Why The Miami Herald's pre-election poll was so far off.

Pre-Election Day polls in Florida predicting Mitt Romney would comfortably win the state’s 29 electoral votes were quite wrong, it turns out.

Though some absentee ballots are still being counted, Romney is narrowly trailing President Barack Obama.

Last week, a survey conducted for the Tampa Bay Times, Bay News 9, Miami Herald and two other media partners showed Florida almost deep red — with Romney winning 51-45.

Three weeks before that, the same polling firm, Mason-Dixon Polling and Research, had Romney ahead 51-44.

A poll conducted for the Florida Times-Union the Sunday before Election Day called it Romney 52, Obama 47.

Mason-Dixon pollster Brad Coker said the shift was not caused by polling error, but because Obama moved the needle with his handling of Hurricane Sandy.

“At the time our poll started, Romney still had the lead in several national polls. Then he stalled late in the week after his campaign all but disappeared from the news for several days,” Coker said. “After that, the gender gap widened, the race fell into a tie nationally and Romney never bounced back.”

But who pollsters call, and how they build their sample is critical, too.

In Mason-Dixon’s case, the polling group ultimately appeared to under-anticipate the number of young voters who would cast ballots in Florida, while over-anticipating the number of people 65 and older who would vote, according to preliminary exit poll data.

The result?

More poll respondents key to Romney’s success, and fewer that were key to an Obama victory.

The Mason-Dixon poll also accounted for more white voters and less Hispanic voters than exit polling data showed.

Read more here: http://www.miamiherald.com/2012/11/07/3086815/how-did-the-presidential-pollsters.html#storylink=cpy

More here

Comments

[none]

you should demand a refund from Mason-Dixon. faulty merchndise

[downballot]

So it started a week before election day? How long was it in the field if it was published on your front page on election day?

[Brillo]

I guess the hurricane somehow only hit their poll, cause a whole lot of other pollsters nailed it.

[FQMiami]

This is nonsense. You should ask Nate Silver why he doesn't even bother with the Mason Dixon poll in his poll aggregate. Mason Dixon was an outlier, showing an advantage for Romney outside the margin of error on the same day the Wall Street Journal poll showed a slim Obama lead. How come Sandy was factored in everyone else's poll except Mason Dixon's?

[Gustavo]

Total BS. You can accept BS from Mason Dixon if you like, but your most informed readers will know that Coker is lying to you and you are either willing to pass the lie on to us or to stupid to know you are being lied to.

The math is not hard on this. Romney's post first debate momentum ended back on 10/12. Coker tells one truth, but it is cloaked in a lie. Romney in fact did continue to lead in a few national polls...however those were the Gallups and Rasmussens who completely failed throughout the cycle. Good pollsters were seeing nothing of the sort. Finally Mr. Caputo, if you believe Hurricane Sandy was responsible for a 7 point turnaround from your poll to actual results, well there is some land in the everglades I'd love to interest you in.

[BradCrompton]

Nate Silver is the guy to read in the NY Times. He uses all the polls and then assigns weight to each poll based on age oof the poll, history of being right or bias, and other important factors like if they call cell phones. He got the first 44 states on the money and his outcome prediction was very close. He is in the Ny Times political section "535". The errors in the Herald are shared by many others as well.

[whasup]

Ya know, it appears Caputo may be dwelling in the same kind of la-la land that caused Chris Matthews to screw up and utter something like "thank God for Sandy"!

Crediting the storm-response for shifting voter opinion in favor of Obama is an insult to Obama, his campaign, and all those workers who fooled the voters into re-electing Obama.

Caputo does at least acknowledge that the demographics of the turnout were very different from what some pollsters expected. That's why poll results should be adjusted to a range that accounts for different turnout scenarios ... the statistical margin of error is different and not adequate to make up for this.

[Alley]

Mr. Coker, you've been put on notice. Sandy had nothing to do with Romney or Obama. Mason-Dixon voter model was an outliar and outdated - and did not reflect Florida's changing demographics.

[CKing]

Here are the other polls at the time of Mason Dixon had Romney +6 in early November: PPP Obama +1; Ipsos tie; Angus Reid tie; YouGov Romney +1; Mellman Obama +2; NBC/Marist Obama +2; We Ask America Romney +0.9; Gravis Marketing Romney +3. In other words, Dix's +6 Romney lead was twice that of the next closest poll, and even that was out of step with the majority. The Herald might have pointed out the discrepancy--isn't that what journalists are supposed to do? Ask questions? But this wasn't journalism. Marc Caputo was ebulliently cheering Romney to a make-believe victory, and Mason Dixon happened to be telling him what he wanted to hear, though other polls were saying something quite different.

[Peter Principle]

Coker is really blowing it out of his you-know-what. How could a hurricane possibly effect a poll taken two weeks before it happened?

The answer is painfully simple: Like the Romney campaign, Gallup and Rasmussen, M-D assumed an electorate that would be much whiter and older than it turned out to be. (At least they didn't bake in a GOP advantage, like Rasmussen).

These guys were all told by the rest of the polling community that they were making a mistake, but they did it anyway. Now they gotta live with the consequences.

But of course, it's always easier to blame the weather.

Coker: I'm sorry teacher, but Sandy ate my homework.