人工智能与发达和新兴经济体的增长：短期影响（英）

Artificial intelligence andgrowth in advanced andemerging economies: byLeonardo Gambacorta,Enisse Kharroubi,AaronMehrotra and Tommaso Oliviero December 2025 JEL classification: E24, O47, O57 Keywords:generative artificial intelligence,emergingmarketeconomies,economic growth,productivity BISWorking Papers are written by members of the Monetary and EconomicDepartment of the Bank for International Settlements, and from time to time by othereconomists, and are published by the Bank. The papers are on subjects of topical This publication is available on the BIS website (www.bis.org). Artificial intelligence and growth in advanced and Leonardo Gambacorta, Enisse Kharroubi, Aaron Mehrotra and 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. Usingan empirical strategy inspired by Rajan and Zingales (1998) and a dataset covering 56economies and 16 industries, we find that the differential growth effects arise fromvariationsin sectoral exposure to cognitive and knowledge-intensive activities, JEL classification: E24; O47; O57. Keywords: generative artificial intelligence; emerging market economies; economicgrowth; productivity differentials; technological readiness, sectoral exposure to AI. Leonardo Gambacorta is with the Bank for International Settlements (BIS) and CEPR; Enisse Kharroubiand Aaron Mehrotra are with the BIS. Tommaso Oliviero is with the University of Naples Federico II,CSEF and Mofir. We thank Ralph De Haas for insightful comments and participants at the ASC –Institutions and Regional Development Conference, L’Aquila (2025). The views expressed are those of 1.Introduction Over the last couple of years, the global adoption of generative artificial intelligence(gen AI) by individuals and organisations has surged dramatically, sparking an intensedebate about its economic effects. The prevailing view is that this technologicalinnovation will enhance worker productivity and spur firm growth and innovation(Brynjolfsson et al., 2023; Noy and Zhang, 2023; Babina et al., 2024). As aggregate Although gen AI represents a general-purpose technology that is transformingcognitive capabilities, its impact is not expected to be uniform across individuals,occupations, and industries. First, gen AI mainly enhances cognitive and knowledge-intensive activities (e.g. professional services such as finance and IT, see Aldasoro et al.,2024b),while tasks with substantial physical components(e.g.construction)areexpected to be less affected. Second, AI adoption may not be practical or beneficial for Productivity gains from AI at the firm or sector level may also fail to translate intoequivalent macroeconomic gains. Broader economic factors—including frictions inadoption, potential disruptions to production networks, and regulation—mean that Moreover, the benefits of gen AI are unlikely to be distributed evenly acrosscountries. The aggregate growth impact of gen AI is likely to depend critically onnational production structures, particularly the relative size of sectors exposed to thenew technology. Advanced economies (AEs), which tend to have a greater share ofvalue added from early-adopting sectors such as finance, healthcare, and advancedmanufacturing,are expected to benefit more than emerging and developing In this paper, we examine whether the short-run growth effects of gen AI differbetween AEs and EMDEs. Our key hypothesis, which we test empirically, is that thegrowth impact of gen AI is driven by sectoral differences in exposure to the technologyand by country-level characteristics that influence its adoption and use. On the one productivity gains associated with gen AI. On the other hand, the heterogeneity in theimpact of AI on growth may reflect countries’ differing levels of readiness to adopt AI,from the quality of digital infrastructure to the regulatory environment. Our analysis Our empirical strategy closely follows the approach of Rajan and Zingales (1998).In their framework, the authors measure the extent of external finance dependence foreach sector (using the US economy as a benchmark) and examine whether sectors moredependent on external finance grow disproportionately faster in countries with moredeveloped financial markets. The parallel in our paper consists in using an industry- We begin by identifying the exposure of each industry to gen AI. Building on Feltenet al. (2021), we use an industry-level measure of AI exposure developed in Aldasoro etal. (2024), which is tailored to the US economy. As in Rajan and Zingales (1998), we treatthe US distribution of sectoral exposures as a benchmark for all countries. This measure To identify countries’ readiness to use AI, we use the AI preparedness index (AIPI)from the IMF (Cazzaniga et al. (2024)). The measure captures four key dimensionsrelevant for AI adoption: digital infrastructure, human ca