Institutional research & analysis

Source: NBER

RESEARCH

Working PaperMay 15, 2026

Data Centers and Local Economies in the Age of AI: A Shift--Share Approach -- by Fernando E. Alvarez, David Argente, Joyce Chow, Diana Van Patten

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...

NBER1 min read

Working PaperMay 15, 2026

Mortality Rates by Race and Ethnicity Among People with Disabilities -- by Madeline S. Helfer, Becky Staiger, Jessica Van Parys

This paper uses Medicaid claims data from 2017-2021 to measure racial/ethnic disparities in mid-life mortality among low-income adults with disabilities receiving Supplemental Security Income (SSI). We find that American Indian and Alaska Native and White SSI recipients have the highest age-adjusted mid-life mortality rates (2.9% and 2.6%, respectively), followed by Black and Hispanic recipients (2.3% and 1.9%), and then Asian recipients (1.6%). We also find differences in diagnosed chronic c...

NBER1 min read

Working PaperMay 15, 2026

Algorithmic Credentialism -- by Peter Q. Blair, Rui Guo

The paper develops a framework for evaluating credential-coded algorithmic screens under existing civil rights law. AI-powered hiring tools trained on historical data often encode and automate bachelor's degree requirements as a proxy for worker skill, producing what this paper terms algorithmic credentialism. Drawing on labor economics research on workers Skilled Through Alternative Routes (STARs), disability theory's critique of the medical model, and disparate-impact doctrine from Griggs v...

NBER1 min read

Working PaperMay 15, 2026

Organization Capital, Large Startups, and the Dearth of IPOs -- by Rüdiger Fahlenbrach, Leandro Sanz, René M. Stulz

Many startups in the 2000s have remained private after achieving large valuations, a pattern that funding availability alone cannot explain. We propose that startups relying heavily on organization capital to achieve economies of scale and network effects through digital technologies are more likely to become large private firms than exit earlier via an IPO or acquisition. Using LinkedIn data, we construct a novel measure of organization capital intensity for startups. Exploiting a legal shoc...

NBER1 min read

Working PaperMay 15, 2026

The Long-Run Effects of the Affordable Care Act: Evidence from a Partially Pre-Committed Research Design Over the COVID-19 Recession and Recovery -- by Jeffrey Clemens, Anwita Mahajan, Joseph J. Sabia

Adjustment frictions can cause the long-run effects of social insurance reforms to differ from their short-run effects. Using pre-committed extensions of event study specifications applied previously for short-run analyses, we test the hypothesis that the Affordable Care Act’s (ACA) impacts on insurance coverage and employment would increase following the substantial churn generated by the COVID-19 pandemic. Contrary to the hypothesis, the ACA’s impacts remained stable through the pandemic. L...

NBER1 min read

Working PaperMay 15, 2026

Crimes Against Campbell-Shiller -- by Itzhak Ben-David, Alex Chinco

The Campbell and Shiller (1988) log-linear approximation is widely viewed as a model-free accounting identity that always holds: in sample, in expectation, and under arbitrary subjective beliefs. None of these claims is true. The formula is far from automatic even in realized data. Many companies do not pay dividends, making the calculation ill-defined. For dividend payers, the results are not always what they seem. The formula registers buybacks and new issuance as phantom cash-flow shocks. ...

NBER1 min read

Working PaperMay 15, 2026

Deep Research on a Loop: Using AI Agents to Construct Economic Datasets -- by Santiago Afonso, Sebastian Galiani, Ramiro H. Gálvez, Raul A. Sosa

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, a...

NBER1 min read

Working PaperMay 15, 2026

Anti-Harassment Policy and the Startup Labor Market -- by Jun Chen, Song Ma, Feng Zhang

This paper examines how anti-harassment legal reforms that weaken non-disclosure agreements (NDAs) in cases of workplace sexual harassment affect startups' hiring and organizational decisions. Using a staggered difference-in-differences design and LinkedIn data on over 50,000 U.S. venture-capital-backed startups from 2014–2022, we find that NDA reforms, although intended for employee protection, reduce female hiring by about 8%, with effects concentrated among junior women, who are statistica...

NBER1 min read

Working PaperMay 15, 2026

AlphaGlass: Interpretable Characteristic-Based Portfolio Choice -- by Sebastian Bell, Ali Kakhbod, Martin Lettau, Abdolreza Nazemi

We propose AlphaGlass, an inherently interpretable machine-learning framework for constructing portfolios that directly optimize investment objectives. AlphaGlass maps stock characteristics into additive signals with sparse interactions and converts these signals into long-short portfolios through a differentiable rank-and-mask layer. This end-to-end design allows the model to optimize objectives such as the Sharpe ratio or mean-variance utility while keeping portfolio weights interpretable a...

NBER1 min read

Working PaperMay 15, 2026

Revealing Life Preferences Through LLMs -- by Omar Abdel Haq, Amitabh Chandra, Tomáš Jagelka, Erzo F.P. Luttmer, Joshua Schwartzstein

Large Language Models (LLMs) are trained on a prodigious corpus of human writing and may reveal human preferences over characteristics of life courses, such as income, longevity, and working conditions. We present OpenAI's GPT-5.4 and a broadly representative sample of Americans with pairs of life stories and ask them to choose the life they would prefer for themselves. A person's choice is better predicted by the LLM's choice than by another person’s choice over the same stories, and LLM val...

NBER1 min read

Working PaperMay 15, 2026

Breaking the Early Bell: Lessons from the First Statewide Mandate on School Start Times -- by Jialu (Gloria) Dou, Rania Gihleb, Osea Giuntella, Jakub Lonsky

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 est...

NBER1 min read

Working PaperMay 15, 2026

What Does A Grade Mean? Informativeness and Strategic Manipulation of Grading Systems -- by Joshua S. Gans, Scott Duke Kominers

When do university grades permit informative comparisons across courses, and how does transcript adjustment affect student and instructor incentives? A raw grade mixes student performance with course-specific conditions, so grade-only comparisons fail whenever course effects are large enough to reverse ability rankings at grade cutoffs. We show that full transcripts can recover comparable student signals through what we call eigengrades: course-adjusted reports that use common or externally a...

NBER1 min read