (by themselves) role in keeping the horse race as apparently close as possible. There's even been some discussion this week because WaPo political reporter Dan Balz wrote a piece essentially arguing that now that Obama is the clear frontrunner, he should be scrutinized more, because he's more likely to become President now. Atrios calls this "throw[ing] up every potentially bad thing around one last time just in case voters didn't hear it the first 300 times they covered it", and attributes it to an approach he calls "presidency as reality TV show". Glenn Greenwald is even more critical. How much all this really affects the polls, though, is questionable.

Wonks often call the phenomenon a kind of buyer's remorse, and it now gets dragged out generally whenever a frontrunner's poll numbers do anything other than shoot upward. (Apparently Democrats have had buyer's remorse several times this year already, judging by a Google.)
posted by dhartung at 1:26 AM on October 15, 2008

FiveThirtyEight.com has a discussion on this subject.
posted by smackfu at 8:49 PM on October 18, 2008