Results of an eagerly awaited poll were withdrawn after members of Democratic challenger Dan Maffei's staff discovered a sampling error that tilted the poll toward Republican incumbent Jim Walsh. This video tells part of the story:
When the results were briefly released last night, they were:
Walsh: 47.5
Maffei: 38
Undecided: 14.5
But it turns out that the weighting of Republicans/Democrats/Other used was 44/31/25, which according to Zogby was the breakdown for NY-24. NY-25's breakdown leans more Democratic, 36/32/31. Pollster Zogby has promised to run the poll over again.
I got this email from a Maffei staffer reporting a reworking of the numbers:
Hey folks,
Good news! I just reweighted the Zogby data and found that if he had done it correctly, it would have come out as follows:
Walsh: 43.8
Maffei: 40.4
Undecided: 15.6
For those of you who are interested, I've attached a spreadsheet that shows how I reweighted things. Basically, Zogby identified people as Republicans, Democrats, and others and then got a sense of what percentage of Republicans were supporting Dan and which were supporting Walsh; then what percentage of Dems were supporting Dan and then Walsh; then the unaligned voters. Then he applied those ratios of Republican split maffei/walsh and Democrat split maffei/walsh to the partisan registration of the district as he understood them.
You can take a look at the reweighed poll numbers spreadsheet here. I'm no polling expert, but the methodology looks plausible to me.
Even if the original numbers were correct, Maffei down by less than 9 points with a 14.5 percent undecided among likely voters has got to be worrisome for Walsh. And these polls would have been taken before the Monday debate, in which (I think) Maffei easily outscored his opponent. A 4 point lead is probably within the margin of error, though I confess I'd feel a lot better being on the other side of that margin.
UPDATE: According to the Mystery Pollster, Zogby's weighting method uses exit polling from the previous election. Well, Walsh ran without a Democratic challenger two years ago, so even though NY-25 voted for Kerry, my guess is that exit polling would have shown a higher proportion of people who identified themselves as Republican voters.