AI and Wage Inequality in the United States, 2010-2024: Why Theory and Data Diverge

This paper investigates the impact of artificial intelligence on wage inequality and labor market equilibria within the United States between the years 2010 and 2024. Task-based economic models predict that artificial intelligence should displace routine middle-skill jobs while enhancing the efficiency of high-skill work, leading to an increase in wage polarization.

However, analysis of data from the Bureau of Labor Statistics (BLS) reveals an unexpected pattern. The 90/10 wage ratio decreased from 4.70 to 4.19 during the period of time that this study covers, even as the AI Adoption Index increased from 0.05 to 0.65. Decomposition by occupational skill group reveals distinct wage growth patterns. High-skill wages grew by approximately 45 percent, middle-skill wages by 48 percent. Low-skill wages rose most sharply at 61 percent, with the lowest-paid workers seeing the greatest gains.

This finding contradicts the hollowing out hypothesis. It particularly suggests that other structural forces such as minimum wage increases, post-COVID labor market tightness, and policy interventions, exceeded automation's polarizing effects during this period.

A lagged robustness check confirms that the negative association between artificial intelligence adoption and wage inequality still holds when AI adoption precedes wage gap movements by one year. The divergence between theoretical predictions and empirical results highlights a key point. Institutional factors mediate technology's labor market effects. This suggests task-based automation theories need refinement. They must account for countervailing policy and market forces.