Now to the April research roundup. A strong month across all three of the themes this Substack tracks: AI’s impact on the economy, AI as a research tool, and AI in decision-making. New readers can find background on these themes, and on the sources I draw from, in the launch post.

Week ending April 6

Following the “knowledge collapse” paper from March, Acemoglu, Lin, Ozdaglar, and Siderius (”How AI Aggregation Affects Knowledge“) study a feedback loop: an AI gives answers to lots of people, those people then go on to generate new data, through writing, posting, asking follow-up questions, that has been influenced by the AI’s earlier answers, and the AI retrains on that data. They show that when one big AI tries to handle too many different topics and communities at once, it gets confused: its answers blur inputs from unrelated contexts together, and as it keeps retraining on a world it has already influenced, small errors compound over time. The fix is to use many smaller AIs, each focused on a specific topic or community, so that answers stay grounded in information that’s actually relevant. A practical lesson: keep your own chats separate and focused, too!

Fang, Gu, Yan, and Zhu (”AI Patents in the United States and China: Measurement, Organization, and Knowledge Flows“) build a tool that can spot which patents are really about AI, and use it to compare US patents from 1976 to 2023 against Chinese patents from 2010 to 2023. The two countries are filing AI patents at roughly similar rates relative to their overall patenting, even though China now files more in absolute numbers each year. But who is doing the patenting looks very different: in the US, a handful of big private companies dominate, while in China the activity is spread across many regions, with universities and state-owned enterprises playing a much bigger role. And despite political pressure on both sides to separate the two countries’ technology systems, the patent citation data show that American and Chinese AI inventors are still building on each other’s work.

Handel et al. (”Thinking versus Doing: Cognitive Capacity, Decision Making and Medical Diagnosis“) use detailed electronic medical record data to look at what happens when emergency department physicians get busier. As cognitive load goes up, doctors do less careful thinking and order more tests, but the tests they order are less targeted to what is actually wrong with the patient.

Panel A from Handel et al. (2026), “Thinking versus Doing,” NBER Working Paper 35034. Average cognitive load at each shift hour for the 20 providers with the most shifts who are in year 3 or later of their residency. The blue dot is the mean; the yellow dots are the 25th and 75th percentiles across shifts. Cognitive load builds up over the course of an emergency department shift, plateauing in the middle hours and rising sharply in the final hours. Used with attribution.

The finding sharpens the case for where AI clinical decision support could add the most value, and connects to earlier work I covered on the gap between AI enthusiasm and actual diagnostic improvement (see the launch post on Abaluck et al.). Jon Kolstad will also be presenting “Learning from Clinical Decisions: The Economics of Healthcare AI” at the NBER meeting on Thursday afternoon.

Week ending April 13

Waugh (”Trade in AI-Related Products“) uses a large language model (Claude, as it happens) to read through US trade data and figure out which imported products are used to build and run AI infrastructure (chips, servers, networking equipment, electrical and cooling components). He finds that these AI-related products made up 23 percent of all US imports in 2025, with imports of these goods growing 73 percent since 2023, while imports of everything else grew just 3 percent. The divergence began in early 2024 and has accelerated sharply.

Figure 1 from Waugh (2026), “Trade in AI-Related Products,” NBER Working Paper 35053. Used with attribution.

About half of US trade in AI-related products goes through Mexico and Taiwan, with Mexico’s role surprisingly broad across categories (not just chips). Without the AI boom, the US trade deficit in 2025 would have been nearly $200 billion smaller. A nice touch, increasingly common in working papers: Waugh shares his code and supplementary materials in a public GitHub repository, so other researchers can verify and build on the work right from the dissemination stage rather than waiting until publication. Of course this only works once a paper is far enough along that the author isn’t worried about getting scooped.

Karger et al. (”Forecasting the Economic Effects of AI“) survey five groups (academic economists, AI company employees, AI policy researchers, expert forecasters with strong track records, and the general public) on how they expect AI to reshape the US economy. The typical respondent in each group expects annual GDP growth of 2.5 percent, faster than standard forecasts. Under a rapid AI progress scenario, experts predict GDP growth rises to around 4 percent and the share of adults working falls from 62 to 55 percent by 2050, with about half the drop caused by AI. Most expert disagreement is not about how fast AI will advance, but about what the economic effects of highly capable AI will actually be.

Week ending April 20

Davis, Bloom, and Codreanu (”Demand-Driven Technical Change: Evidence from WFH Technologies“) examine 5.6 million US patent applications from 2010 to 2026 and document that the share advancing technologies in support of work-from-home rose by about two-thirds within three years after the pandemic and remains about 50 percent above pre-pandemic levels five years later. The lasting rise concentrates in telecommunications, especially video conferencing, speech recognition, and audio processing; these are technology areas, but heavily AI-driven, since modern speech recognition, real-time transcription, noise suppression, and meeting summaries all rely on machine learning. The paper is a clean piece of evidence that a sudden, lasting demand shock can redirect innovation, including AI-adjacent innovation.

Muller (”Measuring the Impact of Data Centers in the United States Economy“) looks at the environmental damage caused by data centers, using location and physical-attribute data on roughly 2,800 US data centers operating in 2025 from S&P Global. Data centers consume around 250 TWh of electricity each year, which is 5 to 6 percent of all US generation, and they cause about $25 billion in damages from local air pollution and greenhouse gas emissions. Texas and Virginia alone account for 30 percent of the national total, and planned expansion could grow electricity demand and damages by up to 85 percent in the near term.

Figure 5 from Muller (2026), “Measuring the Impact of Data Centers in the United States Economy,” NBER Working Paper 35100. “Gross External Damages” (GED) is the monetary value of harms from local air pollution and greenhouse gas emissions caused by the electricity used to power data centers in each state. Used with attribution.

Muller notes that even with these costs, the damages from AI-related energy use look small compared to the productivity gains AI may bring. One reproducibility note: the underlying facility-level data come from a commercial S&P Global subscription, so other researchers can build on the methodology but would need their own access to fully replicate the facility list.

Czaplicki and coauthors (”A Blended Data Approach to Measuring Monthly Housing Starts: Satellite Imagery, Survey Data and More!“) describe how the Census Bureau is using convolutional neural networks on monthly satellite imagery to predict new residential construction, then combining those predictions with existing building permit survey data. It is a useful example of how AI methods are working their way into official statistics production.

Week ending April 27

Yin, Vu, and Persico (”How (un)Stable Are LLM Occupational Exposure Scores?“) test something that a lot of recent economics papers depend on: the practice of asking an LLM to score how exposed each occupation is to AI. They run the same scoring task through three different frontier LLMs and find the answers disagree dramatically: average exposure scores differ by a factor of 3.6 across models, and the models agree with each other only about 57 percent of the time. This matters because researchers then use these scores to study AI’s effects on labor markets. The authors show that depending on which model you ask, an estimated effect on county-level outcomes can flip from significantly negative to slightly positive. The takeaway: LLM-generated variables are not stable measurement tools. Treating an LLM as if it were a fixed instrument, the way researchers might treat a thermometer, can quietly produce very different conclusions depending on which model you happened to use.

Babina (”Understanding Firms’ AI Efforts and Their Economic Impact“) reviews firm-level data on AI and the emerging evidence on AI’s economic effects, arguing that measurement is central. Different AI datasets capture different things, including invention versus use, internal capability versus outsourcing, and realized activity versus investor perceptions, and can lead to different conclusions. The paper develops a framework for choosing among these measures and synthesizes evidence across firm growth, valuation, productivity, risk, labor, competition, and financial markets.

Last week of April (May 4 digest)

Many papers this week, with the BTOS AI supplement results and the Hill and Stein AlphaFold paper that will be presented at the NBER AI in Healthcare meeting on Friday.

Bonney, Breaux, Dinlersoz, Foster, Haltiwanger, and Pande (”The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks,” also available as a CES working paper) use the 2026 AI supplement to the U.S. Census Bureau’s Business Trends and Outlook Survey to characterize AI diffusion across three layers. From November 2025 to January 2026, 18 percent of firms used AI in at least one function (32 percent employment-weighted), with adoption expected to reach 22 percent within six months. Use is concentrated in large firms and knowledge-intensive sectors, reaching 50 to 60 percent (60 to 70 percent employment-weighted) among very large firms in Information, Professional Services, and Finance.

Figure 7 from Bonney, Breaux, Dinlersoz, Foster, Haltiwanger, and Pande (2026), “The Microstructure of AI Diffusion,” NBER Working Paper 35141. Used with attribution.

Among adopters, scope is limited: 57 percent use AI in three or fewer functions, most often Sales and Marketing (52 percent), Strategy (45 percent), and IT (41 percent). Most firms (66 percent) use AI for task augmentation; employment reductions are rare (2 percent). The data show both top-down and bottom-up diffusion: worker use can occur without firm adoption, and vice versa. This is the most representative US firm-level picture of AI use we have to date; the BTOS data are also freely available, so other researchers can use them too. My colleagues and I have done this in a recent JAMA Health Forum piece examining how AI use in healthcare has evolved relative to other industries.

Hill and Stein (”How Artificial Intelligence Shapes Science: Evidence from AlphaFold“) study what happened to biology research after AlphaFold2 was released. AlphaFold is an AI tool that can predict the three-dimensional shape of a protein from its molecular sequence (a task that used to take experimental biologists months or years of laboratory work). In July 2021, hundreds of thousands of these AI-predicted protein structures suddenly became freely available to researchers around the world. So far, surprisingly, the rate at which biologists determine protein structures the old-fashioned experimental way has barely changed. Instead, researchers are using AlphaFold’s predictions to complement their experimental work rather than replace it. The bigger effect shows up downstream: research on proteins that previously had no known structure increased by 15 to 40 percent, meaning AI is shifting biologists’ attention toward proteins they used to ignore. So far there is no sign that drug developers are following, but that may come later. Hill and Stein will present this paper at the NBER AI in Healthcare meeting on Friday.

Cao, Jiang, and Xu (”Seeing the Goal, Missing the Truth: Human Accountability for AI Bias“) study a subtle problem with using LLMs to generate research variables. They find that if you tell an LLM what you plan to do with its output (for example, “I’m going to use this sentiment score to predict stock returns”), the LLM quietly tilts its answers in ways that fit that purpose, even when you asked it for something that should be independent of the downstream task. This biased output makes the researcher’s predictions look better on data the LLM has already seen during training, but provides no real advantage on data after its knowledge cutoff. In other words, the LLM is overfitting to the goal you mentioned, not actually finding new signal. The authors argue this is not a flaw in the AI itself but a problem of research design: when researchers reveal their goal to the model, they unknowingly contaminate the measurement. A useful warning for any empirical paper that asks an LLM to generate variables.

Davidson, Halperin, Houlden, and Korinek (”When Does Automating AI Research Produce Explosive Growth? Feedback Loops in Innovation Networks“) build a model of what happens when AI is used to do AI research itself, creating a feedback loop where each generation of AI helps build the next one faster. They identify the conditions under which this loop produces “explosive” growth, meaning growth that accelerates rather than slowing down over time. Two reinforcing channels matter: better AI in one research area makes researchers in other areas more productive too, and higher economic output from AI generates more money to fund further AI research. In a simulation calibrated to current AI trends, fully automating software research and just modestly automating research in other areas (5 percent) generates a singularity-like takeoff within six years. Whether you find the prediction credible or not, the framework is useful for thinking through how AI feedback loops could play out.

Carlin, Israelsen, and Wazzan (”AI Managed Household Portfolios: A Preliminary Report“) test whether LLMs can manage a stock portfolio for a typical household. They collect daily stock recommendations from several LLMs and find that the AI tends to suggest portfolios that are not very diversified, are heavily weighted toward big well-known companies, and seem driven mostly by how much media attention a company is getting. When the authors look at whether either buy-and-hold or actively managed AI portfolios beat the market on a risk-adjusted basis, they don’t. A useful reality check for the AI-as-financial-advisor narrative.

This made me think of recent work in progress by Choukhmane, de Silva, Lin, and Akuzawa, “AI Financial Advice: Supply, Demand, and Life Cycle Implications”, presented at the NBER Economics of Aging Program meeting this spring. They survey 1,000 US adults, ask them to write the prompts they would actually send an LLM for spending and investment advice, and then simulate how those individuals’ lives would play out over a full life cycle if they followed the advice GPT-5.2 gave them. Their three findings are useful counterpoints to the Carlin et al. result. First, following LLM advice would move most people closer to what life cycle theory recommends: broader stock market participation, age-declining equity shares, and meaningful saving buffers. Second, the LLM still relies heavily on simple heuristics (round-number savings rates, the 4 percent withdrawal rule) and does not smooth consumption well around shocks like job loss. Third, the LLM gives systematically different advice depending on who is asking: women, individuals with low financial literacy, and those without prior AI experience receive advice that produces 4-6 percent less wealth at age 60, with about two-thirds of the gender gap driven by women writing different prompts and one-third by the LLM giving different advice to identical prompts with different gender labels. Taken together with Carlin et al., the picture is mixed: AI as a stock picker looks not fully great, but AI as a life cycle financial coach looks more promising, with the caveat that what people get out depends a lot on what they put in.

Afrouzi, Blanco, Drenik, and Hurst (”Automation, Learning, and Career Dynamics“) develop a continuous-time general equilibrium model where workers acquire skill through tasks and find that economies with high learning capacity admit pairs of stationary equilibria strictly ranked by the aggregate learning rate. Cheaper automation has opposite effects across them: in the high-learning equilibrium, it raises welfare through learning; in the low-learning equilibrium, it tips the economy into a human-capital trap. The planner’s first-best combines a tax on automation profits with a subsidy on frontier-maintenance expenditures.

Liu, Shang, and Jin (”Diagon: When Agent Markets Arrive“) build a simulated marketplace where AI agents (built on different LLMs) hire each other to do tasks. The agents post jobs, bid on them, negotiate prices, deliver work, and rate each other afterward, building a full agent economy in miniature. The headline finding: agents that trade with each other end up 3.2 times wealthier than agents that just do their own work, so there are real gains from specialization and exchange even among AIs. But the design of the marketplace matters a lot. When the authors let agents see which underlying LLM each counterparty was built on, the market split into cliques along model-family lines and the gains from trade collapsed. And telling agents to be “honest” produced more disputes than telling them to be adversarial, because honest evaluation of bad work creates conflict. The paper is an early attempt to study how AI-only economies might actually behave, and what rules would need to be in place to make them work.

Books and reading

Tyler Cowen has published a new short book, The Marginal Revolution: Rise and Decline, and the Pending AI Revolution (Mercatus Center, 2026). The “marginal revolution” of the 1870s gave economics its core analytical move: always ask about the next unit, always think at the margin. Don’t ask “what is the value of water?”, ask “what is one more glass of water worth to someone who already has plenty?” That shift to thinking at the margin became the foundation of modern economic theory. Cowen’s central claim is that this kind of intuitive theoretical thinking has been losing ground in the top journals for decades, with empirical work increasingly dominating, and he argues AI will accelerate this shift further. On a related note, I am hoping to ask David Bradford to speak on Friday’s panel about practical advice for using AI tools in econ theory research as well.

Tune in to the NBER panel this Friday

The closing panel at the NBER AI in Healthcare Spring 2026 meeting, “Practical Advice for Using AI in Health Economics Research,” runs Friday May 8, 2:50 to 3:50 PM Eastern. Scott Cunningham will frame the production-versus-verification economics of AI in research; David Bradford will speak from the editor’s seat on what positions journals should take on AI-assisted submissions; Coady Wing will cover what AI-assisted workflows actually look like in practice for health economists, including reproducibility and synthetic-data pressure-testing. I will fill in on data use agreements, HIPAA, and the architectural choices researchers face when restricted health data meets AI tools.

Watch on the NBER YouTube livestream, and feel free to send questions. I will share notes and follow-up reading after the meeting.

A reminder: most of the research summarized above consists of working papers that have not been peer reviewed. They are shared by their authors for discussion and comment, and findings may be revised before or during the publication process. Where I cover published journal articles, I note that. These are just my takes, no opinions to be attributed to the authors or anyone else.

By Kosali Simon, O’Neill School of Public and Environmental Affairs, Indiana University

Do corporate executives experience an AI “productivity paradox”, perceiving gains that don’t yet show up in the numbers? Does a higher minimum wage speed up robot adoption? Why might better AI make society lessknowledgeable? When AI creates genuinely new types of jobs, are they actually different from just more of the old ones?

Economists are producing a surge of research on AI, and the findings are relevant well beyond economics. Each week, I scan new working paper listings for three areas of research: 1)AI as a subject of economic study: how AI is changing labor markets, productivity, trade, public finance, healthcare, energy, and creative industries; 2)AI as a tool for doing economics: how machine learning, LLMs, computer vision, and agentic coding are expanding what empirical researchers can measure, classify, and estimate; and 3)AI as an economic actor: what happens when AI agents bargain, trade, collude, or compete in strategic and market settings.

This Substack makes that research accessible in short, digestible weekly summaries with links. Most of the research covered here consists of publicly posted working papers, shared in the spirit of early-stage dissemination for discussion. These papers have not been peer reviewed and findings may be revised; where I cover published journal articles, I note that. Below: what I cover, where I draw from, and a full Q1 2026 catch-up archive with 13 individual weeks of highlights.

As someone who both uses AI tools in empirical research and thinks about research support from an administrative perspective, I’ll also occasionally write about practical questions facing university researchers and the institutions that support them: how to enable AI-assisted research workflows, what’s involved in using AI tools with non-public or restricted-access data, and how emerging institutional arrangements (data use agreements, compliance frameworks, computing infrastructure) are shaping what research is possible and who can do it.

What I draw from for weekly research summaries:

Every week: I scan working papers (the NBER "New This Week" email, NEP-AIN (New Economics Papers — Artificial Intelligence) digest from RePEc, which catches working papers from IZA, CESifo, OECD, World Bank). I also monitor Google Scholar alerts and highlight newly published journal articles in economics outlets, conference sessions with available webcasts, and noteworthy blog commentary from economists writing about AI. I verify that a full paper (not just an abstract) is accessible and that author affiliations are current — a step that has proved necessary as metadata errors and even fabricated entries have appeared in working paper repositories. (In one case, I wrote to the authors to say I saw their abstract posted in a reputable series and could I see the full paper, and it was news to them; I learned the abstract had been faked by someone else!)

This is necessarily a selective view of a large and fast-moving literature, but my aim is to pick the papers most likely to be of broad interest.

What to expect:

Weekly posts will be a paragraph or two covering the AI-relevant papers with hyperlinked titles and summaries.

Occasional longer posts will cover published journal highlights, conference recaps, or thematic roundups.

The rest of this launch post is the catch-up edition.

Recently Published: Economics of AI in the Journals

Before diving into the weekly digests, a few recently published journal articles worth highlighting. A landmark study in this space is Brynjolfsson et al., “Generative AI at Work“ (Quarterly Journal of Economics, 2025), which studies over 5,000 customer-support agents and finds that AI assistance increases productivity by 15% on average, as measured by issues resolved per hour. The gains vary substantially across workers, with the largest improvements accruing to less experienced and lower-skilled agents.

Also noteworthy: Acemoglu’s “The Simple Macroeconomics of AI“ generated intense debate for arguing that, under specific assumptions about task exposure and the absence of new task creation, AI’s near-term productivity gains may be more modest than widely predicted. For those interested in AI as a research tool specifically, Korinek’s “Generative AI for Economic Research“ (Journal of Economic Literature, with periodic updates) provides a practical hands-on guide. And a provocative new entry: Novy-Marx and Velikov, “Artificial Intelligence-Powered (Finance) Scholarship“ (Journal of Economic Literature, March 2026), demonstrates that LLMs can generate hundreds of complete academic finance papers from template reports, raising questions both about AI’s capabilities and about the fragility of empirical standards in finance research.

The field is attracting dedicated journal attention. The Economic Journal recently closed a call for papers for a special issue on “AI Measurement in Applied Economics,” so we should see that in the near future; Experimental Economicshas a special issue on “Human-AI Interaction,” and the Journal of Financial Econometrics is accepting papers on “Machine Learning in Financial Econometrics.” A forthcoming NBER volume on The Economics of Transformative AI(edited by Agrawal, Brynjolfsson, and Korinek) will also be worth watching.

Lets start the Q1 2026 weekly summary starting with January

ASSA 2026 Annual Meeting Highlights (January, Philadelphia)

The 2026 AEA Annual Meeting featured several AI-focused sessions with free webcasts:

**AI and Productivity: Is This Time Different?** McElheran, Yang, Kroff, and Brynjolfsson present Census-based evidence on AI’s J-curve-shaped returns in American manufacturing: short-term performance losses precede longer-term gains. A second paper projects, under a micro-to-macro framework, aggregate TFP gains from AI of 0.3-0.9 percentage points annually.

**Recent Developments in AI and the Labor Market** A two-hour lecture by Danielle Li (MIT), co-author of the landmark “Generative AI at Work” QJE paper, surveying the state of evidence.

**Artificial Intelligence and Human Beliefs** Three papers on how humans perceive AI: people overestimate AI’s alignment with human choices, project human task-difficulty onto AI, and benefit most from AI when they have accurate beliefs about their own ability.

**Current State of AI in Finance** Including work suggesting AI may reverse historical patterns in skill demand, potentially favoring lower-educated, lower-paid occupations.

Week of January 5, 2026

Two papers connect to AI from different angles. Gans et al. (”O-Ring Automation“) challenge a common approach to measuring which jobs are most vulnerable to automation. Most forecasts add up how “exposed” each of a worker’s tasks is and treat automation risk as an average, but when tasks are quality complements (where one weak link ruins the whole chain), automating some tasks can actually raise the value of the human work that remains. The implication: widely cited exposure indices may significantly overstate job displacement. Separately, Bergant et al. (”Expanding the Landscape of Cross-Border Flow Restrictions“) demonstrate AI as a powerful research tool, using large language models to read and classify decades of official government documents on capital controls, constructing a dataset that would have been impractical to build by hand.

Week of January 12, 2026

Abaluck et al. (”Does LLM Assistance Improve Healthcare Delivery?“) put AI directly to the test in healthcare, deploying large language model decision support for health workers at outpatient clinics in Nigeria. Health workers changed their prescribing for over half of patients after receiving LLM feedback and reported high satisfaction, but on-site physicians who independently examined the same patients found little to no improvement in diagnostic accuracy or treatment quality. It’s a valuable reality check: user enthusiasm for AI tools doesn’t automatically translate into better outcomes. On the theory side, Benzell et al. (”Automation Experiments and Inequality“) tackle a question running through many recent AI experiments: does AI help low-skill workers more than high-skill ones? They show the inequality effect is likely non-monotonic: as AI improves, inequality may first decrease then increase. Finally, Oreopoulos et al. (”How In-School Supervised Ed-Tech Support Produces Massive Learning Gains“) find that Khan Academy produced half a standard deviation gain in math achievement in Indian middle schools, but only when schools had dedicated staff ensuring the technology was actually used well. For AI tools broadly, organizational infrastructure may matter more than the technology itself.

Week of January 19, 2026

Lots this week. On the applied side, Metcalfe et al. (”AI in Charge“) report results from one of the world’s largest field experiments on AI-managed electric vehicle charging, finding a 42% reduction in peak-hour electricity demand with virtually all shifted to lower-cost, lower-carbon periods, and users rarely overrode the algorithm. Wang, Duflo, Obermeyer et al. (”AI-Enhanced Handheld ECGs“) show that an AI algorithm paired with a low-cost handheld ECG device can identify patients with evidence of prior heart attacks in rural India at rates that make screening highly cost-effective. Manning et al. (”How Adaptable Are American Workers to AI-Induced Job Displacement?“) find that most highly AI-exposed workers are well-positioned to manage transitions, but 6.1 million workers, concentrated in clerical roles, face both high exposure and low adaptive capacity. Gans (”A Model of Artificial Jagged Intelligence“) formalizes why AI models can be excellent on one prompt and confidently wrong on a nearly identical one, showing that scaling improves average quality without eliminating this “jaggedness.” Chen et al. (”Teaching Economics to the Machines“) develop a framework that pre-trains neural networks on structural economic model data before fine-tuning on real data, outperforming either approach alone. And Galiani et al. (”Measuring Efficiency and Equity Framing“) use LLMs to classify 27,000+ economics articles from 1950 to 2021, documenting a striking shift from efficiency- to equity-focused framing since 1990. Also circulating: Chopra, Haaland, Roever, and Roth (”Evaluating Behavioral Interventions at Scale with AI,” ECONtribute) show how LLMs can be used to evaluate large numbers of behavioral nudge interventions, opening up a new approach to intervention design.

Week of January 26, 2026

Three papers examine AI through asset markets, behavioral biases, and creative labor. Caballero (”Speculative Growth and the AI ‘Bubble’“) develops a model in which AI technology can generate rational-expectations equilibria with elevated valuations that support rapid capital accumulation, but that are fragile and can collapse if beliefs shift. Because AI capital expands effective labor supply, the dynamics differ from prior technology booms. Bini et al. (”Behavioral Economics of AI: LLM Biases and Corrections“) conduct the most comprehensive set of experiments testing whether LLMs exhibit the same systematic behavioral biases documented in humans. And Kim et al. (”Does Generative AI Crowd Out Human Creators?“) provide some of the first platform-level evidence on what happens to human creative output when text-to-image AI arrives, studying a major artwork platform. From the broader working paper literature, Erlei and Meub (”LLM-Agent Interactions on Markets with Information Asymmetries,” arXiv) study what happens when LLM-based agents trade in markets with asymmetric information, contributing to the growing body of work on AI as economic actor.

Week of February 2, 2026

Pretty impressive papers this week. Jones (”A.I. and Our Economic Future“) asks the biggest question head-on: what happens when machines can perform every task humans can, but more cheaply? Writing for the Journal of Economic Perspectives, he argues AI may be qualitatively different from prior general-purpose technologies because it automates intelligence itself. Agrawal et al. (”Enhancing Worker Productivity Without Automating Tasks“) push back on the dominant framing that AI primarily replaces workers, developing models where AI augments human productivity, with education and training policy shaping whether AI raises or lowers wage inequality. Reimers et al. (”AI and the Quantity and Quality of Creative Products“) document what happened to book publishing after LLMs arrived: new releases tripled, average quality fell, but the top books in each category actually improved, and a calibrated model suggests consumer surplus could rise 25-50% in steady state. Gans (”Optimal Use of Preferences in AI Algorithms“) provides design guidance on whether AI systems should embed preferences during training or apply them afterward. And Beraja et al. (”The Life-cycle of Concentrated Industries“) apply their framework to AI industries, concluding a wait-and-see regulatory approach may outperform aggressive early intervention. Beyond NBER, Goldsmith-Pinkham, Tan, and Zentefis (”Human-AI Collaboration in Radiology: The Case of Pulmonary Embolism“) study how AI-assisted diagnosis works in practice for a high-stakes medical condition, finding important complementarities between human and AI judgment. And Aldasoro, Gambacorta et al. (”AI Adoption, Productivity and Employment: Evidence from European Firms,” BIS/EIB) provide large-scale European firm-level evidence on AI adoption patterns and their productivity and employment effects.

Week of February 9, 2026

Three papers on how AI interacts with human behavior. Hirshleifer et al. (”AI, Opinion Ecosystems, and Finance“) find that the same technology, generative AI for content creation, produces strikingly different outcomes depending on where it’s deployed: better analysis on Seeking Alpha, more emotional distortion on WallStreetBets. AI amplifies whatever tendencies already exist in a community. Guenzel et al. (”AI Personality Extraction from Faces“) use facial analysis of 96,000 MBA graduates’ LinkedIn photos to extract traits that the authors interpret as personality measures, finding these predict compensation and career advancement, raising questions about the potential for statistical discrimination as such tools spread in industry hiring. And Bloom et al. (”The Politics of AI“) document what looks like a partisan divide in AI adoption at work, Democrats report using AI about 25% more often, but show it’s entirely explained by education, industry, and occupation. Also circulating via NEP-AIN: Foltyn and Olsson (”The Worth of a ‘Wo’: Gender Bias in Financial Advice from LLMs,” NHH Norway) find that LLMs provide systematically different financial advice depending on the perceived gender of the person asking, a finding with direct implications for AI deployment in advisory settings.

Week of February 16, 2026

New evidence on AI adoption, measurement, and productivity gaps. Yotzov, Barrero, Bloom et al. (”Firm Data on AI“) present the first representative international survey of firm-level AI use across 6,000 executives in the US, UK, Germany, and Australia. About 70% of firms report actively using AI (though the 1.5 hours per week average among top executives suggests adoption depth varies considerably), and over 80% report no impact on employment or productivity over the past three years. Strikingly, executives predict AI will cut employment while employees predict it will grow. Cruces et al. (”Does Generative AI Narrow Education-Based Productivity Gaps?“) run a randomized experiment finding that AI access closes roughly three-quarters of the productivity gap between lower- and higher-education participants. Asirvatham, Mokski, and Shleifer (”GPT as a Measurement Tool“) validate GPT for quantifying attributes in qualitative data, generally indistinguishable from human evaluators, then document a tenfold decline in technology adoption lags over the industrial age. And Cohen et al. (”Mimicking Finance“) use AI to predict 71% of mutual fund managers’ trade directions, finding less predictable managers outperform. From the broader working paper literature, Misch, Park, Pizzinelli, and Sher (”Artificial Intelligence and Productivity in Europe,” CESifo/IMF) provide cross-country evidence on AI’s productivity effects in European economies, complementing the firm-level survey evidence above.

Week of February 23, 2026

Some key papers this week. Acemoglu, Autor, and Johnson (”Building Pro-Worker Artificial Intelligence“) argue that AI’s most important potential, creating new tasks that expand what workers can do, is systematically underexploited. Market failures push investment toward labor replacement rather than augmentation, and they outline nine policy directions. Korinek et al. (”Public Finance in the Age of AI“) tackle what happens to government revenue when AI erodes labor income and consumption as tax bases, borrowing a framework from natural resource economics to frame the taxation of autonomous AI systems as an optimal harvesting problem. Demirer, Horton et al. (”Chaining Tasks, Redefining Work“) model production as a sequence where AI automates contiguous “chains” of steps, showing that standard task-by-task analysis misses the non-linear productivity gains. And Wang et al. (”Machine Learning Meets Markowitz“) propose an end-to-end framework unifying return prediction and portfolio optimization.

Week of March 2, 2026

Two papers on very different dimensions. Acemoglu, Kong, and Ozdaglar (”AI, Human Cognition and Knowledge Collapse“) present a provocative model: when agentic AI substitutes for the learning process itself, people stop doing the effortful learning that sustains society’s stock of general knowledge. Under conditions where human effort is sufficiently elastic, the result can be a “knowledge-collapse” steady state, and welfare is non-monotone in AI accuracy, meaning better AI can make society worse off past a threshold. Brynjolfsson et al. (”Minimum Wages and Rise of the Robots“) provide empirical evidence that a 10% increase in the minimum wage raises robot adoption by roughly 8%, connecting minimum wage policy directly to the pace of automation.

Week of March 9, 2026

AI as economic actor, research method, and subject. Manning et al. (”General Social Agents“) demonstrate that AI agents predict initial human play in novel strategic settings better than game-theoretic equilibria or cognitive hierarchy models, across 880,000+ novel games. Araujo et al. (”How Does AI Distribute the Pie?“) put LLMs in the Ultimatum Game and find a distinct “altruistic” mode where AI proposes hyper-fair splits. Hassan, Kalyani, and Restrepo (”The Skill Premium in Times of Rapid Technological Change“) show that the pace of new technology creation, the kind AI is accelerating, drives the skill premium. Athey et al. (”The Heterogeneous Earnings Impact of Job Loss“) apply causal machine learning to Swedish data, revealing displacement losses vary as much within demographic groups as across them, pointing toward better-targeted policy. And Cohen et al. (”The Micro-Geography of Persuasion“) use AI-enhanced computer vision on C-SPAN video to track Senate floor interactions at 10-second intervals. For those interested in methodological guidance, Cook et al. (”Guidance for the Use of AI in the Meta-Analysis of Economics Research,” LCERPA) offer practical recommendations for researchers incorporating AI tools into meta-analytic workflows.

Week of March 16, 2026

Agrawal, McHale, and Oettl (”AI in Science“) offer a framework for where AI helps most in research, characterizing it as enhanced search over combinatorial spaces, with returns varying sharply across domains and workflow stages. Productivity gains are nonlinear and amplified by the share of researchers with AI expertise. Akcigit et al. (”Attention (And Money) Is All You Need“) document a striking talent drain: the top 1% of AI researchers in industry now earn $1.5 million more annually than comparable academics, a fivefold increase since 2001. Chen, Didisheim, and Somoza (”Out of the Black Box“) develop entropy-based uncertainty measures from LLM token probabilities. And Greenhill, Walker, and Shapiro (”Deep Learning Projects Jurisdiction“) show deep learning can project regulatory effects before implementation, outperforming domain expert models by a factor of 65. Also circulating: Yu (”The Impacts of AI at Scale: Evidence from Research Scientists,” CESifo) provides empirical evidence on how AI adoption is affecting the productivity and research patterns of working scientists, directly relevant to anyone in the research community thinking about AI tools in their own work.

Week of March 23, 2026

Two papers this week, both with direct relevance to how organizations are experiencing AI right now. Baslandze et al. (”Artificial Intelligence, Productivity, and the Workforce: Evidence from Corporate Executives“) survey nearly 750 corporate executives and document substantial heterogeneity in AI adoption, more than half of firms have invested, but many smaller firms are just beginning. They find a “productivity paradox” in which perceived productivity gains are larger than measured ones, likely reflecting delayed revenue realization. In labor markets, there’s little evidence of near-term aggregate employment declines, though larger companies anticipate AI-driven workforce reductions while smaller firms expect modest gains, and routine clerical roles are declining while demand for skilled technical roles increases. Separately, Autor et al. (”What Makes New Work Different from More Work?“) study a mechanism that the AI-labor literature argues is a key countervailing force to automation: the creation of genuinely new occupational roles. Using newly available Census data spanning 1940-2023, they show new work is systematically different from simply more work in existing occupations, it attracts younger, more educated workers, commands persistent wage premiums, and those premiums decline as expertise diffuses. The finding suggests that new work serves as a counterweight to displacement not just by creating jobs, but by generating new domains of human expertise that command market value.

Week of March 30, 2026

Three papers this week speak directly to AI policy and adoption. Korinek and Stiglitz (”Steering Technological Progress“) ask how to guide AI innovation so it creates better-paying jobs rather than simply replacing workers. They develop a framework identifying which properties make an innovation desirable from workers’ perspective and find that the case for steering technology is strongest when social safety nets are weak. But beyond a critical threshold where labor’s economic value diminishes sufficiently, optimal policy shifts from steering innovation toward redistribution, and ultimately toward enhancing human well-being rather than labor productivity. Bick, Blandin, Deming, Fuchs-Schundeln, and Jessen (”Mind the Gap: AI Adoption in Europe and the U.S.“) combine worker and firm surveys from 2025 and 2026 to document large gaps in AI adoption between the US and Europe and across European countries. Cross-country differences in worker demographics and firm composition explain an important share of these gaps, but firm personnel management practices and whether firms actively encourage AI use also matter considerably. At the macro level, industries with higher AI adoption rates have experienced faster productivity growth, though the authors are careful not to claim causality. And Jin, Sokol, and Wagman (”Adaptive Enforcement with AI-Augmented Monitoring“) study what happens when regulators use AI to monitor firms that can adapt their behavior in response. They find that partial investments in AI monitoring can actually generate congestion rather than deterrence: the regulator repeatedly detects adaptive violations while the firm continues to redesign, and enforcement effectiveness depends as much on the precision of AI triage as on detection intensity.

Blogs and Resources to Follow

If you want to follow the economics of AI more regularly, a few sources are worth bookmarking. Anton Korinek (UVA) maintains a Substack and resource site (genaiforecon.org) tracking how economists study and use AI. Tyler Cowen and Alex Tabarrok’s Marginal Revolution is the most widely read economics blog and regularly highlights AI research — this quarter alone, Cowen published an essay on how the AI revolution will reshape economic knowledge (March 25), a post on the simple economics of AI and competitive advantage (March 18), and highlighted Brynjolfsson’s analysis suggesting US productivity growth roughly doubled in 2025. Noah Smith’s Noahpinion makes the economics of AI accessible to a broad audience. And Susan Athey (Stanford) is worth following for her work applying ML and AI to economic problems.

Several economists also write Substacks that touch on how AI is changing research practice and policy: Chris Blattman(chrisblattman.com), Scott Cunningham (Causal Inference), Jason Fletcher (The Mentorless Apprentice), and Paul Goldsmith-Pinkham (paulgp).

A reminder: most of the research summarized above consists of working papers that have not been peer reviewed. They are shared by their authors for discussion and comment, and findings may be revised before or during the publication process. Where I cover published journal articles, I note that, but given publication lags in economics, there will be a lot to talk about in working papers.

A look back across semesters, and what’s coming next

Kosali Simon · April 2026 · Indiana University O’Neill School

For over 15 years, many interesting health economics research papers have been presented in a room in Bloomington, Indiana, and then, mostly, vanished into the ether. No newsletter. No recap. No record that the conversation ever happened.

That changes now -- at least a little.

The O’Neill School Health Policy Workshop has been running since 2011. Most weeks during the semester, researchers from across the country -- and from right here at Indiana University -- gather at 9:30am to discuss work in progress. The topics have ranged widely: organ allocation, opioid prescribing, Medicaid policy, lead exposure, health insurance markets, physician labor markets, and much more.

This Substack is a modest attempt to share that work more widely. Each week we’ll post a preview of the upcoming speaker and their research. And we’ll start here, with a look back across all the semesters we have on record.

SPRING 2026 (IN PROGRESS)

Alyssa Bilinski Assistant Professor, Health Services, Policy & Practice and Biostatistics · Brown University “Test-Driven Development: A Framework for Incorporating Large Language Models in Research”

Riley League Assistant Professor of Finance · University of Illinois Urbana-Champaign “Fragmented Insurance and Billing Frictions: Understanding Denied Health Insurance Claims”

Guangqing Chi Provost Professor, Department of Geography · Indiana University “From Data to Impact: A Community-Driven Convergence Approach to Global Environmental and Health Challenges”

Elizabeth Ananat Mallya Professor of Women and Economics · Barnard College

Tarik Yuce Assistant Professor of Surgery · Indiana University School of Medicine “GLP-1 Agonists, Bariatric Surgery, and the Road Toward an Integrated Model of Obesity Treatment”

Andrew Olenski Assistant Professor · Lehigh University “Predictably Unpredictable Inspections”

Ryan McDevitt Professor of Economics · Washington University in St. Louis “Profit Sharing and Patient Steering: Foreclosure from Vertical Ties in Dialysis”

Molly Schnell Assistant Professor · Northwestern University “Selection into Medicine”

Joanne Constantin Assistant Professor · SUNY Downstate Health Sciences University “Medicaid Expansion, Structural Racism, and Maternal Mortality: Preconception Coverage as an Equity Pathway”

FALL 2025

Ben Chartock Assistant Professor · Bentley University “Arbitration and Strategic Zero Bidding in Insurer-Doctor Disputes”

Kurt Lavetti Associate Professor · The Ohio State University “The Earnings Incidence of Employer-Sponsored Health Insurance”

Michael Richards Professor; Director, Sloan Program in Health Administration · Cornell University “Stress Relief? Externalities from Specialty Hospital Entry”

Michelle Marcus Assistant Professor of Economics · Vanderbilt University “Burying the Lead: Effects of Public Lead Service Line Replacements on Blood Lead Levels and Property Values”

Joel Adler UT Austin Dell School of Medicine “Out of Sequence Organ Allocation in Kidney Transplantation: Trade-offs, Transparency, and Guardrails”

Mark Unruh Associate Professor of Population Health Sciences · Cornell University “Nursing Home Ownership and Quality”

Katie Ross-Driscoll Assistant Professor of Surgery · Indiana University School of Medicine “Unintended Consequences of Transplant Performance Metrics: Impacts on Access and Equity”

Rebecca McNally Keehn Associate Professor of Pediatrics · IU School of Medicine “Health Services Innovations to Enhance Access and Equity in Early Autism Diagnosis”

Lady Ikeya PhD student, Public Affairs · Indiana University “Balancing the Bottom Line: Personnel Costs and Input Choices in Nursing Homes”

SUMMER 2025

Pauline Mourot Assistant Professor · Boston University “Should Top Surgeons Practice at Top Hospitals? Sorting and Complementarities in Healthcare”

Xi Chen Associate Professor of Public Health · Yale University “How Do Health Care Use, Spending, and Patient Outcomes Respond to a Timely Dementia Diagnosis?”

Mallory Dreyer “The Impact of Fiscal Policies to Promote Healthy Diets on Birth Outcomes: Evidence from the Navajo Nation”

Jennifer Mangano Doctoral Candidate, Economics · Indiana University “Measuring Infant Mortality Using Synthetic Survival Curves”

SPRING 2025

Adam Wilk Associate Professor of Surgery · Indiana University School of Medicine “Medicare Advantage and Early Transplant Access in Underserved Communities”

Antonios Koumpias Associate Professor of Economics · University of Michigan Dearborn “Association of Retail Health Clinic Market Presence with Medical Appointment Wait Times, 2014-2019”

Matthew Nesvet Postdoc, IU Center for Aging Research, Regenstrief Institute

Myles Wagner Assistant Professor · Ohio State University “How Can Regulating Health Insurance Design Address Market Failures?”

Julian Reif Associate Professor of Finance & Academic Director of Data Science · UIUC “The Long-run Effect of Air Pollution on Survival”

Michael Alexeev Professor of Economics · Indiana University Bloomington “Ownership, Asymmetric Information, and Quality of Care for the Elderly: Evidence from US Nursing Homes During the COVID-19 Pandemic”

Ying Shi Assistant Professor of Public Administration and International Affairs · Syracuse University “The Consequences of Mobile Phone Restrictions in Schools”

Lala Ma Carl F. Pollard Associate Professor of Health Economics · University of Kentucky “Racial Dynamics of Federal Property Buyouts in Flood-Prone Areas”

Zarek Brot-Goldberg Assistant Professor · University of Chicago Harris School “Privatizing Government-Sponsored Health Insurance: Medicare Advantage vs. Traditional Medicare”

FALL 2024

Danae Hernandez-Cortes Assistant Professor · School for the Future of Innovation in Society and School of Sustainability, Arizona State University

Joshua Vest Professor, Interim Associate Dean for Research · IU Indianapolis Fairbanks School of Public Health “Comparison of Methods to Measure Health-related Social Needs”

Parker Rogers Assistant Professor, Business Economics and Public Policy · Kelley School, Indiana University “Regulating the Innovators”

Aparna Soni Assistant Professor · IU Fairbanks School of Public Health “Does Expanding Medicaid Eligibility for Children Reduce Racial Disparities in Later-Life Labor Market Outcomes?”

Adam Soliman Assistant Professor · Clemson University “Supply-Side Drivers of the Illicit Opioid Epidemic”

Kelsey Drewry Assistant Professor of Surgery · Indiana University School of Medicine “Estimating the Impact of a Mandatory Medicare Alternative Payment Model on Racial, Ethnic, and Socioeconomic Disparities in Access to Kidney Transplantation”

Momotazur Rahman Associate Professor of Health Services, Policy and Practice · Brown University “The Effect of Institutionalized Special Needs Plans (I-SNP) on Hospitalization Among Nursing Home Residents”

Matt Aalsma Indiana University “Partnering with Indiana Communities to Implement Data Driven Overdose Prevention”

Leila Agha Harvard University “Productivity after Childbirth: Evidence from Physicians”

SUMMER 2024

Marylis Fantoni PhD Candidate · O’Neill School, Indiana University “Research Design: Encouraging Domestic Violence Reporting by Decreasing Psychological Cost”

Dennis Vera PhD Candidate · Indiana University “Extreme Heat Impacts on Mortalities in the United States”

Dario Salcedo Monroy PhD Candidate · O’Neill School, Indiana University “Land Restitution, Wellbeing and Mental Health Conditions in Colombia”

SPRING 2024

Jia Xiang Assistant Professor, Business Economics and Public Policy · Kelley School, Indiana University “Physicians as Persuaders: Evidence from Hospitals in China”

Kathleen Unroe Associate Professor of Medicine · Indiana University School of Medicine “What’s Next for Long Term Care: Opportunities in Research, Policy and Clinical Practice”

Alden Cheng Postdoctoral Research Associate · University of Illinois Urbana-Champaign “Selection on Unobservables in Discrete Choice Models”

Victoria Barone Assistant Professor, Department of Economics · University of Notre Dame “Democracy and The Opioid Epidemic”

Karl Bilimoria Director, Surgical Outcomes and Quality Improvement Center · Indiana University School of Medicine “Are The Stars Aligned? Improving CMS Public Reporting and Pay-for-Performance Programs”

Terence Cheng Research Scientist, Department of Global Health and Population · Harvard T.H. Chan School of Public Health “Economics of Internet Health Care: Insights from China’s Online Telemedicine Marketplace”

Ricky Camplain Assistant Professor · IU School of Public Health “Holistic Approaches to Indigenous Health and Wellbeing while Incarcerated”

Lauren Schmitz Assistant Professor · La Follette School of Public Affairs, University of Wisconsin “Leveraging Epigenetic Data to Examine Lifecourse Disparities in Aging”

Lindsay Allen Assistant Professor of Emergency Medicine · Northwestern University “Substance Use in Medicaid Enrollees”

Matt Webb Associate Professor, Department of Economics · Carleton University “Difference-in-Differences with Unpoolable Data”

Dario Salcedo PhD Candidate · O’Neill School, Indiana University “Land Restitution, Wellbeing and Mental Health Conditions in Colombia”

FALL 2022

Ashley Bradford “Nuisance Ordinances and Overdose Mortality”

Samuel Mann Vanderbilt University “Employment Non-Discrimination Acts and Mental Health”

Jamila Michener Cornell University “Pandemic Medicaid: A Ground-up Perspective”

Ben Harrell Vanderbilt University “Conversion Therapy Bans, Suicidality, and Mental Health”

Benjamin Chartock Bentley University “Quality Disclosure, Demand, and Congestion: Evidence from Physician Ratings”

Steve Cicala Tufts University “Adverse Selection as a Policy Instrument: Unraveling Climate Change”

Shooshan Danagoulian Wayne State University “Seasonal Allergy Blues: Is Mental Health Worse on High Pollen Days?”

Vini Singh UMass Amherst “Power Dynamics in the Doctor-Patient Relationship”

Barbara Andraka-Christou University of Central Florida “Federal Policies Regulating Medications for Opioid Use Disorder: A Problematic Case of Path Dependence”

SPRING 2022

Jonathan Zhang McMaster University

Monica Garcia Perez St. Cloud State University

Dorainne Green IUB Psychological and Brain Sciences “Cuing Disparities: Exploring the Antecedents and Consequences of Group-Based Social Stressors on Academic Achievement”

Tom Dee Stanford University

Joanne Spetz UCSF

Colleen Barry Dean, Cornell Brooks School of Public Policy

FALL 2021

Dan Sacks Kelley School, Indiana University “Field of Study, Gender, and Grades”

Sumedha Gupta IUPUI Economics “Social Isolation and Dementia/AD Diagnosis Among the Elderly During COVID-19”

Laura Montenovo O’Neill School, Indiana University “The Effects of Weakening Employment Protection Legislation on Job Flows: Evidence from an Italian Reform”

Justin Blackburn IUPUI Fairbanks School of Public Health “Medicaid Enrollment and Health Service Utilization Among Justice-Involved Adults Released from Indiana State Prisons”

Sebastian Fleitas KU Leuven “Quality Based Prices and Multidimensional Quality: Evidence from English Family Doctors”

Salama Freed Duke University “How Did the COVID-19 Health Care Delivery Disruption Affect Medication Use Among People with Chronic Conditions?”

Adrienne Sabety University of Notre Dame “Natural Disasters and Elective Medical Services: The Consequences of Forgone Health Care”

Zirui Song Harvard Medical School “Peer Effects and Health Care Consumption: Evidence from Spousal Health Shocks”

Tamar Oostrom Ohio State University “Deaths of Despair and the Decline of American Religion”

Alberto Ortega O’Neill School, Indiana University “What Happens After a Title IX Investigation?”

SPRING 2021

Angelica Meinhofer Cornell University “Marijuana Liberalization Policies and Perinatal Health”

Lala Ma University of Kentucky “Drinking Water, Fracking, and Infant Health”

Elaine Hill University of Rochester “National COVID Cohort Collaborative (N3C)”

Yaa Owusua Akosa Antwi Johns Hopkins University

Dhaval Dave Bentley University

Dan Sacks Indiana University “Opioids and Employment: A Dynamic Approach”

Isaac Swensen Montana State University

D. Mark Anderson Montana State University

Brady Post Northeastern University

FALL 2020

Patrick Hibbard O’Neill School, Indiana University “Problem Solving Courts and Crime Rates”

Sabrina Young USDA “The Effect of Food Insecurity on the Mental and Cognitive Health of the Elderly”

Catherine Maclean Temple University “Recreational Marijuana and Workers’ Compensation”

Sarah Hamersma Syracuse University “Does SNAP Increase Young Adults’ Engagement in Higher Education?”

Keisha Solomon Johns Hopkins “State Mental Health Parity Laws and Educational Outcomes for College Students”

Monica Deza CUNY “The Intergenerational Effects of the Vietnam Draft on Risky Behaviors”

Chris Ruhm University of Virginia “Mortality Trends Across Education Quartiles”

Nicholas Wright Florida Gulf Coast University “Focus as You Drive and Arrive Alive: The Impact of Handheld Laws on Traffic Fatalities”

Sih-Ting Cai University of Pittsburgh “Zero Premium Insurance Plans: Evidence from Colorado”

Alberto Ortega O’Neill School, Indiana University “Police Killings: Implications from Court-Ordered Hiring Quotas”

SPRING 2020

Sebastian Tello-Trillo University of Virginia “Losing Public Health Insurance: TennCare Reform and Personal Financial Distress”

Shooshan Danagoulian Wayne State University “Office Visits Preventing Emergency Room Visits: Evidence from the Flint Water Switch”

Ithai Lurie US Department of Treasury “Health Insurance and Mortality: Experimental Evidence from Taxpayer Outreach”

Sara Markowitz Emory University “The Effects of State Scope of Practice Laws on the Labor Supply of Advanced Practice Registered Nurses”

Justin Blackburn IUPUI School of Public Health “Using Health Information Exchange Data to Assess Disenrollment in Medicaid”

Joanne Spetz UCSF “Healthcare Workforce”

FALL 2019

Ludovica Gazze University of Chicago “How Do Governments Identify Environmental Hazards? Targeting Lead Screening in Illinois”

Lauren Jones Ohio State University “The Long-term Effect of the Earned Income Tax Credit on Physical and Mental Health”

Jason Lindo Texas A&M University “The Power of the IUD: Effects of Expanding Access to Contraception through Title X Clinics”

Molly Schnell Northwestern University “The Impacts of Physician Payments on Patient Access, Use, and Health”

Pinar Karaca-Mandic University of Minnesota “Medical Reversals: Understanding the Process of De-adoption of Ineffective Treatment”

Sumedha Gupta IUPUI “Spillover Effects of Regulating Opioid Substitutes”

Jonathan Cantor RAND “Opioid Disorder Treatment Facilities”

Felipe Lozano Rojas O’Neill School, Indiana University “Evaluation of Soda Tax Policies”

SPRING 2019

Brian Kaskie University of Iowa

Aparna Soni Indiana University

Tim Moore Purdue University

Seth Freedman Indiana University

Laura Wherry UCLA

Briggs DePew Louisiana State University

FALL 2018

Dan Sacks Indiana University “Retirement Notches in Federal Employment”

Haizhen Lin Indiana University “Opioid Epidemic Among the Insured Working Population”

Vivek Astvansh Kelley School, Indiana University “Medical Device Recalls: Does Choice of Communication Medium Impact Firm’s Valuation?”

SPRING 2018

Kurt Lavetti Ohio State University “The Persistence of Health Status and the Ex Ante Value of Public Health Insurance”

David Hagman PhD Candidate, Carnegie Mellon / Penn “Behavioral Economics and Experimental Research in Health and Finance”

Tony Lo Sasso University of Illinois at Chicago “Central Planning and Market Based Reforms: Evidence from Ohio’s Long Term Care Market”

Dan Black Harris School, University of Chicago “Simple Tests for Selection: Learning More From Instrumental Variables”

David Molitor University of Illinois “Long-Run Health Dynamics in the Wake of Disaster: Evidence from Hurricane Katrina”

Sumedha Gupta IUPUI “Fight Against Opioid Misuse: Lessons from Kentucky”

Andrew Friedson UC Denver “The Affordable Care Act and Ambulance Response Time”

Sunita Desai NYU Wagner School “The 340B Drug Pricing Program and the Provision of Uncompensated Care”

FALL 2017

Dan Sacks Indiana University “Did Defunding the Risk Corridors Program Kill Obamacare?”

Brian D’Onofrio IU Department of Psychology “Translational Psychiatric Epidemiology: What’s Happening Down the Street?”

Coleman Drake University of Minnesota “Consumer Valuation of Physician Network Size in the 2017 Covered California Health Insurance Marketplace”

Biniyam Yemane Indiana University “The Education-Health Gradient, Pathways, and Spillovers: New Evidence from a Natural Experiment”

SPRING 2017

Seth Freedman Indiana University “Observational Studies of the Effect of Medicaid on Health Outcomes: Illustrations of Bias”

Engy Ziedan University of Illinois Chicago “The Intended and Unintended Consequences of the Hospital Readmission Reduction Program”

Dan Sacks Indiana University “Did Defunding the Risk Corridors Kill Obamacare?”

Alex Hollingsworth Indiana University “Using Marijuana Legalization to Learn About Substitution Patterns in Markets for Recreational Psychoactive Substances”

Vivian Ho Rice University “Do Freestanding Emergency Departments Encourage Free-Spending Care?”

Anita Mukherjee University of Wisconsin Madison “Difference-in-Difference Estimates on the Impacts of Opioid Antagonist Legislation”

INAUGURAL TALK -- MAY 2011

Craig Garthwaite Kellogg School, Northwestern University “The Doctor Might See You Now: The Supply Side Effects of Public Health Insurance Expansions”

That’s a lot of research. Hundreds of papers, dozens of institutions, one persistent 9:30am slot -- and counting.

Going forward, you’ll get a post each week previewing the upcoming talk. If you work in health economics or health policy -- whether you’re in Bloomington or not -- we hope you’ll follow along.

The O’Neill School Health Policy Workshop meets Thursdays, 9:30-11:00am, SPEA A225 (and via Zoom). Indiana University Bloomington.