Thread by @kareem_carr

Lets clear up some things about:
- race
- social constructs
- biological constructs
- sociological causation
- biological causation
- predictive accuracy
A thread. 1/n

Let's say X predicts Y. This doesn't mean X is in any way causally related to Y. Therefore, if I say "X predicts Y", it doesn't mean I'm saying "X causes Y". So in particular, if I say race predicts a health outcome, this doesn't mean race caused that outcome. 2/n

The next thing we should talk about is causation. Sociological causation involves entities from a sociological framing of the world. Biological causation involves constructs that originate out of a biological framework. 3/n

From a biology perspective, race is not directly observable. We don't see it explicitly in the genes. From a sociology perspective, people clearly take actions based on racial assumptions and that's visible in their behavior patterns (who people marry, have kids with, and so on).

Who people have kids with is both a sociological and a biological phenomenon. It is directly observable using the tools of biology. In this way, purely sociological phenomena (such as race) can have biologically detectable outcomes on a biological construct (like ancestry). 5/n

Finally, if you're sympathetic to argument that racism causes race and that's why we see disparities, then the most reasonable interpretation of that claim is that race is part of the causal path to these disparities. If "racism" is a cause of the disparity then so is "race". 6/n

Let me repeat this because I think it's an important point. How can a racist person or system act on racial categories without first being aware of those categories on which they must act? Therefore, racism causes disparity by acting *through* the race construct. 7/n

In conclusion:
1. Saying "racism causes a health disparity" and "race predicts a health disparity" is not the same thing.

2. Saying "Race causes health disparities" shouldn't be interpreted as always meaning that the causation is biological.

3. If "racism" is predictive of a health disparity then most likely "race" will be predictive of that health disparity as well. Again, this doesn't necessarily mean there's causation and if there is causation, it doesn't necessarily mean that the causation is biological.

I know it seems like I'm being really pedantic. I'm not. I'm fun at parties. I swear. Most people don't need this level of conceptual clarity about causation but if you want to make sense of what data can teach us about these issues then you need to understand this stuff. 10/n

Hope you enjoyed this thread! Feel free to retweet and like the thread so others can enjoy it as well. I tweet a lot about the kinds of things that we can learn from data. Follow me if you want to see more of this kind of content! 11/11