Thursday, October 30, 2008


I understand why Karl Rove would want the readers of his latest Wall Street Journal column to think John McCain still has a chance -- but (surprise!) he really has to stretch the truth to make his case.

Start with the headline:

Don't Let the Polls Affect Your Vote
They were wrong in 2000 and 2004.

Problem is, he doesn't cite any actual polls of the kind we're looking at now that were wrong in 2000 and 2004. Yes, he cites the '04 exit polls that showed John Kerry winning, but (I'll disregard for the moment the possibility of vote-counting skulduggery) those polls were necessarily incomplete (they were taken in the middle of the day, when many voters were at work).

For 2000, in addition to the (I'd say understandable) calls of Florida by some TV news organizations, Rove cites something that wasn't even a poll at all, a forecast that was announced months before the election:

...academics gathered by the American Political Science Association at the Marriott Wardman Park Hotel in Washington on Aug. 31, 2000, to make forecasts declared that Al Gore would be the winner. Their models told them so. Mr. Gore would receive between 53% and 60% of the two-party vote; Gov. George W. Bush would get between just 40% and 47%. Impersonal demographic and economic forces had settled the contest, they said. They were wrong.

That's a "poll" that didn't ask a single voter about his or her presidential preference. It's not a poll at all. How on earth is that comparable to the polls we're reading now?

(A Washington Post article on that 2000 forecast is here. The political scientists were relying on economic data, the outgoing president's popularity, statistics such as "the number of terms the incumbent party has been in power," and, yes, in at least one case, actual horse-race polling data thrown into the mix. That's not a poll.)

And of course, Rove has to get 1980 in:

Only one time in the past 14 presidential elections has a candidate won the popular vote and the Electoral College after trailing in the Gallup Poll the week before the election: Ronald Reagan in 1980.

Yup -- except that that Gallup poll was conducted before the only presidential debate that year:

In 1980, Carter consented to only one debate with Reagan, held on Oct. 29, less than a week before Election Day. In an Oct. 24-26 Gallup Poll, Carter led Reagan by 3 percentage points, 45% to 42%, among national registered voters. A post-debate registered voter trial-heat figure is not available in Gallup's published records, but in Gallup's final pre-election poll of "likely voters," conducted Oct. 30 to Nov. 2, Reagan led Carter by 3 points, 46% to 43%.


... a review of the late 1980 polls shows that while Reagan soared over the final week (following the campaign's one and only debate on Oct. 29), the contest up until that point was tightly competitive, not trending toward the incumbent Democratic president. At the time, the Associated Press reported "new polls say the race between the two men remains too close to call."

A post-election summary of polls by then-CBS News pollster Warren Mitofsky shows that at no point over the final two weeks did Carter have a lead bigger than three percentage points. There is a published Gallup poll not included in that report showing Carter up six among likely voters in a poll conducted Oct. 24 to 27. Whether six or the eight points cited today, Carter's advantage in Gallup polling was offset by similarly large Reagan leads in NBC-Associated Press or DMI (Reagan's pollsters) polls.

So Reagan's "comeback" win wasn't even a comeback according to some polls.

This year, maybe all the polls -- and I mean all of them -- are wrong. But if that's the case, then there's no precedent for what's happening in the elections Rove cites.

No comments: