either I Hope You Like Matrices or You Better Love Calculus one of the two

Nobody D’s when there is a special election in a Trump district:

--only-skip-permissions-if-claude-feels-like-it

"Even under conservative assumptions, PASEC's benefits exceed its costs by a comfortable margin. Counting just one successful reform in one country per decade with a very conservative learning improvement, yields a benefit-cost ratio of around 31:1." www.cgdev.org/blog/pisa-sh...

Lmao Georgists and spreadsheet libs are gonna end up pro-Mamdani while the dirtbag & populist left turn on him

In a world where many democrats have embraced pie-in-the-sky middle class tax cuts, leave it to a democratic socialist like Mamdani to seriously engage with budgeting tradeoffs and embrace an efficient broad based property tax increase

Enjoyed this story about Kansei "Tokyo Toe" Matsuzawa, a Japanese kicker who decided he wanted to play in the NFL after visiting the Coliseum as a 19 year old tourist.... Now he's hoping to sign with a team this spring 👇 www.espn.com/nfl/draft202...

Very excited for this Friday's Chamberlain Online Seminar, where we'll muse about "The Future of Econometrics." Check out this panelist lineup! Friday 3/27 at Noon ET. Come see! t.co/pE1gZCMI7a

Continuing my occasional “put a whole ounce of Angostura Bitters in, not just a dash” (see: Trinidad Sour), here is a Tradewind Negroni: 2 sweet vermouth; 1 Cointreau; 1 Angostura Bitters.

I am very excited that PyFixest 0.50.0 is on PyPi, including a new graph-based solver for demeaning that makes fixed effects estimation in PyFixest significantly faster for "sparse" fixed effects structures.

It’s not just that there is no requirement, but ASA prohibits any requirement. This is why I am no longer an ASA member and won’t serve on ASA journal editorial boards.

i probably can’t talk too much about what i’m working on at work but i’ll just say this: cutting LLMs tf apart and using them in ways no one dreamed is an underrated strategy that’s quite cost effective

I've been waiting all year to finally drop this pun in class and tomorrow is finally the day

Being sick sucks… Being sick with two young kids SUCKS

Did you know that Arnold Ventures has a standing RFP for causal research proposals related to crime and the criminal justice system? Send us your ideas! We aim to get you an answer fast (within 3 months).

This is why I'm so excited about my new R/Python package, {freestiler}. I connect to a 146M row @duckdb.org database; generate vector tiles for 2.8 million jobs in CO from a query; serve the tiles and visualize on a @maplibre.org map. All in seconds! Get started: walker-data.com/freestiler

I found my first econ history AI one shot paper! Holy shit was I not prepared for how terrible it would be. A quick thread

Big +1 to @jenniferdoleac.bsky.social above - I'd add that we over-weight the value and role of causal inference as a share of meaningful stats to provide to policymakers in decision-making. Valuable and crucial, but the lion's share of evidence is in basic descriptives v amenable to automation

It’s taken everything I have to avoid tweeting ”Through transformers, economics will achieve its destiny of becoming the field of psychohistory”

This is how you can use plotmath expressions in #tidyplots 🤓 https://jbengler.github.io/tidyplots/articles/Advanced-plotting.html#special-characters#tidyplots #rstats #dataviz #phd

My solution to the @paulgp.com AI diff in diff slop challenge is to develop economic world models based on transformers which simulate households, firms, and governments; so policy evaluation requires spinning up some counterfactuals and simulations paulgp.substack.com/p/ai-one-sho...

Good thoughts from Jen

It would be interesting to better understand why AI papers lean towards DID... ...but also (and I say this as an IV person), maybe it is a good thing that AI tries to avoid IV?

7/ In summary: automated AI procedures for empirical work are leaning even MORE on diff-in-diff -- a concerning development. If anything, my view is we are doing too much DiD nowadays for empirical applied causal work.

🧵A thread on "AI one-shot papers". This has not been a focus on Bluesky, but on Twitter/X and LinkedIn, there has been a large discussion of the ability to "one-shot" policy evaluation, which David Yanagizawa-Drott has been pursuing in an interesting way: ape.socialcatalystlab.org

tysm @aka.dad for buying the domain for the new home of this project! now go to bsky.md for the project! (may have to restart browser/flush DNS!) i am also experimenting with hosting this directly on my old Optiplex in my bedroom over Cloudflare tunnels rather than Vercel!

“We need generative models, data models, for prior checking. If you combine the likelihood and prior you get the posterior. If you combine the prior and data model, you get the prior predictive distribution!” 👨‍🍳💋

Markets have turned. S&P down 5.4% YTD — all of that loss in the last two weeks. War premium showing up fast. Watch this space.

In honor of Chuck Norris, one of Conan's best running gags ever

Here are federal employee responses, sorted by dept/agency, to the question about whether their organization's political leaders "maintain high standards of integrity." Results are abysmal. In no agency do even 20% of respondents say their top leaders have integrity

Bernie bros came for me on Muskrat's platform when I quote tweeted this Calude ad saying, "at this point, Bernie’s job seems to be lobbying for Anthropic." So let me elaborate. Bernie has been repeating, almost verbatim, Dario Amodei's talking points, for a while now, even quote tweeting him.🧵

"Your life is worse off because those in power don't like Black people," me ON #VibeCheck TODAY. 😭 I got to join the AMAZING @theferocity.bsky.social and Zach Stafford to discuss history, pop culture, and the Business of Black women. One of my favorite conversations TO DATE.

The Butlerian Jihad is an event referenced in Frank Herbert’s Dune series that describes a revolt by humans to destroy all machines that can “think like man.” But The Butlerian Jihad was never about man vs. machines. The story is about man vs. man — and I think that’s how we should view AI today

I guess the SSRN increase is linked to how Elsevier started posting pre-prints on it. I think it started with the Lancet in 2021.

Hi all, I’m hiring a post doc for next year! Heavy focus on building and designing AI for economic mobility in various domains. Please share widely if you don’t mind! utaustin.wd1.myworkdayjobs.com/en-US/UTstaf...

it is indeed very good that there are open models. their training is still often opaque; what might they be optimized for? there are enough economists and behavioral scientists in tech that I could imagine an open model that somehow maximizes a profit function #linklog

See also psyarxiv - nice dashboard! vuorre.com/psyarxiv-das...

Someone asked how the growth in working papers on SSRN compared to other websites. SSRN vs. NBER vs. ArXiv papers per month -- interesting to see the lack of "ChatGPT" response in NBER relative to Arxiv -- SSRN looks more mixed but clearly trending way up.

New post: "The Butlerian Jihad Was Never About the Machines" paulgp.com/2026/03/19/t... or paulgp.substack.com/p/the-butler...

I love this

Pretty on the nose, Claude:

arXiv📈🤖 On the Equivalence between Neyman Orthogonality and Pathwise Differentiability By Chen, Kennedy, Balakrishnan

Powell: "I have no intention of leaving the board until the investigation is well and truly over with transparency and finality"

I was just on an Anthropic webinar about using Claude Cowork with Excel and PowerPoint. There were 10,000 people on the call. I shit you not. 😳

This is quite the paper, showing elite colleges (in Chile) help low social status individuals move up in society but on balance tighten the links between social and human capital. Amazing work by @andresbafer.bsky.social, Chris Neilson, and Seth Zimmerman. (I'm jealous of Chilean data quality.)

Love this, inspiration to try again in the distant future post tenure :-)

ITS HAPPENING

Hey #rstats & #econsky: fixest v0.14.0 is out! In particular, it fixes a very annoying, false positive, warning message regarding the positive definiteness of the VCOV (when clustered). Please update! github.com/lrberge/fixest cran.r-project.org/web/packages...

Studying where AI actually helps in science by modeling four stages of research. AI excels at screening vast possibilities, but human judgment remains key for new ideas, from Ajay K. Agrawal, John McHale, and Alexander Oettl www.nber.org/papers/w34953