Should Intentions Or Predictions Be Polled ?

(Caricatures above are by DonkeyHotey.)

I present a lot of polls on this blog, but I have repeatedly warned that, at best, they are only an educated guess at where any election stands -- and that it is much better to look at a lot of polls than to put your faith in any one poll. All you have to do is look at the swing state polls I have been posting almost daily to see that different polls can have very different results, even when polling the same states. These differences are because of factors like what questions were asked, how the questions were asked, who responded to the poll, how those respondents were weighted demographically, whether it was a cell phone or landline or internet poll, or various other factors (including the innate bias of the polling organization).

The other day I posted about a different sort of "poll" -- betting parlors, where people actually bet money on which candidate will win an election. A professor from Nottingham University did a 12 year study of American elections -- comparing the predictions of polls to those of the betting parlors. He found that the betting parlors were better at predicting the election winners.

Now there is a new academic study showing that a specific kind of polling is more accurate than the normal polling done and reported by most polling organizations. The study was done by professor Justin Wolfers at the University of Michigan and David Rothschild who is an economist at Microsoft Research.

Most polls ask their respondents who they intend to vote for. Wolfers and Rothschild say that asking a different question yields more accurate results. They think the poll respondents should instead be asked who they think will win the election (instead of who they plan to vote for). A few polls do ask this question every now and then (Gallup is one of them), and when Wolfers and Rothschild compared voter intention polls with voter prediction polls, they found the prediction polls were significantly more accurate. They said:

The question allows people to consider not only their views but also those of their relatives, friends and colleagues, he said. Some voters may also give more-honest answers about their own plans, rather than naming a candidate who briefly intrigues them, as happened in the Republican primaries this year. And an expectations question allows people to take into account speeches, debates and news media reports. More information produces better results.

Take for instance the 2004 election between George W. Bush and John Kerry. While polls showed the race to be fairly close, an October poll by Gallup, asking respondents who they thought would win, showed that 56% thought Bush would win while only 36% thought Kerry would win. Bush won fairly easily. And that's not the only election where the prediction question proven more accurate than the intentions question was. In presidential elections since 1952, when voters were asked to predict the winner they turned out to be right 81% of the time. But when the polls showed only who voters said they would vote for, they turned out to be accurate only 69% of the time. That's a statistically significant difference.

So, who are the voters predicting will win the presidency in this election. Several polling organizations have asked this question (Gallup, ABC/Washington Post, Politico/George Washington University, and Quinnipiac/CBS/New York Times). In all of those polls, Obama was picked to win by margins ranging from 13% to 24%. The most recent, the Gallup Poll taken in mid-October, showed that 54% think President Obama will win, while only 34% think Willard Mitt Romney will win.

While this should give Democrats a little encouragement, it is by no means a guarantee. Remember that the prediction polls have been wrong 19% of the time (and intentions polls wrong 31%). The only poll that really matters is the one on November 6th at voting sites around the nation. So be sure to vote this year. Voting is not just a right, but the duty of every responsible citizen -- and it determines the future we all will have.