Replication Dataset: Generative AI and the Revaluation of Human Capital in Digital Creative Work — Evidence from Upwork

This repository contains the processed datasets, analysis scripts, and supplementary figures for the paper: Zhou, Z. (2026). Generative AI and the Revaluation of Human Capital in Digital Creative Work: Evidence from Upwork. Human Resource Management Journal (under review). The study examines how generative AI diffusion (2022–2024) reshapes the returns to different forms of human capital — skill proficiency, work experience, and reputational signals — in the online creative labour market. Data are drawn from 44,821 Upwork job postings (2022 and 2024) and 140 weekly observations from the Oxford Online Labour Index (OLI). Contents: processed_2022_creative.csv — Upwork 2022 creative job postings (N=222, cleaned) processed_2024_creative.csv — Upwork 2024 creative job postings (N=44,599, cleaned) processed_oli_creative.csv — OLI weekly demand index, Creative & Multimedia category (N=140 weeks) data_processing.py — Full data cleaning and variable construction pipeline oli_trend_analysis.py — OLI time-series and structural break analysis scheme_a_analysis.py — Continuous DID and heterogeneity regressions Figures — Framework diagram and OLI trend visualisations Key findings: AI tool proficiency commoditised within two years (no wage premium, p=0.894); senior workers with AI skills earn 8.9% more than senior non-users; creative strategy and originality roles command 27–51% premiums; high-complexity tasks declined 28.5% vs. 42.1% for low-complexity tasks; OLI demand rose +24.42% alongside a ~30% hourly rate decline.