Some suggestive heterogeneity masked by aggregate results of no effects on marriage: those without a partner in the household at baseline may have been more likely to stay unpartnered, and those with a partner might be more likely to be in a relationship. Cash enables choice. 8/

What didn’t change much? Generally, household composition, partnership/cohabitation, decision-making power, or overall division of labor. Among partnered recipients, there was suggestive evidence of greater trust and improved relationship quality. 7/

Larger households also seemed to have larger increases in expenditures inside the household. These results are consistent with direct expenditures on others such as purchases for children observed in Krause et al. (2026). 3/

What happens when someone receives $1,000/month for 3 years? The effects don't stop with the recipient. In this new NBER working paper, we examine what changed for partners, households, friends & family after 1,000 people received $1,000/month with 2,000 in a control group. 🧵1/

Basically: we are flying blind and just hoping everything works out, and not only works out, but goes as well as possible. I don't think people are taking that seriously enough yet. I signed this statement because it's important to build common knowledge.

I recently presented some results that showed that using LLM forecasts in experimental design could dramatically improve power. Studies in our data start severely under-powered (~0.4). Using LLM forecasts would bring it up to ~0.7! And further improvements coming.

Did you ever wish you could get help with your power calculations? Now you can! earlyreview.ai will give you estimates of treatment effects based on your early project documents (e.g., pre-analysis plans, registered reports, grant proposals, etc.).

I've built a new tool! You can upload your pre-analysis plan or registered report, pre-submission to a registry or journal, and it will screen it for completeness, clarity, and consistency. 1/ 🧵

Our panelists are great! We have a paid forecaster panel that takes the majority of the surveys posted on the platform. And they do very well in comparison to other users.

Interestingly, high self-reported confidence is associated with lower accuracy. This is dissimilar to most of the literature. In our setting, we can track individual forecasters over time. And thus we can observe: this result is driven by overconfident forecasters.

First result: forecasters tend to overestimate treatment effects - but there is a lot of signal in the forecasts made. This means that forecasts can be informative in power calculations or determining which interventions to trial.

🚨 New working paper! How well do people predict the results of studies? @sdellavi.bsky.social and I leverage data from the first 100 studies to have been posted on the SSPP, containing 1,482 key questions, on which over 50,000 forecasts were placed. Some surprising results below.... 🧵👇

3) We previously included quarterly regression results, but we obtained some updated administrative data and made some nicer plots. For example, here is an event study plot looking at employment status. 11/

Yes, there are some negative impacts on labor supply and income excluding the transfers. People also do stuff with that money. This second new figure helps illustrate the overall effects - and what doesn't move. 10/

We've updated a paper on the 3-year, $1000/month U.S. guaranteed income study. New results, in 3 figures: 🧵 1) Subjective well-being significantly improved in the treatment group in year 1, but there were no significant differences between the treatment & control group after that. 1/

Nice quote! One of the reasons for the Social Science Prediction Platform. A few of the cash transfer evaluations (including our own) collected ex ante forecasts there. @sdellavi.bsky.social

Niche, but did you know there is a TTC shop? This is such a perfect fit for my interests (maps, public transit, puzzles) that I have to share. Probably other transit systems make them, too?

We also see no effects on food insecurity or non-parental care. Here is an overall summary figure of the main index measures. We can reject even small changes for most outcomes: 15/

If anything, conditional on a positive pregnancy test, there might be an increase in abortion - but it should be emphasized that this estimate is conditional (i.e., no longer causal) and does not survive false discovery rate corrections. 11/

If you wanted to tell a negative story, you could worry about there maybe being some declines in math (also insignificant after adjusting for multiple hypothesis testing). Is math potentially a bellwether, compared to other subjects? 8/

If you wanted to tell a positive story, you could point to the (insignificant) increases in post-secondary enrollment for the small set of children for which we can observe these outcomes and hope that when more children age into it maybe it will be significant. 7/

🚨 New NBER working paper: "The Impact of Unconditional Cash Transfers on Parenting and Children" This paper estimates the effects of receiving a $1,000/month guaranteed income for 3 years, compared to a control group receiving $50/month, on children and parents in the US. 1/

Good day to dust off this old classic:

Quite the quote.

Apparently the language thing is not a joke. 😬 Should have called them goods surpluses rather than trade deficits.

This figure should never happen.

The Social Science Prediction Platform has a new feature: a leaderboard showcasing those who provided the most forecasts or were the most accurate. Forecasters can opt into displaying their name. You can also see your own rank if logged in. Check it out! @socscipredict.bsky.social

We found that people over-update on research findings with large confidence intervals and under-update on research findings with small confidence intervals. 2/

If anything, there may be a bit of churn in relationships with people outside the household, but this is not significant after accounting for multiple hypothesis testing. 44/