Category: Data Source
One way to benefit adolescents
Have school start later:
We examine the impact of California’s Senate Bill 328 (SB 328), the first statewide mandate requiring later school start times for middle and high schools, on adolescent sleep, mental health, and academic outcomes. Using difference-in-differences and eventstudy designs across five data sources, we find that SB 328 increased the share of students sleeping at least 8 hours per night by 13%, meeting the CDC-recommended minimum for this age group. Average mental health effects are imprecisely estimated, but boys show significant reductions in sadness, hopelessness, and suicidal ideation, and Hispanic students, who experienced the largest sleep-timing shifts, show parallel reductions in difficulty concentrating; together these patterns are consistent with a dose-response relationship between sleep improvement and mental well-being. Math and English scores in grade 8 improved by approximately 0.08–0.10 standard deviations, with the largest gains among Hispanic and economically disadvantaged students. A within-state analysis using teachers’ commute arrival times as a proxy for pre-policy school start times corroborates these findings, and shows academic gains accumulating over 2023–2025 alongside a suggestive decline in high school dropout rates. The absence of effects on chronic absenteeism rules out an attendance-driven mechanism, pointing instead to the direct cognitive benefits of aligning school schedules with adolescents’ biological rhythms.
That is from a new NBER working paper by
The Impact of AI-Generated Text on the Internet
The proliferation of AI-generated and AI-assisted text on the internet is feared to contribute to a degradation in semantic and stylistic diversity, factual accuracy, and other negative developments. We find that by mid-2025, roughly 35% of newly published websites were classified as AI-generated or AI-assisted, up from zero before ChatGPT’s launch in late 2022. We also find evidence suggesting that increases in AI-generated text on the internet bring about a decrease in semantic diversity and an increase in positive sentiment. We do not, however, find statistically significant evidence supporting the hypothesis that an increased rate of AI-generated text on the internet decreases factual accuracy or stylistic diversity. Notably, our findings diverge from public perception of AI’s impact on the internet.
That is from a new paper by Jonas Dolezal, Sawood Alam Mark Graham, and Maty Bohacek. Via Glenn Mercer.
Data centers are good
Data centers are the physical infrastructure behind cloud computing, artificial intelligence, and enterprise software. The rapid diffusion of artificial intelligence (AI) is intensifying demand for compute, accelerating investment in data centers, and raising concerns about the local economic and environmental footprint of these facilities. Their expansion creates a local policy tradeoff. A data center can bring capital investment, construction activity, and specialized employment, but it can also increase demand for electricity, land, and grid capacity. This paper studies these effects at the U.S. county level. We assemble a facility-level panel of global data centers with precise coordinates, scale metrics, and annualized revenue. We map facilities to U.S. counties and combine them with County Business Patterns, county-level IRS income, county-level house prices, and electricity prices. To address endogenous siting, we instrument for data center growth using two shift-share instruments, which leverage pre-existing proximity to InterTubes long-haul fiber nodes and the 1980 county share of U.S. urban college population as shares, and both Chinese and rest-of-the-world data center revenue growth as shifts. The IV estimates show positive effects on total employment, data-processing employment, construction employment, establishments, house prices, and electricity prices at different horizons after data center growth. We also find positive effects on tax returns, adjusted gross income, and wages, while annual payroll responds less robustly. The results suggest that data centers create measurable local activity, increase house prices, and affect local electricity markets through higher prices.
That is from a new NBER working paper by
Early evidence on school smartphone bans and mental health
The word “early” is appropriate here and is to be stressed, nonetheless I am not surprised by these results, given the relative impotence of treatment effects in so many settings:
To provide causal evidence of the effects of these bans, I rely on synthetic difference-in-difference models and the National Survey of Children’s Health (NSCH) from 2016 to 2024. Currently, there are data for only one state with two post-ban periods and two states with one post-ban period, which makes the results preliminary evidence only. The outcome variables are screentime and measures of psychological wellbeing. Overall, these early results provide no clear evidence that the school ban policy reduced screentime or improved psychological wellbeing.
That is from a recent NBER working paper by Henry Saffer.
Using agents to build economic datasets
Constructing datasets from primary sources is one of the costliest tasks in empirical economics. We propose Deep Research on a Loop (DRIL), a methodology that uses AI agents to assemble datasets from publicly available sources. DRIL applies a fixed research instrument across a mapped unit space (e.g., countries by years), with a two-stage architecture separating design from implementation. The instrument specifies variables and coding rules, an evidence policy governs sources and citations, and data quality mechanisms track gaps and uncertainty explicitly. We exercise DRIL on a 2025 update of the Global Tax Expenditures Database for eight Latin American and Caribbean countries. The run produces 129 sources and 136 evidence records, covering 22 qualitative fields fully and 6 quantitative estimate types with documented gaps, at the cost of a standard LLM subscription comparable to a few hours of research-assistant work. We argue that even partial automation of dataset construction can shift the production function of empirical economics.
That is from a new NBER working paper by
USA sectoral shift fact of the day
Healthcare and Social Assistance have added nearly 1.8 million private-sector jobs in the US since the end of 2023 while all of other industries combined have lost 127,800 jobs.
Here is the source (Charlie Bilello) and a graph. In relative terms, is this good or bad for men?
The social media ban in Australia, how is it going?
In December 2025, Australia became the first country to ban youth under 16 years old from holding accounts on major social media platforms, a policy now under consideration in more than a dozen countries and in numerous states. Because social media use is inherently social, the effectiveness of a ban that is easy to circumvent may depend on whether compliance reaches a tipping point: a share of compliant peers high enough to make it optimal for individuals to comply themselves. We surveyed 835 Australian teenagers four months after the ban took effect and find that only about one in four 14–15-year-olds comply. The social environment around use has barely moved: most banned teens believe that their peers are still using banned platforms and cite social reasons for continuing use. Sustaining high compliance requires two ingredients: the share of compliers must be high enough and those who comply must find it preferable to continue complying. The current ban achieves neither. Teenagers report that they require roughly two-thirds of peers to stop using social media to stop themselves, far above the share currently complying. They also perceive compliers as less popular than non-compliers, so the more influential teens disproportionately stay on the platforms. Together, these patterns suggest that compliance is more likely to diminish than to rise. Sustaining higher compliance will likely require pairing the ban with instruments that act on social norms and individual incentives directly.
That is from a new NBER working paper by
A few days ago I was talking with a very smart fifteen year old in Australia (really). He was of the opinion that it was quite ineffective, though he noted he could no longer access LinkedIn. I would note there are more stringent measures, requiring more governmental monitoring and control of the internet, that perhaps could have a greater effect.
Sentences to ponder
Exposure increases interclass (high- and low-parent-income) marriage but has no detectable effect on interracial (White and Black) marriage. A spatial marriage market model predicts that residential segregation—one of many forms of exposure—accounts for more than one third of marital sorting by class but less than 5% by race.
That is from a new NBER working paper by
How Poverty Fell
The share of the global population living in extreme poverty fell dramatically from an estimated 36% in 1990 to 9% in 2015. We describe how this decline happened: the extent to which changes within as opposed to between cohorts contributed to poverty declines, and the key changes in the lives of households as they transitioned out of (and into) poverty. We do so using cross-sectional and panel sources that are representative or near-representative of five countries that collectively accounted for 75% of global poverty decline between 1990 and 2015. The data show that overlapping birth cohorts experienced the decline of poverty together over time, such that poverty decline can be viewed as a primarily within-cohort phenomenon. Within cohorts, the data reveal substantial churn, casting the challenge of escaping poverty as a “slippery slope” more than a long-term trap. The data also illustrate a diversity of pathways out of poverty: sectoral transitions, migration, and changing occupational choices and female labor force participation can all account for some part of poverty reduction, but in all but a handful of cases, a majority of households exiting poverty did so without experiencing these changes.
That is from a new NBER working paper by
Do Americans really hate AI?
We might be heading towards a populist backlash towards AI, but we’re not there yet. Outside the tech bubble, Americans really don’t care about AI yet.
AI is Americans’ 29th most important issue, according to the fantastic survey @davidshor ran that everyone is rightly looking at.
It’s not surprising that Americans will answer sentiment questions about AI negatively, as they’ve been negative towards tech for a while. But it’s a big leap from negative sentiment to meaningful political action.
Americans have been negative on social media for 10 years, and there has been no meaningful political action. And that’s despite all the other hallmarks of backlash people are saying about AI—violent extremists (people forget there was a shooting at YouTube HQ), protests, etc.
My prediction: we will get real populist backlash to AI when the unemployment moves by, say, 2 percentage points and people see it as caused by AI.
That is part of a longer tweet from Andy Hall.
ICE has not improved U.S. labor markets
We provide the first causal, national empirical analysis of the labor market impacts of heightened immigration enforcement during the second Trump administration. Enforcement increased everywhere, but, we take advantage of the fact that the increases have been uneven across geographic areas to classify areas as treated or control and then implement an event study and difference-in-differences design. Areas that experienced particularly large increases in the number of arrests also experienced a decrease in work among likely undocumented immigrants who remain in the U.S., compared to areas with smaller increases in arrests. We find no evidence of positive spillover effects to U.S.-born workers and U.S.-born workers who work in immigrant-heavy sectors are harmed.
That is from a new NBER working paper by Elizabeth Cox & Chloe N. East.
The best study to date on school phone bans
Schools across the U.S. have sharply restricted student use of phones during the school day. We evaluate one type of restriction—lockable phone pouches—using nationwide data combining large-scale surveys, GPS pings, standardized test scores, and school administrative records, along with sales records from the largest pouch provider. Using a staggered difference-in-differences design, we find that pouch adoption substantially reduces phone use as measured by GPS pings and teacher reports. In the first year after adoption, disciplinary incidents increase and student subjective well-being falls, consistent with short-term disruption. However, effects on well-being become positive in later years and disciplinary effects fade. For academic achievement, average effects on test scores are consistently close to zero. High schools see modest positive effects, particularly in math, while middle schools see small negative effects. We find little evidence of effects on school attendance, self-reported classroom attention, or perceived online bullying.
In sum, it is fine to want to run a school that way, but do not expect huge educational gains, if any. The evidence on this is accumulating, but many seem unable to accept the results. In any case it is not worthy of a major moral crusade.
Here is the NBER working paper, with top-tier researchers involved I might add, namely .
Trade and the End of Antiquity
What was the role of trade, and how did economic activity evolve at the End of Antiquity, when political power shifts away from the Mediterranean towards northern Europe and the Middle East? To answer those questions, we assemble a database of hundreds of thousands of ancient coins from the fourth to the tenth century, estimate a dynamic model of trade and money where coins gradually diffuse along trade routes, and recover granular regional trade and real consumption time series. Our estimates suggest that: Mediterranean trade was disrupted by the newly formed border between Islam and Christianity; economic activity shifts away from the Mediterranean starting in the fifth century; real consumption peaks in the Middle East in the eighth century; and by the end of the ninth century, Atlantic regions from Islamic Spain to Frankish northwestern Europe have become the wealthiest regions of the ancient western world.
That is from a new NBER working paper by .
The ‘Ordinary Men’ of the Nazi Party
We digitize and analyze the near-universe of National Socialist German Workers’ Party (NSDAP) membership records and link them to population and industrial censuses. Four findings emerge. First, as the party expanded, its membership came to resemble the broader population more closely in occupational, demographic, and religious terms. Second, SS members’ characteristics remained different: younger, more educated, and more fanatical, as measured by the display of Nazi insignia in membership portraits. Third, within communities, coworkers, and families, early membership generated hysteresis, with subsequent entrants drawn from the same groups. Finally, local increases in party membership are associated with subsequent deportations of Germany’s Jews.
That is from a new NBER working paper by
Do Market Reforms Cause Growth?
Do market-oriented reforms cause economic growth? This paper revisits this question using a cross-country panel of reform episodes identified from various changes in well-known economic freedom and structural reform indices. We exploit the timing of reforms using distributed-lag and event-study frameworks that trace the dynamic response of per-capita GDP. We find little evidence of immediate growth gains and some short-run adjustment costs following reform. However, growth rises gradually and persistently over time, with economically meaningful effects emerging after several years. These patterns are robust across alternative measures of reform and specifications. The results reconcile conflicting findings in the literature by showing that market reforms generate long-run growth gains despite short-run disruptions. Overall, the evidence supports the view that institutional liberalization operates through slow-moving channels that accumulate into sustained improvements in economic performance.
That is from a recent paper by Jon Hartley and Brian Wheaton.