←
scott cunningham99,304 posts
scott cunningham avatar

scott cunningham

@causalinf

Economics professor paying it forward with 55 burgers, 55 fries, 55 tacos, 55 pies, 55 Cokes, 100 tater tots, 100 pizzas, 100 tenders, 100 meatballs

◷ Joined October 2011▧ 5,518 Photos and videos
scott cunningham avatar
scott cunningham@causalinf·Feb 22

I’ve been doing a series on Claude code for “quantitative social scientists” on my substack. It’s not much a series I guess as much as I am regularly documenting what I’m learning. It started mid Dec 2025 and will continue. causalinf.substack.com/s/cla…

Ethan Mollick avatar
⟳ scott cunningham retweeted
Ethan Mollick@emollick·Aug 27

I hear the story about how it took 30 years to gain productivity from electricity a lot in relation to AI (I have even told it myself) But not every technology is that way. Ford went from inventing the assembly line to full deployment in 3 years, cutting car building time by 88%

Ben Moll avatar
⟳ scott cunningham retweeted
Ben Moll@ben_moll·Aug 27

Thanks to @willmacaskill, @SamuelAlbanie, @tomcohen for taking the other side of this bet at 4:1 odds and to @andrewho03 for joining on my side. Here are the final terms we agreed on benjaminmoll.com/growth_bet/ More (within) ~7 years from now 😃

Post attachment

Would anyone like to take the other side of the following bet? I bet you (say) $25k inflation-adjusted that U.S. per capita real GDP growth will be <15% per annum in 2033 or any year before that. (15% implies a doubling time of ~5 years.) Note: I'm extremely bullish on AI and...

scott cunningham avatar
scott cunningham@causalinf·Aug 27

I used to joke that the one thing I loved about summers in texas was the panicked conversations with strangers where all we did was say to one another “can you believe how hot it is???” And they never could believe it. Whereas in Boston, it’s nice.

Post attachmentPost attachment
Giovanni Di Pietra avatar
⟳ scott cunningham retweeted
Giovanni Di Pietra@dipietra·Aug 18

Came for new DiD stuff and updates on matching; wasnt expecting that last note; chills - like old .pdf mixtape with its lyrics.

Post attachment

My new book Causal Inference: the Remix comes out in two weeks (Aug 25). It’s with Yale University Press like the mixtape had been. And it’s basically the 2nd edition too, but it has a different title — go figure. The online version is now up! mixtape.scunning.com

scott cunningham avatar
scott cunningham@causalinf·Aug 27

But the checklist is what I like a lot, and I tend to follow now religiously with my adoption of agents for research. I learned a lot of this from imbens surveys, Rubin, and Pedro Sant’Anna. Anything not useful, that’s maybe likely me.

Post attachment
scott cunningham avatar
scott cunningham@causalinf·Aug 27

And for those who know me, I also have unhinged discussion of my strong opinion of how to calculate pretrend coefficients, rooting it in a falsification argument. And I cover the Rambachan and Roth honest/credible parallel trends paper in Restud.

Post attachmentPost attachment
scott cunningham avatar
scott cunningham@causalinf·Aug 27

I then lay out a detailed decomposition of TWFE with covariates when you simply include X linearly without any interactions. This I hope will give readers a chance to think carefully of what X is doing and what a proper spec looks like.

scott cunningham avatar
scott cunningham@causalinf·Aug 27

And of course staggered diff in diff. And here I discuss a variety of the rogues gallery: bacon decomposition, Callaway and Sant’Anna, Sun and Abraham, de chaisemartin and d’hauœefille, and Borusyak, et al.

scott cunningham avatar
scott cunningham@causalinf·Aug 27

The second diff in diff chapter is what I consider to be complex designs. These are violations of parallel trends due to covariate imbalance and solutions, triple differences, compositional changes in repeated cross sections, missingness, … mixtape.scunning.com/08b-com…

scott cunningham avatar
scott cunningham@causalinf·Aug 27

And the chapter also stresses the applied intuition of what types of evidence I think are compelling features of some of my favorite did papers (and honestly any paper). I also discuss selection and treatment assignment mechanisms, and how they may mess with pre trends.

Post attachment
scott cunningham avatar
scott cunningham@causalinf·Aug 27

This chapter focuses on that, shows where the parallel trends assumption comes from using a few steps from the 2x2, as well as no anticipation and the role that the untreated comparison group plays. I also discuss event studies, and the importance of falsifications

scott cunningham avatar
scott cunningham@causalinf·Aug 27

I like other modern writers found diff-in-diff the 2x2 calculation, of “four averages and three subtractions”. Which can be estimated with four different regression specs (but I only showed three here). I argue 2x2 is the calculation, and can be measured with a specific OLS. 4/n

Post attachment
scott cunningham avatar
scott cunningham@causalinf·Aug 27

Ask ten people who have studied the new diff-in-diff or who have contributed to it how it should be taught or used, and you might get ten answers. This is my answer. Here’s the first one: fundamentals of diff-in-diff mixtape.scunning.com/08a-dif…