Eve, here. Having research prove that AI is working as intended, such as cutting workers’ pay to benefit corporations and tech giants, may seem like a bit of a dog bite. However, this finding is important from a policy perspective. Many of the promoters of AI claim that it will create even more jobs through some magical process that is not yet well explained. Therefore, it is important to debunk those claims.
Written by Antonio Miniti, associate professor at the Bologna University of Economics, Klaus Plettner, professor of economics at the Vienna University of Economics and Business, and Francesco Venturini, visiting researcher at the National Institute of Economic and Social Research. David Bloom, assistant professor at the Perugia School of Economics and the Clarence James Gamble Professor of Economics and Demography at Harvard University; Originally published on VoxEU
The rapid rise of artificial intelligence is raising fundamental concerns about the future of work. This column uses data from 238 regions in 21 European countries to explore how AI-related innovations will impact the distribution of income between labor and capital and between different skill classes of labor. Regions where AI patenting is more concentrated tend to have a lower labor share of income, especially in regions with strong industrial bases, indicating that AI functions as a capital-biased innovation. Without appropriate policy interventions, this trend is likely to exacerbate existing inequalities.
The rapid rise of artificial intelligence (AI) is reshaping production systems and raising fundamental concerns about the future of work (e.g. Korinek and Stiglitz 2019, Webb 2019, Prytkova et al. 2024). Will AI primarily support human labor, or will it make human labor unnecessary? Who will benefit from this transformation? Workers or capital owners? And among workers, who will fare better relative to those with higher or lower skills?
Although these questions dominate the public debate about AI, hard empirical evidence remains limited. A recent article (Minniti et al. 2025) provides new insights into how AI-related innovations will impact the distribution of income between labor and capital and between different skill classes of labor in the European region.
The results showed that labor’s share of income tends to decline in regions where AI patenting is more concentrated, particularly in regions with strong industrial bases. This pattern shows that AI functions as a capital-biased innovation, with returns from technological advances increasingly shifting towards capital.
If not addressed through targeted policies, this trend could exacerbate existing inequalities and pose long-term challenges to social cohesion in developed countries.
AI innovation, skills and changing income distribution
We investigate the relationship between AI-driven innovation and the labor share using data from 238 regions in 21 European countries over the period 2000–2017. 1 This regional perspective is essential because differences in industrial structure, educational background, and regional innovation dynamics strongly mediate the effects of technological change.
We measure AI innovation through a newly constructed dataset of AI-related patents, including technologies such as machine learning, natural language processing, and image processing methods. These data are linked to detailed regional indicators on employment, wages and value added, broken down by skill level.
The findings reveal a strong and statistically significant negative correlation between AI patenting intensity and labor income distribution. Doubling the strength of AI patents reduces the labor share by 0.5 to 1.6 percentage points. Figure 1 illustrates this development, showing the AI intensity across the European region (left panel) and the cumulative change in the labor share from 2000 to 2017 (right panel). Regions with high AI intensity tended to have a decreasing share of labor income over the entire period.
Figure 1
Note: The left side of the graph reports the region’s AI patenting intensity, defined as the technologically manifested comparative advantage calculated as the average number of AI patents compared to other patents filed in the region between 2000 and 2017. The resulting values are expressed as a ratio to the same measurements in other regions. If the index is greater than 1, the region is specialized in AI patenting. The right side shows the cumulative change in the labor share from 2000 to 2017 in percentage points.
Importantly, the decline in the labor share is not uniform across the workforce. We find that middle- and high-skilled workers have experienced the largest decline in their income share, primarily due to wage compression rather than changes in employment. 2 In contrast, the decline in the labor share of low-skilled workers is not as large. This is primarily due to modest employment growth in this sector, partially offsetting stagnant or declining wages. These patterns suggest that AI is reshaping labor markets not only between labor and capital but also within labor by redistributing profits across skill groups (see also Bloom et al. 2024 and 2025 for theoretical frameworks consistent with these results).
Importantly, we do not observe a proportional increase in labor productivity that offsets these distributional changes. In this context, AI not only increases efficiency but also redistributes the profits from innovation and expands the share of income accruing to capital.
From daily rotation to skill compression: Rethinking polarization
Although the decline in the labor share began long before the advent of AI, there is evidence that AI innovations are reinforcing and accelerating this trend. Historically, technological advances have often supplemented skilled labor and replaced mundane, low-skilled tasks. In contrast, modern AI is now able to replicate cognitive tasks, expanding the scope of automation to high- and medium-skill occupations.
This change is particularly evident in the service economy. AI applications are now being used in law, finance, logistics, government, and other white-collar fields previously considered resistant to automation. As a result, the current wave of innovation does not fit the classic narrative of displacement from low-skilled jobs followed by upskilling. Instead, we observe signs of wage compression at the top and middle of the skill distribution.
The net effect is a new form of polarization of the labor market, which does not necessarily benefit highly skilled workers, as in previous waves of technological innovation. Rather, the evidence shows that the redistribution of economic value is moving people away from work across the board, with particularly negative effects on those previously insulated from technological disruption. This even occurs against the background that AI increases average productivity and economic growth (e.g. Acemoglu 2025).
In the European context, there are growing concerns about wage stagnation, inequality and political discontent, and this evolving dynamic requires close attention.
Policy implications
The economic impact of AI adoption is not deterministic. Distributional outcomes will depend on how policymakers respond to the increased adoption and use of this technology. To reduce the negative impact on labor, investment in human capital is critical, particularly through the development of lifelong learning systems and vocational programs that provide workers with the skills to transition into roles that complement AI.
In parallel, tax systems need to adapt to changes in the balance between labor and capital. As capital increasingly captures profits from production, financial frameworks may be developed to ensure the sustainability of capital and returns. A potential policy could be to reduce labor taxes and move to (higher) taxes on pollution, which has negative externalities that would be efficient to reduce in any case. (higher) taxation on land. Supply is fixed so that taxation is not distortionary. and (higher) taxation of growing consumption under AI-driven economic growth (see Prettner and Bloom 2020, especially Chapter 7, for a detailed discussion of such policies). Furthermore, to avoid widening geographic disparities, economic policies could actively promote the spread of AI technologies across regions and sectors. Expanding the adoption of AI will ensure that productivity gains are no longer confined to a few innovation hubs, but contribute to inclusive and regionally balanced growth.
If AI is effectively governed, it has the potential to significantly improve overall economic well-being. However, without appropriate policy interventions, structural disparities between labor and capital and between core and periphery regions may be reinforced.
_______ According to the EU Nomenclature of Statistical Area Units (NUTS), the data refer to the NUTS-2 level (basic area). Skill types are defined using the ISCED classification system, which divides educational attainment into three categories: low (early childhood education, primary education, lower secondary education), intermediate (upper secondary education, post-secondary non-tertiary education), and high (short-term tertiary education, bachelor’s degree or equivalent tertiary level).
See original post for reference
