Max Posch

🚨 Come work with me on the Economics of AI! 🚨 I'm recruiting a fully funded 3.5-year PhD student to study how generative AI is transforming UK labour markets! This position is a unique collab between @KingsCollegeLon and the @AISecurityInst. Deadline: 22 April

Paul and I are teaching a short course on big data and development in a Tuscan castle! (Can finally check this off the bucket list). Pls circulate to anyone who may be interested. There is some funding for students for LMICs so please don't let the cost be a barrier to applying!

🚨 Visiting Faculty position @SFU Fields: Economic History, Political Economy, Development - Summer or 1-2 semesters in 2027-28 - Up to $55k stipend + $7.5k research support - Visa + moving support ⏰ Apply by May 4 🔗👇

found something rather baffling when researching my column this week… I wanted to see if there was any evidence that AI tools were helping economists to make their research more readable. So I analysed the text of NBER working paper abstracts…

Social scientists working with materials requiring digitization can only study what machines can read. In practice, that means printed Latin-script documents from well-funded archives. In a new working paper, I show that Vision Language Models used zero-shot outperform every existing OCR system across every script evaluated, and I propose a pipeline for deploying them on new collections. I apply it to six archival collections spanning 1.8 million pages across six countries for under $1,900.

Every journal editor should read this: causalinf.substack.com/p/claude-code-…

How did the US transform from an agricultural economy to the frontier in science and innovation? A new paper published in Research Policy by a team of researchers at the @HarvardGrwthLab and @CSHVienna shows that this process was not gradual but prompted by abrupt changes clustered in the early 1920s. It was driven by the creation of the industrial research lab, which fostered large-scale, science-based, team-oriented invention. The findings suggest that new technologies, such as AI-powered digital collaboration tools, could play a similar role today by reducing coordination costs and potentially reshaping how collective invention is organized. Co-authors: Matte Hartog, @GomezLievano, @ricardo_hausman, and @FrankNeffke (1/7)

1/ Today I’m excited to share an updated paper with my colleague Friedrich Geiecke 👇👇👇 We develop an open-source platform to conduct high-quality qualitative interviews using LLMs, including voice interviews. Interviews can run GPT, Gemini, Claude, and open-weight LLMs.

Second-best project management skill I've created: Every week, for each of about 10 research projects, I built an automated skill where Claude Code pulls and integrates: -the transcripts of all meetings - all project-related emails - all discussions in the project WhatsApp groups - all new and existing documents (most of which either live in a folder of Markdown files or in a collaborative Google Doc with multiple tabs for many purposes) - any Overleaf or Tex files Then it automatically generates a 3-page project overview covering - strategic overview with a reminder of long run objectives - critical success, factor factors and operational imperatives over the next two weeks and few months - a priority to-do list for every member of the team for that week It then creates a weekly summary of the recent week's most recent discussions and activities and pastes it to a second tab, giving us a detailed documentation of all the decisions we made and everything we discussed that week (and the weekly summary for every previous week lives below that--so we have a project memory of every major decision and discussion) My next step is for it to populate everybody's calendars or reminders system with their to-do list for the week. Yes, I could try to get four different professors from different universities (for each project) and four different research assistants (for each project)to all work on the same software like Slack. Lots of organizations do this. But since they all have their own special processes and programs, this harmonizes everything. I’m training it to create a tutorial and example system for others to copy and adapt this for themselves. All part of “Claude Blattman”. Stay tuned for the link. I just need to get it polished over the next couple of weeks.

It looks like two of the best people in the profession - @HannoLustig and my dear friend and PhD advisor @RWacziarg - have decided to team up for a new econ @Substack called “The Two Cents”. Their first blog is excellent, very insightful analysis - and I bet it’s only gonna get better. Go read it and subscribe!

Hi all, I've uploaded the 2025 update to my PhD Applied Econometrics slides: ➡️ More on regression & causality ➡️ Dynamic panel data models ➡️ Streamlined diff-in-diff extensions ➡️ More on spillover effects ➡️ Results from new papers on many topics Link in the original tweet

Check out the FT cover of our research: Do markets make us moral? ft.com/content/3c3320… via @ft

"AI and Our Economic Future" New paper in preparation for the Journal of Economic Perspectives ==> accessible to a broad audience. web.stanford.edu/~chadj/AIandEc…

Nature, nurture, and next door neighbours (?)

I built a protocol using Claude Code that crawls the web twice a day for new research updates (arxiv, ssrn, etc) and release news and emails me a list and summary every morning at 8 am. Has been useful!

I've always found seminar preparation to be one of the trickiest parts of academic research. Inspired by @ben_golub and @alexolegimas, I vibe coded Dry Run -- a tool to practice presentations with an AI-powered audience that interrupts you in real-time.

On Monday I gave the "New Developments in Language Models for Economics" lecture at AEA! Link to the video and slides in the next post, and then a 🧵.

📢 Call for papers 9th Economics of Media Bias Workshop Keynotes: Ruben Durante (NUS) and Pinar Yildrim (Wharton) 🗓️ May 28th-29th 2026 📍 Berlin, Germany ⏰ Deadline: January 31st 2026 Full CfP and information on how to submit: tinyurl.com/2s3f6tar