China-HQ green FDI readiness / Method appendix
Back to the story

EV case notes

Purpose. Per-cell observations for the five EV country-technology pairs that anchor the China-HQ EV story: Hungary, Mexico, and Thailand (the largest exporter-base bets and the largest project count, first batch 2026-07-02) plus Indonesia and Brazil (the two large home-market entries, added 2026-07-03). The structural fact that separates EV from the solar and battery batches still holds, but now in two variants: every recipient has an existing auto-industry base, so the question is never whether the capital built a factory base from nothing. For the exporter cells (Hungary, Mexico, Thailand, factory-stage RCA above 1.0 before the window) the test is whether the capital is pulling that base toward finished battery-electric vehicles. For the home-market cells (Indonesia, Brazil, factory RCA far below 1.0) the capital arrived to serve protected domestic demand, and the export test is structurally premature by design.

Data sources. Per-cell capability and trade outcomes from data/processed/country_tech_quadrants.csv. Project lists from data/processed/china_outward_fdi.json. HS6 decomposition from data/processed/case_card_hs6_decomposition.csv. Destination decomposition (importer-side) from data/processed/ev_case_destination_evidence.csv. EV capability axis and basket from the dedicated EV model, documented in sources/ev_capability_measurement.md [3]. Manufacturing-only BACI cut by default; factory-stage RCA reported per cell. The finished-vehicle proxy is HS 870380 (battery-electric vehicles); hybrids (870340) are outside the EV basket by design. Four panel records whose own amount-source field reads not publicly disclosed carried implausible multi-billion capex entries (Shedrive/Thailand, TAILG/Indonesia, Yutong/Qatar, GAC/Ethiopia); the curated corrections layer blanks them to undisclosed at ingest, with per-record verification citations in data/manual/fdi_corrections.csv. A fifth EV record, Aima/Indonesia, was blanked on 2026-07-03 under a second contamination class: its amount citation points at an unrelated project, and the company's own SSE filing shows the subsidiary two orders of magnitude smaller than the panel figure [4].

Windows. EV capability pre-window 2018–2019 (the dedicated EV model's series starts in 2018, a documented deviation from the 2015–2019 window used for solar and battery [3]). China-HQ FDI window 2020–2024. Trade pre-window 2017–2019; trade post-window 2022–2024.

A basket caveat that binds hardest on EV. The EV manufacturing basket carries eight multi-use codes flagged in data/manual/ev_stage_types.csv: four integrated-circuit codes (854231/32/33/39), two semiconductor-equipment codes (848610/848620), generic instruments (903089), and generic iron castings (732599). These carry trade far beyond electric vehicles. They stay in the basket because the capability model was trained on it, but per-cell reads that lean on them inherit the commodity caveat. Thailand is the case where this binds hardest.

1. Hungary / EV

The clearest EV capability-formation case in the panel: modeled competitiveness rose from 0.11 to 0.79 across the window, finished battery-electric vehicles entered from near zero to 29.9% of the export-growth delta, and factory-stage RCA crossed further above 1.0. But the anchor megaproject completed late, so the pre-existing auto base carried most of the measured trade.

Hungary entered the FDI window as an incumbent auto exporter with a modest EV capability read: EV PC averaged 0.11 across the 2018–2019 pre-window, at the top-decile floor. It rose steadily afterward, to 0.13, 0.40, 0.80, 0.89, 0.79 across 2020–2024, the strongest sustained EV capability climb of the five cells. Manufacturing-only EV-basket exports were already $10.12B at factory-stage RCA 1.17. The China-HQ FDI signal is 4 projects and $4.79B in disclosed capex, dominated by one record: BYD Szeged ($4.50B, completed 2023), the company's first European passenger-EV plant. The remaining three are components: Chervon ($130M, completed 2020), Zhejiang Shuanghuan ($113M, completed 2023), and Huashuo ($42M, completed 2022). No HQ-audit exclusion applies; all four are mainland-HQ [2].

Manufacturing exports rose to $15.21B (+50%); factory-stage RCA moved from 1.17 to 1.25. The decomposition shows genuine finished-vehicle entry: HS 870380 (battery-electric vehicles) went from $10M pre to $1.53B post, contributing $1.52B or 29.9% of the $5.09B basket increase, the second-highest finished-vehicle share of the five cells after Mexico. Static converters (HS 850440, $1.03B, 20.2%), gearboxes (HS 870840, 9.6%), and AC motors (HS 850153, 6.2%) fill out an EV-drivetrain profile. On the destination side, EU27 + UK take 78.0% of post-window basket exports and 73.1% of the delta; finished battery-electric vehicles run 53.3% to the EU, led by the United States ($371M), Germany ($283M), and the UK ($209M). This is an export-base cell: output leaves the country rather than being absorbed at home.

The timing qualifies the causal read. Hungary's basket already exported $8.94B–$11.09B annually in 2017–2019, so the manufacturing base predates the window. Finished-vehicle exports were genuinely near zero before the window ($10M) and are new, which is the strongest single piece of EV entry evidence in the panel. But BYD Szeged completed only in 2023, late in the trade window, so most of the measured 2022–2024 finished-vehicle growth reflects Hungary's pre-existing auto-assembly base electrifying rather than commercial output from the China-HQ megaproject. The Szeged plant is a next-wave signal: the finished-vehicle entry and the capability climb are real and co-located with the China-HQ bet, but attribution to BYD specifically is a 2025-onward test.

PC pre 0.11 → 0.79 · FDI $4.79B / 4 projects · finished share 29.9% · factory RCA 1.17 → 1.25 · signal moderate_growth · export-base

2. Mexico / EV

Labelled FDI-First only because the EV capability window barely predates Mexico's ramp: modeled competitiveness jumped from 0.09 to 0.84. Finished battery-electric vehicle exports entered at scale ($6.6B of growth, 81.5% into the United States), but this is the existing North American auto industry electrifying, not the China-HQ EV capital, most of which is announced or just-commenced.

Mexico's EV cell sits below the 0.10 capability floor on the 2018–2019 pre-window (PC 0.09), which places it in FDI-First Experiment. That label is a window artifact: EV PC rose to 0.70, 0.82, 0.64, 0.84 across 2021–2024, so on any later window Mexico reads as a confident EV incumbent. It entered with by far the largest EV-basket export base of the five ($41.00B at factory-stage RCA 1.28), the North American auto platform. The China-HQ FDI signal is 11 projects and $3.75B in disclosed capex, but the composition matters: the largest records are announced or newly commenced, not operating: SAIC/MG ($1.05B, announced 2024), Geely-Volvo Nuevo León ($1.00B, commenced 2024), BYD ($600M, announced 2024), Xusheng ($350M, completed 2023), Tuopu ($200M, completed 2022). Under the documented-exclusions HQ screen the cell's disclosed capex falls from $3.75B to $2.75B, because the Geely-owned Volvo record is control-linked rather than mainland-HQ of record; the cell stays FDI-First under both rules [2].

Manufacturing exports rose to $60.42B (+47%); factory-stage RCA held roughly flat at 1.28 → 1.25. Finished battery-electric vehicles (HS 870380) drove the largest single contribution: $95M pre to $6.69B post, a $6.59B delta or 34.0% of the basket increase, the highest finished share of the five cells and by far the largest in absolute terms. Body parts (HS 870829, 14.9%), other vehicle parts (HS 870899, 8.2%), converters (HS 850440, 7.9%), and brakes (HS 870830, 7.7%) fill out the profile. The destination concentration is the tightest in the panel: USA + Canada take 85.1% of post-window basket exports and 96.2% of the delta, and 81.5% of finished battery-electric vehicles, with $4.77B of the $6.69B post to the United States alone. This is a USMCA export platform, not a home-market cell.

The finished-vehicle entry is real, but the FDI attribution is not. Mexico's China-HQ EV projects are overwhelmingly 2023–2024 announcements and commencements; none could have produced the 2022–2024 finished-vehicle export growth. That growth is the incumbent North American auto industry, largely non-China capital, electrifying its Mexican export base for the US market. The China-HQ EV FDI is a next-wave bet layered onto that platform, and the fact that the largest records are still announced rather than operating is itself the finding: the capital has arrived on paper but has not yet shown up in trade. Whether it converts is the 2027-onward test.

PC pre 0.09 → 0.84 · FDI $3.75B panel / $2.75B documented / 11 projects · finished share 34.0% · factory RCA 1.28 → 1.25 · signal moderate_growth · export-base

3. Thailand / EV

The largest China-HQ EV project count in the panel (12 projects, $2.07B disclosed) landed on an incumbent auto-assembly base, but the measured 2022–2024 export growth is 54% multi-use integrated circuits shipped to the United States, not electric vehicles. Finished battery-electric vehicles are 4.7% of the delta and run mostly to Japan; the plants completed too recently to appear in the trade window.

Thailand carries the largest China-HQ EV project count in the panel: 12 projects and $2.07B in disclosed capex, led by Greatwall Motor ($625M, completed 2020), Changan ($570M, completed 2023), and BYD Rayong ($490M, completed 2022). The set also includes Shedrive (completed 2024), a Shanghai Electric Drive (Broad-Ocean Motor) electric-powertrain plant in Chachoengsao whose investment value is not publicly disclosed; its panel entry carried an unsourced $7.6B figure, blanked by the corrections layer after verification against the reported facility scale. Its EV PC read is the noisiest of the five (0.13, 0.15, 0.72, 0.11, 0.27, 0.05, 0.09 across 2018–2024), averaging 0.14 on the pre-window, just above the floor, which places Thailand in Prepared Magnet. Manufacturing-only EV-basket exports were already the second-largest of the five at $22.81B, factory-stage RCA 1.18.

Manufacturing exports rose to $30.36B (+33%); factory-stage RCA edged down to 1.16. But the composition is where the EV read breaks. The $7.55B basket delta is 54.0% multi-use codes: integrated circuits n.e.c. (HS 854239, $3.39B, 44.9% alone) plus IC processors (HS 854231, 10.3%), both flagged multi-use, both shipped predominantly to the United States, and both reflecting Thailand's large general electronics-export base rather than EV-specific capability. Static converters (HS 850440, $2.65B, 35.1%) are a genuine drivetrain contribution. Finished battery-electric vehicles (HS 870380) contribute $352M, or 4.7% of the delta, the lowest finished share among the exporter cells; and on the destination side those finished vehicles run mostly to Japan ($295M), Japanese-brand BEVs assembled in Thailand and shipped back, not to a regional export market. ASEAN + Australia take only 21.6% of post-window basket exports. Thailand is also a net finished-BEV importer, absorbing more than it ships as its new capacity ramps.

The pattern is an incumbent auto-assembly economy where the EV-specific trade signal is not yet legible. Most of the projects completed in 2022–2024, at or past the end of the trade window, so little of the new capacity could have produced the measured growth; and that growth is, in any case, dominated by multi-use electronics rather than EV output. Thailand is the cell where the project count is largest and the trade evidence weakest: the bets are real and recent, the capability read is unstable, and the honest conclusion is that Thailand's EV outcome is unscored until a later window captures the post-2024 capacity. Read the Prepared Magnet label as capability-plus-capital, not as a demonstrated finished-vehicle export base.

PC pre 0.14 (noisy) · FDI $2.07B / 12 projects · finished share 4.7% · factory RCA 1.18 → 1.16 · signal moderate_growth · domestic-market / too-early

4. Indonesia / EV

The panel's clearest home-market experiment: $1.26B of disclosed in-window capital across 6 projects arrived in a country with no measured EV capability, pulled by an explicit import-now, produce-later incentive scheme. The trade data cannot yet see the bet: the EV-basket export delta is 87% alumina and nickel, the same midstream commodity signal as Indonesia's battery and solar cells, and finished battery-electric-vehicle exports are near zero.

Indonesia enters the EV window with no measured capability: EV PC reads 0.00 on the 2018 to 2019 pre-window and stays at 0.00 to 0.01 through 2024. Unlike Mexico, whose FDI-First label is a window artifact on a rising capability series, Indonesia's label describes the cell exactly; this is a genuine low-prior-capability experiment, the EV analogue of the solar and battery FDI-First cells. The in-window China-HQ signal is 6 projects and $1.26B disclosed: BYD Subang ($1.0B, announced 2024, 150,000 units per year, company-stated operations target of early 2026, initial output for the domestic market), Yadea Karawang ($150M across 2024 to 2028, groundbreaking May 2024, a planned 3 million electric two-wheelers per year), Xinri/Sunra Kendal ($108M in the panel; the company described a two-stage plan of roughly $120M and 1 million units per year at its May 2024 groundbreaking), plus Aima (operational March 2024, investment never disclosed), TAILG, and a second SGMW record (both undisclosed). This pass blanked the Aima record's $1,050M panel figure: its amount citation points at an unrelated Elite Solar Egypt project, named Indonesian trade press reported the investment as undisclosed, and Aima's own SSE guarantee filing shows the entire Indonesian subsidiary at roughly $10.7M registered capital with end-2024 total assets near $4.6M [4]. The operating base that predates the window is SGMW Wuling's Cikarang plant (completed 2017, $700M, 120,000 units per year), which built Indonesia's first locally made EV, the Air ev, from August 2022, and the BinguoEV from December 2023 at 47.5% local content.

Manufacturing-basket exports rose from $4.52B to $6.37B (+41%), but the composition is Indonesia's commodity story read a third time. Alumina (HS 281820) contributes $1.02B, 55.0% of the delta; unwrought nickel (750210) 17.6%; nickel sulphates (283324) 14.1%; generic vehicle parts (870899, 15.0%) are the only large component contribution. Finished battery-electric vehicles contribute $9M, 0.5% of the delta, and factory-stage RCA declines from 0.28 to 0.17. The destination side confirms the materials read: the delta runs to Malaysia (31.2%, alumina feeding smelters), China (22.0%, nickel forms), and India (21.6%), not to vehicle markets, and Indonesia is a net finished-BEV importer at $574M per year post-window. Alumina sits in six green-technology baskets and the nickel codes are battery-chain inputs, so this is real industrial output but not EV-specific capability; the cell is assigned to the midstream materials family in the Lens-2 taxonomy, alongside Indonesia's battery and solar cells [1].

What the trade window cannot yet see is the policy machinery assembling behind it. Indonesia pairs nickel downstreaming (hilirisasi; ore export ban since January 2020) with an explicit EV-assembly pull: Perpres 79/2023 and the BKPM incentive regulations let committed manufacturers import built-up BEVs duty-free with government-borne luxury tax through 2025, against a bank-guaranteed obligation to produce locally, one for one, in 2026 to 2027, on a local-content ladder of 40% through 2026 rising to 60% from 2027. The market response is visible in sales, not yet in trade: BEV wholesales grew from 120 units in 2020 to 43,188 in 2024 and 103,900 in 2025, with Chinese brands taking roughly 80% of 2025 sales (BYD about 45%, Wuling 12%) and Hyundai, the incumbent first local BEV assembler, down to 1.8%. Every large China-HQ plant is post-window (BYD Subang targeted early 2026; Yadea 2026; Sunra 2025), so the 2027 trade refresh is the first that can score whether the import-now, produce-later scheme converts into local output, and whether any of it exports; Wuling's cumulative exports of roughly 5,500 vehicles against 12,710 domestic BEV wholesales in 2025 alone suggest the home market absorbs nearly everything. The two-wheeler bet is its own experiment: announced Chinese electric-two-wheeler capacity (Yadea 3 million plus Sunra 1 million units) exceeds current electric-two-wheeler demand, roughly 1% of a 6.3-million-unit annual motorcycle market, by an order of magnitude.

PC pre 0.00 (flat through 2024) · FDI $1.26B / 6 projects · finished share 0.5% · factory RCA 0.28 → 0.17 · signal flat · home-market, midstream trade read

5. Brazil / EV

The cleanest tariff-wall case in the panel: a published EV import-duty ladder (10% in January 2024 stepping to 35% by July 2026) preceded a surge of Chinese vehicle imports and then the conversion of BYD's largest overseas export market into its largest plant outside Asia. Both in-window plants began output in mid-2025, after the trade window closed, so the measured export delta is the incumbent auto industry's parts trade plus alumina, and finished-vehicle exports are near zero.

The in-window signal is 2 projects and $1.72B, both verified as sourced in the amount audit [4]: BYD Camacari ($1.0B; announced July 2023 at R$3.0B and later scaled to R$5.5B by the company; sited partly on the former Ford Camacari plant in Bahia) and GWM Iracemapolis ($720M; the former Mercedes-Benz plant in Sao Paulo state acquired in 2021, within a R$4B phase-one program through 2026). The documented sequence is consistent with tariff-induced localization. Chinese BEVs had entered at zero duty; the government reinstated import duties on a published ladder (10% January 2024, 18% July 2024, 25% July 2025, 35% July 2026), with Reuters reporting the schedule as explicitly aimed at encouraging local production; importers front-ran each step (Brazil overtook Belgium as the largest export market for Chinese NEVs in April 2024, importing 40,163 units in that month alone, thirteen times the year before; BYD registered 76,713 vehicles in Brazil in 2024, up 328%, its largest overseas market); and the plants followed, with the first Dolphin Mini rolling off the Camacari line on July 1, 2025 and GWM's Haval H6 hybrid production starting August 15, 2025. In July 2025 the government also brought the kit-import (CKD/SKD) schedule forward to 35% by January 2027 after BYD requested relief and incumbent automakers protested, closing the semi-knocked-down route the new plants opened with.

The 2022 to 2024 trade window closes before any of that output exists. Manufacturing-basket exports rose from $6.06B to $7.63B (+26%, delta $1.57B), and the composition is the pre-existing auto industry plus a commodity: body parts (870829), multi-phase AC drive motors (850153 and 850152, $253M combined), and brakes (870830) lead the Product Component share (66% of the delta), running mostly to Argentina (39.6% of the delta) inside the Mercosur auto system, while alumina (281820, 29.8%) flows to smelters in Norway, Canada, and Iceland, the same multi-use commodity that inflates Indonesia's cells. Finished battery-electric vehicles contribute $1M, 0.1% of the delta; factory-stage RCA reads 0.19 falling to 0.15; and Brazil is a net finished-BEV importer at $918M per year post-window. Brazil produced 2.55 million vehicles in 2024, eighth in the world, with exports falling 1.3%: a home-market and regionally oriented industry, and the new EV capacity is aimed the same way.

Attribution therefore runs entirely through timing, and the record carries real caveats. BYD's Camacari buildout was disrupted by a labor investigation: in December 2024 inspectors removed 163 workers employed by contractor Jinjiang under findings of degrading, slavery-like conditions, work visas were suspended, and labor prosecutors filed a R$257M civil action in May 2025; the plant opened officially in October 2025 working initially from semi-knocked-down kits, with Bahia state officials expecting full operation by December 2026. The announced next wave is large but out-of-window and softer than the trackers suggest: GAC's own communications pledge R$6B (roughly $1.0B to $1.1B, below the $1.3B tracker figure) with a plant conditional on market response, GWM's Aracruz second plant (announced June 2026, up to 200,000 units per year) targets 2029, and the Chery Anapolis expansion is Brazilian Caoa capital with Chinese manufacturing partners, not China-HQ FDI [4]. Like Mexico, Brazil is a next-wave cell; unlike Mexico, whose modeled EV competitiveness jumped to 0.84 after the window, Brazil's stays at 0.01 to 0.05 through 2024, so the bet rides on the protected market, not on measured capability. The 2027 refresh scores whether tariff-wall assembly deepens into components and, eventually, Mercosur exports.

PC pre 0.01 (flat through 2024) · FDI $1.72B / 2 projects · finished share 0.1% · factory RCA 0.19 → 0.15 · signal moderate_growth · home-market

Cross-cell observations

  1. Every EV recipient has an existing auto-industry base, but the batch splits into two variants. Hungary, Mexico, and Thailand enter as established vehicle exporters with factory-stage RCA already above 1.0 (1.17, 1.28, 1.18), so their test is whether the capital pulls that base toward finished battery-electric vehicles: Hungary (29.9%) and Mexico (34.0%) show real entry, Thailand (4.7%) does not yet. Indonesia and Brazil enter far below specialization (factory RCA 0.28 and 0.19) as home-market producers behind incentive and tariff walls; their capital is market-seeking, both are net finished-BEV importers, and the export test is structurally premature rather than failed.
  1. Trade legibility does not track capital size or project count. Thailand carries the largest project count (12) yet the weakest EV-specific trade signal; Mexico's finished-vehicle growth is attributable to the pre-existing (largely non-China) auto industry; Hungary carries the largest disclosed anchor (BYD Szeged, $4.5B) and the cleanest capability-formation read, but that anchor completed too late in the window to be the cause; Indonesia's and Brazil's measured deltas are commodities and incumbent parts trade, with the actual bets invisible until plants that opened in 2025 and 2026 reach the data. Across all five, the bets are too recent for trade to score.
  1. Timing dominates all five. Shedrive (Thailand, completed 2024, value undisclosed), BYD Szeged ($4.5B, Hungary, 2023), Mexico's SAIC / Geely-Volvo / BYD records (all 2024), BYD Camacari and GWM Iracemapolis (Brazil, first output July and August 2025), and BYD Subang (Indonesia, targeted early 2026) all completed or were announced at or after the end of the 2022–2024 trade window. None could have produced the measured outcome. The EV batch is therefore a next-wave read; the real test is a 2027-onward trade window, exactly as for the largest solar and battery FDI-First bets.
  1. The multi-use caveat binds on three of the five cells, through two different mechanisms. Thailand's delta is 54% integrated circuits, basket codes that are not clean-tech-specific. Indonesia (87%) and Brazil (30%) carry alumina and nickel forms, codes the EV basket shares with the battery chain and, in alumina's case, with six green-technology baskets. Where these dominate, the "EV export growth" reading is not trustworthy: Thailand is reported as too-early, Indonesia's cell is assigned to the midstream family outright, and Brazil's component read survives because parts, not alumina, dominate its delta [1].
  1. Destination decomposition separates the variants cleanly. Mexico is a USMCA platform (96.2% of the delta, 81.5% of finished BEVs to USA + Canada); Hungary is an EU platform (73.1% of the delta, 53.3% of finished BEVs to EU27 + UK). Thailand is a net finished-BEV importer whose parts and ICs serve external demand while finished-vehicle output is home-absorbed during the ramp. Indonesia's delta runs to Malaysia, China, and India as industrial materials, not vehicles; Brazil's runs to Argentina as Mercosur parts trade plus alumina to Atlantic smelters. Neither home-market cell exports finished battery-electric vehicles at any scale ($9M and $1M respectively).
  1. The HQ-audit sensitivity is EV-concentrated but touches only Mexico among the five. Mexico/EV moves from $3.75B to $2.75B under the documented-exclusions screen because the Geely-owned Volvo record is control-linked, not mainland-HQ; the cell's FDI-First label survives. Indonesia's SGMW records belong to the audit's one flagged joint-venture firm (SAIC-GM-Wuling), but its in-window record carries no disclosed capex, so no magnitude changes. Thailand, Hungary, and Brazil carry no exclusions; panel-rule and documented-screen figures are carried side by side [2][3].
  1. The corrections layer is load-bearing for the EV panel's home-market cells. Five of the six corrected-or-blanked EV amounts sit in this batch's countries (Shedrive/Thailand, TAILG/Indonesia, Aima/Indonesia, plus the out-of-window BYD Sao Paulo and Yadea Thailand records), and the Aima case surfaced a second contamination class: a real source URL attached to the wrong project. Panel figures for market-seeking EV FDI appear systematically noisier than for the exporter cells, consistent with these projects being announced through local channels with weaker disclosure [4].

References

Markdown source on GitHub