Countercyclical capital buffers (CCyBs) are a centrepiece of the post-crisis macroprudential toolkit. A growing literature finds that building capital in good times strengthens bank resilience without discouraging lending at longer horizons and can even support credit in downturns (Muñoz and Smets 2025). Yet a harder question remains: does the regulation reach the risks it is designed to contain, or does it merely shift them elsewhere?

Evidence of unintended spillovers is mounting. Auer and Ongena (2020) show that Switzerland’s sectoral CCyB generated compositional spillovers into commercial lending. Danisewicz et al. (2015, 2017) document that branches contract lending more sharply than subsidiaries when capital requirements tighten. Buch et al. (2016), summarising the International Banking Research Network, find that roughly one-third of specifications detect significant cross-border spillovers. Albuquerque et al. (2025) show that macroprudential tightening redirects corporate credit from banks to nonbanks — a substitution margin outside regulators’ direct control.

What connects these findings is that the leakage channels all run through financial institutions: bank structure, nonbank intermediaries, global banking networks (De Haas and Van Lelyveld 2011). We report evidence that a different, largely overlooked channel can fully neutralise a foreign CCyB at the firm level: the internal capital market of the multinational corporation itself. Consistent with the idea that multinationals actively reallocate funds across affiliates (Bilir et al. 2019a, 2019b), we find that when host-country regulators raise the CCyB, bank credit to the local subsidiary falls, but parent companies step in, borrowing domestically and on-lending to their affected subsidiaries. The CCyB travels back to the parent’s home jurisdiction.

Countercyclical capital buffers, reciprocity, and why internal capital markets matter

CCyBs raise banks’ capital requirements when credit growth in a jurisdiction is strong. Under mandatory reciprocity in Basel III, both domestic and foreign banks must apply the activating country’s buffer rate to their exposures there. The immediate channel is straightforward: higher capital requirements raise the marginal cost of lending, so bank credit to borrowers in the activating country should fall.

What CCyBs do not cover is intra-group financing within multinational firms. A parent is free to lend to its foreign subsidiary, regardless of the host-country buffer. If the group’s internal capital market is sufficiently deep, the parent can simply replace the bank credit the CCyB has withdrawn. The policy then reduces local bank credit without reducing the subsidiary’s overall funding: the risk originally located in the host country does not disappear — it is reallocated.

A clean empirical setting

We exploit a clean quasi-experimental setting. Between 2013 and 2019, Germany did not activate a CCyB, while many host countries of German-owned subsidiaries did. This asymmetry lets us isolate how German multinationals and their lenders respond to foreign CCyB activations, free of confounding domestic buffer changes.

We combine three granular German data sources: the credit register (bank and non-bank credit), firm-level ownership data linking German parents to their foreign subsidiaries, and balance-sheet information on both sides of the group. This lets us track, loan by loan, how lending, internal debt flows, leverage, and default risk respond to foreign CCyB increases — at both the subsidiary and the parent.

Countercyclical capital buffers work at the bank–firm level…

Our first result confirms that the CCyB works where it is implemented. When a subsidiary’s host country raises its CCyB by one percentage point, lending by German banks to that subsidiary falls by roughly 10%, on both the intensive and extensive margins: existing loans shrink and bank–firm relationships are more likely to be terminated.

Critically, non-bank lending — which is not subject to CCyB regulation — does not respond, consistent with Albuquerque et al. (2025). This confirms a regulatory effect, not a general repricing of credit risk.

At face value, the policy appears to work.

…but internal capital markets fully offset the shock

The central result is that this contraction does not reduce subsidiaries’ overall funding. Subsidiaries replace the lost bank credit one-for-one by borrowing more from their German parents. A one-percentage-point CCyB increase raises internal parent debt by 2.3 percentage points of subsidiary total liabilities and 1.2 percentage points of total assets — roughly one-third of the average internal debt level.

Three features stand out:

• The adjustment runs entirely through debt, not equity. Parents do not recapitalise subsidiaries; they lend to them.
• The substitution is effectively complete. Total liabilities, leverage, and default probabilities of affected subsidiaries are unchanged.
• From the subsidiary’s perspective, the CCyB is fully neutralised.

This is the internal-capital-market mechanism of Bilir et al. (2019a, 2019b), applied to a macroprudential shock: the group reallocates funds internally to preserve the affiliate’s financing plan, insulating it from host-country regulation.

Parents borrow more at home — and their risk rises

Internal support must itself be funded. How do parents finance it? Two facts stand out.

First, affected parents borrow more externally in Germany. Bank borrowing rises by about 4.1%; non-bank borrowing by around 15%. Parents do not redirect funds from other subsidiaries — they tap domestic markets.

Second, this additional leverage raises risk. The probability of default of affected parents rises by around ten basis points — roughly 25% of the average parent probability of default. The stability gained at the subsidiary is, in effect, paid for by higher leverage and default risk at the parent.

At the group level, the picture changes qualitatively. Although bank credit to subsidiaries abroad contracts, total lending by German banks and non-banks to the group increases. German lenders’ exposure to these multinationals rises, and the exposure-weighted default risk of the group goes up.

Summarising the mechanism

Figure 1 summarises the mechanism. A one-percentage-point CCyB increase cuts cross-border bank lending to the affected subsidiary by 10.6%, while non-bank credit is unchanged. German parents fully offset this via internal lending, refinanced through 4.1% more domestic bank and 15% more non-bank credit. Other subsidiaries are not drawn upon; subsidiary leverage is unchanged; parent and group risk rise.

Figure 1 Summary of the results

 

Notes: Foreign CCyB activation, the multinational group, and cross-border risk migration. Green arrows: increases in credit. Red arrow: contraction in bank credit to the affected subsidiary. Grey arrows: no significant response. Source: Imbierowicz et al. (2026), Bundesbank Discussion Paper 12/2026, Figure 1.

Why do parents over-borrow? Precautionary insurance of internal capital markets

Why don’t parents simply match the subsidiary-level contraction, rather than raising group-level external borrowing? Our evidence is consistent with precautionary over-borrowing: CCyB activations typically signal a broader, persistent tightening abroad. Facing the prospect of further constraints, parents borrow defensively at home to insure their internal capital market.

Because domestic and foreign bank credit are imperfect substitutes — each embedded in its own relationships and covenants — domestic borrowing becomes the natural buffer. The behaviour is individually rational but generates a systemic externality: risk migrates toward jurisdictions that did not themselves tighten, even under full reciprocity on the banking side.

Reinterpreting the cross-border spillover literature

Our findings speak to, and extend, the literature on macroprudential spillovers in two ways.

First, prior work has mostly studied bank-driven leakages: branches versus subsidiaries (Danisewicz et al. 2015, 2017), regulatory arbitrage across organisational forms, and bank-to-nonbank substitution (Albuquerque et al. 2025). We show that a firm-driven channel exists alongside these and can fully neutralise a bank-side policy even when reciprocity rules out bank-side arbitrage.

Second, existing evidence of CCyB effects on bank lending (Auer and Ongena 2020, Muñoz and Smets 2025) is consistent with our subsidiary-level result — but the net effect on the borrower, and on the distribution of risk, can be qualitatively different. Reading effectiveness from the bank–firm loan-level response alone overstates the policy’s reach.

Policy implications

Three implications follow:

1. National macroprudential tools have firm-level limits in integrated economies. When multinational groups have deep internal capital markets, national CCyBs can be partially or fully neutralised at the borrower level, even under full reciprocity.
2. Risk does not disappear; it relocates. In our setting, tighter CCyBs abroad increase risk in the parent’s home jurisdiction. Monitoring frameworks focused only on domestic borrowers or bank–firm lending will miss this reallocation.
3. Effective macroprudential policy in integrated economies requires broader coordination. At a minimum, this means sharing information on multinational groups’ consolidated exposures across home and host authorities, and considering intra-group debt flows in the design of macroprudential instruments.

None of this argues against CCyBs. The policy works where it is designed to work — at the bank–firm loan level in the activating country. But in an integrated corporate world, firms’ internal capital markets can carry the policy’s effects across borders just as readily as banks’ (De Haas and Van Lelyveld 2011). Recognising this firm-driven leakage channel is an important next step for the macroprudential agenda.

Concluding remarks

Countercyclical capital buffers make banking systems safer at the bank–firm level. But multinational firms respond. Parents borrow more at home, lend more to affected subsidiaries, and leave the subsidiaries’ funding, leverage, and default risk unchanged — while raising their own. Under reciprocity, the regulation binds German banks’ exposure to the host country, but the group’s internal capital market reallocates both funding and risk back to the parent’s home jurisdiction.

The result is not that CCyBs fail. It is that they travel. Where risk ends up depends on the structure of the firms the policy acts upon, not only on the structure of the banks it directly regulates.

References

Albuquerque, B, E M Cerutti, N Chen and M Firat (2025), “From banks to nonbanks: Macroprudential and monetary policy effects on corporate lending”, VoxEU.org, 16 June.

Auer, R and S Ongena (2020), “Spillovers and the optimal design of macroprudential measures: Evidence from Switzerland”, VoxEU.org, 14 January.

Bilir, L K, D Chor and K Manova (2019a), “Host-country financial development and multinational activity”, VoxEU.org, 6 May.

Bilir, L K, D Chor and K Manova (2019b), “Host-country financial development and multinational activity”, European Economic Review 115: 192–220.

Buch, C M, M Bussière and L S Goldberg (2016), “Prudential policies crossing borders: Evidence from the International Banking Research Network”, VoxEU.org, 9 December.

Danisewicz, P, D Reinhardt and R Sowerbutts (2015), “Macroprudential spillovers: The role of organisational structure”, VoxEU.org, 5 March.

Danisewicz, P, D Reinhardt and R Sowerbutts (2017), “On a tight leash: Does bank organizational structure matter for macroprudential spillovers?”, Journal of International Economics 109: 174–194.

De Haas, R and I Van Lelyveld (2011), “How did multinational banks weather the previous perfect storm?”, VoxEU.org, 14 December.

Imbierowicz, B, A Loeffler, S Ongena and U Vogel (2026), “How CCyBs travel — Internal capital markets & domestic borrowing”, Deutsche Bundesbank Discussion Paper No. 12/2026 and CEPR Discussion Paper DP21413.

Muñoz, M A and F Smets (2025), “Macroprudential policy for banks: Build the countercyclical capital buffer when there is headroom for doing so”, VoxEU.org, 25 July.