Saturday, September 14, 2013

Heritage of dreck

And what, if I may be so bold, was O. Paul Krugman, commonly known as Nobel Prize–winning economist Dr. Paul Krugman, wrong about?
Song by the Frankfurt punk band Die Strassenjungs 
Seems he publicly doubted some poll results that Heritage has been pushing, from August 15: [jump]

How do you feel about defunding Obamacare? A new poll of American voters shows a majority—57 percent—support defunding this unfair, unaffordable, unworkable law.
This was so far off base from what was known at the time—Kaiser Family Foundation had just released a poll showing 57% support for the Affordable Care Act programs—that Talking Points Memo assumed they must have been reading the Kaiser poll backwards, and sent out a Tweet to that effect. But apparently Heritage really had its own poll, commissioned from a firm called Basswood Research, and the 57% number was just a coincidence. Although you'd think it might have struck them that Kaiser is known for doing pretty good work, and if Basswood was getting diametrically opposite results, it might just be the case that there was something amiss with their methodology.

Meanwhile, Dr. Krugman was not impressed, and in a column of September 9 on the terrible shortage of social science knowledge in the Republican party, he used the case as an example:
in an echo of the Romney camp’s polling fantasies, other conservative “experts” are creating false impressions about public opinion. Just after Kaiser released a poll showing a strong majority — 57 percent — opposed to the idea of defunding health reform, the Heritage Foundation put out a poster claiming that 57 percent of Americans want reform defunded. Did the experts at Heritage simply read the numbers upside down? No, they claimed, they were referring to some other poll. Whatever really happened, the practical effect was to delude the right-wing faithful.
Heritage had a BIG hissy fit on September 10:
it is Mr. Krugman who is doing a disservice to his readers by knowingly trading in these falsehoods. His readers, incidentally, really ought to know that he uses the liberal Talking Points Memo as the source of his “research”—in a column decrying that conservatives read only their own press, no less.

In fact they were clearly wrong about that "knowingly", at least. All Krugman knew, as far as I can see, is that their number was wrong, as was clear to anybody who was following the subject; he didn't need to look at the poll to find out. But if he had known what sort of poll they were reporting he would have given them a much harder time than he did.

The sampling procedure they used was like this:
On August 7-8, 2013, Basswood Research conducted a nationwide survey of likely general election voters in ten different Congressional districts.  Six of those House districts are presently held by Republicans, four are held by Democrats.  They broadly represent a cross-section of Republican-leaning but not safe-Republican districts.  The Republican held seats are FL-2, IL-18, NJ-7, NC-2, OH-12, and OR-2.  The Democratic held seats are GA-12, NC-7, UT-4, and WV-3.  The survey was conducted by live professional interviewers by telephone.  The sample had 100 interviews in each of the ten districts, yielding an overall sample size of 1000, with a margin of error of +/- 3.1%, at a 95% confidence interval for the overall sample.
In other words, a sample designed to be biased in favor of people who are more or less likely to vote Republican. It's not even a valid sample of that, since it is not in any way randomized, and since it obviously includes plenty of people who are not likely to vote Republican. But in no way can it be said to provide a representative group of "American voters".

The purpose of the thing, in fact, was to lobby Republican congresspersons into supporting a government shutdown by showing them that voters in Republican-leaning districts were so distressed at the prospect of being forced to have decent health insurance at subsidized rates that a shutdown would not bother them. Even with those limited aims it was a remarkably dishonest exercise, because the populations they interviewed were highly abnormal, and the questions were loaded.

The subjects were 16% urban, 38% suburban, and 43% rural! They were 43% Republican and 33% Democrat! A full 80% were satisfied with the health insurance they already had (including 27% over 65 and therefore on Medicare), but the questionnaire referred to
None of which, to the (pretty limited) extent it is true at all, would have any effect on at least that 80%. In other words it was a common or garden push poll, with no scientific value and no legitimate use.

Indeed, the poll was so badly done even Jennifer Rubin could see what was wrong with it, and said so, quite clearly, a full month ago. And yet here they are, still pushing this dreck, and complaining about Krugman. It's pretty amazing.
Image from

No comments:

Post a Comment