Pedro’s checklist 4c: Ashenfelter Dip, Parallel Trends but Non-zero Pre-trends

Part 1: Introduction and Theoretical Background

In this substack, I’m going to simulate a treatment assignment mechanism that I’ll be calling Ashenfelter’s Dip. I define it below. But the punchline of the substack is this:

1. This assignment mechanism will still satisfy parallel trends

2. But this assignment mechanism will also create problems with pre-trends

3. A…