Paul Goldsmith-Pinkham
@paulgp.com
16K followers 1.5K following 2.8K posts

Yale SOM professor & Bulls fan. I study consumer finance, and econometrics is a big part of my research identity. He/him/his

Economics 47%
Mathematics 9%
Posts Media Videos Starter Packs

paulgp.com
Feel like there has always been weird hints but the last 5 years have seen a strong shift towards stuff like this

paulgp.com
Yes you and Kevin are a great example of folks diving deeply into them (heck even starting your own company!!)

paulgp.com
Yes that’s a great example!

paulgp.com
I think I agree, but it’s really useful for me to keep abreast of how these things are changing — helps me keep track of how “AI” might actually work..

paulgp.com
For example, are they working with agents, are they using claude code (or codex, or any other command line tool), are they building out their own MCPs or trying out others, are they trying running models locally, are they tuning their own LLMs, etc.

paulgp.com
Extremely curious how many AI researchers in econ are actually using cutting edge AI tools and not just chatbot web interfaces?
yewonbyun.bsky.social
💡Can we trust synthetic data for statistical inference?

We show that synthetic data (e.g., LLM simulations) can significantly improve the performance of inference tasks. The key intuition lies in the interactions between the moment residuals of synthetic data and those of real data
itsafronomics.bsky.social
Good MORNING #bluesky!!!!!!! Catch me on CBS News Boston PRIMETIME live in studio today online. Thank you Courtney Cole for featuring me on the show and for such a warm welcome.

If you haven't already, make sure to grab your copy of #TheDoubleTax: annagifty.com! #blacksky #news #booksky

paulgp.com
Thanks for the link, this is quite interesting.

paulgp.com
This is quite remarkable. From a monetary standpoint, I interpret this as saying "labor market is much less weak than expected -- we should be tightening to deal with inflation."
peark.es
Absolute banger of a blog post from Dallas Fed economist Anton Cheremukhin who estimates breakeven payrolls have fallen to *30k* thanks to rapid declines in population growth.
www.dallasfed.org/research/eco...
Chart showing estimates of US population growth at monthly frequencies from 2017-2025 from BLS, BEA/Census, CBO, and author's calculations Estimated monthly contributions to monthly breakeven payroll growth across population, changes in institutionalization/civilian status, and labor force participation 2017 - 2025 Estimated monthly breakeven payrolls rate from 2022-2025
peark.es
Absolute banger of a blog post from Dallas Fed economist Anton Cheremukhin who estimates breakeven payrolls have fallen to *30k* thanks to rapid declines in population growth.
www.dallasfed.org/research/eco...
Chart showing estimates of US population growth at monthly frequencies from 2017-2025 from BLS, BEA/Census, CBO, and author's calculations Estimated monthly contributions to monthly breakeven payroll growth across population, changes in institutionalization/civilian status, and labor force participation 2017 - 2025 Estimated monthly breakeven payrolls rate from 2022-2025

paulgp.com
Added Finance job decomposition, and Federal Reserve Bank/Board jobs (zero so far this year...)

paulgp.com
Added Fed analysis (also FDIC and OCC) -- you're right that there's nothing. And that's way below the usual level. paulgp.com/2025/10/08/j...
Economics Job Market Tracker: Visualizing JOE Posting Trends 2015-2025
paulgp.com

paulgp.com
I’ll try to look. This is a great idea

paulgp.com
You’re ready for the Times editorial team!!
paulgp.com
Posted a short blog post with updated data (and public repo with data) of current state of Econ job market:

paulgp.com/2025/10/08/j...
vgel.me
if you're interesting in gaining a better intuition for how llms behave at inference time, you should try logitloom🌱, the open-source tool i made for exploring token trajectory trees (aka looming) on base and instruct models! more info in thread

🌱 vgel.me/logitloom
💻 github.com/vgel/logitloom
jackjenkins.me
Gonna be thinking about this lede for a minute.
(RNS) — Last month, the Rev. David Black stood in front of a Chicago-area U.S. Immigration and Customs Enforcement facility and spread his arms wide. Adorned in all black and wearing a clerical collar, the pastor looked up at a group of masked, heavily armed ICE agents on the roof and began to pray.

“I invited them to repentance,” Black, a minister in the Presbyterian Church (USA), said in an interview. “I basically offered an altar call. I invited them to come and receive that salvation, and be part of the kingdom that is coming.”

But when Black began to lower his arms a few seconds later, the agents responded to his spiritual plea by firing pepper balls, or chemical agents that cause eye irritation and respiratory distress, video footage shows. One struck Black in the head, exploding into a puff of white pepper smoke and forcing him to his knees. Fellow demonstrators rushed to his aid, and as the pastor rubbed his face in pain, the agents continued to fire.

“We could hear them laughing,” Black said.
paulisci.bsky.social
A Brief History of Men are Becoming Less Manly

🧵

paulgp.com
I just don't think of that as counterfactually changing race. It counterfactually changes how the *reader* of the resume perceives race, which is very manipulable.

paulgp.com
I don’t love counter factual statements that don’t have a practical manipulation associated with them (eg I think race counter factual is not a reasonable idea à la Holland 2005)
seema.bsky.social
We just spent 6 months to add 1 figure to this paper. Some people said, "Couples aren't prioritizing men's careers. Men just have better earnings opportunities when moving."

Earnings effects of moves for couples on the left, singles on the right. Negligible gap between single men and women.
Event study coefficients that show that men's earnings rise more than women's among couples following a cross-commuting zone move (left panels). The pattern is muted or reverses among single men and single women (right panels).

Reposted by Joshua Goodman

paulgp.com
Working on writing up a short note (and blog post) trying to explain the distinction between model-based and design-based identification.

Any questions regarding this that are burning for you?

paulgp.com
I was only just barely understanding Sen in college, this would’ve been over my head