Wednesday, March 28, 2007

Reporting On Feminism And Health

Do you remember the big fuss the media made over the 1999 study by Kawachi and others which found that greater gender equality appeared to be correlated with better health for both sexes in the United States? How about the even bigger media fuss caused by the 2005 study by Chen and others which found that gender equality appeared to be correlated with better mental health for women? And surely you remember the excitement in the media last year when we all learned about the Swedish study which showed that both men and women have better health when roles are shared more equally at home?

You don't recall? Neither do I, because there was no such fuss at all. Studies with those findings are not mentioned in the popular media at all or only fleetingly. But when a Swedish study in 2007 suggests that greater gender equality leads to less health for both sexes, what happens? You guessed it. The media is on the study right away:

Warning: feminism is bad for your health
By Roger Dobson
Published: 25 March 2007

Since before Germaine Greer published The Female Eunuch in 1970, and even before Mary Wollstonecraft wrote A Vindication of the Rights of Woman in 1792, campaigners have fought for sexual equality, convinced it is the key to a better society. Now researchers have discovered that gender equality may make people unwell.

It is most interesting, is it not? Consider this: Hundreds of studies are published each month in the social science literature, and only a very few of these are ever publicized extensively. How do those lucky studies get picked? Some of them are obviously important in their findings, but many are selected because they might sell more newspapers or get more television watchers glued to their sets. And I'm beginning to suspect (heh) that there is an ideological point to deciding which studies are to be given more advertising. It will not be studies which suggest that feminism is a good thing.

This has two important consequences. The first one is that the general audience obtains a biased understanding of what the studies show in general. The second one is that people like me have to spend an awful lot of time criticizing and analyzing the mispopularization of studies. It doesn't matter how well I do that, because it LOOKS like all the studies out there are proving points for the anti-feminist side. What is urgently needed is some sort of a way of getting a more representative sample of studies into the popular debate. But this is not something the anti-feminists want to do.

Sigh. I am bitter, bitter.

Here is my take on the Swedish study which has made Rush Limbaugh so happy. The study tries to establish whether increasing gender equality causes illness levels for men and women to equalize. What it states it found is that it does, to some extent, but that increasing gender equality is also correlated with lower health levels for both women and men.

The data the study uses is on the level of a municipality. Municipalities are very small geographical units. For each municipality, the study gathered information on measures of gender equality in occupations, in public spending, in who has positions of political power in the availability of parental leave and the relative rate of gender segregation in various industries. These measures are all about the society and tell nothing about the gender-equality of individual lives. The measures are then analyzed as possible determinants of ill health.

Ill health is measured in two ways: with sick days and with data on age-adjusted death rates for each municipality. This is the first econometric problem of the study: the death rates are from the same time period as the data on gender-equality.

It is hard to see how this year's data on gender equality could have much effect on this year's death rates. Death rates depend on past events, possibly on events over a long period of time. An even bigger problem (but related to the mortality measure) is that people don't stay in one tiny municipality all their lives and the one they die in may not be the one in which they lived most of their lives, and that one would be the place where gender equality might have had an influence. The researchers had no data on migration patterns, however, so they had no way of checking that possibility out.

The study includes some variables that are controlled for because they are known to have an effect on illness, separately from what the researchers want to study. These are the age distribution of the municipality, the average education level and some variables relating to income and labor market conditions.

Here comes my second major econometric problem with the study: These confounding extra variables do not include the rate of urbanization. Some of those municipalities are in Stockholm, some are rural places in the north. In general urban areas have lower health levels because of pollution and urban stress and the related lifestyle aspects. But note that urban areas are also going to have greater levels of gender equality, or so I would suspect. Omitting the urbanization variable doesn't allow us to disentangle the two effects. What it does instead is make the urbanization effects work through the gender equality measures.

Perhaps adding the urbanization measure wouldn't change the results. There is no way of knowing, but it sure would be interesting to check.

I find it odd that one would expect gender equality to differ in its effects on such a low level of measurement as a municipality. Most laws and customs in Sweden are country-level ones and already very egalitarian, and although recent immigrants might have different sex role expectations the study did control for that among the confounding variables.