The First and Second Waves of Difference-in-Differences

Event study and parallel trends are more recent than you think

Difference-in-differences is a tool like the tractor or an Erlenmeyer flask. It's a scientific instrument that enables researchers to make more credible inferences about the impact of policies and programs on historical events. And just like the tractor has a history, diff-in-diff has one too. In this substack I will tell you a part of that history that maybe has not been told before which is that the age of difference-in-differences started around 1984 and evolved over 40 years in two distinct waves characterized by different practices, words, pictures (or lack thereof), reasoning and assumptions.¹

The first wave of difference-in-differences (hereafter, “the first wave”) begins in 1984 when Orley Ashenfelter and David Card post an NBER working paper coining the phrase “difference-in-differences”. While others such as John Snow and Ignaz Semmelweis have been connected with this method, Ashenfelter and Card gave it its name and so for that reason and for this substack, I mark the start of the first wave in or around 1984. The technique built momentum with steady and broad adoption, climbing empirically upward in a picture you’ll see, before ebbing at a high point that lasted from 2007 to 2011.

The second wave of diff-in-diff (hereafter, “the second wave”) begins in 2011 and lasts until the end of my dataset.² The second wave is characterized both by an exponential growth in popularity and with simultaneously new descriptions of the method that had not been there before. The second wave was when researchers en masse referenced both “parallel trends” and “event study” in their studies. As I’ll show, both terms had been almost entirely absent from the first wave, but were a deafening roar in the second.

As this is a long substack, and probably not for everyone, I’ve decided to put it behind the paywall. Not everyone really enjoys reading a story about the scientific history of difference-in-differences (go figure). The intended audience for this is someone who loves the stories of modern causal inference and applied social sciences and just wants to understand where all this material came from. If you want to read this, you can subscribe to the substack for free for 7 days and then cancel before it’s over.

So buckle up as go through the story of the first and second waves of difference-in-differences using pictures, data, screenshots and video clips with a few Nobel Laureates and their advisor!