Posted: 9/29/2004 6:26:55 AM EDT
[#15]
It's not the statistics, but the statisticians that are flawed. This seemed like a good explanation of the current weighted polling methods. ******************************** From the WEEKLY STANDARDHeavy Weight SpinExcerpt from this weeks Weekly Standard "Scrapbook" Remember, back in June, how the Los Angeles Times released a poll showing John Kerry with a sizable 51-44 percent lead over George W. Bush in the presidential horserace? Weirdly enough, the same poll also showed Bush simultaneously ahead among independents, and running better among Republicans than Kerry was running among Democrats. How could this be?
Matthew Dowd, the Bush campaign's top pollster, offered one obvious answer: Democrats were wildly overrepresented--there were one-third fewer Republicans--in the pool of voters the Times had surveyed. Had the Times properly "weighted" its results--adjusted them for the actual party-ID breakdown thought to exist across the country--things would have come out differently. Dowd called the Times poll a "mess."
Piffle, replied wonkishly inclined Demollectuals. Ruy "Emerging Democratic Majority" Teixeira, for instance, insisted that there was no need to "weight" the Times poll's peculiar party ID numbers because "there are ample grounds for thinking there is, in fact, a surge toward the Democrats . . . among the broad electorate" (props to blogger and Democratic pollster Mark Blumenthal for resurrecting this quote).
But that was then. Modern statistical science has apparently since discovered that party ID "surges" are only real when they benefit Democrats, and that it therefore is sometimes appropriate to "weight" poll results--like whenever that's the only way to make an apparent Republican advantage disappear.
Two weeks ago, for example, a new CBS/New York Times poll put Bush on top of Kerry by eight points, based on responses from a pool of voters with slightly more Republicans than Democrats. Teixeira cried foul: CBS and the Times had failed to "weight" their poll!
Actually, the race is tied, Teixeira argued on his website--"if you weight their data to conform to the 4-point Democratic party ID lead which we have good reason to believe is the underlying distribution in the voting electorate. . . . Once you adjust for the apparent overrepresentation of Republican identifiers in some samples, the polls all seem to be saying the same thing: The race is a tie or very close to it."
A man can do wondrous things with statistics. Especially if he's not too worried about looking like a hypocrite.© Copyright 2004, News Corporation, Weekly Standard, All Rights Reserved.
|
|