Intro — Waitan Ludi

01 · THE CORPUS

Every level of government, in one searchable record

Most China monitoring stops at Beijing. Most of the money and the local preference sits lower down, in provincial, city and county documents. Each document here carries its 文号, issuer, date, source link and attachments. Copies are folded into one record.

Search in Chinese or English: charging subsidy finds 充电补贴, CATL finds 宁德时代.
Filter by issuer, province, date, topic tree or measure, and download the result.
See exactly what we hold of each province, and what we cannot measure.

riskcomplianceinvestorsresearch

Documents →Provinces Coverage register

50,270 documents, by the level of the issuing government

Central3,734

Provincial29,218

City15,607

County1,697

Documents held, not documents issued: coverage differs by province, and the register says by how much.

02 · CENTRAL → LOCAL TRANSMISSION

Watch a State Council directive travel down to the provinces

Chinese documents cite the documents they implement by number. Parsing those citations gives a graph from each central directive to every provincial and city document that restates it, with the time it took.

The directive tracer: one directive, every province's response, on a clock.
Which central directives a province has not yet been seen to restate.
Typical restatement windows, used to forecast when the 15th-plan wave lands.

riskcomplianceresearch

Directive tracer →Promulgation graph

151days median time from a State Council directive to a province's implementing document. The shaded band is the middle half of directives.

03 months6 months9 months1 year

Middle half84–259 days

Answered in a year59% of directives (653 of 1,112)

Measured on the provinces whose documents are listed in full. A missing restatement is "not observed", never non-compliance.

03 · CLAUSE-LEVEL EVIDENCE

The money and the favouritism, quoted from the text

A screen reads each document for support measures (grants, cheap loans, tax relief, land, guidance funds) and for favouring provisions: clauses that tilt a programme towards local firms, local products or domestic suppliers. Every hit carries its quote and a link to the clause.

Whether a measure is operative or only a directive to lower levels, and when it ends.
Who a favouring clause disadvantages: firms from other provinces, foreign firms, imports.
Checked precision printed next to every measure class.

riskcomplianceresearch

Monitor →Insights Who may apply

14,963 support measures found and quoted, and 3,786 favouring provisions. The six most common kinds:

Capital grant6,774

Tax relief2,337

Interest subsidy1,195

Credit guarantee705

Guidance fund / state equity559

Purchase subsidy494

Screen output: leads with a measured precision, not findings. A document can carry several measures.

04 · FIRMS

Who is named, and who gets paid

Policies publish lists: subsidy settlements with amounts, tax-relief catalogues, demonstration and award lists, type approvals. Each firm on a list gets a page that follows it across documents. For Hong Kong-listed groups, the government grants they report in their own annual reports sit beside those lists.

Every list a firm is on, with the amount where one is printed.
Ownership (state, private, foreign, joint venture) from a curated lexicon.
The foreign-subsidy screen: a group's Chinese entities on funding lists in the three years before an EU FSR notification.

complianceinvestors

BYD →Firms FSR screen

12.6bn RMB government grants BYD reported for 2025 in its own annual report (19% on the year before). Largest reported grants among listed groups held:

BYD 比亚迪12.6bn

Great Wall Motor 长城汽车2.2bn

Geely Auto 吉利汽车1.1bn

GAC Group 广汽集团0.8bn

Zoomlion 中联重科0.6bn

From HKEX annual reports, machine-extracted with the page cited; 2025 figures. Grants are not all industrial policy.

05 · SECTOR RADAR

Where Chinese support, EU dependence and export push coincide

16 strategic sectors on one scorecard: policy momentum, provincial spread, money and local preference from the documents; export growth from UN Comtrade; China's share of EU imports from Eurostat. You can change the weights, and every rank shows how far it moves under random weights.

A brief per sector: EU price and volume, where in China it is pushed, the money quotes, the dates ahead.
The risk brief: one printable page for a desk officer.
The sectors pushing out after their support peaked, and the next wave.

riskinvestorsresearch

Sector radar →Risk brief Batteries brief

6of 16 sectors in which China supplies more than half of what the EU imports from outside the EU (Eurostat Comext, by value):

Solar PV98%

Batteries and storage90%

Critical minerals83%

Drones and low-altitude76%

EV charging59%

AI and computing51%

Bar length is China's share of extra-EU imports of the sector's HS headings, latest year.

06 · THE NEXT FIVE YEARS

What the new Five-Year Plan outlines name, before any trade figure shows it

Plan watch compares the provincial and city outlines of the 15th Five-Year Plan (2026–30) with those of the 14th: which industries they name, and how densely. It also forecasts when provinces will restate the new central sector plans, and follows that wave in a feed.

New since the 14th plan: low-altitude economy (14 of 15), embodied AI and humanoid robots (11 of 15), brain-computer interfaces (8 of 15).
A quote from the outline that names each term most.
An exploratory check: the 14th-plan outlines led local policy, not exports.

riskinvestorsresearch

Plan watch →Future industries Atom feed

28× as often per 100,000 characters: how much more densely the 15 outlines of the 15th plan held name drones and low-altitude than the 7 of the 14th did. The sectors rising, each one where the EU already depends on China:

Drones and low-altitude27.7×

AI and computing5.0×

Batteries and storage2.6×

Solar PV1.7×

Log scale. A mention is not a commitment; the outlines held are not a sample of China.

07 · YOUR EXPOSURE

What in all this applies to your operation, as a memo and a calendar

Describe an operation (sectors, provinces, ownership, home jurisdiction, HS codes). The exposure check returns the clauses that name foreign firms, programmes open only to local firms, support in force, Chinese export-control and countermeasure acts touching your goods, and the next twelve months' dates. The subsidy file sorts a sector's support under the headings of an EU anti-subsidy investigation.

Download a Markdown memo and an .ics calendar, or subscribe to the feed.
A CSV annex per sector for a trade-defence file.

compliancerisk

Example: US chip-tool supplier in Shanghai →NEV subsidy file

140of 376 NEV documents since 2020 that carry a support measure under an EU anti-subsidy heading; 6 of 6 headings documented, 6 matched in China's WTO notifications, 2 cited by the EU's BEV act.

Preferential financing4

Grants92

Equity and guidance funds1

Revenue foregone41

Land, power and inputs below adequate remuneration5

Demand-side subsidies28

A starting point for a file, not a finding. Not matched in the notifications held never means not notified.

08 · ASK THE EVIDENCE, OR PLUG IT IN

Ask in plain English; get answers that cite the documents

The research assistant answers from the corpus and shows its sources, so every claim can be checked against the clause. The same checks run as a JSON API and as tools for Claude or any assistant that speaks MCP.

Questions such as "Which provinces offer land discounts for battery plants?"
/api/v1/: radar, exposure, plan watch, subsidy file, FSR screen.
MCP tools: sector_radar, exposure_check, plan_watch, subsidy_file, foreign_subsidy_screen.

riskcomplianceinvestorsresearch

Ask the evidence →Exposure as JSON

$ curl '/api/v1/exposure/?sector=semis&province=31&ownership=foreign&home=US'
{ "profile": {…}, "documents_in_scope": …, "favouring": […],
  "eligibility": […], "province_restrictiveness": {…},
  "support_in_force": […], "coming_into_force": […], "expiring": […],
  "export_controls": […], "eu_cases": […], "central_directives": […],
  "firms": […], "note": "…" }

claude> Which battery programmes in Jiangsu are open only to local firms?
→ calls exposure_check(sectors=["storage"], provinces=["32"])
→ answers with each clause, its 文号 and a link to the document

The assistant searches the corpus and cites what it finds; it does not answer from memory. Each tool returns the same figures as the page.

09 · HOW SURE WE ARE

Every figure says how sure it is, and the nulls are shown as nulls

The tests were specified before they were run, and two came back null; both are on the page with the design that would answer them. Screens print their checked precision. A coverage register shows what we hold of each province. Every table downloads as CSV, and the method report is generated from code.

Co-occurrence, never causation. Documents held, not documents issued.
"Not observed" is never "non-compliance"; "proposed" is never "paid".

researchrisk

Robustness and tests →Data Method report

ρ -0.24Pre-specified test: policy pressure at end-2021 against export growth 2021–25 across 14 sectors (p = 0.39). Null, reported as null.
p 0.99Second test, from the EU side: China's share of EU imports rose no more in the headings its policy targets (0.0 points). The exposure is broad.
135/160Sampled sector-set titles read against the sector, on topic (round 1); 52/60 after the fixes. AI-checked, not an independent coder.
56/60Local-registration clauses, hand-checked.
500Random weightings behind every rank's 10th–90th percentile range.

Ask the evidence.

China's industrial policy, read document by document, from the State Council down to the county.