Thread by @aspaglayan

After reading @anthlittle, @ethanbdm and @rodrikdani views about whether concerns about causal identification should be weighed more heavily than the importance of research questions, I'm left thinking we are missing a big issue in this discussion 1/

Social scientific research is not just about assessing causal claims. I know, others have already pointed to the importance of descriptive research so that we can identify what causal questions are important, 2/

And I for sure agree with that. E.g., did you know most of the global expansion of public primary schooling took place *before* democracy? This implies it's really important to study how different factors affected educ provision under autocracy (which we haven't done much of) 3/

But descriptive research of the kind that identifies important patterns than then require an explanation is not the only important non-causal form of research we should focus on. There is a lot of work related to describing *processes* we should be doing as well. E.g., 4/

What strategies do autocrats use to survive? How do unions attempt to influence policy? What are the main policies governments turn to when they want to foster nationalism? Etc., etc., etc.

LOTS of our theories make *assumptions* about these things, yet these 5/

Assumptions can be tested. Unlike in causal identification, where we are making assumptions about an unobservable counterfactual, we can actually study and identify the *processes* by which autocrats try to survive, govts try to foster nationalism, etc. 6/

Our understanding of the world would be much better if we did more of this type of descriptive work. And then we can focus on identifying if these strategies, policies, etc. actually succeed in getting actors what they want. (That's the causal identification question.) 7/

So why don't we have more of this? I *don't* think the main problem are incentives to publish. The problem starts with the training we provide. In many top political science departments today, and certainly in econ, we do not train our students to do this type of research. 8/

Students do the type of research they are taught to do, because that is what they know how to do well. They know what the standards for a good diff-in-diff/ RD/ etc. study are. But they have no clue what good research looks like outside quantitative causal inference. 9/

I had to travel from @Stanford to #IQMR at @SyracuseU to start getting this type of training -- after being told at my alma mater that good qualitative historical research is "just common sense." It's as much common sense as doing good quantitative causal inference. 10/

Bottom line: I agree w/@ethanbdm that we need to do much better in training students to use causal inference tools in *responsible* ways. But equally, we should be teaching them how to employ other research methods rigorously. Our knowledge will be so much better as a result. End