Research Method
Method
Primary-source research, structured for comparison.
Primary sources, structured as comparable evidence.
WTS distinguishes between collecting public information and building an analytical dataset. Records are verified, normalized, classified and evidence-graded before they are used in analysis.
Topic Methodologies
China’s dual-use export controls
Data Methodology
Depending on the research question, WTS works across Chinese customs administrative enforcement records and no-penalty decisions; Ministry of Commerce and other central-government regulatory information; government procurement, tender and inspection-capacity records; judicial and prosecutorial records; corporate registration, ownership and credit information; and other relevant official public records. Trade statistics, tariff data and mirror statistics are being progressively integrated as the analytical base expands.
For China Customs administrative enforcement data, WTS currently relies on human-led discovery, retrieval, and direct review of source records. Earlier in the project, automated retrieval and AI-assisted collection were tested, but the structure, access constraints, and instability of China Customs source systems did not provide the reliability, reproducibility, or completeness required for a case-level research dataset. Manual verification is therefore a deliberate quality-control choice: WTS prioritizes case-level completeness, duplicate control, direct source review, source traceability, and evidence quality over collection speed.
China Customs administrative enforcement records are identified, retrieved, and verified through human-led research. This approach is designed to preserve the integrity of each case record, distinguish duplicates and related notices, confirm the underlying source where available, and maintain traceability from analytical claims back to evidence.
Original official materials are preferred where available. Corroborated public evidence and reconstructed records are separately identified.
Records are included only when the facts required for the relevant analytical use can be established to the required evidentiary threshold.
The analytical unit is defined at case level so that one matter is not counted multiple times merely because it generated several public documents.
Re-announcements, corrections, reposts and related notices are consolidated where they refer to the same underlying case or project.
Directly reviewed original records are distinguished from cases reconstructed after an original source becomes unavailable.
Recurring fields such as entities, authorities, products, destinations and legal grounds are standardized for comparison while source-level distinctions are preserved where analytically relevant.
Destination is treated as confirmed only when supported by the source evidence used for the dataset. Estimated or inferred destinations are separately labeled where used.
Decision, publication, declaration and violation dates are kept conceptually separate. Proxy dates are used only where expressly allowed by the methodology.
Unknown, unconfirmed and unavailable are not converted into “none.”
Analytical labels are applied according to documented research rules and separated from source-confirmed facts where interpretation is involved.
Research datasets are versioned so additions and corrections can be tracked, reviewed and reflected in the relevant analytical outputs.
Published records are not assumed to represent the full universe of enforcement, and raw case counts are not treated as enforcement rates without appropriate denominators and disclosure controls.
Evidence Quality & Reliability Guide
Government, customs, court or other official source directly reviewed and sufficient to confirm the material facts used in analysis.
The complete original document is unavailable, but key facts can be confirmed with high confidence through official search records, related official documents or preserved public material.
The original record is no longer publicly accessible and the case has been reconstructed from multiple preserved or official-source traces.
The existence of a case may be identifiable, but the evidence is insufficient for the relevant substantive or quantitative analysis.
Research principles
Source traceability ≠ source replication. Claims should be traceable to the evidence on which they rely. Traceability does not require publication of WTS’s internal source corpus or proprietary collection methods.
Unknown is not the same as none. Missing, inaccessible or unconfirmed information is treated as a data condition, not converted into a negative finding.
Published case counts are not enforcement rates. Observed records are interpreted within disclosure, denominator and comparability limits.
The strength of a claim should not exceed the strength of the evidence. Source quality, reconstruction status, missing data and analytical limitations shape how findings are stated.
Method is public; recipe is not. Analytical claims should be traceable to evidence, while proprietary research know-how remains protected. WTS does not publish specific search queries, search routes, discovery patterns, internal dictionaries, or detailed methods used to reconstruct missing or unavailable source records.