Key facts
- AI adoption is linked to a 6.7% average decline in real wage growth for highly exposed jobs since 2023.
- The impact of AI on overall employment levels was not statistically detectable.
- Lower-earning workers and service sector employees experienced the most significant reductions in earnings growth due to AI.
- High-income workers showed no significant wage impact from AI adoption.
- Approximately 5.8 million workers are in roles highly exposed to AI, with this number projected to increase.
Artificial intelligence poses a greater threat to the job market through wage suppression rather than outright job elimination, according to new research from Apollo Global Management. The study analyzed wage and employment data for 321 occupations in the U.S., finding that jobs with the highest exposure to AI experienced an average 6.7% decrease in real wage growth after 2023, coinciding with the viral spread of tools like ChatGPT.
The report, authored by analyst Sania Edlich and chief economist Torsten Sløk, concluded that AI's impact on overall employment levels was not statistically detectable. Instead, the primary effect observed was a dampening of wage growth, particularly for lower earners. Service workers saw their earnings growth decline by an average of 24.3%, and those in the bottom quartile of earners experienced a 10.7% wage reduction over the period.
Conversely, the paper noted no significant wage effects among the highest-paid workers. The research drew upon data from the Bureau of Labor Statistics and utilized Anthropic's Economic Index to gauge job exposure to AI based on the proportion of tasks performable by AI tools. Occupations like computer programmers and statistical assistants, identified as highly exposed, showed notable drops in real wages between 2022 and 2024.
While some wage declines were attributed to broader industry factors, the study highlighted that approximately 5.8 million workers are currently in roles highly exposed to AI. Edlich and Sløk anticipate this figure will grow substantially, potentially exacerbating income inequality and influencing future labor market policies. The findings align with other analyses suggesting that automation of a significant portion of intelligence tasks could lead to wage stagnation.