Abstract

The labor market effects of artificial intelligence (AI) have been examined largely through changes in aggregate employment. Yet AI may alter workers' relative wages even where total employment remains unchanged. Because wage inequality depends on industry-level factors such as productivity, skill composition, and compensation structures, this study examines the relationship between AI exposure and wage inequality at the industry level. Using occupation-level AI exposure from Webb (2020), we aggregate the index to the industry level via a shift-share approach based on pre-adoption industry occupational composition, and estimate a long-difference model between 2013 and 2023.
The main findings are as follows. AI exposure has a positive and significant effect on the within-industry P90/P10 wage ratio, suggesting that within-industry inequality tends to widen. No significant effect is found for the P80/P20 ratio or for median-based ratios, implying that AI affects mainly the tails of the wage distribution. Industries with higher AI exposure also show faster average wage growth, indicating a rising wage premium and widening between-industry wage gaps. In terms of employment composition, total and high-wage employment show no significant change, whereas industries with higher AI exposure exhibit significantly slower growth in low-wage employment and a significantly lower within-industry share of low-wage workers. This decline would mechanically narrow the wage gap, yet the P90/P10 ratio still widens, indicating that inequality expands on net.
These findings show that AI diffusion may widen both within-industry inequality and between-industry wage gaps. This calls for industry-specific policies that reflect each industry’s employment and wage structure. In particular, the slower growth in low-wage employment points to the need for job transition support and AI literacy programs for low-wage workers, along with monitoring of low-wage job risks and wage gaps across industries.