Mechanism vs Narrative: An Analytical Method

The Triggering Case

Chinese wind. Headline frame: "China builds wind turbines for export but burns coal at home — performative greenwashing." Aggregates flip the story.

~49% of global cumulative wind capacity (640.5 GW of 1,299 GW, 2025); ~70% of new installations in 2024–25 (GWEC 2025, WWEA H1 2025).
Offshore: 4.4 GW added 2024 (~55% global); <5 GW (2018) → 42.7 GW (Mar 2025), 41% CAGR; ~50% of global offshore pipeline (Global Energy Monitor).
Solar manufacturing: 83% polysilicon, 97% wafers, ~85% modules; H1 2025 cell exports +73% YoY (IEA, Wood Mackenzie, Ember).
Generation mix: coal 59% (H1 2024) → 56% (H1 2025); solar 8.4% → 11.5%; renewables ~42% (Ember, Carbon Brief).

Mechanism: industrial-input cost minimization to defend factory-of-the-world status. Cheap power, owned supply chain, captive demand. Climate is second-order. The frame gets the causal arrow backwards.

The Method

What supply-chain analysts (Eric Cracks China, Brad Setser, Patrick McKenzie) do that pundits don't:

Aggregate flows, not anecdotes. Customs data, capacity-factor reports, utility filings. Setser: China's $1.2T 2025 goods surplus kills "China is collapsing" headlines — the goods are leaving.
What is the actor optimizing for? Not the slogan. The budget constraint, throughput target, input cost driven to zero.
Separate signaling from operations. The CCP signals climate leadership and coal sovereignty. Neither tells you what's being built.
Capacity has gravity. Built factories want to run. Built grids want to fill.

Intellectual Lineage

Adam Tooze (Chartbook) — material/financial flows over ideology. Wages of Destruction: Nazi Germany as resource-constrained. Crashed: 2008 as transatlantic dollar-funding mechanism.
Brad Setser (CFR Follow the Money) — balance-of-payments forensics. China's "true" surplus ≈ $1T.
Patrick McKenzie (Bits about Money) — institutions seen through their data models. "Seeing Like a Bank."
Helen Thompson (Disorder, 2022) — energy/money/geopolitics as substrate; democracy as the thing on top.
Noah Smith (Noahpinion) — industrial policy via Studwell's How Asia Works.
Kyla Scanlon — inverse case: "vibecession" = narrative is the mechanism when self-fulfilling. Caveat against pure materialism.
Wolfgang Streeck — democratic capitalism as a structural bargain in default.

Common posture: data first; assume actors are optimizing something specific; refuse the dignity of "intent" when "incentive" will do.

Seven 2025–26 Cases Where Frame Diverges from Mechanism

1. CHIPS Act — "America is reshoring chips." $7.86B Intel + $6.6B TSMC + $4.75B Samsung finalized; TSMC Fab 1 producing 4nm at yields 4 points above Taiwan; Fab 2 installing 3nm/2nm; Intel 18A in HVM at Fab 52 (Futurum, IEEE Spectrum). But Intel Ohio slipped to 2030; Samsung Taylor 4nm cancelled. Mechanism: leading-edge enclave, not "manufacturing renaissance."

2. IRA — "Massive clean-energy buildout." $115B of actual manufacturing investment Q3 2022–Q1 2025; quarterly rate $2.5B → $14B. But 68% batteries, 14% EVs, 13% solar — battery industrial policy in climate clothing. Q1 2025: $6.9B cancellations, a record (Clean Investment Monitor).

3. Russia sanctions — "Crushing Russia." Shadow fleet 900–1,400 vessels (~600 sanctioned); 250k artillery shells/month (~3M/yr → 4M); 250 new T-90M tanks in 2025 → 428/yr by 2028; defense = 4.5M jobs (~20% of mfg employment) (RFE/RL, Frontelligence, KSE). Revenue is squeezed via discount widening, not volume.

4. India / Apple — "India is the new China." 55M iPhones in India in 2025 (+53% YoY); ~26% of global iPhone shipments; Tata 15k → 75k workers in two years (TechWire Asia). Real but narrow: iPhone assembly, not deep component supply chain. Apple-shaped, not India-shaped.

5. AI datacenter buildout — "Grid melting." IEA: datacenter electricity +17% in 2025 vs 3% overall — real. But utilities flag the same projects shopped to multiple utilities; forecasts span 449 → 980 TWh by 2030. Mechanism: hyperscalers double-booking interconnect queues as a real option. See ai-infrastructure-endgame-indicators.

6. Climate rhetoric vs. fossil-fuel capex. IEA 2025: $3.3T total, $2.2T clean vs $1.1T fossil — clean is 2x fossil; electricity capex now 50% above oil/gas/coal combined (IEA WEI 2025). Spend has flipped; "we're failing on climate" is lagging the books.

7. US–China decoupling — "Decoupling is happening." US imports from China: 13% → 7% in 2025; -45% YoY in 12 months to Nov 2025. But the aggregate US deficit didn't shrink — Mexico/Vietnam/Taiwan absorbed it (Bruegel, SCMP). Mechanism: rerouting, not decoupling. China's global surplus = $1.2T.

Bonus — Germany. Production -3.9% YoY Aug 2025; auto -18.5% MoM August; 248k industrial jobs lost through Dec 2025 (Destatis). But Nordic green-industry capex intact. "European decline" is German-auto, not pan-European.

Failure Modes of Mechanism Analysis

Over-determinism. Assumes actors are rational-profit-maximizers. They're not always. Boards do empire-building, regimes do prestige projects, founders do revenge.
Politics inside the mechanism. Xi can override the supply chain (zero-COVID did). Modi can override the labor market. Trump can tariff against the carry trade. The mechanism is conditional on the political ceiling not falling on it.
Black-swan principals. Putin (2022 invasion against every macro incentive), Trump (April 2 "Liberation Day" tariffs), Erdoğan (2021 unorthodox monetary policy) — single decision-makers who don't behave as the aggregate would predict.
When narrative beats mechanism: election outcomes (Brexit, 2016, 2024), cultural shifts (post-#MeToo HR cost), religious movements, bank runs (where narrative is the mechanism), wartime morale. Anything where coordination is the binding constraint.

The honest version is mechanism is the prior, narrative is the residual. When the data is dense, lead with mechanism. When the system is undetermined or coordination-dominant, narrative gets a vote.

Other Recent Examples of Political Claim Collapsed by Aggregate Data

"China is collapsing." Setser's $1.2T 2025 surplus, 6M+ car exports, lapping global trade growth. Bears were arguing demographics and Evergrande; mechanism shows export production capacity expanding.
"Sanctions destroyed Russia's military." NATO's own intel: 3M shells/year, 250 new tanks/year, defense = 20% of manufacturing employment.
"IRA isn't building anything real." $115B already invested, ~190 facilities operational by Q1 2025. Real, but skewed to batteries — both the political left and right narratives are partially wrong.
"Mexico nearshoring is a boom." Record $41B FDI through Q3 2025, but Dallas Fed points out it's only marginally above pre-pandemic trend. Real but smaller than headlines.

How To Run It

When you encounter a confident political claim:

Identify the headline frame (one sentence).
Ask: what aggregate would have to be true for the frame to be right? What aggregate would falsify it?
Find the primary-source data (IEA, GWEC, CFR, Destatis, Census, central bank releases).
Identify the actor's budget constraint and throughput target.
Write the mechanism in one sentence.
If frame ≠ mechanism, ask why the frame persists. (Usually: domestic politics, lazy media, or the frame's beneficiary.)

The wind-turbine case is the canonical worked example. The method generalizes.