What exactly is being measured?
The objective is not to create a hierarchy in which “concentration” mechanically becomes “dependency” and then “strategic vulnerability”. The empirical literature instead uses several complementary filters to identify products for which a disruption in foreign supply may be difficult to absorb. Each criterion captures a distinct margin of exposure, and the classifications used on this website are simply the intersections of those criteria.
How five different methodologies define a "vulnerable product"
Statistical methodologies to identify import vulnerabilities have proliferated since 2020. The table below (criteria compared as in Vicard and Wibaux, 2024) lines up the five main approaches used in France and at EU level.
| Criterion | European Commission (2021) | Bonneau & Nakaa — French Treasury (2020) | Jaravel & Mejean — CAE (2021) | Baur & Flach — CESifo (2022) | Mejean & Rousseaux (2024) |
|---|---|---|---|---|---|
| 1 | Import concentration: HHI > 0.4 | Origin of imports: >50% from outside the EU | Origin of imports: majority extra-EU | Relevance for domestic production: top-3 inputs used in the 5 largest sectors | Import concentration: HHI > 0.4 |
| 2 | Importance in demand: extra-EU / total imports > 0.5 | Import concentration: HHI > 0.5 | Import concentration: HHI > 0.5 | Import concentration: HHI > 0.33 | Importance in demand: extra-EU / total imports > 0.5 |
| 3 | Substitutability by EU production: extra-EU imports / EU exports > 1 | Diversification potential: centrality risk > 2.5 | Demand granularity: a single French firm accounts for ≥90% of imports | Substitutability: imports / exports > 1 | Substitutability by EU production: extra-EU imports / EU exports > 1 |
| 4 | — | — | — | — | Exposure in domestic absorption: imports / (Y + M − X) > 0.5 |
| 5 | — | — | — | — | Low ex-post substitutability: stickiness (Martin et al., 2024) > 3rd quartile (HS6) |
Source: Rousseaux (2025), building on Vicard and Wibaux (2024)'s comparison. Data: CEPII-BACI, Prodcom-Eurostat.
The European Commission's bottom-up approach (Commission Staff Working Document, 2021) is the reference starting point for almost the entire literature. A product is flagged if it simultaneously meets all three conditions: import concentration (HHI > 0.4), a majority extra-EU sourcing share, and extra-EU imports that exceed total EU exports of the same product. All three rely purely on observed trade flows — no production data is needed, which makes the approach easy to replicate across time and countries, but also means it can only proxy for domestic substitution capacity indirectly, through EU export volumes.
Bonneau and Nakaa (2020), for the French Treasury (Direction générale du Trésor), were among the first post-Covid methodologies. They combine an import-origin criterion with an import-concentration criterion (HHI > 0.5) and a network-centrality measure of diversification potential, rather than the EC's export-based substitutability proxy.
Jaravel and Mejean (2021), for the French Council of Economic Analysis (CAE), add a firm-level granularity criterion specific to France: a product is flagged as especially fragile if a single French importing firm accounts for 90% or more of national imports of that product — a check on concentration at the level of individual firms, not just countries.
Baur and Flach (2022), at CESifo, start instead from production relevance: which imported inputs are actually used intensively in the sectors that matter most for the domestic economy. They then apply a somewhat looser concentration threshold (HHI > 0.33) and the same substitutability logic as the EC.
Mejean and Rousseaux (2024) keep the EC's three criteria intact and add two more, aimed specifically at closing the gap between dependency and strategic vulnerability: a domestic-absorption criterion using actual production data (Prodcom-Eurostat) rather than export volumes as the substitution proxy, and an ex-post substitutability criterion built on inter-firm relationship stickiness (Martin et al., 2024). See the funnel below for what each addition removes from the list.
The five Mejean & Rousseaux (2024) criteria, one at a time
Applied at the HS6 product level (2017 nomenclature), using CEPII-BACI trade data and Eurostat-Prodcom production data.
Intuition: if a handful of countries supply almost all of a product, a shock to any one of them cannot easily be absorbed by switching to another supplier in real time.
Data: bilateral trade values by HS6 product and exporting country (CEPII-BACI).
What it eliminates: the vast majority of the roughly 5,000+ HS6 products traded by the EU — most goods are sourced from a reasonably diversified set of countries.
Intuition: if most of what the EU imports of a product already comes from other member states, the EU has an intra-bloc supply cushion even if the product is concentrated among a few original (possibly non-EU) sources further upstream.
Data: intra-EU vs. extra-EU trade flows (CEPII-BACI).
Intuition: if the EU exports at least as much of a product as it imports from outside the bloc, that is indirect evidence of a domestic production base large enough to plausibly substitute for the foreign supply.
Limitation this motivates criterion 4: export volumes are only an indirect, and sometimes misleading, proxy for production capacity — round-tripping and re-exports can inflate EU export figures without reflecting genuine local capacity.
Intuition: complements the export-based proxy in criterion 3 with actual EU production data. A product only qualifies if extra-EU imports account for more than half of EU absorption, defined as production plus imports minus exports. This criterion directly asks whether foreign supply is quantitatively important relative to the amount of the product absorbed in the EU.
Data: Eurostat-Prodcom production statistics, matched to HS6 trade codes.
What it eliminates: products for which the EU already produces enough to cover most of its own needs even though import flows look concentrated — for example, many agricultural products protected by the Common Agricultural Policy. This single criterion cuts the EC-list roughly in half in the 2015–2019 reference period.
Intuition: even a highly concentrated, poorly substitutable-by-production product may not be a genuine vulnerability if buyers can quickly switch to a different foreign supplier after a shock. The stickiness indicator is a product-level measure derived from the persistence of firm-to-firm trade relationships. Products with especially persistent relationships are interpreted as having lower ex-post supplier substitutability: when a relationship is disrupted, finding and qualifying an alternative supplier is likely to be more difficult. Mejean and Rousseaux retain products in the upper quartile of the HS6 stickiness distribution (a value above 3.2 in their application).
Why this matters empirically: Boehm, Flaaen and Pandalai-Nayar (2019) find that Japanese firms hit by the 2011 Tōhoku earthquake cut production almost one-for-one with the drop in imports — consistent with rigid, Leontief-like input structures. But Lafrogne-Joussier, Martin and Mejean (2023) find that French firms exposed to the first China lockdown in January 2020 fared no worse if they were undiversified beforehand, because undiversified firms were also the fastest to find new foreign partners after the shock. Ex-ante diversification and ex-post substitutability do not always move together, which is exactly why criterion 5 is measured directly rather than assumed.
Applying the criteria sequentially: how many products are actually vulnerable?
Click through the steps below to see what each additional criterion removes. Toggle the period to compare the paper's 2015–2019 reference window with the live 2021–2024 replication.
These are the headline figures from Mejean and Rousseaux (2024). The EU diagnostic page presents a live replication of this funnel and its full sectoral and geographic breakdown using the underlying data, together with the year-by-year evolution since 2015.
Data sources
Product positioning within global value chains uses the upstreamness metric of Antràs, Chor, Fally and Hillberry (2012); end-use categories follow the UN Broad Economic Categories (BEC) classification; sector groupings follow the NACE classification.