A Little Matching Never Hurt Anyone (Maybe)
Even though most of you may never use matching, that doesn't mean it's not fun to learn it.
Introduction
The only estimation procedure I can think of with a worse reputation among empirical economists than instrumental variables is matching. It isn’t matching itself that is to blame so much as it is the class of methods using conditional independence but still, the apple does not fall too far from the tree. And it does make sense that the two — IV and matching — might be maligned given how highly dependent both are to prior theory, not in the economic theory sense but in the data generating processes themselves. Both IV and matching involve untestable assumptions that are best defended with models and graphical approaches, but as Card notes in his 2014 speech at Michigan, the trajectory of empirical micro was away from models, not towards them, over the last 50 years. Ironic, in a way, given difference-in-differences, without any appeal to randomization, has in its engine the parallel trends assumption that one could argue is a model based approach too. But alas, history doe…
Keep reading with a 7-day free trial
Subscribe to Scott's Substack to keep reading this post and get 7 days of free access to the full post archives.