Occasional Paper Series
Completing the Banking Union with a
European Deposit Insurance Scheme:
who is afraid of cross-subsidisation?
Jacopo Carmassi, Sonja Dobkowitz,
Johanne
Evrard, Laura Parisi, André Silva,
Michael Wedow
Disclaimer: This paper should not be reported as representing the views of the European Central Bank
(ECB).
The views expressed are those of the authors and do not necessarily reflect those of the ECB.
No 208 / April 2018
ECB Occasional Paper Series No 208 / April 2018
1
Contents
Abstract 2
Non-technical summary 3
1 Introduction 6
2 Deposit insurance: rationale, objectives and challenges 8
3 A European Deposit Insurance Scheme: key features and the way
forward 10
Box 1 Comparison between EDIS and the US FDIC 15
4 EDIS exposure 16
4.1 Data on covered deposits 16
4.2 Methodology 17
4.3 Main results 21
5 Risk-based contributions to EDIS 24
5.1 Rationale 24
5.2 Methodology 24
5.3 Contributions 26
Box 2 Aggregate risk weight construction – DGS vs. SRF
methodology 28
5.4 Distribution of contributions to the DIF across banks 31
5.5 Cross-subsidisation 32
6 A mixed deposit insurance system 42
6.1 Contributions 42
6.2 Cross-subsidisation 44
7 Conclusions 48
References 51
Appendix 54
Abbreviations 55
Acknowledgements 56
ECB Occasional Paper Series No 208 / April 2018
2
Abstract
On 24 November 2015, the European Commission published a proposal to establish
a European Deposit Insurance Scheme (EDIS). The proposal provides for the
creation of a Deposit Insurance Fund (DIF) with a target size of 0.8% of covered
deposits in the euro area and the progressive mutualisation of its resources until a
fully-fledged scheme is introduced by 2024. This paper investigates the potential
impact and appropriateness of several features of EDIS in the steady state. The
main findings are the following: first, a fully-funded DIF would be sufficient to cover
payouts even in a severe banking crisis. Second, risk-based contributions can and
should internalise specificities of banks and banking systems. This would tackle
moral hazard and facilitate moving forward with risk sharing measures towards the
completion of the Banking Union in parallel with risk reduction measures; this
approach would also be preferable to lowering the target level of the DIF to take into
account banking system specificities. Third, smaller and larger banks would not
excessively contribute to EDIS relative to the amount of covered deposits in their
balance sheet. Fourth, there would be no unwarranted systematic
cross-subsidisation within EDIS in the sense of some banking systems
systematically contributing less than they would benefit from the DIF. This result
holds also when country-specific shocks are simulated. Fifth, under a mixed deposit
insurance scheme composed of national deposit insurance funds bearing the first
burden and a European deposit insurance fund intervening only afterwards,
cross-subsidisation would increase relative to a fully-fledged EDIS.
The key drivers behind these results are: i) a significant risk-reduction in the banking
system and increase in banks' loss-absorbing capacity in the aftermath of the global
financial crisis; ii) a super priority for covered deposits, further contributing to protect
EDIS; iii) an appropriate design of risk-based contributions, benchmarked at the euro
area level, following a "polluter-pays" approach.
Keywords: European Deposit Insurance Scheme (EDIS), risk-based contributions,
cross-subsidisation.
JEL codes: G21, G28.
ECB Occasional Paper Series No 208 / April 2018
3
Non-technical summary
This paper provides five analytical contributions to the discussion on the
establishment of a European Deposit Insurance Scheme (EDIS). First, the exposure
of a fully mutualised EDIS to bank failures
1
is estimated, examining how the
European Deposit Insurance Fund (DIF), with a target size of 0.8% of covered
deposits of participating banking systems, would be affected under different stress
and bail-in scenarios as well as under different methodological assumptions.
Second, the paper provides a quantitative analysis of how the calibration of deposit
insurance risk-based contributions (based on current banks’ risk profiles) affects the
distribution of contributions across countries and banks. Third, the paper investigates
how the collection of contributions would be spread across small, medium and large
banks. Fourth, the analysis aims to verify whether EDIS would produce any
systematic cross-subsidisation between banking sectors in different Member States,
also taking into account potential country-specific shocks. Finally, a mixed deposit
insurance scheme, with national funds bearing the first burden and a European fund
intervening only afterwards, is tested to investigate its potential implications for
contributions and cross-subsidisation.
This paper first focuses on a fully-fledged EDIS with a target level of 0.8% of covered
deposits of the participating banking systems. The analysis is based on Bankscope
data as well as supervisory data for 2015:Q4 on covered deposits and balance sheet
indicators to estimate EDIS exposure to bank failures and contributions to EDIS at
bank level. The conclusions of the analysis are therefore based on the assumption
that banks’ balance sheet structures remain the same until EDIS has been fully
introduced. The sample scrutinised comprises 1,675 euro area banks with total
assets of €22.14 trillion, representing approximately 75% of total assets of credit
institutions in the euro area, and €4,744 billion of covered deposits, corresponding to
approximately 83% of covered deposits in the euro area. The sample can be
considered as representative at the euro area level, both in terms of total assets and
covered deposits.
2
The target size of the DIF for the sample is approximately
€38 billion.
The exposure of EDIS is measured for banking crises of a different magnitude,
where the riskiest 3% or 10% of banks fail simultaneously according to their
estimated probabilities of default, in combination with different magnitudes of loss
severity
3
, ranging from 5% to 25% of total assets in resolution and between 7.5%
and 37.5% in insolvency, and two variations of banks’ loss-absorbing capacity. The
results indicate that a fully-funded DIF would be sufficient to cover payouts even in
very severe crises - even more severe than the 2007-2009 global financial crisis. It
should be stressed that the loss scenarios used for the analysis are extremely
1
The term “EDIS exposure” refers to the potential need for an EDIS intervention in case of a bank
failure.
2
The degree of representativeness of the sample at country level is, however, heterogeneous.
3
Losses in insolvency are assumed to be always 50% higher than losses in resolution.
ECB Occasional Paper Series No 208 / April 2018
4
conservative. First, the loss ranges considered are extremely conservative, as the
lower loss rate of 5% total assets, for example, is higher than the upper bound of
4.7% reported by the FSB for G-SIBs during the last crisis, and twice the average
estimates reported by the European Commission for the period 2007-2010. Second,
the high loss rates tested in this study are applied simultaneously to all the banks
assumed to fail.
The comparison of banks' risk-based contributions based on different indicator sets
suggests that specificities of a banking system can be taken into account in the
risk-based contributions to the DIF: this approach would allow to take into account
country-specific features, e.g. the likelihood of a significant share of banks within a
banking system to be subject to resolution instead of insolvency or the degree of
bank interconnectedness. This is preferable to a lowering of the EDIS target level as
is currently allowed under the Deposit Guarantee Scheme Directive (DGSD) since it
maintains the Fund's level of resilience and also preserves the level playing field. In
addition, adjusting the indicators in the risk-based contributions would make it
possible to take into account the relative riskiness of the banking sectors while the
DIF is being built up and while risk reduction measures are implemented.
4
The
analysis also finds that contributions depend on the calculation method used. Most
importantly, rescaling of the aggregate risk scores leads to a change in the
distribution across banks and countries in favour of the most risky banks.
Furthermore, a comparison of contributions that would be paid by banks of different
size reveals that smaller and larger banks would not excessively contribute to EDIS
relative to the amount of covered deposits in their balance sheet, suggesting that
measures to reduce contributions for the smallest and/or largest banks would be
unwarranted.
Banks’ contributions to EDIS are estimated and compared to the EDIS exposures
obtained in the previous part of the paper. This comparison aims to identify possible
unwarranted cross-subsidisation across euro area countries. Given the very high
loss rates necessary to produce cross-subsidisation, which would be considerably
higher than those experienced in the last global financial crisis, the findings suggest
no unwarranted systematic cross-subsidisation via EDIS, in the sense of some
banking systems systematically contributing less than they would benefit from the
DIF. This result holds also when country-specific shocks are simulated. In general, it
should be noted that cross-subsidisation can be seen as a form of desirable
risk-sharing in more severe crises. This is different from a systematic unwarranted
cross-subsidisation and is in line with the purpose of an EDIS: pooling resources to
enhance the ability of the deposit insurance system, particularly in the more severe
crises, to withstand shocks better than under a system of national stand-alone
DGSs.
Finally, the analysis shows that under a mixed deposit insurance scheme, composed
of national funds bearing the first burden and a European fund intervening only
4
In this context, see the 2015 European Commission Communication towards the completion of banking
Union (European Commission, 2015) and the 2017 Commission Communication on completing the
Banking Union (European Commission, 2017).
ECB Occasional Paper Series No 208 / April 2018
5
afterwards, each with a target level of 0.4% of covered deposits, cross-subsidisation
would increase relative to a fully-fledged EDIS. The reason is that the contributions
paid to the national deposit insurance fund would be fixed at 0.4% of covered
deposits in a Member State and would not be risk-based relative to the euro area
banking system.
In conclusion, EDIS would offer major benefits in terms of depositor protection while
posing limited risks in terms of EDIS exposure, since the probability and magnitude
of interventions are likely to be low. EDIS will play a key role in terms of confidence
building, also avoiding risks of self-fulfilling prophecies on bank runs. Additionally,
based on the results shown in this paper, there is no risk of unwarranted systematic
cross-subsidisation.
The key drivers behind these results are the following: first, a significant
risk-reduction in the banking system and increase in loss-absorbing capacity have
taken place in the aftermath of the global financial crisis; second, a super priority for
covered deposits will further contribute to protect EDIS; third, following a "polluter-
pays" approach, appropriately-designed risk-based contributions, benchmarked at
the euro area level, are crucial to establish the right incentives and strike the right
balance between ensuring adequate and credible deposit protection and minimising
cross-subsidisation across countries. Risk-based contributions can and should
internalise the specificities of banks and banking systems. This would address moral
hazard and facilitate moving forward with risk-sharing measures in parallel with risk-
reduction measures; this approach would also be preferable to lowering the target
level of the DIF to take into account banking system specificities.
ECB Occasional Paper Series No 208 / April 2018
6
1 Introduction
On 24 November 2015, the European Commission published a proposal for a
European Deposit Insurance Scheme (EDIS). The appropriateness of the Deposit
Insurance Fund's (DIF) target size – in the proposal, 0.8% of covered deposits, as
already envisaged for national deposit guarantee schemes – is one crucial aspect of
the design of the scheme. In addition, the distribution of risk-based contributions
across different types of banks and the possibility of a reduction of contributions for
specific types of banks is a subject of discussion. Some of the concerns raised upon
the publication of the proposal were that the EDIS would lead to unwarranted
cross-subsidisation, i.e. banking sectors in one Member State would have to pay for
bank failures in other Member States.
5
This paper investigates the validity of these concerns by first assessing the
exposure
6
of a fully mutualised EDIS with a target level of 0.8% of covered deposits
of the participating banking systems to bank failures under different stress and bail-in
scenarios. Second, the paper provides a quantitative analysis of how the calibration
of risk-based contributions (based on current banks’ risk profiles) affects the
distribution of contributions across countries and banks. Third, we show how the
collection of contributions would be distributed across small, medium and large
banks. Fourth, the analysis aims to investigate whether EDIS would produce any
systematic cross-subsidisation between banking sectors in different Member States,
also taking into account hypothetical country shocks. Finally, the paper compares the
results on contributions and cross-subsidisation under a fully mutualised EDIS with a
"mixed" deposit insurance scheme, composed of national compartments with a
target level of 0.4% of domestic covered deposits - intervening first - and a European
compartment with a target level of 0.4% of euro area covered deposits - intervening
only after the national compartment is depleted.
The analysis is based on two steps for both the fully-fledged EDIS and the mixed
deposit insurance system. In the first step, the exposure of EDIS to bank failures is
calculated using covered deposits as well as banks’ estimated probabilities of default
(PD), loss given default (LGD) and banks’ loss-absorbing capacity. Crises of different
magnitudes are considered where the riskiest 3% or 10% of banks fail
simultaneously according to their estimated PDs in combination with different
magnitudes of loss severity (LGD)
7
and two variations of banks’ loss-absorbing
capacity. This first step makes it possible to assess the resilience of EDIS to
potential loss scenarios of different severity and under different assumptions on loss
absorption by banks’ liabilities. In a second step, banks’ contributions to EDIS are
5
In its June 2016 Conclusions, the Economic and Financial Affairs Council (Ecofin) agreed that political
negotiations on EDIS would start as soon as sufficient further progress had been made on a set of
measures to reduce risks in the banking system.
6
The term “EDIS exposure” refers to the potential need for an EDIS intervention in case of a bank
failure. It is calculated, first, as losses minus loss-absorbing capacity at bank level and, second,
summed up for all banks.
7
Losses in insolvency are assumed to be always 50% higher than the losses in resolution.
ECB Occasional Paper Series No 208 / April 2018
7
estimated and compared to the EDIS exposures obtained in the first step. This
comparison aims to identify possible unwarranted cross-subsidisation across euro
area countries. In addition, the distribution of contributions across banks of different
size is studied.
The paper is organised as follows. Section 2 provides an overview of the rationale,
objectives and challenges related to deposit insurance. Section 3 illustrates the key
features of EDIS and sets the stage for the empirical analysis. Section 4 presents the
model for the estimation of banks' default probabilities, describes the loss-absorbing
mechanism and assumptions, and reports the findings on EDIS exposure. Section 5
discusses the rationale, methodology and findings of the contributions and cross-
subsidisation analysis under a fully-fledged EDIS. Section 6 illustrates the results on
contributions and cross-subsidisation under a mixed deposit insurance scheme.
Section 7 sets out conclusions.
ECB Occasional Paper Series No 208 / April 2018
8
2 Deposit insurance: rationale, objectives
and challenges
The guiding rationale behind a deposit insurance scheme is to enhance financial
stability by increasing depositors’ confidence in the safeness of their deposits
(Diamond and Dybvig, 1983). By removing the depositors’ incentives to withdraw
their money when concerned about a bank’s solvency, deposit insurance ultimately
reduces liquidity risk which may in turn reduce the likelihood of financial crises and
their severeness. Bernet and Walter (2009) argue that a deposit insurance scheme
can also be seen as a means to strengthen the competitiveness of smaller banks, to
foster growth by encouraging savings, and to have banks financing the deposit
insurance thereby reducing costs otherwise borne by taxpayers or depositors in case
of a resolution or insolvency.
As documented by Demirgüç-Kunt et al. (2008), the number of countries with an
explicit deposit insurance system rapidly increased over the last few decades, from
twenty in 1980 to eighty-seven by the end of 2003. Therefore, already before the
2008 global financial crisis, deposit insurance had become a key tool of the financial
safety net. After the crisis, this trend was reinforced: Demirgüç-Kunt et al. (2014)
reported that, out of 189 countries covered in their 2013 updated deposit insurance
dataset, 112 countries (or 59%) had explicit deposit insurance by yearend 2013 while
there were 84 (or 44%) in 2003.
8
The widespread adoption of explicit deposit insurance schemes around the world
signals a general belief that deposit insurance will bring benefits for depositor
protection and financial stability. However, deposit insurance schemes can also have
unintended consequences such as a shift in incentives towards risk-taking by banks
(moral hazard) and a decrease in effective monitoring exercised by depositors
(market discipline) (see, for example, Calomiris and Jaremski 2016a, 2016b).
Indeed, theory predicts that, once insured, there is a higher incentive for bankers to
enlarge the value of the deposit insurance by increasing risk (Merton, 1977). On the
depositor side, there is a lower incentive to search for the best bank to entrust their
savings or to demand higher interest rates in return for higher risk. For uninsured
deposits and where doubts on the credibility of the insurance arise, the economic
literature finds that banks’ risk is in fact reflected in their interest rates (e.g., Brewer
and Mondschean, 1994; Ellis and Flannery, 1992; Cook and Spellmann, 1994).
However, different conclusions emerge from Allen et al. (2017), who analyse the
effects of government guarantees on financial stability: the authors find that
guarantees are welfare improving since they induce banks to improve liquidity
8
Although all deposit insurance schemes share the goal of protecting depositors' confidence and
financial stability, their key features differ across countries, as shown by Demirgüç-Kunt et al. (2008,
2014). Differences may relate, among other features, to the coverage level, the source of funding of
deposit insurance, the institutional arrangements, the coverage of foreign currency deposits and
interbank deposits, or the presence of a co-insurance mechanism (under which depositors are insured
only for a fraction of their insured deposits and thus retain a portion of the possible losses).
ECB Occasional Paper Series No 208 / April 2018
9
provision although they can have (but not always) negative indirect effects on banks'
risk-taking decisions.
Calomiris and Jaremski (2016b) examine the state deposit insurance experiments of
the early 20th century in the US and find that deposit insurance increased both
insured banks’ default risk and overall systemic risk by removing market discipline
that had been constraining banks that were uninsured. Anginer et al. (2014) show
empirically that generous deposit insurance schemes increase systemic risk via
moral-hazard mechanisms before a crisis, and reduce risk and increase stability
during crisis times. Demirgüç-Kunt and Detragiache (2002) examine 61 countries
from 1980 to 1997 and also find that deposit insurance increases the likelihood of a
banking crisis and that this negative impact is intensified with a broader insurance.
The negative aspects of a deposit insurance scheme can however be mitigated by a
strong banking supervision framework correcting risk-taking incentives, e.g. a
supervisory authority equipped with the power to take corrective and preventive
actions (Anginer et al., 2014).
Demirgüç-Kunt and Huizinga (2004) shed light on the relationship between market
discipline, on the one hand, and deposit insurances and their design, on the other
hand. When looking at 30 developed and developing countries in the period from
1990 to 1997, they find that deposit insurance decreases market discipline and
causes lower interest rates. Regarding the design of deposit insurance, they
examine over 50 countries and show that co-insurance, in the sense of a deductible
for depositors, strengthens market discipline and increases deposit rates. On the
contrary, a higher explicit coverage reduces market discipline and lowers deposit
rates. With regard to the timing of the funding of a deposit insurance fund, there is
evidence that an ex-ante funded scheme deteriorates market discipline the most as
opposed to an ex-post funding mechanism, i.e. unfunded or an unfunded but callable
mechanism.
Funding of the deposit insurance fund may come from banks only, banks and
governments or just governments. The source of funding also affects market
discipline and interest rates: the more involved the government, the lower the
interest rates and market discipline. Hovakimian et al. (2003) focus their analysis on
56 countries and uncover that the introduction of explicit deposit insurance
encourages banks to shift risk onto the deposit insurer by increasing either their
leverage or the volatility of the return on assets. However, this effect can be
moderated by including risk-based premiums, coverage limits and co-insurance in
the insurance design. A similar alleviating effect can be attributed to a sound legal
system. In fact, Laeven (2002) finds that a strong legal system largely corrects for
the adverse consequences of an explicit deposit insurance.
ECB Occasional Paper Series No 208 / April 2018
10
3 A European Deposit Insurance Scheme:
key features and the way forward
In November 2015 the European Commission presented a proposal to establish an
EDIS as the third pillar of the banking union.
9
EDIS would be set up in three stages:
first, for three initial years from July 2017 to July 2020
10
, a reinsurance scheme
would cover up to 20% of the liquidity shortfall
11
and up to 20% of the excess loss
12
of a participating DGS whenever payouts and losses exceed the DGS’s available
financial means. The liquidity would be provided by means of a loan which the DGS
has to pay back, while the reinsured part of the excess loss, i.e. 20%, would not
have to be paid back. Furthermore, in order to limit moral hazard, the reinsurance
funding would be capped at 20% of the DIF’s initial target level or ten times the
target level of the insured DGS, whichever is lower. In addition, the benchmark for
calculating whether and to what extent a DGS can access the EDIS during the re-
insurance phase is the hypothetical level of liquidity the DGS should have if it had
complied with all its obligations (e.g. collecting ex ante contributions to reach the
target level), and not the actual level of liquidity in a DGS. Finally, other sources
available to the DGS (e.g. raising short-term ex post contributions) have to be tapped
before resorting to EDIS and the SRB is mandated to monitor the way the DGSs
pursue their claims during insolvency proceedings. During the reinsurance stage,
banks' risk-based contributions to the DIF would be calculated with reference to the
national banking system, i.e. relative to the riskiness of banks in the same country
and not of all banks in the banking union.
In the second stage, for four years after the end of the reinsurance stage and until
July 2024, a co-insurance scheme would be set up where the DIF would cover a
gradually increasing share (20% in year 1, 40% in year 2, 60% in year 3, 80% in year
4) of the liquidity needs and losses of participating DGSs. Co-insurance would kick-in
“as of the first euro”, so independently of the national DGSs’ resources being
exhausted. As it is the DGS which has the claim against the DIF (as opposed to
depositors or banks), any payout would be channelled through the national DGS.
While the liquidity provided to the DGS would have to be repaid, this is not the case
for the covered loss, which would be shared pro rata between the national DGSs and
DIF in line with the gradually increasing coverage ratio. No cap would be provided for
the amount due by the DIF. During the co-insurance stage, and differently from the
9
Proposal for a Regulation of the European Parliament and Council amending Regulation (EU)
806/2014 in order to establish a European Deposit Insurance Scheme.
10
To note, due to delays in the on-going negotiations, the timeline foreseen by the Commission is not
expected to be met.
11
According to the Commission proposal, the liquidity shortfall is the amount of covered deposits in the
failing bank which exceeds the total available financial means in the DGS (i.e. under the DGSD,
available funding plus extraordinary contributions that the DGS can raise within 3 days of the payout
event).
12
The excess loss is the loss remaining once the insolvency procedure is over (after recovery) and
long-term ex-post contributions have been called.
ECB Occasional Paper Series No 208 / April 2018
11
reinsurance stage, banks' risk-based contributions would be calculated with
reference to the riskiness of all banks in the banking union.
In the third and final stage, starting in July 2024 (after the seven years of re- and
co-insurance), a full insurance scheme would be in place: the EDIS would cover all
liquidity needs and losses of participating DGSs. In other words, the final stage
consists of a 100% mutualisation with national DGSs being fully insured by the DIF.
Also in this case there is no cap provided for the amount due by the DIF. Finally, to
limit the potential moral hazard linked to a full EDIS, the Commission proposal also
foresees that DGSs can be disqualified from EDIS coverage if they do not comply
with their obligations under the framework.
13
As in the coinsurance stage, banks'
risk-based contributions would be calculated taking into account the riskiness of all
banks in the banking union.
In October 2017 the European Commission published a Communication on the
completion of Banking Union (European Commission, 2017), including a proposed
new approach on EDIS aimed to address diverging views in the European
Parliament and the Council. The Commission proposed to introduce EDIS more
gradually relative to its original proposal. During re-insurance, differently from the
November 2015 proposal, there would be no coverage of losses, although the
coverage of the liquidity shortfall would be higher, increasing progressively up to
90% in the third year. The move to the second phase, i.e. co-insurance, would not be
automatic but contingent on a set of conditions to be assessed by the Commission,
for example related reduction of banks’ portfolios of Non-Performing Loans and Level
3 assets (illiquid assets which cannot be evaluated on the basis of market prices or
models). If these conditions are met and co-insurance starts, EDIS would also
provide coverage for losses, starting with a 30% coverage which should
progressively increase. However, the Communication does not provide any
information on how such an increase would take place and therefore the path of
mutualisation is unclear. Depending on the end point of the progressive increase in
losses coverage, the final stage could be closer or distant from a fully-fledged EDIS
with full insurance. As the Communication indicated that the original proposal
“remains on the table unchanged”, full-insurance in the steady state remains a
possibility to be discussed by co-legislators.
The main focus of this paper is on a fully-fledged EDIS with full insurance at the
European level, as in the steady state of the November 2015 Commission proposal -
that the European Central Bank supported in its public opinion in April 2016.
14
This
choice is made also taking into account that the October 2017 Communication is not
formally a new proposal and that full-insurance in the steady state is still possible.
However, recent policy discussions have also considered different approaches to
EDIS, notably referring to a possible design under which national DGSs would
13
Article 41 (i) of the Commission proposal envisages that a DGS shall not be covered in the
reinsurance, co-insurance or full insurance phase if disqualifying conditions are met.
14
Opinion of the European Central Bank of 20 April 2016 on a proposal for a Regulation of the European
Parliament and of the Council amending Regulation (EU) No 806/2014 in order to establish a European
Deposit Insurance Scheme (CON/2016/26).
ECB Occasional Paper Series No 208 / April 2018
12
intervene first, and the European deposit insurance fund would only step in as a
second line of defence. Despite some technical differences, the idea of national
DGSs or national compartments of EDIS bearing the first burden is common across
several proposals, e.g. respectively the November 2016 draft report of the
Committee on Economic and Monetary Affairs, drafted by Member of the European
Parliament Esther de Lange,
15
and the proposal by a group of French and German
economists in January 2018 (Bénassy-Quéré et al., 2018). For this reason, this study
also includes an additional analysis based on a mixed deposit insurance scheme,
composed of national compartments with a target level of 0.4% of domestic covered
deposits - intervening first - and a European compartment with a target level of 0.4%
of covered deposits - intervening only after depletion of the national compartment -
which is substantially in line with the De Lange's report.
Creating an EDIS is a logical step in completing the European Banking Union: while
the supervision and resolution pillar have already started to operate, respectively
with the Single Supervisory Mechanism (SSM) and the Single Resolution
Mechanism (SRM), the deposit insurance pillar is still missing. Ensuring a uniform
protection of depositors across the entire banking union, regardless of geographic
location, is a crucial element to preserve depositors’ trust and thus avoid bank runs
and protect financial stability. Establishing a fully-fledged EDIS would therefore
strengthen the banking union and reinforce the single currency.
Not all deposits should be afforded protection. Retail deposits of small, less
sophisticated investors should be insured, but wholesale deposits of large
sophisticated investors should not, in order to strike an appropriate balance between
protecting depositors’ trust and financial stability, on the one hand, and preserving
market discipline and limiting moral hazard, on the other hand. The European
legislation has achieved such a balance by restricting deposit insurance only to
“eligible deposits” and by setting a maximum level of coverage at €100,000 per
depositor per bank, which is harmonised throughout Europe.
Directive 2014/49/EU defines as “covered deposits” the deposits which are eligible
16
for the deposit guarantee and up to the €100,000 coverage level. It should be noted
that deposits from other banks and financial institutions are not eligible for coverage.
Therefore, the combination of a coverage cap and the exclusion of deposits from
certain types of investors aims to narrow the deposit insurance protection only to
15
Draft report on the Proposal for a regulation of the European Parliament and of the Council amending
Regulation (EU) 806/2014 in order to establish European Deposit Insurance Scheme (COM(2015)0586
– C8-0371/2015 – 2015/0270(COD)) Committee on Economic and Monetary
Affairs, European
Parliament, rapporteur: Esther de Lange, 4 November 2016.
16
“Eligible” deposits are defined by exclusion under Article 5 of Directive 2014/49/EU, which excludes
from the protection the following items: a) interbank deposits; b) own funds; c) deposits arising out of
transactions in connection with which there has been a criminal conviction for money laundering as
defined in Article 1(2) of Directive 2005/60/EC; (d) deposits by financial institutions as defined in point
(26) of Article 4(1) of Regulation (EU) No 575/2013; (e) deposits by investment firms as defined in point
(1) of Article 4(1) of Directive 2004/39/EC; (f) deposits the holder of which has never been identified
pursuant to Article 9(1) of Directive 2005/60/EC, when they have become unavailable; (g) deposits by
insurance undertakings and by reinsurance undertakings as referred to in Article 13(1) to (6) of
Directive 2009/138/EC of the European Parliament and of the Council; (h) deposits by collective
investment undertakings; (i) deposits by pension and retirement funds; (j) deposits by public authorities;
(k) debt securities issued by a credit institution and liabilities arising out of own acceptances and
promissory notes.
ECB Occasional Paper Series No 208 / April 2018
13
those deposits and investors deemed in need of protection, i.e. assumed to be small
and non-sophisticated. The lack of protection for wholesale investors and for
amounts exceeding the coverage level should tackle moral hazard and enhance
market discipline, for investors which are considered non-retail and more
sophisticated. In the empirical analysis on EDIS developed in this paper, the focus
will thus be on covered deposits only, i.e. deposits which are eligible for deposit
insurance protection and up to the coverage level of €100,000.
A second crucial issue for the functioning of EDIS is the interaction with the new
requirement for Total Loss-Absorbing Capacity (TLAC) introduced by the Financial
Stability Board for Global Systemically Important Banks (G-SIBs) and the new
Minimum Requirement for own funds and Eligible Liabilities (MREL) introduced for all
European and banking union banks by the BRRD (Bank Recovery and Resolution
Directive
17
) and the Single Resolution Mechanism Regulation (SRMR)
18
,
respectively. Both TLAC and MREL will in fact play a loss-absorbing function which,
in principle, could shield the EDIS from losses related to resolution and enhance the
resilience of the banking system in general. The amount of TLAC and MREL
liabilities, in proportion to the overall bank balance sheet, as well as the amount of
covered deposits in a bank will play a crucial role in determining whether the deposit
insurance fund would be effectively protected or should step in to bear part of the
losses – in lieu of covered deposits – in a bank resolution context. While the
involvement of the deposit insurance fund for loss coverage is less likely in a
resolution, it might still be possible, especially if a bank does not have sufficient loss-
absorbing capacity and/or if the resolution authority decides to exercise its power to
exclude on a discretionary basis some liabilities from the scope of bail-in, e.g. for
financial stability purposes. This could narrow the cushion of loss-absorbing capacity
which protects covered deposits/deposit insurance fund.
However, the possible intervention of the resolution fund, which is subject to certain
preconditions and caps, would lower the probability of exposures of the deposit
insurance fund to losses. Finally, the choice by the resolution authorities between
resolving or liquidating a bank under an ordinary insolvency procedure will also play
a crucial role: losses are likely to be higher under an insolvency procedure than in
resolution, meaning that the deposit insurance fund could be potentially more
exposed to losses in insolvency than in resolution (although this depends on the
loss-absorbing capacity of liabilities in resolution and liquidation; in any case, the
reverse cannot happen because losses for the deposit insurance fund in resolution
cannot be higher than under insolvency; see Article 109.1 BRRD and Article 79.5
SRMR).
All these elements have fundamental implications for the exposure of the DIF which
would be established under the EDIS. A key objective of this paper will be to assess
the EDIS exposure under different scenarios, obtained by combining different
assumptions on the features discussed above.
17
Directive 2014/59/EU.
18
Regulation (EU) No. 806/2014.
ECB Occasional Paper Series No 208 / April 2018
14
A crucial, related issue concerns the appropriate size of the deposit insurance fund.
In the European Commission proposal, the target size for the DIF in the steady state
is set at 0.8% of covered deposits. This is in line with the European Directive on
Deposit Guarantee Scheme (DGSD), which harmonised the target level for national
DGSs.
19
At the time of the proposed review of the DGSD in 2010, the Commission
had prepared an impact assessment looking among other things at the potential
target for national DGS. The impact assessment of the Commission concluded that
the target level should be sufficiently high to ensure that schemes are credible and
capable of dealing with medium-sized bank failures, while maintaining banks'
profitability.
20
The Commission proposed a target of 1.5% of eligible deposits to be
reached with ex-ante contributions. The target was subsequently reduced to 0.8% of
covered deposits during the co-legislative process between the European Parliament
and the Council of the European Union. Additionally, a possibility was introduced for
Member States to authorise a lower minimum target level of 0.5% of covered
deposits, where the banking system in the Member State is highly concentrated with
a large quantity of assets held by a small number of credit institutions or banking
groups, which are more likely to go into resolution than insolvency. Finally, the DGSD
contains the possibility for Member States to agree to a mechanism of voluntary
lending between national DGSs, but it has never been used so far.
In the context of the EDIS proposal, the Commission also published an effect
analysis on EDIS which concludes that pooling risk at the European level delivers a
significantly stronger deposit guarantee system than a system of purely national
schemes, or a scheme with voluntary lending between DGSs.
21
However, the study of the Commission did not take into account several aspects
related to MREL and bail-in as well as scenarios including combinations of
probability of default and loss given default for banks. Therefore, this paper conducts
further empirical analysis to test whether and under what scenarios and conditions a
0.8% target would be appropriate and sufficient. In addition, the paper analyses how
different calibrations of risk-based deposit insurance affect the distribution of
contributions across countries, investigates how the collection of contributions would
be distributed across small, medium and large banks and assesses whether an EDIS
would produce any systematic cross-subsidisation between banking sectors in
different Member States.
19
Directive 2014/49/EU.
20
Commission staff working document - Impact Assessment - Accompanying document to the Proposal
for a Directive …/…/EU of the European Parliament and of the Council on Deposit Guarantee Schemes
[recast] and to the Report from the Commission to the European Parliament AND to the Council
Review of Directive 94/19/EC on Deposit Guarantee Schemes, published 12 July 2010.
21
See Effects analysis on the European deposit insurance scheme, published 11 October 2016.
ECB Occasional Paper Series No 208 / April 2018
15
Box 1
Comparison between EDIS and the US FDIC
The deposit insurance scheme set up by the Federal Deposit Insurance Corporation (FDIC) in the
US has performed a crucial role in ensuring financial stability since it was created in the 1930s. In
addition to deposit protection, the FDIC also has a key role in bank supervision and resolution
(Beck and Laeven, 2006). In fact, following the financial crisis of 2007-2009, the Dodd-Frank Act
gave the FDIC even more responsibility in this regard, notably on resolution.
First, regarding the degree of protection of deposits, while the EDIS aims to continue covering
deposits up to €100,000 in case of a bank failure, the FDIC insurance amount is currently
USD250,000 per depositor per insured bank.
Second, similarly to Europe where the target level for ex-ante funds was set at 0.8% of covered
deposits after the initial proposal to set it at 1.5% of eligible deposits, there has been a substantial
debate in the US about the optimal size of the deposit insurance fund. From its creation until 1989
there was no target size for the FDIC fund. However, to address concerns about the viability of the
fund after 2,900 bank and thrift failures from 1980 through 1994 (Ellis, 2013), a target size in the
form of a Designated Reserve Ratio (DRR) equal to at least 1.25% of insured deposits was
introduced in 1989. However, the FDIC deposit insurance fund went negative again following the
2007-2009 financial crisis and, as a result, the Dodd-Frank Act of 2010 increased the minimum
reserve ratio to 1.35%, to be reached by 2020. Since 2011, the FDIC has set the DDR at 2%.
Following the greater authority that was given to the FDIC in 2010, the Corporation moved from
portfolio management techniques and loss distribution simulation models to determine the size of its
fund to a much simpler approach determining the size of the fund on the basis of two severe
banking crises to prevent it from going negative. This framework underpins the current fund-
management strategy of the FDIC and corresponds to a long-term target of 2% of insured deposits.
The US FDIC deposit insurance fund had a balance of USD 83 billion at the end of 2016,
corresponding to 1.20% of the total amount of insured deposits (USD 6.9 trillion). Based on the
volume of insured deposits at year-end 2016, a 2% DRR would translate into a USD 138 billion
fund.
When assessing the appropriateness of the European target level, it has to be noted that the DIF
will coexist with the SRF, which itself has a target level of 1% of covered deposits. Thus, a total
amount of 1.8% of covered deposits will be dedicated to resolution and deposit insurance purposes,
corresponding to about €100 billion. In the US, the deposit insurance fund is also used to perform
resolution functions. As a result, its target size could be compared with the target size in the steady
state of both the SRF and the DIF.
Third, the DGSD specifies that contributions to the EDIS should be made ex ante and be based not
only on the amount of covered deposits but also on banks’ individual degree of risk. Similarly, the
FDIC fund has always been financed ex ante by the banks since alternative arrangements such as
ex-post assessments and contributions could arguably increase the risk of costly delays and
undermine the confidence in the financial system (Ellis, 2013). Nevertheless, the FDIC evolved from
simple rules into a more sophisticated system to determine what premium banks have to pay. In
fact, from its inception to 1991 all banks paid the same rate set by the US Congress and, as a
result, less risky banks were effectively subsidising banks with a higher risk profile. However, in
1991 the US Congress required the FDIC to adopt a risk-based premium framework which started
being implemented in 1993 (Ellis, 2013).
ECB Occasional Paper Series No 208 / April 2018
16
4 EDIS exposure
4.1 Data on covered deposits
The analysis uses Bankscope data and supervisory data from COREP (Common
Reporting) and FINREP (Financial Reporting) for 2015:Q4 on covered deposits and
balance sheet indicators to estimate EDIS exposure to bank failures and
contributions to EDIS at bank level. The conclusions of the analysis are therefore
based on the assumption that the banks’ balance sheet structure remains the same
until EDIS has been fully introduced. The sample scrutinised comprises 1,675 euro
area banks with total assets of €22.14 trillion, which amounts to about 75% of total
assets of credit institutions in the euro area, and €4,744 billion of covered deposits,
corresponding to about 83% of covered deposits in the euro area. The sample can
be considered as representative, both in terms of total assets and covered
deposits.
22
The target size of the DIF for the sample is approximately €38 billion. The
box plot in Chart 1 shows the distribution of covered deposits per total assets within
each country in the euro area as of year-end 2015. There is generally heterogeneity
across countries both in terms of median and in terms of dispersion. German banks
have the highest median of covered deposits per total assets in the euro area, but
also considerable variation in the amount of covered deposits relative to their
balance sheet size.
Chart 1
Distribution of covered deposits to total assets by country
(in %)
Sources: ECB calculations based on COREP for 1,675 banks, reporting date 2015:Q4.
Notes: The bottom and top of the box represent the first and third quartiles of the within-country distribution, while the band inside the
box is the second quartile (median). The ends of the whiskers are the maximum and minimum values excluding outliers. Outliers are
represented by diamonds.
22
The degree of representativeness of the sample at country level is, however, heterogeneous. For
example, the coverage ratio in terms of total assets is 77% for Germany and France, 62% for Italy, 83%
for the Netherlands, 88% for Greece, 65% for Belgium, 43% for Austria, 34% for Ireland and 24% for
Cyprus.
0
10
20
30
40
50
60
70
80
90
100
AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK
ECB Occasional Paper Series No 208 / April 2018
17
4.2 Methodology
4.2.1 Probability of default estimation
The analysis of the exposure of EDIS is based on various crisis simulations which
follow the methodology for calculating the Probability of Default (PD) for banks
provided in Betz et al. (2014) as well as Lang et al. (2018). The PD of around 5,000
euro area banks is estimated via an early warning model that accounts for different
bank-specific, aggregate banking sector and macro-financial variables, using panel
data from 1999 to 2013.
To estimate the coefficients used to calculate bank-specific PDs, banks in the sample
were classified as either in distress/default or not in distress/default. The
identification of distress events can be challenging given that actual bank failures
have not been frequent in the euro area. As a result, following Betz et al. (2014), a
bank is defined as in distress/default if: (i) the status of the bank in the Bankscope
database is either “bankruptcy”, “dissolved” or “in liquidation”; (ii) the bank has
negative capital; (iii) the bank was involved in a distressed merger, i.e. the merged
entity has a negative coverage ratio (capital equity and loan loss reserves minus
non-performing loans to total assets) one year before the merger; or (iv) the bank
received state aid based on the data from the European Commission.
23
A forecast
horizon of two years prior to bank distress/default events is used to identify the build-
up of vulnerabilities with a sufficient lead time.
Table 1 shows the estimated coefficients from the logit regressions using either five
or eight bank specific controls and five or eight country-level characteristics –
specifications (A) and (B), respectively. All coefficients are statistically significant in
both models. In detail, less profitable banks, banks with lower tangible equity as a
percentage of assets, higher share of trading income, lower share of deposits and
higher provisions for NPLs have a higher probability of default. In terms of country-
level characteristics, in line with Jordà et al. (2017), who analyse 17 advanced
economies from 1870 to 2013 and find that credit growth on the asset-side of banks’
balance sheet and liquidity indicators such as the loan-to-deposit ratio are key crises
predictors, the reported coefficients show that banks’ probability of default is higher
in countries with higher 1-year change in the loan-to-deposit ratios. Also in line with
economic theory and previous empirical evidence (e.g. Lang et al., 2018), the
estimates indicate that banks’ probability of default is higher in countries with higher
financial assets as a percentage of GDP, lower 1-year change in the ratio of issued
debt securities to total liabilities, as well as higher unemployment and inflation.
23
This information is publicly available at: state aid. In this analysis, resort to central bank Emergency
Liquidity Assistance is not considered as an indicator of distress/default, because it concerns banks
which are facing temporary liquidity problems but are solvent.
ECB Occasional Paper Series No 208 / April 2018
18
Finally, the relatively high area under the receiving operating curve (AUROC)
indicates that the models can explain and predict the data well.
24
This finding is also
confirmed by the in-sample test (see Appendix). Given the very high correlation
between the two sets of predicted PDs (above 90% both in terms of estimated PDs
and rankings) specification (A) below is used throughout the study to maximise the
number of banks in the analysis.
Table 1
Estimated coefficients for the early warning models (1999 - 2013)
A B
Tangible equity / Total assets
-0,009** (0,004) -0,011** (0,005)
Return on Equity
-0,045*** (0,004) -0,022*** (0,007)
Interest expenses / Total liabilities
0,139*** (0,033) 0,195*** (0,039)
Share of trading income
0,042*** (0,008) 0,043*** (0,009)
Deposits to Assets
-0,022*** (0,002) -0,024*** (0,002)
Provisions for NPLs / Total assets
0,345*** (0,081)
Cost to Income
0,012*** (0,003)
Loans to Deposits
-0,000* (0,000)
Financial assets / GDP
0,000*** (0,000) 0,000* (0,000)
Loans / Deposits (1-year change)
0,010*** (0,003) 0,014*** (0,004)
Issued debt / Total liabilities (1-year
change)
-0,052*** (0,008) -0,036*** (0,011)
Unemployment
0,077*** (0,011) 0,065*** (0,014)
Inflation
0,018*** (0,006) 0,030*** (0,008)
House price index
-0,037** (0,015)
10-year yield (1-year change)
0,126*** (0,046)
Bank concentration (HHI)
3,382*** (0,882)
No. Observations
47 775 45 167
No. Banks
5 082 4 881
No. Distressed Banks
293 241
Pseudo R2
0,128 0,153
AUROC
0,795 0,818
Source: ECB staff calculations based on Bankscope, ECB Statistical Data Warehouse and European Commission dataset on state aid
measures, 1999:Q1 – 2013:Q4.
Note: Standard errors robust to heteroskedasticity in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by
*, ** and ***, respectively.
The bank-specific PDs for the year 2015 are calculated on the basis of the
coefficients estimated above. Given the distribution of the estimated 2015 PDs for
1,675 banks (93 of which are Significant Institutions - SIs), the 3% (using the 97th
percentile as a threshold) and 10% (corresponding to the 90th percentile) of banks
with the highest PDs are singled out as banks most likely to fail on the basis of the
described PD methodology. The analysis assumes that, for each crisis simulation, all
banks belonging to the riskiest 3% or 10% fail simultaneously. The 97th and 90th
percentiles correspond to a PD of 5.32% and 3.04%, respectively.
24
The AUROC (Area Under the Receiver Operating Characteristics Curve) is used to measure the
performance of the early-warning model. The greater the predictive power, the more bowed the curve,
and hence the area beneath the curve is often used as a measure of the predictive power. A perfect
indicator has an AUROC of 1, while an uninformative indicator has an AUROC of 0.5.
ECB Occasional Paper Series No 208 / April 2018
19
As a result of the empirical set-up for the estimation of PDs, crisis simulations in this
analysis strongly depend on observed banking failures and crises. Additionally, while
the data on independent variables used to calculate banks’ probabilities of failing are
point in time (year-end 2015), the coefficients for the regressions to obtain PDs are
calculated taking the economic and financial cycle into account. The inclusion of the
recent financial crisis may influence this paper’s results leading to potential
discrepancies between simulated failures and those possibly materialising in the
steady state.
4.2.2 Loss given default
To define a range for the Loss Given Default (LGD), i.e. the amount of losses for
each bank when it fails, we have considered the historical bank losses observed in
the past, particularly during the recent financial crisis. The European Commission
25
estimated average losses for 23 banks over the period of 2007-2010 to be 2.5% of
total liabilities (maximum of 46.4%; minimum of 0.2%) while losses plus
recapitalisation needs were on average 6% of total liabilities (maximum of 50.7%;
minimum of 2.6%). The Financial Stability Board found that for G-SIBs losses as a
fraction of total assets ranged from less than 1% to 4.7%, with most banks in a 2-4%
range. The maximum ratio of losses and recapitalisation amounts relative to total
assets was 8.8%, with most banks between 3.9% and 6.1% (Financial Stability
Board, 2015).
Given the evidence above and the significant variation in the estimates of LGDs, a
wide range for average LGD for banks in distress is considered. Since one of the
objectives of the simulations in this paper is to also test an extremely stressful
scenario, the analysis considers losses in resolution (with bail-in) from 5% to 25% of
total assets. This is an extremely conservative range if compared to the values
observed in the last crisis. The lower threshold of 5% total assets, for example, is
higher than the upper bound reported by the FSB for G-SIBs during the last crisis,
and twice the average estimates reported by the Commission for the period
2007-2010. In addition, the analysis presented in this paper not only simulates
extremely severe shocks, but also considers the same amount of losses
simultaneously affecting a wide sample of banks, e.g. the 3% and the 10% of the
euro area riskiest banks. This is a very conservative assumption, also when
compared to events in the last financial crisis.
Insolvency proceedings can be particularly complex, costly and time-consuming,
thus often resulting in higher average losses than when a bank is resolved. This is
true even if one ignores externalities such as spillovers into the rest of the financial
system and the losses that could ultimately be borne by the bank’s depositors
(Hardy, 2014). Therefore, to consider the complexity, cost and length of insolvency
proceedings, losses in insolvency are assumed to be always 50% higher than the
losses in resolution, and therefore range from 7.5% to 37.5%.
25
Commission staff working document impact assessment.
ECB Occasional Paper Series No 208 / April 2018
20
4.2.3 Methodology to model resolution and loss-absorption capacity
The need for an EDIS contribution in case of a bank failure depends on whether the
bank goes into resolution or insolvency, and on its level of loss-absorption capacity.
For the purpose of the analysis, it is assumed that a bank would be resolved if it has
(i) a balance sheet size of more than €20 billion, or (ii) more than 40,000
transactional accounts, and that it would otherwise be liquidated.
26
Since there is no
available data on the number of transactional accounts for the banks in the sample,
the assumption is that any bank with more than €4 billion in covered deposits is
above the 40,000 threshold, i.e. that each account has €100,000 (corresponding to
the maximum amount covered by deposit insurance per depositor per bank). This
assumption is conservative, since on average each account has less than €100,000.
Therefore, the analysis may overestimate the number of banks going into insolvency
rather than resolution, which is overall conservative in terms of losses since
insolvency may cause more destruction of value than resolution. In resolution, a
contribution from the Single Resolution Fund (SRF) is considered in each scenario,
respecting the conditions set out in the legislation. This means that SRF
contributions are capped at a maximum of 5% of bank’s total assets after
shareholders and creditors have absorbed losses and recapitalisation costs
corresponding to at least 8% of total assets.
27
In addition, cumulative SRF
contributions cannot exceed the overall size of the SRF (i.e. 1% of total covered
deposits in the sample which equals €47.4 billion
28
). Furthermore, the DIF
contribution in resolution cannot be higher than the contribution it would have paid in
insolvency.
29
Regarding banks’ loss-absorption capacity in resolution, two scenarios are used in
the analysis:
A. All liabilities except for secured liabilities and covered deposits absorb losses;
B. Only regulatory capital, subordinated debt and senior unsecured bonds with a
remaining maturity of at least 12 months absorb losses.
Scenario A is always used to model banks' loss-absorption capacity in liquidation.
26
These assumptions broadly follow the Bank of England’s proposed approach to direct institutions to
maintain a minimum requirement for own funds and eligible liabilities, in their December 2015
consultation paper (Bank of England, 2015). Following feedback to the consultation, the Bank of
England made two changes regarding the transactional accounts threshold: first, it clarified that
accounts are defined as “transactional” on the basis of the frequency of their use (at least nine
withdrawals over the previous three months); second, a range of between 40,000 and 80,000 accounts
replaced a fixed threshold of 40,000; see Bank of England (2016).
27
The BRRD (art. 44.5 and 49.3) requires that the 8% threshold to allow SRF intervention and the 5%
cap on the SRF intervention are calculated on total liabilities and own funds (TLOF), taking into account
netting agreements for derivatives. However, this analysis uses total assets as a proxy for TLOF due to
data availability issues.
28
For the purpose of the calculation of this cap, only ex-ante contributions are taken into account. Ex-post
contributions to the SRF could raise the cap.
29
See Article 79 of the Single Resolution Mechanism Regulation. To ensure that this condition is met, the
EDIS exposure for a bank subject to resolution is compared to that bank’s hypothetic loss posed to
EDIS if it had been liquidated. The EDIS exposure is then set to be the lower of the two.
ECB Occasional Paper Series No 208 / April 2018
21
The analysis follows the existing creditor hierarchy, where covered deposits have a
super-priority, both in resolution and liquidation. Indeed, the BRRD and the SRMR
make it possible to subject a wide range of unsecured liabilities to losses, e.g. via a
bail-in, and give a super-priority to covered deposits in the ranking of creditors.
However, in practice, it is unlikely that all liabilities within the scope of bail-in will be
fully loss-absorbing at the point of resolution. Therefore, scenario B considers a
bail-in scenario in which only MREL-eligible liabilities are considered to be fully
loss-absorbing (but deposits of large corporates above €100,000 are not included,
despite being MREL-eligible, to make the scenario more conservative). It should be
noted, however, that senior unsecured liabilities currently rank alongside other
liabilities classes, e.g. derivatives. Thus, it is unlikely that they would be fully
loss-absorbing given the “no-creditor-worse-off” principle.
4.3 Main results
Table 2 presents the simulation results showing the estimated exposure of EDIS
under loss absorbency scenarios A and B in resolution, respectively. These
scenarios are estimated for different levels of LGD and different PD thresholds, i.e.
the 90th and 97th percentiles. Additionally, for all scenarios the simulations show the
estimated EDIS exposure with and without SRF contribution assuming the
requirements mentioned in the previous section are satisfied.
30
30
The analysis takes into account the following caps: (1) a ceiling for the SRF contribution in order not to
deplete the SRF (SRF contribution is capped at 1% of covered deposits of participating banks); (2)
EDIS exposure for each bank cannot exceed the amount of covered deposits held by that bank; (3) a
resolution cap not allowing EDIS exposure in resolution to exceed the theoretical exposure if the bank
had been subject to insolvency; and 4) SRF contribution cannot exceed 5% of a bank’s total assets.
Note that, in this exercise, the fact that DGS contribution in resolution cannot exceed 50% of its target
level as per Article 109 BRRD and Article 79 SRMR is not considered. As a result of the different caps
which limit EDIS exposure, the exposure numbers are similar with and without SRF contribution
calculations. Furthermore, it should also be noted that only bank losses are considered for the
calculations on the EDIS exposure, while bank recapitalisation needs (which would not be borne by
EDIS) are not included.
ECB Occasional Paper Series No 208 / April 2018
22
Table 2
EDIS exposure in EUR billion for scenario A and B, with and without SRF contribution
(EDIS target size amounts to €38 billion)
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: Resolution scenario A: all liabilities except for secured liabilities and covered deposits absorb losses. Resolution scenario B: only capital, subordinated debt and senior
unsecured bonds with a remaining maturity of at least 12 months absorb losses.
The table depicts different levels of losses for scenario A (B) in columns 3 to 6
(columns 7 to 10) as a percentage of total assets, both in resolution and insolvency
(columns 1 and 2). Columns 3(7) and 5(9) represent the DIF exposure when the
riskiest 3% of euro area banks in the sample fail, without and with SRF contribution,
respectively. For the sample of 1,675 banks, this implies the failure of 51 banks
which hold 12.66% of the total assets in the sample. 18 of those failing banks enter
into resolution and 33 into insolvency, corresponding to 12.42% and 0.24% of total
assets, respectively. Columns 4(8) and 6(10) show the equivalent numbers for an
extremely severe crisis where the riskiest 10% of euro area banks in the sample fail,
respectively, both without and with SRF contribution: this implies the failure of 167
banks which hold approximately 40% of the total assets in the sample
31
The table shows that EDIS exposure never exceeds the DIF target size in the
sample, equal to € 38bn: this means that despite the severity of the simulated losses
(up to 25% of the total assets in resolution, simultaneously applied to all the 3% and
10% riskiest banks in the sample), and the conservative assumptions on the
loss-absorbing capacity in resolution (Scenario B), the fund is never depleted.
The results also suggest that a fully-funded DIF with ex-ante contributions of 0.8% of
covered deposits would have no exposure in cases of loss rates up to 20% in
resolution and 30% in insolvency, in both scenarios A and B. This finding reflects the
strengthening of banks’ loss-absorbing capacity and the effects of the risk-reducing
measures that have been implemented so far.
Under scenario A, EDIS exposure remains very limited in all the crisis-loss
combinations, with and without SRF contribution in resolution: even with a loss rate
equal to 25% of total assets in resolution, for instance, the EDIS exposure is capped
at € 40mln.
31
As a term of comparison, according to the definition of default/distress used in this analysis, in 2007-
2009 the share of failed/distressed banks was 0.8% in terms of number of banks and 39.1% in terms of
total assets.
Scenario A in resolution Scenario B in resolution
Without SRF With SRF Without SRF With SRF
1) 2) 3) 4) 5) 6) 7) 8) 9) 10)
Loss in Resolution
(as % of total
assets)
Loss in Insolvency
(as % of total
assets)
3 Percent
riskiest
banks fail
10 Percent
riskiest
banks fail
3 Percent
riskiest
banks fail
10 Percent
riskiest
banks fail
3 Percent
riskiest
banks fail
10 Percent
riskiest
banks fail
3 Percent
riskiest
banks fail
10 Percent
riskiest
banks fail
5% 7,50% 0 0 0 0 0 0 0 0
10% 15% 0 0 0 0 0 0 0 0
15% 22,50% 0 0 0 0 0 0 0 0
20% 30% 0,003 0,003 0,003 0,003 3 3,6 3 3,6
25% 37,50% 0,04 0,04 0,04 0,04 23,9 30,4 23,8 30,4
ECB Occasional Paper Series No 208 / April 2018
23
Under scenario B, EDIS exposure would become material only in case of losses at
least equal to 25% of total assets in resolution. It should be emphasised that,
according to the FSB estimates (see Section 4.2.2), during the last crisis only one
bank reported losses higher than 8% of total assets: in order to trigger high EDIS
exposures, in this study it is necessary to assume that the entire 3% (or 10%) of the
banks in the sample is affected by the 25% loss rate - a scenario considerably
harsher than the historical cases occurred both in Europe and the US. It should be
noted, in fact, that the analysis in this paper assumes a simultaneous failure of the
riskiest banks and a fixed level of losses relative to total assets rather than a
distribution, which is extremely more conservative than what was observed in past
crises, notably for high loss levels.
Finally, it should be emphasised that the main benefit of an EDIS derives from
reducing the sovereign-bank nexus as well as from pooling resources across
Member States. 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: this effect would be even more beneficial considering that the
banking system would be exposed to EDIS only in case of extremely severe crises.
Spillovers are not modelled in the analysis given the confidence enhancing role of an
EDIS.
ECB Occasional Paper Series No 208 / April 2018
24
5 Risk-based contributions to EDIS
5.1 Rationale
One concern which is frequently voiced regarding EDIS relates to the possibility that
the pooling of resources could lead to cross-border subsidies, i.e. the eventuality of
one or several banking systems structurally contributing more and benefitting less
from the scheme than other, potentially riskier, systems. The pooling of resources
could also lead to increased moral hazard and incentivise risk-taking behaviour by
banks given the existence of a larger deposit guarantee scheme and fund. There
might also be the risk that certain banking systems would be more likely to tap into
the EDIS funds than others, even though all banking systems would benefit from the
enhanced capacity of the deposit scheme to withstand larger crises.
The post-crisis review of the European Deposit Guarantee Scheme Directive
(DGSD) applied the concept of risk-based contributions to national DGSs.
32
The
Commission proposal on EDIS also foresees the use of risk-based contributions to
the DIF, the methodology of which would be determined in a Commission Delegated
Act. The use of a “Banking Union methodology”, i.e. a methodology comparing
banks across the banking union rather than within each national banking system,
would have the potential to reduce the risk of cross-border subsidies compared to a
system where banks’ contributions would be calculated only relative to their national
peers. This is because, following a “polluter-pays” principle, a national banking
system would contribute more to the DIF overall if it is riskier relative to other
banking systems in the banking union. This approach would have the benefit of
aligning incentives and tackling moral hazard, since banking systems which include
riskier banks would contribute more to the DIF overall than they would if
contributions were solely based on the amount of deposits or calculated only taking
into account the riskiness of banks within the national banking system.
5.2 Methodology
Given that the exact methodology for the calculation of banks’ contributions to the
DIF is yet to be developed, the analysis below is based on a modified version of the
methodology developed by the EBA for national DGSs in which banks’ contributions
are risk-based.
33
It must be stressed that, while the EBA methodology for national
DGSs applies the risk adjustment at a national level, the risk adjustment in this
analysis is carried out at the banking union level.
According to the EBA Guidelines, the calculation of an institution’s contribution is
based on five risk categories: (1) capital, (2) liquidity and funding, (3) asset quality,
32
Directive 2014/49/EU.
33
See EBA Guidelines on methods for calculating contributions to deposit guarantee schemes to be
found here : Guidelines.
ECB Occasional Paper Series No 208 / April 2018
25
(4) business model and management, and (5) potential losses for the DGS (this
factor is not considered in this analysis due to limited data on unencumbered
assets). For the purpose of this study, the leverage ratio and the total risk-based
capital ratio are included for category (1), liquid assets per total assets
34
are included
for category (2), and the Return On Equity (ROE) and Risk Weighted Assets (RWA)
per total assets are used for the category representing an institution’s business
model and management (4). Furthermore, the analysis includes a measure of (part
of) MREL-eligible liabilities.
35
The higher the MREL, the higher the likelihood of a
bank going into resolution rather than liquidation, the higher the bank's expected
capacity to absorb losses and, all else being equal, the lower the potential exposure
for EDIS.
36
The combination of these indicators shall hereinafter be referred to as
“DGS-baseline indicators” and is comparable to the list of indicators proposed for
EDIS. 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. In a first modification of the baseline list of indicators,
the indicator for MREL-eligible liabilities is excluded to test the impact of this
indicator on the contributions. In a second modification, the established baseline set
of indicators is extended by additionally including an indicator for interconnectedness
measured as the sum of loans and advances from and to banks relative to the total
amount of these items in the sample. The non-performing loans ratio (category (3)) is
not included in the baseline analysis because of data limitations. However, it is
reported separately in column 7 in Table 3, where it is added to the baseline
indicators to indicate its potential relevance for the purpose of future analyses.
Finally, the established baseline indicator set is extended by including the World
Bank index for the strength of insolvency frameworks, since the proper functioning of
the insolvency framework will have an impact on the deposit insurance's capacity to
recover money in insolvency after a payout.
37
The EBA Guidelines suggest two alternative approaches to constructing aggregate
risk weights (ARW) that are used in the contribution calculation: a bucket approach
and a sliding scale approach. The results presented are those for the sliding scale
approach, since this approach requires fewer assumptions and uses a normalisation
method that is better suited to preserving the level of information of the indicators.
38
The 25th and 75th percentiles are used as lower and upper boundaries,
34
Defined as: (Cash & balances with central banks + Net loans and advances to banks + Level 1 assets
(fair value hierarchy)) / Total Assets.
35
Senior unsecured bonds only: regulatory capital is not included to avoid double consideration, given
that it is already used for category (1) on capital.
36
The EDIS exposure could be lower for several reasons: for example, MREL-eligible liabilities cannot be
suddenly withdrawn, e.g. in a run, because they must have residual maturity of at least one year;
losses in insolvency tend to be higher than in resolution; the losses for the deposit guarantee scheme
in resolution cannot be higher than in insolvency (see Article 109 of the BRRD and Article 79 of the
SRMR).
37
See Resolving Insolvency
38
See OECD and JRC (2008). 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 to calculate the ARW. The aforementioned work of the OECD and JRC gives an
insightful overview of indicator construction in general.
ECB Occasional Paper Series No 208 / April 2018
26
respectively.
39
The normalisation transforms each individual indicator value into an
individual risk score (IRS) such that a lower score corresponds to a better performing
bank in the respective field. In a second step, the IRSs are aggregated using a
weighted arithmetic average to obtain a single aggregate risk score (ARS) for each
bank.
40
The relative size of the weights used in the analysis follows the EBA
Guidelines. In this paper’s analysis the ARS is not rescaled before it is used as an
ARW in the contribution calculation.
41
In a last step, banks’ contributions are calculated as the product of the contribution
rate, the ARW, the total covered deposits held by a bank and an adjustment factor
that ensures that contributions add up to the target size, which is set to 0.8% of the
total covered deposits in the sample.
To gain some insight into the importance of the calculation method used, the results
are also compared with the risk-based contributions calculated under the SRF
approach
42
.
5.3 Contributions
Table 3 gives an overview of the amounts (in billions of euro) and the share
contributed by each banking system based on the different indicator sets described
above. All columns are obtained using the DGS sliding scale methodology according
to the EBA Guidelines. Column 3 shows the amount and the non-risk based share
contributed only on the basis of covered deposits amounts. Column 4 reports
contributions and the share for each banking system on the basis of the
DGS-baseline indicators. The comparison between columns 3 and 4 provides an
indication of the impact of the risk factor in the calculation of contributions. A
reduction in contributions following the “polluter-pays” principle is visible for banks in
Austria, Belgium, Finland, France, Lithuania, Latvia and the Netherlands; on the
contrary, banks in Germany, Spain, Greece, Italy, Portugal and Slovakia would see
an increase in contributions due to their higher risk with respect to the European
benchmark. Columns 5 to 8 show the modifications of the baseline indicators as
described in the previous section and column 9 presents the results based on the
SRF approach. The impact of the MREL indicator on the contributions becomes
apparent when comparing column 4 to column 5. This indicator can be perceived as
a proxy for a bank’s likeliness to go into resolution instead of insolvency. The
39
Except for the NPL ratio, where missing values are set to zero in order to keep the sample size large,
the first and third quartile thresholds are both zero. To avoid division by zero and to produce a more
differentiated IRS, the upper bound is set to the highest value observed.
40
An alternative to this aggregation method would be a geometric average which, in contrast to the
arithmetic average, does not allow for the compensation of a poor performance in one field by a very
good performance in another field (OECD and JRC, 2008).
41
Rescaling has a distorting effect on aggregate risk scores by, in this case, reducing risk scores for
riskier banks more than for safer banks. See Box 2.
42
The construction of the aggregate risk weight for contributions to the SRF differs in several aspects
from the EBA Guidelines. For instance, for the normalisation method the relevant Commission
delegated regulation prescribes a bucketing approach, applies several rescalings and makes use of
geometric and linear aggregation methods. For a full description of the methodology see Commission
Delegated Regulation (EU) 2015/63.
ECB Occasional Paper Series No 208 / April 2018
27
inclusion of this variable means that banks which are likely to go into resolution
experience a reduction in contributions; the rationale being that these banks are
expected to cause less exposure for EDIS. A reduction in contributions following the
inclusion of the MREL indicator can be observed, for instance, for French, Spanish
and Dutch banks, and could potentially constitute an alternative to a target level
reduction in favour of countries in which a significant portion of banks is likely to go
into resolution rather than insolvency. This approach would have the advantage of
not reducing the overall target level of EDIS, thereby maintaining its level of
resilience. It should be noted that, as our analysis is based on banks’ current risk
profiles, and given this composition of indicators, contributions from larger banks are
likely to decrease further in the future when MREL buffers are built up.
Column 6 shows the results under the same DGS methodology, but also including an
indicator for interconnectedness. This would allow the risks posed by the failure of
interconnected banks to the rest of the banking system, and hence to EDIS, to be
taken into account. Column 7 shows the results with the inclusion, on top of the
baseline, of an indicator for NPLs, the NPL ratio (non-performing loans and
advances over total gross loans and advances as reported in FINREP). The ratio is
only included for 134 banks, including 89 SIs and 45 LSIs, and for the other banks in
the sample it is set to zero. While the inclusion of this indicator does not allow us to
draw definitive conclusions regarding the banks’ contributions due to data limitations,
NPLs are likely to be an important indicator affecting the possibility of cross-
subsidisation within EDIS.
In addition, column 8 reports risk-based contributions when the World Bank index for
the strength of insolvency frameworks is included. This is relevant given the impact
of insolvency regimes in allowing DGSs to effectively recoup resources in insolvency
proceedings. The inclusion of the index modifies the contributions of banking
systems in most Member States.
A comparison of contributions based on the SRF method in column 9 with non-risk
adjusted contributions (based on covered deposits only) in column 3 and with
contributions calculated using the EBA Guidelines (column 4) suggests that
contributions based on the SRF methodology could be less risk sensitive than those
based on the DGS method. SRF-based contributions always fall in the range
between risk-unadjusted and DGS-method-based contributions. The former are
closer or equally close to the risk-unadjusted contributions as DGS-method-based
contributions for all countries. This points towards a lower risk sensitivity of the SRF
methodology. The reason for this finding is the effect of the rescaling of the
aggregate risk scores. While it is used to calculate contributions with the SRF
methodology, in this analysis a final rescaling to the risk scores under the DGS
approach is not applied.
43
Indeed, when rescaling risk scores, prior to the
contribution calculation, to the same range which is used in the SRF methodology
[0.8, 1.5], the contributions converge towards non-risk adjusted contributions.
43
The EBA Guidelines do not explicitly require risk scores to be rescaled. Furthermore, the boundaries
can be chosen freely. Choosing a range of [0,100] would leave the risk scores unchanged.
ECB Occasional Paper Series No 208 / April 2018
28
Furthermore, they are even closer to non-risk adjusted contributions than those
based on the SRF methodology. Box 2 provides further insights into this topic.
Table 3
Contributions by country based on DGS sliding scale methodology
Source: ECB staff calculations based on COREP, FINREP and Bankscope data, 2015:Q4.
Note: The DGS-baseline calculation includes the following indicators: total risk-based capital ratio, leverage ratio, highly liquid assets per total assets, RWA per total assets, MREL-
eligible liabilities (only senior unsecured bonds), and ROE.
Box 2
Aggregate risk weight construction – DGS vs. SRF methodology
The method used for the construction of the aggregate risk weight (ARW) is a decisive part of the
determination of banks' contributions to the Fund as was shown by comparing contributions based
on the EBA method to those determined using the SRF methodology in Table 3. This box aims to
investigate the impact of several steps taken under the two methodologies to obtain the ARW,
which is a composite indicator.
The two methodologies differ, first, in the normalisation method used. The normalisation method is
applied to all individual indicators before aggregation into one composite indicator. While the DGS
sliding scale method uses the distance from some benchmark value (here: the 75th and 25th
percentile), the SRF makes use of a bucketing method, where banks are assigned to different bins
based on their relative performance regarding the various individual indicators. The two approaches
1) 2) 3) 4) 5) 6) 7) 8) 9)
Country
Number
of
banks
0.8% of covered
deposits DGS - Baseline
DGS without
MREL indicator
DGS plus
Interconnectedness
DGS plus NPL
ratio
DGS plus
insolvency
indicator
SRF methodology
with baseline
indicator set
In EUR
billion
In
percent
of fund
In EUR
billion
In
percent
of fund
In EUR
billion
In
percent
of fund
In EUR
billion
In
percent
of fund
In EUR
billion
In
percent
of fund
In EUR
billion
In
percent
of fund
In EUR
billion
In
percent
of fund
AT
33 0,6 1,56% 0,5 1,38% 0,6 1,46% 0,6 1,45% 0,5 1,38% 0,6 1,57% 0,6 1,50%
BE
11 0,7 1,87% 0,5 1,28% 0,5 1,33% 0,6 1,47% 0,5 1,28% 0,6 1,59% 0,6 1,69%
CY
3 0,1 0,24% 0,1 0,34% 0,1 0,25% 0,1 0,27% 0,2 0,42% 0,1 0,34% 0,1 0,31%
DE
1180 9,8 25,72% 12,5 32,92% 11,1 29,17% 11,3 29,65% 11,8 31,19% 9,5 24,99% 10,5 27,71%
EE
6 0,002 0,01% 0,002 0,01% 0,001 0,00% 0,001 0,00% 0,002 0,01% 0,002 0,01% 0,002 0,01%
ES
20 7,2 19,00% 7,9 20,85% 8,3 21,93% 7,9 20,72% 7,9 20,80% 8,4 22,13% 7,4 19,55%
FI
8 0,2 0,48% 0,1 0,34% 0,1 0,35% 0,1 0,36% 0,1 0,33% 0,1 0,31% 0,2 0,43%
FR
23 9,6 25,38% 6,6 17,52% 7,1 18,80% 7,7 20,33% 6,6 17,47% 8,2 21,71% 8,6 22,67%
GR
6 0,8 2,12% 1,4 3,68% 1,5 3,94% 1,2 3,27% 1,7 4,38% 1,3 3,49% 1 2,76%
IE
5 0,3 0,70% 0,3 0,70% 0,3 0,69% 0,3 0,71% 0,3 0,79% 0,3 0,76% 0,3 0,69%
IT
297 4 10,50% 4,2 11,13% 4,5 11,85% 4,2 10,99% 4,5 11,98% 4,5 11,93% 4,1 10,90%
LT
5 0,1 0,19% 0,04 0,11% 0,01 0,05% 0,03 0,10% 0,04 0,11% 0,1 0,14% 0,1 0,16%
LU
17 0,1 0,23% 0,1 0,19% 0,1 0,16% 0,1 0,20% 0,1 0,18% 0,1 0,22% 0,1 0,22%
LV
16 0,05 0,13% 0,03 0,10% 0,01 0,05% 0,03 0,09% 0,03 0,10% 0,04 0,12% 0,04 0,12%
MT
6 0,01 0,03% 0,01 0,03% 0,01 0,03% 0,01 0,03% 0,01 0,03% 0,01 0,04% 0,01 0,03%
NL
18 3,7 9,67% 2,3 6,02% 2,4 6,39% 2,8 7,27% 2,3 5,94% 3 7,78% 3,3 8,64%
PT
12 0,7 1,93% 1,2 3,18% 1,3 3,33% 1,1 2,86% 1,3 3,34% 1 2,63% 0,9 2,37%
SI
6 0,1 0,22% 0,1 0,16% 0,1 0,16% 0,1 0,16% 0,1 0,19% 0,1 0,20% 0,1 0,20%
SK
3 0,01 0,04% 0,02 0,06% 0,02 0,05% 0,01 0,05% 0,02 0,06% 0,02 0,06% 0,01 0,04%
Total
1675 38 100% 38 100% 38 100% 38 100% 38 100% 38 100% 38 100%
ECB Occasional Paper Series No 208 / April 2018
29
differ in that the former one preserves information on the relative difference in observations of the
raw indicators, whereas the latter one only preserves some. The bucketing method can be thought
of as discretising the observations by clustering them on different levels; in contrast, the DGS
approach preserves continuity of the indicators (see OECD and JRC (2008) for further
normalisation methods).
As regards the choice of weights, both approaches rely on expert opinion and prescribe weights to
indicators. As noted by OECD and JRC (2008), weights are a crucial part of aggregation. In general
weights express the individual indicator’s importance for the relevant object of measurement.
Statistical methods can be used to determine weights from the data in order to correct for
correlation between individual variables. This might be desirable to avoid double counting of the
correlated part of the variables.
Another difference in the two methodologies lies in the choice of the aggregation method. In
contrast to the SRF methodology, which uses a linear aggregation method for the within pillar
aggregation and a geometric one for the across pillar aggregation
44
, the EBA Guidelines prescribe a
linear aggregation method only. Linear aggregation allows for making up for bad performance in
one pillar by a better performance in another pillar. The compensability indeed depends on the
pillar-specific weights, but it is independent of the indicator value of the pillar.
45
When using a
geometric average the marginal impact of a change in an indicator on the composite indicator value
depends on the size of the indicator. An increase in a low indicator has a higher impact on the
composite indicator than an increase in a high indicator value.
46
Since under the SRF methodology
the geometric average is used to aggregate indicators where higher values signify a better
performance in the respective category, bad performance in one field cannot easily be offset by an
even better performance in a pillar where the bank already performs well. This gives incentives to
improve the weaknesses in a bank's risk profile. The use of a geometric average for the DGS
method would be counterintuitive since the indicators to be aggregated are such that a higher value
identifies a riskier bank. Thus, an improvement in a risky field would have a relatively small
decreasing impact on the composite indicator.
Both guidelines consider a rescaling of the composite indicator (which was not applied under the
DGS methodology in this paper). While rescaling to a fixed range of [0.8, 1.5] constitutes a
mandatory step in the SRF methodology, the EBA Guidelines leave it to the discretion of the
responsible authority to decide on the range of rescaling and whether rescaling is applied in the first
place. One option presented in the EBA Guidelines for a rescaling of the aggregate risk score
(ARS) to obtain the ARW is a linear transformation of the form ARW = (a b) ARS/100 + b,
where b is the lower and a the upper bound.
44
The linear aggregation looks as follows: PI
=
where
determines the aggregated
indicator for pillar , and
is the weight for the -th indicator
in pillar . The geometric aggregation
method means application of the following formula: I =
with I denoting the composite risk
indicator for a given bank, and
is the weight assigned to pillar .
45
This can be seen when looking at the derivatives of the linear aggregation formula. We have:
=
so that an increase in a pillar value has a marginal impact on the composite indicator equal to its weight
independent of the initial size of the pillar.
46
Compare the 2nd derivative:
=
…
(
1)
< 0. Thus, the
higher the initial pillar value, the lower its marginal impact on the composite indicator.
ECB Occasional Paper Series No 208 / April 2018
30
When rescaling the ARSs a change in contributions can be observed. Chart A shows the change in
contributions due to rescaling as a function of ARS. The change in contribution is defined as the
fraction of contributions based on rescaled risk scores [
] with respect to contributions based
on non-rescaled risk scores [
]:
/
. The chart shows the effect of
rescaling for four different rescaling ranges: [0.75, 1.5] (used as an example in the EBA Guidelines),
[0.1, 1.5], [0.75, 5], and [0.1, 5]. The black line indicates no change in contributions due to rescaling.
Chart A
Change in contributions (
/
) due to rescaling
(y-axis: C_rescaled/C_nonrescaled; x-axis: ARS)
Source: ECB staff calculations based on COREP, FINREP and Bankscope data, 2015:Q4.
All graphs reveal a negative relationship between the change in contributions due to rescaling, and
riskiness (measured by the ARS). Focusing on the interval given as an example in the EBA
Guidelines [0.75, 1.5], for instance, the safest banks are confronted with a contribution after
rescaling which is up to 21 times as high as without rescaling, while the riskiest bank sees a
contribution after rescaling that is 0.59 times its non-rescaled contribution. In other words, the
higher the ARS (meaning the riskier a bank is), the higher is the reduction in the aggregate risk
score. The reason is that the change in the composite indicators due to rescaling differs across
banks ARW_i/ARS_i ARW_j / ARS_j . This difference then feeds through to the contributions.
The degree of distortion differs with the rescaling range. There is no change in contributions when
the lower bound is set to zero, irrespective of the value chosen for the upper bound. The reason is
that a zero lower bound only causes the change in contributions to be non-bank-specific such that
ARW_i/ARS_i = ARW_j / ARS_j for all banks. A negative lower bound leads to a reversed effect: the
safest banks profit while the riskiest banks pay more. However, this transformation leads to
negative contributions, a payout from the Fund, for the safest banks.
For intervals with a lower bound above zero the broader the interval, the less pronounced the
negative effect on safer banks and also the more muted the positive effect on riskier banks. For
example, the orange dots, which refer to the broadest range depicted, [0.1, 5], can be shown to be
always closer to the red line than the blue ones, representing the narrowest interval in the sample
[0.75, 1.5]. This can be interpreted as a lower degree of distortion for the broader interval as
contributions based on rescaled risk scores are closer to contributions based on non-rescaled risk
scores for all ARS. A decrease in the degree of distortion can also be seen when only broadening
0
5
10
15
20
0 20 40 60 80 100
Interval [0.75, 1.5]
Interval [0.1, 1.5]
Interval [0.75, 5]
Interval [0.1, 5]
ECB Occasional Paper Series No 208 / April 2018
31
one side of the interval at a time. The green and red dots display a broadening of the interval on
one side of the narrowest interval each and show both a decrease in distortion as compared to the
narrowest interval.
All in all, rescaling diminishes risk sensitivity of the risk weights and contributions. It would,
however, be desirable to design risk weights that are able to precisely distinguish between risky and
safe banks in order to give the right incentives. In their article on deposit insurance pricing options
for the FDIC, Bloecher et al. (2003) define risk sensitivity to be the “primary objective of deposit
insurance pricing reform” as it would reduce subsidies paid by low-risk institutions to riskier ones
and moderate incentives for increased risk taking and it “would achieve the goal of making the
deposit insurance system more equitable and economically efficient”.
5.4 Distribution of contributions to the DIF across banks
Another key issue relates to the distribution of EDIS contributions across banks and
to the question of whether the risk-based contributions should be further lowered for
smaller banks (to ensure the proportionality principle) or whether the target level
should be lowered for larger banks (as these are less likely to benefit from EDIS).
Table 4 sheds light on these issues showing, for each decile of banks’ total assets in
our sample of 1,675 banks, the sum of contributions, the average contribution per
euro of covered deposits and the smallest and largest value of contributions per
covered deposits. Column 3 provides the aggregate amount of contributions paid by
banks in each decile. The numbers suggest that the smallest 10 percent of banks in
our sample would pay €0.11 billion or 0.28% of the €38 billion target size of EDIS. In
contrast, the largest 10% of banks would pay €28.5 billion or 75.09% of the overall
EDIS target. These numbers need to be put in relation to the actual covered deposits
of the banks in each decile to check whether the largest banks bear the brunt of the
cost of EDIS. In fact, column 4 suggests that the smallest and the largest banks’
contribution per covered deposits on their balance sheet is relatively low on average
with 1 cent (approximately) and 0.83 cent per euro, respectively.
47
It is rather the
banks in the intermediate decile range that pay slightly more ranging from 1 to 1.14
cents per euro of covered deposits. This finding is further underpinned by the range
of the largest and the smallest contributions per euro of covered deposits in column
5 which demonstrates that the range for each decile is by and large comparable. In
sum, the evidence in Table 4 indicates that smaller and larger banks would not
excessively contribute to EDIS, relative to their amount of covered deposits.
47
To note, this contribution is not a “one-off” but is rather built-up over the years as EDIS is being filled.
ECB Occasional Paper Series No 208 / April 2018
32
Table 4
Contribution distribution across different bank sizes
1) 2) 3) 4) 5)
Decile group by
total assets Interval total assets
Total
contribution
per decile
group
Contribution
per covered
deposits,
average
Contribution per covered
deposits, interval
Smallest in
EUR billion
Largest in EUR
billion
In EUR billion
(in % of EDIS
target size) Lower bound Upper bound
10 0,02 0,15 0,11 (0,28%) 0,0097 0,0024 0,0183
20 0,15 0,26 0,25 (0,65%) 0,0107 0,0035 0,0176
30 0,26 0,38 0,34 (0,90%) 0,0104 0 0,0181
40 0,38 0,56 0,48 (1,25%) 0,011 0,003 0,0184
50 0,56 0,76 0,61 (1,62%) 0,01 0,0003 0,0182
60 0,76 1,08 0,96 (2,52%) 0,0109 0 0,019
70 1,09 1,66 1,39 (3,66%) 0,0114 0,0024 0,0185
80 1,66 2,77 2,03 (5,34%) 0,0109 0,0024 0,0178
90 2,77 6,49 3,3 (8,69%) 0,0104 0,0007 0,0183
100 6,6 1807,57 28,5 (75,09%) 0,0083 0,0003 0,0165
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: Each decile corresponds to 167 or 168 banks; Contribution based on DGS baseline indicator set. The value in column 1
indicates the upper bound of the interval the observation belongs to. For example, "20" refers to all banks with total assets being
above the 10th and below or equal to the 20th percentile.
5.5 Cross-subsidisation
Tables 5, 6 and 7 provide information about the magnitude of cross-subsidisation
across banking systems under simulated crises of different magnitude. First, the 3%
riskiest banks (51 banks, representing the 12.66%of the total assets of the entire
euro area sample) are assumed to fail and to be hit by losses from 5% (7.5%) to
25% (37.5%) of total assets in resolution (insolvency). Such loss rates,
simultaneously applied to a set of banks, are significantly higher than the averages
observed in the last crisis (see Section 4.2.2). The same loss levels are then applied
to the 10% riskiest banks failing (167 banks, representing approximately the 40% of
the total assets of the entire euro area sample), thus simulating an extremely severe
crisis.
Table 5 reports the results referred to the 3% riskiest banks failing, with contributions
based on the DGS-baseline indicators and resolution scenario B with SRF
contribution. The first two columns refer to a given LGD value under resolution and
insolvency. Rows 1, 3, 5, 7 and 9 show the EDIS exposure per euro contributed for
banking systems in each country: this is equal to the sum of the EDIS exposure
posed by all banks located in a country divided by the sum of the contributions paid
by the same banks. A value exceeding one (red cells in the Tables below) indicates
that the banks in a country would contribute less than what they would receive from
ECB Occasional Paper Series No 208 / April 2018
33
EDIS in the simulated crisis when 3% of the riskiest banks fail.
48
Rows 2, 4, 6, 8 and
10 show EDIS exposure in euro billion for given LGD values. For a crisis in which 3%
of the banks in the sample fail simultaneously, the analysis shows that there is never
an EDIS exposure and, consequently, no cross-subsidisation for loss rates in
resolution up to 15% of total assets. Some EDIS exposure only starts to materialise
in Spain and Greece for loss rates in resolution of 20% of total assets, but without
cross-subsidisation. The only evidence of cross-subsidisation is found, in Spain and
Greece, for loss rates of 25% in resolution (37.5% in insolvency), which reproduce a
scenario much harsher than the 2007-2009 global financial crisis.
Table 6 reports the results referred to the 10% riskiest banks failing and loss rates in
resolution (insolvency) equal to 5% (7.5%) and 10% (15%): even though the number
of banks affected by losses is now approximately three times bigger than in Table 5,
with these loss rates the EDIS exposure continues to be equal to zero and there is
no evidence of cross-subsidisation in any country. It must be stressed that the 167
banks which are assumed to fail under this scenario have assets corresponding to
about 40% of the total assets of euro area banks in the sample: hitting all these 167
banks simultaneously with losses in resolution (insolvency) of 5% (7.5%) and 10%
(15%) of total assets creates a scenario much more severe than the last crisis.
Therefore, the 10% scenario can be described as a black swan event.
49
Table 7 reports the results referred to a "big black swan" scenario: in particular, it
shows the EDIS exposure in case the 10% riskiest banks simultaneously fail and are
simultaneously affected by losses in resolution (insolvency) from 15% (22.5%) to
25% (37.5%) of total assets. The red cells indicate exposures per euro contributed
above 1, meaning that banks in a country, for a certain loss rate, would contribute
less than what they would receive from EDIS in case of a crisis, thus indicating
cross-subsidisation.
The first evidence of EDIS exposure is only found for loss rates of 20% in resolution
(30% in insolvency); but only one country, Belgium, would receive more funds from
EDIS than it would contribute, thus showing very limited cross-subsidisation despite
the severity of the simulated crisis. Losses equal to 25% of total assets in resolution
(37.5% in insolvency) are necessary to observe cross-subsidisation in more
countries, precisely in Belgium, Cyprus, Spain and Greece. Given the high loss
rates
50
necessary to produce cross-subsidisation (and given that such rates are set
equally high for all banks, making the crisis more severe than in actual crises where
48
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. Among the
various caveats, as explained in Section 4.2.1, is the fact that the coefficients to calculate PDs are
estimated using through-the-cycle data while PDs are obtained using point-in-time data for the
independent variables. 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.
49
The black swan theory has been introduced by Nassim Nicholas Taleb in its 2007 book "The Black
Swan: The Impact of the Highly Improbable". According to Taleb, a black swan event is characterised
by the following three features: rarity, extreme impact, and retrospective (though not prospective)
predictability.
50
As a term of comparison, see the loss rates referred to the last crisis in Section 4.2.2.
ECB Occasional Paper Series No 208 / April 2018
34
loss rates vary across banks), these findings suggest that there is no unwarranted
systematic cross-subsidisation via EDIS in the steady state.
51
51
An assessment of the possible cross-subsidisation in the transition to the steady state would require a
different quantitative analysis and is not the object of this paper. Furthermore, due to the lack of
granular information on the nationality of deposit holders, no specific consideration and treatment have
been given to cross-border deposits, i.e. deposits at one bank headquartered in a Member State which
are held by counterparts from other Member States. This implies that the results could in principle
overestimate the extent of cross-subsidisation.
ECB Occasional Paper Series No 208 / April 2018
35
Table 5
Cross-subsidisation: Fund exposure per euro contributed and in EUR billion - 3% riskiest banks failing
Loss-absorbency scenario B in resolution with 97th-percentile crisis simulation with SRF contribution; banks’ contributions to EDIS based on DGS sliding scale method and DGS-baseline indicators
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
in EUR bn
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure
5% 7,50% (1) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(2) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
10% 15% (3) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(4) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
15% 22,50% (5) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(6) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
20% 30% (7) per EUR
contributed
0 0 0 0 0 0,26 0 0 0,63 0 0 0 0 0 0 0 0 0 0
-
(8) in EUR bn 0 0 0 0 0 2,1 0 0 0,9 0 0 0 0 0 0 0 0 0 0
3
25% 37,50% (9) per EUR
contributed
0 0 0 0 0
1,04
0 0
11,2
0 0,01 0 0 0 0 0 0 0 0
-
(10) in EUR bn 0 0 0 0 0
8,2
0 0
15,5
0 0,04 0 0 0 0 0 0 0 0
23,8
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country.
ECB Occasional Paper Series No 208 / April 2018
36
Table 6
Cross-subsidisation: Fund exposure per euro contributed and in EUR billion - 10% riskiest banks failing
Loss-absorbency scenario B in resolution with 90th-percentile crisis simulation with SRF contribution; banks’ contributions to EDIS based on DGS sliding scale method and DGS-baseline indicators
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
in EUR bn
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure
5% 7,50% (1) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(2) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
10% 15% (3) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(4) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country.
ECB Occasional Paper Series No 208 / April 2018
37
Table 7
Cross-subsidisation: Fund exposure per euro contributed and in EUR billion - 10% riskiest banks failing
Loss-absorbency scenario B in resolution with 90th-percentile crisis simulation with SRF contribution; banks’ contributions to EDIS based on DGS sliding scale method and DGS-baseline indicators
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
in EUR bn
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure
15% 22,50% (1) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(2) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
20% 30% (3) per EUR
contributed
0
1,37
0 0 0 0,26 0 0 0,63 0 0 0 0 0 0 0 0 0 0
-
(4) in EUR bn 0
0,7
0 0 0 2,1 0 0 0,9 0 0 0 0 0 0 0 0 0 0
3,7
25% 37,50%
(5) per EUR
contributed
0
6,15 2,03
0 0
1,45
0 0
11,2
0 0,01 0 0 0 0 0 0 0 0
-
(6) in EUR bn 0
3 0,3
0 0
11,5
0 0
15,6
0 0,04 0 0 0 0 0 0 0 0
30,4
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country.
ECB Occasional Paper Series No 208 / April 2018
38
5.5.1 Cross-subsidisation with country shocks
The results presented in the previous Section refer to simulated losses applied to the
3% and 10% riskiest banks at the banking union level. The selection of the 3% and
10% riskiest banks in the sample of euro area banks is not based on any
country-specific shocks and is carried out only on the basis of the riskiness of banks
relative to all the other euro area banks in the sample. This Section provides
additional and complementary information, showing the results on
cross-subsidisation in a scenario where country-specific shocks are simulated.
The country-specific shocks are designed to mimic the European shock simulated
above, on the basis of the share of total assets of the banks assumed to fail.
Therefore, the first set of results has been obtained by assuming that the riskiest
banks in each country, representing up to the 13% of the domestic banking system
total assets, simultaneously fail. The choice of the 13% total assets threshold is
based on the fact that the 3% riskiest banks assumed to fail in the euro area
scenario (results reported in Table 5) represent 13% of the total assets of the euro
area banks in the sample. The second set of results is obtained under the
assumption that the riskiest banks in each country, representing up to the 40% of the
domestic banking system total assets, simultaneously fail. This choice is in line with
the scenario with the 10% riskiest banks failing at the euro area level (results
reported in Table 6 and 7), whose assets represent approximately 40% of total
assets of euro area banks in the sample.
Table 8 presents the results, expressed in terms of EDIS exposure and
cross-subsidisation, referred to the assumption that the riskiest banks in each
country, representing the 13% of the domestic bank total assets, simultaneously fail.
First, regardless of the severity of the loss rates imposed to the failing banks, the DIF
would be sufficient to cover all financial needs, even in case the country-specific
shocks occur all at the same time. Second, cross-subsidisation becomes material in
Cyprus and Spain only for loss rates equal to 25%; on the other hand,
cross-subsidisation in Luxembourg appears to be higher, and already visible with
losses equal to the 15% of total assets in resolution (22.5% in insolvency).
Table 9 reports the results referred to the assumption that the riskiest banks in each
country, representing the 40% of the domestic bank total assets, simultaneously fail.
Even if the crisis simulated in this scenario is extremely severe - the 40% of the total
assets of each country simultaneously stressed with losses from 5% to 25% (in
resolution) of the banks' balance sheet - the overall exposure of the deposit
insurance fund remains well below its target size (equal to € 38bn for the sample). In
terms of cross-subsidisation, only for very high loss rates five countries would
receive payouts from EDIS of more than what they would pay into EDIS: Cyprus,
Spain and Luxembourg (consistent with Table 8) and Belgium and Malta (starting
with loss rates equal to 20% in resolution).
Even though the 13% and 40% thresholds calibrated on total assets are consistent
with the crisis simulated, respectively, with the 3% and 10% riskiest banks failing at
ECB Occasional Paper Series No 208 / April 2018
39
the euro area level (Section 5.5, Tables 5, 6 and 7), the results differ in some
countries: in particular, and differently from the results in Tables 5 and 7, Tables 8
and 9 reveal no evidence of cross-subsidisation in Greece. This can be explained by
the fact that when a sample of riskiest banks is selected at the euro area level, such
banks may be concentrated in some countries, thus representing a very high portion
of the domestic bank total assets; when the same share of riskiest banks is instead
selected at the national level, it may happen that some banks, even if risky at the
euro area level, are not included among the riskiest ones at the domestic level.
Finally, the comparison of the results reported in Tables 8 and 9 reveals that, even
though the total assets of the failing banks in the second scenario (Table 9) is much
higher than the total assets of the failing banks in the first scenario (Table 8), the
overall (aggregate across the euro area) increase in the EDIS exposure is not
substantial (in the most severe crisis - loss rate equal to 25% in resolution - the EDIS
exposure increases from € 8.9 bn to € 11.9 bn): this can be explained by the fact that
EDIS exposure is mainly due to the 3% riskiest banks in each country. For extremely
high loss rates (20% and 25% in resolution), cross-subsidisation would materialise
also for Belgium and Malta in Table 9, but there would be no change for Cyprus,
Spain and Luxembourg relative to Table 8.
Overall, the key results of the analysis are consistent across both a banking union
shock and country-specific shocks: first, an EDIS exposure would only materialise for
very high loss rates - much higher than experienced in the 2007-2009 global
financial crisis; second, cross-subsidisation would generally materialise only for
extremely high loss rates, and only few countries would be affected
52
. Therefore, this
analysis confirms that there would be no unwarranted systematic cross-subsidisation
via EDIS in the steady state.
52
This conclusion is confirmed by an additional analysis that defines the country shock as the 10%
riskiest banks failing in each country, consistently with the scenario simulated at the euro area level
(see Tables 6 and 7). Even in this adverse situation (on average, the 10% riskiest banks failing
correspond to the 30% of domestic bank total assets), cross-subsidisation materialises only in Belgium
for loss rates in resolution (insolvency) of at least 20% (30%).
ECB Occasional Paper Series No 208 / April 2018
40
Table 8
Cross-subsidisation: Fund exposure in EUR billion and per euro contributed - riskiest banks in each country, representing up to the 13% of the domestic total assets, failing
Loss-absorbency scenario B in resolution with SRF contribution; banks’ contributions to EDIS based on DGS sliding scale method and DGS-baseline indicators
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
in EUR bn
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure
5%
7,50%
(1) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(2) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
10%
15%
(3) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0,76 0 0 0 0 0 0
-
(4) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0,05 0 0 0 0 0 0
0,05
15%
22,50%
(5) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0
2,22
0 0 0 0 0 0
-
(6) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0
0,16
0 0 0 0 0 0
0,16
20% 30% (7) per EUR
contributed
0 0 0 0 0 0,26 0 0 0 0 0 0
3,68
0 0 0 0 0 0
-
(8) in EUR bn 0 0 0 0 0 2,07 0 0 0 0 0 0
0,26
0 0 0 0 0 0
2,33
25% 37,50% (9) per EUR
contributed
0 0
2
0 0
1,04
0 0 0 0 0,01 0
5,14
0 0 0 0 0 0
-
(10) in EUR bn 0 0
0,26
0 0
8,25
0 0 0 0 0,04 0
0,36
0 0 0 0 0 0
8,91
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country.
ECB Occasional Paper Series No 208 / April 2018
41
Table 9
Cross-subsidisation: Fund exposure in EUR billion and per euro contributed - riskiest banks in each country, representing up to the 40% of the domestic total assets, failing
Loss-absorbency scenario B in resolution with SRF contribution; banks’ contributions to EDIS based on DGS sliding scale method and DGS-baseline indicators
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
in EUR bn
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure
5%
7,50%
(1) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
-
(2) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
10%
15%
(3) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0 0,76 0 0 0 0 0 0
-
(4) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0 0,05 0 0 0 0 0 0
0,05
15%
22,50%
(5) per EUR
contributed
0 0 0 0 0 0 0 0 0 0 0 0
2,22
0 0,66 0 0 0 0
-
(6) in EUR bn 0 0 0 0 0 0 0 0 0 0 0 0
0,16
0 0,01 0 0 0 0
0,16
20% 30% (7) per EUR
contributed
0
1,38
0 0 0 0,26 0 0 0 0 0 0
3,68
0
1,99
0 0,01 0 0
-
(8) in EUR bn 0
0,66
0 0 0 2,07 0 0 0 0 0 0
0,26
0
0,02
0 0,01 0 0
3,02
25% 37,50% (9) per EUR
contributed
0
6,21 2
0 0
1,04
0 0 0 0 0,01 0
5,14
0
3,32
0 0,02 0 0
-
(10) in EUR bn 0
2,98 0,26
0 0
8,25
0 0 0 0 0,04 0
0,36
0
0,03
0 0,02 0 0
11,94
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country.
ECB Occasional Paper Series No 208 / April 2018
42
6 A mixed deposit insurance system
As discussed in Section 3, several proposals on EDIS have introduced the principle
of national DGSs bearing the first burden before the European deposit insurance
fund steps in. This Section provides quantitative information on how the risk-based
contributions would change under such a "mixed" deposit insurance scheme, and
how EDIS exposure and cross-subsidisation would be affected.
6.1 Contributions
First, the mixed deposit insurance scheme is assumed to be funded, with an equal
share of 0.4% of covered deposits, at the national level and at the European level
(the overall target level remains 0.8% of covered deposits). The key change relative
to a fully-fledged EDIS is that the contributions paid by banks to reach the national
0.4% target would be risk-based relative to their national benchmark, and not to the
riskiness of the entire euro area banking system. Therefore, for the purpose of the
national 0.4% target, contributions would still be risk-based, but this would only affect
the distribution of contributions among domestic banks, while the overall national
target would be risk-insensitive and would be fixed at 0.4% of covered deposits in
the domestic banking system. This implies that some banking systems would end up
paying overall less than they would under a fully-fledged EDIS; on the other hand,
some banking systems would pay overall more than they would under a fully-fledged
EDIS. This is illustrated in Table 10, which shows the contributions paid by banking
systems under both a fully-fledged EDIS and under a mixed deposit insurance
scheme: the latter is divided into contributions to the national funds (equal to 0.4% of
domestic covered deposits) and to the European compartment (equal to the
remaining 0.4% of euro area covered deposits). The last two columns report the
overall contributions banking systems would pay under the mixed scheme: the red
cells, in particular, identify the banking systems that would pay more under the mixed
scheme.
The results indicate that under a mixed deposit insurance scheme Cyprus, Germany,
Spain, Greece, Ireland, Italy and Portugal would pay less than under a fully-fledged
EDIS, while Austria, Belgium, Finland, France, Lithuania, Luxembourg, Malta, the
Netherlands and Slovenia would pay more.
ECB Occasional Paper Series No 208 / April 2018
43
Table 10
Contributions by country based on DGS sliding scale methodology and DGS baseline indicators: fully-fledged EDIS and mixed deposit insurance scheme
1) 2)
3) 4)
Fully-fledged EDIS
Mixed deposit insurance system
National target level European target level National + European target level
Country Number of banks
In EUR billion
In percent of
fully-fledged EDIS fund
(DIF) In EUR billion
In percent of national
compartments In EUR billion
In percent of European
compartment In EUR billion In percent of mixed fund
AT
33 0,52 1,37% 0,3 1,58% 0,26 1,37%
0,56 1,48%
BE
11 0,48 1,26% 0,36 1,90% 0,24 1,26%
0,6 1,58%
CY
3 0,13 0,34% 0,05 0,26% 0,06 0,32% 0,11 0,29%
DE
1180 12,49 32,90% 4,88 25,71% 6,25 32,93% 11,13 29,32%
EE
6 0,002 0,01% 0,001 0,01% 0,001 0,01% 0,002 0,01%
ES
20 7,92 20,86% 3,61 19,02% 3,96 20,86% 7,57 19,94%
FI
8 0,13 0,34% 0,09 0,47% 0,06 0,32%
0,15 0,40%
FR
23 6,65 17,52% 4,82 25,40% 3,32 17,49%
8,14 21,44%
GR
6 1,39 3,66% 0,4 2,11% 0,7 3,69% 1,1 2,90%
IE
5 0,27 0,71% 0,13 0,68% 0,13 0,68% 0,26 0,68%
IT
297 4,23 11,14% 1,99 10,48% 2,11 11,12% 4,1 10,80%
LT
5 0,04 0,11% 0,04 0,21% 0,02 0,11%
0,06 0,16%
LU
17 0,07 0,18% 0,04 0,21% 0,04 0,21%
0,08 0,21%
LV
16 0,04 0,11% 0,02 0,11% 0,02 0,11% 0,04 0,11%
MT
6 0,01 0,03% 0,01 0,05% 0,01 0,05%
0,02 0,05%
NL
18 2,29 6,03% 1,83 9,64% 1,14 6,01%
2,97 7,82%
PT
12 1,21 3,19% 0,37 1,95% 0,6 3,16% 0,97 2,56%
SI
6 0,06 0,16% 0,04 0,21% 0,03 0,16%
0,07 0,18%
SK
3 0,02 0,05% 0,01 0,05% 0,01 0,05% 0,02 0,05%
Total
1675 38 100% 19 100% 19 100% 38 100%
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: The contributions calculation includes the following indicators: total risk-based capital ratio, leverage ratio, highly liquid assets per total assets, RWA per total assets, MREL-eligible liabilities (only senior unsecured bonds), and ROE
ECB Occasional Paper Series No 208 / April 2018
44
6.2 Cross-subsidisation
The change in the overall contributions paid by banking systems produces
implications for cross-subsidisation. Given the same simulated shocks (3% or 10%
riskiest banks failing) and, consequently, given the same exposures of a deposit
insurance scheme, cross-subsidisation under a mixed system would increase in
those countries that would pay more contributions under a fully-fledged EDIS than
under the mixed system. This is illustrated in Table 11, which shows a comparison
between cross-subsidisation under a fully-fledged EDIS and under a mixed system in
the scenario with the 3% riskiest banks failing. Under the mixed system, cross-
subsidisation is measured as the ratio between the exposure of the European fund
and the contributions paid to the European fund only. The red cells indicate an
increase in cross-subsidisation under the mixed system, and this can be observed in
Spain and Greece for losses in resolution (insolvency) equal to 25% (37.5%) of total
assets.
ECB Occasional Paper Series No 208 / April 2018
45
Table 11
Cross-subsidisation: Fund exposure per euro contributed under a "mixed" deposit insurance scheme and fully-fledged EDIS - 3% riskiest banks failing
Loss-absorbency scenario B in resolution with SRF contribution; banks’ contributions based on DGS sliding scale method and DGS-baseline indicators, calibrated at the European level for the fully-fledged EDIS and for the European compartment in the "mixed" scheme,
calibrated at the national level for the national compartments in the "mixed" scheme
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
to "mixed"
scheme
(EUR bn)
0,56 0,6 0,11 11,13 0,002 7,56 0,16 8,14 1,1 0,26 4,11 0,06 0,08 0,04 0,01 2,98 0,97 0,07 0,02 38
Contributions
to fully-fledged
EDIS
(EUR bn)
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure per EUR contributed
5%
7,50%
(1) Mixed 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
(2) Full EDIS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
10%
15%
(3) Mixed 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
(4) Full EDIS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
15%
22,50%
(5) Mixed 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
(6) Full EDIS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
20% 30% (7) Mixed 0 0 0 0 0 0 0 0 0,68 0 0 0 0 0 0 0 0 0 0
(8) Full EDIS 0 0 0 0 0 0,26 0 0 0,63 0 0 0 0 0 0 0 0 0 0
25% 37,50% (9) Mixed 0 0 0 0 0
1,17
0 0
21,81
0 0 0 0 0 0 0 0 0 0
(10) Full EDIS 0 0 0 0 0
1,04
0 0
11,2
0 0,01 0 0 0 0 0 0 0 0
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country. Under the mixed system, cross-subsidisation is measured as the ratio between
the exposure of the European fund and the contributions paid to the European fund only.
ECB Occasional Paper Series No 208 / April 2018
46
Table 12 shows a comparison between cross-subsidisation under a fully-fledged
EDIS and under a mixed system in the scenario with the 10% riskiest banks failing.
Cross-subsidisation under a mixed scheme would increase in Belgium, Cyprus,
Spain and Greece in case of losses in resolution (insolvency) equal to 25% (37.5%)
of total assets. On the other hand, cross-subsidisation would decrease in Belgium for
losses in resolution equal to 20% of total assets: this can be explained by the fact
that there is a certain level of exposure below which the national compartment is
enough, or almost enough, to cover all the financial needs, thus reducing
cross-subsidisation.
ECB Occasional Paper Series No 208 / April 2018
47
Table 12
Cross-subsidisation: Fund exposure per euro contributed under a "mixed" deposit insurance scheme and fully-fledged EDIS - 10% riskiest banks failing
Loss-absorbency scenario B in resolution with SRF contribution; banks’ contributions based on DGS sliding scale method and DGS-baseline indicators, calibrated at the European level for the fully-fledged EDIS and for the European compartment in the "mixed" scheme,
calibrated at the national level for the national compartments in the "mixed" scheme
Loss
resolution
(%TA)
Loss
Insolvency
(%TA)
Country AT BE CY DE EE ES FI FR GR IE IT LT LU LV MT NL PT SI SK TOT
Contributions
to "mixed"
scheme
(EUR bn)
0,56 0,6 0,11 11,13 0,002 7,56 0,16 8,14 1,1 0,26 4,11 0,06 0,08 0,04 0,01 2,98 0,97 0,07 0,02 38
Contributions
to fully-fledged
EDIS
(EUR bn)
0,52 0,48 0,13 12,49 0,002 7,92 0,13 6,65 1,39 0,27 4,23 0,04 0,07 0,04 0,01 2,29 1,21 0,06 0,02 38
EDIS exposure per EUR contributed
5%
7,50%
(1) Mixed 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
(2) Full EDIS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
10%
15%
(3) Mixed 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
(4) Full EDIS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
15%
22,50%
(5) Mixed 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
(6) Full EDIS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
20% 30% (7) Mixed 0 1,27 0 0 0 0 0 0 0,68 0 0 0 0 0 0 0 0 0 0
(8) Full EDIS 0 1,37 0 0 0 0,26 0 0 0,63 0 0 0 0 0 0 0 0 0 0
25% 37,50% (9) Mixed 0
10,84 3,33
0 0
1,99
0 0
21,81
0 0 0 0 0 0 0 0 0 0
(10) Full EDIS 0
6,15 2,03
0 0
1,45
0 0
11,2
0 0,01 0 0 0 0 0 0 0 0
Source: ECB staff calculations based on COREP and Bankscope data, 2015:Q4.
Note: EDIS exposure per euro contributed for each country is calculated as the sum of EDIS exposures to all banks within a country divided by the sum of contributions of all banks' of that country. Under the mixed system, cross-subsidisation is measured as the ratio between
the exposure of the European fund and the contributions paid to the European fund only.
ECB Occasional Paper Series No 208 / April 2018
48
7 Conclusions
The analysis in this paper provides five main insights which support the policy
discussion on the introduction of a European Deposit Insurance Scheme.
First, a fully-funded deposit insurance fund with ex-ante contributions of 0.8% of
covered deposits (€38 billion in the sample analysed) would be sufficient to cover
payouts even in case of hypothetical losses much higher than the losses
experienced during the last crisis (2007-2009). Considering a scenario where the
riskiest 3% of euro area banks fail simultaneously and only MREL-eligible liabilities
are bailed-in (with the exception of large corporate deposits above €100,000), losses
in resolution and insolvency up to, respectively, 15% and 22.5% of banks’ total
assets are not enough to create exposures of the DIF in any country. The same
conclusion is reached when considering a scenario where the 10% riskiest banks in
the euro area simultaneously fail and are all affected by losses in resolution equal to
5%, 10% and 15% of total assets (corresponding to 7.5%,15% and 22.5% losses in
insolvency). These loss scenarios are considerably more severe than historical
losses both in Europe and in the United States, including the recent global financial
crisis. Exposures of the DIF are triggered only in case of an extremely severe crisis,
where the 3% or 10% riskiest banks simultaneously fail and are all affected by losses
in resolution at least equal to 20% of total assets, (corresponding to losses in
insolvency of 30%). Even in this extreme case, however, the DIF is never depleted.
Second, the specificities of banks and, as a result, also of banking systems can be
taken into account in the risk-based contributions to the deposit insurance fund,
which is preferable to the lowering of the EDIS target level. The methodology used in
this paper follows the EBA Guidelines for national DGS contributions but is applied to
all participating banks within the scope of EDIS to ensure that a bank’s risk profile is
compared to its peers across the entire banking union rather than only within each
national banking system. The features of banks and banking systems can be
appropriately reflected in the risk-based contributions using a “polluter pays”
approach. This would have the benefit of keeping the credible target level of EDIS,
which has been shown in Section 4 to be appropriate in dealing also with severe
banking crises. Furthermore, risk-based contributions would allow a wide range of
factors which are likely to have an impact on EDIS to be taken into account. For
example, including an indicator for MREL-eligible liabilities provides an indication of
banks’ loss-absorbing capacity and could also be a proxy for the likelihood of a bank
going into resolution rather than insolvency. Therefore, including this variable means
that banks that are likely to go into resolution may have their contributions reduced
because of their higher loss-absorbing capacities and the resulting potentially lower
exposure for EDIS. This could cater for the fact that a banking system composed of
larger institutions would be less likely to benefit from EDIS as these are more likely
to be resolved, thus limiting the possible contribution needed from the deposit
insurance fund. Also, the inclusion of an interconnectedness indicator would permit
the impact of a bank’s failure on the banking system as a whole, and therefore on
ECB Occasional Paper Series No 208 / April 2018
49
EDIS, to be taken into account. This would be particularly relevant for a banking
system composed mainly of interconnected institutions.
In addition, contributions are impacted by the choice of the calculation method.
Contributions based on the SRF methodology appear less risk sensitive than those
based on the EBA Guidelines without rescaling. However, when rescaling the latter
to the same range as under the SRF, contributions are even closer to non-risk-
adjusted contributions. This paper shows that rescaling has a diminishing effect on
risk sensitivity of the contributions: for intervals with positive lower bound the riskiest
banks profit at the expense of the safest banks.
Third, the analysis indicates that smaller and larger banks would not excessively
contribute to EDIS relative to the amount of covered deposits in their balance sheet,
suggesting that measures to reduce contributions for the smallest and/or largest
banks, as had been proposed by some Member States, would be unwarranted. As
regards the distribution of contributions across different banks, the 10% smallest
banks in the sample would pay €0.11 billion or 0.28% of the €38 billion target size of
EDIS. In contrast, the 10% largest banks would pay €28.5 billion or 75.09% of the
overall EDIS target. However, when comparing the contributions to the level of
covered deposits, the smallest and the largest banks’ contribution per covered
deposits on their balance sheet is relatively low with approximately 1 cent and 0.83
cent per euro of covered deposit respectively, while the banks in the intermediate
deciles range pay slightly more ranging from 1 to 1.14 cents per euro of covered
deposits.
Fourth, there is no unwarranted systematic cross-subsidisation within EDIS, in the
sense of some banking systems systematically contributing less than they would
benefit from the deposit insurance fund. A comparison of banks’ risk-based
contributions to the DIF exposure shows that, while there are some cases in which
the contributions of a banking system are lower than the amounts which would be
received from EDIS, this is only the case for very high loss rates that have a low
probability of occurring and for crises which would be much more severe than the
2007-2009 global financial crisis. This result holds also when considering country-
specific shocks, i.e. when the same share of banks in terms of total assets is
assumed to fail in each country, rather than at the euro area level. It should also be
noted that cross-subsidisation can be seen as a form of desirable risk-sharing in
more severe crises. This is different from a systematic unwarranted cross-
subsidisation and is in line with the purpose of an EDIS: pooling resources to
enhance the ability of the deposit insurance system, particularly in the more severe
crises, to withstand shocks better than under a system of national stand-alone
deposit guarantee schemes.
Fifth, the comparison between a fully-fledged EDIS and a mixed deposit insurance
scheme (where the national funds intervene before the European insurance fund)
reveals that the latter would increase cross-subsidisation. This result is the
consequence of some banking systems paying less under a mixed scheme, thus
building up a smaller pool of resources: the reason is that national target levels
depend only on the amount of covered deposits and are thus risk-insensitive.
ECB Occasional Paper Series No 208 / April 2018
50
In conclusion, EDIS would offer major benefits in terms of depositor protection while
posing limited risks in terms of EDIS exposure since the probability and magnitude of
interventions are likely to be low. It should be emphasised that EDIS will play a key
role in terms of confidence building, also avoiding risks of self-fulfilling prophecies on
bank runs. Additionally, based on the results shown in this paper, there is no risk of
unwarranted systematic cross-subsidisation.
The key elements driving these results are the following: first, a significant risk-
reduction and increase in loss-absorbing capacity have taken place in the aftermath
of the global financial crisis, including higher levels of capital, the build-up of other
loss-absorbing liabilities (which can absorb losses e.g. via bail-in) and the new
resolution framework. As the data in the analysis refer to year-end 2015, the banks'
loss-absorbing capacity is expected to further increase over time. Second, a super
priority for covered deposits will further contribute to protect EDIS. Finally, following a
"polluter-pays" approach, appropriately-designed risk-based contributions,
benchmarked at the euro area level, are crucial to establish the right incentives and
strike the right balance between ensuring adequate and credible deposit protection
and minimising cross-subsidisation across countries. Risk-based contributions can
and should internalise the specificities of banks and banking systems. This would
facilitate moving forward with risk-sharing measures in parallel with risk-reduction
measures, address moral hazard and avoid lowering EDIS capacity.
ECB Occasional Paper Series No 208 / April 2018
51
References
Allen, F., Carletti, E., Goldstein, I. and Leonello, A. (2017), "Government guarantees
and financial stability", ECB Working Paper Series No 2032, February.
Anginer, D., Demirgüç-Kunt, A. and Zhu, M. (2014), “How does deposit insurance
affect bank risk? Evidence from the recent crisis”, Journal of Banking & Finance, Vol.
48, pp. 312-321.
Bank of England (2015), The Bank’s approach to setting a minimum requirement for
own funds and eligible liabilities (MREL), Consultation Paper, December.
Bank of England (2016), The Bank of England’s approach to setting a minimum
requirement for own funds and eligible liabilities (MREL), Responses to Consultation
and Statement of Policy, November.
Beck, T. and Laeven, L. (2006), Resolution of failed banks by deposit insurers:
Cross-country evidence, World Bank Publications, 3920.
Bénassy-Quéré, A., Brunnermeier, M., Enderlein, H., Farhi, E., Fratzscher, M., Fuest,
C., Gourinchas, P., Martin, P., Pisani-Ferry, J., Rey, H., Schnabel, I., Véron, N.,
Weder di Mauro, B. and Zettelmeyer, J. (2018), Reconciling risk sharing with market
discipline: A constructive approach to euro area reform, CEPR Policy Insight No. 91,
January.
Bernet, B. and Walter, S. (2009), Design, structure and implementation of a modern
deposit insurance scheme, SUERF – The European Money and Finance Forum,
Vienna.
Betz, F., Oprica, S., Peltonen, T. and Sarlin, S. (2014), “Predicting distress in
European banks”, Journal of Banking & Finance, 2014, Vol. 45 (C), pp. 225–241.
Bloecher, E.P., Seale, G.A. and Vilim, R.D. (2003), “Options for pricing federal
deposit insurance”, FDIC Banking Review, Vol. 15, No 4.
Brewer III, E. and Mondschean, T.H., (1994), “An empirical test of the incentive
effects of deposit insurance”, Journal of Money, Credit, and Banking, Vol. 26, pp.
146-164.
Calomiris, C.W. and Jaremski, M. (2016a), “Deposit Insurance: Theories and Facts”,
Annual Review of Financial Economics, Vol. 8, pp. 97-120.
Calomiris, C.W. and Jaremski, M. (2016b), Stealing Deposits: Deposit Insurance,
Risk-Taking and the Removal of Market Discipline in Early 20th Century Banks,
Working Paper.
Cook, D.O. and Spellman, L.J. (1994), “Repudiation risk and restitution costs: toward
understanding premiums on insured deposits”, Journal of Money, Credit, and
Banking, Vol. 26, pp. 439-459.
ECB Occasional Paper Series No 208 / April 2018
52
Demirgüç-Kunt, A. and Detragiache, E. (2002), “Does deposit insurance increase
banking system stability? An empirical investigation”, Journal of Monetary
Economics, Vol. 49, No 7, pp. 1373-1406.
Demirgüç-Kunt, A. and Huizinga, H. (2004), “Market discipline and deposit
insurance”, Journal of Monetary Economics, Vol. 51, No 2, pp. 375-399.
Demirgüç-Kunt, A. Kane, E. and Laeven, L. (2008), Deposit Insurance Design and
Implementation: Policy Lessons from Research and Practice, in Demirgüç-Kunt, A.
Kane, E. and Laeven, L. (eds.), Deposit Insurance around the World - Issues of
Design and Implementation, The MIT Press, Cambridge, Massachussets, London,
England, pp. 3-26.
Demirgüç-Kunt, A., Kane, E. and Laeven, L. (2014), Deposit Insurance Database,
IMF Working Paper, WP/14/118, July.
Diamond, D. and Dybvig, H. (1983), “Bank runs, deposit insurance, and liquidity”,
The Journal of Political Economy, Vol. 91, No 3, pp. 401-419.
Ellis, D. (2013), “Deposit insurance funding: assuring confidence”, Federal Deposit
Insurance Corporation Staff Paper.
Ellis, D.M. and Flannery, M.J. (1992), “Does the debt market assess large banks’
risk?”, Journal of Monetary Economics, Vol. 30, No 3, pp. 481-501.
European Commission (2015), Communication from the Commission to the
European Parliament, the Council, the European Central Bank, the European
Economic and Social Committee and the Committee of the Regions "Towards the
completion of the Banking Union" , COM(2015) 587 final, Strasbourg, 23 November.
European Commission (2017), Communication to the European Parliament, the
Council, the European Central Bank, the European Economic and Social Committee
and the Committee of the Regions on completing the Banking Union, COM(2017)
592 final, Brussels, 11 October.
Financial Stability Board (2015), Historical losses and recapitalisation needs,
Findings Report.
Hardy, D. (2014), “Bank resolution costs, depositor preference and asset
encumbrance”, Journal of Financial Regulation and Compliance, Vol. 22, No 2, pp.
96-114.
Hovakimian, A., Kane, E. and Laeven, L. (2003), “How country and safety-net
characteristics affect bank risk-shifting”, Journal of Financial Services Research, Vol.
23, No 3, pp. 177-204.
Jordà, Ò., Richter, B., Schularick, M. and Taylor, A.M. (2017), Bank Capital Redux:
Solvency, Liquidity, and Crisis, National Bureau of Economic Research Working
Paper No 23287.
ECB Occasional Paper Series No 208 / April 2018
53
Laeven, L. (2002), “International evidence on the value of deposit insurance”, The
Quarterly Review of Economics and Finance, Vol. 42, No 4, pp. 721-732.
Lang, J., Sarlin, P. and Peltonen, T. (2018), A Framework for Early-Warning
Modelling with an Application to Banks, ECB Working Paper, forthcoming.
Merton, R.C. (1977), “An analytic derivation of the cost of deposit insurance and loan
guarantees - An application of modern option pricing theory”, Journal of Banking &
Finance, Vol. 1, No 1, pp. 3-11.
OECD and JRC (2008), Handbook on constructing composite indicators:
Methodology and user guide, OECD publishing, Paris.
ECB Occasional Paper Series No 208 / April 2018
54
Appendix
PD estimation and in-sample test
Starting from the specification A of the model implemented to estimate the default
probability of banks, an in-sample test has been performed to compare the predicted
results with the real, observed ones. According to the estimated regressors (see
Table 1), the default probability has been calculated for the entire sample of banks,
and for the period ranging from 2000 to 2013 (same period used to estimate the
model). Given that the outcome of the model is a probability (continuous variable,
between 0 and 1) while the observed defaults are classified as a dummy variable (0
in case of non-default, 1 in case of default), two different thresholds have been used
to transform the estimated PDs into dichotomic values:
(i) When the estimated PD is higher than the PD corresponding to the
97th percentile of the distribution, the bank is classified as in default.
This choice is consistent with the 3% riskiest banks failing scenario;
(ii) When the estimated PD is higher than the PD corresponding to the
90th percentile of the distribution, the bank is classified as in default.
This choice is consistent with the 10% riskiest banks failing scenario.
Table A1 summarises the performance of the model specification A under options (i)
and (ii).
Table A1
Performance of the model used to estimate default probabilities
Diagnostic
(i) (ii)
97th percentile as threshold 90th percentile as threshold
TP 99 155
TN 46 496 45 384
FP 767 1 879
FN 413 357
Accuracy 98% 95%
Sensitivity 19% 30%
Specificity 98% 96%
Source: ECB staff calculations based on Bankscope, ECB Statistical Data Warehouse and European Commission dataset on state aid
measures, 1999:Q1 – 2013:Q4.
Note: TP abbreviates True Positive (default events correctly estimated as defaults); TN abbreviates True Negative (non-default events
correctly estimated as non-defaults); FP abbreviates False Positive (or false alarms, non-default events estimated as defaults); FN
abbreviates False Negative (or II type errors, default events estimated as non-defaults). Accuracy is calculated as the ratio between
TP+TN and the overall population; Sensitivity is calculated as the ratio between TP and TP+FN; Specificity is calculated as the ratio
between FP and FP+TN.
On the one hand, the overall accuracy of the model is extremely high under both
options (i) and (ii); on the other hand, the discrimination between failing and non-
failing banks based on the 90th percentile (option (ii)) seems to substantially
increase the sensitivity of the model.
ECB Occasional Paper Series No 208 / April 2018
55
Abbreviations
ARS Aggregate Risk Score
ARW Aggregate Risk Weight
AUROC Area Under the Receiver Operating Characteristics Curve
BRRD Bank Recovery and Resolution Directive
COREP Common Reporting
DGS Deposit Guarantee Scheme
DGSD Deposit Guarantee Scheme Directive
DIF Deposit Insurance Fund
EBA European Banking Authority
EDIS European Deposit Insurance Scheme
FDIC Federal Deposit Insurance Corporation
FINREP Financial Reporting
G-SIB Global Systemically Important Bank
IRS Individual Risk Score
LGD Loss Given Default
MREL Minimum Requirement for own funds and Eligible Liabilities
NPL Non-Performing Loans
OECD Organisation for Economic Cooperation and Development
JRC Joint Research Centre
PD Probability of Default
ROE Return on Equity
RWA Risk Weighted Assets
SRF Single Resolution Fund
TLAC Total Loss-Absorbing Capacity
Acknowledgements
We would like to thank for useful comments Iñigo Arruga Oleaga, Barbara Attinger, Andreas Baumann, Thorsten Beck, Inês Cabral,
Simona Dodaro, Joachim Eule, Malte Jahning, Luc Laeven, Johannes Lindner, Luis Molestina Vivar, Sergio Nicoletti-Altimari, Fátima
Pires, Tanguy Poelman, Ad van Riet, Antonio Riso and Pär Torstensson.
Jacopo Carmassi
European Central Bank, Frankfurt am Main, Germany; email: [email protected]
Sonja Dobkowitz
Bonn Graduate School of Economics, University of Bonn, Bonn, Germany; email: s3sodobk@uni-bonn.de
Johanne Evrard
European Central Bank, Frankfurt am Main, Germany; email: [email protected]
Laura Parisi
European Central Bank, Frankfurt am Main, Germany; email: [email protected]
André Silva
Cass Business School, London, United Kingdom; email: andre.sil[email protected].ac.uk
Michael Wedow
European Central Bank, Frankfurt am Main, Germany; email: [email protected].eu
© European Central Bank, 2018
Postal address 60640 Frankfurt am Main, Germany
Telephone +49 69 1344 0
Website www.ecb.europa.eu
All rights reserved. Any reproduction, publication and reprint in the form of a different publication, whether printed or produced
electronically, in whole or in part, is permitted only with the explicit written authorisation of the ECB or the authors.
This paper can be downloaded without charge from www.ecb.europa.eu, from the Social Science Research Network electronic library or
from RePEc: Research Papers in Economics. Information on all of the papers published in the ECB Occasional Paper Series can be
found on the ECB’s website.
ISSN xxxx-xxxx (pdf) DOI xx.xxxx/xxxxxx (pdf)
ISBN 978-xx-xxx-xxxx-x (pdf) EU catalogue No QB-xx-xx-xxx-EN-N (pdf)
