Tinsel Wing

Monday, November 27, 2006

The margin of error in the margin of error

If you aren't a poll junkie like me, you can skip this one.

Though he leans well to the right, Rasmussen is a very good, very professional pollster, who bothers to publish a new job approval figure for Bush every day. (It's a tracking poll, which is to say, in order to get his 3.5 percent margin of error, he polls 300 odd people each day, and the day's published figure is the rolling average of the three most recent days.) He is also commendably transparent in discussing his methodology.

Recently, he explained why particular polls consistently show Bush's job approval higher or lower than other polls do. It depends on which of three usual ways of posing the question are used. Rasmussen's way, which requests a "Strongly approve", "Somewhat approve", "Somewhat disapprove", or "Strongly disapprove", regularly adds several points to the Prez's numbers. The trick is to ignore the absolute levels, and just follow the ups and downs within any given poll. Those changes will track each other closely, regardless of the form of the question.

My own interpretation, after reading R's essay, is that respondents, even when they think the Chief is sucking more than a little, tend to want to give him the benefit of the doubt if the poll lets them do that. If their feelings are negative but not strongly so, they'll pick whatever answer looks to them like the lowest passing grade.

0 Comments:

Post a Comment

<< Home