Methodological Foundations

International Sovereignty Index — Measurement Framework & Computational Architecture

1. Institutional Purpose

The International Sovereignty Index (ISI) is a structural measurement framework designed to quantify the concentration of external supplier relationships across strategic dependency domains. The inaugural release covers the EU-27 as the founding cohort.

The ISI functions as a descriptive statistical instrument. It does not advocate policy positions, prescribe diversification targets, or rank countries by desirability of outcome. Its purpose is to provide a reproducible, transparent, and methodologically stable measurement of structural concentration exposure — enabling downstream analysis by researchers, policymakers, and institutions operating under their own analytical mandates.

The framework is maintained under principles of institutional neutrality. Published outputs reflect observed concentration patterns derived from public data. No editorial judgment is applied to the interpretation of scores.

2. Operational Definition

For the purposes of this framework, sovereignty exposure is defined operationally as the degree to which a state's access to a critical external input is concentrated among a small number of foreign suppliers or counterparties.

This definition is intentionally narrow. It measures structural concentration — the distributional shape of supplier relationships — rather than qualitative attributes such as political alignment, reliability, or substitutability of partners. A high concentration score indicates structural dependence on few sources regardless of who those sources are.

Definitional boundary. The ISI does not measure sovereignty in the normative, legal, or political sense. It measures the statistical concentration of bilateral exposure in defined economic domains.

3. Scope of Measurement

3.1 What the ISI Measures

Concentration of external supplier shares within defined dependency channels
Channel-level and axis-level aggregation of Herfindahl-type concentration indices
Cross-country comparison of structural concentration profiles within the current release cohort (EU-27)
Composite concentration across six strategic axes

3.2 What the ISI Does Not Measure

Economic efficiency or competitiveness. A concentrated supplier structure may reflect efficient specialisation. The ISI makes no judgment on optimality.
Geopolitical risk or threat assessment. Concentration is measured irrespective of the identity, alignment, or stability of counterparties.
Resilience, adaptability, or substitution capacity. The framework captures static concentration at the observation window. It does not model dynamic adjustment paths.
Domestic production capacity or self-sufficiency. Import concentration does not account for domestic alternatives. A country that produces 90% of an input domestically and imports the remaining 10% from one supplier will register high concentration on that import channel.
Intra-EU vs. extra-EU distinction. In v0.1, all bilateral partners — including EU member states — are treated as external suppliers. The framework does not currently distinguish between intra-Union and extra-Union dependency.
Ordinal ranking or relative positioning. The ISI produces cardinal scores on a continuous [0, 1] interval. No transformation to ordinal ranks is performed. Published rank displays are derived convenience outputs and do not form part of the core measurement.

4. Concentration Doctrine

The mathematical backbone of the ISI is the Herfindahl-Hirschman Index (HHI), a standard concentration measure used in competition economics, trade analysis, and regulatory assessment. The ISI applies Herfindahl-type logic uniformly across all axes.

4.1 Core Formula

For a given reporting country, within a defined dependency channel, let s_i denote the share of total exposure attributed to supplier i, where shares are normalised such that Σ s_i = 1. The channel-level concentration index is:

C = Σ s_i² for all suppliers i = 1, …, N

4.2 Boundary Properties

Condition	Score	Interpretation
Single supplier (s₁ = 1)	1.0000	Maximum concentration
N equal suppliers (s = 1/N)	1/N	Symmetric diversification
N → ∞ with equal shares	→ 0.0000	Perfect diversification (theoretical limit)

4.3 Classification Thresholds

For interpretive convenience, concentration scores are assigned to four named bands. These thresholds are aligned with established HHI conventions and are applied uniformly across all axes and channels.

Range	Classification
≥ 0.50	Highly Concentrated
0.25 – 0.49	Moderately Concentrated
0.15 – 0.24	Mildly Concentrated
< 0.15	Unconcentrated

These classifications are convenience labels. They do not carry normative weight and should not be interpreted as risk thresholds. The underlying continuous score is the primary measurement output.

5. Layered Computational Architecture

The ISI processes bilateral data through a four-layer pipeline. Each layer is deterministic: identical inputs produce identical outputs with no stochastic component.

5.1 Layer 1 — Supplier Share Computation

For each reporting country r and dependency channel c, bilateral flows are aggregated from the underlying data source. The share of each supplier i is computed as:

s_i = flow(r, i, c) / Σ_j flow(r, j, c)

Shares are computed from the reporter-side perspective. All bilateral observations for the selected channel and observation window are included. Aggregate, re-export, and unallocated entries are excluded at this stage (see Section 8).

5.2 Layer 2 — Channel-Level Concentration

Supplier shares are squared and summed to produce the Herfindahl concentration index for each channel:

C_c = Σ_i s_i²

This produces a single scalar on [0, 1] for each (country, channel) pair. No adjustments, caps, or smoothing functions are applied. The raw HHI is the channel output.

5.3 Layer 3 — Axis-Level Aggregation

Each axis comprises one or more dependency channels. Channel-level concentrations are aggregated to the axis level using a volume-weighted mean:

A = Σ_c (V_c × C_c) / Σ_c V_c

where V_c is the total bilateral volume for channel c (denominator of the share computation in Layer 1). This ensures that channels with larger exposure volumes exert proportionally greater influence on the axis score. The weighting reflects economic materiality, not editorial judgment.

5.4 Layer 4 — Composite Score

The ISI composite is computed as an unweighted arithmetic mean of all axis-level scores for which a country has valid data:

ISI_composite = (1/N) × Σ_a=1…N A_a

No cross-axis weighting is applied. This is a deliberate design constraint. The framework does not presume which dependency domains carry greater strategic significance — that judgment belongs to the analyst. The composite provides a first-order summary; axis-level scores provide the analytical substance.

6. Axis Coverage

The v0.1 framework covers six strategic dependency domains. Each axis is independently constructed from domain-specific bilateral data sources and follows the shared computational architecture described in Section 5.

Axis	Domain	Observation Basis
Energy	Fossil fuel and electricity import concentration	Bilateral trade flows
Critical Inputs	Raw materials and critical mineral import concentration	Bilateral trade flows
Technology	Semiconductor and advanced technology import concentration	Bilateral trade flows
Defense	Military equipment and arms transfer concentration	Bilateral transfer records
Financial	Banking, investment, and financial services exposure concentration	Bilateral financial positions
Logistics	Freight, shipping, and supply chain routing concentration	Bilateral trade flows

Detailed channel definitions, data source citations, inclusion/exclusion criteria, and known limitations for each axis are published on the corresponding axis detail pages, which are generated from the backend axis registry.

7. No Normalisation Principle

ISI scores are not normalised, rescaled, z-scored, percentile-ranked, or subjected to any distributional transformation. This is a foundational design decision with the following rationale:

Cardinal interpretability. A score of 0.3200 means the sum of squared supplier shares equals 0.3200. This interpretation is stable across time, across axes, and across framework versions.
Cross-temporal comparability. If normalisation were applied relative to the current distribution, a country's score would change when other countries' data changes — even if its own supplier structure remained constant. Unnormalised scores avoid this contamination.
Audit transparency. Any analyst with access to the underlying bilateral data can reproduce the published score by computing shares and summing squares. No knowledge of distributional parameters is required.
No artificial discrimination. Normalisation would inflate differences in regions of the distribution where true variance is low, creating false separation between structurally similar countries.

8. Reproducibility & Audit Standards

The ISI is designed for deterministic reproducibility. The following standards apply to all published outputs:

8.1 Deterministic Pipeline

All computation is performed by a deterministic export pipeline. No random sampling, Monte Carlo estimation, or machine learning inference is used at any stage. Identical source data produces identical outputs on every execution.

8.2 Public Data Provenance

All underlying bilateral data is sourced from publicly accessible datasets maintained by international statistical authorities. Source identifiers, observation windows, and extraction parameters are documented in the axis registry metadata for each published version.

8.3 Audit Artifacts

The API layer publishes structured audit breakdowns at the country-axis level, exposing intermediate values including channel-level concentrations, supplier counts, volume weights, and top-partner shares. These artifacts enable third-party verification without requiring access to the computation pipeline.

8.4 Exclusion Documentation

All systematic exclusions are documented. Common exclusions across axes include:

Aggregate partner codes (e.g., “World”, “Unspecified”, regional totals)
Re-export and transit-trade entries where identifiable
Observation records with zero or negative reported values
Countries with insufficient bilateral coverage for reliable share computation

Axis-specific exclusions are documented on the relevant axis detail pages.

9. Structural Limitations

The following limitations apply to the v0.1 framework and should be considered when interpreting published outputs:

Static observation window

Scores reflect concentration at a single observation period. Temporal dynamics, trend direction, and adjustment velocity are not captured.

Reporter-side perspective only

All shares are computed from the importing country's reported data. Mirror statistics (exporter-reported) are not cross-validated in v0.1. Discrepancies between reporter and partner records are a known limitation of bilateral trade statistics.

No intra-/extra-EU distinction

All bilateral partners are treated equivalently. Imports from another EU member state are treated identically to imports from a non-EU supplier. This may overstate external dependency for countries with high intra-EU trade concentration.

Domestic production not modelled

The framework measures concentration of imports, not overall supply. A country that produces 95% of an input domestically and imports 5% from one supplier will register a concentration score of 1.0 on that channel. The domestic share is outside the measurement boundary.

Data coverage heterogeneity

Not all countries report bilateral flows with equal completeness across all channels and axes. Coverage gaps may result in missing axis scores or artificially elevated concentration where only partial partner data is available. The number of scored axes per country is published alongside all outputs.

Equal composite weighting

The composite score weights all axes equally. This does not imply that all dependency domains are of equal strategic significance. The equal-weight design is a methodological neutrality constraint, not an analytical claim.

10. Versioning & Forward Compatibility

The ISI framework is versioned to ensure methodological traceability. All published outputs are tagged with a framework version identifier. The following versioning principles apply:

Version immutability. Once a version is published, its methodological specification is frozen. Corrections to published versions are issued as errata, not as silent modifications.
Backward traceability. Each version documents all methodological parameters, data source identifiers, observation windows, and exclusion rules in sufficient detail for full reproduction.
Forward extensibility. The layered architecture is designed to accommodate additional axes, alternative aggregation rules, and expanded geographic scope without requiring changes to existing axis computations. New axes extend the framework; they do not retroactively alter existing scores.
Breaking change protocol. Changes to the concentration formula, share computation methodology, or classification thresholds constitute breaking changes and require a major version increment.

Framework Version

ISI Baseline Methodology — Unweighted HHI Mean

Version: v1.0

Frozen at: 21 February 2026

Years available: 2024

Axes: 6

Aggregation: Unweighted Arithmetic Mean

Scope: EU-27 (founding release cohort)

Concentration measure: Herfindahl-Hirschman Index (C = Σ s_i²)

Normalisation: None

Classification thresholds: 0.15 / 0.25 / 0.50

This document describes the methodological foundations of the ISI as published. It does not constitute a forecast of future framework development.

← Back to Overview