Saturday, June 30, 2012

The Oracle Speaks

My views of polling are perhaps familiar. I wrote on the subject about three years ago on the old LaMarotte (link). My chief problems are that the samples are exceedingly small—thus each respondent quite frequently speaks for 226,500 other adults—and the quality of the response is at best dubious. As I put it in that post:

Poll results have zero value. They have no basis in fact. Respondents have no specialized knowledge; they may or may not have thought about the subject. If they have thought about it, their reasoning may be muddled. They’re probably echoing snippets they’ve heard on the media. The practical value? Also nil. If no immediate vote is impending, politicians can’t gain anything at all by changing message with every whim of the media. Opinion is a Rorschach ink blot—read into it whatever you like.

My best mentor in the use of statistics was the chief statistician for Anheuser Bush, a very able and amusing man. Back in the good-old-days, when calculators had some size, he used to wear one strapped to his trouser belt as if it were a six-shooter. He used to say: “The basic requirement for credible work in our field? When you publish something, any one of your readers should be able to check your results at least arithmetically. You’ve got to give them all the numbers.”

Today comes a Gallup Poll on “Americans’” opinion of the Supreme Court ruling Thursday (on the Patient Protection and Affordable Care Act). Results? 46 percent agree and 46 percent disagree with the Court’s ruling. The sample? 1,012 people aged 18 or older. These results are then also subdivided by party affiliation into Democrats, Independents, and Republicans. But here comes the sticking point. Gallup publishes all of the questions asked of this random sample. Those questions, however, don’t include any query about the respondent’s political affiliation. How that sample breaks down by affiliation is not provided. And, yes. I followed every link Gallup provided offering additional information. Nothing there.

In my older post I urge that, given this situation, we might as well return to reading the flight of birds or the shape of the entrails of animals. Might be cheaper.

In the process of trying to figure out what Gallup’s numbers meant, I cam across this article (link) in USA Today which gives a picture of how many registered voters there were in 2004. The numbers are 72 million Democrats, 55 million Republicans, and 42 million Independents. The numbers have changed since, I am told, but that’s a recent benchmark. Gallup having failed me, I think I’ll go outdoors now and see if I can practice augury. There is a grey dove that comes this time of the morning…

No comments:

Post a Comment