Thursday, April 10, 2008

Kevin Drum reports an appalling discovery based on an interesting method

STUDY: 26% "ANGRY OR UPSET" BY FEMALE PRESIDENT....Last month I blogged about a poll showing that while only (only!) 17% of respondents said they'd have trouble voting for a woman for president, 45% claimed that "most" of the people they know wouldn't vote for a woman.

So which is it? Is the 17% number too low because some people won't fess up their true feelings to a pollster? Is the 45% number too high because people are too cynical about their neighbors? Or what?

Via John Sides, some political science types at Northern Illinois University and Loyola Marymount provide an interesting way of measuring something closer to the true answer. They presented a control group with a list of four items and asked how many of those items made them "angry or upset." The average was 2.16 items. (Respondents didn't have to say which items they were.) Then they presented a second group with the same set of items except they added one more: "5. A woman serving as president." This time the average response was 2.42 items. The poll was conducted in March 2006 (so it's probably not merely a reaction to Hillary Clinton personally), sample size was large, and all the usual statistical controls were in place. The full paper is here and has all the details if you're interested.

So what does it mean? The arithmetic is simple: (2.42 - 2.16) x 100 = 26%. This means that 26% of the respondents were angered or upset by the notion of a woman serving as president.


The methodology is interesting, but I see a possible problem. People might tend to answer half of questions with a yes. Certainly it would seem odd to answer yes to all or no to all. A way of addressing my concern would be to compare scores with 3 sets of 5 questions one with the woman president question and 4 others, one with something which doesn't seem likely to make many people angry (say people praying*) and upset and the same 4 others and one which would make almost anyone angry and upset (say needless war**) and the 4 others.

The number of upsetting things with the woman president question minus that with prayer would be a lower bound to the percent upset at the idea of a woman President. One plus the number of upsetting things with woman President minus that with pointless war would be a lower bound (who knows Dick Cheney might be in the sample).

My guess is that the lower number would be well below the upper, not because there are lots of bloodthirsty warmongers out there, but because people would score other questions differently as they anchor at 0. Alternatively, the one extreme question might pull the other answers along as people get the idea that they are supposed to say yes.


* I know someone who is furious with Barack Obama for saying he prays to Jesus every night but I think he would be cool if Obama just did it without talking about it. However, there is an anonymous commenter.I chose the example of prayer just so I could write this footnote, then I forgot to.

** I'm assuming that Richard Cheney would refuse to participate in the experiment.

4 comments:

reason said...

People praying upsets me!

Brendan said...

Do you have any idea how infuriating it is to see asterisks and no corresponding footnotes?

Robert said...

No***

Brendan said...

Heh.

I'm reminded of an essay on punctuation by Lewis Thomas. He ended it like this,