fb-pixelIn defense of polls - The Boston Globe Skip to main content

In defense of polls

Pablo Martinez Monsivais/ AP

The Catholic religion has its own profession of faith. Politicians and pundits need one too, and it should begin, “I believe in the power of polls . . .” The meaning of the 2012 election will likely be subject to interminable argument. The Tea Party, for instance, believes it proved “weak-kneed Republicans do not get elected.” Others think the contrary — hard-right posturing threatens to marginalize the party altogether. But one thing it seems all sides should acknowledge: Pollsters know what they’re doing.

On election morning, New York Times blogger Nate Silver predicted 313 electoral votes for Obama along with a popular vote favoring the president 50.8 percent to 48.3 percent. The Real Clear Politics average of polls just before the election showed Obama up by 0.7 points. The Signal looked at the same polls and predicted at least 303 electoral votes for Obama. The Huffington Post's national tracking model estimated a 1.1 point Obama edge. Nate Cohn of the New Republic reached similar conclusions. The actual results: Obama received 50 percent of the popular vote and 332 electoral votes. In essence, all of these folks were right on the mark.

Advertisement



Contrast that those who got it wrong. One-time Clinton advisor Dick Morris foresaw a Romney landslide — 325 electoral votes to 213. Political analyst Michael Barone gave Romney 315 electoral votes; columnist George Will, 321. Karl Rove — the Democrats' favorite demon — projected Romney with 51 percent of the popular vote. Former Reagan speechwriter Peggy Noonan and the Weekly Standard's Fred Barnes also foresaw a Romney victory.

It wasn't only conservative commentators who got it wrong. Jim Cramer, host of CNBC's Mad Money, saw Obama taking an amazing 440 electoral votes. (Those of you who follow Cramer's investing advice may want to reconsider your allegiance.) And there were also many observers — such as Philip Klein from The Examiner and Ross Douthat of The New York Times — who, while correctly predicting an Obama victory, were far off from the final results.

Those who got it right didn't have any particular insight. Instead, they simply looked at the polls and believed the data. Indeed, Nate Cohn prefaced all of his forecasts by saying "my 'predictions' aren't really mine (they're the polls)."

Everyone else tried to play analyst. Some simply saw pollsters as politically biased and so tried to "fix" their numbers by adjusting them up or down. Others saw evidence that seemed to confound what the polls were saying. Peggy Noonan, for instance, pointed to the crowds at Romney's final rallies and the proliferation of GOP yard signs in Ohio. "The Republicans have the passion now, the enthusiasm," she wrote. "The Democrats do not." Polls, in this view, were dry statistical tools that failed to look into the hearts and minds of voters.

Advertisement



Others questioned methodology. Pollsters were not accounting for differences in turnout from 2008: This year, for example, evangelicals would turn out while young people, disillusioned by Obama, would not. Why, analysts wondered, didn't pollsters get that? So too, they theorized other factors were distorting results: The rise of cellphones, polling burnout (Americans refusing to answer questions), and deliberate lying (some voters would falsely say they were voting for Obama).

There are two problems with all of this analysis. The first is that people see what they want to see. Those favoring one candidate or another would seize on anecdotes contrary to the polls and think of them as some broad new truth. They probably didn't do this intentionally — no one likes egg on the face — but it's an all-too-human failing. It's the exact failing, by the way, that polling is designed to avoid.

The second is that pollsters aren't idiots. Methodological concerns such as turnout and demographics and cellphones were all well-known to the industry and in different ways have been addressed. Polling firms make their money on being right and are aware — more than any armchair analyst — of the pitfalls they need to avoid. This year, quite clearly, they succeeded.

Advertisement



When it comes to the prediction game, data and statistics rule. Analysis is too risky. Next election, if you to want to seem smart to your friends, forget about what you hope will happen and just predict exactly what the polls are saying. You'll be right.


Tom Keane writes weekly for The Globe. He can be reached at tomkeane@tomkeane.com.