Triple difference part 3: triple differences is a design, not a falsification
Case 2: Two Biased DiDs might equal an unbiased Triple Diff
Triple Differences is a Design, not a Falsification
Wednesdays are my “free content explainer” day, and I’m trying to stick to that. So today is my third entry in the “Triple Differences” series, and frankly, I think it’s kind of the most valuable one so far of the series.
Robert Frost once criticized writing poems that didn’t rhyme as like playing tennis with the net down. And I suspect that I suffer from that too; when nothing is forcing me to be succinct, I just ramble. Trying not to ramble, I just ended up rewriting the substack three times. I woke up early this morning and decided to rewrite it a fourth time to make it simpler and more digestible.
Today’s substack has a simple goal and that’s to convince you that triple differences is not a falsification test; it’s a research design. Which means it has its own identifying assumption, and that assumption is not parallel trends; it’s parallel bias from two diff-in-diffs. I am going to walk you through some tables and a regres…
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.