Resilience is a network property: Europe’s integration test for 2050

Europe is converging on a definition of economic security as reduced dependence – de-risking, reshoring, buying European. This column argues that the definition is mistaken. Three decades of evidence on financial and production networks shows that resilience is a property of a node’s position in the network – the depth, diversity, and quality of its connections – not of the number of connections it has severed for self-sufficiency. Applying this test to critical input supply chains such as energy and defence – with Turkey, the EU’s largest customs union partner, as the defining case – the author argues that Europe passes the initiative’s ‘Singapore test’ only if it treats integration depth, rather than self-sufficiency, as the union’s security variable.

Europe has spent the last few years redefining economic security as reduced dependence. The instruments differ – ‘buy European’ procurement clauses, domestic content targets, strategic autonomy in its harder readings, with tariffs on Chinese EVs – but the underlying policy variable is the same: fewer external links, more security. The question this initiative poses is whether that variable is the right one on a thirty-year horizon (Blanchard et al. 2026). The evidence from research on global financial and production networks says it is not. If Europe optimises for self-sufficiency between now and 2050, it will arrive poorer and, on its own chosen criterion, less secure.

Resilience is a network property

The intuition behind the current consensus is not wrong; it is incomplete. Shocks do propagate through input-output linkages and concentrate at critical nodes (Acemoglu et al. 2012). Interdependence can be weaponised by whoever controls the hubs (Hirschman 1945, Farrell and Newman 2019). And fragmentation, once under way, is costly – plausible estimates run to several percentage points of global GDP, with losses concentrated in the most open economies (Aiyar et al. 2023, Gopinath et al. 2025). So Europe has decided that exposure itself is the bigger risk.

The network evidence supports a different inference. Whether a shock is absorbed or amplified depends on network topology. A node with many diversified connections spreads any given shock thinly; a node with few connections concentrates its entire exposure in whichever links remain. Severing ties does not remove risk – it removes insurance. In earlier work with Bent Sørensen and Oved Yosha, I showed that regions and countries which share risk through cross-border ownership of productive assets can afford to specialise — the mechanism Obstfeld (1994) formalised, whereby better risk-sharing shifts economies toward higher-risk, higher-return activities — and that specialisation raises income precisely because the risk it creates is diversified away through integration (Kalemli-Özcan et al. 2001, 2003). Note that these gains differ from the standard efficiency gains of the static trade literature: they arise from dynamic, intertemporal trade in financial assets that diversifies risk and can change the original calculation of Lucas on the costs of business cycles on welfare and growth as shown by Obstfeld (1994). Autarky forces the opposite trade: each country self-insures by replicating capacity, forgoing these gains while retaining undiversified exposure to its own shocks. That is what strategic autonomy means when implemented literally, and it is one reason the scale deficits documented by Draghi (2024) will not be closed from within Europe.

Integration also fails when it is layered on unrepaired domestic frictions. When Southern Europe received large capital inflows after monetary union, the flows were intermediated as debt through shallow domestic financial markets and were misallocated towards low-productivity firms (Gopinath et al. 2017). The lesson of that episode is not that integration failed; it is that shallow integration layered on existing domestic financial frictions fails. Deep integration – equity-like, diversified, intermediated through developed capital markets of the kind the capital markets union is meant to build – is what buys resilience through financial and trade integration. The policy variable for 2050 is therefore not the quantity of Europe’s external links but their depth and composition.

Directed threats, chosen partners, and deterrence

The strongest objection to this framing is that threats are directed based on asymmetric trade relations. Trade coercion comes from specific places, and when the coercer is identifiable – tariffs from Washington, gas from Moscow – reducing exposure to that particular counterparty is simply prudent. This is correct, and it is worth being precise about what it implies. Rebalancing away from a likely coercer is not decoupling; it is rewiring. The substitute for a dangerous link is another link – a different supplier, a new ownership stake, a deeper partnership – not domestic replication of capacity. The policy error the network evidence warns against is answering a risky counterparty with self-sufficiency rather than with substitution. The deeper difficulty with organising policy entirely around identified adversaries is that alignment is time-varying. The state whose coercion Europe now hedges most actively was, a decade ago, the anchor of its security and the architect of the trading system it now defends against. If ‘reliable partner’ can be reclassified within ten years, a portfolio concentrated in today’s friends recreates concentration risk under a more comforting label. Breadth is the hedge against being wrong about who turns hostile.

There is also a complementary security argument for maintaining dense links, including with potential adversaries: deterrence by entanglement. A Europe embedded in connections that its potential coercers themselves value is a Europe that can retaliate in kind – the economic analogue of mutual assured destruction (MAD), the defensive mirror of the weaponised interdependence that Farrell and Newman (2019) describe. And the two conditions that make such deterrence credible are both network properties depending on linkages. Exposure must be roughly symmetric, because interdependence deters only when the adversary also has something to lose. And retaliation must be survivable, because a country that would be devastated by the first blow cannot credibly threaten a second. Resilience via production, trade and finance networks via diversified links, mapped dependencies, insured single points of failure – is the second-strike capability of economic MAD. Deterrence and resilience are therefore not rival arguments for integration; resilience is the credibility condition on which the deterrent rests.

The euro: Coordination, not grievance

The same logic governs the monetary question. The dollar’s dominance is a coordination equilibrium: it is used because it is used, across invoicing, funding, and reserves simultaneously (Gourinchas and Rey 2007, Gopinath and Stein 2021). Coordination equilibria do not flip because the incumbent financial hegemon behaves badly; they flip when a rival offers a deeper, safer, more open network. The euro’s path to co-anchor status by 2050 therefore runs through the supply of a genuine European safe asset (Blanchard and Ubide 2025, Dorrucci and Rossi 2026) and through capital markets that are open to the world rather than shielded from it. A Europe that fragments outwardly shrinks the very network externalities on which monetary status rests. Financial de-risking and monetary ambition are not complements; they are substitutes.

Supply chains: The reshoring mirage

At the firm level, the reshoring agenda is already failing on its own terms, quietly. Aggregate trade data show Chinese value added continuing to reach Western markets through connector economies rather than disappearing (Alfaro and Chor 2023, Freund et al. 2024). In recent work using firm-level customs and ownership data, my co-authors and I find that dependence is being re-labelled rather than reduced: the same upstream exposure is routed through third countries, frequently through affiliates of the same original firms, at higher cost and with less transparency (Buckberg et al. 2026). Measured self-sufficiency rises while actual network position deteriorates, because the efficiency cost of decoupling is paid without the insurance it was meant to purchase. The operational content of ‘open strategic autonomy’ should be the opposite: map the network at the firm level, identify genuine single points of failure, and diversify them by adding links – new suppliers, new ownership stakes, new partners – rather than deleting old ones.

The test case: Turkey

Every framework needs a case where it bites, and for Europe that case is Turkey. As NATO convenes in Ankara this summer, Europe is embracing Turkey in exactly one layer of the network – defence – and only under duress, while capping its participation in the EU’s SAFE instrument and leaving its economic anchoring untouched. Yet Turkey is a high-centrality node in at least four layers simultaneously: defence-industrial capacity, energy transit, migration management, and manufacturing supply chains bound to Europe through a customs union essentially unmodernised since 1996 – all this atop the EU’s longest-standing membership application, pending since 1987. Trading goods and security services with a node like Turkey while excluding it from economic and monetary union is decoupling logic in mirror image – and it produces the same result, a connection that is load-bearing but uninsured.

This is not a new argument, and the mechanism has already run once as a natural experiment. Turkey’s convergence in the first decade of this century coincided precisely with a live accession anchor: exports grew roughly 15% a year, inward FDI quadrupled, and more than half of European economic legislation came to be mirrored in Turkish law – before the European anchor lapsed after 2004–05, through fault on both sides, and reform effort decayed with it. A decade ago, Stuart Eizenstat and I argued that the security crisis of that moment – then Islamic State, not Russia – should trigger a comprehensive re-anchoring of Turkey: a modernised customs union extended to services, public procurement and agriculture; Turkish participation in the EU’s Energy Community; and a reinvigorated accession process alongside deeper NATO engagement (Eizenstat and Kalemli-Özcan 2015). Every element of that agenda remains pending a decade later, while the fault on Turkey’s side is getting bigger, and the cost is visible in Ankara’s hedging between blocs. Nodes that are excluded do not remain idle; they rewire, with unintended consequences.

The obvious objection is that Turkey’s economic troubles since 2015 were entirely homegrown – the product of unorthodox monetary policy and institutional erosion, not of European exclusion. That is true, and it is the mechanism rather than the refutation. Anchors operate on the domestic incentive regime: the reform era coincided with a live accession process because conditionality made good policy politically profitable at home, and the drift into bad policy followed the anchor’s collapse. Causality ran in both directions – deteriorating governance also made anchoring harder – but the policy-relevant asymmetry is that the anchor is the one lever Europe controls. And for Europe’s own security calculus, the assignment of blame is beside the point: an uninsured load-bearing connection is a risk to Europe whoever is responsible for its current condition.

The founding document of this initiative allows that integration may proceed at different speeds and depths. Turkey is the defining test of that variable geometry. Rule-of-law and democratic conditionality are boundary conditions, not bargaining chips. The alternative futures are stark. By 2050 the European core is surrounded either by deeply integrated partners with institutional depth, or by armed, capable, resentful middle powers that Europe trained itself to need and taught itself to exclude.

The Singapore test

This initiative asks where Europe should stand in thirty years, not where current trends lead it – and whether Europe can pass what its founding document calls the Singapore test: being secure and prosperous not despite openness but because of it. A Europe that ties its security to what it can produce alone will be smaller in every dimension that matters: poorer without specialisation, monetarily peripheral without network externalities, and less safe without insurance. A Europe that treats integration depth as the security variable – completing its internal capital market, supplying a safe asset to the world, diversifying rather than deleting its supply-chain links, and anchoring its neighbourhood through calibrated institutional depth – passes the test. The geometry of peace, power, and prosperity is a network. Europe’s position in it is a choice, and it is being made now.

Source : VOXeu