Frequently on Dana's website, Jesurgislac will argue that John Kerry won in 2004 because early exit polls showed him ahead. No matter how frequently we try to explain that the polls were done by an inexperienced firm and they were, well, wrong, Jes refuses to accept it. This isn't surprising, given that Jes still whines about me deleting its lying comments (for explanations, see here and here).
The problem with polling is that it is only as good as the people doing the polls. Plus, there's always the possibility of a variety of flaws in the data gathering that can skew results.
Take this poll that has lefties all aflutter.
It’s hard to say which is worse news for Republicans: that George W. Bush now has the worst approval rating of an American president in a generation, or that he seems to be dragging every ’08 Republican presidential candidate down with him. But According to the new NEWSWEEK Poll, the public’s approval of Bush has sunk to 28 percent, an all-time low for this president in our poll, and a point lower than Gallup recorded for his father at Bush Sr.’s nadir. The last president to be this unpopular was Jimmy Carter who also scored a 28 percent approval in 1979. This remarkably low rating seems to be casting a dark shadow over the GOP’s chances for victory in ’08. The NEWSWEEK Poll finds each of the leading Democratic contenders beating the Republican frontrunners in head-to-head matchups.
There are a couple of big problems with this poll (although I'm sure the Newsweek people were clapping excitedly at the results).
First, the election is 18 months away and a lot of things can happen between now and then to change this data. For example, Dana frequently points out that a party typically holds on to the White House for eight years and then there's a change of parties. However, this didn't hold up in 1988 because, while George H.W. Bush wasn't the most attractive candidate, Michael Dukakis was a terrible candidate. The fact is, anything could happen in the next 18 months to make this particular poll completely irrelevant.
The second point is something that Captain Ed points out.
Yes, this would be a devastating poll, if one could rely on it. It contradicts nearly every other poll, which has consistently shown Giuliani beating Obama, Clinton, and Edwards. How could Newsweek get the results they have published?
Well, for one thing, it helps when you poll 50% more Democrats than Republicans. If one reads the actual poll results all the way to the end, the penultimate question shows that the sample has 24% Republicans to 36% Democrats. Compare that to the information given by Newsweek's NBC partners in February, which showed that party affiliation had shifted from a difference of less than a percentage point to a gap of 3.9 points -- 34.3% to 30.4%, with 33.9% independents.
Does it really surprise Newsweek that a sample where half again as many Democrats as Republicans were polled tend to prefer Democrats for President? Do they find it all that surprising that George Bush isn't terribly popular when surveys oversample Democrats? They knew that the poll had to have some problems; the margins of error for the poll were 7% for the Democrats and 8% for the Republicans, quite high for these kinds of polls.
Newsweek apparently doesn't employ people like editors and fact checkers before rushing their analyses to print. Thankfully, the blogosphere can take the time and effort to have these layers of correction so that we can provide the best possible information to our readership.
Don't believe what the polls tell you.
|