Thursday, January 18, 2007

Social Science Scavenger Hunt

by Drek

One of the things that I love about the scientific process is that it creates puzzles. Now, don't get me wrong, science is a great way to answer questions as well, but the journey to an answer often involves the generation of new questions. So, as a result, trying to figure something out turns into a sort of empirical scavenger hunt. First, we locate all the pieces of the puzzle, and then we kill ourselves trying to put them together. Maybe it doesn't sound like much fun to most people, but to me it's a great way to spend a career.

Recently I came across a great example of a scientific puzzle and, knowing how my readers run towards the intellectual and all, I've decided to share it with y'all. The hum-dinger in question comes to us via Tara Smith's blog Aetiology and will likely make poor Plain(s)Feminist's head spin around like that girl in the exorcist. The puzzle is as follows: a large study of European women has found that moderate physical activity seems to reduce the rate of breast cancer. This probably won't surprise anyone, except that the only physical activity that seems to matter is housework. Yes, ladies and gentlemen, that's right: physical activity at work and in recreation don't seem to matter- only physical activity as a part of domestic labor.

It goes without saying that certain sources are using this study to argue that feminism is responsible for breast cancer. This explanation, however, sounds like crap to me. This isn't because of some sort of ideological stance- it's a skepticism born of a certain amount of experience.

To understand what I mean, let's think about an activity I often give my students. When talking about research methods I tell them about the very real finding that the level of ice cream production is associated with levels of forcible rape, and then ask them to figure out why. Over the years I've gotten some pretty fascinating answers, including that ice cream contains a chemical that increases aggression, that "brain-freeze" makes people angry, and that after a rape women turn to ice cream for consolation. Needless to say, most students are pretty baffled because, really, how the hell could ice cream cause rape? Well, the answer is that it doesn't: both are controlled by a third-variable. As discussed elsewhere, both ice cream production and criminal offending rise in the summer when more people leave their homes and, so, both appear to be correlated with each other. Correlation, as we have been taught over and over, is not the same thing as causation.

As I explain to my students, statistical tests and relationships are a little like the forensic tests used in the television show CSI. The tests allow us to say certain things with certainty but, by themselves, these tests don't usually answer our questions. If a police officer finds out that a certain person's fingerprints were at a crime scene, they still don't know why. Perhaps that person committed the crime, but then again, perhaps they simply passed through earlier in the day. Similarly, these results tell us something about labor and cancer, but maybe that something isn't obvious. Odds are, actually, that it isn't.

We have a finding here that housework appears to protect women against breast cancer but we're left with the simple question: why? Why is it that housework appears to act as a protective factor and not other kinds of labor? Is there something about scrubbing a bathtub, or dusting a shelf that is intrinsically more cancer-protecting than, say, playing tennis or walking? Maybe but, then again, like ice cream and rape, perhaps what we're seeing is the impact of another sneaky third variable. Minus a wrathful, and exceedingly slow, god who really likes to see women doing menial jobs, I tend to doubt that breast cancer is the wages of feminism.

So what could it be? Well, that's the question for all of us today. The study used multivariate statistical methods to control for a variety of factors including age at menarche, age at first pregnancy, drinking frequency, smoking status, use of hormone replacement therapy, and use of oral contraception. None seemed to make a difference. It also attempted to control for the relative strenuousness of particular physical activities. In most respects the methods are solid, although most of the data is self-report and, in some cases, retrospective, which can introduce quite a bit of error. Additionally, as Tara points out, they ask about current levels of activity, hormone replacement and oral contraception use, but not historical levels. So, a woman who has used oral contraception for twenty years and stopped yesterday looks like a woman who has never used oral contraception. That's certainly an issue.

Is this it? Do these failings explain the results? Maybe. Then again, maybe we can think of some alternatives that are slightly more plausible than "feminism gives you breast cancer." Family income wasn't tracked- is it possible that women who do more housework are often in a higher socioeconomic class where nutrition and medical care are generally better? Does it matter, as Tara comments and as Arlie Hochschild would agree, that all women do a lot of housework in their presumptive second shift? What about women who work as maids or nannys? They should be doing a whole lot of housework on a daily basis and, logically, should be about immune to breast cancer if these results are correct. Are breast cancer rates for female domestic servants, indeed, dramatically lower than for other low-income women? Is it possible that breast cancer has some sort of bacterial or environmental origin and women who spend less time out of the home have a lower probability of exposure? Is it something I haven't even thought of yet?

Give those analytical minds of yours a workout and see if you can't crack this puzzle. Or, at the very least, keep me company while I try.

God I love the smell of a conundrum in the morning!

Note that this was cross-posted over on Total Drek.
One other thing that occurred to me:

Did they control for the number of pregnancies? I know I've seen stats that suggest that more menstrual cycles leads to higher risk of breast cancer. In the past, a married woman without access to birth control (i.e., throughout most of human existence) could expect to be pregnant or breastfeeding for the majority of her adult (premenopausal) life. The difference in number of menstrual cycles is really shocking. Perhaps women who perform a lot of domestic labor have more "traditional" relationships, and thus are spending more time not menstruating.
While the multivariate analysis would remove some confounding-factors issues, it doesn't look like this could be said to have dealt with all of them, based on Mrs. C.'s comment and some of the other discussion cited in the post.

Investigating domestic help might be the Freakonomic thing to do, though in many ways it doesn't have the hallmarks of a good natural experiment (i.e., there are self-selection issues).
Thank you for citing my blog. I am not making any hard statistical claims, I am just siezing a chance to satirise feminism, and draw public attention to the offensive crap talked by the likes of de Beauvoir and Steinem. I find I am expected to hold these women in some kind of reverence. Not me. It's time for men to start getting political on their own behalf.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?