China Government Work Reports
A structured read of China’s annual Government Work Reports (2006–2026) — the premier’s once-a-year speech to the National People’s Congress that sets the year’s economic targets and surfaces the leadership’s policy vocabulary. Each headline target is extracted with its modifier (around, at least, range) and paired with the realised series from China’s National Bureau of Statistics; tracked phrases show how the rhetoric evolves year-on-year. Every value links back to the paragraph it came from.
Target vs. realised
GWR-stated targets paired with realised data from China’s National Bureau of StatisticsMethodology
how the signals on this page were producedCorpus
English translations of the annual Government Work Report delivered to the National People’s Congress, 2006 onward. Each report is paragraph-chunked so every extracted value can be traced to a specific source paragraph (see the coverage manifest above for per-year provenance). Different years use different translators (State Council / Xinhua / NPC); raw phrase counts are therefore over English variants, not over the underlying Chinese terms.
Quantitative targets
Each headline target (GDP growth, CPI, fiscal deficit % of GDP, M2, urban surveyed unemployment, new urban jobs) is extracted with an explicit modifier — around, at least, at most, range, or point — and an is_explicitly_absent flag for years where the target was deliberately not stated (2020 GDP, COVID). That distinguishes “not stated” from “extraction failure”.
Tracked phrases
An n-gram analysis is run over the corpus to identify and extract key phrases. An n-gram is a contiguous run of n words — “reform” is a 1-gram, “common prosperity” is a 2-gram, “high-quality development” is a 3-gram. We extract 1- to 5-gram counts per year across the corpus and flag phrases by their year-over-year frequency profile as emerging (rising in recent years), persistent (consistent across the window), or fade-out (declining). In addition, a small a_priori list of policy slogans (e.g. “dual circulation”) was added by hand. For each phrase, a Chinese concept anchor is identified and a list of their English variants observed across documents are mapped; counts are computed over those English variants. The Chinese term is manually selected and not an extraction from the actual text.
Caveats
- These are stated targets and political rhetoric — not realised macro data. For realised China GDP, CPI, unemployment, etc., use the underlying National Bureau of Statistics data directly.
- Phrase frequency is computed over English translations. Different years use different translators, so the same Chinese term may be rendered several different ways; the tracked-phrase list keys phrases by Chinese term and sums English variants.
- The Chinese term shown in each phrase’s detail panel is a hand-selected concept anchor, not extracted from the translated source.
- Year coverage and translator metadata are surfaced in the coverage manifest above so gaps are visible.