🧵1/4
🚨 We have finalized our DID textbook!
You’ve run out of excuses for sketchy pre-trends and mysterious TWFE coefficients.
📘 Credible Answers to Hard Questions: Differences-in-Differences for Natural Experiments
By Chaisemartin & D’Haultfœuille.
papers.ssrn.com/sol3/papers....
Are you a PhD student or Postdoc working on topics in Empirical Economics? Please apply to our 9th Potsdam PhD Workshop 👇
⏰ Deadline to apply: June 30, 2025.
*Hidden gem in the paper for econometrics fans: Angrist and Imbens warned us about the core of this DiD problem in their LATE’94 Econometrica paper!
On this hopeful note:
If you are interested in using it for teaching, happy to share seminar slides, lecture slides, and coding assignment for students that replicates our analysis (they end up with stacked DiD code ready to use) - just write me an email to v.melentyeva@tilburguniversity.edu
Most likely, their responses to policies differ as well, which is why we argue that analyses should be conducted by age at birth
Key take-away: heterogeneity by age at birth is not just a source of problems - it is an opportunity for better understanding!
Younger mothers mostly forego earnings progression, as they stop climbing the concave career ladder at its very beginning
Older mothers tend to experience level losses, struggling to sustain the careers they had built before childbirth
When apply it to German admin data:
- it produces 30% larger post-birth earnings losses than conventional event study (!)
- motherhood effects turn out to be so different in magnitude and interpretation (losses in levels vs growth) by age at birth, that we should not in fact average them
Understanding the problem, we build a new solution: basically transfer the analysis to age-at-birth level with stacked DiD + include only close ages at birth in control group for comparability
We illustrate how the issues materialize in a very simple example and then carry on to common dynamic DiDs (aka “event studies“) -> useful for understanding and teaching
Not a problem on its own, but becomes problematic if we apply DiD to such setting: already-treated mothers end up joining the control group (definitely not what we wanted)
As mothers of different ages at birth are very different (that we have known), it means we are in a situation of staggered treatment adoption over age and heterogenous effects (that we have not realized)
Turns out that estimation of motherhood effects on women’s outcomes (aka “child penalties”) is a perfect illustration of all these issues that the new DiD literature warns us about
If you
- work on #gender inequality in the labor market
- want to finally understand what this never-ending #DiD literature is talking about
- teach recent advances in DiD
then take a look at our WP with Lukas Riedel! #EconSky @rfberlin.bsky.social
www.rfberlin.com/research/chi...
📢 We are looking for a postdoctoral researcher at the Institute for Employment Research (@iabnews.bsky.social) for our research project on the gig economy! #EconSky
Link: econjobmarket.org/positions/11...
start date: October 2025
Some German knowledge is required.
Just posted updated version of our DID textbook! We now have drafts of all chapters, including the one on general designs! Now you can tell your friends still on X that they are DID-outdated :-) Happy easter for those of you that celebrate it. papers.ssrn.com/sol3/papers....
Slowly but surely coming along: we have a new version of our working textbook on diffs in diffs!
papers.ssrn.com/sol3/papers....
Chapters 2 to 4, which cover the set up, classical DIDs, and relaxations of the parallel trends assumptions have been thoroughly revised and are now almost finished.
I really liked this idea of using a histogram as a legend in a choropleth map (since land isn't unemployed; people are), so I made a little guide to doing it with #rstats, {ggplot2}, and {patchwork}
www.andrewheiss.com/blog/2025/02...
@Antoinedeeb.bsky.social and I have new paper and Stata commands, to estimate & predict treatment-effect heterogeneity in multi-site RCTs! papers.ssrn.com/sol3/papers....
Estimators computed by multisite_varITT multisite_regITT and multisite_varLATE Stata packages, available from SSC.
Our practical guide to shift-share IV is now out in the JEP!
www.aeaweb.org/issues/793
(Ungated version: www.dropbox.com/scl/fi/el9yn...)
I put together an overview of recent developments in the literature on gender-based violence for @AEAjournals
There's enough material for 2x 1.5hr lectures. I've covered the material in labor & gender economics classes
I hope they can be helpful!
www.dropbox.com/scl/fi/ea7v9...
Recently accepted by #QJE, “The Impact of Being Denied a Wanted Abortion on Women and their Children,” by Londoño-Vélez (@jlondonovelez.bsky.social) and Saravia: doi.org/10.1093/qje/...
"The whole field of economics, or what used to be called political economy, was really founded by people who thought they should be telling other people what to do."
This recent talk by Nobel laureate David Card at @rfberlin.bsky.social, on Immigration and Minimum Wages, is magisterial.
I've decided to collect my DiD materials in a single place.
psantanna.com/did-resources
There, you will find
- 14 lectures of my comprehensive DiD course
- Shorter lectures/talks I have given on DiD
- My DiD R/Stata/Python packages
- Some DiD checklists
- DiD materials from my friends
Enjoy!
Here are the first five sets of slides:
01 Introduction: psantanna.com/DiD/01_Intro...
02 Classical 2x2 setup: psantanna.com/DiD/02_two_b...
03 Clustering issues: psantanna.com/DiD/03_Clust...
04 Functional form: psantanna.com/DiD/04_Funct...
05 Covariates: psantanna.com/DiD/05_Covar...
📢 #CallForPapers
@cepr.org and RFBerlin invite submissions for the RFBerlin-CEPR Annual #Symposium in #Labour Economics.
Keynotes by: John Friedman (Brown U. & Opportunity Insights)
and Heather Sarsons (UBC Economics & CEPR).
📆 Deadline: 31 Jan 2025
More details & submit here: ow.ly/wZGK50U923u
Job announcement for Postdoctoral Researcher in Economic History
Job announcement for Postdoctoral Researcher in Economic History
📢We are hiring 📢
We look to fill a Postdoc position in Economic History!
5-year position funded by ERC project „Entrenched: Elite Capture and Social Mobility“
jobportal.uni-koeln.de/bewerben/1964
Deadline: Jan 17, 2025
Please spread widely!
@econtribute.bsky.social
#EconSky
If you are working on economics of aging, take a look at the upcoming conference at #RFBerlin!
Economists love using linear regression to estimate treatment effects — it turns out that there are perils to this method, but also amazing perks
Come with me in this 🧵 if you want to learn about our now-published paper "Contamination Bias in Linear Regressions!"
1/ (Twitter rerun!)
We've created a starter pack for all of the economics journals on Bluesky. Please nominate other journals to join the list.
go.bsky.app/4kR21vX
Ok guys here's the point.
Diff in Diff is a model based assumption that identifies the ATT (unless you're using random timing of rollouts, but almost noone is).
It is model-based because you make assumptions to identify the OUTCOME MODEL in the *absence* of the treatment.
1/