{"slug": "apollo-research-links-ai-exposure-to-slower-wage-growth", "title": "Apollo Research Links AI Exposure to Slower Wage Growth", "summary": "Apollo Research published a July 30 white paper estimating that real-wage growth in high-AI-exposure occupations was 6.7 percentage points slower after 2023 than in lower-exposure occupations, with no statistically significant employment effect. The difference-in-differences analysis by analysts Sania Edlich and Torsten Sløk matched 321 US occupations and measured exposure using observed Claude use from Anthropic's Economic Index. The estimated wage-growth difference was larger for service occupations and the lowest-paid quartile, while top earners showed no significant effect.", "body_md": "# Apollo Research Links AI Exposure to Slower Wage Growth\n\nApollo published a July 30 white paper estimating that real-wage growth in high-AI-exposure occupations was 6.7 percentage points slower after 2023 than in lower-exposure occupations, while the study detected no statistically significant employment effect. The difference-in-differences analysis matched 321 occupations and measured exposure using observed Claude use from Anthropic's Economic Index.\n\nApollo published a July 30 white paper estimating that real-wage growth in occupations with high AI exposure was **6.7 percentage points slower after 2023** than in lower-exposure occupations. The study did not find a statistically significant employment effect.\n\nApollo analysts Sania Edlich and Torsten Sløk used a difference-in-differences design with occupation and year fixed effects across **321 matched US occupations**. Their panel covers 2015 through 2025 and combines Bureau of Labor Statistics wage and employment data with Anthropic's Economic Index.\n\nThe paper defines a high-exposure occupation as one with an Anthropic index score of at least 0.5. The index is based on the share of occupational tasks observed being performed with Anthropic's AI tools, weighted by task importance.\n\n### The estimated effect was concentrated among lower earners\n\nThe paper estimates that workers in the bottom wage quartile experienced about **10.7% slower real-wage growth** relative to low-exposure occupations. It reports a roughly **24.3% relative decline in wage growth** for service occupations and no statistically significant wage effect among the highest-paid workers.\n\nThese are relative growth estimates, not a finding that every exposed worker received a 10.7% or 24.3% pay cut. The distinction matters because the model compares changes over time between occupation groups after accounting for occupation and year effects.\n\nThe employment coefficient was not statistically significant in the main model. That result supports the paper's narrower claim that the early measured effect appears in wages rather than headcount; it does not prove that AI has no employment effect in every company, occupation, or future period.\n\n### Exposure is a proxy, not an employer deployment record\n\nUsing observed Claude activity gives the study a behavioral input that theoretical task-exposure scores lack. It also creates a limitation: the index measures activity on one provider's tools and assigns exposure at the occupation level. It does not show whether a particular employer deployed AI across an entire role or whether workers used other providers.\n\nThe authors also note challenges in matching occupational classifications and data collection over time. Like any difference-in-differences analysis, the result depends on whether the comparison groups provide a credible counterfactual after accounting for fixed effects and trends.\n\nFor data and ML practitioners, the study broadens the labor-market question beyond job counts. Replication with other model-use datasets, employer-level adoption measures, and longer post-adoption periods would help test whether the reported wage pattern persists and how much of it can be attributed to AI rather than other changes in the labor market.\n\n## Key Points\n\n- 1Apollo's 321-occupation analysis estimates 6.7 percentage points slower real-wage growth in high-AI-exposure occupations after 2023, with no statistically significant employment effect.\n- 2The estimated wage-growth difference was larger for service occupations and the lowest-paid quartile, while top earners showed no significant effect.\n- 3Anthropic usage data provides a behavioral exposure measure, but it remains an occupation-level proxy from one provider rather than a record of employer deployment.\n\n## Scoring Rationale\n\nThe white paper presents a notable empirical estimate of AI exposure and wage growth using occupational data and a Claude usage-based index. Its quasi-experimental design is relevant to practitioners and policymakers, but the exposure measure is occupation-level, limited to one provider's observed use, and drawn from a single early analysis.\n\n## Sources\n\nPrimary source and supporting public references used for this report.\n\nPractice interview problems based on real data\n\n1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.\n\n[Try 250 free problems](/problems)", "url": "https://wpnews.pro/news/apollo-research-links-ai-exposure-to-slower-wage-growth", "canonical_source": "https://letsdatascience.com/news/apollo-research-links-ai-exposure-to-slower-wages-65991142", "published_at": "2026-07-30 16:36:09+00:00", "updated_at": "2026-07-30 18:28:23.098948+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "ai-research"], "entities": ["Apollo Research", "Anthropic", "Sania Edlich", "Torsten Sløk", "Bureau of Labor Statistics", "Claude"], "alternates": {"html": "https://wpnews.pro/news/apollo-research-links-ai-exposure-to-slower-wage-growth", "markdown": "https://wpnews.pro/news/apollo-research-links-ai-exposure-to-slower-wage-growth.md", "text": "https://wpnews.pro/news/apollo-research-links-ai-exposure-to-slower-wage-growth.txt", "jsonld": "https://wpnews.pro/news/apollo-research-links-ai-exposure-to-slower-wage-growth.jsonld"}}