How QinMetrics data is built

QinMetrics is organised around a single chain of public-sector activity:

1. Fiscal capacity. What a government can raise and borrow, measured by indicators observed for a jurisdiction and a period.
2. Spending. How that capacity is allocated, measured by expenditure indicators and the budget documents behind them.
3. Procurement. How spending becomes a market decision: the contracting processes that governments publish, with their buyers, methods, categories and outcomes.
4. Contracts. The agreements those processes produce, linking a buyer to a supplier with a value, a period and a type.
5. Payments. The money that actually moves under a contract, and when.
6. Performance. What happens to the contract over its life: milestones, delays, amendments, disputes, penalties, terminations and reviews.

Each stage is an entity type with its own records, and every record points to the stage before it. A payment belongs to a contract; a contract belongs to a procurement process and a buyer; a buyer belongs to a jurisdiction whose fiscal indicators are observed over time. The model is described in detail in Entity model.

Two data lines

Data is published in two parallel lines, Global and China. They are equals: neither is the default and neither contains the other. Each line has its own catalogue of datasets, its own sources and its own coverage notes. What they share is the entity model, the identifier rules and the provenance requirements, so that a contract from either line has the same fields and the same guarantees.

Entity pages are line-independent. A supplier that appears in both lines has one page, and that page shows which datasets it appears in.

Rules for every record

• Source-linked. A record is published only with a link to the document or publication it came from, together with the date it was retrieved. See Sources and provenance.
• Verbatim where possible. Values are stored as published. Amounts keep their original currency; names keep their original script alongside an English form; fields that do not fit the core model are preserved under attributes rather than discarded.
• Empty rather than guessed. A field the source does not state is left empty. Estimates and imputations are allowed only as separately flagged values, never in place of an observation.
• Stable addresses. Every entity has a permanent URL and a permanent slug. See Identifiers and URLs.
• Versioned releases. Datasets change through dated releases with record counts and checksums, never by silent edits. See Data quality, revisions and releases.

What is published today

The dataset catalogue lists every dataset with its status. A dataset is listed as planned or in development before any of its records exist, so that its scope and methodology can be reviewed in advance.