@WorldBank Economist. I study markets in everything. Liberalism | Humanism | Moderation
Washington, DC
Joined August 2009
- Tweets9.3K
- Following1.3K
- Followers1.8K
- Likes23.6K
Great thread below 👇 summarizing our new WP studying how cities create upward mobility for women in South Asia
The AEA has featured our recent paper on zoning reform in São Paulo. We use unusually rich block-level variation and administrative land and permits records to show that upzoning:
🏠 Increases housing supply 📈
💰Reduces prices 📉
A win for the YIMBYs!
aeaweb.org/research/zoning-r…
If you’ve got a pre-2026 top 5, pat yourself on the back. That’s basically a pre-embargo Cuban cigar.
PPML gang forever, log(y+1) believers in shambles
PPML estimator of misspecified models gives an optimal approximation & can identify average marginal effects. In the September issue, by J.M.C. Santos Silva and Rainer Winkelmann zurl.co/q2INp
The extent to which people are revealing fundamentally anti-human beliefs in AI debates is frightening.
AI is an incredible tool. But it’s just that — a tool, to be wielded by man for whatever great things we can dream up.
Thankfully the majority — mostly off Twitter — agrees.
This is an important observation. The marginal project without AI is higher quality, since the hurdle is greater. But AI also enables many ideas that were high quality but infeasible. Impact on avg qual is ambiguous. Slop might be the price we pay for new research opportunities.
Replication is solved.
I wonder if this will speed up processing times for data editors at top journals. I’ve had replication reviews take 6+ months in the past.
Somehow, I’m not holding my breath.
From my non-social-media spouse, a new paper, using LLMs to replicate ... everything.
nber.org/papers/w35782
Headlne: "Across 4,452 published replication packages for five economics journals, the workflow flags discrepancies in 3,460 articles or their appendices."
A lot of incredibly bad takes on this list but for my money one of the worst is the criminal under-ranking of True Detective S1. That’s a top 10 season of tv of all time.
Our list of the 100 Best TV Shows of the 21st Century is out now. See what came in at No. 1 here. nyti.ms/4joj3Dm
Gotta admit, Codex with Astra is really good. It doesn’t necessarily feel “smarter” than earlier 5-series models, but it just makes a lot fewer mistakes. It just does things right.
Forget about AI doom for a second. If you read one thing today, read this ⬇️. This is a day to remember Rick Rescorla and all of those who died on September 11, 2001. May their memories be a blessing.
washingtonpost.com/archive/l…
A lot of the drivel published in the Economist begins to make sense when you realize it’s just three 20-something Oxbridge PPE grads in a trenchcoat
For some time Daron Acemoglu has been the colossus of economics. Give an economist a few drinks, however, and some of them will venture their true opinions about this giant economist.com/finance-and-ec…
Jonah Rexer retweeted
Highly AI-exposed software and back office jobs in India can be displaced either by int’l buyers “reshoring” these services with AI, or by local AI adoption. In fact, AI could create scale effects, growing demand for services but shrinking jobs. Data so far support this story.
These aren’t jobs, the plot shows exports.
We find that jobs have been slowing in the Indian software / outsourcing sector, even as exports have risen.
openknowledge.worldbank.org/…
Highly AI-exposed software and back office jobs in India can be displaced either by int’l buyers “reshoring” these services with AI, or by local AI adoption. In fact, AI could create scale effects, growing demand for services but shrinking jobs. Data so far support this story.
These aren’t jobs, the plot shows exports.
We find that jobs have been slowing in the Indian software / outsourcing sector, even as exports have risen.
openknowledge.worldbank.org/…
We need more of these head-to-head, open-ended “replication” exercises where both AI and human are tasked with the same research question. My prior would be that for economics, AI agents cannot yet autonomously produce high quality original research.
Recursive Self Improvement is a possibility worth taking seriously. Our new paper adds empirical evidence to the understanding of current capabilities, avoiding both extremes — hype based on naive trend extrapolation and knee-jerk dismissal based on hand-wavy arguments. normaltech.ai/p/ai-agents-ca…
