Tuesday, June 10, 2014

More Iowa Poll silliness

Shame on the Iowa Des Moines Register and the Iowa Poll for peddling nonsense from the latest iteration of the poll.

The newspaper and its pollster reported in great detail on their survey about almost a dozen possible 2016 Republican presidential candidates, without mentioning that the survey results are technically meaningless.

Selzer & Co., the pollster, interviewed 400 Republicans who “planned to vote” in the June 3 primary in Iowa. Hardly a credible sample frame, for starters….

Selzer reported the survey had a margin of error of plus/minus 4.9 percentage points.

OK, so here’s the up-front problem: nearly all of the different survey results—for “favorability” and “most electable” and so on—for most of the 11 candidates are with the range of the margin of error, meaning that the only honest interpretation of the findings is “we can’t say there’s any reliable difference between the candidates.”

Technically, based on these poll findings, there are no obvious frontrunners. With a margin of error of 4.9 points, any two results less than 9.9 percentage points apart do not legitimately represent any real difference.

And by the way: the news release I saw made no mention of the computerized statistical “weighting” that pollsters must do these days to even pretend that they’re reporting results from a satisfactory random sample, which of course they aren’t because no pollster can reach a true random sample these days. All political survey results are adjusted and privately cooked by the pollsters to make the results seem more reasonable—but the “weighting” makes them less reliable.

No comments:

Post a Comment