VoxEU Column Europe's nations and regions Global crisis International Finance

Coordinating bank-failure costs and financial stability

The dissent brewing throughout Europe hinges on the question of whether the financial burdens of the Eurozone crisis should be shared between weak and strong. This column presents a new paper arguing that the wealthier, more stable economies don’t have much choice.

During the financial crisis, failure or distress of cross-border firms has been met by ad hoc coordinated solutions (eg Fortis and Dexia) or national solutions (eg UK and US banks). However, economic theory (such as Freixas 2003) shows that ‘improvised coordination’ is inefficient as it leads to a general under-provision of the public good (ie financial stability). The European Financial Stability Facility (EFSF) for the Eurozone (EZ) constitutes the first example instead of an ex ante burden-sharing agreement.

But little research has been done on sharing mechanisms, and the possible effects on financial stability. An exception is Goodhart and Schoenmaker (2009), who estimate each country’s contributions for a general fund mechanism and specific burden-sharing mechanism. The authors prefer a specific burden-sharing mechanism, as it is “closer to an efficient solution of the coordination problem”. However, existing research fails to answer two important questions that are particularly important to policymakers:

  • What mechanism should we choose according to a determined criterion? And
  • Are these mechanisms bringing financial-stability benefits?

In a recent paper (van Lelyveld and Spaltro 2011) we attempt to answer these two questions by constructing a detailed dataset of major international cross-border banks, which allows us to estimate the expected losses for banking defaults in every sovereign. Moreover, we try to give a preliminary understanding of the financial-stability benefits of burden-sharing mechanism by using a Monte Carlo simulation.

We have constructed banks’ balance sheets accounting for mergers and acquisition by adding all the subsidiaries that constitute the group in a given year. In this way we are able to map banks’ activities across jurisdictions in detail. Using these data we have estimated the contingent liabilities by sovereign for the ‘national solution’, ie when a sovereign recapitalises the banking group(s) across all jurisdictions.

We have also done the same exercise for various burden-sharing agreements. In this way we can compare costs with and without burden-sharing agreements and single out winners and losers from the establishment of burden-sharing agreements. In addition, we have extracted probability of default from banks’ CDS premia to estimate the expected cost of default by sovereign and for different burden-sharing agreements. Here we aim to estimate the expected loss for each sovereign given the cross-border exposures of its banking system.

In the national solution world, the US would have to inject around $250 billion if all its major cross-border banks needed a capital injection, followed by the UK with around $235 billion.1 France and Germany also have significant potential costs in absolute terms, even though more limited than the UK as a proportion of GDP. These numbers ignore sharing and (market) default expectations.

Figure 1. Total exposure with the national solution, 2008

With a general fund we see that, regardless of which banking group is involved, a country’s share of the total cost does not vary. For example, Figure 2 shows that in our sample Germany will have to contribute 8.1% of the total cost to recapitalise a bank like Allied Irish Banks (AIB) even though AIB does not have any significant activity in Germany. In such a mechanism, every country will be called to contribute when a public capital injection or liquidity provision is needed.2

Figure 2. General fund - population key, 2008

With a specific burden-sharing agreement the total recapitalisation costs are shared according to the activities of the individual bank in the various countries. In our analysis we present estimates using the asset key but other keys are conceivable. If these agreements were established, the US would have the largest total exposure (around $280 billion), while the UK would have the second largest exposure with around $250 billion. It is important to note that that these figures are independent of where these losses are incurred. For example, the UK would have to participate in the group’s recapitalisation of Bank of America Merrill Lynch even if losses materialised entirely in the US. And small nominal exposures may translate into high relative burden on a country in terms of GDP.

Whereas Figure 1 and Figure 2 showed the total contingent liability for different burden-sharing agreements, they do not give us an idea of the expected cost involved by each sovereign. Figure 3 weights the total exposures in the specific burden-sharing agreement according to bank riskiness using CDS premia. It shows how much each country would be expected to pay for ‘insurance’ in any given year, broken down by bank. The US still has the largest outlay with around $4.5 billion, while the UK should be expected to pay slightly less than this amount.

Figure 3. Risk-weighted exposure with specific burden-sharing – total assets key, 2008

We have then compared costs for various sovereigns with the national solution and burden-sharing mechanisms. Figure 4 shows that the Netherlands, France, Switzerland, Belgium, and the UK would gain with the establishment of a general burden-sharing mechanism as they are net exporters of banking services (they have large banks with foreign operations whose rescue capital will be provided by other sovereigns as well).

Comparing the differences across different types of burden-sharing agreements, our data suggests that the reallocation of contingent liabilities under the general fund mechanism is more ‘unfair’. This is because the standard deviation of the difference between the national solution and a burden-sharing agreement is much larger under the general fund than specific burden-sharing agreements. Reallocating costs according to assets would lead to the fairest redistribution of costs.

Figure 4. A comparison between the national solution and a general Fund (GDP), 2008

A burden-sharing mechanism can also be seen as an insurance scheme against bad states of the world. Countries would pay a relatively small premium every year to insure against defaults in their jurisdictions. To illustrate the general fund mechanism insurance property we perform a Monte Carlo simulation using 2008 default expectations. This provides us with a state of the world where generally none, sometimes some, and occasionally many banks fail. We then compute the cost for each country according to the national solution and to different cost sharing rules.


The key result of this analysis is that outlays under both schemes can be sizeable, but the right tail (ie the national solution being more costly than the general fund) is much longer than the negative tail.3 This implies that, under the national solution, some countries might have to pay extremely large amounts to support their banking systems. When these outlays are also large compared to a country’s GDP this can cause widespread financial instability, as the sovereign may not be able to support the ailing financial institution. These results support the findings of economic theory that sharing the burden should bring about financial-stability benefits.

Editor’s Note: The views expressed in this article are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Bank of England or that of the De Nederlandsche Bank.


Freixas, X (2003), “Financial Supervision in Europe,” in Crisis Management in Europe, Kremers, J, D Schoenmaker, and P Wierts (eds), Cheltenham: Edward Elgar.

Goodhart, C, and D Schoenmaker (2009), “Fiscal Burden Sharing in Cross-Border Banking Crises,” International Journal of Central Banking 5: 141–165.

van Lelyveld, Iman, and Marco Spaltro (2011), “Coordinating Bank Failure Costs and Financial Stability”, DNB Working Paper No. 306.


1 For our estimates we assume that in case of crisis sovereign fiscal authorities have to inject equity capital equal to 2% of total assets in line with recent experience.

2 For the EFSF the ECB key, an average of GDP and population shares, is used.

3 We assume that defaults are independent and that the existence of a sharing rule does not affect the behaviour of the stakeholders involved (moral hazard). 


210 Reads