Despite growing concerns about technological unemployment, existing evidence remains fragmented, country-specific, and largely silent on whether productivity growth driven by artificial intelligence (AI) can decouple from human labour globally. This paper provides new global evidence on AI-labour substitution using an unbalanced panel of 65 countries over 2013–2023. Building on a task-based theoretical framework, we conceptualize AI as a potentially labour-replacing technology that reallocates economic tasks from humans to machines. We test this framework empirically using AI development, combined with cross-country data on labour force participation, productivity, human capital, and institutional characteristics. The results reveal three main findings. First, AI development is associated with a statistically significant decrease in labour force participation, indicating the presence of labour displacement effects. Second, AI simultaneously raises productivity, pointing to productivity-labour decoupling, whereby output growth becomes less dependent on human labour. Third, human capital and governance quality play crucial moderating roles in weakening labour-displacing effects of AI. Remarkably, dynamic panel estimates suggest that AI induces short-run labour displacement, while long-run outcomes depend on countries’ capacity to adapt through education and institutional quality.
Lê et al. (Sun,) studied this question.