This ECB Occasional Paper (no. 208, April 2018) attempts to justify the necessity of introducing a pan-European deposit insurance scheme. This attempt fails completely, as was already the case in the European Commission’s paper “Effects on the European Deposit Insurance Scheme (October 2016)“. Both analyses completely neglect the fact that probabilities of indemnity events (especially in times of crisis and insecurity among depositors) are correlated and that there is therefore no diversification or pooling effect from an actuarial point of view.
The overall inference which has to be drawn is the following:
- Omitting cross-border contagion effects caused by a pan-European deposit insurance scheme leading to cross-border insecurity among depositors constitutes a grave methodological defect which leads, in turn, to misleading results.
- The notion that a pan-European deposit insurance scheme would benefit from a pooling or diversification effect is simply wrong.
I. Objective of the study
Using a model simulation, the study seeks to address the following questions:
- Will the target volume of the EDIS suffice to indemnify losses?
- How much of a burden is a risk-oriented mode of calculating contributions likely to impose on the banking systems in the individual countries?
- How will risk-oriented contributions be distributed among smaller, medium-sized and larger financial institutions?
- Will the EDIS entail systematic transfer payments between the banking systems in the various member states?
II. The study does not convincingly justify the need for a pan-European deposit insurance scheme
Claims made in the study:
“EDIS will play a key role in terms of confidence building, also avoiding risks of self-fulfilling prophe-cies on bank runs.“ (p. 5)
“Creating an EDIS is a logical step in completing the European Banking Union… the deposit insurance pillar is still missing. Ensuring a uniform protection of depositors across the entire banking union, regardless of geographic locations…” (p. 12)
Criticism: This justification does not carry conviction.
The case is argued and justified dogmatically rather than in economic terms. The argument that an EDIS is needed to ensure uniform protection of depositors is wrongly-conceived. The DGSD (Deposit Guarantee Scheme Directive) has already harmonised all claims by depositors; every Euro in deposits is protected right across the European Monetary Union to an identical extent and on the basis of uniform standards. What is more, the target level of the insurance schemes is identical.
To that extent, there is no reason to fear that the probability of a bank run would be higher or lower in a given country within the currently valid regulatory framework. Just how a pan-European scheme would aid the goal of confidence building on the part of depositors is not explained by the authors, merely asserted.
III. Methodology behind the simulation
The study is based on a sample of 1,635 euro area banks, thus representing 75% of total assets of credit institutions in the euro area and 83% of covered deposits in the euro area. As to the banks in the sample, data from Bankscope and COREP (Common Reporting) / FINREP (Financial Reporting) data was drawn on.
Criticism: The sample of credit institutions used is not sufficiently broad.
Although the sample may be broad enough in relation to the euro area as a whole, it is unsuited to the task of assessing possible transfers between states because the coverage ratios for the individual national banking markets diverge sharply. The authors themselves point to this problem in FN 22 and FN 2 but fail to draw any interpretative conclusions from this: “The degree of representativeness of the sample at country level is, however, heterogeneous“.
For example, the sample only covers approximately 24% of the banking market in Cyprus and only 34% of total assets in the Irish banking market. Yet both these countries would have been candidates for cross-border transfers during the financial-market crisis.
If “problem candidates“ – and therefore potential risks and the exposure of the EDIS – are so seriously under-represented, it is methodologically reprehensible to wish to draw inferences about possible transfer effects.
Probability of Default (PD)
For the credit institutions in the sample, a PD has been calculated on the basis of different bank-specific, aggregate banking sector and macro-financial variables. The calculation follows the methodology provided in Betz et al. (2014) and Lang et al. (2018)*.
Criticism: The authors do not provide a well-founded underpinning for the methodology employed to calculate bank-specific probabilities of default.
In addition, the methodology partly derives from an ECB Working Paper which has not yet been subjected to the usual scientific evaluation process.
Loss Given Default (LGD)
The European Commission has estimated that average losses for banks which became distressed between 2007 and 2010 amounted to 2.5% of total assets (maximum of 46.4%) while losses plus recapitalisation needs were on average 6% of total assets (maximum of 50.7%). The Financial Stability Board found that losses as a fraction of total assets mostly ranged from 2% to 4% for G-SIBs which had become distressed (maximum of 8.8%).
Following on from this, the study considers losses in resolution from 5% to 25% of total assets which would have to be borne by the deposit insurance scheme if the bank were to go into resolution. The simulation assumes that credit institutions with a balance-sheet total of more than EUR 20 billion or covered deposits of more than EUR 4 billion would be wound up.
Furthermore, the interval of loss ratios is defined in terms of loss contributions which have indeed materialised in the more recent past.
Results of the simulation
The study assumes that between 3% and 10% of the credit institutions with the highest PDs would default. An LGD of between 5% and 25% of total assets (losses in resolution) or of 7.5% to 37.5% of total assets (losses in liquidation) is applied. It emerges from the simulations that the EDIS would have to bear costs of EUR 30.4 billion (“EDIS exposure“) under a scenario in which 10% of the banks with the highest PDs failed, with the highest assumed loss ratios (25% loss in resolution and 37.5% loss in liquidation) applying.
Assuming (as the simulation does) that the pan-European deposit insurance fund had a volume of EUR 38 billion, this volume would therefore be enough to completely cover necessary payouts. The authors accordingly conclude that a target level for ex-ante contributions of 0.8% of covered deposits would be sufficient.
Criticism: The analysis omits the dynamic effects of a cross-border deposit insurance scheme leading to cross-border depositor insecurity
The methodology underlying the simulation is plausible. However, the simulation completely omits cross-border contagion effects deriving from growing depositor insecurity. The study therefore proceeds from a questionable ceteris paribus assumption with regard to depositor behaviour.
Especially in cases involving heavy costs for the EDIS, it is to be assumed that depositors, motivated by fear, would withdraw their deposits even from healthy credit institutions or else switch them into other asset classes (for example cash).
The insecurity among depositors spawned by the supposedly secure deposit insurance scheme would be particularly great because the backstop fund would be located at a “remote“ political level, and because the decision-makers involved could not be sanctioned by elections or, even if they could, only to a limited extent.
In the absence of national protection ceilings, there would be a danger of a pan-European deposit insurance scheme leading to depositor insecurity right across the euro area.
It is methodologically incorrect to assume that the EDIS would be viable on the basis of historical data which do not capture changes in behaviour and growing cross-border contagion effects rendering depositors more insecure, or to seek to appraise a possible inter-country transfer element through a comparison of contribution volumes for the various banking systems.
IV. Calibration of contributions
Contributions are calculated using the risk-adjusted DGS sliding scale methodology according to the EBA Guidelines. However, only selected risk categories and indicators are actually taken into account due to the fact that data is not always available. What is more, a bank’s risk profile is compared not to its peers within its national banking market but rather to its counterparts across the entire banking union.
Alternatively, the study calculates the aggregate risk weight (ARW) by using the SRF methodology. The SRF approach to determining the risk weight uses different indicators and a geometric aggregation method (the EBA Guidelines governing contributions to deposit insurance schemes prescribe a linear aggregation method only).
Criticism of the methodological approach underlying the simulation
A rudimentary approximation using few indicators; it is unclear whether they would be in line with the actual methodology used in an EDIS. The calculation of contributions, and especially their risk weight, is based on a very slender set of indicators; not even all the indicators named in the EBA Guidelines are covered. What is more, there are data gaps, the upshot being that the authors set individual indicators to zero on an ad hoc basis if data is not available.
Whether such a rudimentary set of risk indicators would be in line with the risk-adjusted benchmarks actually used in an EDIS is unclear. To that extent, the forecasts offered about possible systematic transfers between banking systems cannot be reliable.
The authors themselves concede these methodological weaknesses but then go on to defend their conclusions as though they were inviolable truths: “As these indicators are still under discussion, the set used here does not prejudge the final calculation method that will be decided by the Council of the EU and the European Parliament.“ (p. 25); “The construction of the composite risk indicator is a crucial topic in the calculation of risk-based contributions as contributions strongly depend on the choice and design of the various steps taken…“ (footnote 38).
Calculated distribution of contributions: Germany would contribute 33% (EUR 12.5 billion) and France 17% (EUR 6.6 billion) towards the aggregate volume of an EDIS. This assumed distribution of contributions would entail heavy cost burdens for the German banking market.
If calculations were based on covered deposits alone, the German banking sector would have to bear a shade over 26% of the overall volume (EUR 9.8 billion), with its French and Italian counterparts shouldering 25% (EUR 9.6 billion) and 10.5% (EUR 4 billion), respectively.
The aggregate risk weights (ARWs), based on the sliding scale methodology according to the EBA Guidelines, which are used in the study result in a considerable shift in contribution burdens, at the expense, for example, of Germany, Greece and Italy but in favour of France. On a risk-adjusted basis, the German banking sector would have to bear 33% of the overall cost (EUR 12.5 billion), France 17.5% (EUR 6.6 billion) and Italy 11.1% (EUR 4.2 billion).
In the light of this – at least on the basis of the methodology employed in the study – the German banking market appears to be highly risky, the Italian banking market marginally risky and the French banking market very low-risk.
The sharp reduction in the French contribution results, amongst other things, from the inclusion of the MREL indicator. The argument here runs as follows: the higher the MREL ratio fixed by the supervisory authority, the greater the likelihood of a bank going into resolution (because systemically relevant) and the smaller the likelihood of it causing exposure to the EDIS.
The authors of the study advocate including the MREL indicator in order to help to compensate France for the loss of its rebate (which, naturally, has to be paid for by the others). The authors prefer this variant to reducing the target level for individual member countries.
If risk weights are determined on the basis of the SRF methodology, the costs incurred by the German banking sector would remain at roughly the level which would result if the methodology were geared exclusively to covered deposits.
Conclusion: The exact method shaping the risk adjustments is the main driver for the distribution of contributions and this is political highly suggestible.
V. Distribution of contributions across “smaller“ and “larger“ banks
Were the risk-based contribution approach delineated above to be applied, the smallest ten percent of banks in the sample (balance-sheet total of up to EUR 0.15 billion) would have to pay 0.97 cents for each euro of covered deposits. By contrast, the largest ten percent of banks in the sample (balance-sheet total of EUR 6.6 billion or more) would have to fork out 0.83 cents for each euro of covered deposits.
Meanwhile, medium-sized banks would have to stump up 1.1 cents for each euro of covered deposits. This would mean that Germany’s Savings Banks would have to pay the highest average contributions per euro of covered deposits. In contrast large credit institutions would tend to pay lower contributions per euro of covered deposits.
The authors regard the fact that larger banks would tend to have a lower contribution burden as warranted because such credit institutions would be less likely to benefit from the EDIS, tending to go into resolution instead. Unfortunately, the authors do not mention the possibility of adding a mark-up to the bank levy in the case of large institutions, even though it would be only consistent to argue along such lines.
VI. Distribution effects between the member countries
As already clarified under II., the authors run a simulation for the costs incurred by the EDIS if the 3% (and, alternatively, the 10%) of banks with the highest PDs were to fail. In the first case, this would involve 51 euro area credit institutions; in the second case, 167 would be involved.
In this connection, the authors assume various loss rates which would have to be borne by the EDIS (up to 25% of total assets for losses in resolution and up to 37.5% of total assets for losses in insolvency). These costs per country are put in relation to the risk-based contributions by that country’s credit institutions. It emerges in this context that increasing loss rates to be borne by the EDIS result in redistribution effects to the advantage of Spain and Greece.
A scenario simulating country-specific shocks, with instabilities only occurring in individual countries, results in redistribution effects to the advantage of Belgium, Cyprus, Spain, Luxembourg and Malta. The results tally with the expectation that above all small member countries or those with banking systems fraught with risk would profit from the EDIS.
However, such redistribution only kicks in if the EDIS has to absorb loss rates of 15% (losses in resolution) and 22.5% (losses in insolvency) or higher. Given that redistribution only comes into play if loss rates are higher, the authors infer that the EDIS would not entail any systematic redistribution.
The statements about possible inter-country transfer elements are based on far-reaching assumptions. What is very problematic in this context is that a PD is ascertained for each individual credit institution. Precisely the financial-market crisis – and especially the case of Lehman Brothers – has demonstrated that there need not be any congruence between a theoretically unimpeachable PD and the situation on the ground when it comes to real-life cases of insolvency or resolution.
To that extent, the study considers a quite specific cost scenario and its spatial distribution. To wish to deduce generally valid conclusions about inter-country transfers from this is methodologically unsound. What is more, the analysis only provides a purely static snapshot; the situation in the future could look different.
Looking at Germany in particular, ascertaining the PD on a model basis neglects the special protection of an institution guarantee against insolvency. To that extent, potential claims on an EDIS by German credit institutions are over-estimated, and Germany’s position as net contributor is under-estimated.
In a footnote (FN 48), the authors do indeed concede that they have run into methodological problems:
“This methodology is used as a proxy for cross-subsidisation and is based on several assumptions, including those for the estimation of PDs and the calculation of risk-based contributions … Risk-based contributions are also based on point-in-time data. The effectiveness of the risk-based contributions as a tool to mitigate cross-subsidisation is therefore subject to the aforementioned limitations.“
It is curious that the authors admit the methodological weaknesses of their study, on the one hand, while formulating the conjectural conclusions of their analysis in an emphatically unambiguous manner, on the other. In view of the methodological weaknesses which the authors themselves admit to, the conclusion to be found on p.34 is not tenable: “… these findings suggest that there is no unwarranted systemic cross-subsidisation via EDIS…“. The underlying assumptions are too far-reaching for such an axiomatic inference to be drawn.
What is more, the simulations confirm – albeit only on the basis of higher loss rates to be borne by the EDIS – that it is above all small countries (for example, Belgium, Cyprus, Luxembourg or Malta) or else countries with banking systems fraught with risk (for example Spain or Greece) which would profit from the setting-up of an EDIS.
VII. Faulty inference: The ceteris paribus assumption about depositor behaviour is not valid
… the main benefit of an EDIS derives from reducing the sovereign-bank nexus as well as from pooling resources across Member States“. (p. 23)
For one thing, this statement is factually incorrect; for another thing, it proceeds from an inaccurate ceteris
An EDIS would not sever the link between sovereigns and banks in their jurisdictions for as long as there were still claims on states without large-credit limits. States which have difficulties placing debt instruments in the capital market will be inclined to coax them into their domestic banking system. If there are no regulatory volume limits, national budgetary risks will therefore get transferred to the local banking system.
If such risks materialise, for example within the framework of a haircut, the national banking system would get into trouble. The costs of rectifying the situation would be distributed via the EDIS across the whole of the euro area banking system.
In view of the fact that such a situation would put a severe strain on the Deposit Insurance Funds (DIF) capacity, the EDIS would logically require an (unspecified) lender of last resort. This, in turn, would give rise to fiscal burdens – on whichever national budgets. The state-bank nexus certainly does not only run from the banking system to the national sovereign, but in the other direction as well.
In contrast to classic insurance configurations, a deposit insurance scheme does not benefit from a diversification effect. The actuarial background here is that the risks to which a deposit insurance scheme is exposed – indemnity events – are not always stochastically independent from one another.
A positive correlation between the probabilities of indemnity events would result especially in the event of cross-border depositor insecurity spawned by claims depleting the financial resources of the cross-border deposit insurance fund.
Depositor insecurity would be particularly marked because the backstop fund would be located at a level of the (pan-European) political hierarchy far removed from the world of citizens / depositors.
To that extent, the further inference drawn by the authors is incorrect because the contagion effects arising from a pan-European deposit insurance scheme leading to depositor insecurity are left out of the equation: “A European Deposit Insurance Scheme would enhance depositor confidence and reduce the risk of wider deposit withdrawals which may also spill over to other banks“ (S. 23).
The authors themselves admit to these serious methodological shortcomings: “Spill-overs are not modelled in the analysis given the confidence enhancing role of an EDIS“ (S. 23); however, they do not bear them in mind when framing their conclusions although this is the cardinal weak point of a pan-European deposit insurance scheme.
The authors neglect to explain where the depositor confidence they think would be enhanced by a pan-European system should come from. General surveys do not indicate that citizens have significantly greater confidence in the capacity for action of decision-makers at the pan-European as opposed to the national level.
Ultimately, there would be a threat of high fiscal costs, incurred by the need to redeem depositors from the insecurity sweeping the euro area. Or else – and this is a more probable scenario – there would have to be an explicit deposit guarantee by states for their respective jurisdiction, which would leave Europe’s single financial market architecture permanently fragmented.
*Betz, F./S. Oprica/T. Peltonen/S. Sarlin 2014, “Predicting Distress in European Banks“, Journal of Banking & Finance, Vol. 45(C), 225-241
Lang, J./P. Sarlin/T. Peltonen 2018, “A Framework for Early-Warning Modelling with an Application to Banks“, ECB Working Papers, forthcoming