Research Paper : Generative AI - Seniority Biased Tech change in Employment

 The sources analyze whether generative AI (GenAI) constitutes seniority-biased technological change (SBTC), meaning it disproportionately affects junior workers compared to senior workers. The findings provide early, broad-based evidence suggesting that the diffusion of GenAI since 2023 is indeed associated with these seniority-biased employment effects within firms.

This framework extends the classic literature on skill-biased technological change—which focuses on education or occupation groups—to the related but distinct dimension of seniority within firms.

Key Evidence of Seniority Bias

The analysis, which uses U.S. résumé and job posting data covering nearly 62 million workers, tracks within-firm employment dynamics by seniority over the period 2015–2025.

1. Divergence in Employment Trends: Beginning in 2023Q1, coinciding with the sharp increase in GenAI adoption, junior employment in adopting firms declined steeply relative to non-adopters. Conversely, senior employment in adopting firms continued to rise, showing no sign of a break in trend after 2022.
2. Triple-Difference Results: A triple-difference analysis, which controls for shocks specific to firms, time, and industry-by-seniority dynamics, confirms this widening gap. The coefficients remained flat until 2022Q4 but declined sharply starting in 2023Q1, reaching roughly a 10 percent drop in junior employment relative to senior employment after six quarters.
3. Concentration in High-Exposure Jobs: The decline in junior employment is not a broad phenomenon but is concentrated specifically in occupations with high exposure to GenAI (above the 75th percentile of exposure measures). Junior roles in low-exposure occupations showed no comparable decline.

Mechanisms Driving the Decline

The sources investigate the mechanism behind the reduction in junior headcount, leveraging linked employer-employee data to decompose workforce changes into inflows (hires), outflows (separations), and internal promotions.

• Slower Hiring: The sharp contraction in junior employment among GenAI adopters is driven primarily by a sharp slowdown in hiring after 2023Q1. Adopting firms hired an average of 5.0 fewer junior workers per quarter relative to non-adopters.
• Separations and Promotions: Interestingly, separation rates (exits) for juniors also fell relative to non-adopters, suggesting the decline is not the result of elevated attrition or layoffs. Promotions of juniors into more senior roles did not change significantly.

This pattern—driven by reduced hiring—is interpreted as firms making forward-looking adjustments, scaling back recruitment for roles they predict will be automated in the near future, as they may view hiring cuts as less costly than future layoffs.

Affected Tasks and Potential Implications

GenAI is likely substituting for tasks typically performed by those starting their careers. In many high-skill, white-collar jobs, junior workers begin by performing intellectually mundane tasks—routine yet cognitively demanding activities such as debugging code or reviewing legal documents—that are highly exposed to recent GenAI advances. If GenAI disproportionately substitutes for these entry-level tasks, the lower rungs of career ladders may be eroding.

The sources emphasize the potential long-term consequences of this seniority bias:

• Upward Mobility and Inequality: Since a large share of college graduates’ lifetime wage growth comes from within-firm advancement, this shift may have lasting consequences for upward mobility, income disparities, and the college wage premium.
• Career Ladders: GenAI adoption appears to shift work away from entry-level tasks, impacting how firms cultivate talent and how careers begin.

Heterogeneity in Effects

The decline in junior hiring among GenAI adopters was examined across different educational backgrounds, revealing a pronounced U-shaped pattern:

• Juniors from mid-tier institutions (Tier 3 and Tier 4) experienced the steepest relative declines in employment.
• Declines were smaller for juniors from elite (Tier 1), strong (Tier 2), and lowest-tier (Tier 5) schools.

Furthermore, the sources found that the decline in junior hiring is broad-based across industries and is not limited to or driven by specific dynamics in the information technology sector.

Caveats to Interpretation

The researchers acknowledge that GenAI adoption is not random; adopting firms are systematically larger, more technologically oriented, and concentrated in fields requiring highly educated labor. While the triple-difference design helps account for many observable and unobservable differences, the possibility of alternative explanations driven by confounding factors cannot be fully ruled out. Additionally, the study relies on a conservative measure of adoption (job postings for "GenAI integrator" roles), which may miss informal or "silent" adoption by employees.