The impact of artificial intelligence on labor income share: From the perspective of technological progress bias

Artificial intelligence (AI), a general-purpose technology with the characteristics of a new type of infrastructure, can augment both capital and labor. Its impact on the labor income share (LIS) remains debated. From the perspective of biased technological progress, this paper investigates how AI affects the LIS through both theoretical and empirical analyses. First, using a factor-augmenting CES production function, we incorporate AI and biased technological progress into a unified framework for the determination of the LIS and clarify the sources of variation in labor’s share. Second, we use machine learning to build an AI dictionary, perform patent text analysis, and construct a city-level AI index. Based on panel data on 287 Chinese cities from 2008 to 2022, we then examine the effects of AI and biased technological progress on the LIS. The results show that AI development significantly reduces the LIS. The finding remains robust to alternative measures, different sample windows, alternative estimators, and IV (2SLS) strategies that address endogeneity. Mechanism and channel analyses suggest that AI promotes capital-biased technological change, thereby turn depresses the LIS. Industrial upgrading and ownership structure significantly raise the LIS and mitigate AI’s adverse impact. Heterogeneity analysis shows that the negative effect of AI on the LIS is more pronounced in labor-intensive, inland, small and medium-sized, and central and western cities. This study provides policy insights for developing economies such as China in mitigating the labor-market consequences of AI, narrowing income gaps, and advancing common prosperity.