DeepSeek AI–Fintech Regulatory Shock Panel Dataset: Comparator-Adjusted Firm-Quarter Data, 2019–2024

This dataset supports the study “Regulatory Shocks and Cost Efficiency in China’s AI–Fintech Sector: Quasi-Experimental Evidence from DeepSeek and Comparator Firms.” It provides a comparator-adjusted firm-quarter panel for 2019–2024 constructed to examine how China’s 2021 data-governance shock, centered on the Data Security Law (DSL) and Personal Information Protection Law (PIPL), affected compliance/legal cost share, infrastructure cost share, and modeled gross-margin resilience in an AI–fintech setting.
The focal firm is DeepSeek, treated as a high-exposure private AI firm operating in a finance-related, data-sensitive domain. Public comparator firms include SenseTime, Baidu AI, and iFlyTek, with segment extraction and scope-adjustment procedures used to improve comparability across firms that differ in scale, diversification, and business-model breadth. The dataset is designed for firm-quarter analysis and for replication of the event-study, difference-in-differences, and robustness procedures reported in the associated manuscript.
The package includes: (1) the main firm-quarter panel; (2) a variable dictionary/codebook; (3) source-mapping and provenance materials; (4) comparator-scaling and haircut assumptions; (5) validation sheets; and (6) simulation-related inputs used for the paper’s Monte Carlo risk-translation extension. Key variables include compliance/legal cost share (ℓ), infrastructure cost share (cinfra), modeled gross-margin proxy (GM*), pre-shock exposure intensity, software-process intensity (SPI), customer churn proxies, competitor pricing-pressure proxies, and related benchmarking fields.
All data are derived from public or public-facing materials, including company filings, policy documents, industry and analyst sources, public operating signals, and documented transformations. No proprietary internal company records are included. DeepSeek-related values are inferred through transparent public-data proxy construction rather than direct access to internal accounts.
This deposit is intended to improve transparency and reproducibility in a context where private AI firms disclose limited internal financial information. It should be interpreted as a structured research dataset for empirical replication, robustness analysis, and comparative benchmarking rather than as a source of official company accounts.