Guido Imbens and Josh Angrist won the Nobel Prize in economics for two papers: their 1994 Econometrica presenting the LATE theorem, and their 1996 JASA with Don Rubin which fleshed out some of these subpopulations a bit more. And it is now well known that instrumental variables when cast in potential outcomes and allowing for heterogenous treatment effects does not identify the ATE. It identifies something called the LATE — the local average treatment effect. The LATE is the average treatment effect for a mostly invisible and unknown subpopulation called “compliers”. And who are they? Well that’s interesting.
The potential outcomes framework is well known to readers I think. For a binary treatment, D, it describes two ex ante fixed but unknown hypothetical states of the world in which there are two “potential outcomes”:
Y(1) if D=1 for person i at time t
Y(0) if D=0 for person i at time t
This is associated with Jerzy Neyman’s 1923 article, but it’s mostly associated with a ton of …
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.