Policy-semantics analysis translates the action verbs in policy documents into “intensity levels and implementation timelines.” In Chinese policy language, a single item carries a nine-level vocabulary ladder running from weak to strong (研究 → 积极研究 → 推进 → 积极稳妥推进 → 积极推进 → 加快推进 → 择机推动 → 推动 → 积极推动). The further right on the scale, the stronger the semantic signal and the closer the implementation. One must also understand the “medium-term certain, short-term vague” (Greenspan-style) linguistic feature — short-term language is deliberately vague to prevent disturbing market expectations. Big-data policy-semantics analysis (AI reading spokespersons’ facial expressions, a 20-year official-media language corpus) represents the technical frontier of this field.
The Framework As It Stands
This section is compiled from the research draft: the original framework’s structure, terminology, and key formulations are preserved, including editorial bridges and externally fact-checked supplements; diagrams are drawn by the compiler according to the original structure.
The Position of Policy Analysis in Macro Research
This framework holds that policy analysis must be introduced only after the growth-driver framework (exports, real estate, manufacturing) has been built and data verification completed; adding the policy dimension significantly raises the complexity of macro analysis. Macro quantification and policy analysis are two areas not yet adequately covered by mainstream macro frameworks. A complete macro team comprises seven types of specialization; policy-semantics analysis is an extension of empirical regularity-type thinking to the big-data scale.
Technical Frontier (Course Reference Point: 2019-01)
U.S.-side frontier: Certain hedge funds already describe themselves as “technology companies,” using AI to observe Federal Reserve spokespersons’ facial expressions to judge the true intent behind their statements.
China-side frontier: University research groups have already extracted policy language from major official media over the past 20 years, using big data to find patterns — the correspondence between specific words/word frequencies and policy shifts. This is beyond human capacity and represents the direction of the future.
Two Properties of Policy Language
Medium-term certainty: In standard policy language (following the Fed’s model), medium-term expressions are relatively definite — rate-hike plans for the next 3–4 years are communicated clearly to the market so as not to disturb market expectations.
Short-term vagueness: Short-term language is deliberately vague — the typical Greenspan style: after a speech, no one can tell whether he is optimistic or pessimistic, because overly explicit statements risk disturbing expectations and causing asset price bubbles or excessive declines. This framework holds that Chinese policy language fundamentally follows the same pattern.
The Nine-Level Intensity Ladder (Weak to Strong)
| Level | Wording | Meaning |
|---|---|---|
| ① | 研究 (study) | Not preparing to act for now; positioned beyond five years into the future |
| ② | 积极研究 (actively study) | Time window narrows to approximately three years |
| ③ | 推进 (advance) | Listed as a policy priority, but no timeline |
| ④ | 积极稳妥推进 (actively and prudently advance) | Recognized as the direction of development, but current constraints acknowledged |
| ⑤ | 积极推进 (actively advance) | Assessment is relatively smooth; preparation to bring to implementation |
| ⑥ | 加快推进 (accelerate advancement) | May be driven to implementation in the near future |
| ⑦ | 择机推动 (push forward at the right moment) | “Advance” (推进) upgrades to “push forward” (推动); semantic intensity rises sharply; essentially at the stage of imminent implementation |
| ⑧ | 推动 (push forward) | More certain than “push forward at the right moment” |
| ⑨ | 积极推动 (actively push forward) | Can be put into practice within the foreseeable policy window |
Timing caveat: The year-range mappings above (研究 ≈ five years out; 积极研究 ≈ approx. three years) are experience-based descriptions from the course at the time (2019-01) (course estimate), not a fixed schedule that holds for every policy; reading current policy wording requires checking the current official text and context for that wording, not mechanically applying these year ranges.
Applied habit: The Federal Reserve also follows a similar set of policy-language conventions. Doing macro research requires building the habit of watching the CCTV evening news and reading People’s Daily — policy will announce future direction in advance; the problem is that at the time it is often not recognized as such (“at the time it all seemed ordinary”).
Compiler’s Perspective
Coordinates: China and Great-Power Rivalry · Shu · What It Is
Bridge Layer
The exclusive increment this entry contributes is the invisible watershed between “推进” and “推动.” From ③ 推进 (priority work, no timeline) to ⑦ 择机推动 (imminent implementation), the textual difference is just a few characters, but the time gap can be years. The intensity jump across the nine-level ladder is uneven at the two most critical nodes: ③ → ⑦ is the shift from “registered” to “off the waiting list,” while ⑦ → ⑨ is from “imminent implementation” to “already in the operating room” — these two segments carry far more weight for asset pricing than the early ① → ③ study stage, and are also more prone to confusion.
The specific error of the old approach: treating any mention of an issue in a policy document as a signal, without identifying which rung of the nine-level ladder the action verb occupies — interpreting “积极研究” (approx. three-year window, 2019-01 experience description) as imminent implementation leads to systematically early positioning; treating “积极推动” as ordinary wording causes one to miss the highest-confidence implementation signal. An even more common error is judging intensity from the title’s “attaches great importance to X” rather than from the body’s action verb, ignoring the fact that nominal rhetoric and verb level can be completely at odds.
AI reading facial expressions and big-data extraction of 20 years of official-media policy language both point to the same cognitive gap: most macro researchers’ policy-semantic interpretation remains at the level of subjective qualitative judgment, while pattern recognition and frequency analysis can find regularities invisible to individual researchers across large samples. This framework, when put forward in 2019-01, positioned this as “the direction of the future” — after large language models became widespread, that judgment has become more timely rather than less.
The Nature of Macro Research and the Sense of Position’s meta-goal of “narrowing the range of choices” and this entry form a division of labor: the sense-of-position framework tells you the ultimate goal of macro analysis; the policy-semantics ladder tells you when and with what force the policy variable affects the answer to that goal. The order of combined use is: first use positional sense to judge macro coordinates, then use the semantics ladder to judge when policy momentum will be realized.
See Also
-
The Four Fiscal Accounts and Broad Fiscal Aggregates: Three-Step Infrastructure Estimation — Fiscal landing point of policy intensity signals: how the broad fiscal infrastructure measure follows up quantitatively after semantics upgrade to “accelerate advancement”
-
The Nature of Macro Research and the Sense of Position — Upstream of the meta-method: policy analysis must be introduced only after the growth-driver framework is built
-
Three Modes of Macro Research Thinking and the Primacy of Empirical Regularity — Methodological positioning: big-data policy-semantics is an extension of empirical-regularity thinking to machine scale
-
The Inflation Triangle Model: Two-Layer Forecasting and High-Frequency Tracking — Monetary policy semantics level directly affects the inflation-expectations term of the inflation triangle
-
Policy Pricing by Objective Effect: Not the Intent, Only the Actual Effect
-
Property Indicator Chain Tracking: The Three Sales Drivers and Investment Transmission
Sources
- Compiled research draft · collected 2026-07
Compiled draft z-0109 · collected 2026-07