– I teach a course on natural experiments. It’s a course on, sort of
integrating the standard approach to natural experiments, which is heavily quantitative. Statistical with emphasis
on how qualitative methods are crucial part of natural experimenting and can make them stronger. Typically in methodology
classes, for good reason, the statistics is front and
center and subject matter is in the background and
I like to reverse that. I think that one of the key insights in the physical literature
is that the validity of the design is completely dependent on the main causal assumption that a causal variable is as if random or exongenously applied to the subject in the study. But there’s basically very little guidance on how one establishes that and
I try to put that little bit on firmer ground and
bring some of the tools that sort of qualitative
methodologists have developed to think about how we can
improve natural experiment. I really enjoy the
participatory section we do where we have students come up with their own natural experiments and defend them to the class.
That’s always a lot of fun. It can be useful for students
to see how they can actually use their own deep knowledge to produce convincing natural experiments. So I think that’s helpful.