|
on Artificial Intelligence |
| By: | Foltyn, Richard; Olsson, Jonna |
| Abstract: | Do large language models (LLMs) provide gender-neutral financial advice? We answer this question by prompting 33 widely used LLMs from five vendors, varying only a single word in otherwise identical prompts: “man†versus “woman.†We find that women are advised to allocate 1.8 percentage points less to equity funds than men; this gap persists across vendors, model generations, and model complexity. Providing richer investor information attenuates but does not entirely eliminate the gender gap. Since even modest allocation differences imply persistent return differentials, algorithmic financial advice can shape wealth accumulation across demographic groups. |
| JEL: | C1 G11 J16 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:21323 |
| By: | Ryota IWAMOTO; Takunori ISHIHARA; Takanori IDA |
| Abstract: | This study empirically investigates the differences in risk preferences and reference dependence between humans and generative AI. We conduct a nationwide online survey of 4, 838 individuals and generate AI responses under identical conditions by using personas constructed from demographic attributes. The results show that in gain domains, both humans and the AI select risk-averse options and exhibit similar preference patterns. However, in loss domains, AI shows a stronger risk-loving tendency and responds more sharply to individual attributes such as gender, age, and income. We retrain the AI by fine-tuning it based on human choice data. After fine-tuning, the AI’s preference distribution moves closer to that of humans, with loss-related decisions showing the greatest improvement. Using the Wasserstein distance, we also confirm that fine-tuning reduces the behavioral gap between AI and humans. |
| Keywords: | bias, risk preference, reference dependence, generative AI, persona, fine-tuning, Wasserstein distance |
| JEL: | D91 C91 |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:kue:epaper:e-25-006-v2 |
| By: | Matthieu Bunel; Elisabeth Tovar; Marie-Noëlle Lefebvre |
| Abstract: | This paper studies how large language models (LLMs) trade off moral norms against economic incentives in discriminatory hiring decisions. Bridging discrimination economics and the literature on the moral alignment in computer science, we submit 18 frontier and local LLMs to the factorial vignette experiment of a published human survey that manipulates the motive of discrimination (customer taste-based versus statistical), the cost of non-discrimination, and explicit moral injunctions. We extend this design with LLM-relevant factors: model and user personas, reasoning instructions, scenario realism, and conversational memory. In line with the literature, we find that LLMs align with humans in the direction of the effects manipulated in the survey; we also find important inter-model heterogeneity. Beyond, we contribute to the literature with, to the best of our knowledge, five novel results: 1) the market-oriented motive (customer-taste) overwhelmingly sways models in favour of discrimination, much more than what happens for human respondents; 2) models are more polarised than humans in response to moral injunctions; 3) classic prompt engineering interventions (user stated motives and model personas) have a weak impact on the models’ “moral compass”; 4) post-training alignement, not scale, shape inter-model heterogeneity, which means that de-biasing is possible but must be explicitly implemented by model providers and 5) memory effects suggest that moral permissivity in the models can be induced by conversational contextual effects. |
| Keywords: | Moral judgment on discrimination ; Artificial intelligence ; LLM audit |
| JEL: | D63 D91 C90 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:drm:wpaper:2026-16 |
| By: | Akcigit, Ufuk; Chikis, Craig A.; Dinlersoz, Emin; Goldschlag, Nathan |
| Abstract: | We construct a novel dataset linking academic publication records to U.S. Census employer–employee data to track 42, 000 AI researchers over two decades. We document systematic changes in the allocation of AI talent. Industry increasingly attracts younger and foreign-born researchers, while gender representation improves more in academia. The top 1% of publishing industry scientists now earn $1.5 million more annually than comparable academics, a fivefold increase since 2001. Rising wage premia coincide with greater sorting into large incumbent firms. Researchers who move to industry publish less but patent more, consistent with a shift from open science toward proprietary innovation. |
| Keywords: | Artificial intelligence; open science; Innovation; Research and development |
| JEL: | I23 J45 L33 O31 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:21293 |
| By: | Koren, Miklós; Békés, Gábor; Hinz, Julian; Lohmann, Aaron |
| Abstract: | Generative AI is changing how software is produced and used. In vibe coding, an AI agent builds software by selecting and assembling open-source software (OSS), often without users directly reading documentation, reporting bugs, or otherwise engaging with maintainers. We study the equilibrium effects of vibe coding on the OSS ecosystem. We develop a model with endogenous entry and heterogeneous project quality in which OSS is a scalable input into producing more software. Users choose whether to use OSS directly or through vibe coding. Vibe coding raises productivity by lowering the cost of using and building on existing code, but it also weakens the user engagement through which many maintainers earn returns. When OSS is monetized only through direct user engagement, greater adoption of vibe coding lowers entry and sharing, reduces the availability and quality of OSS, and reduces welfare despite higher productivity. Sustaining OSS at its current scale under widespread vibe coding requires major changes in how maintainers are paid. |
| JEL: | O33 L86 D85 |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:21145 |
| By: | Mert Demirer; John J. Horton; Nicole Immorlica; Brendan Lucier; Peyman Shahidi |
| Abstract: | Production is a sequence of steps that can be executed (1) manually, (2) augmented with AI, or (3) fully automated within contiguous AI-executed steps called ''chains.'' Firms optimally bundle steps into tasks and then jobs, trading off specialization gains against coordination costs. We characterize the optimal assignment of humans and AI to steps and the firm's resulting job structure, showing that comparative advantage logic can fail with AI chaining. The model implies non-linear productivity gains from AI quality improvements and admits a CES representation at the macro level. Empirical evidence supports the model's key predictions that (1) AI-executed steps co-occur in chains, (2) dispersion of AI-exposed steps lowers AI execution at the job level, and (3) adjacency to AI-executed steps increases the likelihood that a step is AI-executed. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.15960 |
| By: | Alex Chan |
| Abstract: | This paper studies market design for generative AI intermediation. AI answer systems can improve user experience while diverting visits that finance publisher content and generate source-level quality signals. I show that an AI platform that underinternalizes future content reproduction retains too little referral traffic and can make costly open-web information subcritical, even with truthful content, accurate answers, and rational users. The mechanism can be self-reinforcing: less source-level measurement weakens conventional search, inducing further AI reliance. Sustainable repair requires replacing displaced revenue and deleted measurement through visitor-replacement royalties, audited provenance, human-information audits, and keystone-topic compensation. |
| JEL: | D4 D43 D47 D49 D62 D8 D82 D83 L82 L86 O3 O33 |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35344 |
| By: | Richard Archer (Yale University); Soheil Ghili (Yale University); Nima Haghpanah (Yale University) |
| Abstract: | As AI systems shift from directing users to content toward consuming it directly, publishers need a new revenue model: charging AI crawlers for content access. This model, called pay-per-crawl, must solve a problem of mechanism selection at scale: content is too heterogeneous for a fixed pricing framework. Different sub-types warrant not only different price levels but different pricing rules based on different unstructured features, and there are too many to enumerate or design by hand. We propose the LM Tree, an adaptive pricing agent that grows a segmentation tree over the content library, using LLMs to discover what distinguishes high-value from low-value items and apply those attributes at scale, from binary purchase feedback alone. We evaluate the LM Tree on real content from a major German technology publisher, using 8, 939 articles and 80, 451 buyer queries with willingness-to-pay calibrated from actual AI crawler traffic. The LM Tree achieves a 65% revenue gain over a single static price and a 47% gain over two-category pricing, outperforming even the publisherÕs own 8-segment editorial taxonomy by 40%Ñrecovering content distinctions the publisherÕs own categories miss. |
| Date: | 2026–04–01 |
| URL: | https://d.repec.org/n?u=RePEc:cwl:cwldpp:2516 |
| By: | Olivier Bos; Stefano Bosi |
| Abstract: | We study how artificial intelligence (AI) interacts with social communication networks to shape the stability of collective knowledge. Agents exchange information through a network while receiving AI-generated content, and AI systems retrain on the aggregate social information they influence. This interaction generates two feedback forces: an AI contagion channel, through which distortions diffuse across the network, and an AI social distortion multiplier, through which retraining amplifies past errors. Despite the high dimensionality of the environment, we show that the long-run behavior of the system admits a two-dimensional representation whose spectral radius determines whether AI-mediated information systems are dynamically stable or unstable. We characterize a sharp regulatory frontier identifying the minimum filtering required for stability and show how network topology shapes systemic informational risk. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.15206 |
| By: | James Wabenga Yango |
| Abstract: | This paper develops a general equilibrium overlapping-generations model with endogenous fertility, in which firms accumulate both physical and artificial intelligence (AI) capital, and uses it to study the macroeconomic transmission of two structural disturbances: an AI technology shock and a longevity shock. The AI shock acts as a capital-demand disturbance: it raises all rates of return, most sharply the return to AI capital, reallocates investment from physical to AI capital, and produces a front-loaded output expansion that decays monotonically. The longevity shock acts as a saving-supply disturbance: it deepens the aggregate capital stock, compresses returns and the real interest rate, and generates hump-shaped, persistent dynamics. The two shocks move fertility in opposite directions: AI raises it modestly through an income effect, while longevity lowers it by strengthening the life-cycle saving motive and the cost of childrearing. A forecast-error variance decomposition attributes most aggregate volatility to the longevity shock, while the AI shock dominates the variance of the return to AI capital. Fertility is strongly countercyclical and almost perfectly negatively correlated with hours worked, placing household time allocation at the center of the mechanism. Robustness checks across the capital share, the shock persistence, and the utility specification show that only an empirically implausible labor-AI elasticity reverses the wage and fertility signs. A welfare analysis finds the AI shock welfare-improving under complementarity, whereas longevity produces a short-run welfare loss that recedes as capital deepening raises wages, since households initially compress consumption and fertility to finance a longer retirement. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.22037 |
| By: | Cantarella, Michele (Technical University of Denmark - DTU); Molinari, Giuseppe (University of Modena and Reggio Emilia); Strozzi, Chiara (University of Modena and Reggio Emilia) |
| Abstract: | This paper investigates how Artificial Intelligence reshapes the human capabilities that jobs require. Using longitudinal O*NET data for the U.S. labour market over 2011–2025, we distinguish among three types of human capabilities - abilities, skills, and knowledge - and construct two measures of human capabilities’ exposure to AI: one based on observed progress in Generative AI benchmark performance and one based on the broader evolution of AI-related scientific and public attention. We document a dual pattern. Within occupations, greater AI exposure is associated with higher proficiency requirements for selected capabilities. At the occupational level, more exposed occupations exhibit a compression in the overall breadth of capabilities required. Together, these findings suggest that AI is driving a process of occupational restructuring, leading to more specialized and less diverse capability profiles embedded in jobs. |
| Keywords: | artificial intelligence, AI exposure, skill reallocation, task content, deskilling |
| JEL: | J24 J21 O33 |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18751 |
| By: | Aldasoro, Inaki; Gambacorta, Leonardo; Pál, Rozália; Revoltella, Debora; Weiss, Christoph; Wolski, Marcin |
| Abstract: | This paper provides new evidence on how the adoption of artificial intelligence (AI) affects productivity and employment in Europe. Using matched EIBIS-ORBIS data on more than 12, 000 non-financial firms in the European Union (EU) and United States (US), we instrument the adoption of AI by EU firms by assigning the adoption rates of US peers to isolate exogenous technological exposure. Our results show that AI adoption increases the level of labor productivity by 4%. Productivity gains are due to capital deepening, as we find no adverse effects on firm-level employment. This suggests that AI increases worker output rather than replacing labor in the short run, though longer-term effects remain uncertain. However, productivity benefits of AI adoption are unevenly distributed and concentrate in medium and large firms. Moreover, AI-adopting firms are more innovative and their workers earn higher wages. Our analysis also highlights the critical role of complementary investments in software and data or workforce training to fully unlock the productivity gains of AI adoption. |
| Keywords: | Artificial intelligence; Firm productivity; Europe; Digital transformation |
| JEL: | D22 J24 L25 O33 O47 |
| Date: | 2026–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:21082 |
| By: | Ide, Enrique |
| Abstract: | Recent advances in Artificial Intelligence (AI) have sparked expectations of unprecedented economic growth. Yet, by enabling senior workers to accomplish more tasks independently, AI may reduce entry-level opportunities, raising concerns about how future generations will acquire expertise. This paper develops a model to examine how automation and AI affect the intergenerational transmission of tacit knowledge — practical, hard-to-codify skills critical to workplace success. I show that the competitive equilibrium features socially excessive automation of early-career tasks, and that improvements in such automation generate an intergenerational trade-off: they raise short-run productivity but weaken the skills of future generations, slowing long-run growth — sometimes enough to reduce welfare. Back-of-the-envelope calculations suggest that AI-driven entry-level automation could reduce the long-run annual growth rate of U.S. per-capita output by 0.05 to 0.35 percentage points, depending on its scale. I further show that AI co-pilots can partially offset lost learning by assisting individuals who fail to acquire skills early in their careers. However, they may also weaken juniors’ incentives to develop such skills. These findings highlight the importance of preserving and expanding early-career learning opportunities to fully realize AI’s potential. |
| Keywords: | Artificial intelligence; Automation; Long-run economic growth; Tacit knowledge |
| JEL: | O40 O33 J24 |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20940 |
| By: | Ilse Lindenlaub (Department of Economics, Yale University); Ryungha Oh (Department of Economics, Yale University); Mar’a Alejandra Rodr’guez Vega (Department of Economics, Yale University); Laura Veldkamp (Columbia Business School, Columbia University) |
| Abstract: | We document and explain the gap between measures of AI exposure and measures of AI adoption in the workplace. This leads us to propose a new AI adoption index based on comparative advantage. Using the representative German DiWaBe employee survey linked to worker and establishment information, we compare worker-reported AI use to prominent exposure measures and find that the relationship is weak. Motivated by this gap, we develop a framework in which adoption depends not only on technical feasibilityÑAIÕs absolute advantage measured by exposureÑbut on profitabilityÑAIÕs comparative (dis)advantage relative to a specific workerÑbalancing AI productivity against AI user costs and worker productivity against wages. We operationalize this framework at the task level by (i) estimating worker productivity relative to pay, (ii) mapping exposure indices into AI productivity, and (iii) inferring task-specific AI user costs from revealed-preference adoption. The resulting occupation-level index accounts for 60% of cross-occupation variation in observed AI adoption, compared to 14% for an exposure-only model. The two approaches diverge substantially for approximately 30% of workers, highlighting that comparative advantageÑnot exposure aloneÑis crucial for assessing AIÕs labor-market impact. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:cwl:cwldpp:2532 |
| By: | Cullen, Zöe; Faia, Ester; Guglielminetti, Elisa; Perez-Truglia, Ricardo; Rondinelli, Concetta |
| Abstract: | We present the first large-scale field experiment test of strategic complementarities in firms' technology adoption. Our experiment was embedded in a Bank of Italy survey covering around 3, 000 firms. We elicited firms' beliefs about competitors' adoption of two advanced technologies: Artificial Intelligence (AI) and robotics. We randomly provided half of the sample with accurate information about adoption rates. Most firms substantially underestimated competitors' current adoption, and when provided with information, they updated their expectations about competitors' future adoption. The information increased firms' own intended future adoption of robotics, although we do not observe a significant effect on AI adoption. Our findings provide causal evidence on coordination in innovation and illustrate how information frictions shape technology diffusion. |
| Keywords: | Innovation |
| JEL: | O33 D22 C93 L21 |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20894 |
| By: | Sébastien Houde; Wenjun Wang |
| Abstract: | This paper investigates the relationship between AI adoption and carbon emission intensity. Using micro-level data from Chinese firms, we find that carbon intensity decreases following the adoption of AI. The effect is particularly pronounced among large firms, those headquartered in AI hubs, and those in high-carbon intensity sectors. We investigate several mechanisms and find that AI adoption is also associated with increases in energy management processes, green innovation, inventory efficiency, overall productivity, and the share of specialized labor. We find that AI-induced carbon reductions are subject to a large rebound effect of approximately 70%. |
| Keywords: | artificial intelligence, carbon emissions, energy intensity, green innovation |
| JEL: | D22 L11 O33 Q54 Q55 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_12803 |
| By: | Gambacorta, Leonardo; Kharroubi, Enisse; Mehrotra, Aaron; Oliviero, Tommaso |
| Abstract: | This paper investigates whether the positive effects of generative artificial intelligence (gen AI) on growth rate of value added differ across countries in the short run. Using an empirical strategy inspired by Rajan and Zingales (1998) and a dataset covering 56 economies and 16 industries, we find that the differential growth effects arise from variations in sectoral exposure to cognitive and knowledge-intensive activities, differences in production structures, and countries’ AI preparedness. Our results suggest that, on average, gen AI is likely to benefit advanced economies more than emerging market economies, thereby widening global income disparities in the near term. |
| Keywords: | General artificial intelligence; Emerging market and developing economies; Economic growth; Productivity differences; Technology diffusion |
| JEL: | E24 O47 O57 |
| Date: | 2026–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20992 |
| By: | Matthew O. Jackson; Zafer Kanik |
| Abstract: | We examine the economic impact of increasingly productive AI and policies that spread its benefits across the economy. Improvements in AI productivity trigger labor reallocation and changes in absolute and relative wages for different types of labor. Wages of labor that is essential for building AI increase faster than overall GDP. Wages of labor that is substituted for by AI decrease in both absolute and relative terms. Wages of labor that is used only in final goods production and is not displaced by AI increase in line with overall GDP. We contrast the impact of productivity gains depending on whether AI production is competitive or monopolistic. Monopoly production of AI restricts its deployment, slowing the transition and impact of AI. Optimal tax and regulatory policies that achieve Pareto-improvements differ depending on whether there is competition in AI production. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.01101 |
| By: | Loschiavo, David; Armantier, Olivier; Dalla-Zuanna, Antonio; Gambacorta, Leonardo; Moscatelli, Mirko; Supino, Ilaria |
| Abstract: | This paper explores the household adoption of Generative Artificial Intelligence (GenAI) in the United States and Italy, leveraging survey data to compare usage patterns, demographic influences, and employment sectoral composition effects. Our findings reveal higher adoption rates in the US, driven by socio-demographic differences between the two countries. Despite their lower usage of GenAI, Italians are more confident in its potential to improve their well-being and financial situation. Both Italian and US users tend to trust GenAI tools less than humanoperated services, but Italians report greater relative trust in government and institutions when handling personal data with GenAI tools. |
| Keywords: | Generative AI; Technology adoption; Cross-country comparisons; Socio-demographic factors; Trust in technology; Cultural attitudes |
| JEL: | O33 D10 J24 |
| Date: | 2026–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:21083 |
| By: | Alam, M. Jahangir; Boyle, Shane; Li, Huiyu; Sekhposyan, Tatevik |
| Abstract: | Recent research suggests that generic large language models (LLMs) can match the accuracy of traditional methods when forecasting macroeconomic variables in pseudo out-of-sample settings generated via prompts. This paper assesses the out-of-sample forecasting accuracy of LLMs by eliciting real-time forecasts of U.S. inflation from ChatGPT. We find that out-of-sample predictions are largely inaccurate and stale, even though forecasts generated in pseudo out-of-sample environments are comparable to existing benchmarks. Our results underscore the importance of out-of-sample benchmarking for LLM predictions. |
| Date: | 2026–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:21057 |
| By: | Andreas Ferrara |
| Abstract: | Large language models (LLMs) are lowering the entry barriers to working with exciting data sources that used to require strong data science skills, such as handwritten ledgers, text, images, or sound recordings. This guide provides an introduction for researchers who are new to LLMs. It sets out a step-by-step workflow for turning a research idea into working code and data, and describes the four main ways of interacting with an LLM: the chat window, editor-integrated assistants, agentic coding tools, and the API. It then works through the decisions a practitioner meets in sequence, beginning with whether an LLM is the right tool and whether the data are allowed to be sent to one, then how to select models, write prompts, manage context limits, and control costs, and finally how to validate, reproduce, document, and correct LLM-generated measures in regression settings. A review of recent research shows how these tools already extract, link, harmonize, and classify historical data at scale. Four worked examples with replication files illustrate the use of LLMs. They classify emotions in paintings, link census records without names, measure newspaper salience and sentiment around the 1882 Chinese Exclusion Act, and score the emotional delivery of Franklin D. Roosevelt's wartime speeches. The guide also condenses the workflow, the best-practice recommendations, and the preparation of replication packages into summary tables and checklists to aid applied economists. |
| JEL: | C55 C8 N0 |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35374 |
| By: | Ershov, Daniel; Lyons, Elizabeth |
| Abstract: | The use of autonomous pricing algorithms has grown across markets in recent years, and many firms outsource their pricing algorithms to third-party developers. While recent evidence highlights the potential for pricing algorithms to influence competition, the design of these algorithms and how the risks of these algorithms can be managed is less clear. We collect pricing algorithms from both programmers, via an RCT on Upwork.com, and from an LLM to characterize how programmers think about pricing algorithms, and the extent to which third-party programmer decisions can be adjusted using simple non-technical prompts. We show that, on average, programmer and LLM-written algorithms are less sophisticated than the Q-learning algorithms used in the theoretical literature. We also show that a prompt aimed to focus programmer attention on economic fundamentals can help human programmers to produce algorithms that better match competitive prices. Finally, we find the biggest threat of supra-competitive prices is generated via mis-specification of demand models. |
| Keywords: | Price competition; Management; Artificial intelligence |
| JEL: | L22 L24 O32 L41 D43 |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20901 |
| By: | Fang, Tony (Memorial University of Newfoundland, NL, Canada); Lin, Carl (Bucknell University Lewisburg, PA, USA); Liu, Qing (Hefei University of Technology, Anhui, China) |
| Abstract: | We construct city–year measures of AI labor demand from 1.6 million online job postings between 2016 and 2024, and merge them with nationally representative microdata from the China Family Panel Studies (2016–2022). Fixed-effects estimates show that local AI labor demand has positive impacts on individual wages: a one-unit increase in AI demand (1, 000 postings, firms, or job titles) raises wages by about 0.2–0.3 percent. Women experience stronger gains—about 0.5–0.7 percent per unit increase—while men show no measurable effect. Wage effects are largest in Western provinces, and in China’s major AI-cluster cities where complementary production factors and digital infrastructure are most developed. Occupational analyses further show that women’s gains are concentrated in service-oriented, less skill-intensive jobs where AI complements interpersonal and coordination tasks rather than substituting them. Overall, AI diffusion generates meaningful but unequal labor market spillovers, with wage gains concentrated among women, dynamic regions, and human–AI complementary occupations, underscoring both the opportunities of technological transformation and the challenges of achieving inclusive growth. |
| Keywords: | artificial intelligence (AI), labor market, wages, productivity, China |
| JEL: | I23 J24 |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18740 |
| By: | David Deming, Christopher Ong, Lawrence H. Summers |
| Abstract: | This paper explores past episodes of technological disruption in the US labor market, with the goal of learning lessons about the likely future impact of artificial intelligence (AI). The authors measure changes in the structure of the US labor market going back over a century in two ways. First, they examine the relative frequencies of occupations from 1880-2020. Over that period, the structure of the US labor market underwent two large shifts: From 1880 to 1960, workers moved out of agriculture jobs. In 1880, 41 percent of all workers in the US economy were employed as farmers or farm laborers. This share fell consistently by 4 percentage points per decade, and by 1960 only 6 percent of US employment was in agriculture. From 1960 to 1980, jobs moved from the factory to the office. The share of workers employed in blue-collar jobs like manual labor, construction, production and manufacturing, transportation, and maintenance and repair remained relatively constant at 40 percent from 1880 to 1960, then fell ten percentage points by 1980. It has experienced a slower decline since, reaching 20 percent by 2010. They also find that the pace of change, as measured by occupational churn , has slowed over time: the years spanning 1990 to 2017 were less disruptive than any prior period we measure, going back to 1880. This comparative decline is not because the job market is stable today but rather because past changes were so profound. These changes were caused by general-purpose technologies (GPTs), like steam power and electricity, which dramatically disrupted the twentieth-century labor market over the course of several decades. The authors argue that AI could be a GPT on the scale of prior disruptive innovations and suggest that there are two patterns in the data that might indicate that AI is leading to labor market disruptions along the lines of past GPTs. First, increased investment in new technologies and a J-curve pattern of productivity growth in AI-exposed sectors. Second, large but steady declines in employment share for AI-exposed jobs, especially jobs in sectors where consumers don’t increase consumption with rising income. They present early evidence of such signs in four stylized facts: The labor market is no longer polarizing— employment in low- and middle-paid occupations has declined, while highly paid employment has grown. Employment growth has stalled in low-paid service jobs. The share of employment in STEM jobs has increased by more than 50 percent since 2010, fueled by growth in software and computer-related occupations. Retail sales employment has declined by 25 percent in the last decade, likely because of technological improvements in online retail. The post pandemic labor market is changing very rapidly, and a key question is whether this faster pace of change will persist into the future. |
| Keywords: | labor economics, AI, automation, technology |
| Date: | 2024–10–01 |
| URL: | https://d.repec.org/n?u=RePEc:cxx:wpaper:technological-disruption-in-the-us-labor-market |
| By: | Gu, Gyun Cheol |
| Abstract: | We replicate and extend the ultimatum game experiment of Araujo and Uhlig (2026) to five consumer-facing large language models (LLMs)—ChatGPT, Claude, Copilot, Gemini Flash, and Gemini Pro—across four scenario types (HH, HA, AH, AA), four stake levels ($10 to $10, 000), and ten repetitions per configuration, yielding 6, 832 proposer and 8, 532 responder observations. Four findings emerge. First, all five models propose shares in the 30–47% range, squarely within the human empirical benchmark and absent the extreme behavioral modes documented in earlier research-grade models, suggesting that alignment training has compressed the behavioral distribution toward human norms. Second, every model exhibits twosided sensitivity to human presence: proposed shares rise when the Responder is human (+4 to +25 p.p.) and minimum acceptable thresholds rise when acting on behalf of a human (+11 to +26 p.p.), a pattern that survives even when no human principal is being served and is inconsistent with simple principal–agent alignment. Third, all five models forgo 25–63% of feasible payoff, confirming that consumer LLMs are not payoff-maximizing agents. Fourth, responder thresholds decline significantly with stake size across all models— consistent with rational expected-utility behavior—while proposer stake sensitivity is heterogeneous. We interpret these patterns as evidence of identity internalization: successive rounds of reinforcement learning with human feedback cause models to behave as if they are human rather than merely as if they prefer human-like outcomes. |
| Keywords: | ultimatum game, large language models, human identity, alignment, RLHF, behavioral economics |
| JEL: | C70 C90 |
| Date: | 2026–06–07 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:129505 |
| By: | Ide, Enrique; Talamas, Eduard |
| Abstract: | Artificial Intelligence (AI) is reshaping offshoring and globalization by automating knowledge work and altering trade patterns. We analyze this transformation in a two-region world where firms structure work hierarchically to use knowledge efficiently: the most knowledgeable individuals specialize in problem-solving, while others perform routine work. Before AI, the Advanced Economy specializes in problem-solving services, while the Emerging Economy focuses on routine knowledge work. We model AI as a technology that converts compute into autonomous “AI agents, †which serve as perfect substitutes for humans with a given level of knowledge. Reflecting the concentration of AI infrastructure in advanced economies, we assume that all compute is located in the Advanced Economy. We show that basic AI reduces the Advanced Economy’s net exports of problem-solving services, potentially reversing pre-AI trade patterns. In contrast, sophisticated AI increases the Advanced Economy’s net exports of problem-solving services, reinforcing existing trade patterns. We also examine the effects of restricting AI autonomy, finding that a global restriction redistributes AI’s benefits toward lower-skilled workers, while a regional restriction—such as banning autonomous AI in the Emerging Economy—does little to benefit lower-skilled workers and harms the most knowledgeable individuals in that region. Our results underscore the need for a coordinated global approach to AI regulation. |
| Keywords: | Artificial intelligence; Offshoring; International trade |
| JEL: | F16 F66 O33 |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20801 |
| By: | Kwon, Byeungchun; Park, Taejin; Rungcharoenkitkul, Phurichai; Smets, Frank |
| Abstract: | Macroeconomic indicators provide quantitative signals that must be pieced together and interpreted by economists. We propose a reversed approach of parsing press narratives directly using Large Language Models (LLM) to recover growth and inflation sentiment indices. A key advantage of this LLM-based approach is the ability to decompose aggregate sentiment into its drivers, readily enabling an interpretation of macroeconomic dynamics. Our sentiment indices track hard-data counterparts closely, providing an accurate, near real-time picture of the macroeconomy. Their components–demand, supply, and deeper structural forces–are intuitive and consistent with prior model-based studies. Incorporating sentiment indices improves the forecasting performance of simple statistical models, pointing to information unspanned by traditional data. |
| JEL: | E30 E44 E60 C55 C82 |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20828 |
| By: | Christopher W. Karvetski; Sheldon S. Huang; Simas Ku\v{c}inskas; Nadja Flechner; Jingyu Hu; Philip Tetlock; Ezra Karger |
| Abstract: | Decision-makers routinely rely on expert judgments accompanied by written explanations, yet explanation quality is difficult to measure at scale. Forecasting tournaments offer a natural testing ground: probabilistic judgments are paired with natural-language rationales and scored against realized outcomes. We introduce Explanation Quality Markers (EQMs), a set of sixty theory-guided reasoning patterns scored by large language models (LLMs). In a pre-registered analysis of over 55, 000 forecast-rationale pairs from a multiyear forecasting tournament, EQMs predict accuracy at both the forecast and forecaster levels, consistently outperforming pre-LLM text-analysis methods. More than 90% of statistically significant pattern-level EQM-accuracy correlations match our directional hypotheses. The signal is asymmetric: EQMs identify likely underperformers more reliably than they distinguish the very best forecasters. Benchmarked against traditional indicators of forecasting skill, EQMs are the strongest predictor at the forecast level and competitive at the forecaster level, though weaker than prior accuracy. Human ratings of rationale quality are less consistently correlated with accuracy and place disproportionate weight on rationale length. Results transfer to an independent forecasting study. EQMs provide a scalable, interpretable method for extracting judgment-relevant information from written explanations. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.30987 |