ATT Estimation using Regression and Matching with Heterogeneous Treatment Effects
A Deep Dive into Saturated Regressions and Nonparametric Matching Approaches (with simulations!)
Act III: Regression specifications for the ATT
In this Substack, I will address several critical concepts that I believe warrant our attention, despite the fact that my thoughts may not be entirely original. After multiple attempts to refine my writing and arrange my ideas, I have decided to dive in and share my perspective on these matters, as I believe they hold significant importance. The following points will be the focus of our discussion:
The necessity for a "parameters first" principled approach in specifying regressions, as some of us may not be doing so correctly.
The implication of heterogeneous treatment effects on regression specifications, specifically the presence of both average treatment effects (ATE) and average treatment effects on the treated (ATT) within the same specification.
The potential pitfalls of recovering the ATT from a regression specification, and why matching might be a more reliable alternative.
Now, let us delve into each of these points and explore their …
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.