{"id":1359,"date":"2026-03-31T13:00:00","date_gmt":"2026-03-31T13:00:00","guid":{"rendered":"https:\/\/forecastingresearch.org\/?post_type=research&p=1359"},"modified":"2026-05-21T14:18:36","modified_gmt":"2026-05-21T14:18:36","slug":"economic-effects-of-ai","status":"publish","type":"research","link":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai","title":{"rendered":"Forecasting the Economic Effects of AI"},"content":{"rendered":"\n

Abstract<\/h2>\n\n\n\n

We elicit forecasts of how AI will affect the U.S. economy, comparing the beliefs of five groups: academic economists, employees at AI companies, policy researchers focused on AI, highly accurate forecasters, and the general public. The median respondent in each group expects substantial advances in AI capabilities by 2030, small declines in labor force participation consistent with demographic shifts, and an annual GDP growth rate of 2.5%, which exceeds both the typical medium-run (2.0%) and long-run (1.7%) baseline forecasts from government agencies and private-sector forecasters. Conditional on a \u201crapid\u201d AI progress scenario, in which AI systems surpass human performance on many cognitive and physical tasks, experts forecast substantial, though not historically unprecedented, economic shifts: annualized GDP growth rising to around 4% and the labor force participation rate falling from its current level of 62% to 55% by 2050, with roughly half of that decline\u2014equivalent to around 10 million lost jobs\u2014attributable to AI. A variance decomposition suggests that expert disagreement about these effects is driven primarily by different beliefs about the economic effects of highly capable AI systems rather than by disagreement about the pace of AI progress. These forecasts map onto notably different policy preferences across groups: experts strongly favor targeted measures such as worker retraining, whereas the general public supports both targeted programs and broader interventions, including a job guarantee and universal basic income.<\/p>\n\n\n\n

\n
View the full PDF Report \n \n<\/svg>\n\n \n<\/svg><\/a><\/div>\n<\/div>\n\n\n\n
\n
Acknowledgments<\/summary>\n

We thank seminar attendees at the Congressional Budget Office for thoughtful feedback on this project. We also thank Morgane Bascle, Oskari Luoma, Aashish Reddy, Sam Holton, and Victoria Schmidt for outstanding research assistance.<\/p>\n<\/details>\n\n\n\n

Disclaimer<\/summary>\n

This research was funded by a grant from Open Philanthropy (now known as Coefficient Giving). The views expressed in this paper do not necessarily reflect the views of the Federal Reserve Bank of Chicago or the Federal Reserve System.<\/p>\n<\/details>\n<\/div>\n\n\n\n

1. Introduction<\/h2>\n\n\n\n

1.1 Background<\/h3>\n\n\n\n

The diffusion of generative artificial intelligence into workplaces,\nconsumer products, and public services has renewed a familiar set of\neconomic questions: Will automation raise productivity enough to change\nthe economy\u2019s long-run growth path? What will happen to work\u2014employment,\nparticipation, wages, and occupational structure\u2014as machines become\ncapable of performing an increasingly broad set of cognitive tasks? And\nif the economic gains from AI are large, will they be broadly shared or\nconcentrated among owners of capital and workers whose skills complement\nAI?<\/p>\n\n\n\n

Despite intense attention, the evidence on AI\u2019s economic effects to date remains mixed. Recent work finds signs of labor-market adjustment concentrated among the parts of the workforce most exposed to generative AI: Brynjolfsson, Chandar, et al. (2025)<\/span> document a sizable decline in early-career employment in AI-exposed occupations following the widespread rollout of generative AI, with limited effects on older workers and less-exposed fields.1<\/a><\/sup> Complementary evidence, however, complicates this interpretation. Humlum and Vestergaard (2025)<\/span> find similar early-career declines in Danish data, but do not find that the declines are strongly linked to firm-level adoption of generative AI,2<\/a><\/sup> raising the possibility that measured effects reflect broader changes in demand, task organization, or labor supply. Davis (2026)<\/span> adds further context to these findings, showing that while employment in highly AI-exposed sectors has lagged the rest of the economy since late 2022, wages in those same sectors have not fallen.3<\/a><\/sup> Related evidence in Gimbel et al. (2025)<\/span> likewise points to limited near-term aggregate impacts.4<\/a><\/sup> Taken together, the emerging empirical record suggests that some groups of workers and occupations may already be experiencing measurable effects from AI, but the aggregate macroeconomic signal remains difficult to isolate from typical fluctuations in a dynamic economy.<\/p>\n\n\n\n

At the same time, prominent voices in business and the AI industry warn of transformative upheaval: Jamie Dimon, CEO of JPMorgan Chase, argues that AI \u201cwill eliminate jobs\u201d and that \u201cpeople should stop sticking their head[s] in the sand\u201d;5<\/a><\/sup> Sam Altman, CEO and cofounder of OpenAI, predicts that \u201cwhole classes of jobs\u201d will disappear even as unprecedented wealth is created;6<\/a><\/sup> and Dario Amodei, CEO and cofounder of Anthropic, suggests that AI could push overall unemployment to 10\u201320% within the next five years.7<\/a><\/sup> Quantitative analyses span a similarly broad range. Arnon (2025)<\/span>, using the Wharton Budget Model, projects AI-driven productivity and GDP gains (in levels) of just 1.5% by 2035 and 3.7% by 2075\u2014modest figures that translate to less than 0.04 percentage points of additional annual productivity growth in the long run.8<\/a><\/sup> The OECD, by contrast, estimates that AI could add 0.4\u20131.3 percentage points to annual aggregate labor productivity growth over a ten-year horizon in high-exposure countries (Filippucci et al. 2025)<\/span>.9<\/a><\/sup> The gap between these assessments is wide, and the policy stakes are correspondingly high.<\/p>\n\n\n\n

A central challenge in debates about the economic effects of AI is\nthat such forecasts are, unavoidably, joint forecasts about the\ncapabilities of AI systems and the diffusion of AI-related technology\ninto the economy. In practice, these discussions often combine answers\nto three distinct questions into one. First, will AI capabilities\nadvance meaningfully, such that AI systems become capable of\nindependently performing, or assisting with, a large quantity of\neconomically valuable work? Second, if such progress occurs, what will\nhappen to key macroeconomic outcomes, including GDP growth,\nproductivity, labor-force participation, and inequality? And third,\ngiven predictions and uncertainty about the effects of AI on the\neconomy, what are the optimal policy responses?<\/p>\n\n\n\n

The first question about the progression of AI capabilities is primarily addressed by two research fields: (1) AI benchmarking, where computer scientists develop tasks that track the limits of AI systems\u2019 capabilities over time in work such as Russakovsky et al. (2015), Hendrycks et al. (2020), and Jain et al. (2024)<\/span>;10<\/a><\/sup> and (2) capability forecasting, where researchers collect predictions about AI capabilities in projects like Grace et al. (2024)<\/span> and the Longitudinal Expert AI Panel (Murphy et al. 2025)<\/span>.11<\/a><\/sup> The second question on how AI diffuses into the economy is the core focus of a large and growing economics literature, as discussed above, yet standard models yield ambiguous predictions: the impact of rapid technological change on employment is theoretically indeterminate in the short run and neutral in the long run under canonical frameworks,12<\/a><\/sup> in part because automating some human tasks often augments the value of others.13<\/a><\/sup> Diffusion and adoption introduce further uncertainty, as even the most powerful technologies can take decades to reshape aggregate outcomes (Comin and Hobijn 2010)<\/span>.14<\/a><\/sup> On the third question\u2014which policies to optimally pursue\u2014progress is difficult because of unresolved disagreement about the answers to the first two questions, since the optimal policy interventions depend critically on both the pace of technological change and its economic consequences.<\/p>\n\n\n\n

1.2 This paper<\/h3>\n\n\n\n

We elicit expert beliefs on all three of these questions simultaneously, surveying four carefully selected expert groups alongside a general public sample. We draw our expert economist panel from those publishing on the intersection of AI and economics, speakers at major conferences about AI and the economy, and faculty at top-100 economics departments\u2014with a dedicated sub-sample of prominent economists. We recruit AI industry professionals from the organizations behind frontier AI models, and we sample AI policy professionals from U.S.-based think tanks and research institutions. We draw highly accurate forecasters (superforecasters) from a set of individuals with a verified track record of exceptional predictive accuracy.15<\/a><\/sup> See Section 2.1<\/a> for more information about our survey respondents.<\/p>\n\n\n\n

First, we collect unconditional (all-things-considered) forecasts of\nkey U.S. economic variables\u2014such as annual GDP growth, total factor\nproductivity (TFP) growth, the labor-force participation rate (LFPR),\nand wealth inequality\u2014at both near-term (2030) and long-term (2050)\nhorizons. These forecasts reflect each respondent\u2019s current beliefs\nabout the likely trajectory of the economy given their overall views\nabout both AI and other trends and shocks. Second, we elicit forecasts\nof the same variables conditional on three AI-progress scenarios by\n2030\u2014slow, moderate, and rapid\u2014and each respondent\u2019s subjective\nprobability that the world will end up in each of these scenarios. This\nconditional structure allows us to separate disagreement about AI\ncapabilities from disagreement about economic impacts, given a\nparticular level of capabilities. Third, we ask respondents to predict\nthe marginal effect of six specific policy proposals on GDP and LFPR\nunder both unconditional and rapid-progress conditions, and to indicate\ntheir normative support for each policy. This multi-layered design\nallows us to trace out how beliefs about AI progress propagate through\nto economic forecasts and, ultimately, to policy preferences.<\/p>\n\n\n\n

The three scenarios for AI progress describe capabilities progress achieved by 2030, across the domains of research, problem-solving, creativity, agency, and robotics. While the full scenarios are described in Appendix H.2.1<\/a>, to summarize:<\/p>\n\n\n\n

    \n
  1. In the \u201cslow\u201d scenario, AI is a capable assisting technology for\nhumans: writing literature reviews at the level of a capable PhD\nstudent, handling half of all freelance software-engineering jobs that\nwould take an experienced human a day to complete, topping up your\nonline grocery cart, and physically being able to unload dishwashers in\nsome homes.<\/li>\n\n\n\n
  2. In the \u201cmoderate\u201d scenario, AI is an effective collaborator\nacross domains: autonomous lab systems can make rapid advances in\nsolar-cell technology; almost all freelance software-engineering jobs\nrequiring 5 days of effort from an experienced human are automatable;\nrobots can do dishes as quickly as humans; robo-taxis can drive anywhere\nthat humans can.<\/li>\n\n\n\n
  3. In the \u201crapid\u201d scenario, AI systems surpass humans in most\ncognitive and physical tasks. Autonomous researchers can collapse\nyears-long research timelines into months or even days. AI systems can\nsurpass all freelance software engineers, customer service agents,\nparalegals, and clerical workers. Models can write 2025-Pulitzer-caliber\nbooks\u2014and negotiate the resulting book contract. Robots can assist in an\narbitrary home or factory anywhere in the world.<\/li>\n<\/ol>\n\n\n\n

    In addition, forecasters were advised that the scenarios above were\nintended to describe AI capabilities, not adoption, and that they should\nconsider that regulation, social norms, or integration challenges could\ndelay real-world deployment of systems with these capabilities. They\nwere further advised that reasonable people may disagree with our\ncharacterization of what constitutes slow, moderate, or rapid AI\nprogress and may expect slow progress in some AI capabilities alongside\nmoderate or rapid progress in others. Nevertheless, they were asked to\nselect the scenario that, on balance, best represented their views.\nFinally, capabilities were defined as \u201cachieved\u201d only if they could be\ndone by an AI system as inexpensively and as reliably as humans\ntoday.<\/p>\n\n\n\n

    1.3 Key findings<\/h3>\n\n\n\n

    A detailed analysis of the resulting forecasts yields six key\nfindings:<\/p>\n\n\n\n

    A majority of survey respondents predicted significant AI progress by 2030.<\/strong> All five groups surveyed anticipate substantial AI capability advances\u2014even if real-world adoption lags\u2014with the average economist assigning a 61.4% probability to moderate or rapid progress.<\/p>\n\n\n\n

    Despite expecting significant AI progress, unconditional economic forecasts are close to historical trends.<\/strong> Although economists\u2019 unconditional GDP forecasts exceed almost all government and private-sector projections, most do not forecast major departures from recent macroeconomic baselines, citing historical base rates, adoption lags, demographic headwinds, policy responses, potential infrastructure bottlenecks, and longstanding patterns in how general-purpose technologies affect the economy.<\/p>\n\n\n\n

    Conditional on the rapid scenario, economists expect significant economic shifts, but not the transformative acceleration some have predicted.<\/strong> If the rapid scenario materializes, economists forecast GDP growth of 3.5%, LFPR falling to 55.0%, and the fraction of wealth held by the wealthiest 10% of households rising to 80.0% by 2050\u2014large shifts, but with historical parallels, such as to GDP growth post-WWII, the LFPR before women entered the workforce en masse, or pre-WWII inequality.<\/p>\n\n\n\n

    Unconditional consensus masks significant uncertainty about rapid scenario outcomes.<\/strong> Economists\u2019 unconditional forecasts are relatively tightly clustered, but under the rapid scenario the range of plausible outcomes expands, suggesting that experts have far less confidence about what would happen if AI proves truly transformative.<\/p>\n\n\n\n

    Between-group differences are small relative to within-group disagreement, and most disagreement reflects uncertainty about economic effects rather than AI capabilities.<\/strong> In contrast to the view, articulated most clearly by Cunningham (2025)<\/span>, that the primary source of disagreement about AI\u2019s economic effects is disagreement about the pace of AI capability progress,16<\/a><\/sup> a variance decomposition suggests that expert disagreement about long-run macroeconomic outcomes is primarily driven by different beliefs about the economic effects of highly capable AI systems rather than disagreement about the pace of AI development itself.<\/p>\n\n\n\n

    Economists and the general public disagree on how to respond to AI\u2019s economic impacts.<\/strong> Economists strongly favor targeted policy interventions such as AI-focused worker retraining (71.8% support) over broad structural interventions like job guarantees (13.7%) or universal basic income (37.4%), whereas the general public supports both targeted and broad interventions.<\/p>\n\n\n\n

    1.4 Prior Work<\/h3>\n\n\n\n

    A growing body of economic research examines how advanced AI may\nreshape productivity and growth, labor markets, and inequality. While\nthere is near-universal agreement that more capable AI tends to raise\nproductivity, the magnitude and timing of this and other effects are\ncontested. The core disagreement reflects a debate about two previously\ndiscussed questions: how fast will AI capabilities progress, and how\nfast will capable AI systems diffuse through the economy? This paper\nelicits expert beliefs on both dimensions, as well as on policy\nresponses, and our survey design maps directly onto this debate.<\/p>\n\n\n\n

    Productivity and\nGrowth<\/h4>\n\n\n\n

    When it comes to productivity and growth, the research largely falls into three schools of thought, each defined by how broadly they believe AI can replace human labor across economic tasks. At the most optimistic end, Trammell and Korinek (2023)<\/span> review theoretical models which suggest that, if AI can automate both production and research and development, self-reinforcing feedback between capital accumulation and idea generation could produce double-exponential or hyperbolic growth, marking a structural break from the near-linear growth of the past two centuries.17<\/a><\/sup> Erdil and Besiroglu (2024)<\/span> synthesize the arguments for \u201cexplosive growth,\u201d which refers to economic growth rates significantly higher than recent historical trends, and assign roughly 50% odds of this occurring by 2100 if AI capable of broadly substituting for human labor is developed, while acknowledging the many potential regulatory, physical, and alignment constraints that create wide uncertainty in this forecast.18<\/a><\/sup><\/p>\n\n\n\n

    A more moderate view anticipates meaningful but smaller bumps in productivity and economic growth from AI. Based on Eloundou et al. (2024)<\/span> estimates of the tasks GPT-4 could perform with scaffolding,19<\/a><\/sup> Aghion and Bunel (2024)<\/span> estimate annual TFP gains of 0.5\u20131.3 percentage points depending on assumptions about adoption and the share of tasks that are profitably automatable, noting that gains are likely transitory unless AI also automates idea production.20<\/a><\/sup> Filippucci et al. (2024)<\/span> embed sector-level AI exposure in a general equilibrium model and find a similar 0.25\u20130.6 percentage point increase in TFP,21<\/a><\/sup> while highlighting that Baumol Cost-related effects from sectors with limited AI penetration may constrain aggregate gains.22<\/a><\/sup> Beraja and Zorzi (2022)<\/span> add an important caveat to this discussion by explaining that when private incentives diverge from social welfare, which often happens in the rollout of a new technology, automation may be inefficiently allocated, slowing welfare gains even when measured productivity rises.23<\/a><\/sup><\/p>\n\n\n\n

    The most skeptical view is described by Acemoglu (2024)<\/span>, who uses task-based growth accounting calibrated to data to conclude that what he considers to be realistic near-term AI adoption adds only around 0.07 percentage points per year to TFP growth.24<\/a><\/sup> His key finding is that most currently exposed tasks are either low-value, not yet reliably automatable, or both. Consistent with this view, Comunale and Manera (2024)<\/span> confirm that firm-level productivity effects are positive, but that macro magnitude remains highly sensitive to adoption speed and institutional context.25<\/a><\/sup> This \u201cproductivity without prosperity\u201d framing has broader implications: income for many may fall, since even if aggregate output rises, labor\u2019s share of that output may fall, concentrating gains at the top of the income or wealth distribution.<\/p>\n\n\n\n

    Labor Markets<\/h4>\n\n\n\n

    The empirical labor market literature has struggled to keep pace with a technology diffusing faster than standard data-collection cycles can capture. Brynjolfsson, Chandar, et al. (2025)<\/span> document a 13% relative employment decline for early-career workers (ages 22\u201325) in the most AI-exposed occupations following the widespread rollout of generative AI.26<\/a><\/sup> This may be evidence of concentrated displacement of entry-level workers; an early indicator of eventual macroeconomic effects. In this analysis, the mechanism appears to resemble a negative labor demand shock rather than simple labor substitution. Wages in exposed roles rose alongside employment declines, consistent with firms using fewer, more experienced workers rather than replacing all workers with AI. In contrast, Gimbel et al. (2025)<\/span> find the overall occupational mix (by AI exposure) to be broadly stable,27<\/a><\/sup> and Humlum and Vestergaard (2025)<\/span> look for and find Brynjolfsson, Chandar, et al. (2025)<\/span>\u2019s employment decline28<\/a><\/sup> in a Danish sample, but show that the decline is uncorrelated with firm-level generative AI adoption,29<\/a><\/sup> raising questions about potential confounders.<\/p>\n\n\n\n

    These conflicting early results are consistent with an economics literature that often struggles to find technological displacement in aggregate data over short time horizons. Frey and Osborne (2017)<\/span>, whose framework pre-dates generative AI, remains the standard baseline for classifying occupations as being exposed to automation from computerization; they summarize the literature finding such effects in data from the prior decades.30<\/a><\/sup> On the other hand, Korinek and Juelfs (2022)<\/span> provide a theoretical counterpoint to pessimism about displacement.31<\/a><\/sup> They argue that a utilitarian social planner would phase out work gradually and only for workers with low productivity and weak non-monetary attachment to employment. The implication is that the appropriate response to automation is a managed transition rather than resistance.<\/p>\n\n\n\n

    Inequality<\/h4>\n\n\n\n

    A recurring concern in the literature is that AI\u2019s productivity gains will be distributed unequally, potentially decoupling aggregate economic growth from broad-based welfare improvement for most people. Acemoglu (2024)<\/span> argues that rising TFP alongside a falling labor share is the most plausible scenario given task-based substitution patterns.32<\/a><\/sup> Korinek (2024)<\/span> extends this argument, modeling a post-AGI economy33<\/a><\/sup> in which productivity growth could exceed 18% annually while workers\u2019 wages collapse unless there is active redistribution.34<\/a><\/sup> This model motivates a set of interconnected policy responses that the authors propose, ranging from UBI to global AI governance. Korinek and Stiglitz (2021)<\/span> echoes this inequality concern on a global scale, arguing that AI strengthens superstar-firm dynamics and concentrates rents in the most developed economies, weakening the development pathways that allowed developing countries to grow quite fast in the 20th century.35<\/a><\/sup> Lastly, Abbott and Bogenschneider (2018)<\/span> identify a mechanism that could amplify these trends: existing tax systems treat automation-related capital as deductible while taxing labor, creating a potential bias toward over-automation that erodes the labor tax base.36<\/a><\/sup><\/p>\n\n\n\n

    Policy Responses<\/h4>\n\n\n\n

    Several studies emphasize that optimal policy responses to AI (or technological innovation more generally) depend on assumptions about the pace and breadth of AI adoption. Slow, uneven diffusion may imply that targeted worker adjustment programs are optimal; rapid, broad-based automation may require structural reforms to taxation and redistribution at a scale with no peacetime precedent. Comunale and Manera (2024)<\/span> and the OECD (OECD 2023)<\/span> both catalog existing regulatory approaches and stress that optimal interventions vary sharply with assumptions about adoption speed and displacement risk.37<\/a><\/sup><\/p>\n\n\n\n

    On the labor-market adjustment side, Hyman et al. (2024)<\/span> use a regression discontinuity in the U.S. Trade Adjustment Assistance program to show that this type of program shortens unemployment spells by 1.26 quarters on average, raises cumulative earnings by over $18,000, and is self-financing once one accounts for fiscal externalities from affected workers.38<\/a><\/sup> This result speaks directly to the debate about modernized unemployment insurance. Korinek and Juelfs (2022)<\/span> argue that broader institutional reform is ultimately needed, shifting social insurance from work-conditioned benefits to unconditional income support as automation capital replaces labor income.39<\/a><\/sup><\/p>\n\n\n\n

    On the tax and redistribution side, Abbott and Bogenschneider (2018)<\/span> propose neutralizing the current potential bias towards capital (and against labor) by taxing capital income at the same rate as labor income, instead of at an implicitly lower rate.40<\/a><\/sup> Bastani and Waldenstr\u00f6m (2024)<\/span> provide a more systematic optimal-tax approach, concluding that capital taxation becomes increasingly important as the labor tax base erodes, though they note practical constraints on feasibility in open economies.41<\/a><\/sup> At the far end of the spectrum, Korinek (2024)<\/span> and Anthropic (2025)<\/span> advance comprehensive policy frameworks organized by scenario severity, from modest workforce training grants to sovereign wealth funds and new revenue structures, an approach that directly motivates this survey\u2019s conditional policy questions.42<\/a><\/sup><\/p>\n\n\n\n

    Prior Surveys<\/h4>\n\n\n\n

    Several prior surveys have elicited expert beliefs about AI capabilities and economic impacts. Grace et al. (2024)<\/span> survey 2,778 AI researchers on their predictions about the pace of AI progress, finding a median estimate that AI will outperform humans at all tasks by 2047, with substantial disagreement across respondents.43<\/a><\/sup> The Clark Center\u2019s US Economic Experts Panel (Clark Center Forum 2025)<\/span> has polled leading economists on whether AI adoption will substantially increase growth rates or unemployment over the next decade, with most economists expressing uncertainty about the magnitude though agreeing on the direction of productivity effects.44<\/a><\/sup> Murphy et al. (2025)<\/span> establish the Longitudinal Expert AI Panel (LEAP), a three-year panel survey tracking the views of computer scientists, economists, AI industry professionals, and policy researchers through monthly surveys, providing a dynamic alternative to one-time cross-sectional surveys.45<\/a><\/sup> (We compare our results directly to LEAP\u2019s in Appendix C<\/a>.) Our survey differs from these prior studies in three key ways: (1) we elicit forecasts of specific quantitative outcomes rather than qualitative agreement or disagreement; (2) we condition on explicit AI-progress scenarios, thereby allowing us to separate beliefs about capabilities from beliefs about economic effects; (3) and finally, we simultaneously collect forecasts and normative policy preferences, allowing us to link empirical expectations to policy support.<\/p>\n\n\n\n

    The Source of\nDisagreement and Our Contribution<\/h4>\n\n\n\n

    A key open question that our survey is designed to address is whether disagreement about future economic outcomes is primarily about AI capabilities progress or about economic mechanisms conditional on that progress. Cunningham (2025)<\/span> argues it is predominantly the former: forecasters largely agree on the economic logic but diverge on whether transformative AI capabilities will actually arrive.46<\/a><\/sup> This view is consistent with the observation that Acemoglu (2024)<\/span>\u2019s skepticism and Trammell and Korinek (2023)<\/span>\u2019s optimism are in part reconcilable as they agree on the theoretical mechanisms, but apply to very different scenarios for AI capabilities and adoption.47<\/a><\/sup> Here, we examine forecasts based on scenarios that span a wide range of AI development. Contra Cunningham (2025)<\/span>, we find that disagreement centered on whether new AI capabilities will have an economic impact, rather than disagreement over whether such capabilities will arise.<\/p>\n\n\n\n

    2. Methods<\/h2>\n\n\n\n

    This section describes the strategy for recruiting survey\nparticipants, the survey instrument, and the data processing procedures\nused in this study. The survey was launched in October 2025 and\nconcluded in February 2026.<\/p>\n\n\n\n

    2.1 Participant Recruitment<\/h3>\n\n\n\n

    We surveyed a diverse panel of experts and non-experts. Our sampling\nstrategy targeted five participant groups: (i) economists, (ii) AI\nindustry professionals, (iii) AI policy professionals, (iv)\nsuperforecasters, and (v) members of the general public. These groups\nwere selected to capture complementary forms of expertise relevant to\nforecasting technological progress, economic outcomes of that progress,\nand policy responses. For ease of presentation, in most analyses, we\ngroup AI industry and policy professionals under \u2018AI experts\u2019.<\/p>\n\n\n\n

    In total, we contacted 4,866 experts, 54 superforecasters, and 1,395 members of the general public. Of the invited, 90 economists, 30 AI industry professionals, 27 AI policy professionals, 38 superforecasters and 401 members of the general public completed the survey. After filtering our expert respondents to ensure they meet our population criteria, our final sample contains 69 economists, 27 AI industry professionals, and 25 AI policy professionals. For details on participant recruitment, see Appendix A<\/a>.<\/p>\n\n\n\n

    Economists, AI industry professionals and AI policy professionals\nwere compensated at $100\/hour for a minimum of five and a maximum of ten\n(self-reported) hours. Superforecasters were paid $60\/hour, with the\nsame time limits on compensation. Participants from the general public\nwere initially paid $30 for completing the survey, but this was\nincreased to $40 for later batches of participants. In addition to these\npayments, we incentivized participants to give insightful rationales by\npromising to award ten $500 prizes among the expert participants, two\n$500 prizes among superforecasters, and twenty $100 prizes among the\ngeneral public to participants with the highest-quality rationales.<\/p>\n\n\n\n

    Economists<\/h4>\n\n\n\n

    Our primary population of interest was economists, who are best positioned to understand and provide quantitative forecasts of macroeconomic outcomes. We further subdivided this target population into three sub-populations: (i) economists working on AI-related topics, (ii) economists working on economic growth and technological changes more broadly, and (iii) well-known economists, such as Nobel prize winners. In the final sample, these groups have 24, 43, and 2 respondents, respectively. Figure 1<\/a> shows summary statistics about our economist respondents. Our respondents were younger, less experienced, more male, and more European than the sampling frame. They also had fewer high-quality publications and were less likely to have high levels of citations. That being said, we still have significant coverage across the less represented groups in our participant set. For example, 45% of our respondents had at least 16 years of experience, 38% were age 45 and older, 17% were female, 35% were from North America, 37% had at least one Tier A publication (a so-called \u2018top-5\u2019 publication in economics), and 58% had at least 1,000 citations. In our main results, we reweight the participants on experience and geography to match the sampling frame. This has minimal effects on our results (see Appendix G<\/a>).<\/p>\n\n\n\n

    \"\"\/
    Figure 1: Summary statistics of the economist sample.<\/strong> The bars on the left show the distribution for participants, and the bars on the right show the distribution in our sampling frame (both participants and non-participants). The figure is based on publicly available information collected by human assistants and an AI-based system. Tier A and B publications are defined in Table 58<\/a>.<\/figcaption><\/figure>\n\n\n\n

    Economists working on AI were identified using three primary sampling\npools: (i) a literature-based pool of authors publishing on the\neconomics of AI, identified through Research Papers in Economics (RePEc)\nusing relevant Journal of Economic Literature (JEL) codes and AI-related\nkeywords, (ii) an event-based pool of speakers and participants at major\nacademic and policy conferences focused on AI and economic outcomes, and\n(iii) an institution-based pool targeting top-100 economics institutions\naccording to RePEc. In the literature pool, invitations were extended in\ndescending order of citation counts adjusted by paper age. In other\npools, we sampled randomly.<\/p>\n\n\n\n

    Economists working on growth and technological change were similarly\nidentified through (i) publications indexed in RePEc using JEL codes\nrelated to technological innovation and economic growth, and (ii) an\ninstitution-based pool. In the literature pool, authors were ranked by\nage-adjusted citation counts, and invitations were sent in descending\norder.<\/p>\n\n\n\n

    Well-known economists were identified using a combination of Nobel\nPrize recipients, RePEc author rankings, and participation in the Clark\nCenter U.S. Economic Experts Panel. Invitations were extended to all\nindividuals meeting these criteria.<\/p>\n\n\n\n

    AI Industry\nProfessionals<\/h4>\n\n\n\n

    AI industry professionals were sampled from companies developing or\napplying frontier AI models. We constructed sampling pools using three\nsources: (i) institutions associated with frontier models ranked by\ntraining compute, (ii) institutions associated with organizations\nproducing top-performing models on public evaluation leaderboards, and\n(iii) highly funded AI startups identified via fundraising databases.\nWithin each institution, we randomly sampled research and engineering\nstaff. Sampling was stratified to ensure maximum coverage across chosen\norganizations.<\/p>\n\n\n\n

    AI Policy\nProfessionals<\/h4>\n\n\n\n

    AI policy experts were sampled from U.S.-based think tanks, research\ninstitutions, and government-affiliated organizations engaged in AI\ngovernance and technology policy. Participants were identified via\ninstitutional staff directories and professional networking platforms,\nfocusing on researchers and policy practitioners working directly on AI\ngovernance.<\/p>\n\n\n\n

    Superforecasters<\/h4>\n\n\n\n

    Superforecasters were recruited through our existing connections. All\nindividuals in this pool have a demonstrated track record of forecasting\naccuracy.<\/p>\n\n\n\n

    Forecasters are denoted \u201csuperforecasters\u201d if they (1) were in the top 2% of the accuracy distribution in a given year of the Intelligence Advanced Research Projects Activity (IARPA) Aggregative Contingent Estimation (ACE) tournament48<\/a><\/sup> or (2) they were a highly accurate performer on Good Judgment Open, an online continuous geopolitical forecasting tournament. Good Judgment Inc., which runs Good Judgment Open, then adds these top forecasters to the \u201csuperforecaster\u201d pool. Most superforecasters come from the first selection criterion. Mellers et al. (2015)<\/span> find persistent performance of these superforecasters across several years of geopolitical forecasting.49<\/a><\/sup><\/p>\n\n\n\n

    General Public<\/h4>\n\n\n\n

    We included a general public sample to compare expert beliefs against the broader public\u2019s expectations. Members of the general public were recruited through CloudResearch Connect.50<\/a><\/sup><\/p>\n\n\n\n

    2.2 Survey Instrument<\/h3>\n\n\n\n

    The survey instrument elicited probabilistic forecasts of AI\nprogress, economic growth, labor-market outcomes, inequality, and policy\neffects over medium- and long-term horizons (by 2030 and 2050).\nRespondents provided both unconditional forecasts and conditional\nforecasts under three AI development scenarios\u2014slow, moderate, and rapid\nprogress, defined using concrete capability benchmarks.<\/p>\n\n\n\n

    For forecasts of economic outcomes, participants reported 50th-percentile forecasts and, for selected questions, 10th- and 90th-percentile forecasts. The survey also collected qualitative rationales to capture underlying reasoning and mechanisms. Participants received background information, historical data, and detailed resolution criteria for all forecasting questions. The survey instrument was administered online and required 8 hours for completion, on average. The survey questions are shown in Appendix H<\/a> and the forecasting interface in Figure 80<\/a>.<\/p>\n\n\n\n

    2.3 Data Processing<\/h3>\n\n\n\n

    Coherence Checks<\/h4>\n\n\n\n

    To ensure forecasts were both logically coherent and representative of participants\u2019 intended forecasts, we applied a series of consistency checks to each participant\u2019s forecasts. These include verifying that scenario probabilities summed to one and that conditional forecasts were not inconsistent with the unconditional forecasts for a given economic outcome. We reached out to participants to give them the opportunity to update their forecasts if their forecasts were flagged by these checks. In all analyses, we present results using these updated forecasts. If participants did not update their forecasts, we removed their forecasts for some checks, while we left them in place for others. For a full list of these checks, see Appendix F<\/a>. The appendix also describes the impact of the coherence check process on aggregate forecasts: differences between pre- and post-intervention results are small.<\/p>\n\n\n\n

    Reweighting<\/h4>\n\n\n\n

    We assign each participant in the economist group a weight to correct for non-response bias in participant recruitment. This ensures that the sample we derive our results from is representative of those we invited. We reweighted on years of relevant expertise and continent (North America, Europe, other), since our sample was biased towards European and more junior participants compared to the sampling frame (see Figure 1<\/a>). These weights are used in all results presented below. For a full list of variables we considered for reweighting, see Appendix G<\/a>. The appendix also compares key results with and without reweighting. The procedure does not systematically make results more conservative or extreme: reweighted GDP growth aggregate forecasts are more conservative (small-to-moderate differences) and LFPR forecasts more extreme (moderate-to-large differences). Even the larger differences, observed for longer-horizon LFPR forecasts, are well within aggregate uncertainty bounds.<\/p>\n\n\n\n

    For the remaining groups, sample sizes are smaller and reweighting\nhas not been applied; results for these groups should accordingly be\ninterpreted as characterizing the recruited samples rather than their\nbroader populations.<\/p>\n\n\n\n

    Aggregation<\/h4>\n\n\n\n

    We use two approaches for aggregating forecasts. For our main\nresults, we calculate a weighted (using the participant weights) median\nfor each percentile separately. We see these as our main results, since\nthis aggregates the noisy estimates of the underlying 50th, 10th, and\n90th percentiles to arrive at a more accurate estimate.<\/p>\n\n\n\n

    To understand forecaster disagreement, we additionally performed an alternative aggregation procedure for questions where the additional 10th and 90th percentile forecasts were available. We first fit a distribution to an individual participant\u2019s quantile forecasts (separately for unconditional forecasts and forecasts conditional on the rapid scenario) and then calculated a weighted average over the participants. This has the benefit of showing disagreement among participants. Section 4.3<\/a> contains more details about this procedure. For a comparison of these two aggregation approaches and evidence that the former may lead to more accurate forecasts, see Lichtendahl et al. (2013)<\/span>.51<\/a><\/sup><\/p>\n\n\n\n

    Rationale Analysis<\/h4>\n\n\n\n

    We used an LLM to extract the key drivers mentioned in rationales\naccompanying forecasts for the four main results (GDP, TFP, LFPR and\nwealth inequality). We supplemented these LLM-curated drivers with a\nmanual analysis of the rationales, checking for inconsistencies and\nadjusting the identified drivers to fully capture those offered in the\nrationales. We then used a second LLM agent to tag each occurrence of\nthe identified drivers, and then analyzed the frequency with which\ndifferent drivers were mentioned by various groupings of\nparticipants.<\/p>\n\n\n\n

    The rationales accompanying the policy results received similar\ntreatment. An initial partial human review identified the likely main\ndrivers and checked for systemic inconsistencies or question\nmisinterpretations; this was supplemented by LLM analyses to confirm and\nexpand upon these drivers, roughly quantify the frequency of driver\nmentions across participant groups, and extract rationales that best\nexemplified each driver.<\/p>\n\n\n\n

    See Appendix D.2<\/a> and Appendix I<\/a> for more details on rationale analysis.<\/p>\n\n\n\n

    3. Results<\/h2>\n\n\n\n

    3.1 AI Progress<\/h3>\n\n\n\n

    In the survey, we elicited participants\u2019 beliefs about the pace and scope of AI capability progress by the end of 2030 using three descriptive scenarios: slow, moderate, and rapid progress. These scenarios, described in full in Appendix H.2.1<\/a>, are intended as broad benchmarks rather than precise descriptions of how the world will unfold. Participants were asked to judge which scenario a panel of experts would judge as the best overall match to the true state of AI capabilities, recognizing that progress may be uneven across domains. Across all groups, we find that a majority of respondents place most of their probability weight on the moderate or rapid progress scenarios, expecting AI capabilities to advance significantly by 2030 even if real-world adoption lags.<\/p>\n\n\n\n

    Figure 2<\/a> summarizes the scenario descriptions and participants\u2019 probability forecasts. On average, economists assigned the highest probability to the moderate scenario (47.4%), with the remaining probability split between the slow (38.6%) and rapid (14.0%) scenarios. The rapid scenario was judged to be the least likely according to all groups, including AI policy and industry professionals (here grouped together under \u2018AI experts\u2019), superforecasters, and the general public. Superforecasters placed more weight on slow progress when compared to economists, assigning 45.0% to the slow scenario, and were also the most skeptical about the rapid scenario, assigning it 12.6% on average. In contrast, the general public\u2019s probabilities were more evenly distributed across scenarios (40.8% slow, 41.0% moderate, 18.1% rapid).<\/p>\n\n\n\n

    \"\"\/
    Figure 2: Average probability of AI progress scenarios by 2030.<\/strong> The average probability assigned by respondents in each group to the likelihood of a given AI progress scenario most closely describing the real world in 2030. The full scenario definitions can be found in Appendix H.2.1<\/a>.<\/figcaption><\/figure>\n\n\n\n

    Figure 3<\/a> summarizes the distribution of scenario probability forecasts across expertise groups. The slow scenario category demonstrates the highest variance among respondents, most pronounced among AI experts, whose interquartile range (IQR) stretches from 16.5\u201360.0%. Variance was slightly lower for the moderate scenario probabilities, with economists having the widest interquartile range, stretching from 31.5\u201364.5%. Variance across respondent groups was lowest for the rapid scenario, with interquartile ranges across groups spanning from 10 percentage points (superforecasters) to 20 percentage points (the general public).<\/p>\n\n\n\n

    \"\"\/
    Figure 3: Distribution of AI progress scenario forecasts.<\/strong> Vertical lines show the 25th, 50th, and 75th percentiles of the scenario forecasts. The 25th and 75th percentiles are labeled. The distribution is shown in two ways: as a histogram and as a density estimate.<\/figcaption><\/figure>\n\n\n\n

    3.2 Growth<\/h3>\n\n\n\n

    Gross Domestic\nProduct (GDP)<\/h4>\n\n\n\n

    Figure 4<\/a> compares historical measures of GDP growth (shown in black) to respondents\u2019 forecasts of annualized GDP growth over the five years leading to 2030 and 2050. Participants provided unconditional forecasts, reflecting their overall views, as well as forecasts conditional on the three AI progress scenarios described throughout the survey. Across all groups, unconditional forecasts for 2030 were clustered tightly around the current rates, spanning from 2.4% to 2.5%. By 2050, unconditional medians remain in roughly the same range, ranging from the economists\u2019 prediction of 2.5% to AI experts\u2019 prediction of 3.0%. However, disagreement between groups, as well as uncertainty within groups, slightly increases (as reflected in the shaded regions). These forecasts suggest that most respondents do not expect strong trends of acceleration in GDP growth absent major shifts.<\/p>\n\n\n\n

    In contrast, the rapid scenario is associated with significantly\nhigher expected GDP growth for both time horizons. For 2025\u20132029, the\nmedian economist predicted annual GDP growth to be around 3.3% (with\nmedian 10th and 90th percentile forecasts of 1.2% and 5.5%), and\nsuperforecasters showed a similar median (3.7%, 2.0\u20136.0%). AI experts\u2019\nrapid scenario median is slightly higher in this period, at 3.7%. The\ndivergence between groups is wider by the 2045\u20132049 period, with annual\nGDP growth reaching 3.5% for economists (1.0\u20137.0%), 4.0% for\nsuperforecasters (1.0\u20137.0%), 5.3% for AI experts (2.3\u20139.3%), and 4.5%\nfor the general public (2.5\u20137.1%). The moderate scenario falls between\nunconditional and rapid.<\/p>\n\n\n\n

    \"\"\/
    Figure 4: Forecasts for five-year annualized change in the Gross Domestic Product (GDP).<\/strong> Lines show the medians of 50th percentile forecasts across participants. Shaded regions span from the median 10th to the median 90th percentile forecast. The results for economists are reweighted to adjust for non-response bias (see Section 2.3<\/a>). See Appendix H.4.1<\/a> for question details and the source of the historical data.<\/figcaption><\/figure>\n\n\n\n

    Figure 5<\/a> summarizes uncertainty using pooled distributions, shown for the unconditional and rapid scenarios, for which participants provided 10th and 90th percentile predictions in addition to their best-guess 50th percentile predictions. Variance increases at the longer time horizon (see Table 10<\/a>), consistent with greater uncertainty about growth outcomes, and with this broadening happening especially under the rapid scenario. For example, economists\u2019 pooled distribution for GDP growth in 2030 under the rapid scenario spans 1.2%\u20136.4% (10th\u201390th percentile), compared to 0.7%\u20134.6% for the unconditional scenario. Similar patterns hold for other groups. Notably, the 90th percentiles of the 2030 and 2050 pooled distributions for AI experts reach the 10% winsorization cap (a bound applied to the distribution tails to prevent outlier forecasts from disproportionately skewing the pooled variance) and some experts forecast growth well above our cap. While superforecasters\u2019 medians once more fall close to economists\u2019, they assign higher weight to >10% growth in the rapid scenario, even in 2030.<\/p>\n\n\n\n

    \"\"\/
    Figure 5: Distribution of forecasts for five-year annualized change in Gross Domestic Product (GDP)<\/strong>. Distribution is pooled across participants to summarize the full distribution of participant beliefs. Tail mass outside of figure bounds shown as ball-and-stick at 0% and 10%, with numbers in boxes indicating the proportion of the pooled distribution that lies below 0% or above 10%. Interior points show 10th \/50th \/90th percentiles of the distribution. See Appendix H.4.1<\/a> for question details.<\/figcaption><\/figure>\n\n\n\n

    We compare the median of economists\u2019 50th percentile GDP forecasts to other forecasts in Figure 6<\/a>.52<\/a><\/sup> Our economist sample\u2019s estimate for 2025\u20132029 of 2.5% is higher than eight of the nine other forecasts we consider, which range from 1.9% to 2.1%. The only forecast that is higher is OMB\u2019s of 2.8%. When we consider 2045\u20132049, this paper\u2019s estimate of 2.5% is higher than both other estimates (1.3% and 1.6%).<\/p>\n\n\n\n

    \"\"\/
    Figure 6: Comparison of economist GDP growth forecasts to other forecasts.<\/strong> In this figure, we compare the weighted median forecast for economists, which is indicated by the red dashed line, to nine other forecasts for 2025\u20132029 (left panel) and to two other forecasts for 2046\u20132050 (right panel). We use the baseline (or analogous) estimates. Where yearly estimates are given, we average these estimates. Where a single estimate is given for a yearly range, we use this estimate. Exceptions are: 1) for 2025\u20132029 Federal Reserve, we use the \u201cLonger run\u201d value for 2029; 2) for 2025\u20132029 Conference Board, we use the 2028\u20132032 value for 2028 and 2029; 3) for 2025\u20132029 Deloitte, we linearly interpolate values for 2027\u20132029 based on the 2026 and 2030 values; 4) for 2025\u20132029 (2045-2049) OECD, we use the value for the range of 2025\u20132030 (2045\u20132050). The type of forecast (real GDP or potential real GDP) is indicated.<\/figcaption><\/figure>\n\n\n\n

    Compounded over decades, small differences in growth rates can produce large differences in prosperity; see Figure 7<\/a>. Consider the 3.5% growth rate in the rapid forecast possesses a doubling time of 20 years versus 28 years for the 2.5% growth in the unconditional forecast. Extrapolating to 2050, the rapid scenario produces a real GDP of $54.7T, 25% larger than $43.7T in the unconditional scenario.53<\/a><\/sup> This is roughly equivalent to the GDP difference between 2016 and today.<\/p>\n\n\n\n

    Towards the end of the forecast period, growth approaches the\npost-World War 2 economic boom. From 1951-1973, growth averaged 4.06%,\nclose to the rapid scenario; because population growth today is lower\nthan during the postwar era, this suggests the rapid scenario could\nmeaningfully exceed the per capita growth experienced from\n1951-1973.<\/p>\n\n\n\n

    \"\"\/
    Figure 7: Projected GDP trajectory under different scenarios.<\/strong> Based on economists\u2019 GDP growth forecasts; growth rates between 2030-2045 are linearly interpolated from 2025\u20132029 and 2045\u20132049 forecasts. Shaded areas are determined using 10th and 90th percentile forecasts for the unconditional and rapid-scenario forecasts. Note the logarithmic scale on the y-axis.<\/figcaption><\/figure>\n\n\n\n

    Total Factor\nProductivity<\/h4>\n\n\n\n

    As a separate measure of economic growth, participants were asked to provide forecasts for the annualized change in total factor productivity (TFP) over 5 years. Figure 8<\/a> compares participant forecasts to historical measures of this outcome. Median forecasts are homogenous across groups, clustering tightly for the unconditional scenario around 1.0% (superforecasters and the general public)\u20131.2% (economists and AI experts) for 2030 and 1.0%\u20131.7% for 2050, with economists at 1.5%.54<\/a><\/sup> Under the rapid scenario, forecasts roughly double relative to the unconditional baseline, growing to 1.9%\u20132.0% for 2030 and 2.2%\u20132.5% for 2050.<\/p>\n\n\n\n

    \"\"\/
    Figure 8: Forecasts for five-year annualized change in total factor productivity (TFP).<\/strong> Lines show medians of 50th percentile forecasts across participants. Because we elicited only 50th percentile predictions for TFP growth, this figure does not show uncertainty. The results for economists are reweighted to adjust for non-response bias (see Section 2.3<\/a>). See Appendix H.4.3<\/a> for question details and the source of the historical data.<\/figcaption><\/figure>\n\n\n\n

    Key\nfinding: Despite expecting significant AI progress, most unconditional\neconomic forecasts are close to historical trends<\/h4>\n\n\n\n

    The results above reveal an apparent tension: economists assign a 61.4% probability to moderate or rapid AI progress by 2030 (see Figure 2<\/a>), yet their unconditional GDP and TFP forecasts do not substantially depart from recent baselines. The median unconditional GDP forecast of 2.5% for both 2030 and 2050 is only marginally above the 2021\u20132025 baseline of 2.39%, and TFP forecasts of 1.2% for 2030 and 1.5% for 2050 similarly represent noticeable, but incremental, gains relative to the 2025 baseline of 0.97%. While it is true that these forecasts are not quite as conservative as they may first appear when considered in the context of the medium- and long-run projections we report in Figure 6<\/a>, the implied AI-driven productivity acceleration is modest by historical standards: the IT boom of the late 1990s arguably led to an additional 0.65 p.p. jump in annual TFP growth in the late 1990s,55<\/a><\/sup> and the economists\u2019 forecasts of TFP growth in our study imply an acceleration in growth from AI that is roughly half that size by 2030 and on par with it by 2050.<\/p>\n\n\n\n

    Economists\u2019 written rationales, however, lend insight into this tension. The most frequently cited reason why transformative technology would translate into only modest economic growth was uneven and time-lagged diffusion, with economists drawing on analogies to electrification, automobiles, and personal computers to argue that multi-decade lags routinely separate a general-purpose technology\u2019s arrival from its measurable productivity impact. Geopolitical, structural, and demographic headwinds\u2014including trade wars, climate change, an aging population, and declining immigration\u2014that could offset AI-driven gains were also cited, as were constraints on energy, chips, and data center construction. Collectively, these factors were deemed likely to cap the pace at which AI capabilities could be deployed regardless of how quickly they advance. Under the rapid scenario, the possibility that large-scale workforce exit, societal unrest, or existential risk could become drags on GDP was also considered. A deeper analysis of these rationales can be found in Appendix D.2.1<\/a> (GDP growth) and Appendix D.2.3<\/a> (TFP growth).<\/p>\n\n\n\n

    3.3 Labor Markets<\/h3>\n\n\n\n

    Overall Impact<\/h4>\n\n\n\n

    Our main metric for understanding labor market impacts is the Labor\nForce Participation Rate (LFPR). It measures the fraction of the adult\npopulation participating in the labor force, either through employment\nor active jobseeking. The LFPR has historically been relatively stable,\nmostly shaped by demographic trends, and is much less cyclical than the\nunemployment rate. It captures discouragement and long-term exits from\nthe labor force.<\/p>\n\n\n\n

    Forecasts for the LFPR are shown in Figure 9<\/a>. Economists expect the decreasing trend of the last two decades to continue. By the beginning of 2030, the median economist forecasts LFPR of 61.0%, a decrease from the January 2025 rate of 62.6%.56<\/a><\/sup> This forecast has already been affected by economists\u2019 expectations about AI capabilities: conditional on the slow scenario, the median forecast is 61.5%, 0.5 p.p. higher. However, when asked to assume the rapid scenario, their forecast drops to 59.3%. If this were to occur, it would be the first time since the 1970s that a monthly LFPR reading fell below 60% in the U.S., and not since the 1960s has the annual rate fallen below 60%. The potential of AI to influence the LFPR\u2014a difference between the slow and rapid scenarios of 2.2 p.p.\u2014rivals the dip seen during the Covid-19 pandemic (from 63.3% in February 2020 to 60.1% in April 2020), but would reflect a more permanent impact on the labor force. By the beginning of 2050, economists expect a LFPR of 58.3% in the unconditional scenario57<\/a><\/sup> and 55.0% when assuming the rapid scenario, with roughly half of that decline\u2014equivalent to around 10 million lost jobs\u2014likely attributable to AI rather than demographics and other non-AI trends.58<\/a><\/sup><\/p>\n\n\n\n

    \"\"\/
    Figure 9: Forecasts for the labor force participation rate (LFPR).<\/strong> Lines show medians of 50th percentile forecasts across participants. Shaded regions span from the median 10th to the median 90th percentile forecast. The results for economists are reweighted to adjust for non-response bias (see Section 2.3<\/a>). See Appendix H.4.4<\/a> for question details and the source of the historical data.<\/figcaption><\/figure>\n\n\n\n

    Directionally, all groups agree: faster progress implies lower LFPR\nvalues. The general public is slightly more optimistic than economists,\nwith unconditional median forecasts of 62.0% and 60.0% for 2030 and\n2050, respectively. Superforecasters are slightly more pessimistic in\nthe long term: unconditionally, they give a median forecast of 58.2% for\n2050 and 54.6% assuming the rapid scenario.<\/p>\n\n\n\n

    In the written rationales economists offered to explain their LFPR forecasts, the most frequently mentioned themes were AI-driven job substitution and loss, reallocation and reskilling, demographic changes, and historical LFPR baselines. A deeper analysis of the rationales can be found in Appendix D.2.4<\/a>.<\/p>\n\n\n\n

    Key\nfinding: Unconditional consensus masks significant uncertainty about\nrapid scenario outcomes<\/h4>\n\n\n\n

    The median LFPR forecasts reported above may give a misleading\nimpression of expert agreement. Across groups, unconditional forecasts\ncluster in a narrow band\u2014economists at 61.0% for 2030 and 58.3% for\n2050, with other groups nearby\u2014and the decline from the current 62.6%\nbaseline looks orderly. But when we examine the distributions underlying\nthese forecasts, particularly under the rapid scenario, the range of\nplausible outcomes expands.<\/p>\n\n\n\n

    Figure 10<\/a> shows this uncertainty using pooled distributions for the unconditional and rapid scenarios, for which participants provided 10th and 90th percentile predictions in addition to their best-guess 50th percentile predictions. The pooled distribution represents both within-forecaster uncertainty and between-forecaster disagreement. The variance of the distribution (shown in detail in Table 10<\/a>) increased at the longer time horizon and in the rapid scenario, consistent with the pattern observed for GDP growth. While economists\u2019 pooled distribution for LFPR in 2030 spans 56.9\u201365.2% (10th\u201390th percentile) in the unconditional scenario, it widens to 53.1\u201364.4% in the rapid scenario. Variance increases even more in the 2050 rapid scenario, spanning 44.8\u201364.7%, with a significant amount of probability mass (14.3%) below the 45% winsorization floor (note that the 10th percentile falls below the winsorization floor), indicating that, in aggregate, economists predict a non-trivial chance of extreme low-LFPR outcomes. An LFPR at or near the median economist\u2019s 10th percentile forecast for 2050 in the rapid scenario would indicate tens of millions had left the workforce and reflect an unprecedented shift in the structure of the economy.<\/p>\n\n\n\n

    \"\"\/
    Figure 10: Distribution of forecasts for labor force participation rate (LFPR).<\/strong> Distribution is pooled across participants to summarize the full distribution of participant beliefs. Tail mass outside of figure bounds shown as ball-and-stick at 45% and 70%, with numbers in boxes indicating the proportion of the pooled distribution that lies below 45% or above 70%. Interior points show 10th \/50th \/90th percentiles of the distribution. See Appendix H.4.4<\/a> for question details.<\/figcaption><\/figure>\n\n\n\n

    This heightened uncertainty in the rapid scenario is also visible in\nGDP forecasts, where the 90th percentile of the pooled economist\ndistribution under the rapid 2050 scenario reaches 8.43%, suggesting\nthat experts have narrow priors for a world in which AI augments the\neconomy incrementally, but far less confidence about what will happen if\nthe technology proves truly transformative.<\/p>\n\n\n\n

    Impacts by Sector<\/h4>\n\n\n\n

    Economists\u2019 median forecasts for the sizes of different sectors as shares of the labor force are shown in Figure 11<\/a>. In the unconditional scenario, economists expect the share of business and analytical (\u201cwhite-collar\u201d) roles to continue to rise slowly, reaching 21.0% in 2030 and 22.0% in 2050 compared to a 2025 baseline of 20.4%. The share of care and service workers is forecast to increase more rapidly, reaching 48.0% in 2030 and 52.5% in 2050 in the unconditional scenario, while skilled trade and industrial (\u201cblue-collar\u201d) occupations would fall to 12.5% in 2030 and 11.0% in 2050.<\/p>\n\n\n\n

    \"\"
    Figure 11: Economists\u2019 forecasts for different groups of occupations as shares of the labor force.<\/strong> Areas show the median 50th percentile forecast. The \u2018Other\u2019 category is derived by subtracting the sum of the other sectors\u2019 median forecasts from 100%. It consists primarily of public sector and agricultural workers. Labeled historical values correspond to the beginning of 2025. See Appendix H.4.7<\/a> for question details and the source of the historical data.<\/figcaption><\/figure>\n\n\n\n

    In the rapid scenario, the growth in white-collar occupations would\nstall, with their share remaining flat at around 20.0%\u201321.0% in 2030 and\n2050. Care and service occupations are forecast to see larger increases\ncompared to the unconditional scenario (reaching 57.1% in 2050), while\nblue-collar occupations would sharply decline, with a median share of\n8.0% by 2050\u2014a historical low.<\/p>\n\n\n\n

    Directionally, other groups largely agree with these aggregate economist forecasts, although there are some areas of disagreement. Under the rapid scenario, by 2050 superforecasters and AI experts expect a decline in white-collar jobs to 16.0% and 18.0%, respectively. AI experts predict a much smaller increase in the share of care and service occupations, growing to only 49.8% by 2050 in the rapid scenario, while the general public is the only group expecting a decline in this sector in the rapid scenario, forecasting a share of 42.0%. The results for other groups are shown in Appendix D.2<\/a>.<\/p>\n\n\n\n

    Impacts by\nOccupation<\/h4>\n\n\n\n

    Each survey participant was randomly shown 10 of the 43 International Standard Classification of Occupations (ISCO-08) sub-major occupation groups and asked to rank them by predicted percent change in employment from 2025 to 2030, indicating whether each would grow or decline relative to a 0% marker, both unconditionally and conditional on the rapid scenario.59<\/a><\/sup><\/p>\n\n\n\n

    Figure 12<\/a> plots the percent of respondents who expect there to be growth in employment in a given occupation between the beginning of 2025 and the beginning of 2030. In the unconditional scenario, occupations economists expect will see the strongest employment growth are personal service workers, personal care workers, health professionals, and military occupations\u2014roles characterized by in-person physical presence, human interaction, or security functions that are difficult to automate. At the bottom of the ranking, economists expect declines in general and keyboard clerks, other clerical support workers, stationary plant and machine operators, and assemblers\u2014routine cognitive and manual roles thought to be particularly vulnerable to technological displacement. Point estimates for blue-collar occupations generally fell below the 50% threshold, indicating that a majority of economists placed these occupations in the job-loss category. This pattern was more pronounced than for white-collar occupations, which clustered closer to or above 50%, and for service occupations, which spanned a wide range but included some of the survey\u2019s most optimistic forecasts.<\/p>\n\n\n\n

    Conditioning on the rapid scenario did not significantly alter these\nrankings. While most occupation groups shifted modestly leftward\u2014that\nis, fewer respondents predicted positive growth\u2014the differences between\nthe two scenarios were not statistically distinguishable (also note the\nrelatively small sample size). It is worth noting that the unconditional\nscenario should not be read as a counterfactual with no LLM exposure;\nrespondents were presumably already incorporating LLM effects into their\nunconditional predictions.<\/p>\n\n\n\n

    \"\"\/
    Figure 12: The fraction of economists predicting a positive change in employment for each occupation between the beginning of 2025 and the beginning of 2030.<\/strong> Gray points correspond to the unconditional scenario, while red points correspond to the rapid scenario. Dashed lines indicate statistically non-significant results at the 5% level.<\/figcaption><\/figure>\n\n\n\n

    In Appendix D.2.8<\/a>, we compare the fraction of economists who predict each occupation will experience growth with the measure of AI exposure from Eloundou et al. (2024)<\/span>.60<\/a><\/sup>,<\/sup>61<\/a><\/sup>,<\/sup>62<\/a><\/sup> We do not observe a relationship between our results and AI exposure. We also compare the AI exposure measure to the difference in the fraction predicting positive growth between the rapid and unconditional scenarios, and again find no relationship.<\/p>\n\n\n\n

    3.4 Economic Inequality<\/h3>\n\n\n\n

    We measure economic inequality as the fraction of wealth held by the\n10% wealthiest households. This metric has had a moderate upward trend\nsince the 1980s, reaching 71.2% in 2023. Like the LFPR, this metric has\nbeen relatively stable historically, ranging from 62.7% (in 1985) to\n73.2% (in 2013-2014).<\/p>\n\n\n\n

    \"\"\/
    Figure 13: Forecasts for wealth held by the 10% wealthiest households.<\/strong> Lines show medians of 50th percentile forecasts across participants. Shaded regions span from the median 10th to the median 90th percentile forecast. The results for economists are reweighted to adjust for non-response bias (see Section 2.3<\/a>). See Appendix H.4.9<\/a> for question details and the source of the historical data.<\/figcaption><\/figure>\n\n\n\n

    These historical values, as well as forecasts by different groups, are shown in Figure 13<\/a>. Economists expect wealth inequality to increase, with a median forecast of 73.2% in 2030 and 75.0% in 2050. In the rapid scenario, the increase is faster: 75.0% in 2030 and 80.0% in 2050. Our secondary measure of inequality, the labor share of income, leads to similar conclusions: inequality is increasing, and faster AI progress is predicted to lead to greater increases in inequality. These results appear in Appendix D.2.10<\/a>.<\/p>\n\n\n\n

    Other participant groups agree with economists in directional terms.\nHowever, superforecasters give notably more conservative forecasts,\nespecially for longer-term outcomes. For 2050, their unconditional\nforecast is 74.5%, and the forecast conditional on the rapid scenario is\n75.0%.<\/p>\n\n\n\n

    Although these forecasts indicate inequality is set to increase, economists also predict that median household income will continue to increase (see Appendix Figure 52<\/a>). Since we are measuring median household income in real terms (adjusted for inflation), this reflects an increase in the purchasing power of the median household. In addition, according to these forecasts, faster AI progress will lead to larger increases. Setting aside that we are measuring different concepts with these two measures (income vs. wealth), a possible interpretation of these results is that while both the top-10% of households and the median household will see gains as a result of faster AI progress, these gains may tend to disproportionately accumulate to the top-10%.<\/p>\n\n\n\n

    In the written rationales economists offered to explain their wealth inequality forecasts, the most frequently mentioned themes were the shift from labor to capital, historical wealth inequality trends, redistribution and tax policy, ownership of AI infrastructure and IP, and uncertainty around 2050 outcomes. A deeper analysis of these rationales can be found in Appendix D.2.9<\/a>.<\/p>\n\n\n\n

    Key\nfinding: If the rapid scenario materializes, economists expect\nsignificant economic shifts, but not the transformative acceleration\nsome have predicted.<\/h4>\n\n\n\n

    The wealth inequality forecasts under the rapid scenario describe a\nU.S. economy that would be substantially more unequal than today\u2014but not\nunrecognizably so. The median economist forecast of 80.0% of national\nwealth held by the top 10% in 2050 would represent the highest\nconcentration since the late 1930s, and yet that is a level the U.S. has\nreached before, albeit under very different technological and\ninstitutional conditions. This result fits a broader pattern we observe\nunder the rapid scenario in which economists forecast large shifts\u2014GDP\ngrowth of 3.5%, LFPR falling to 55.0%, wealth inequality at 80.0%\u2014but\nshifts that still have historical parallels, such as GDP growth\npost-WWII, or the LFPR before women entered the workforce en masse, or\npre-WWII inequality.<\/p>\n\n\n\n

    These forecasted impacts stand in marked contrast to the \u201ctransformative\u201d economic impacts proposed by some technologists (Amodei, 2024)<\/span> and highlighted as possible by some economists (Brynjolfsson, Korinek, and Agrawal, 2025).63<\/a><\/sup> Even in this rapid scenario where AI models can surpass humans at most cognitive tasks and robots can surpass humans at most physical tasks, economists do not forecast\u2014and no other group forecasts\u2014anything like the tenfold increase in economic growth to around 30% discussed in the literature.64<\/a><\/sup><\/p>\n\n\n\n

    3.5 Other Outcomes<\/h3>\n\n\n\n

    In addition to the main survey questions we report on above, we elicited forecasts on eleven additional measures, the results of which are summarized below. The full results can be found in Appendix D.2<\/a>.<\/p>\n\n\n\n

    Overall, economists project that unemployment will remain remarkably\nstable and that real median household incomes will continue to grow, but\nthat the nature of work will change due to increasing use of AI. Labor\u2019s\nshare of economic output is expected to drop.<\/p>\n\n\n\n

    Key Results<\/h4>\n\n\n\n
    Labor Productivity.<\/h5>\n\n\n\n

    Economists forecast only modest increases in the annualized change in\nlabor productivity. For 2030, their median forecasts are 2.0% for the\nunconditional scenario, 2.0% for the slow, 2.5% for the moderate, and\n3.2% for the rapid scenario\u2014this compared to a 2025 baseline of 1.94%.\nFor 2050, these expectations shift slightly to 2.5% (unconditional),\n2.0% (slow), 3.0% (moderate), and 4.0% (rapid). AI experts, however, are\nsignificantly more optimistic about the rapid scenario in 2050,\nforecasting a 5.0% annualized change in 2050.<\/p>\n\n\n\n

    Unemployment Rate.<\/h5>\n\n\n\n

    Economists expect the overall unemployment rate to remain relatively\nstable, even under the rapid scenario. For 2030, they forecast 5.0% for\nthe unconditional, 5.0% for the slow, 5.0% for the moderate, and 6.0%\nfor the rapid. For 2050, their estimates are 5.0% (unconditional), 5.0%\n(slow), 6.0% (moderate), and 6.0% (rapid). AI policy and industry\nprofessionals, however, project higher long-term unemployment under the\nrapid scenario: for 2050, they forecast 8.0%.<\/p>\n\n\n\n

    Youth Unemployment Rate.<\/h5>\n\n\n\n

    Even for 20-to-24-year-old workers, economists expect the\nunemployment rate to remain relatively stable. They forecast 2030 youth\nunemployment rates of 9.5% for the unconditional scenario, 9.0% for the\nslow, 10.0% for the moderate, and 11.0% for the rapid\u2014predictions that\nfall well within the historical range. By 2050, their forecasts drop\nslightly across the board to 9.0% (unconditional), 8.2% (slow), 9.8%\n(moderate), and 10.0% (rapid). As with the overall employment rate, AI\npolicy and industry professionals were more pessimistic under the rapid\nscenario, predicting 11.4% for 2050.<\/p>\n\n\n\n

    Labor Share.<\/h5>\n\n\n\n

    Economists project a slight downward trend in the share of economic\noutput going to workers, particularly if AI advances quickly. For 2030,\nthey forecast 54.3% for unconditional scenario, 55.0% for the slow,\n54.0% for the moderate, and 52.0% for the rapid, compared to a 2025\nbaseline of 55.48%. For 2050, this drops to 50.0% (unconditional), 52.0%\n(slow), 50.0% (moderate), and 45.0% (rapid). AI experts, however, expect\na drastic collapse in the labor share under the rapid 2050 scenario,\nforecasting just 40.0%.<\/p>\n\n\n\n

    Median Household Income.<\/h5>\n\n\n\n

    Economists forecast steady real income growth. For 2030, their median\nestimates are $83,967 for the unconditional scenario, $83,046 for the\nslow, $85,000 for the moderate, and $87,000 for the rapid, compared to a\n2023 baseline of $80,610. By 2050, their forecasts rise to $93,175\n(unconditional), $91,864 (slow), $95,142 (moderate), and $100,000\n(rapid). AI experts are markedly more pessimistic than economists in the\nrapid 2050 scenario, expecting a median household income of $95,000.<\/p>\n\n\n\n

    Life Satisfaction.<\/h5>\n\n\n\n

    Economists do not expect AI progress to drastically alter average life satisfaction in the U.S. For 2030, forecasts on the Cantril ladder65<\/a><\/sup> are 6.6 for the unconditional scenario, 6.7 for the slow, 6.6 for the moderate, and 6.5 for the rapid, compared to a 2024 baseline of 6.72. In 2050, forecasts are 6.7 (unconditional), 6.7 (slow), 6.5 (moderate), and 6.5 (rapid). Forecasts across all groups and scenarios are remarkably clustered, generally hovering between 6.4 and 6.8 with very little deviation. Superforecasters are an exception: for 2050, they forecast an average life satisfaction of 7.0, independent of the AI scenario.<\/p>\n\n\n\n

    Work Hours Assisted by\nGenerative AI.<\/h5>\n\n\n\n

    Economists foresee significant adoption of generative AI. In 2030,\nthey estimate the percentage of assisted work hours will be 10.1% for\nthe unconditional scenario, 8.0% for the slow, 12.9% for the moderate,\nand 24.2% for the rapid, compared to a 2024 estimate of 3.35%. By 2050,\nthey expect this to surge to 40.0% (unconditional), 25.0% (slow), 44.7%\n(moderate), and 62.0% (rapid). AI experts also project high utilization\nin the rapid scenario, hitting 60.0% in 2050. By comparison,\nsuperforecasters are much more conservative, predicting just 33.0% in\n2050 under the rapid scenario.<\/p>\n\n\n\n

    AI Electricity Consumption.<\/h5>\n\n\n\n

    Economists predict growing energy demands for AI. Their 2030 median\nforecasts for the share of U.S. electricity consumption used by AI are\n4.0% for the unconditional scenario, 2.3% for the slow, 4.9% for the\nmoderate, and 7.4% for the rapid, compared to a 2024 baseline estimate\nof 1%. For 2050, these jump to 8.0% (unconditional), 5.0% (slow), 8.3%\n(moderate), and 15.0% (rapid). AI experts and superforecasters\nanticipate somewhat higher electricity usage in 2050 under the rapid\nscenario, both forecasting 19.5%.<\/p>\n\n\n\n

    3.6 Policy Responses<\/h3>\n\n\n\n

    We asked survey participants to predict the marginal impact of six policy proposals on GDP growth and the LFPR, under both the unconditional and rapid scenarios, and for 2030 and 2050. Participants were instructed to estimate the effect of each policy in isolation, setting aside any consideration of conditions, political or otherwise, under which each policy might be adopted. We also asked respondents to indicate their support for implementing each policy without conditioning on any specific AI progress scenario. Finally, we asked respondents to estimate the probability that each policy (or a similar policy, as judged by an economist panel) would be implemented by the U.S. by the end of 2026. We present a summarized version of each policy, along with our key findings, below. Findings reported are for impacts on GDP growth and the LFPR in 2030. The full policy descriptions, results, and rationale analyses can be found in Appendix D.3<\/a>.<\/p>\n\n\n\n

      \n
    1. Retraining Support: Offers displaced workers in high-automation-risk industries up to $25,000 per year (for up to two years) in training credits, career counseling, and relocation assistance, funded by a small payroll tax.<\/em><\/li>\n\n\n\n
    2. Modernized Unemployment Insurance: Expands unemployment benefits to 75% of prior salary for up to 18 months for automation-displaced workers, with added wage loss insurance and streamlined administration, funded by higher employer payroll taxes.<\/em><\/li>\n\n\n\n
    3. Universal Basic Income: Gives every American adult $1,000 per month unconditionally, funded by a 15% VAT on all goods and services.<\/em><\/li>\n\n\n\n
    4. Manhattan Project for AI: Deploys 0.4% of U.S. GDP annually66<\/a><\/sup> in federal spending to accelerate AI research and infrastructure development, funded by a 0.7% VAT.<\/em><\/li>\n\n\n\n
    5. Compute Tax: Taxes heavy AI electricity users $50 per MWh above a set threshold and redistributes the revenue to consumers as stimulus checks.<\/em><\/li>\n\n\n\n
    6. Job Guarantee Program: Guarantees a federally funded job paying at least $15 per hour (indexed to inflation) with full benefits to any adult who wants one, funded by a 0.5% VAT.<\/em><\/li>\n<\/ol>\n\n\n\n

      Key Results<\/h4>\n\n\n\n

      The Manhattan Project for AI had the highest projected GDP impact, and the Job Guarantee Program had the highest projected LFPR impact.<\/strong> A comparison of median marginal impacts across policies is shown in Figure 14<\/a> for GDP growth and in Figure 15<\/a> for the LFPR. However, economists expressed exceptionally low support for the Job Guarantee Program (13.7%), and only middling support for the Manhattan Project for AI (55.8%), suggesting that broader economic and societal considerations, beyond headline GDP and LFPR projections, are the primary drivers of economists\u2019 policy preferences.<\/p>\n\n\n\n

      \"\"\/
      Figure 14: The median marginal impact of the policies on GDP growth in 2030 (5-year annualized change), according to economists.<\/strong> Marginal impact is relative to none of the policies being implemented; the legend shows the predicted GDP growth conditional on this baseline. Several policies have zero predicted impact relative to baseline. See Appendix H.5<\/a> for question details.<\/figcaption><\/figure>\n\n\n\n
      \"\"\/
      Figure 15: The median marginal impact of the policies on the LFPR in 2030, according to economists.<\/strong> Marginal impact is relative to none of the policies being implemented; the legend shows the predicted LFPR conditional on this baseline. Several policies have zero predicted impact relative to baseline. See Appendix H.5<\/a> for question details.<\/figcaption><\/figure>\n\n\n\n

      Retraining Support was the consensus favorite among economists.<\/strong> It drew 71.8% support and only 19.9% opposition (with the remainder registering as \u201cunsure\u201d), received modestly positive forecasts on both GDP (+0.1 p.p.) and LFPR (+0.5 p.p.) for the unconditional scenario, and attracted little concern about its funding mechanism\u2014a small payroll tax. Retraining support was also broadly popular with AI industry and policy professionals, superforecasters, and the general public. The fraction of participants in each group supporting the implementation of the different policies is shown in Figure 16<\/a>.<\/p>\n\n\n\n

      \"\"\/
      Figure 16: Proportion of respondents in each group supporting implementation of each policy.<\/strong> Support reflects responses of \u2019Yes, with at most minor alterations\u2019 to the question \u2019Do you think [policy] should be implemented?\u2019 Respondents could also answer \u2019No\u2019 or \u2019Unsure.\u2019 Policy support was elicited without conditioning on any specific AI progress scenario. See Appendix H.5<\/a> for question details.<\/figcaption><\/figure>\n\n\n\n

      Support for the six policies varied significantly between groups.<\/strong> The Job Guarantee Program produced the survey\u2019s largest economist-general public divergence: whereas only 13.7% of economists supported it, 57.1% of the general public did. In general, economists\u2019 normative support tilted toward more incremental, targeted policies, while the general public was willing to entertain broader interventions.<\/p>\n\n\n\n

      Universal Basic Income was predicted to have no impact on GDP across both the unconditional and rapid scenarios, and carried a substantial projected LFPR decline for both scenarios.<\/strong> A plurality of economists (38.2%) opposed it\u2014the second-highest opposition rate for any policy in the survey\u2014citing labor supply disincentives and the drag of a 15% VAT as primary concerns; by contrast, a plurality of the general public (47.9% for versus 30.3% opposed) supported it.<\/p>\n\n\n\n

      Modernized Unemployment Insurance was forecast to have zero impact on both GDP and LFPR.<\/strong> However, it still drew a robust 62.3% support from economists, with some viewing it as a social stabilizer independent of its predicted macroeconomic effects.<\/p>\n\n\n\n

      All cohorts assigned low probabilities to real-world implementation of any policy by the end of 2026.<\/strong> Economists\u2019 median estimates ranged from sub-1% (UBI, Job Guarantee) to 15.0% (Manhattan Project for AI), reflecting a pervasive view that near-term political obstacles are considerable, regardless of a policy\u2019s perceived merit.<\/p>\n\n\n\n

      Many respondents in all cohorts view rapid AI progress as increasing the likelihood of policy enactment in the long term.<\/strong> This is particularly true for safety-net and redistributive policies like UBI, which some respondents\u2014in their rationales\u2014predicted would become inevitable under the rapid scenario. Others indirectly implied the rapid scenario would trigger stabilizing policy responses by arguing near-term action is unlikely because<\/em> AI has not yet caused enough disruption.<\/p>\n\n\n\n

      Under the rapid scenario, economists expressed heightened uncertainty about whether labor market policies could keep pace with automation.<\/strong> Several noted that workforce interventions might be overwhelmed by structural displacement, and that policies designed for a normal economy could prove inadequate in a world where AI is advancing faster than humans can retrain.<\/p>\n\n\n\n

      4. Drivers of Disagreement<\/h2>\n\n\n\n

      4.1 Framing the Debate<\/h3>\n\n\n\n

      Debate on the future economic impacts of AI can largely be reduced to\ntwo questions:<\/p>\n\n\n\n

        \n
      1. Will AI capabilities progress meaningfully, such that AI systems\nare capable of completing a large quantity of economically meaningful\nwork?<\/li>\n\n\n\n
      2. If this progress in capabilities occurs, what will happen to\nimportant economic indicators?<\/li>\n<\/ol>\n\n\n\n

        In this section, we emphasize disagreement<\/em> on these two\nquestions. Specifically, we ask: is disagreement about the path for\neconomic indicators driven by disagreement on AI capabilities progress,\nor disagreement on the effects of capabilities progress on economic\nindicators? Cunningham (2025)<\/span> asks\nthis question and concludes:<\/p>\n\n\n\n

        \n

        The disagreement is about the AI, not about the economics. The primary reason for the disagreement seems to be about the future rate of AI capabilities progress, not about the more directly economic questions such as (1) the current economic impact of AI; (2) the rate of diffusion & adoption over time; (3) the substitutability between AI-produced and human-produced services.<\/em>67<\/a><\/sup><\/p>\n<\/blockquote>\n\n\n\n

        Below, we develop a quantitative approach and reach a different conclusion. When forecasting the future path of key economic indicators, disagreement centered on the economic impacts of AI capabilities progress<\/em>, rather than the degree to which progress will be made. The rationale analyses in Appendix D.2<\/a> dovetail with this conclusion. They suggest the factors economists weigh most heavily\u2014historical base rates, adoption lags, demographics, policy responses, macroeconomic headwinds, and structural views on how economies absorb technology\u2014are shaped more by priors about economics and institutions than by the specific AI capability scenario assumed. These rationales also support our assumption that respondents shared a broadly similar understanding of what each scenario entailed, although we discuss the limits of this assumption at the end of this section. We focus our discussion in this section on economists, but we include similar analyses for all other respondent groups in Appendix B<\/a>.<\/p>\n\n\n\n

        4.2 Disagreement on Capabilities Progress<\/h3>\n\n\n\n

        First, we ask how much disagreement there is on AI capabilities progress; see Table 1<\/a>. While the average economist assigns a probability of 14.0% (median 10.0%) to the rapid scenario, the bottom quartile of economists gives the rapid scenario a probability of 6.7% or less, while the top quartile gives forecasts of 20.0% or higher. For the slow scenario, the interquartile range is 16.8% to 60.0%, compared to a mean of 38.6% (median 38.3%). Accordingly, average and median forecasts mask substantial disagreement about the progress of AI capabilities.<\/p>\n\n\n\n

        Group<\/td>Scenario<\/td>Mean<\/td>25th percentile<\/td>50th percentile<\/td>75th percentile<\/td><\/tr>
        Economists<\/td>Slow<\/td>38.6<\/td>16.8<\/td>38.3<\/td>60.0<\/td><\/tr>
        Economists<\/td>Moderate<\/td>47.4<\/td>31.5<\/td>50.0<\/td>64.5<\/td><\/tr>
        Economists<\/td>Rapid<\/td>14.0<\/td>6.7<\/td>10.0<\/td>20.0<\/td><\/tr><\/tbody><\/table><\/div>
        Table 1:<\/strong> AI Progress Scenario Probabilities: Economists<\/figcaption><\/figure>\n\n\n\n

        4.3 Total Variance in Possible Outcomes<\/h3>\n\n\n\n

        Measuring total variance.<\/h4>\n\n\n\n

        The disagreement on AI capabilities progress documented in the\nprevious section, while extensive, is not necessarily the primary driver\nof disagreement between forecasters on their unconditional forecasts of\neconomic outcomes. In order to answer our question\u2014\u201cis disagreement\nabout the path for economic indicators driven by disagreement on AI\ncapabilities progress, or disagreement on the effects of capabilities\nprogress on economic indicators?\u201d\u2014we now assess the total<\/em>\nvariance in possible outcomes.<\/p>\n\n\n\n

        On select questions, we ask respondents to express their uncertainty in the form of quantile forecasts for the unconditional case and the rapid scenario.68<\/a><\/sup> We fit a distribution to each respondent\u2019s forecasts to fully characterize their beliefs.69<\/a><\/sup> Given the mean and variance for the unconditional and the rapid scenario distributions, as well as the forecasted probability of the rapid scenario, we can obtain the mean and variance for a bundled slow\/moderate scenario.70<\/a><\/sup> It is possible that the variance of the unconditional outcome distribution is too low\u2014and the probability placed on the rapid scenario too high\u2014to yield a coherent conditional distribution for the bundled slow\/moderate scenario.71<\/a><\/sup> The proportion of incoherent distributions72<\/a><\/sup> range from 11% to 21% across questions and time horizons for economists. Other groups, notably the general public, tend to have higher proportions of incoherent distributions. Superforecasters, on the other hand, tend to have lower rates of incoherence. We summarize the proportion of incoherent distributions in Table 8<\/a>. The incoherent distributions were removed from this analysis.<\/p>\n\n\n\n

        With forecaster-level distributions for the slow\/moderate scenario, rapid scenario, and unconditional case, we create \u2018pooled\u2019 distributions by taking a mixture distribution over forecasters, using our derived weights to govern selection probabilities across forecasters in the unconditional case.73<\/a><\/sup> In effect, we are considering a hierarchical mixture: we first select a scenario according to the aggregated scenario probabilities. Within each scenario, we then select a forecaster. Lastly, we can then draw an outcome from this forecaster-scenario distribution. Accordingly, these distributions represent the full variation in expert views, rather than a distribution created with calibration in mind. Indeed, this pooled distribution is not necessarily the optimal way to aggregate forecasts in terms of forecasting accuracy. For example, Ranjan and Gneiting (2010)<\/span> show that combining well-calibrated forecasts in this fashion yields forecasts that are miscalibrated.74<\/a><\/sup> Similar results are reported by Lichtendahl et al. (2013)<\/span>.75<\/a><\/sup> Nevertheless, approaches that aim to generate calibrated aggregate distributions will collapse disagreement between forecasters and are thus ill-suited to the questions posed above. In summary, our approach captures four sources of variation:<\/p>\n\n\n\n

          \n
        1. Within-scenario variance: how much does the outcome vary within<\/em> a given scenario?<\/li>\n\n\n\n
        2. Between-scenario variance: how much does the expected<\/em> outcome vary between<\/em> scenarios?<\/li>\n\n\n\n
        3. Within-forecaster uncertainty: how uncertain is a given\nforecaster about the outcome?<\/li>\n\n\n\n
        4. Between-forecaster disagreement: how much do expected<\/em> outcomes differ between<\/em> forecasters?<\/li>\n<\/ol>\n\n\n\n

          We quantify these four sources of variance with a hierarchical law of total variance decomposition. We first decompose the total variation in the outcome into between- and within-scenario components, corresponding to the first two items above. Next, we further decompose each of these two components into within- and between-forecaster components, yielding the four components of the total variance. Lastly, on the within-scenario variance branch, we can assess how much the rapid and slow\/moderate scenarios contribute to each component.76<\/a><\/sup> See Appendix B<\/a> for the associated derivations.<\/p>\n\n\n\n

          Understanding\ntotal variance in GDP forecasts: raw data.<\/h4>\n\n\n\n

          We now describe in detail a decomposition of 2030 GDP growth forecasts, focusing only on economists. After this discussion, we summarize the results for the other questions. Results for all groups, questions, and time horizons are in Appendix B<\/a>.<\/p>\n\n\n\n

          First, before imposing any assumptions, we can explore the quantile\nforecasts directly. The median of 50th percentile forecasts\nin the unconditional scenario is 2.5% (IQR: 2.0, 3.0). We can compare\nthese forecasts to those for the three scenarios: the slow scenario\nmedian is 2.0% (IQR: 1.5, 2.2), the moderate scenario median is 2.6%\n(IQR: 2.2, 3.0), and the rapid scenario median is 3.3% (IQR: 2.9,\n4.5)<\/p>\n\n\n\n

          Next, we can consider the quantile forecasts given in the\nunconditional and rapid scenarios. For the 10th percentile,\nthe unconditional median is 1.0% (IQR: 0.0, 1.2) and the rapid scenario\nmedian is 1.2% (IQR: 0.5, 2.6). For the 90th percentile, the\nunconditional median is 4.0% (IQR: 3.5, 5.0) and the rapid scenario\nmedian is 5.5% (IQR: 4.0, 7.0).<\/p>\n\n\n\n

          While there are noticeable differences in the central tendencies\nacross scenarios, the 10th and 90th percentile\nforecasts demonstrate that there remains substantial overlap in outcomes\nacross the scenarios; within-scenario variance is a critical driver\nof total variance<\/em>.<\/p>\n\n\n\n

          Nevertheless, the above decomposition does not assess how much of the\nvariance within a scenario owes to forecasters\u2019 uncertainty, and how\nmuch to disagreement between forecasters. The variance decomposition\nbelow addresses this question directly.<\/p>\n\n\n\n

          Understanding\nvariance in GDP forecasts: fitted distributions.<\/h4>\n\n\n\n

          We next assume functional forms for forecaster-level outcome distributions, as described above. We pool these distributions across forecasters to assess total variance in the unconditional and rapid scenarios. The result is shown in Figure 5<\/a>. In the unconditional pooled distribution, the median growth rate is 2.5%, with an interquartile range of 1.6% to 3.4%. In the rapid scenario, the median growth rate is 3.4%, with an interquartile range of 2.3% to 4.7%. Again, we see notable overlap. The rapid scenario median sits between the 50th and 75th percentiles of the unconditional distribution, despite economists assigning the rapid scenario a small 14.0% probability on average.<\/p>\n\n\n\n

          Understanding\nvariance in GDP forecasts: implied distributions for the slow\/ moderate\nscenario and a decomposition.<\/h4>\n\n\n\n

          We next assume that the forecasts for the bundled slow\/moderate\nscenario would be consistent with the forecasts for the rapid and\nunconditional scenarios. Again, this assumption yields the mean and\nvariance for each forecaster\u2019s conditional outcome distribution in the\nbundled slow\/moderate scenario.<\/p>\n\n\n\n

          Lastly, we explore the variance of these three distributions. The standard deviation of 2030 GDP growth in the unconditional case is 1.49%. Conditioning on the slow\/moderate scenario reduces this by 9% to 1.36%. In contrast, conditioning on the rapid scenario increases the standard deviation by 33% to 1.98%. Thus, the rapid scenario not only raises expected GDP growth but also significantly increases uncertainty about this economic outcome. While moving to the slow\/moderate scenario reduces the variance of the outcome distribution, substantial within-scenario variance remains.<\/p>\n\n\n\n

          4.4 Decomposition Results<\/h3>\n\n\n\n

          Before presenting the decomposition results, we note that Appendix B.1<\/a> provides a detailed illustrative example using two hypothetical forecasters, demonstrating how the decomposition separates within-scenario from between-scenario variance and how within-forecaster uncertainty can dominate between-forecaster disagreement even when forecasters hold different beliefs about AI progress.<\/p>\n\n\n\n

          We perform the decomposition, as described in Appendix B.1<\/a>, and report the results for economists\u2019 forecasts of GDP growth, labor force participation rate, and wealth inequality in Figure 17<\/a> below. Specifically for forecasts of GDP growth in 2030, within-scenario (WS) variance\u2014comprising 94.9% of total variance\u2014is the dominant driver of total variance, relative to between-scenario (BS) variance (5.1% of the total). On both paths, within-forecaster uncertainty comprises a larger share of the variance: 83% (78.7% \/ 94.9%) and 96% (4.9% \/ 5.1%) of the within- and between-scenario variance, respectively. Forecasters are quite uncertain about the future, and this noise is larger than the variance driven by disagreement.<\/p>\n\n\n\n

          \"\"\/
          Figure 17: Decomposition of the variance in economists\u2019 forecasts of GDP growth, labor force participation rate, and wealth inequality in 2030 and 2050.<\/strong><\/figcaption><\/figure>\n\n\n\n

          Lastly, we now consider how this analysis relates to the question of\nwhether forecasters disagree on capabilities progress or outcomes\nconditional on capabilities progress.<\/p>\n\n\n\n

            \n
          1. If forecasters disagree noticeably on capabilities progress, we expect between-scenario, between-forecaster<\/em> variance to be large.<\/li>\n\n\n\n
          2. If forecasters disagree noticeably on outcomes conditional on capabilities progress, we expect within-scenario, between-forecaster<\/em> variance to be large.<\/li>\n<\/ol>\n\n\n\n

            For GDP growth, the first component (0.3%) is small in absolute terms\nand relative to the second component (16.1%). This suggests that\ndisagreement about outcomes conditional on various levels of progress is\na more important driver of total variance in 2030 GDP growth forecasts\nthan disagreement on capabilities progress per se.<\/p>\n\n\n\n

            This analysis also quantifies the contributions of disagreement versus uncertainty to total variance in outcomes. The within-forecaster<\/em> components all reflect uncertainty, which is outside of this question of where<\/em> disagreement<\/em> is most prevalent. Hence, we subtract out those components to isolate disagreement. Nevertheless, uncertainty dominates total variance in forecasts, explaining approximately 80% of total variance in Figure 17<\/a>. Forecasters also highlight uncertainty in their rationales. One forecaster highlights uncertainties around adoption, driving greater variance in the rapid scenario:<\/p>\n\n\n\n

            \n

            Rapid capability progress creates a genuine upside tail via (i) higher TFP, (ii) faster innovation cycles, and (iii) automation of broader task bundles. But it also widens uncertainty because the same world can feature large transitional frictions and misallocation: costly retooling, brittle deployment in complex domains, and the possibility that measured GDP gains are damped by adjustment costs (even if capabilities are \u201cthere\u201d).<\/em><\/p>\n<\/blockquote>\n\n\n\n

            Outcome<\/td>Year<\/td>Total Std. Dev.<\/td>WS<\/td>BS<\/td>WS-WF<\/td>WS-BF<\/td>BS-WF<\/td>BS-BF<\/td><\/tr>
            GDP<\/td>2030<\/td>1.494<\/td>0.949<\/td>0.051<\/td>0.787<\/td>0.161<\/td>0.049<\/td>0.003<\/td><\/tr>
            GDP<\/td>2050<\/td>2.166<\/td>0.947<\/td>0.053<\/td>0.753<\/td>0.194<\/td>0.050<\/td>0.003<\/td><\/tr>
            LFPR<\/td>2030<\/td>3.978<\/td>0.968<\/td>0.032<\/td>0.852<\/td>0.116<\/td>0.030<\/td>0.002<\/td><\/tr>
            LFPR<\/td>2050<\/td>6.731<\/td>0.963<\/td>0.037<\/td>0.567<\/td>0.397<\/td>0.035<\/td>0.002<\/td><\/tr>
            Median HH Income<\/td>2030<\/td>10,568<\/td>0.999<\/td>0.001<\/td>0.928<\/td>0.070<\/td>0.001<\/td>0.000<\/td><\/tr>
            Median HH Income<\/td>2050<\/td>16,327<\/td>1.000<\/td>0.000<\/td>0.679<\/td>0.321<\/td>0.000<\/td>0.000<\/td><\/tr>
            Unemployment Rate<\/td>2030<\/td>2.534<\/td>0.963<\/td>0.037<\/td>0.810<\/td>0.153<\/td>0.034<\/td>0.002<\/td><\/tr>
            Unemployment Rate<\/td>2050<\/td>3.192<\/td>0.990<\/td>0.010<\/td>0.737<\/td>0.253<\/td>0.009<\/td>0.001<\/td><\/tr>
            Wealth Inequality<\/td>2030<\/td>3.726<\/td>0.939<\/td>0.061<\/td>0.801<\/td>0.138<\/td>0.056<\/td>0.005<\/td><\/tr>
            Wealth Inequality<\/td>2050<\/td>6.424<\/td>0.981<\/td>0.019<\/td>0.790<\/td>0.190<\/td>0.018<\/td>0.001<\/td><\/tr><\/tbody><\/table><\/div>
            Table 2:<\/strong> Variance Decomposition: Economists
            Note: W = within, B = between, S = scenario, F = forecaster. GDP = Gross Domestic Product, LFPR = Labor Force Participation Rate, HH = household.<\/figcaption><\/figure>\n\n\n\n

            We conduct this analysis across all questions, time horizons, and groups. While we report more results in Appendix B<\/a>, we report results for economists in Table 2<\/a>. The same patterns hold as above. First, since within-scenario, between-forecaster variance dominates between-scenario, between-forecaster variance, we conclude that disagreement on the impacts of AI is primarily driven by disagreement on the conditional impacts of various capability levels, rather than divergence view on AI capabilities progress\u2014although we acknowledge that noise from discretization and scenario ambiguity, as discussed at the end of this section, may account for a small portion of the within-scenario variance. Second, within-forecaster uncertainty dwarfs between-forecaster disagreement. Regardless of realized AI capability progress, forecasters express substantial uncertainty and provide overlapping outcome distributions.<\/p>\n\n\n\n

            4.5 Limitations of the Decomposition<\/h3>\n\n\n\n

            One caveat is that our scenario descriptions bundled multiple dimensions of AI capability\u2014cognitive tasks such as research and coding alongside physical tasks such as robotics\u2014into single composite scenarios. Respondents were asked to select which scenario, in sum, best represented their views, and were advised that progress might be uneven across domains. Therefore, two respondents who both selected the rapid scenario may have held meaningfully different assumptions about the specific capability profile underlying that label. If respondents interpreted the scenarios differently in this way, some of the within-scenario, between-forecaster variance that our decomposition attributes to disagreement about economic mechanisms could potentially reflect disagreement about the effective capability level assumed within a given scenario. This effect would correspondingly weaken the contrast with Cunningham (2025)<\/span>.77<\/a><\/sup> In addition to ambiguity, our chosen discretization of progress could introduce some noise in the scenario forecasts. Two respondents might share identical scenario interpretations, but, conditional on rapid progress, one could expect progress to noticeably exceed the threshold between the moderate and rapid scenarios, while another expects progress to fall just above the threshold.<\/p>\n\n\n\n

            We acknowledge this limitation but believe that its effect was\nlimited. Respondents\u2019 written rationales suggest a broadly consistent\npattern: forecasters who deviated from the literal scenario descriptions\nmost commonly assumed faster cognitive AI progress and slower robotics\nprogress, rather than adopting wildly divergent capability profiles.\nThis consistency suggests that scenario ambiguity is unlikely to account\nfor much of the within-scenario variance we observe.<\/p>\n\n\n\n

            5. Discussion<\/h2>\n\n\n\n

            As we documented in our Section 4<\/a> variance decomposition, the primary source of disagreement among economists is likely not about whether AI capabilities will advance significantly\u2014majorities assign meaningful probability to the moderate or rapid scenario\u2014but about how quickly the economy can absorb these potentially transformative capabilities, and what absorption will lead to in terms of economic impacts. Specifically, economists who share similar views on the likelihood of rapid AI progress nevertheless diverge on the likely rate of diffusion, the extent to which new job creation will offset displacement, the degree to which lags will occur between adoption and productivity gains, and how institutional and regulatory responses will shape the transition. We also see this in the analysis in Appendix E.1<\/a>: while economists who assign above-median probabilities to the rapid scenario tend to have slightly higher unconditional forecasts for GDP growth and lower forecasts for the LFPR, the outcome distributions of these above- and below-median participants overlap significantly.<\/p>\n\n\n\n

            However, despite disagreement on the magnitude of these effects, the\nmajority of experts agree that their net direction will be to attenuate\nrather than accelerate AI\u2019s impact on the economy. Indeed, even under\nthe rapid scenario, where AI systems surpass human performance on most\ncognitive and physical tasks by 2030, experts do not forecast economic\noutcomes outside the range of historical experience. Instead, their\nwritten rationales point repeatedly to diffusion lags, infrastructure\nbottlenecks, political instability, and demographic headwinds as\nmechanisms that will likely prevent even highly capable AI from\nproducing unprecedented economic outcomes in the near term.<\/p>\n\n\n\n

            This finding stands in marked contrast to warnings, raised by some prominent voices in the AI industry, about rapid economic transformation. Our sample partially captured this view, especially in the AI expert group, which forecast a GDP growth rate of 3.7% in 2030 and 5.3% in 2050 under the rapid scenario. Their 90th percentile forecast is 6.5% in 2030, meaning that they assign a 10% probability to growth equal to or larger than this. These predictions are higher than the economists\u2019 but lower than those of some in the industry.78<\/a><\/sup> One possible explanation is that forecasters in our study over-anchored to historical data, which were prominently displayed in the survey interface (see Figure 80<\/a>). Additionally, some of the difference (though not all, as described above) could be due to different capability progress beliefs. Our sample and scenario discretization may not capture the most extreme end of rapid progress beliefs, such that even the forecasts conditional on the rapid scenario remain unrepresentative of those in the industry with the most extreme beliefs about capabilities. In Appendix C<\/a>, we present some early evidence that beliefs about AI timelines may be speeding up across multiple groups: respondents in the Longitudinal Expert AI Panel (LEAP), which ran in February 2026, assigned higher probabilities to the rapid scenario than those in this study (October 2025\u2013February 2026). Even so, we find little difference in economic outcome forecasts between these two samples.<\/p>\n\n\n\n

            While the aggregate forecasts we find in this study are at the more\nmoderate end of the spectrum of discourse, we note that our survey\nexperts express the most uncertainty under the rapid AI progress\nscenario where the stakes for policy design are the highest. Experts\u2019\nunconditional forecasts\u2014the ones that reflect their actual\nall-things-considered beliefs\u2014cluster around historical baselines, but\ntheir uncertainty under the rapid scenario widens significantly for the\nLFPR, GDP growth, and inequality. While experts only assign the rapid\nscenario a 14% probability of occurring, given the magnitude of\npotential consequences, that number is far from negligible. This matters\nbecause, if AI progress is slow, existing institutions making\nincremental adjustments to current policies may prove adequate. But if\nprogress is rapid, the breadth of the outcome distribution implies that\npolicymakers cannot simply plan for the median outcome; they must\ncontend with tail risks, including the potential for a deep contraction\nin labor force participation.<\/p>\n\n\n\n

            On the question of which policies might best mitigate negative\nimpacts from AI, our survey revealed a marked divergence between\neconomists and the general public: economists favor targeted,\nincremental interventions such as retraining support and modernized\nunemployment insurance, while the general public expressed support for\nbroader interventions such as job guarantees and universal basic income.\nThis gap is disproportionate to the differences in economic forecasts\nbetween the two groups\u2014the general public\u2019s median unconditional GDP\ngrowth and LFPR projections only differ modestly from economists\u2019\u2014and\nyet the general public is nearly four times as likely as economists to\nsupport a job guarantee. If AI progress accelerates and labor market\ndisruption becomes more visible, these underlying disagreements may\nbecome the central fault lines of policy debates.<\/p>\n\n\n\n

            Notes<\/h2>\n\n\n
            1. Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence<\/em>. Stanford Digital Economy Lab. URL: https:\/\/digitaleconomy.stanford.edu\/publications\/canaries-in-the-coal-mine\/<\/a>. \u21a9\ufe0e<\/a><\/li>
            2. Humlum, Anders and Emilie Vestergaard (May 2025). Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI<\/em>. Working Paper 33777. Revised March National Bureau of Economic Research. DOI: 10.3386\/w33777<\/a>. URL: https:\/\/www.nber.org\/papers\/w33777<\/a>. \u21a9\ufe0e<\/a><\/li>
            3. Davis, Scott (Feb. 2026). AI is Simultaneously Aiding and Replacing Workers, Wage Data Suggest<\/em>. Dallas Fed Economics. Accessed: 2026-03-26. Federal Reserve Bank of Dallas. URL: https:\/\/www.dallasfed.org\/research\/economics\/2026\/0224<\/a>. \u21a9\ufe0e<\/a><\/li>
            4. Gimbel, Martha, Molly Kinder, Joshua Kendall, and Maddie Lee (2025). Evaluating the Impact of AI on the Labor Market: Current State of Affairs<\/em>. URL: https:\/\/budgetlab.yale.edu\/research\/evaluating-impact-ai-labor-market-current-state-affairs<\/a>. \u21a9\ufe0e<\/a><\/li>
            5. Gerut, Amanda (2025). Jamie Dimon on AI: \u201cPeople Should Stop Sticking Their Head in the Sand\u201d<\/em>. URL: https:\/\/fortune.com\/2025\/10\/08\/jamie-dimon-head-in-the-sand-ai-will-take-jobs-tip-iceberg\/<\/a>. \u21a9\ufe0e<\/a><\/li>
            6. Altman, Sam (June 2025). The Gentle Singularity. https:\/\/blog.samaltman.com\/thegentle-singularity<\/a>. Blog post, published June 10, 2025. Accessed March 30, 2026. \u21a9\ufe0e<\/a><\/li>
            7. VandeHei, Jim and Mike Allen (May 2025). \u201cBehind the Curtain: A white-collar bloodbath\u201d. In: Axios<\/em>. URL: https:\/\/www.axios.com\/2025\/05\/28\/ai-jobs-white-collar-unemployment-anthropic<\/a> (visited on 03\/30\/2026). \u21a9\ufe0e<\/a><\/li>
            8. Arnon, Alexander (Sept. 2025). The Projected Impact of Generative AI on Future Productivity Growth. Tech. rep. Penn Wharton Budget Model. URL: https:\/\/budgetmodel.wharton.upenn.edu\/p\/2025-09-08-the-projected-impact-of-generative-ai-on-futureproductivity-growth<\/a>. \u21a9\ufe0e<\/a><\/li>
            9. Filippucci, Francesco, Peter Gal, Karin Laengle, and Matthias Schief (2025). Macroeconomic Productivity Gains from Artificial Intelligence in G7 Economies<\/em>. 41. OECD. DOI: 10.1787\/a5319ab5-en<\/a>. URL: https:\/\/doi.org\/10.1787\/a5319ab5-en<\/a> \u21a9\ufe0e<\/a><\/li>
            10. Russakovsky, Olga, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei (2015). \u201cImageNet Large Scale Visual Recognition Challenge\u201d. In: International Journal of Computer Vision<\/em> 115, pp. 211\u2013252. DOI: 10.1007\/s11263-015-0816-y<\/a>; Hendrycks, Dan, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt (2020). \u201cMeasuring Massive Multitask Language Understanding\u201d. In: arXiv<\/em>. DOI: 10.48550\/arxiv.2009.03300<\/a>; Jain, Naman, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica (2024). \u201cLiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code\u201d. In: arXiv<\/em>. DOI: 10.48550\/arxiv.2403.07974<\/a>. \u21a9\ufe0e<\/a><\/li>
            11. Grace, Katja, Harlan Stewart, Julia Fabienne Sandk\u00fchler, Stephen Thomas, Ben Weinstein-Raun, Jan Brauner, and R. C. Korzekwa (2024). Thousands of AI Authors on the Future of AI<\/em>. DOI: 10.48550\/arXiv.2401.02843<\/a>. URL: https:\/\/arxiv.org\/abs\/2401.02843<\/a>; Murphy, Connacher, Josh Rosenberg, Jordan Canedy, Zach Jacobs, Nadja Flechner, Rhiannon Britt, Alexa Pan, Charlie Rogers-Smith, Dan Mayland, Cathy Buffington, Simas Ku\u010dinskas, Amanda Coston, Hannah Kerner, Emma Pierson, Reihaneh Rabbany, Matthew Salganik, Robert Seamans, Yu Su, Florian Tram\u00e8r, Tatsunori Hashimoto, Arvind Narayanan, Philip E. Tetlock, and Ezra Karger (2025). The Longitudinal Expert AI Panel: Understanding Expert Views on AI Capabilities, Adoption, and Impact<\/em>. Working paper 5. Forecasting Research Institute. URL: https:\/\/forecastingresearch.org\/s\/the-longitudinalexpert-ai-panel.pdf<\/a> (visited on 03\/18\/2026). \u21a9\ufe0e<\/a><\/li>
            12. Aghion, Philippe, Benjamin F. Jones, and Charles I. Jones (2019). \u201cArtificial Intelligence and Economic Growth\u201d. In: The Economics of Artificial Intelligence: An Agenda. Ed. by Ajay Agrawal, Joshua Gans, and Avi Goldfarb. Chicago: University of Chicago Press, pp. 237\u2013290. ISBN: 9780226613338. DOI: 10.7208\/chicago\/9780226613475.003.0009<\/a>; Agrawal, Ajay, Joshua Gans, and Avi Goldfarb (2018). Prediction Machines: The Simple Economics of Artificial Intelligence. Harvard Business Review Press. \u21a9\ufe0e<\/a><\/li>
            13. Agrawal, Ajay, Joshua Gans, and Avi Goldfarb (2022). Power and Prediction: The Disruptive Economics of Artificial Intelligence. Harvard Business Review Press. \u21a9\ufe0e<\/a><\/li>
            14. Comin, Diego and Bart Hobijn (2010). \u201cAn Exploration of Technology Diffusion\u201d. In: American Economic Review<\/em> 100.5, pp. 2031\u20132059. DOI: 10.1257\/aer.100.5.2031<\/a> \u21a9\ufe0e<\/a><\/li>
            15. The term \u2018superforecaster\u2019 was coined in early research on subjective probability elicitation in a geopolitical context, as described in Mellers, Barbara, Lyle Ungar, Jonathan Baron, Jaime Ramos, Burcu Gurcay, Katrina Fincher, Sydney E. Scott, Don Moore, Pavel Atanasov, Samuel A. Swift, Terry Murray, Eric Stone, and Philip E. Tetlock (2014). \u201cPsychological Strategies for Winning a Geopolitical Forecasting Tournament\u201d. In: Psychological Science<\/em> 25.5, pp. 1106\u20131115. DOI: 10.1177\/
              0956797614524255<\/a>.               \u21a9\ufe0e<\/a><\/li>
            16. Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>. \u21a9\ufe0e<\/a><\/li>
            17. Trammell, Philip and Anton Korinek (2023). Economic Growth under Transformative AI<\/em>. w31815. National Bureau of Economic Research. DOI: 10.3386\/w31815<\/a>. URL: https:\/\/www.nber.org\/papers\/w31815<\/a>. \u21a9\ufe0e<\/a><\/li>
            18. Erdil, Ege and Tamay Besiroglu (2024). Explosive Growth from AI Automation: A Review of the Arguments.<\/em> DOI: 10.48550\/arXiv.2309.11690<\/a>. URL: http:\/\/arxiv.org\/abs\/2309.11690<\/a>. \u21a9\ufe0e<\/a><\/li>
            19. Eloundou, Tyna, Sam Manning, Pamela Mishkin, and Daniel Rock (2024). \u201cGPTs are GPTs: Labor market impact potential of LLMs\u201d. In: Science<\/em> 384.6702, pp. 1306\u20131308. \u21a9\ufe0e<\/a><\/li>
            20. Aghion, Philippe and Simon Bunel (2024). AI and Growth: Where Do We Stand?<\/em> Federal Reserve Bank of San Francisco. URL: https:\/\/www.frbsf.org\/wp-content\/uploads\/AIand-Growth-Aghion-Bunel.pdf<\/a>. \u21a9\ufe0e<\/a><\/li>
            21. Filippucci, Francesco, Peter Gal, and Matthias Schief (2024). Miracle or Myth? Assessing the Macroeconomic Productivity Gains from Artificial Intelligence<\/em>. OECD. DOI: 10.1787\/b524a072-en<\/a>. URL: https:\/\/www.oecd.org\/en\/publications\/miracle-or-myth-assessing-the-macroeconomic-productivity-gains-from-artificialintelligence_b524a072-en.html<\/a>. \u21a9\ufe0e<\/a><\/li>
            22. For more on Baumol\u2019s Cost Disease, see Baumol, William J (2012). The cost disease: Why computers get cheaper and health care doesn\u2019t.<\/em> Yale university press. \u21a9\ufe0e<\/a><\/li>
            23. Beraja, Martin and Nathan Zorzi (2022). Inefficient Automation<\/em>. w30154. National Bureau
              of Economic Research. DOI:               10.3386\/w30154<\/a>. URL: https:\/\/www.nber.org\/papers\/w30154<\/a>. \u21a9\ufe0e<\/a><\/li>
            24. Acemoglu, Daron (2024). \u201cThe Simple Macroeconomics of AI\u201d. In: Economic Policy 40.121, pp. 13\u201358. DOI: 10.1093\/epolic\/eiae042<\/a>. \u21a9\ufe0e<\/a><\/li>
            25. Comunale, Mariarosaria and Andrea Manera (2024). The Economic Impacts and the Regulation of AI: A Review of the Academic Literature and Policy Actions<\/em>. DOI: 10.5089\/9798400271663.001<\/a>. URL: https:\/\/papers.ssrn.com\/abstract=4774326<\/a>. \u21a9\ufe0e<\/a><\/li>
            26. Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence<\/em>. Stanford Digital Economy Lab. URL: https:\/\/digitaleconomy.stanford.edu\/publications\/canaries-in-the-coal-mine\/<\/a>.<\/span> \u21a9\ufe0e<\/a><\/li>
            27. Gimbel, Martha, Molly Kinder, Joshua Kendall, and Maddie Lee (2025). Evaluating the Impact of AI on the Labor Market: Current State of Affairs<\/em>. URL: https:\/\/budgetlab.yale.edu\/research\/evaluating-impact-ai-labor-market-current-state-affairs<\/a>. \u21a9\ufe0e<\/a><\/li>
            28. Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence<\/em>. Stanford Digital Economy Lab. URL: https:\/\/digitaleconomy.stanford.edu\/publications\/canaries-in-the-coal-mine\/<\/a>.<\/span> \u21a9\ufe0e<\/a><\/li>
            29. Humlum, Anders and Emilie Vestergaard (May 2025). Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI<\/em>. Working Paper 33777. Revised March National Bureau of Economic Research. DOI: 10.3386\/w33777<\/a>. URL: https:\/\/www.nber.org\/papers\/w33777<\/a>. \u21a9\ufe0e<\/a><\/li>
            30. Frey, Carl Benedikt and Michael A. Osborne (2017). \u201cThe Future of Employment: How Susceptible Are Jobs to Computerisation?\u201d In: Technological Forecasting and Social Change<\/em> 114, pp. 254\u2013280. DOI: 10.1016\/j.techfore.2016.08.019<\/a>. \u21a9\ufe0e<\/a><\/li>
            31. Korinek, Anton and Megan Juelfs (2022). Preparing for the (Non-Existent?) Future of Work.<\/em> w30172. National Bureau of Economic Research. DOI: 10.3386\/w30172<\/a>. URL: https:\/\/www.nber.org\/papers\/w30172<\/a>. \u21a9\ufe0e<\/a><\/li>
            32. Acemoglu, Daron (2024). \u201cThe Simple Macroeconomics of AI\u201d. In: Economic Policy 40.121, pp. 13\u201358. DOI: 10.1093\/epolic\/eiae042<\/a>. \u21a9\ufe0e<\/a><\/li>
            33. AGI stands for Artificial General Intelligence and refers to a world where AI systems can perform tasks and bundles of tasks at a level that exceeds the vast majority of current workers, especially for white-collar workers whose jobs are not particularly physical in nature. \u21a9\ufe0e<\/a><\/li>
            34. Korinek, Anton (2024). Economic Consequences of Frontier AI<\/em>. NBER Working Paper No. 32980. Cambridge, MA. DOI: 10.3386\/w32980<\/a>. URL: https:\/\/www.nber.org\/papers\/w32980<\/a>. \u21a9\ufe0e<\/a><\/li>
            35. Korinek, Anton and Joseph E. Stiglitz (2021). Artificial Intelligence, Globalization, and Strategies for Economic Development<\/em>. w28453. National Bureau of Economic Research. DOI: 10.3386\/w28453<\/a>. URL: https:\/\/www.nber.org\/papers\/w28453<\/a>. \u21a9\ufe0e<\/a><\/li>
            36. Abbott, Ryan and Bret Bogenschneider (2018). \u201cShould Robots Pay Taxes? Tax Policy in the Age of Automation\u201d. In: Harvard Law & Policy Review<\/em> 12.1, pp. 145\u2013175. DOI: 10.2139\/ssrn.2932483<\/a>. URL: https:\/\/heinonline.org\/HOL\/P?h=hein.journals\/harlpolrv12&i=147<\/a> \u21a9\ufe0e<\/a><\/li>
            37. Comunale, Mariarosaria and Andrea Manera (2024). The Economic Impacts and the Regulation of AI: A Review of the Academic Literature and Policy Actions<\/em>. DOI: 10.5089\/9798400271663.001<\/a>. URL: https:\/\/papers.ssrn.com\/abstract=4774326<\/a>; OECD (2023). OECD Employment Outlook 2023<\/em>. Report. Accessed: 2026-03-30. Organisation for Economic Co-operation and Development. URL: https:\/\/www.oecd.org\/content\/dam\/oecd\/en\/publications\/reports\/2023\/07\/oecd- employment- outlook- 2023_904bcef3\/08785bba-en.pdf<\/a>. \u21a9\ufe0e<\/a><\/li>
            38. Hyman, Benjamin G., Brian K. Kovak, and Adam Leive (2024). Wage Insurance for Displaced Workers<\/em>. DOI: 10.3386\/w32464<\/a>. URL: https:\/\/www.nber.org\/papers\/w32464<\/a>. \u21a9\ufe0e<\/a><\/li>
            39. Korinek, Anton and Megan Juelfs (2022). Preparing for the (Non-Existent?) Future of Work.<\/em> w30172. National Bureau of Economic Research. DOI: 10.3386\/w30172<\/a>. URL: https:\/\/www.nber.org\/papers\/w30172<\/a>. \u21a9\ufe0e<\/a><\/li>
            40. Abbott, Ryan and Bret Bogenschneider (2018). \u201cShould Robots Pay Taxes? Tax Policy in the Age of Automation\u201d. In: Harvard Law & Policy Review<\/em> 12.1, pp. 145\u2013175. DOI: 10.2139\/ssrn.2932483<\/a>. URL: https:\/\/heinonline.org\/HOL\/P?h=hein.journals\/harlpolrv12&i=147<\/a> \u21a9\ufe0e<\/a><\/li>
            41. Bastani, Spencer and Daniel Waldenstr\u00f6m (2024). AI, Automation and Taxation. URL: https:\/\/papers.ssrn.com\/sol3\/papers.cfm?abstract_id=4811796<\/a>. \u21a9\ufe0e<\/a><\/li>
            42. Korinek, Anton (2024). Economic Consequences of Frontier AI<\/em>. NBER Working Paper No. 32980. Cambridge, MA. DOI: 10.3386\/w32980<\/a>. URL: https:\/\/www.nber.org\/papers\/w32980<\/a>; Anthropic (2025). Preparing for AI\u2019s Economic Impact: Exploring Policy Responses<\/em>. URL: https:\/\/www.anthropic.com\/research\/economic-policy-responses<\/a>. \u21a9\ufe0e<\/a><\/li>
            43. Grace, Katja, Harlan Stewart, Julia Fabienne Sandk\u00fchler, Stephen Thomas, Ben Weinstein-Raun, Jan Brauner, and R. C. Korzekwa (2024). Thousands of AI Authors on the Future of AI<\/em>. DOI: 10.48550\/arXiv.2401.02843<\/a>. URL: https:\/\/arxiv.org\/abs\/2401.02843<\/a>. \u21a9\ufe0e<\/a><\/li>
            44. Clark Center Forum (2025). AI and Growth<\/em>. URL: https:\/\/kentclarkcenter.org\/surveys\/ai-and-growth\/<\/a>. \u21a9\ufe0e<\/a><\/li>
            45. Murphy, Connacher, Josh Rosenberg, Jordan Canedy, Zach Jacobs, Nadja Flechner, Rhiannon Britt, Alexa Pan, Charlie Rogers-Smith, Dan Mayland, Cathy Buffington, Simas Ku\u010dinskas, Amanda Coston, Hannah Kerner, Emma Pierson, Reihaneh Rabbany, Matthew Salganik, Robert Seamans, Yu Su, Florian Tram\u00e8r, Tatsunori Hashimoto, Arvind Narayanan, Philip E. Tetlock, and Ezra Karger (2025). The Longitudinal Expert AI Panel: Understanding Expert Views on AI Capabilities, Adoption, and Impact<\/em>. Working paper 5. Forecasting Research Institute. URL: https:\/\/forecastingresearch.org\/s\/the-longitudinalexpert-ai-panel.pdf<\/a> (visited on 03\/18\/2026). \u21a9\ufe0e<\/a><\/li>
            46. Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>. \u21a9\ufe0e<\/a><\/li>
            47. Acemoglu, Daron (2024). \u201cThe Simple Macroeconomics of AI\u201d. In: Economic Policy 40.121, pp. 13\u201358. DOI: 10.1093\/epolic\/eiae042<\/a>; Trammell, Philip and Anton Korinek (2023). Economic Growth under Transformative AI<\/em>. w31815. National Bureau of Economic Research. DOI: 10.3386\/w31815<\/a>. URL: https:\/\/www.nber.org\/papers\/w31815<\/a>. \u21a9\ufe0e<\/a><\/li>
            48. IARPA (2011). IARPA Aggregative Contingent Estimation (ACE) Program<\/em>. URL: https:\/\/www.iarpa.gov\/research-programs\/ace<\/a>; Mellers, Barbara, Lyle Ungar, Jonathan Baron, Jaime Ramos, Burcu Gurcay, Katrina Fincher, Sydney E. Scott, Don Moore, Pavel Atanasov, Samuel A. Swift, Terry Murray, Eric Stone, and Philip E. Tetlock (2014). \u201cPsychological Strategies for Winning a Geopolitical Forecasting Tournament\u201d. In: Psychological Science<\/em> 25.5, pp. 1106\u20131115. DOI: 10.1177\/
              0956797614524255<\/a>.               \u21a9\ufe0e<\/a><\/li>
            49. Mellers, Barbara, Eric Stone, Pavel Atanasov, Nick Rohrbaugh, S. Emlen Metz, Lyle Ungar, Michael M. Bishop, Michael Horowitz, Ed Merkle, and Philip E. Tetlock (2015). \u201cIdentifying and Cultivating Superforecasters as a Method of Improving Probabilistic Predictions\u201d. In: Perspectives on Psychological Science<\/em> 10.3, pp. 267\u2013281. DOI: 10.1177\/1745691615577794<\/a>. \u21a9\ufe0e<\/a><\/li>
            50. Hartman, Rachel, Aaron J. Moss, Shalom N. Jaffe, Cheskie Rosenzweig, Leib Litman, and Jonathan Robinson (2023). \u201cIntroducing Connect by CloudResearch: Advancing Online Participant Recruitment in the Digital Age\u201d. In: DOI: 10.31234\/osf.io\/ksgyr<\/a>. URL: https:\/\/doi.org\/10.31234\/osf.io\/ksgyr<\/a>. \u21a9\ufe0e<\/a><\/li>
            51. Lichtendahl, Kenneth C., Yael Grushka-Cockayne, and Robert L. Winkler (2013). \u201cIs It Better to Average Probabilities or Quantiles?\u201d In: Management Science<\/em> 59.7, pp. 1594\u20131611. DOI: 10.1287\/mnsc.1120.1667<\/a>. \u21a9\ufe0e<\/a><\/li>
            52. These forecasts are: FOMC (Federal Reserve (FOMC) 2025), CBO (Congressional Budget Office 2026), OMB (Office of Management & Budget (OMB) 2025), IMF (International Monetary Fund 2025), SPF (Survey of Professional Forecasters 2025, 2026), OECD (OECD 2025), Goldman Sachs (Goldman Sachs 2025), The Conference Board (The Conference Board 2026), and Deloitte (Deloitte 2025). We focus on the baseline estimates; some forecasts consider other cases, such as Deloitte\u2019s \u201cDownside\u201d and \u201cUpside\u201d scenarios and OECD\u2019s \u201cEnergy Transition\u201d scenarios. \u21a9\ufe0e<\/a><\/li>
            53. These estimates assume growth rates in 2030-2045 are linearly interpolated between the 2025\u20132029 and 2045\u20132049 forecasts, as in Figure 7<\/a>. \u21a9\ufe0e<\/a><\/li>
            54. For comparison, the average CBO forecast of TFP between 2025 and 2029 is 0.82%, and between 2044 and 2049 it is 1.1% (Congressional Budget Office (Feb. 2026). The Long-Term Budget Outlook Data: 2026 to 2056<\/em>. Tech. rep. Data available at: https:\/\/www.cbo.gov\/system\/files\/2026-02\/57054-2026-02-LTBO-econ.xlsx<\/a>. Accessed: March 19, 2026. URL: https:\/\/www.cbo.gov\/publication\/62044<\/a>.). \u21a9\ufe0e<\/a><\/li>
            55. Jorgenson, Dale W and Kevin J Stiroh (2000). \u201cRaising the speed limit: US economic growth in the information age\u201d. In: Knowledge economy, information technologies and growth.<\/em> Routledge, pp. 335\u2013424. \u21a9\ufe0e<\/a><\/li>
            56. For comparison, the average forecast LFPR between 2025 and 2029 for CBO (using the Census Through 2020 Plus CBO Projection) is 62.4% and for Deloitte (after converting unemployment rate and employment-to-population rates to LFPR) is 62.0% (Congressional Budget Office (Feb. 2026). The Long-Term Budget Outlook Data: 2026 to 2056<\/em>. Tech. rep. Data available at: https:\/\/www.cbo.gov\/system\/files\/2026-02\/57054-2026-02-LTBO-econ.xlsx<\/a>. Accessed: March 19, 2026. URL: https:\/\/www.cbo.gov\/publication\/62044<\/a>; Deloitte (Dec. 2025). US Economic Forecast 2026-2030<\/em>. Tech. rep. Accessed: March 13, 2026. URL: https:\/\/www.deloitte.com\/us\/en\/insights\/topics\/economy\/us-economicforecast\/united-states-outlook-analysis.html<\/a>.). \u21a9\ufe0e<\/a><\/li>
            57. For comparison, the average forecast LFPR between 2046 and 2050 for CBO is 62.1% (Congressional Budget Office (Feb. 2026). The Long-Term Budget Outlook Data: 2026 to 2056<\/em>. Tech. rep. Data available at: https:\/\/www.cbo.gov\/system\/files\/2026-02\/57054-2026-02-LTBO-econ.xlsx<\/a>. Accessed: March 19, 2026. URL: https:\/\/www.cbo.gov\/publication\/62044<\/a>.). \u21a9\ufe0e<\/a><\/li>
            58. Both estimates are derived by comparing slow and rapid scenario LFPR forecasts. Under the rapid scenario, economists forecast a 7 p.p. decrease in the LFPR from the February 2026 reading of 62% to the 2050 median prediction of 55.0%. Under the slow scenario, where AI-driven displacement considerations are likely to be minimal, the decrease is forecasted to be 2.8 p.p., from 62% to 59.2%. This implies that (7 \u2212 2.8)\/7 \u2248 60% of the projected rapid-scenario decline is attributable to AI rather than demographics or other non-AI trends. The AI-attributable share of the decline is thus roughly 4.2 p.p. (the difference between the rapid-scenario forecast of 55.0% and the slow-scenario forecast of 59.2%). Applying this to the current U.S. civilian non-institutional population aged 16 and over of approximately 270 million yields roughly 11 million fewer labor force participants. We round to 10 million to reflect the imprecision inherent in this calculation, including uncertainty about the size of the 2050 population and the simplifying assumption that the slow scenario captures all non-AI sources of LFPR decline. We also note that worlds with rapid AI progress likely differ from worlds with slow progress in ways that extend beyond AI capabilities themselves, so this gap should be understood as the estimated effect of moving between scenarios that are focused on AI rather than a cleanly identified causal effect of AI capabilities alone. That said, the U.S. civilian population is projected to grow through 2050, which would increase the number of displaced workers, making this estimate conservative. \u21a9\ufe0e<\/a><\/li>
            59. There were 10 blocks, each with 10 occupations, and each person was randomly assigned to see one block. Two blocks mistakenly contained the same occupation twice. In these cases, we consider only one job if the respondent answered consistently for the duplicates and do not consider either if they answered inconsistently. \u21a9\ufe0e<\/a><\/li>
            60. When comparing results, it is important to note that Eloundou et al. (2024) focus on LLM-exposure, while our survey respondents predicted job growth. While one component of job growth is LLM exposure, there are, of course, many other things that the survey respondent may be considering. \u21a9\ufe0e<\/a><\/li>
            61. Eloundou et al. (2024) quantify how exposed the occupations are to LLMs by obtaining ratings from humans (and ChatGPT) on how exposed a job\u2019s tasks (from O*NET) are. They present several versions of AI exposure; we focus on the version using human raters and Direct Exposure plus 0.5x LLM+ Exposed, where Direct Exposure is defined as \u201cusing the described LLM via ChatGPT or the OpenAI playground can decrease the time required to complete the DWA or task by at least half (50%)\u201d and LLM+ Exposed is defined as \u201caccess to the described LLM alone would not reduce the time required to complete the activity\/task by at least half, but additional software could be developed on top of the LLM that could reduce the time it takes to complete the specific activity\/task with quality by at least half. Among these systems, we count access to image generation systems.\u201d \u21a9\ufe0e<\/a><\/li>
            62. Eloundou, Tyna, Sam Manning, Pamela Mishkin, and Daniel Rock (2024). \u201cGPTs are GPTs: Labor market impact potential of LLMs\u201d. In: Science<\/em> 384.6702, pp. 1306\u20131308. \u21a9\ufe0e<\/a><\/li>
            63. Amodei, Dario (Oct. 2024). Machines of Loving Grace: How AI Could Transform the World
              for the Better<\/em>. Essay, darioamodei.com. URL:               https:\/\/www.darioamodei.com\/essay\/machines-of-loving-grace<\/a>; Brynjolfsson, Erik, Anton Korinek, and Ajay Agrawal (2025). A Research Agenda for the Economics of Transformative AI<\/em>. w34256. National Bureau of Economic Research. DOI: 10.3386\/w34256<\/a>. URL: https:\/\/www.nber.org\/papers\/w34256<\/a>. \u21a9\ufe0e<\/a><\/li>
            64. Davidson, Tom (2021). Could Advanced AI Drive Explosive Economic Growth?<\/em> https:\/\/coefficientgiving.org\/research\/could-advanced-ai-drive-explosive-economic-growth\/<\/a>. Coefficient Giving, accessed March 31, 2026. \u21a9\ufe0e<\/a><\/li>
            65. Cantril, Hadley (1965). The Pattern of Human Concerns<\/em>. New Brunswick, NJ: Rutgers University Press. \u21a9\ufe0e<\/a><\/li>
            66. This would have corresponded to approximately 120 billion dollars in 2025. For context, Epoch estimates that the combined capital expenditures at Alphabet, Amazon, Meta, Microsoft, and Oracle in 2025 were 448.2 billion dollars: https:\/\/epoch.ai\/data-insights\/hyperscaler-capex-trend<\/a> \u21a9\ufe0e<\/a><\/li>
            67. Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>. \u21a9\ufe0e<\/a><\/li>
            68. Respondents provide forecasts for the 10th , 50th , and 90th percentiles. \u21a9\ufe0e<\/a><\/li>
            69. We fit a normal distribution to outcomes with unbounded support (GDP growth, although strictly speaking it is bounded from below by -100%), a gamma distribution to outcomes that are bounded only above or only below (median household income), and a beta distribution to variables that are bounded both above and below (LFPR, unemployment rate, wealth inequality). \u21a9\ufe0e<\/a><\/li>
            70. We express the unconditional distribution for the outcome as a mixture distribution over the scenarios. The sub-distributions are given by the distribution of the outcome conditional on the rapid or bundled slow\/moderate scenario. We use a respondent\u2019s scenario probabilities to define the mixture weights. \u21a9\ufe0e<\/a><\/li>
            71. In this case, the implied variance of the outcome conditional on the slow\/moderate scenario will be negative. \u21a9\ufe0e<\/a><\/li>
            72. Incoherent distributions could reflect incoherence in beliefs or misspecification in the distribution fitting procedure. \u21a9\ufe0e<\/a><\/li>
            73. We must accordingly adjust the weights in the rapid and slow\/moderate scenarios. For example, a forecaster\u2019s weight in the rapid conditional outcome distribution will be proportional to the product of their original weight and their forecasted probability for the rapid scenario. \u21a9\ufe0e<\/a><\/li>
            74. Ranjan, Roopesh and Tilmann Gneiting (2010). \u201cCombining Probability Forecasts\u201d. In: Journal of the Royal Statistical Society: Series B<\/em> 72.1, pp. 71\u201391. \u21a9\ufe0e<\/a><\/li>
            75. Lichtendahl, Kenneth C., Yael Grushka-Cockayne, and Robert L. Winkler (2013). \u201cIs It Better to Average Probabilities or Quantiles?\u201d In: Management Science<\/em> 59.7, pp. 1594\u20131611. DOI: 10.1287\/mnsc.1120.1667<\/a>. \u21a9\ufe0e<\/a><\/li>
            76. The within-scenario variance is the probability-weighted sum of the variance in each scenario, so we can calculate directly each scenario\u2019s contribution to within-scenario variance. \u21a9\ufe0e<\/a><\/li>
            77. Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>. \u21a9\ufe0e<\/a><\/li>
            78. See Cunningham (2025)<\/span>. \u21a9\ufe0e<\/a><\/li><\/ol>\n\n\n
              \n
              References and the Appendix are provided in the full PDF Report \n \n<\/svg>\n\n \n<\/svg><\/a><\/div>\n<\/div>\n\n\n\n

              <\/p>\n","protected":false},"excerpt":{"rendered":"We surveyed academic economists, AI experts, and superforecasters to produce the most comprehensive picture yet of how AI could reshape the U.S. economy. Respondents also evaluated the impact of six different policies intended to address the impact of rapid AI progress.","protected":false},"featured_media":856,"template":"","meta":{"footnotes":"[{\"id\":\"7d0b09ab-4816-4b53-955a-37ecb2ba9477\",\"content\":\"Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence<\/em>. Stanford Digital Economy Lab. URL: https:\/\/digitaleconomy.stanford.edu\/publications\/canaries-in-the-coal-mine\/<\/a>.\"},{\"id\":\"c5438462-e1e8-49c0-bd03-c23e7e0aaec1\",\"content\":\"Humlum, Anders and Emilie Vestergaard (May 2025). Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI<\/em>. Working Paper 33777. Revised March National Bureau of Economic Research. DOI: 10.3386\/w33777<\/a>. URL: https:\/\/www.nber.org\/papers\/w33777<\/a>.\"},{\"id\":\"113842d8-a4e4-45f3-adcf-e721aaa339b5\",\"content\":\"Davis, Scott (Feb. 2026). AI is Simultaneously Aiding and Replacing Workers, Wage Data Suggest<\/em>. Dallas Fed Economics. Accessed: 2026-03-26. Federal Reserve Bank of Dallas. URL: https:\/\/www.dallasfed.org\/research\/economics\/2026\/0224<\/a>.\"},{\"id\":\"7bb2c008-3960-4982-a8c9-1a51139a60ce\",\"content\":\"Gimbel, Martha, Molly Kinder, Joshua Kendall, and Maddie Lee (2025). Evaluating the Impact of AI on the Labor Market: Current State of Affairs<\/em>. URL: https:\/\/budgetlab.yale.edu\/research\/evaluating-impact-ai-labor-market-current-state-affairs<\/a>.\"},{\"id\":\"59424be4-c89f-47a5-b895-3593941db212\",\"content\":\"Gerut, Amanda (2025). Jamie Dimon on AI: \u201cPeople Should Stop Sticking Their Head in the Sand\u201d<\/em>. URL: https:\/\/fortune.com\/2025\/10\/08\/jamie-dimon-head-in-the-sand-ai-will-take-jobs-tip-iceberg\/<\/a>.\"},{\"id\":\"24dba128-90f9-4de4-9737-19f8a4a023c9\",\"content\":\"Altman, Sam (June 2025). The Gentle Singularity. https:\/\/blog.samaltman.com\/thegentle-singularity<\/a>. Blog post, published June 10, 2025. Accessed March 30, 2026.\"},{\"id\":\"1a00f4f1-dbdf-4b92-9e84-9e3caf4054d1\",\"content\":\"VandeHei, Jim and Mike Allen (May 2025). \u201cBehind the Curtain: A white-collar bloodbath\u201d. In: Axios<\/em>. URL: https:\/\/www.axios.com\/2025\/05\/28\/ai-jobs-white-collar-unemployment-anthropic<\/a> (visited on 03\/30\/2026).\"},{\"id\":\"d44136e6-d729-4c3a-9a4c-3e64a7244581\",\"content\":\"Arnon, Alexander (Sept. 2025). The Projected Impact of Generative AI on Future Productivity Growth. Tech. rep. Penn Wharton Budget Model. URL: https:\/\/budgetmodel.wharton.upenn.edu\/p\/2025-09-08-the-projected-impact-of-generative-ai-on-futureproductivity-growth<\/a>.\"},{\"id\":\"be27a04d-4831-45f9-b3e1-d005d53ebd06\",\"content\":\"Filippucci, Francesco, Peter Gal, Karin Laengle, and Matthias Schief (2025). Macroeconomic Productivity Gains from Artificial Intelligence in G7 Economies<\/em>. 41. OECD. DOI: 10.1787\/a5319ab5-en<\/a>. URL: https:\/\/doi.org\/10.1787\/a5319ab5-en<\/a>\"},{\"id\":\"bf59ef57-30df-4411-ba7c-18251efd558d\",\"content\":\"Russakovsky, Olga, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei (2015). \u201cImageNet Large Scale Visual Recognition Challenge\u201d. In: International Journal of Computer Vision<\/em> 115, pp. 211\u2013252. DOI: 10.1007\/s11263-015-0816-y<\/a>; Hendrycks, Dan, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt (2020). \u201cMeasuring Massive Multitask Language Understanding\u201d. In: arXiv<\/em>. DOI: 10.48550\/arxiv.2009.03300<\/a>; Jain, Naman, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica (2024). \u201cLiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code\u201d. In: arXiv<\/em>. DOI: 10.48550\/arxiv.2403.07974<\/a>.\"},{\"id\":\"d2c185a0-3982-46a6-80dd-0cc03afe9e90\",\"content\":\"Grace, Katja, Harlan Stewart, Julia Fabienne Sandk\u00fchler, Stephen Thomas, Ben Weinstein-Raun, Jan Brauner, and R. C. Korzekwa (2024). Thousands of AI Authors on the Future of AI<\/em>. DOI: 10.48550\/arXiv.2401.02843<\/a>. URL: https:\/\/arxiv.org\/abs\/2401.02843<\/a>; Murphy, Connacher, Josh Rosenberg, Jordan Canedy, Zach Jacobs, Nadja Flechner, Rhiannon Britt, Alexa Pan, Charlie Rogers-Smith, Dan Mayland, Cathy Buffington, Simas Ku\u010dinskas, Amanda Coston, Hannah Kerner, Emma Pierson, Reihaneh Rabbany, Matthew Salganik, Robert Seamans, Yu Su, Florian Tram\u00e8r, Tatsunori Hashimoto, Arvind Narayanan, Philip E. Tetlock, and Ezra Karger (2025). The Longitudinal Expert AI Panel: Understanding Expert Views on AI Capabilities, Adoption, and Impact<\/em>. Working paper 5. Forecasting Research Institute. URL: https:\/\/forecastingresearch.org\/s\/the-longitudinalexpert-ai-panel.pdf<\/a> (visited on 03\/18\/2026).\"},{\"id\":\"511079d6-f10a-4bfe-a84f-27da9866791b\",\"content\":\"Aghion, Philippe, Benjamin F. Jones, and Charles I. Jones (2019). \u201cArtificial Intelligence and Economic Growth\u201d. In: The Economics of Artificial Intelligence: An Agenda. Ed. by Ajay Agrawal, Joshua Gans, and Avi Goldfarb. Chicago: University of Chicago Press, pp. 237\u2013290. ISBN: 9780226613338. DOI: 10.7208\/chicago\/9780226613475.003.0009<\/a>; Agrawal, Ajay, Joshua Gans, and Avi Goldfarb (2018). Prediction Machines: The Simple Economics of Artificial Intelligence. Harvard Business Review Press.\"},{\"id\":\"69bfe079-2814-498f-a12e-1b190061344d\",\"content\":\"Agrawal, Ajay, Joshua Gans, and Avi Goldfarb (2022). Power and Prediction: The Disruptive Economics of Artificial Intelligence. Harvard Business Review Press.\"},{\"id\":\"4d9867f5-0bef-4e47-9f7b-f5a8680cd0ea\",\"content\":\"Comin, Diego and Bart Hobijn (2010). \u201cAn Exploration of Technology Diffusion\u201d. In: American Economic Review<\/em> 100.5, pp. 2031\u20132059. DOI: 10.1257\/aer.100.5.2031<\/a>\"},{\"id\":\"e849cf7d-ddbe-4384-a666-2edeca7cb117\",\"content\":\"The term \u2018superforecaster\u2019 was coined in early research on subjective probability elicitation in a geopolitical context, as described in Mellers, Barbara, Lyle Ungar, Jonathan Baron, Jaime Ramos, Burcu Gurcay, Katrina Fincher, Sydney E. Scott, Don Moore, Pavel Atanasov, Samuel A. Swift, Terry Murray, Eric Stone, and Philip E. Tetlock (2014). \u201cPsychological Strategies for Winning a Geopolitical Forecasting Tournament\u201d. In: Psychological Science<\/em> 25.5, pp. 1106\u20131115. DOI: 10.1177\/
              0956797614524255<\/a>.\"},{\"id\":\"dec75bc6-4459-42cd-97a0-aab1c2da43a8\",\"content\":\"Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL:               https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>.\"},{\"id\":\"dff4ef6e-c78c-4ba2-957a-51e7c7ebe561\",\"content\":\"Trammell, Philip and Anton Korinek (2023). Economic Growth under Transformative AI<\/em>. w31815. National Bureau of Economic Research. DOI: 10.3386\/w31815<\/a>. URL: https:\/\/www.nber.org\/papers\/w31815<\/a>.\"},{\"id\":\"1f607843-9668-4971-a5d3-b1a0f2184f25\",\"content\":\"Erdil, Ege and Tamay Besiroglu (2024). Explosive Growth from AI Automation: A Review of the Arguments.<\/em> DOI: 10.48550\/arXiv.2309.11690<\/a>. URL: http:\/\/arxiv.org\/abs\/2309.11690<\/a>.\"},{\"id\":\"5ec0e99d-609c-45bc-8e4b-8d9442776788\",\"content\":\"Eloundou, Tyna, Sam Manning, Pamela Mishkin, and Daniel Rock (2024). \u201cGPTs are GPTs: Labor market impact potential of LLMs\u201d. In: Science<\/em> 384.6702, pp. 1306\u20131308.\"},{\"id\":\"c12cc757-e7bd-4f20-845f-14a19b86cbeb\",\"content\":\"Aghion, Philippe and Simon Bunel (2024). AI and Growth: Where Do We Stand?<\/em> Federal Reserve Bank of San Francisco. URL: https:\/\/www.frbsf.org\/wp-content\/uploads\/AIand-Growth-Aghion-Bunel.pdf<\/a>.\"},{\"id\":\"9e6f8a05-bf72-452b-897e-b7f91a4c7045\",\"content\":\"Filippucci, Francesco, Peter Gal, and Matthias Schief (2024). Miracle or Myth? Assessing the Macroeconomic Productivity Gains from Artificial Intelligence<\/em>. OECD. DOI: 10.1787\/b524a072-en<\/a>. URL: https:\/\/www.oecd.org\/en\/publications\/miracle-or-myth-assessing-the-macroeconomic-productivity-gains-from-artificialintelligence_b524a072-en.html<\/a>.\"},{\"id\":\"c5cfe566-fadb-4034-b8dd-5411d1098899\",\"content\":\"For more on Baumol\u2019s Cost Disease, see Baumol, William J (2012). The cost disease: Why computers get cheaper and health care doesn\u2019t.<\/em> Yale university press.\"},{\"id\":\"141535b4-0713-4679-848b-121de49aeb80\",\"content\":\"Beraja, Martin and Nathan Zorzi (2022). Inefficient Automation<\/em>. w30154. National Bureau
              of Economic Research. DOI:               10.3386\/w30154<\/a>. URL: https:\/\/www.nber.org\/papers\/w30154<\/a>.\"},{\"id\":\"41434218-d1ca-4049-ac54-89a26ca4a26a\",\"content\":\"Acemoglu, Daron (2024). \u201cThe Simple Macroeconomics of AI\u201d. In: Economic Policy 40.121, pp. 13\u201358. DOI: 10.1093\/epolic\/eiae042<\/a>.\"},{\"id\":\"26fdcdf4-07f1-4d6f-a9e1-cc1a657d5772\",\"content\":\"Comunale, Mariarosaria and Andrea Manera (2024). The Economic Impacts and the Regulation of AI: A Review of the Academic Literature and Policy Actions<\/em>. DOI: 10.5089\/9798400271663.001<\/a>. URL: https:\/\/papers.ssrn.com\/abstract=4774326<\/a>.\"},{\"id\":\"eba81395-1b68-4199-afc8-24c15d44251f\",\"content\":\"Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence<\/em>. Stanford Digital Economy Lab. URL: https:\/\/digitaleconomy.stanford.edu\/publications\/canaries-in-the-coal-mine\/<\/a>.<\/span>\"},{\"id\":\"1502acf0-353a-46da-949c-c45bfbecec09\",\"content\":\"Gimbel, Martha, Molly Kinder, Joshua Kendall, and Maddie Lee (2025). Evaluating the Impact of AI on the Labor Market: Current State of Affairs<\/em>. URL: https:\/\/budgetlab.yale.edu\/research\/evaluating-impact-ai-labor-market-current-state-affairs<\/a>.\"},{\"id\":\"4bfb5d6d-5e97-4baf-be08-701f90e49007\",\"content\":\"Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence<\/em>. Stanford Digital Economy Lab. URL: https:\/\/digitaleconomy.stanford.edu\/publications\/canaries-in-the-coal-mine\/<\/a>.<\/span>\"},{\"id\":\"c8ae48d4-385e-445c-b2a9-53733cec4bb0\",\"content\":\"Humlum, Anders and Emilie Vestergaard (May 2025). Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI<\/em>. Working Paper 33777. Revised March National Bureau of Economic Research. DOI: 10.3386\/w33777<\/a>. URL: https:\/\/www.nber.org\/papers\/w33777<\/a>.\"},{\"id\":\"e23353bb-d88a-4e20-a168-b9f6d99ba341\",\"content\":\"Frey, Carl Benedikt and Michael A. Osborne (2017). \u201cThe Future of Employment: How Susceptible Are Jobs to Computerisation?\u201d In: Technological Forecasting and Social Change<\/em> 114, pp. 254\u2013280. DOI: 10.1016\/j.techfore.2016.08.019<\/a>.\"},{\"id\":\"6b7ad398-fa0b-4f52-b6b9-6f4065360adf\",\"content\":\"Korinek, Anton and Megan Juelfs (2022). Preparing for the (Non-Existent?) Future of Work.<\/em> w30172. National Bureau of Economic Research. DOI: 10.3386\/w30172<\/a>. URL: https:\/\/www.nber.org\/papers\/w30172<\/a>.\"},{\"id\":\"7ede70b6-f469-4074-b56c-1d2c27765756\",\"content\":\"Acemoglu, Daron (2024). \u201cThe Simple Macroeconomics of AI\u201d. In: Economic Policy 40.121, pp. 13\u201358. DOI: 10.1093\/epolic\/eiae042<\/a>.\"},{\"id\":\"55f4fae7-da4a-4e36-b818-913b342af14b\",\"content\":\"AGI stands for Artificial General Intelligence and refers to a world where AI systems can perform tasks and bundles of tasks at a level that exceeds the vast majority of current workers, especially for white-collar workers whose jobs are not particularly physical in nature.\"},{\"id\":\"16da4810-8be6-401e-8f2a-575b57d9126a\",\"content\":\"Korinek, Anton (2024). Economic Consequences of Frontier AI<\/em>. NBER Working Paper No. 32980. Cambridge, MA. DOI: 10.3386\/w32980<\/a>. URL: https:\/\/www.nber.org\/papers\/w32980<\/a>.\"},{\"id\":\"82add757-055e-4c80-84a1-f722f3cd7a1c\",\"content\":\"Korinek, Anton and Joseph E. Stiglitz (2021). Artificial Intelligence, Globalization, and Strategies for Economic Development<\/em>. w28453. National Bureau of Economic Research. DOI: 10.3386\/w28453<\/a>. URL: https:\/\/www.nber.org\/papers\/w28453<\/a>.\"},{\"id\":\"0464f0e3-e006-4adb-a645-be13937b3aaa\",\"content\":\"Abbott, Ryan and Bret Bogenschneider (2018). \u201cShould Robots Pay Taxes? Tax Policy in the Age of Automation\u201d. In: Harvard Law & Policy Review<\/em> 12.1, pp. 145\u2013175. DOI: 10.2139\/ssrn.2932483<\/a>. URL: https:\/\/heinonline.org\/HOL\/P?h=hein.journals\/harlpolrv12&i=147<\/a>\"},{\"id\":\"bfede75c-9a32-4d2f-abd4-e4007ce96e5e\",\"content\":\"Comunale, Mariarosaria and Andrea Manera (2024). The Economic Impacts and the Regulation of AI: A Review of the Academic Literature and Policy Actions<\/em>. DOI: 10.5089\/9798400271663.001<\/a>. URL: https:\/\/papers.ssrn.com\/abstract=4774326<\/a>; OECD (2023). OECD Employment Outlook 2023<\/em>. Report. Accessed: 2026-03-30. Organisation for Economic Co-operation and Development. URL: https:\/\/www.oecd.org\/content\/dam\/oecd\/en\/publications\/reports\/2023\/07\/oecd- employment- outlook- 2023_904bcef3\/08785bba-en.pdf<\/a>.\"},{\"id\":\"b5803623-436a-4c18-83b5-b6b939cdba3b\",\"content\":\"Hyman, Benjamin G., Brian K. Kovak, and Adam Leive (2024). Wage Insurance for Displaced Workers<\/em>. DOI: 10.3386\/w32464<\/a>. URL: https:\/\/www.nber.org\/papers\/w32464<\/a>.\"},{\"id\":\"17726ad3-14e9-41cd-a0a9-3dba3e0d966b\",\"content\":\"Korinek, Anton and Megan Juelfs (2022). Preparing for the (Non-Existent?) Future of Work.<\/em> w30172. National Bureau of Economic Research. DOI: 10.3386\/w30172<\/a>. URL: https:\/\/www.nber.org\/papers\/w30172<\/a>.\"},{\"id\":\"5e025a24-e9a8-43ff-826f-b8f663573cf4\",\"content\":\"Abbott, Ryan and Bret Bogenschneider (2018). \u201cShould Robots Pay Taxes? Tax Policy in the Age of Automation\u201d. In: Harvard Law & Policy Review<\/em> 12.1, pp. 145\u2013175. DOI: 10.2139\/ssrn.2932483<\/a>. URL: https:\/\/heinonline.org\/HOL\/P?h=hein.journals\/harlpolrv12&i=147<\/a>\"},{\"id\":\"a31682f9-e5da-4c61-a706-6721bfb75fb9\",\"content\":\"Bastani, Spencer and Daniel Waldenstr\u00f6m (2024). AI, Automation and Taxation. URL: https:\/\/papers.ssrn.com\/sol3\/papers.cfm?abstract_id=4811796<\/a>.\"},{\"id\":\"5544c52b-3518-48dc-ae36-bd6bf00ad82e\",\"content\":\"Korinek, Anton (2024). Economic Consequences of Frontier AI<\/em>. NBER Working Paper No. 32980. Cambridge, MA. DOI: 10.3386\/w32980<\/a>. URL: https:\/\/www.nber.org\/papers\/w32980<\/a>; Anthropic (2025). Preparing for AI\u2019s Economic Impact: Exploring Policy Responses<\/em>. URL: https:\/\/www.anthropic.com\/research\/economic-policy-responses<\/a>.\"},{\"id\":\"097894f2-d836-4043-8f77-df17b9d628bb\",\"content\":\"Grace, Katja, Harlan Stewart, Julia Fabienne Sandk\u00fchler, Stephen Thomas, Ben Weinstein-Raun, Jan Brauner, and R. C. Korzekwa (2024). Thousands of AI Authors on the Future of AI<\/em>. DOI: 10.48550\/arXiv.2401.02843<\/a>. URL: https:\/\/arxiv.org\/abs\/2401.02843<\/a>.\"},{\"id\":\"8a2e8bfd-c2d5-451c-90a5-25291f25d3a0\",\"content\":\"Clark Center Forum (2025). AI and Growth<\/em>. URL: https:\/\/kentclarkcenter.org\/surveys\/ai-and-growth\/<\/a>.\"},{\"id\":\"1df40bf4-be53-45a5-a613-8786aa591a9d\",\"content\":\"Murphy, Connacher, Josh Rosenberg, Jordan Canedy, Zach Jacobs, Nadja Flechner, Rhiannon Britt, Alexa Pan, Charlie Rogers-Smith, Dan Mayland, Cathy Buffington, Simas Ku\u010dinskas, Amanda Coston, Hannah Kerner, Emma Pierson, Reihaneh Rabbany, Matthew Salganik, Robert Seamans, Yu Su, Florian Tram\u00e8r, Tatsunori Hashimoto, Arvind Narayanan, Philip E. Tetlock, and Ezra Karger (2025). The Longitudinal Expert AI Panel: Understanding Expert Views on AI Capabilities, Adoption, and Impact<\/em>. Working paper 5. Forecasting Research Institute. URL: https:\/\/forecastingresearch.org\/s\/the-longitudinalexpert-ai-panel.pdf<\/a> (visited on 03\/18\/2026).\"},{\"id\":\"7973fd83-e25a-43e6-84cd-4d6a2d2f6cbd\",\"content\":\"Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>.\"},{\"id\":\"da268163-3720-41ff-b8e3-8c46dde66257\",\"content\":\"Acemoglu, Daron (2024). \u201cThe Simple Macroeconomics of AI\u201d. In: Economic Policy 40.121, pp. 13\u201358. DOI: 10.1093\/epolic\/eiae042<\/a>; Trammell, Philip and Anton Korinek (2023). Economic Growth under Transformative AI<\/em>. w31815. National Bureau of Economic Research. DOI: 10.3386\/w31815<\/a>. URL: https:\/\/www.nber.org\/papers\/w31815<\/a>.\"},{\"id\":\"11408350-f2e0-4d90-98c3-e0fca56c58ab\",\"content\":\"IARPA (2011). IARPA Aggregative Contingent Estimation (ACE) Program<\/em>. URL: https:\/\/www.iarpa.gov\/research-programs\/ace<\/a>; Mellers, Barbara, Lyle Ungar, Jonathan Baron, Jaime Ramos, Burcu Gurcay, Katrina Fincher, Sydney E. Scott, Don Moore, Pavel Atanasov, Samuel A. Swift, Terry Murray, Eric Stone, and Philip E. Tetlock (2014). \u201cPsychological Strategies for Winning a Geopolitical Forecasting Tournament\u201d. In: Psychological Science<\/em> 25.5, pp. 1106\u20131115. DOI: 10.1177\/
              0956797614524255<\/a>.\"},{\"id\":\"05d58cfa-7472-46d3-92c4-db21f05b9e11\",\"content\":\"Mellers, Barbara, Eric Stone, Pavel Atanasov, Nick Rohrbaugh, S. Emlen Metz, Lyle Ungar, Michael M. Bishop, Michael Horowitz, Ed Merkle, and Philip E. Tetlock (2015). \u201cIdentifying and Cultivating Superforecasters as a Method of Improving Probabilistic Predictions\u201d. In: Perspectives on Psychological Science<\/em> 10.3, pp. 267\u2013281. DOI:               10.1177\/1745691615577794<\/a>.\"},{\"id\":\"cec9b96f-7b19-4198-93f6-14e454f97032\",\"content\":\"Hartman, Rachel, Aaron J. Moss, Shalom N. Jaffe, Cheskie Rosenzweig, Leib Litman, and Jonathan Robinson (2023). \u201cIntroducing Connect by CloudResearch: Advancing Online Participant Recruitment in the Digital Age\u201d. In: DOI: 10.31234\/osf.io\/ksgyr<\/a>. URL: https:\/\/doi.org\/10.31234\/osf.io\/ksgyr<\/a>.\"},{\"id\":\"2bfadadb-87d8-4fef-8d83-1d0798d0566c\",\"content\":\"Lichtendahl, Kenneth C., Yael Grushka-Cockayne, and Robert L. Winkler (2013). \u201cIs It Better to Average Probabilities or Quantiles?\u201d In: Management Science<\/em> 59.7, pp. 1594\u20131611. DOI: 10.1287\/mnsc.1120.1667<\/a>.\"},{\"id\":\"d5452222-1919-4952-b61c-a5f9315ccb76\",\"content\":\"These forecasts are: FOMC (Federal Reserve (FOMC) 2025), CBO (Congressional Budget Office 2026), OMB (Office of Management & Budget (OMB) 2025), IMF (International Monetary Fund 2025), SPF (Survey of Professional Forecasters 2025, 2026), OECD (OECD 2025), Goldman Sachs (Goldman Sachs 2025), The Conference Board (The Conference Board 2026), and Deloitte (Deloitte 2025). We focus on the baseline estimates; some forecasts consider other cases, such as Deloitte\u2019s \u201cDownside\u201d and \u201cUpside\u201d scenarios and OECD\u2019s \u201cEnergy Transition\u201d scenarios.\"},{\"id\":\"8dba68bd-bbf2-4cc1-bf43-41d3902a0eb1\",\"content\":\"These estimates assume growth rates in 2030-2045 are linearly interpolated between the 2025\u20132029 and 2045\u20132049 forecasts, as in Figure 7<\/a>.\"},{\"id\":\"bf2ca64c-3537-4efd-8b44-4206b196160c\",\"content\":\"For comparison, the average CBO forecast of TFP between 2025 and 2029 is 0.82%, and between 2044 and 2049 it is 1.1% (Congressional Budget Office (Feb. 2026). The Long-Term Budget Outlook Data: 2026 to 2056<\/em>. Tech. rep. Data available at: https:\/\/www.cbo.gov\/system\/files\/2026-02\/57054-2026-02-LTBO-econ.xlsx<\/a>. Accessed: March 19, 2026. URL: https:\/\/www.cbo.gov\/publication\/62044<\/a>.).\"},{\"content\":\"Jorgenson, Dale W and Kevin J Stiroh (2000). \u201cRaising the speed limit: US economic growth in the information age\u201d. In: Knowledge economy, information technologies and growth.<\/em> Routledge, pp. 335\u2013424.\",\"id\":\"1fe47a81-f6de-42d2-a263-182d17538877\"},{\"id\":\"169d582b-c560-401c-9bd8-85cbf948fffb\",\"content\":\"For comparison, the average forecast LFPR between 2025 and 2029 for CBO (using the Census Through 2020 Plus CBO Projection) is 62.4% and for Deloitte (after converting unemployment rate and employment-to-population rates to LFPR) is 62.0% (Congressional Budget Office (Feb. 2026). The Long-Term Budget Outlook Data: 2026 to 2056<\/em>. Tech. rep. Data available at: https:\/\/www.cbo.gov\/system\/files\/2026-02\/57054-2026-02-LTBO-econ.xlsx<\/a>. Accessed: March 19, 2026. URL: https:\/\/www.cbo.gov\/publication\/62044<\/a>; Deloitte (Dec. 2025). US Economic Forecast 2026-2030<\/em>. Tech. rep. Accessed: March 13, 2026. URL: https:\/\/www.deloitte.com\/us\/en\/insights\/topics\/economy\/us-economicforecast\/united-states-outlook-analysis.html<\/a>.).\"},{\"id\":\"d562954a-9e71-4e39-ac90-402eee9ee995\",\"content\":\"For comparison, the average forecast LFPR between 2046 and 2050 for CBO is 62.1% (Congressional Budget Office (Feb. 2026). The Long-Term Budget Outlook Data: 2026 to 2056<\/em>. Tech. rep. Data available at: https:\/\/www.cbo.gov\/system\/files\/2026-02\/57054-2026-02-LTBO-econ.xlsx<\/a>. Accessed: March 19, 2026. URL: https:\/\/www.cbo.gov\/publication\/62044<\/a>.).\"},{\"id\":\"273f119e-ad65-4ea1-9410-2a458ece805e\",\"content\":\"Both estimates are derived by comparing slow and rapid scenario LFPR forecasts. Under the rapid scenario, economists forecast a 7 p.p. decrease in the LFPR from the February 2026 reading of 62% to the 2050 median prediction of 55.0%. Under the slow scenario, where AI-driven displacement considerations are likely to be minimal, the decrease is forecasted to be 2.8 p.p., from 62% to 59.2%. This implies that (7 \u2212 2.8)\/7 \u2248 60% of the projected rapid-scenario decline is attributable to AI rather than demographics or other non-AI trends. The AI-attributable share of the decline is thus roughly 4.2 p.p. (the difference between the rapid-scenario forecast of 55.0% and the slow-scenario forecast of 59.2%). Applying this to the current U.S. civilian non-institutional population aged 16 and over of approximately 270 million yields roughly 11 million fewer labor force participants. We round to 10 million to reflect the imprecision inherent in this calculation, including uncertainty about the size of the 2050 population and the simplifying assumption that the slow scenario captures all non-AI sources of LFPR decline. We also note that worlds with rapid AI progress likely differ from worlds with slow progress in ways that extend beyond AI capabilities themselves, so this gap should be understood as the estimated effect of moving between scenarios that are focused on AI rather than a cleanly identified causal effect of AI capabilities alone. That said, the U.S. civilian population is projected to grow through 2050, which would increase the number of displaced workers, making this estimate conservative.\"},{\"id\":\"99c9d48a-60be-4078-8802-d71c0576c00a\",\"content\":\"There were 10 blocks, each with 10 occupations, and each person was randomly assigned to see one block. Two blocks mistakenly contained the same occupation twice. In these cases, we consider only one job if the respondent answered consistently for the duplicates and do not consider either if they answered inconsistently.\"},{\"id\":\"dd2d9aa7-be57-4966-8350-022412afa796\",\"content\":\"When comparing results, it is important to note that Eloundou et al. (2024) focus on LLM-exposure, while our survey respondents predicted job growth. While one component of job growth is LLM exposure, there are, of course, many other things that the survey respondent may be considering.\"},{\"id\":\"c29b1c2e-8104-4713-845c-fdbd0f38ed86\",\"content\":\"Eloundou et al. (2024) quantify how exposed the occupations are to LLMs by obtaining ratings from humans (and ChatGPT) on how exposed a job\u2019s tasks (from O*NET) are. They present several versions of AI exposure; we focus on the version using human raters and Direct Exposure plus 0.5x LLM+ Exposed, where Direct Exposure is defined as \u201cusing the described LLM via ChatGPT or the OpenAI playground can decrease the time required to complete the DWA or task by at least half (50%)\u201d and LLM+ Exposed is defined as \u201caccess to the described LLM alone would not reduce the time required to complete the activity\/task by at least half, but additional software could be developed on top of the LLM that could reduce the time it takes to complete the specific activity\/task with quality by at least half. Among these systems, we count access to image generation systems.\u201d\"},{\"id\":\"4369bedf-0621-49e3-a86a-2bb59c0c9a9c\",\"content\":\"Eloundou, Tyna, Sam Manning, Pamela Mishkin, and Daniel Rock (2024). \u201cGPTs are GPTs: Labor market impact potential of LLMs\u201d. In: Science<\/em> 384.6702, pp. 1306\u20131308.\"},{\"id\":\"3fe2a1c5-ad94-497b-bd23-e092761b60a8\",\"content\":\"Amodei, Dario (Oct. 2024). Machines of Loving Grace: How AI Could Transform the World
              for the Better<\/em>. Essay, darioamodei.com. URL:               https:\/\/www.darioamodei.com\/essay\/machines-of-loving-grace<\/a>; Brynjolfsson, Erik, Anton Korinek, and Ajay Agrawal (2025). A Research Agenda for the Economics of Transformative AI<\/em>. w34256. National Bureau of Economic Research. DOI: 10.3386\/w34256<\/a>. URL: https:\/\/www.nber.org\/papers\/w34256<\/a>.\"},{\"id\":\"dfbb633a-28c0-43ec-a9ce-b2c7dd5a79ff\",\"content\":\"Davidson, Tom (2021). Could Advanced AI Drive Explosive Economic Growth?<\/em> https:\/\/coefficientgiving.org\/research\/could-advanced-ai-drive-explosive-economic-growth\/<\/a>. Coefficient Giving, accessed March 31, 2026.\"},{\"id\":\"d92fa119-bf5c-45cf-a3af-ddd0a6c7bdf2\",\"content\":\"Cantril, Hadley (1965). The Pattern of Human Concerns<\/em>. New Brunswick, NJ: Rutgers University Press.\"},{\"id\":\"8926b457-50f7-4f9e-aaf0-cef18a613967\",\"content\":\"This would have corresponded to approximately 120 billion dollars in 2025. For context, Epoch estimates that the combined capital expenditures at Alphabet, Amazon, Meta, Microsoft, and Oracle in 2025 were 448.2 billion dollars: https:\/\/epoch.ai\/data-insights\/hyperscaler-capex-trend<\/a>\"},{\"id\":\"22f83121-b574-42ac-b1b2-1af4ae087975\",\"content\":\"Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>.\"},{\"id\":\"3e150656-59a6-4706-9441-33f63459677e\",\"content\":\"Respondents provide forecasts for the 10th , 50th , and 90th percentiles.\"},{\"id\":\"9e7268e9-b7c5-4c6d-bed4-bd375fa81b56\",\"content\":\"We fit a normal distribution to outcomes with unbounded support (GDP growth, although strictly speaking it is bounded from below by -100%), a gamma distribution to outcomes that are bounded only above or only below (median household income), and a beta distribution to variables that are bounded both above and below (LFPR, unemployment rate, wealth inequality).\"},{\"id\":\"d8b72be8-9f8a-4838-9993-c30549705080\",\"content\":\"We express the unconditional distribution for the outcome as a mixture distribution over the scenarios. The sub-distributions are given by the distribution of the outcome conditional on the rapid or bundled slow\/moderate scenario. We use a respondent\u2019s scenario probabilities to define the mixture weights.\"},{\"id\":\"c478885e-d624-4329-9ace-b021a037c9f7\",\"content\":\"In this case, the implied variance of the outcome conditional on the slow\/moderate scenario will be negative.\"},{\"id\":\"8c6a85ef-9c3e-4067-958a-4325d18a5f2b\",\"content\":\"Incoherent distributions could reflect incoherence in beliefs or misspecification in the distribution fitting procedure.\"},{\"id\":\"977a9a2e-4acf-478e-8adf-1c4008611257\",\"content\":\"We must accordingly adjust the weights in the rapid and slow\/moderate scenarios. For example, a forecaster\u2019s weight in the rapid conditional outcome distribution will be proportional to the product of their original weight and their forecasted probability for the rapid scenario.\"},{\"id\":\"4be83e18-fbe6-4d1e-bef3-ae5034a199aa\",\"content\":\"Ranjan, Roopesh and Tilmann Gneiting (2010). \u201cCombining Probability Forecasts\u201d. In: Journal of the Royal Statistical Society: Series B<\/em> 72.1, pp. 71\u201391.\"},{\"id\":\"1b3cb02b-d7a5-4223-899e-b1993f392764\",\"content\":\"Lichtendahl, Kenneth C., Yael Grushka-Cockayne, and Robert L. Winkler (2013). \u201cIs It Better to Average Probabilities or Quantiles?\u201d In: Management Science<\/em> 59.7, pp. 1594\u20131611. DOI: 10.1287\/mnsc.1120.1667<\/a>.\"},{\"id\":\"07bb23b7-bdde-4522-bc8d-81f063dff225\",\"content\":\"The within-scenario variance is the probability-weighted sum of the variance in each scenario, so we can calculate directly each scenario\u2019s contribution to within-scenario variance.\"},{\"id\":\"cf9e5e4f-8886-49fa-a8c1-c28353aa2b27\",\"content\":\"Cunningham, Tom (2025). Forecasts of AI and Economic Growth<\/em>. URL: https:\/\/tecunningham.github.io\/posts\/2025-10-19-forecasts-of-AI-growth.html<\/a>.\"},{\"id\":\"b6c548b9-cfb8-416f-b8b8-678d49512e84\",\"content\":\"See Cunningham (2025)<\/span>.\"}]"},"research_type":[4],"class_list":["post-1359","research","type-research","status-publish","has-post-thumbnail","hentry","research_type-working-paper"],"acf":[],"yoast_head":"Forecasting the Economic Effects of AI – Forecasting Research Institute<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Forecasting the Economic Effects of AI – Forecasting Research Institute\" \/>\n<meta property=\"og:description\" content=\"We surveyed academic economists, AI experts, and superforecasters to produce the most comprehensive picture yet of how AI could reshape the U.S. economy. Respondents also evaluated the impact of six different policies intended to address the impact of rapid AI progress.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai\" \/>\n<meta property=\"og:site_name\" content=\"Forecasting Research Institute\" \/>\n<meta property=\"article:modified_time\" content=\"2026-05-21T14:18:36+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/forecastingresearch.org\/wp-content\/uploads\/2025\/09\/FRI-illustration-library-1.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1232\" \/>\n\t<meta property=\"og:image:height\" content=\"928\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai\",\"url\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai\",\"name\":\"Forecasting the Economic Effects of AI – Forecasting Research Institute\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/forecastingresearch.org\\\/wp-content\\\/uploads\\\/2025\\\/09\\\/FRI-illustration-library-1.jpg\",\"datePublished\":\"2026-03-31T13:00:00+00:00\",\"dateModified\":\"2026-05-21T14:18:36+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai#primaryimage\",\"url\":\"https:\\\/\\\/forecastingresearch.org\\\/wp-content\\\/uploads\\\/2025\\\/09\\\/FRI-illustration-library-1.jpg\",\"contentUrl\":\"https:\\\/\\\/forecastingresearch.org\\\/wp-content\\\/uploads\\\/2025\\\/09\\\/FRI-illustration-library-1.jpg\",\"width\":1232,\"height\":928},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/research\\\/economic-effects-of-ai#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/forecastingresearch.org\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Forecasting the Economic Effects of AI\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/forecastingresearch.org\\\/#website\",\"url\":\"https:\\\/\\\/forecastingresearch.org\\\/\",\"name\":\"Forecasting Research Institute\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/forecastingresearch.org\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"}]}<\/script>","yoast_head_json":{"title":"Forecasting the Economic Effects of AI – Forecasting Research Institute","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai","og_locale":"en_US","og_type":"article","og_title":"Forecasting the Economic Effects of AI – Forecasting Research Institute","og_description":"We surveyed academic economists, AI experts, and superforecasters to produce the most comprehensive picture yet of how AI could reshape the U.S. economy. Respondents also evaluated the impact of six different policies intended to address the impact of rapid AI progress.","og_url":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai","og_site_name":"Forecasting Research Institute","article_modified_time":"2026-05-21T14:18:36+00:00","og_image":[{"width":1232,"height":928,"url":"https:\/\/forecastingresearch.org\/wp-content\/uploads\/2025\/09\/FRI-illustration-library-1.jpg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai","url":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai","name":"Forecasting the Economic Effects of AI – Forecasting Research Institute","isPartOf":{"@id":"https:\/\/forecastingresearch.org\/#website"},"primaryImageOfPage":{"@id":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai#primaryimage"},"image":{"@id":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai#primaryimage"},"thumbnailUrl":"https:\/\/forecastingresearch.org\/wp-content\/uploads\/2025\/09\/FRI-illustration-library-1.jpg","datePublished":"2026-03-31T13:00:00+00:00","dateModified":"2026-05-21T14:18:36+00:00","breadcrumb":{"@id":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai#primaryimage","url":"https:\/\/forecastingresearch.org\/wp-content\/uploads\/2025\/09\/FRI-illustration-library-1.jpg","contentUrl":"https:\/\/forecastingresearch.org\/wp-content\/uploads\/2025\/09\/FRI-illustration-library-1.jpg","width":1232,"height":928},{"@type":"BreadcrumbList","@id":"https:\/\/forecastingresearch.org\/research\/economic-effects-of-ai#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/forecastingresearch.org\/"},{"@type":"ListItem","position":2,"name":"Forecasting the Economic Effects of AI"}]},{"@type":"WebSite","@id":"https:\/\/forecastingresearch.org\/#website","url":"https:\/\/forecastingresearch.org\/","name":"Forecasting Research Institute","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/forecastingresearch.org\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"}]}},"_links":{"self":[{"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/research\/1359","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/research"}],"about":[{"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/types\/research"}],"version-history":[{"count":130,"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/research\/1359\/revisions"}],"predecessor-version":[{"id":2255,"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/research\/1359\/revisions\/2255"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/media\/856"}],"wp:attachment":[{"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/media?parent=1359"}],"wp:term":[{"taxonomy":"research_type","embeddable":true,"href":"https:\/\/forecastingresearch.org\/api\/wp\/v2\/research_type?post=1359"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}