Posted 8:22 AM by Robert

A Word Of Caution

Barry Casselman from RealClearPolitics makes some good points:

A word of caution to Republicans: Even if there is a GOP surge in the closing days, most of the critical House and Senate races remain too close to call. If some candidates who were written off only a few days ago now have their races "at play," there is no guarantee that Republicans will actually win many or most of these races, or many of the "battleground" contests so important to control of each house of the Congress. In spite of the apparent surge, it might not be a good night for Republicans when the votes ar counted. ....

The Democrats have perhaps taken the greatest risk by allowing a "fellow-traveling" media and even non-partisan pundits to create the expectation of an historic Democratic victory that will restore their party to congressional power and rebuke President George W. Bush. A landslide could happen, but the only "objective evidence" of this has been a myriad of volatile political polls, few of which may have been accurate or instructive. Political polling has become a lucrative industry as we begin the 21st century, but the quality of polls has been in decline for decades.

John Podhoretz has recently said that most political polls this year are "garbage," and that the pollsters know it. The low initial response to pollsters by the public is one of the reasons why the "margin of error" is much larger than pollsters will admit. With hundreds of millions of dollars at stake, it is no wonder that most pollsters maintain a cabal of silence about the true value of their product. In virtually all of the contested major races this year, there are numerous polls which have attempted to measure voter sentiment. There are local and state polls (mostly by newspapers and universities), and national polls conducted by consultants and pollsters for cash and for media attention that promotes their polls. ...

I am convinced (and believe it can be undeniably demonstrated) that the so-called margins of error in most political polls is not the usual three to five points, but really ten points or more. If that is so, the dependence on polls by the media and politicians is not only unjustified, but is likely to misinform the public. This year the evidence is rather overwhelming in that similar polls taken at the same time with similar samples have come up with such divergent results. Who is to be believed? ....

I agree with Barry on this as far as he goes. But I think the new problems with the polls are far worse than their concealed margins of error. Many public polls appear to have acquired, perhaps fromtheir media paymasters, a new concealed bias towards the left. Barry points out that at least one poll (the Star Tribune private poll) has confessed to such a naked bias in the past. But there is lots of evidence that other pollsters-for-hire are giving the left-wing media what it wants and pays for: Left-leaning polls that support left-leaning articles in left-leaning media outlets. For example, how else does one convincingly explain a Washington Post/ABC News Poll from a polling company (Sampling, data collection and tabulation by TNS of Horsham, Pa.) in the aftermath of the Foley pseudo-scandal - a survey whose sample base included 41% more Democrats than Republicans? Of course, pollsters delivering biased results have a problem close to election day: The vote will expose the bias unless the pollster finds a "last minute surge." Gee, that seems to be happening now. Zogby has been particularly embarrassed by this kind of trickery in the past.

Hey, it's a cheap way to fill newscolumns as long as the suckers don't catch on!

I have some serious issues with another of Barry's comment from the same column

:It has been observed by some that in the closing days of this year¹s elections the Republicans appear to be surging. If this is so, it is because most of the undecided voters this year were those who usually vote Republican, and they are coming home to vote for their party on Election Day.

What does this analysis do for one's understanding? Attributing causation to some voters "coming home" to their historical party affiliation is not very useful, especially in this election where putative shifts in party affiliation in the past 2 years seem to be a highly controversial aspect of the polls themselves. Why were normally GOP voters disaffected in the first place? - have those factors changed? If they were so disaffected, why are they "coming home" now instead of not voting at all? After all, "not voting" is what many pollsters and media had emphatically predicted for these putatively disaffected voters. Why does that prediction seem to be failing? For example, some fiscally conservative voters were supposedly disaffected by a Republican Congress spending too much. Have those voters suddenly realized that Congress has not been spending so much after all? I don't think so. Were they really so disaffected in the first place? Attributing causation to historical party affiliation is accounting or history, not causation - and it certainly does nothing to help one make predictions even one day in advance.