Steven Glazerman

Your regular reminder that AI continues to routinely produce incorrect word salad. Sometimes having an "assistant" like this can be exhausting.

😢 when you look in the mirror and you see your own image reflected back in an NBER paper

A People-First AI Fund: $50M to support nonprofits Applications are now open for OpenAI’s People-First AI Fund, a $50M initiative supporting U.S. nonprofits advancing education, community innovation, and economic opportunity. Apply by October 8, 2025,...

Outstanding funding opportunity for U.S. based/focused nonprofits that are smallish (<$10m revenue) to advance AI education and innovation. h/t @alexnawar.bsky.social

openai.com/index/people...

Hmm. I don't think even the strongest advocates for RCTs think they are the only option.

Table shows that most authors of papers on sub-Saharan Africa in top economics journals are at American institutions, followed by the UK, Canada, and European countries. 5.2 percent are authored by researchers in African institutions.

Nice @markusgold.bsky.social and Juan Manuel Menendez post about how far empirical economics still needs to go in studying sub-Saharan Africa www.cgdev.org/blog/researc...

Motivating for @poverty-action.bsky.social's initiative promoting LMIC research collabs poverty-action.org/research-col...

Would be an honor to be on this list!

How-to guides on topics like randomization data.poverty-action.org/research-des...
Statistical power
data.poverty-action.org/research-des...
Version control with Github: data.poverty-action.org/software/git...
Coding environments. Are you Quarto-curious? data.poverty-action.org/software/qua...

One page gets you set up with software tools, like virtual environments for Python and R.
data.poverty-action.org/software/gui...

Screenshot from the IPA Knowledge Hub home page. Headers are "Research Design", "Research Ethics", "Data Quality", "Data Collection", "Data Cleaning", and "Software Guides"

Really useful resource for RAs and others interested in doing field work in international settings, or any rigorous empirical work in social sciences
@poverty-action.bsky.social just launched the IPA Knowledge Hub
data.poverty-action.org

The private sector can’t replace official statistics—but could be a great partner After President Donald Trump fired Bureau of Labor Statistics commissioner Erika McEntarfer, claiming that jobs data were “faked” and “rigged,” an urgent question arose: Could private sector data fill...

How can survey researchers and survey methodologists step in to shore up faith in official statistics? Unclear, but this blog by @jedkolko.bsky.social brings useful commentary about private sector's role in public data

www.piie.com/blogs/realti...

Anybody else out there in an NGO spending thousands of dollars to prove to funders and auditors that you’re not wasting hundreds of dollars?

10 Things to Know About Adaptive Experimental Design buff.ly/Klwp446 #EGAPMethodsGuide

Yesterday, June 2, was an interesting day at work. I was reporting a story about how the Education Department hadn't delivered a report to Congress on the Condition of Education, despite the June 1 deadline mandated by law. 🧵 (1/9)

This is odd because the same bill has marriage penalty provisions like the Child Tax Credit that deny credits if one of the parents does not have a SSN. If you have a child with a partner on a visa (legal), the bill provides disincentive to marry.

Is anyone on #EconTwitter or another social science aware of any evaluations of Promise Pay (in its use as a utility debt relief app--I know there can be other uses)? I'd appreciate any leads. Thanks!

Academics: how many hours would you set aside for prepping a class you've taught before in different forms, but not this particular audience and format, and level of background?

Assume 32 hours of instructional time (lecture and prof-led discussion).

Should there be a neutral response option or do you mean to force polarity?

Our community must take this issue very seriously!

ideally, we'd post these publicly:
raw_no_pii.csv
cleaning.R
clean.rds
analysis.R

but destroy or archive privately:
raw.csv
remove_pii.R

(was glad to see in this thread political science apparently does better than other fields but still)

Exclusive: NSF faces radical shake-up as officials abolish its 37 divisions Changes seen as a response to presidential directives on what research to fund

Exclusive: National Science Foundation staff were told today that the agency’s 37 divisions—across all eight directorates—are being abolished and the number of programs within those divisions will be drastically reduced.

NGOs Aren't the Enemy One oddity about the recent political environment is the flaming hostility to so-called NGOs, or non-governmental organizations.

@stuartbuck.bsky.social regularly puts out very thoughtful level-headed pieces on substack about science and social impact. This is just one of his gems.

goodscience.substack.com/p/ngos-arent...

Stylized image of an eccentric billionaire holding a chainsaw over an outline of the U.S. Capitol. Credit: The New Yorker magazine

That sinking feeling when you're trying to access a really useful online repository of measurement tools and... it doesn't load. Then you remember it was USAID-funded.

www.youthpower.org/resources/me...

Talking about dolls deflects from real people who have to forgo real things like insulin or food.

Pitching austerity as a promise for a better future would be more convincing from someone who made sacrifices themself

Aggregation bias? Household measures averaged to country level and then compared seems fishy.

despite the all caps this is the correct take. The refs just wanted to go home. The NBA has turned into MMA.

Hey, some good news, ERIC lives!

Is your causal question “what is the impact of an intervention on people who will never get or want the intervention?”

Null findings are so important! ..as long as the study is well design, meaning interesting counterfactual, not a straw man. Through null finding studies we can learn tons about prevailing practice if we're as careful measuring control group experiences as treatment experiences w/interventions.

Congrats to the authors on this long-term followup of a famous trial that has influenced thinking on ed-tech for a while.

New working paper: Could AI leapfrog the web?

Only 37% of sub-Saharan Africans use the internet. Cost is the #1 constraint. In a study with 469 teachers in Sierra Leone, we find AI works better than the web--and is 87% cheaper.

arxiv.org/abs/2502.12397

Let's dive in... 🧵(1/X)