Prabhat Barnwal

The mass replication studies published in Nature today are insane, exemplary and an enormous pile of work to improve science. It's just so awesome how many people spent time on this in return to be 1/100 coauthors on a thing. Just outstanding 🙌🙌🙌 1/

Solar is going to turbocharge African development

@dylantmoore @ivankorolev89 @Jon_Hartley_ Totally agree

@ivankorolev89 @Jon_Hartley_ Agreed, but for some projects, I am tragically stuck with it... There's also probably something to be said for continuing to use Stata initially if it is all you've ever known and you are in the trust building/skill development phase of using coding agents.

Don't have Stata run Claude. Have Claude run Stata. I made a Claude Code skill just for this: github.com/dylantmoore/st… Instructions for Claude on how to run and write Stata code + compressed Stata documentation in the form of md files.

🚨Replication alert🚨 I'm pleased to announce that my replication of Moretti (2021) is now accepted as a comment at AER. I find ten issues in the paper. My comment focuses on two major problems; in the appendix, I document eight (relatively) minor problems. 1/

Reading Hal Varian’s 2018 paper on AI is a useful reminder: one can be wrong about the path of technology and still be right about the economics of technology. He did not foresee ChatGPT or the exact shape of the foundation-model era, but he understood early that AI would be about scarcity, firm boundaries, cloud infrastructure, data access, organizational capability, and market structure. That is a major intellectual achievement, and the paper remains one of the most insightful early economic readings of AI. sebastiangaliani.substack.com/p/hal-varian-s…

@danilaserra_eco Congratulations!!!

📣Popping in to share some happy news as I update my CV! 📣 Look out economics profession - there will be one more female FULL professor in September 2026! 🎉🍾👇

@ksangani8 @otis_reid Congrats! This is a great paper.

"Incomplete pass-through" (when costs rise 10%, prices rise <10%) is often "complete pass-through in levels" in disguise (costs rise 10 cents/unit, so prices rise 10 cents/unit). Turns out this simple pattern can explain several other features of the data.

PhD students: submit to LSE Environment Camp 2026 (London, 20–21 May). Deadline 6 March 2026. Great chance for feedback + to meet other env/energy economists. lse-environment-week.com/environment-ca… #EconTwitter #EnvEcon

Academic relaxation ladder: Undergrad: relaxes from homework by doing extracurriculars PhD student: relaxes from research by doing homework Professor: relaxes from admin by doing research

@soumitrashukla9 On point 3, the review process is also going to use the same technology that production process is using. The weight would then fall more on “taste” that should not take much time either. Totally agree on 1 and 2.

Some predictions based on this thread: 1) Access to novel and hard-to-get datasets will increasingly differentiate top publications from others 2) The old boys' network and journal gatekeeping will become worse, advantaging the already privileged 3) Standards for publishing across all journals will go way up, leading to longer review times

@mattkahn1966 Give the current pace, it is not hard to imagine even sending humanoid bots on Waymo to do household surveys and implement RCT interventions.

Yujung is right. In my AI and undergraduate economics paper posted in the next tweet, I argue that undergrad RAs will be redeployed to generate and collect data. The robot can't sample. The robot must be fed!

Added a lecture on Econ of AI to our Rotman PhD innovation seminar this week, joined by @ansonwhho & @DavidDuvenaud. Tons of interesting new work - looked at GDPVal, Mullainathan/Ludvig, @bfjo's bottlenecks paper, but mostly learned that there are *many* huge open questions. 1/2

The largest randomized trial of medical A.I. —Over 100,000 women in Sweden —radiologist + AI vs 2 radiologists, in follow-up —AI added led to 29% more cancer detected, 44% reduced workload, and —Less cancer dx in subsequent 2 years, and, when found, less aggressive thelancet.com/journals/lance…

Corporate lawyer, yesterday: "LLMs have increased my workload because every client thinks they're a lawyer now."

Hello! I am hiring for two positions. One based at UCSD (first review of applicants next week) and one based in Mozambique (rolling review). Please see information here: saralowes.com/resources.html

@Tyler_Menzer I think the “typical adoption rate” comes from the very same table and paper, it is asked to replicate.

@PrabhatBarnwal I like that it says "based on typical adoption rates" and then just creates a completely random binomial

Claude Code Confessions-- Opus 4.5 failed to replicate the summary stats table exactly, even after detailed step-wise instructions in the planning model. When grilled, regretfully confesses to generating a random variable. The RA we wish for, but def not the one we can trust?

econ job market candidates who work on india, check out this fully-funded 2-year post-doc from gareth nellis, a friend and former colleague and new director of penn's CASI. a wonderful opportunity, esp in a really tough job market year! deadline feb 2. casi.sas.upenn.edu/casi-postdocto…

The AEA has posted eight "Recent Developments" lectures exploring highly topical issues in economics, presented by the best scholars in the field: aeaweb.org/conference/web… Well worth a watch!

@ben_golub My PhD was ~50% procrastinating, 25% suffering from imposter syndrome, 25% being amazed by American culture. AI now accelerates each by at least 10×. Nothing will ever be the same.