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

Capability measurement

Purpose. Specifies the country-technology capability measures used as Lens-1 inputs, the floor that separates Prepared Magnet from FDI-First Experiment, and the design choices made to defend each.

Primary direct measure: PC (predicted competitiveness)

The primary capability measure is pc_level_pre: a country-technology score in [0, 1] produced by goodenough's predicted-competitiveness model for 10 green technologies over 2003-2024 [1]. PC is fit on revealed-comparative-advantage patterns across the full HS6 product space and represents the probability that a country will be a competitive exporter of the technology given its current export portfolio. The measure is lagged to the 2015-2019 pre-FDI window. Using a same-period PC would be circular: FDI-driven output increases in the 2020-2024 window would feed back into the PC score, contaminating it as an input to the diagnostic.

PC is preferred over direct technology RCA as the headline measure because PC reflects both current technology exports and the export structure that makes a country well-positioned to expand into the technology. RCA alone misses the broader industrial substrate, especially for cells like Mexico/Battery (auto-supplier base that PC partially picks up but direct battery RCA misses).

Secondary measure: broad RCA footprint

The secondary measure is broad_rca_footprint_pre: the median RCA across six broad green-category baskets, per country-technology cell, in the 2015-2019 window [2]. The decision to use median rather than mean is load-bearing. Mean-across-categories is contaminated by undiversified-economy bias: a country with one extreme high-RCA category and five very low ones reads as having industrial breadth when it has the opposite. Median is robust to this bias.

This measure is not a textbook Hidalgo-style phi-weighted product-space relatedness density. Building that measure properly requires running the full HS6 lift on BACI export data and is a v2 task. The pre-aggregated six-category cut captures the same intuition (industrial breadth into related categories) at far lower computational cost and is defensible as a v1 placeholder. The cells where the choice matters substantively, such as Mexico/Battery and Saudi/Solar, are flagged in the per-cell observations.

Quadrant floor: PC = 0.10

The Prepared Magnet vs FDI-First Experiment distinction depends on a single threshold: pc_level_pre >= 0.10. The floor is absolute, not a percentile of the panel. It corresponds to roughly the top decile of incumbent capability in each scored technology's universe (11 of 155 solar cells, 11 of 155 battery cells, and 16 of 158 EV cells sit at or above it). The floor was chosen intentionally strict so the label "Prepared Magnet" connotes confident incumbency, not "above the median." A 30th-percentile threshold would have inflated the Prepared Magnet count and weakened the diagnostic distinction.

The trade-off is documented: cells with substantial adjacent industrial bases but weak direct technology exports fall below the floor by design. Mexico/Battery (PC 0.014) is the load-bearing example. The station handles this by treating supply-chain adjacency as a Lens-2 finding on the trade outcome side rather than a quadrant input, and by carrying an explicit v2 backlog item for an adjacent-capability axis.

Composites for ranking, not classification

A composite capability_momentum_index (50% PC level, 25% PC momentum, 25% broad RCA footprint) is computed for within-quadrant ranking and case selection. The composite is never used for quadrant membership; that depends only on PC and the FDI signal.

The 50/25/25 weighting is pre-registered. PCA on the standardized capability matrix yields data-derived weights of 38/30/33, and the pooled Spearman rho between the two weightings is 0.998 [3]. The ranking is therefore not sensitive to the pre-registered weights, which is why the simpler transparent weighting is used in public surfaces.

Scope: Solar, Battery, and EV

Quadrant scoring covers Solar, Battery, and EV. The v1 release scored only Solar and Battery, because goodenough's PC model trains on 10 green technologies and does not include EV. EV joined the scored panel in July 2026 via a dedicated EV predicted-competitiveness model; the companion note "EV capability measurement" in this rail documents the model's provenance, its shortened 2018-2019 pre-window, and the EV basket's stage-type curation. This note's PC and broad-RCA design choices apply to the goodenough-backed Solar and Battery axes; the EV note records where the EV axis matches them and where it deviates.

Wind remains excluded because the 2020-2024 China-HQ FDI window holds only 15 wind projects with under $0.5B in disclosed capex, which is much thinner than the scored sectors; Wind appears in the narrative as a prose finding (China-HQ wind firms have not globalized like solar and battery firms) rather than as a quadrant cell. Wind records remain in the underlying FDI panel for ad-hoc analysis and do not influence any cell-level claim. The ten-cell narrative batch remains Solar + Battery.

Time windows

Capability is measured over 2015-2019. FDI is observed over 2020-2024. Trade outcomes are pre-windowed at 2017-2019 and post-windowed at 2022-2024. The capability pre-window avoids same-period FDI leakage into PC; the trade pre-window smooths year-on-year volatility and matches conventional gravity-equation timing.

The capability window does not overlap the FDI window. The trade pre-window overlaps the tail end of the capability window, which is intentional: the trade pre-window measures the export base immediately before the FDI window opens, while the capability window measures the deeper industrial substrate from which that base was generated.

Lagged-window false negatives

The pre-FDI capability window is a deliberate choice with a cost: a country whose technology takeoff postdates 2019 reads as low-capability even when it is rapidly becoming competitive. India/Solar is the cleanest example. Its 2015-2019 PC averages 0.015, far below the 0.10 floor, so it cannot be a Prepared Magnet regardless of its FDI signal. It carries two China-HQ solar projects with undisclosed capex, which places it in the Active (magnitude below or unknown) category. Its PC then accelerates sharply: 0.06 in 2021, 0.32 in 2022, 0.62 in 2023, 0.67 in 2024, among the fastest-rising solar trajectories in the panel.

The acceleration is invisible to a lagged measure by construction. A momentum term computed on the same pre-window would not recover India either, because its 2015-2019 PC momentum is also near zero: the takeoff happens entirely after the window closes. Only a later observation window catches it. This is the temporal mirror of the USA/Solar "rule-correct but substantively different" caveat: a static label can both overstate a mature incumbent and understate a fast riser. A forward-looking flag or a PC-trajectory inset for the most salient false negatives, India/Solar first among them, is a v2 candidate.

References

Markdown source on GitHub