A Simple Argument on Race and Class

Most civically minded people agree that race affects health (and if you have trouble with that argument, this is probably not the blog for you). More specifically, a lot of medical conditions, not to mention death, come more often or earlier or worse for people of color than they do for whites. That doesn’t necessarily mean that because Phyllis is white and Enid is black, Phyllis is healthier than Enid.  But it does mean that if you randomly select a person from the population of African-Americans and another person from the population of whites, the randomly chosen African-American has a higher probability of being sick than does the randomly chosen white person. Leaving things as general as possible, you could represent the population-level relationship like this:

When preventable differences in health result from something like race that is inherent to a person, Public Health types refer to the difference as a disparity. Racial disparities are one of the great shames of the United States. Here in Wisconsin, the infant mortality rate for African-American mothers is 2.8 times the mortality rate for white mothers–a particularly heartbreaking estimate, but just one selection from a long, long list.

Let me present another uncontroversial (I bloody well hope) model:

Class and its components, especially poverty, are incredibly strong predictors of health. In fact, class is an even stronger predictor of health outcomes than race is. Put another way, when your statistical model controls for socioeconomic status (SES), race becomes a much weaker predictor of health. So what does that mean?

Let me pose to the reader a question: Bearing in mind that the direction of each arrow represents causality, which of the following models relating the effects of race and SES on health, seems correct to you?

In Model A, SES plays the role of what epidemiologists call a confounder. If this model makes sense to you, you believe that SES distorts the relationship of race to health.  Since SES is a stronger predictor of health, Model A implies that race only looks like an important factor because of its association with class, which is the real determinant of health.

In Model B, SES plays the role of a mediator. If you choose Model B, you believe that SES is a part of the mechanism of racial health disparities. One of the most important ways that systemic racism affects people of color is that it has created multigenerational obstacles to education, good jobs, equal pay for equal work, and an infinite list of factors that together construct class. In this model, SES does not distort the relationship of race to health–it is the heart of it.

It matters which interpretation you choose because statistically, these two models look the same. I’ve sat through several lectures in the past few weeks during which it was argued that controlling for SES diminishes the estimated effect of race in such and such a model because SES is the most important determinant of health. This is typically followed by a hasty disclaimer that there are true race effects, they are just small, but the implication of embracing Model A is that time, money and effort are being wasted on racial disparities when addressing class disparities would make most or all of the racial disparities vanish. There is only so much public health pie, it says, and minorities are gobbling up more than their fair share. This is a serious argument being pushed by serious people, some of them very important and smart.

Here’s what’s wrong with it. If race affects health because people of color are more likely to be of lower SES than white people are*, then controlling for SES erases race’s true effect. Same numbers, different story. And moreover, one of the causal mechanisms I drew for you above makes no sense. Let’s look just at the relationship of race and SES. Which of the four models below makes sense?

Model C posits that there is no true association between race and class. One explanation consistent with Model C is that it is just a coincidence that, say, black children are more likely to be born into poverty than white children. That is theoretically possible, but if you buy that then frankly you suffer from a lack of reality testing. Another explanation would be that some third factor causes both. The premise of eugenics, for example, is that genes determine both race and class. I hope you understand why that is incorrect.

Model D suggests that SES affects race. That really doesn’t make a lot of sense. Having, say, an income above the national median, or a master’s degree, does not lead to whiteness (despite what you may have heard).

At first glance, Model E seems like an appealingly holistic and complex compromise. The problem is that if Model D is wrong, Model E is also wrong. Class does not cause race.

That leaves us with Model F, which is both logical and intuitive. One of the most important effects of systemic racism is that it limits the social and economic opportunities given to people of color. Anybody want to disagree with that?

Now that we have established that race affects SES rather than the other way around, let us return to our original competing models A and B.

It is clear that Model A, the confounding model, must be wrong, because it gets the causality of race and class backwards. Model B, the mediating model, makes perfect sense.

I have a feeling I’m going to have to repeat this: class is a very important predictor of health. My objection to the class vs race argument is not that class is unimportant, it’s that the idea that race and class are competing interests is inherently wrong.

Though I can’t see into (most) other people’s minds, I have my suspicions about why the confounding model is so widely promoted. It gives cover to people who would rather not publicly say that they don’t think racial disparities deserve priority. Instead they can say that addressing class disparities inherently addresses race disparities at the same time. But in fact the reverse is probably true. You can’t hope to address the impacts of classism for people of color without addressing racism.

Unless you are heavily invested in denying the existence of white privilege, though, the implications of Model B are cheering. It means that efforts to reduce racial disparities and efforts to reduce class disparities work together synergistically. There is plenty of pie for everyone. Or there would be if we lived in a country that invested sensibly in public health, but we’ll get to that one another time.

Edit: Doubtless some people will argue that race and class end up as covariates in the same model because race confounds the relationship to class, not the other way around. If so, you’re treating race as a nuisance variable to be adjusted away. It’s your prerogative to do so, but come out and admit that you don’t want to know how race affects health. Otherwise, assume that race modifies the effect of class until proven otherwise.

*That’s actually not the whole story. Race remains a predictor of health effects independent of SES. That infant mortality piece I linked to above? Not only does the race disparity persist after controlling for SES, it is even stronger in mothers with the most education.


Your turn...

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s