The Uses and Abuses of Polling

Polls, strictly speaking, just generate data, so they can be used or abused according to the needs of those who package and propagate them–or who consume them. Among serious consumers, there are some who use polls to better understand what it happening politically, and there are others who are simply trolling for ammunition for Our Team, and ignore or dispute contrary information.

I’d like to think I’m generally in the former camp, but can’t resist occasionally publicizing polling data that are discomfiting to people I don’t like. Since I’ve been predicting a very good Republican year since about two seconds after the 2012 elections, I’m not alarmed by or tempted to suppress good Republican polling news, though I am deeply annoyed by those who refuse to look at the wildly and very temporarily pro-GOP landscape and instead postulate some sort of epochal and permanent shift in public opinion.

In any event, we’ll be talking about polls enough between now and November 4 (I’ll have a polling update this very afternoon, as a matter of fact) that it’s a good idea to pass along some educated thoughts about how much reliance to put on them. And there’s a good offering today at Politico Magazine from Larry Sabato, that concludes with some warnings about over-confident poll-reading:

First, we can probably expect up to a baker’s dozen of Senate contests to remain highly competitive right up to Election Day. Second, polling averages are likely to mislead us about the eventual winner in one or two cases. And finally, if there should be multiple Senate contests where the pre-election polling average has the candidates separated by three percentage points or less, the polling leader in about a third of these cases may well lose.

Therefore, if we’re headed for an election that produces a Senate divided by only a seat or two, don’t expect polls to precisely predict the outcome. Even well-conducted, large-sample surveys are blunt instruments with a margin of error.

One can be too critical of polls and polling averages, because the truly remarkable thing is that polling is as accurate as it is. Think about the problems plaguing pollsters today, from skyrocketing refusal and non-response rates among potential interviewees to the methodological complications presented by both cell phones and online surveys.

And in the end, we may be waiting on December and January runoffs before any bets can be collected on Senate control. There will, however, be fresh batches of polls.

Ed Kilgore

Ed Kilgore, a Monthly contributing editor, is a columnist for the Daily Intelligencer, New York magazine’s politics blog, and the managing editor for the Democratic Strategist.