Capital mobility has been legally guaranteed within the EU since the 1993 Maastricht Treaty. Yet de facto integration remains strikingly limited. Using AnaCredit, we document that cross-border loans to firms account for only 1% of all euro area loans and for just 5.6% of total euro area lending since 2019. This is the smallest cross-border share across all major intermediaries and asset classes, well below the foreign shares for equities.

Figure 1 makes the fragmentation visible. It plots the full matrix of bilateral lending positions (in orders of magnitude, log₁₀), with lending (creditor) countries on the vertical axis and borrowing (firm) countries on the horizontal axis for Q3 of 2024. The darkest cells lie squarely on the diagonal: in other words, lending overwhelmingly stays within national borders. The off-diagonal terms are not only lighter but frequently empty – many country pairs have no cross-border lending relationship at all (Estonian banks, for instance, did not lend to firms outside the Baltics and Finland at that time). The takeaway is one of a core-periphery network with pervasive home bias rather than an integrated single market.

Figure 1 Bilateral lending matrix

The same story holds for the decision of banking groups to open branches and subsidiaries abroad. Using group ownership and control information to link branches and subsidiaries to their parent groups, we find very few banking groups operate across borders. Most off-diagonal cells are empty.

Measuring the frictions using micro-data

To move from description to measurement, we develop what we call ‘micro-econometric wedge accounting’. The idea is to recover bilateral, country-pair frictions directly from bank-firm data and regressions, while purging the confounding effects of borrower risk and bank characteristics. Exploiting within-firm and within-bank variation – essentially asking whether a given firm is less likely to match with a foreign than a domestic bank, and vice versa – we estimate non-parametric country-pair wedges along three margins: (i) relationship formation, (ii) loan pricing and quantities conditional on a relationship, and (iii) bank entry into foreign markets.

The headline findings are stark. Barriers operate overwhelmingly on the extensive margin. Cross-border bank-firm links are orders of magnitude less likely than domestic ones, implying very large relationship-formation wedges (an unweighted mean of 65 percentage points relative to domestic links). Entry barriers are larger still: across specifications, foreign entry carries an implicit ‘tax’ of roughly 99 percentage points relative to domestic presence. By contrast, the intensive margin is benign – conditional on a relationship existing, cross-border loans carry interest rates only about 28 basis points lower (8.5% of the sample mean) and modestly different quantities. Where European banking is fragmented, it is fragmented at the door, not at the price.

A new dataset: Regulatory distances

Why are those doors so hard to open? Policymakers have long pointed to regulatory heterogeneity (Enria 2021, Buch 2024, Schnabel 2024). To test this directly, we construct a novel dataset of bilateral policy distances across eight regulatory domains – supervision, micro-prudential, macroprudential, resolution, entry, governance, deposit insurance, and corporate bankruptcy – harmonising dispersed information from the European Banking Authority (EBA), European Systemic Risk Board (ESRB), International Association of Deposit Insurers (IADI), the World Bank, and the OECD. Using Gower (1971) distance metrics, we summarise each country pair by the share of dimensions on which they differ.

Figure 2 shows the result. The average bilateral policy distance is 25%: a typical pair of euro area countries still differs on roughly one in four regulatory variables, despite a decade of banking union reforms – the Single Supervisory Mechanism, the Single Resolution Mechanism, and partial harmonisation of deposit insurance. Fragmentation persists in the corners the Single Rulebook leaves open: Options and Discretions, national resolution and bankruptcy regimes, governance rules, and incomplete deposit-insurance integration.

Figure 2 Average policy distance by country pair

Crucially, policy distances are correlated with the estimated wedges. Controlling for standard gravity forces (geography, language, culture, shared borders, and legal origin), regulatory distance correlates strongly with the extensive-margin wedges – relationship formation and bank entry – with the largest roles for macroprudential rules, deposit insurance, and bankruptcy regimes. These results suggest that regulatory differences across countries raise the compliance costs, informational acquisition costs, and legal uncertainty of cross-border activity.

How much output is at stake?

To quantify the macroeconomic costs of these barriers, we embed the estimated wedges in a quantitative spatial general-equilibrium model of the euro area with heterogeneous banks and firms. Households supply deposits, firms borrow to finance capital, and banks intermediate funds while choosing where to open branches and subsidiaries and at what rates to lend, subject to cross-border frictions in relationship formation, pricing, and entry. We invert the model using AnaCredit micro-data to recover firm productivity and default risk, bank-firm demand and pricing shifters, and bilateral entry costs.

Our benchmark counterfactual cuts cross-border relationship wedges by 10% across all country pairs. Euro area GDP rises by 1.6%. The surprise lies in the composition: we find roughly 96% of the gain comes from factor accumulation – capital and labour increase by 2.4% and 1.6%, respectively – and only about 4% from improved allocative efficiency, with total factor productivity (TFP) rising less than 0.1%. The mechanism is intuitive once stated: broader access to foreign intermediaries raises the cross-border credit share by 1.2 percentage points, lowering firms’ effective user cost of capital and inducing them to invest and hire more. This level effect operating through the cost of capital trumps the reallocation effect operating through who gets the credit and where capital flows.

The gains are highly uneven (Figure 3) and unrelated to initial income. Financial centres – Ireland, Luxembourg, the Netherlands, and to a degree Belgium – gain most, alongside small economies such as Malta, Slovenia, and Estonia that draw funds from larger neighbours. Spain, Greece, Portugal, and Italy gain least. Reassuringly, cross-country inequality narrows somewhat: the variance of output per hour worked falls by 2.7%.

Figure 3 Output gains by country

Conclusion and key takeaways

Broadly speaking, three messages follow for the debate on completing the banking union and building a Savings and Investment Union.

First, the binding constraint is the extensive margin – i.e. forming relationships and entering markets – not prices or quantities on existing loans. Policy aimed at compressing markups or fine-tuning loan pricing potentially targets the wrong friction.

Second, these barriers correlate with regulatory differences. Whilst these correlations do not establish watertight causation, they are highly suggestive. Differences in macroprudential frameworks, deposit insurance, and bankruptcy regimes are first-order impediments to integration. Harmonisation is where the leverage lies.

Third, completing the banking union can deliver meaningful output gains. The gains are real and arrive through a different channel than conventional wisdom assumes: a broad-based decline in the cost of capital that lifts investment and employment everywhere, rather than a sharp reallocation of credit towards the most productive uses.

Authors’ note: The views expressed in the paper are those of the authors and do not necessarily represent the views of the ECB or Eurosystem, the IMF, the IMF’s Executive Board, or the IMF’s Management. Krüger acted as a consultant to DG Research at the ECB during the work on this project.

References

Buch, C (2024), “Financial integration in Europe: Where do we stand after the banking union’s first decade?”, speech by Claudia Buch, Chair of the Supervisory Board of the ECB.

Capelle, D, A Fernandes, J J Kruger and P McAdam (2026), “Barriers to a European Banking Union”, IMF Working Paper 123 (also available as ECB Working Paper 3249). 

Draghi, M (2024), The Future of European Competitiveness, European Commission.

Enria, A (2021), “How can we make the most of an incomplete banking union?”, Speech by Andrea Enria, Chair of the Supervisory Board of the ECB.

Gower, J C (1971), “A general coefficient of similarity and some of its properties”, Biometrics 27(4): 857–871.

Schnabel, I (2024), “From laggard to leader? closing the euro area’s technology gap”, inaugural lecture of the EMU Lab by Isabel Schnabel, Member of the Executive Board of the ECB, at the European University Institute.