Thursday, November 19, 2009

Lying with (other people's) stats, ch. LXXVII

What are you supposed to do when you want to report on a survey, but the numbers don't quite say what you want them to? Listen and attend as the Fair 'n' Balanced Network offers two solutions:
1) Ignore inconvenient numbers
2) Lie about the other ones
This is too good a front page to pass up. The "Santa Clause" in the picture isn't the "volunteer" purportedly found to have been a sex offender, who's on the other side of the country (what's a little visual libel among friends?), and the "a little turkey" hed seems to have produced exactly the comments that sort of prime is intended to, but we're going to focus on the third bullet in the second item there: "Poll: Majority of Americans dislike Obama policies." Here are the hed and the lede as they appear inside:

Majority of Americans like Obama personally
but not his policies, poll finds
An overwhelming number of American voters say they like President Obama as a person but disapprove of most of his policies, according to a Quinnipiac University poll released Thursday.
So it's not just a majority but an "overwhelming" number who like the guy but dislike the policy -- fair enough? I wonder what the poll itself says!
Three-quarters of American voters - 74 percent - like President Barack Obama as a person, but only 47 percent like most of his policies, and voters disapprove 51 - 35 percent of the health care overhaul passed by the House of Representatives which he has endorsed, according to a Quinnipiac University national poll released today.
Hmm. Seems to be a bit of discord there. The poll offers four choices (like both, like A but not B, like B but not A, and dislike both). The "overwhelming number" who like the president but not the policies is actually, erm, 28%. Add together the groups who dislike the policies and you get 48%, compared with 47% who report liking most of the policies. An appropriate way of describing that finding would be "about even." Back to Fox:
The poll, which surveyed 2,518 registered voters nationwide from Nov. 9 to 16, found that Obama's approach to health care reform is among the president's most unpopular domestic priorities -- with 53 percent saying they disapprove of his policy on health reform while 41 percent said they approve.
Hard to see how you could draw a conclusion like "among the president's most unpopular domestic priorities" when it's the only "domestic priority" on the survey instrument Fox links to (there is a question about executive pay, but it'd be a stretch at best to expect respondents to interpret that as a priority). Nor is the paraphrasing very precise: "Do you approve or disapprove of the way Barack Obama is handling health care?" isn't the same as asking about "his policy on health care reform" (this question about Congress -- "Do you approve or disapprove of this health care reform plan?" -- finds a significant majority disapproving).
These findings are kind of interesting, if you're fond of the incremental, not-very-exciting stuff that makes survey research worthwhile. But if health care is the big domestic deal at stake here, you have to wonder why Fox doesn't report some more of the results. Respondents support "giving people the option of being covered by a government health insurance plan that would compete with private plans," 57%-35%.* Opposition to the opt-out and the trigger is significantly higher than support. Most respondents think it's very (39%) or somewhat (22%) important that Congress "approve of a health care overhaul plan this year." People in general have unfavorable views of both parties, but they're much more unfavorable to the GOP (28% favorable, 53% unfavorable) than to the Democrats (39%-46%).

Here's the point. There are (broadly) two reasons for running stories about public opinion on core policy issues. One, to give people information they need to have a good sense of their place in a democratic society. Two, to support cultural preconceptions about how the world works: bad people and policies are punished in the court of public opinion, and good people and policies are rewarded. If you think public opinion about health care policy is relevant, you're going make one set of selections about which bits of information are more important than others. If you think it's important to make one party look good and another look bad, you make different selections.

That, in short, is how you can tell Fox from a real news organization. Well, that and flatly making stuff up about the numbers.**

* Confidence interval of +/- 2 points at 95% confidence, N = 2,518 registered voters, in the field 8 days.
** All right, that and using the wrong footage to illustrate points about public support for Sarah Palin and the tea parties.

Labels: ,


Post a Comment

<< Home