AI Adoption Parity Index & Exposure-Adjusted Distributional Analysis
Open analysis measuring state-level AI adoption against AI-exposed employment rather than population using the Anthropic Economic Index, BLS OES, and O*NET.
Project Overview
Conceived and engineered an open-source analysis measuring state-level AI adoption against AI-exposed employment rather than raw population. Utilizes the Anthropic Economic Index, BLS Occupational Employment and Wage Statistics (OES), and O*NET occupational exposure models, evaluating geographic concentration through Lorenz curves and Gini coefficients on both raw and exposure-adjusted distributions.
My Role
Lead Quantitative Analyst. Conceived the research question, engineered the exposure weighting model, calculated concentration metrics, built the interactive audit visualization, and open-sourced the analysis.
Tools / Stack
Project Details
Implementation Roadmap
Development Process.
Ingest state-level AI activity telemetry from the Anthropic Economic Index (releases 2025-09 through 2026-06).
Extract and harmonize employment distributions across detailed occupational categories from BLS OES.
Map O*NET occupational exposure scores to estimate state-level AI-exposed labor forces on fixed working-age population denominators.
Construct Lorenz curves and compute Gini inequality and Theil indices across 6 official US government geographical groupings.
Build live interactive Plotly audit interface allowing researchers to inspect geographic aggregation sensitivity and counterfactual shifts.
Publish open-access Jupyter notebooks, interactive tools, and documentation enabling reproducible public replication.
Case Study Analysis
Challenges & Outcomes.
Anthropic's AI Usage Index reports state Claude usage relative to population. However, raw population normalization and geographic aggregation boundaries create significant distortion: changing the geography (e.g. from states to BEA or Census regions) causes measured inequality (Theil/Gini) to fall dramatically, concealing substantial regional variation.
Built an interactive audit and exposure-weighted benchmark integrating the Anthropic Economic Index with BLS OEWS and O*NET occupational exposure models. Rendered dynamic Lorenz curves, Gini metrics, and Theil decompositions across multiple geographic tiers.
Delivered a public open-source analysis proving that geographic aggregation hides up to 60%+ of measured variation, providing external researchers and policymakers with an interactive tool and exposure-adjusted benchmark for understanding true regional AI workforce dynamics.