One of my empirical-work econ takes is that I think academia should be *more* open to publishing work based on best-we-can-do weak data, as long as there's really no other way to address the question and that work is very clear about what it can and can't actually say.

There are lots of cities where driving is difficult because of heavy traffic or nutty drivers. Portland is the only city that's extremely confusing to drive in and would be equally confusing if you were the only car on the road.

This episode is interesting as it's about a research group explicitly trying dozens to hundreds of quasi experiments, progressing only with the ones that "work", backing them up with a bunch of robustness tests, and then extrapolating a lot from the results. podcastaddict.com/freakonomics...

the "gotta hand it to em" the first time around was rapid approval and aided development of a vaccine that saved hundreds of thousands of lives, and this time it's "uhhh i guess they got rid of the penny"

As a guy whose name comes up when you search "many-analyst" now I've gotten quite a few referee requests now for papers where the "many analysts" are AI bots, or undergrads, or random people, rather than professional researchers.

For the second time since this baby has been born I am the only well person in a house of three other people. My life advice to all of you is to get a worse immune system

when you have a really good idea for a paper and a solid identification plan and access to data, it should be illegal for the data to be too noisy or small for your estimate to be any good

It is not until you have a child that you learn just how in your way it is possible for someone to stand

New paper! w/ @cocoscilab.bsky.social🧵Can large language models reason flexibly, or have they learned what reasoning looks like? We introduce a new paradigm to test this question—the riddle riddle—and find that humans and LLMs show opposite patterns of performance. 📜

i enjoyed the Bad Guys movie but man the books are *awful*

I have been doing econometrics and data science consulting in academia and the private sector for a decade, and for the first time in a while I am open to new clients. If you are looking for someone to help you with data stuff, drop me a line! nickchk.com/consulting.h...

the baby was up from 2AM-5:30AM this morning and i found this great channel that has just been jamming out these videos with some great recommendations i'd never heard of. my to-listen queue is gettin real long www.youtube.com/@birdmusiclo...

birdslibrary

youtube.com

i think there are a lot of people who would find it very hard to believe that the field of economics could produce this paper. and yet! Acemoglu is like the mainstreamest and most successful economist working today. some priors could use an update. www.nber.org/papers/w35336

Automation and Repression

Founded in 1920, the NBER is a private, non-profit, non-partisan organization dedicated to conducting economic research and to disseminating research findings among academics, public policy makers, an...

nber.org

Causal inference theory and methods are absolutely essential these days. Just ask @jpinasanchez.eurosky.social or @iainbrennan.bsky.social. Personally, my favourite causal inference book is 'The Effect' by @nickchk.com! theeffectbook.net

I now have a video up about this - if you want to see how magic intersects with 300 year old probability paradoxes, now's your chance www.youtube.com/watch?v=QeCg...

This Magic Card Sums Up a 300-Year-Old Paradox

YouTube video by Econometrics, Causality, and Coding with Dr. HK

youtube.com

Nick Huntington-Klein@nickchk.com · 5mo ago

Saw @gavinverhey.bsky.social in drag on Spell Slayers which was v fun and also learned about this card which is a great demonstration of the St. Petersburg Paradox! This creature has an average power equal to infinity, and yet is realistically much smaller than measly 10k-power Jumbo Cactuar

it's wild when i think about how much better food is now than when i was a kid. so much more variety and quality. like, in my tiny hometown there's ethiopian and korean places, that's totally new. i can get a bagged salad cheaply at the grocery store and it's not wilted garbage.

New paper with @nickchk.com Do LLMs act as repositories of causal knowledge? Short answer: No! Longer answer: www.degruyterbrill.com/document/doi...

I have my students using AI assistance for their ggplot coding in data viz this year and one thing I've noticed is that all AI chatbots are absolutely addicted to putting the x-axis labels on an angle.

In a new Methodology and Research Practice Commentary, @mkarhulahti.bsky.social, @nickchk.com, @nballou.bsky.social, and Ivan Ropovik find Zendle et al. (2023) cannot support its claims due to a flawed design doi.org/10.1525/coll...

I have a new working paper up today, "Partial Identification of Causal Effects that Vary by Setting" This looks at cases where you want to estimate the same causal effect across many settings (geographic, demographic, over time, etc.). arxiv.org/abs/2605.25483

Partial Identification of Causal Effects that Vary by Setting

The estimation of causal effects using quasiexperiments often relies on the use of unusual or serendipitous sources of exogenous variation. When the goal is estimating the same causal effects across m...

arxiv.org

See, yesterday I just thought it was plausible that 3-5 of you decided to post out of the blue about how cool Sonny Rollins is, but I guess this is how long it takes until finally someone puts "RIP" in the post

derek guy@dieworkwear.bsky.social · 2mo ago

RIP Sonny Rollins, a great jazz talent and the last surviving musician in one of the best photos of men in suits during the 20th century

This paper is now out in the Journal of Causal Inference with some updates but the same general message: LLMs were not particularly good at matching expert judgment on whether a given variable pair represented a causal relationship. www.degruyterbrill.com/document/doi...

Nick Huntington-Klein@nickchk.com · 2y ago

New working paper out today with @epiellie.bsky.social called "Do LLMs Act as Repositories of Causal Knowledge?" Can LLMs (ie ChatGPT) build for us the causal models we need to identify an effect? There are reasons to expect they could. But can they? Well, not really, no. arxiv.org/html/2412.10...

I like that "that's a steal" implies it is cheap, and "this is robbery" implies it is expensive

I have a new working paper out today, looking at the impact of educational use of ChatGPT/AI on student test scores. Importantly, this is overall student-led use (incl for cheating!), not specialized AI tutoring tools. To my surprise, I found... no effect at all! arxiv.org/abs/2605.08812

Little Impact of ChatGPT Availability on High School Student Test Score Performance

In educational settings, AI can be used as a learning aid, but can also be used to avoid schoolwork, thereby passing classes while learning little. Many existing studies on the impact of AI on educati...

arxiv.org

Got a new working paper and just noticed I left something out and now I gotta wait for the arxiv process again nooo

i admit that i have been wrong about canvas all this time. it *could* have been worse

always a complete crapshoot what a kid's gonna think when you show them a piece of media anyway we're now on roughly our thirtieth listen of Karma Police in the past three days

Remember how right after ChatGPT was released you'd see a huge dip in usage during the summer months when school let out? That didn't happen in 2025! At least not looking at mobile/web traffic or Google Trends. This graph is of Google Trends during school year vs. summer across US states.

i found a coding error in an almost-finished paper and thought "oh no oh god this is going to totally destroy the paper isn't it" and then i fixed it and it changed nothing except for one weird inconvenient result i had trouble explaining which totally went away lol

From @mkarhulahti.bsky.social, me, @nballou.bsky.social, and Ivan Ropovik, newly accepted at Collabra: Psychology. This concerns a Nature: Human Behavior publication on the impacts of a Chinese anti-gaming policy. The paper has serious flaws making its results unsupportable. osf.io/preprints/ps...