Monday, May 17, 2004

Some Bad Science



Or quasi-science. I came across two examples today. The first was in a Women's Magazine! I bought it for a study I'm doing of magazines intended for women and men, respectively, and tomorrow I'm going to buy some more mags. What should I get, by the way? I have Elle, and I'm planning to buy a few others of similar nature plus something like GQ or Maxim's? I'm a babe in the woods in this field.

Anyway, this study in Elle was about gender differences in people's views about internet sex and pornography. It argued that though the percentage of men who have used internet porn is much higher (around 80% of respondents) the number of women who have also done so is unexpectedly high (around 60% of respondents). The problem with this study is that it is utterly meaningless: it is not based on random sampling. To be admitted into the study, you need to take the initiative. This means that only people who a) care enough to answer and b) have spare time on their hands will participate. Also, one needs to be a reader of the magazines who carried out the study. The results can't be generalized into anything at all, so they're of no real interest. Just as an idea of the problems, note that only 25% of the respondents were female.

Of course, most people probably are aware of the uselessness of these sorts of studies. What is a lot more dangerous is the spread of some factoids from old, poorly done academic studies. I came across one today in an NPR program about the Bush administration Marriage Initiative (not for gays or lesbians, of course). The program described one of the religious courses on marital preparation that we as taxpayers support, and one of the participants in this course stated that she was so worried to find out that people who cohabit before marriage have higher divorce rates, as she and her husband had done so.

I sat there listening to this stuff that my taxes finance, and steam rose out of my ears. I happen to remember the study from which this horrible factoid sprung: it was published in the 1980's and is a very good example of very bad research.
In a nutshell, the study found out that average divorce rates were higher in the group of couples who had cohabited before marriage than in the group of couples who had not done so (or admitted to it). No problem yet. But here comes the problem: the researcher then went on to argue that living together causes higher divorce rates! There is no evidence for this argument, and much evidence against it. The most likely real reason for the observed correlation is that people who are opposed to living together before marriage are also opposed to divorce. The results say nothing about how good or bad these marriages are, and it's inane to argue that knowing your future spouse better before marriage would make you more likely to get divorced than not knowing about the smelly feet or the snoring before you stand in front of the Justice of Peace or priest or minister.

Yet obviously this is what is being fed to prospective marriage partners as 'scientific findings'. And not only fed to them, but at the taxpayers' expense. I shouldn't be surprised by anything this administration does anymore, yet the fact that I do feel some new outrage shows how terribly naive I have been. A naive goddess! Whoever heard of such a thing?