The rapid adoption of artificial intelligence (AI) is clearly changing economies and how work is organised, but its long-term effects remain unclear as productivity benefits may be compromised by unintended costs such as labour market disruption and economic disparities. Some, including those covered by VoxEU (e.g. Albanesi et al. 2023a), have found that occupations that are more exposed to AI have expanded in European countries – a result replicated with firms (Babina et al. 2024) – whereas others have found that the rise of AI has led to some substitution of non-AI jobs (Acemoglu et al. 2022).
While the productivity effects of AI are often discussed in the context of national economy and firm-level outcomes, the implications of AI labour investments on local economies and municipal finance have been largely overlooked and the range of estimates, even among economists, varies greatly (Ilzetzki and Jain 2023). In a recent study (Andreadis et al. 2025), we exploit plausibly exogenous shocks to AI job postings and assess their local effects on municipal capital markets. Municipal bond yields serve as a forward-looking gauge of local economic and fiscal health, reflecting investor expectations about a municipality’s ability to generate revenue and repay debt. Since bond markets aggregate vast amounts of information, changes in yields provide a real-time assessment of how AI investments influence local government finances. Using a Bartik-like estimator that exploits pre-sample exposure, we find that increases in AI job postings are linked to declines in municipal bond yields, particularly for riskier and longer-term bonds, translating into higher tax revenues.
Artificial intelligence labour investments and local economic effects
While macroeconomic studies have explored the broad impact of AI on productivity and employment (Aghion et al. 2017, Farboodi and Veldkamp 2021, Eloundou et al. 2024), our research shifts the focus to the local economy level. AI-related job postings have surged in the past decade, rising from 0.5% of all job listings in 2014 to 2.05% in 2022 (Maslej et al. 2024).
However, the breadth of AI investments remains geographically diverse, with some areas growing much more rapidly than others (Andreadis et al. 2025). The intensity of AI job adoption is concentrated in the West (e.g. California) and Northeast (e.g. Massachusetts). Notably, there is substantial variation even within states, suggesting that different regions could experience diverse economic benefits from AI attributed to their inherent diversity.
Figure 1 Spatial heterogeneity in the share of and change in artificial intelligence jobs
Note: Panel A presents the proportion of artificial intelligence (AI) job postings by county averaged across 2014 and 2023. Panel B presents the percentage point change in the share of AI jobs between 2018-19 and 2022-23 averages within the same county. The redder the county, the higher the share and percentage point growth. Source: Lightcast.
Our study leverages county-level job posting data to track the localised diffusion of AI and its effects on economic conditions. A key challenge in identifying the causal impact of AI is that investments may be concentrated in regions already experiencing economic growth. To address this, we employ a Bartik-type shift-share instrument that captures pre-sample county-level exposure to national trends in AI-related occupations (Goldsmith-Pinkham et al. 2020). This isolates plausibly exogenous variation in national AI demand shocks on counties with a larger historical share of AI-exposed occupations.
Every additional AI job posting per establishment is associated with a 0.118 percentage point decline in bond yields. This amounts to an aggregate economic disparity of approximately 14 basis point reduction in bond yields between counties with greater versus lower AI adoption. We replicate that result using variation within counties with similar economic conditions based on per capita income and proximity (adjacent counties), ruling out alternative explanations based on the economic environment. These effects are particularly strong for riskier bonds, suggesting that AI adoption helps reduce perceived economic uncertainty.
As an additional robustness check, we exploit the launching of ChatGPT by OpenAI as a plausibly exogenous shock to investor attention about the potential value of AI technologies to examine implications within the secondary bond market. Using an event study approach, we compare yields within a narrow event window of two to three months before and after the introduction of ChatGPT in areas with high versus low exposure to cumulative AI labour investments. On average, we find a decline in bond yields after the introduction of ChatGPT, but the larger decline is observed for bonds of municipalities located in counties with higher cumulative AI labour investments. Economically, the introduction of ChatGPT led to a decrease in yields in the secondary market with a magnitude of 3.7 basis points (2.1 basis points) when exploiting cross-sectional (within) bond variation.
Despite the expectation that lower bond yields would incentivise increased municipal borrowing, we find no significant effect of AI investments on the total volume of bond issuance. Instead, municipalities adjust their financing strategy by shifting toward longer-term bonds. This suggests that local governments may view AI investments as a stabilising force, allowing them to lock in favourable borrowing terms for extended periods
Productivity and capital market implications
So, does this mean that investors are just responding to positive signals about AI investments, or are these counties becoming more productive? We find that counties with higher AI labour investments experience significant improvements in local economic conditions. Real productivity, measured as GDP per employee, rises by 1% over three years in AI-intensive counties. This growth is driven by GDP expansion rather than employment declines, suggesting that AI complements rather than displaces workers at the local level. Indeed, this contributes to the debate on whether AI will cause labour market disruptions or augment productivity (Acemoglu and Restrepo 2019, Brynjolfsson et al. 2021).
AI-intensive counties also experience positive spillovers into local public finance. Using an entropy balancing approach that non-parametrically matches counties ranking higher versus lower in AI intensity, we find that counties in the top third of AI labour investments see an 8.8% rise in total revenues in the first year, increasing to 17.2% after three years. This revenue growth is driven primarily by higher property taxes rather than sales taxes, indicating that AI-related economic expansion boosts real estate values and local tax bases.
Our findings quantify the impact of AI in shaping local economic and financial outcomes. AI investments not only contribute to local productivity gains but also improve municipal credit conditions, lowering borrowing costs for local governments and reinforcing public sector finances, increasing tax revenues. While concerns persist about the uneven distribution of AI’s positive impact, our results provide evidence that its benefits extend beyond major tech hubs, potentially supporting broader economic development. However, as AI adoption continues to grow, policymakers must consider how to ensure that its gains are widely shared and acknowledge that local governments should be offered the resources to manage its long-term implications. Furthermore, the types of AI investments vary, so we will need to be intentional about the composition of investments and transparent on their effects.
References
Acemoglu, D and P Restrepo (2019), “Automation and new tasks: How technology displaces and reinstates labor”, Journal of Economic Perspectives 33(2): 3–30.
Acemoglu, D, D Autor, J Hazell and P Restrepo (2022), “Artificial intelligence and jobs: Evidence from online vacancies”, Journal of Labor Economics 40(S1): S293–S340.
Albanesi, S, A D da Silva, J F Jimeno, A Lamo and A Wabitsch (2023a), “Artificial intelligence and jobs: Evidence from Europe”, VoxEU.org, 29 July
Albanesi, S, A D da Silva, J F Jimeno, A Lamo and A Wabitsch (2023b), “New Technologies and Jobs in Europe”, CEPR Discussion Paper No. 18220.
Andreadis, L, M Chatzikonstantinou, E Kalotychou, C Louca and C Makridis (2025), “Local Heterogeneity in Artificial Intelligence Jobs Over Time and Space”, AEA Papers & Proceedings, forthcoming.
Aghion, P, B F Jones and C I Jones (2017), “Artificial intelligence and economic growth”, NBER Working Paper 23928.
Babina, T, A Fedyk, A He and J Hodson (2024), “Artificial intelligence, firm growth, and product innovation”, Journal of Financial Economics 151, 103745.
Brynjolfsson, E, D Rock and C Syverson (2021), “The productivity J-curve: How intangibles complement general purpose technologies”, American Economic Journal: Macroeconomics 13(1): 333–72.
Farboodi, M and L Veldkamp (2021), “A model of the data economy”, NBER Working Paper 28427.
Goldsmith-Pinkham, P, I Sorkin and H Swift (2020), “Bartik instruments: What, when, why, and how”, American Economic Review 110(8): 2586–2624.
Eloundou, T, S Manning, P Mishkin and D Rock (2024), “GPTs are GPTs: Labor market impact potential of LLMs”, Science 384(6702): 1306–1308.
Ilzetzki, E and S Jain (2023), “The impact of artificial intelligence on growth and employment”, VoxEU.org, 20 June.
Maslej, N, L Fattorini, R Perrault, V Parli, A Reuel, E Brynjolfsson, J Etchemendy, K Ligett, T Lyons, J Manyika, J Niebles, Y Shoham, R Wald and J Clark (2024), “The AI index annual 2024 report”, Institute for Human-Centered AI, Stanford University, Stanford, CA.