Too bad so much social science shields itself from similar evaluations

All colleges and universities are re-considering (“reimagining”) the essential skills we need to be teaching and how students can demonstrate expertise during their work-lives.

But current rules and infrastructure are not ready

Another lightning strike---two teams working on the same question

Economic history & related topics: The man who woke up eager, and the discipline of working as hard as you can to get the numbers right...

demography weighs in

Andrew Gelman of course has a great blog/substack and has been talking about statistical power issues in the social sciences for a long time.

I’m a good way through sketching out a book length treatment of the work I’ve been doing on longevity (esp with Hamid Noghanibehambari).

Peter Hull asked, so...

Two teams, nearly the same question, nearly the same data—and a revealing difference in what the papers ask the evidence to do

A searchable library of the NBER Summer Institute 2026 recordings: jump straight to any talk's video and paper.

This is a new role for many of us. We have some of the tools to do it, but must develop others

monopolists everywhere, funded by NIH

ASHEcon 2026 wrapped up in Minneapolis a couple weeks ago, and I thought I’d post a few of the main questions I heard in the early scholars mentoring sessions on Sunday.

Under powered studies are a problem; a bigger problem is data monopolists.

To the extent my views on AI assistance in research are distinctive among economists, one reason is that I have a history of running something closer to a lab of PhD students.

On AI agents, grunt work, and the part of science that isn't replaceable.

ASHEcon 2026 wrapped up in Minneapolis last week, and I thought I’d post a few of the main questions I heard in the early scholars mentoring sessions on Sunday.

In this post, I consider the question of completely autonomous AI research.

I’ve outlined an idea in the past about “blaming AI” for what really are “user errors”.

Let's not abandon our evidentiary standards when asking important questions about AI impacts

The promise is cheap, fast, and many. The danger is a fire hose of plausible output that no one actually owns.

TLDR: Don’t Let Mature Papers Sit on Your Desk and Rely on Mentors

ASHEcon 2026 is wrapping up today in Minneapolis, and I thought I’d post a few of the main questions I heard in the early scholars mentoring sessions on Sunday.

Pick your poison

Lots of thoughts on AI-automation and AI assistance in research, especially whether it will lead to paper factories and autonomous policy evaluation.

A step in the Paper Factory model of semi-autonomous paper creation by Nathan Wilmers and Per Engzell caught my eye.

I thought I’d note a few ideas that I observed in a CS class this spring.

Why AI will reshape graduate training very differently across institutions

A really useful session at NBER “Applications of Artificial Intelligence in Healthcare” led by Kosali Simon here:

I’m getting ready to drive down to St Louis for the annual demographic conference (PAA) and thought I’d send out some notes I worked on last month about the split personality of demographic research.

Seamus Heaney’s complete poems, following on editions of his letters, prose, and translations, confirm the extent of his achievement.

Kathryn Paige Harden’s new book offers a defense of a kind of Scientific Calvinism that converges with the eugenicist ideals of the contemporary Far Right.

I think there is often a lot to learn from rejection— detailed comments from reviewers or editors who read your work carefully.

There are major limitations in 'interpretability' of LLMs'