A dataset of firm-level geopolitical risk perception for Chinese listed companies

This dataset provides firm-level measures of corporate geopolitical risk perception for Chinese A-share listed companies (2010–2024), constructed using NLP techniques applied to Management Discussion and Analysis (MD&A) sections of annual reports.
Methodology:
The index employs TF-IDF weighting based on a hierarchical "Four Pillars, Twelve Dimensions" lexicon containing 449 core terms, expanded via Word2Vec semantic modeling (vector size=300, similarity threshold=0.6). Text preprocessing includes jieba segmentation and stop-word filtering. Scores represent normalized keyword frequency per 100 words, winsorized at 1% and 99% levels.
Theoretical Framework:
Economic & Trade (Econ): Trade Barriers (tariffs, anti-dumping), Green Barriers (CBAM, carbon footprints), Investment & Financial (CFIUS, SWIFT sanctions)
Technology & Innovation (Tech): Sanction Blacklists (Entity List, EAR), Bio/Hard Tech Decoupling (CHIPS Act, biosecurity), Defensive Substitution (IT innovation, domestic substitution)
Supply Chain (Chain): Ethical Compliance (UFLPA, forced labor), Strategic Reconstruction (China Plus One, de-risking), Resource & Logistics (rare earths, energy security)
Macro-Political & Data (Macro_Data): Data Sovereignty (GDPR, cybersecurity), Ideological & Political (systemic rivalry), Regional Stability (geopolitical conflicts)
Data Content:
Primary file: GRPI_China_Firm_Level_2010_2024.csv (54,524 firm-year observations)
Key variables: Stock code (Stkcd), Year, Total_Words, GRPI_Total (aggregate index), four pillar scores (Econ, Tech, Chain, Macro_Data), twelve sub-dimension scores, industry classification (6 sectors), ownership type (Private/SOE)