|
on Artificial Intelligence |
| By: | Ali Merali |
| Abstract: | This paper derives `Scaling Laws for Economic Impacts' -- empirical relationships between the training compute of Large Language Models (LLMs) and professional productivity. In a preregistered experiment, over 500 consultants, data analysts, and managers completed professional tasks using one of 13 LLMs. We find that each year of AI model progress reduced task time by 8%, with 56% of gains driven by increased compute and 44% by algorithmic progress. However, productivity gains were significantly larger for non-agentic analytical tasks compared to agentic workflows requiring tool use. These findings suggest continued model scaling could boost U.S. productivity by approximately 20% over the next decade. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2512.21316 |
| By: | Jian, Wenze; Lu, Hang; Yang, Zimo; Zhong, Ziqi |
| Abstract: | Using data on Chinese A-share listed firms and regions from 2011–2023, this paper employs a difference-in-differences (DID) framework to evaluate the productivity returns to artificial intelligence (AI) application from both firm-level and societal perspectives. The findings are as follows: First, AI intensity significantly increases firms' total factor productivity (TFP). Second, AI intensity significantly increases social TFP. Third, green financial innovation exerts a significant positive mediating effect on the pathway from AI intensity to firm TFP. Fourth, green financial innovation also partially mediates the pathway from AI intensity to social TFP. Substantively, the paper links micro-level firm transformation with macro-level regional performance, providing empirical evidence and policy implications for understanding the transmission mechanism from digitalization to greening to high-quality growth. |
| Keywords: | AI intensity; TFP; green financial innovation |
| JEL: | F3 G3 J50 |
| Date: | 2026–02–28 |
| URL: | https://d.repec.org/n?u=RePEc:ehl:lserod:130473 |
| By: | Mario Passalacqua (UQAM - Université du Québec à Montréal = University of Québec in Montréal); Robert Pellerin (MAGI - Département de Mathématiques et de Génie Industriel - EPM - École Polytechnique de Montréal); Florian Magnani (MAGELLAN - Laboratoire de Recherche Magellan - UJML - Université Jean Moulin - Lyon 3 - Université de Lyon - Institut d'Administration des Entreprises (IAE) - Lyon); Laurent Joblot (LISPEN - Laboratoire d’Ingénierie des Systèmes Physiques et Numériques - Arts et Métiers Sciences et Technologies); Frédéric Rosin (LISPEN - Laboratoire d’Ingénierie des Systèmes Physiques et Numériques - Arts et Métiers Sciences et Technologies); Esma Yahia (LISPEN - Laboratoire d’Ingénierie des Systèmes Physiques et Numériques - Arts et Métiers Sciences et Technologies); Pierre-Majorique Léger (HEC Montréal - HEC Montréal) |
| Abstract: | Advancements in artificial intelligence (AI) have ushered in the era of the fourth industrial revolution, transforming workplace dynamics with AI's enhanced decision-making capabilities. While AI has been shown to reduce worker mental workload, improve performance, and enhance physical safety, it also has the potential to negatively impact psychosocial factors, such as work meaningfulness, worker autonomy, and motivation, among others. These factors are crucial as they impact employee retention, well-being, and organizational performance. Yet, the impact of automating decision-making aspects of work on the psychosocial dimension of human-AI interaction remains largely unknown due to the lack of empirical evidence. To address this gap, our study conducted an experiment with 102 participants in a laboratory designed to replicate a manufacturing line. We manipulated the level of AI decision support-characterized by the AI's decision-making control-to observe its effects on worker psychosocial factors through a blend of perceptual, physiological, and observational measures. Our aim was to discern the differential impacts of fully versus partially automated AI decision support on workers' perceptions of job meaningfulness, autonomy, competence, motivation, engagement, and performance on an error-detection task. The results of this study suggest the presence of a critical boundary in automation for psychosocial factors, demonstrating that while some automation of decision selection can nurture work meaningfulness, worker autonomy, competence, self-determined motivation, and engagement, there is a pivotal point beyond which these benefits can decline. Thus, balancing AI assistance with human control is vital to protect psychosocial well-being. Practically, industry and operations managers should keep employees involved in decision making by adopting partial, confirm-or-override AI systems that sustain motivation and engagement, boosting retention and productivity. |
| Keywords: | Human-centred AI, Motivation, Engagement, Psychosocial, Industry 4.0, Industry 5.0, Human-centered AI |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05345071 |
| By: | Evgenidis Anastasios; Fasianos Apostolos; Papapanagiotou George; Lazarou Nicholas Joseph (European Commission - JRC) |
| Abstract: | Recent advances in Artificial Intelligence (AI) and the growing role of these technologies in enhancing productivity have attracted significant research and policy attention, yet the determinants of AI innovation remain relatively understudied. This study contributes to this emerging literature by examining the role of public R&D spending in fostering AI-related innovation across EU regions. Our analysis draws on bibliographic information from all patents registered at the European Patent Office (EPO) between 1980 and 2023. Using textual analysis of patent abstracts, we identify the share of AI patents among total patents and construct a novel dataset that allocates AI patents to NUTS-2 regions based on inventor addresses. This regional mapping enables us to assess the impact of public R&D funding on AI innovation while addressing endogeneity concerns by instrumenting regional public R&D spending with national defence-related R&D expenditure. The results show that public R&D plays a significant role in driving AI innovation: a 1% increase in public R&D spending raises AI patent output by approximately 0.27%. These findings speak directly to Europe’s innovation policy framework, providing evidence that public investment remains a powerful lever for stimulating AI development. They also reinforce the rationale for sustained funding under Horizon Europe, the Digital Europe Programme, and national innovation strategies aimed at building technological and reducing regional disparities in AI advancement. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:ipt:termod:202512 |
| By: | Guanyu Lu (Faculty of Political Science and Economics, Waseda University, 1-6-1 Nishiwaseda, Shinjuku-ku, Tokyo 169-8050.); Hajime Katayama (Faculty of Commerce, Waseda University, 1-6-1 Nishiwaseda, Shinjuku-ku, in the Tokyo, 169-8050.); Toshi H. Arimura (Faculty of Political Science and Economics, Waseda University, 1-6-1 Nishiwaseda, Shinjuku-ku, Tokyo 169-8050.); Shohei Morimura (Research Institute for Environmental Economics and Management, Waseda University, 1-6-1 Nishiwaseda, Shinjuku-ku, Tokyo 169-8050.); Tomoichi Ishiwatari (iGRID SOLUTIONS Inc., 3-7-4 Kojimachi, Chiyoda-ku, Tokyo 102-0083.); Tetsu Iwasaki (iGRID SOLUTIONS Inc., 3-7-4 Kojimachi, Chiyoda-ku, Tokyo 102-0083.) |
| Abstract: | This study examines the impact of “Enudge, †an artificial intelligence (AI) energy management system (EMS), on electricity consumption in the retail sector. As retail installations increasingly contribute to nonindustrial CO₂ emissions, conventional EMSs frequently fail to manage the complex and variable energy demands in these settings. By leveraging a difference-in-differences framework on store-level data from over 1, 700 retail stores in Japan between November 2018 and December 2023, this study finds that installation of AI EMS-Enudge reduces electricity consumption by an average of 1.9%. However, this reduction effect declines over time, with electricity savings diminishing within five to ten months. This decay effect is consistent with the decrease in user interaction with the recommendations provided by AI, suggesting that user engagement may play a crucial role in reducing electricity consumption. Heterogeneity analyses reveal that the system’s performance varies across retail establishments and seasonal contexts. Moreover, a cost-benefit analysis aimed at exploring break-even tariffs and implied abatement costs highlights that the installation of an AI EMS can contribute to cost savings, especially under high tariffs and higher-carbon grids. |
| Keywords: | Artificial intelligence energy management system, electricity consumption, difference in differences, energy savings |
| JEL: | Q41 Q48 C23 |
| Date: | 2026–01 |
| URL: | https://d.repec.org/n?u=RePEc:was:dpaper:2502 |
| By: | Leonardo Gambacorta; Enisse Kharroubi; Aaron Mehrotra; Tommaso Oliviero |
| 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: | generative artificial intelligence, emerging market economies, economic growth; productivity differentials, technological readiness, sectoral exposure to AI |
| JEL: | E24 O47 O57 |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:bis:biswps:1321 |
| By: | Siddhartha Chib; Fei Tan |
| Abstract: | We show how state-of-the-art large language models (LLMs), seemingly inapplicable to the small samples typical of macroeconomics, can be trained to learn the language of macroeconomy. We estimate a large-scale dynamic stochastic general equilibrium (DSGE) model on an initial segment of the data and obtain a posterior distribution over structural parameters. We sample from this posterior to generate millions of theory-consistent synthetic panels that, when mixed with actual macroeconomic data, form the training corpus for a time-series transformer with attention. The trained model is then used to forecast out-of-sample through 2025. The results show that this hybrid forecaster, which combines the theoretical coherence of DSGE models with the representational power of modern LLMs, successfully learns the macroeconomic language. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2512.21031 |
| By: | Jian Xue; Qian Zhang; Wu Zhu |
| Abstract: | We study how generative artificial intelligence (AI) transforms the work of financial analysts. Using the 2023 launch of FactSet's AI platform as a natural experiment, we find that adoption produces markedly richer and more comprehensive reports -- featuring 40% more distinct information sources, 34% broader topical coverage, and 25% greater use of advanced analytical methods -- while also improving timeliness. However, forecast errors rise by 59% as AI-assisted reports convey a more balanced mix of positive and negative information that is harder to synthesize, particularly for analysts facing heavier cognitive demands. Placebo tests using other data vendors confirm that these effects are unique to FactSet's AI integration. Overall, our findings reveal both the productivity gains and cognitive limits of generative AI in financial information production. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2512.19705 |
| By: | Gang Li; Dandan Qiao; Mingxuan Zheng |
| Abstract: | We find that event features extracted by large language models (LLMs) are effective for text-based stock return prediction. Using a pre-trained LLM to extract event features from news articles, we propose a novel deep learning model based on structured event representation (SER) and attention mechanisms to predict stock returns in the cross-section. Our SER-based model provides superior performance compared with other existing text-driven models to forecast stock returns out of sample and offers highly interpretable feature structures to examine the mechanisms underlying the stock return predictability. We further provide various implications based on SER and highlight the crucial benefit of structured model inputs in stock return predictability. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2512.19484 |
| By: | ZOHA, MAMUN UZ (University of Technology Sydney) |
| Abstract: | Artificial intelligence (AI) is rapidly transforming financial services and, in doing so, is exposing gaps between technological adoption and existing regulatory frameworks. This paper examines why the Australian Securities and Investments Commission (ASIC) has moved to provide guidance on AI use in financial services, and explores how different degrees of regulation—from none to very heavy—would shape market behaviour and systemic risk. Using Australia as the primary case, and Japan and the United States as benchmarks, the discussion considers AI both as an investment theme that challenges superannuation chief investment officers (CIOs) and as a potential investor or advisory agent in its own right. A qualitative, scenario-based analysis is used to reflect on implications for capital markets, volatility, portfolio composition and the risk of market failure. The paper argues for a balanced, principle-based regulatory approach that maintains innovation and efficiency while preserving market integrity and investor protection. Keywords: Artificial intelligence; financial regulation; ASIC; superannuation; systemic risk; market volatility; herding; governance; Australia; Japan; United States |
| Date: | 2025–12–21 |
| URL: | https://d.repec.org/n?u=RePEc:osf:lawarc:xer7g_v1 |
| By: | Iñaki Aldasoro; Sebastian Doerr; Daniel Rees |
| Abstract: | Investment related to artificial intelligence (AI) is surging – both in nominal amounts and as a share of GDP – and currently accounts for a substantial share of economic growth. The size of anticipated investment needs will require firms to shift the source of financing from operating cash flows to debt, with private credit playing a rapidly increasing role. While macroeconomic and financial stability risks from the AI boom appear moderate, the boom's sustainability hinges on AI firms meeting high earnings expectations. The fact that equity prices have run far ahead of debt market pricing underscores this tension. |
| Date: | 2026–01–07 |
| URL: | https://d.repec.org/n?u=RePEc:bis:bisblt:120 |
| By: | Christopher Regan; Ying Xie |
| Abstract: | We introduce obfuscation testing, a novel methodology for validating whether large language models detect structural market patterns through causal reasoning rather than temporal association. Testing three dealer hedging constraint patterns (gamma positioning, stock pinning, 0DTE hedging) on 242 trading days (95.6% coverage) of S&P 500 options data, we find LLMs achieve 71.5% detection rate using unbiased prompts that provide only raw gamma exposure values without regime labels or temporal context. The WHO-WHOM-WHAT causal framework forces models to identify the economic actors (dealers), affected parties (directional traders), and structural mechanisms (forced hedging) underlying observed market dynamics. Critically, detection accuracy (91.2%) remains stable even as economic profitability varies quarterly, demonstrating that models identify structural constraints rather than profitable patterns. When prompted with regime labels, detection increases to 100%, but the 71.5% unbiased rate validates genuine pattern recognition. Our findings suggest LLMs possess emergent capabilities for detecting complex financial mechanisms through pure structural reasoning, with implications for systematic strategy development, risk management, and our understanding of how transformer architectures process financial market dynamics. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2512.17923 |
| By: | Niclas Griesshaber; Jochen Streb |
| Abstract: | We leverage multimodal large language models (LLMs) to construct a dataset of 306, 070 German patents (1877-1918) from 9, 562 archival image scans using our LLM-based pipeline powered by Gemini-2.5-Pro and Gemini-2.5-Flash-Lite. Our benchmarking exercise provides tentative evidence that multimodal LLMs can create higher quality datasets than our research assistants, while also being more than 795 times faster and 205 times cheaper in constructing the patent dataset from our image corpus. About 20 to 50 patent entries are embedded on each page, arranged in a double-column format and printed in Gothic and Roman fonts. The font and layout complexity of our primary source material suggests to us that multimodal LLMs are a paradigm shift in how datasets are constructed in economic history. We open-source our benchmarking and patent datasets as well as our LLM-based data pipeline, which can be easily adapted to other image corpora using LLM-assisted coding tools, lowering the barriers for less technical researchers. Finally, we explain the economics of deploying LLMs for historical dataset construction and conclude by speculating on the potential implications for the field of economic history. |
| Date: | 2025–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2512.19675 |