iiinova — where your digital self keeps evolving

8/ There’s still a lot we don’t know about how or why AI is affecting early-career workers. Following new grads over time adds useful context, helps reconcile recent AI findings, and confirms a weak labor market that’s been widely described but often hard to see in aggregate data. Thoughts welcome!

120

7/ The broad picture is a weak early-career labor market for highly AI-exposed graduates: lower initial employment and earnings, movement into lower-paying sectors, and more job switching. For the most exposed majors, these differences persist beyond the first year.

6/ We find little evidence that the results are driven by changes in remote work, the changing supply of new graduates, or our choice of AI exposure measure. Graduate school and self-employment may explain some of the employment decline, but not most of it.

120

5/ Job switching also changed after ChatGPT. Pre-ChatGPT graduates became less likely to switch jobs, while post-ChatGPT graduates were comparatively more likely to switch—possibly as they tried to improve on weaker initial outcomes.

4/ Persistence also differs sharply by exposure. Outside the top decile, effects are largely transitory. But among the most exposed graduates, substantial employment and earnings differences remain visible for at least two years.

3/ A major part of the earnings decline - about half - comes from reallocation of where graduates work. Highly exposed grads shifted away from higher-paying sectors like Information and Prof./Sci/Tech. Services, and toward lower-paying sectors like Retail and Accommodation and Food Services.

2/ Among graduates in the top exposure decile: • Initial employment fell ~5 percentage points • Full-quarter earnings fell ~13% The earnings loss is comparable in magnitude for these graduates to estimates associated with graduating into a large recession.

1/ Employment and earnings for graduates from more AI-exposed majors began to diverge immediately after the release of ChatGPT in Nov 2022. Effects appear across much of the exposure distribution, but are largest for the top decile—mostly computer and software-related majors.

Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors

#EconSky 📈📉 Excited to share a new paper w/ @cody-orr.bsky.social & Lee Tucker: “Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors” www.census.gov/library/work... How did new grads fare after ChatGPT? We compare labor mkt outcomes of college majors by AI exposure. 🧵:

census.gov

The most AI-exposed new college grads have initial employment declines of 5pp and earnings declines of 13%, partially driven by shifts to lower-paying sectors.

152

Reposted by Larry Warren

NBER @nber.org · 10/08/2026

Open call for proposals, Grants to Support Data for Economic Measurement Research Projects. Submit proposals by 11:59pm EDT on September 30, 2026. More information: www.nber.org/calls-papers...

089

Reposted by Larry Warren

Alexander Berger @albrgr.bsky.social · 21/07/2026

New CBO paper: every $1 of federal R&D returns ~$3–$4 of GDP in present value over 30 years www.nber.org/papers/w35472

27945

Reposted by Larry Warren

The Review of Economic Studies (REStud) @reveconstudies.bsky.social · 28/04/2026

Manager pay and pay inequality have increased substantially since the 1980s. Manager pay reflect that managers increase the firm’s productivity which in turn affects market power. New paper by Eeckhout, Bao, and De Loecker www.restud.com/manager-pay-... #REStud #EconSky

093

Reposted by Larry Warren

John Haltiwanger @jchaltiwangerum.bsky.social · 23/04/2026

New BTOS AI Supplement from Census released today. AI use for any business function varies dramatically across sectors and firm size. In AI intensive use sectors, employment-weighted AI use 60-70% -- compared to 32% economy-wide employment weighted. See www.census.gov/library/work...

census.gov

The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks

Using new Business Trends and Outlook Survey data, we find AI use prevalent in large firms and knowledge-intensive sectors; augments tasks; labor declines rare.

095

Larry Warren @lfwarren.bsky.social · 23/03/2026

We have two predoc positions opening. Work with us to build cool public data products and gain a ton of research experience using administrative data at Census!

000

Reposted by Larry Warren

Andrew Foote @afoote-econ.bsky.social · 23/03/2026

🚨 Pre-Doc Opportunity at LEHD 🚨 Do you know graduating seniors or recent graduates in economics, math or statistics? The LEHD group at Census is hiring up to 2 positions in this category. Apply following the instructions here: www.census.gov/about/census... If you have questions, DM me!

census.gov

Economic Measurement and Research Internship (EMRI) Program

The EMRI program seeks to hire interns in economics, accounting and related fields to support the nation’s growing demand for information.

176

Reposted by Larry Warren

Per Engzell @pengzell.bsky.social · 19/12/2025

All research is exploratory if you’re confused enough

524555

Reposted by Larry Warren

David Wessel @davidmwessel.bsky.social · 13/11/2025

NBER launches initiative on economic measurement. nber.org/news/nber-la...

nber.org

NBER Launches Initiative on Economic Measurement

01710

Main takeaway: UI generosity modestly raises reservation & reemployment wages, but much of action is in increased collection of UI. This take-up margin dampens wage response. Policy levers on eligibility and costs of claiming and collecting UI benefits important, and interact with benefit levels.

000

We build a directed search model with endogenous take-up. Key insight: Higher generosity draws in lower-wage claimants, dampening the average wage response. In the absence of a collection margin, the wage response is stronger.

100

From CPS: • Reemployment wage premium for UI-eligible ≈ +9% vs. ineligible during expansions. • Take-up rate jumped from ~30% to ~40%, entirely due to higher benefits — not from changes in composition of unemployed or eligibles.

100

From BAM data we find: • Cross-section elasticity of reservation wages to UI is very small • Pandemic expansions raised reservation wages by only 8–12%. • Composition shifts: Relative to pre-pandemic, collectors were younger, nonwhite, and leisure/hospitality workers.

During COVID, UI generosity increased significantly ($600/$300 weekly supplements, plus expanded eligibility). A natural experiment to measure the pass-through from benefits to wages and who collects UI.

100

How do more generous unemployment benefits affect job search, reemployment wages, and the decision to claim UI benefits? We study this using detailed admin & survey data from BAM and CPS — and a directed search model with UI and endogenous take-up.

100

AEA Journals @aeajournals.bsky.social · 09/09/2025

Hey #EconSky 📈📉 Excited to finally share this working paper with Roozbeh Hosseini and Martin Gervais. "Unemployment Insurance, Wage Pass-Through, and Endogenous Take-up" www2.census.gov/library/work... A Short 🧵:

161

Reposted by Larry Warren

Forthcoming in the AER: "Changing Income Risk across the US Skill Distribution: Evidence from a Generalized Kalman Filter" by J. Carter Braxton, Kyle Herkenhoff, Jonathan Rothbaum, and Lawrence Schmidt. www.aeaweb.org/articles?id=...

aeaweb.org

Changing Income Risk across the US Skill Distribution: Evidence from a Generalized Kalman Filter

(Forthcoming Article) - For whom has earnings risk changed, and why? We answer these questions by combining the Kalman filter and EM-algorithm to estimate persistent and temporary earnings for every i...

042

Reposted by Larry Warren

Olivier Simard-Casanova @o.simardcasanova.net · 21/08/2025

I love economic history because for instance, you discover that a bicycle bubble happened in Great Britain during the 1890s

1527

Reposted by Larry Warren

David Wiczer @davidwiczer.bsky.social · 21/08/2025

A fun/important little piece with @AtlantaFed colleagues. Our survey center asked firms how exposed their supply chain was to imports. More import-intensive firms expect to raise prices more *and* the more their competitors import,the more they'll raise. www.atlantafed.org/research/pub...

atlantafed.org

Will Tariffs Touch Off an Inflationary Impulse? Business Execs Think So.

This Policy Hub paper uses survey data to explore tariffs' potential inflationary impact and how it could spill over beyond those firms directly affected.

132

Reposted by Larry Warren

Tony Corke @matterofstats.bsky.social · 14/07/2025

ROFL = Real Outliers Fatally Lanced TLDR = Too Lazy; Didn't Research SMH = So Many Hypotheses BRB = Bayes ... Really, Bayes? IMHO = Is My Hypothesis Obvious? DND = Distribution Not Defined FAFO = Fanciful Analysis of Four Observations JC = Jointly Comical NFI = Never Forget Independence

164

Reposted by Larry Warren

John Holbein @johnholbein1.bsky.social · 10/07/2025

N = 16 workers Workers were assigned to do multiple tasks The authors don't cluster standard errors I don't think we learn much from this study.

8435

Reposted by Larry Warren

METR @metr.org · 10/07/2025

We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.

11968702993

Reposted by Larry Warren

Clément de Chaisemartin @cdechaisemartin.bsky.social · 18/04/2025

Just posted updated version of our DID textbook! We now have drafts of all chapters, including the one on general designs! Now you can tell your friends still on X that they are DID-outdated :-) Happy easter for those of you that celebrate it. papers.ssrn.com/sol3/papers....

papers.ssrn.com

Credible Answers to Hard Questions: Differences-in-Differences for Natural Experiments

This book introduces applied researchers to modern Differences-in-Differences (DID) methods, that they can use to obtain credible answers to hard causal inferen

1225678

Larry Warren @lfwarren.bsky.social · 17/03/2025

The work is mysterious and important.