Description
The purpose of this paper is to assist the numerous regulators around the globe who are
currently considering ways to impose domestic systemic importance-based capital requirements on
banks
Journal of Financial Economic Policy
Capital buffers based on banks’ domestic systemic importance: selected issues
Michal Skorepa J akub Seidler
Article information:
To cite this document:
Michal Skorepa J akub Seidler , (2015),"Capital buffers based on banks’ domestic systemic
importance: selected issues", J ournal of Financial Economic Policy, Vol. 7 Iss 3 pp. 207 - 220
Permanent link to this document:
http://dx.doi.org/10.1108/J FEP-07-2014-0040
Downloaded on: 24 January 2016, At: 21:52 (PT)
References: this document contains references to 27 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 50 times since 2015*
Users who downloaded this article also downloaded:
Kangbok Lee, Wenling Lu, (2015),"Do bank regulation and supervision matter?: International
evidence from the recent financial crisis", J ournal of Financial Economic Policy, Vol. 7 Iss 3 pp.
275-288 http://dx.doi.org/10.1108/J FEP-03-2015-0019
Gurbachan Singh, (2015),"Thinking afresh about central bank’s interest rate policy", J ournal of
Financial Economic Policy, Vol. 7 Iss 3 pp. 221-232 http://dx.doi.org/10.1108/J FEP-08-2014-0051
Abdullah Noman, Mohammad Nakibur Rahman, Atsuyuki Naka, (2015),"Portfolio investment outflow
and the complementary role of direct investment", J ournal of Financial Economic Policy, Vol. 7 Iss 3
pp. 190-206 http://dx.doi.org/10.1108/J FEP-03-2014-0024
Access to this document was granted through an Emerald subscription provided by emerald-
srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.
*Related content and download information correct at time of
download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Capital buffers based on banks’
domestic systemic importance:
selected issues
Michal Skorepa
Czech National Bank, Prague, Czech Republic, and
Jakub Seidler
Financial Stability Department, Czech National Bank, Prague,
Czech Republic
Abstract
Purpose – The purpose of this paper is to assist the numerous regulators around the globe who are
currently considering ways to impose domestic systemic importance-based capital requirements on
banks.
Design/methodology/approach – The article discusses in some detail a number of issues from the
viewpoint of regulatory practice, mentioning relevant literature where available. Comments partly
refect the experience that the Czech National Bank gathered over the past two years while preparing its
own regime of domestic systemic importance-based capital requirements on banks.
Findings – The authors stress, among other points, one weakness of the (otherwise well-designed)
method suggested by the Basel Committee for Banking Supervision (BCBS) for assessment of banks’
systemic importance: the method is “relative” in that it does not refect the absolute importance of the
banking sector for the economy. The paper also explains that in some cases, use of individual-level
rather than consolidated-level data may be preferable, in contrast to what the BCBS guidance suggests.
Further, implications of the buffers over a longer term are pointed out.
Originality/value – As far as the authors are aware, this article is the frst to comprehensively discuss
the main issues surrounding both key steps (systemic importance assessment and determination of
buffer level) in the process of introducing buffers based on domestic systemic importance. A number of
questions related to these two steps are raised which regulators may appreciate to be reminded of, even
if some of the questions are such that it is not possible to give a generally applicable answer to them.
Keywords Banks, Regulatory change, Capital, Financial risk and risk management
Paper type General review
1. Introduction
The fact that bank regulation may need to be sensitive to a bank’s systemic importance
has been acknowledged for a number of years (BCBS, 2002). The debate on appropriate
ways to measure systemic importance and to adapt regulation to it intensifed in light of
JEL classifcation – G21, G28
The authors would like to thank Christian Castro, Jan Frait, Václav Hausenblas, Jan Kubíc?ek,
Sergio Masciantonio, Gor Sahakian, Jan Sobotka, Boøek Vašíc?ek and an anonymous referee for
helpful comments and useful recommendations. However, all errors and omissions are those of the
authors. The views expressedinthis paper are those of the authors andnot necessarilythose of the
Czech National Bank.
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1757-6385.htm
Banks’
domestic
systemic
importance
207
Received1 July2014
Revised5 November 2014
6 January2015
Accepted6 January2015
Journal of Financial Economic
Policy
Vol. 7 No. 3, 2015
pp. 207-220
©Emerald Group Publishing Limited
1757-6385
DOI 10.1108/JFEP-07-2014-0040
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
the dramatic repercussions caused by the failure of Lehman Brothers in September of
2008. The result of this debate is the Basel Committee for Banking Supervision’s
(BCBS’s) guidance on the setting of buffers for globally systemically important banks
(G-SIBs) and domestically systemically important banks (D-SIBs) – see BCBS (2013,
2012), respectively. As a result, implementation of systemic importance-based buffers is
currently a highly topical regulatory issue across the globe.
The BCBS guidance suggests measuring systemic importance based on a set of
indicators which are likely to correlate with the bank’s systemic importance; none of
the model-based approaches suggested in the academic literature (surveyed by
Bisias et al., 2012; de Bandt et al., 2013) is viewed as robust enough (yet) to be used
as a basis for actual regulatory measures (Berg, 2011; Löffer and Raupach, 2013).
Once individual banks’ systemic importance is assessed – expressed quantitatively as
their “SIB scores” – the guidance relies primarily on the principle of “equal expected
impact” (see BCBS, 2013, Appendix II) to determine the appropriate level of the buffer.
In this article, we discuss some of the issues that arise when a regulator intends to
introduce, more or less along the lines of the above BCBS guidance, capital buffers based
on banks’ domestic systemic importance. Appropriate measurement of the degree of a
fnancial institution’s systemic importance and the question of whether and how to
refect this degree in regulation and supervision are issues whose study faces a number
of conceptual as well as empirical challenges. For this reason, our comments, while
partly refecting the hands-on experience that the Czech National Bank gathered over
the past two years during the process of preparing its own regime of domestic systemic
importance-based capital requirements on banks, will at multiple points have to refer to
common sense or intuition, rather than robust conclusions of theoretical or empirical
literature. Indeed, literature focusing on this topic is still relatively scarce as the D-SIB
regulation is a rather recent addition to the regulatory framework. Thus, our
step-by-step discussion of establishing D-SIB regulation based on the experience of a
particular regulatory authority is the main value added of this article and contribution to
the on-going policy debate.
The article is structured as follows. Section 2 discusses some issues concerning the
measurement of systemic importance. Section 3 then focuses on estimation of
appropriate capital buffers for individual banks based on their systemic importance.
Section 4 summarises the main conclusions.
2. Calculation of systemic importance
2.1 Some possible indicators beyond the BCBS guidance
In contrast to the G-SIB methodology, the BCBS does not go into detail on the indicators
that belong to each category or on the method for calculating the D-SIB score itself (for
more on the G-SIB guidance and the D-SIB guidance, see BCBS, 2013; 2012). Given the
general tone that the BCBS guidance on D-SIBindicators has, one possibility is to use the
indicator list applied within the G-SIB methodology, summarised in Table I, as a basis
and modify it by dropping or, conversely, taking on board various indicators and
adjusting the weights of the indicators and/or indicator categories (such as by moving
cross-jurisdictional claims and liabilities under complexity) as deemed appropriate from
the perspective of the domestic fnancial system or for other relevant reasons[1].
For example, the list of indicators within the interconnectedness category might be
augmented with a measure of concentration of interbank connections (measured, for
JFEP
7,3
208
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
example, as the ratio of the bank’s top three interbank claims to its total interbank
claims, and similarly for liabilities). The idea – which seems relevant even at the G-SIB
level – is that a higher concentration of assets or liabilities implies stronger contagion:
the higher is the proportion of Bank A’s assets that have been lent to Bank B, the lower
is the probability that Bank A will survive Bank B’s failure to repay the loan; an
analogous proportionality holds for concentration of liabilities (for some empirical
evidence pointing in this direction, see Craig et al., 2013).
In countries where the banking sector is the dominant source of funding for the
corporate sector, the substitutability category may, in turn, be augmented with the share
of banks’ total loans provided to non-fnancial corporations. This indicator refects the
importance of the particular bank from the perspective of providing funds to the real
sector. The higher share the bank has, the more signifcant would be the problemcaused
to the real economy in the event of distress of the bank. Conversely, in some countries,
the banking sector might be exposed to some sectors to such an extent that any bank
distress and resulting fre sale of exposures might lead to a signifcant drop in prices of
the instruments concerned, adversely affecting the whole sector (Shleifer and Vishny,
2011). If this is the case, the list of indicators might be augmented to refect this
specifcity of the domestic banking sector as well, e.g. a high concentration of the
domestic banking sector towards sovereign bonds of domestic government.
In highly concentrated banking sectors, another possible substitutability indicator is
the share in total insured deposits (or a proxy for this share where the data are not
available) because bank failure and reimbursement of insured depositors may entail
high costs for public fnance. This risk is high in a number of situations: when there is no
deposit guarantee fund, when it is small relative to the deposits paid back, or when it is
of the ex post type and there are doubts about the possibility to collect enough
contributions from banks suffciently quickly ex post. The share-of-insured-deposits
Table I.
List of indicators
applied within the
G-SIB methodology
Category (and weighting) Individual indicator Indicator weighting (%)
Cross-jurisdictional activity (20%) Cross-jurisdictional claims 10
Cross-jurisdictional liabilities 10
Size (20%) Total exposures as defned for use
in the Basel III leverage ratio
20
Interconnectedness (20%) Intra-fnancial system assets 6.67
Intra-fnancial system liabilities 6.67
Securities outstanding 6.67
Substitutability/fnancial
institution infrastructure (20%)
Assets under custody 6.67
Payment activity 6.67
Underwritten transactions in debt
and equity markets
6.67
Complexity (20%) Notional amount of over-the-
counter derivatives
6.67
Level 3 assets 6.67
Trading and available-for-sale
securities
6.67
Source: BCBS
209
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
indicator might therefore refect the possible costs incurred by taxpayers if banks
experience distress.
If it is decided that systemic importance will be measured on an individual basis (an
issue to which we return below), then information on the relative size of the bank’s
sub-consolidated group can still be included via a new complexity indicator which will
express the sub-consolidated group’s assets as a multiple of the bank’s solo assets. Our
approximation of the level of complexity of a given bank’s liquidation may be improved
by means of data on its number of branches and number of employees, as this indirect
information refects the fragmentation of the bank’s business and the increased
diffculty of resolving the bank in the event of the bank’s liquidation (Parker, 2011).
Furthermore, data on the share of non-performing loans might also be a useful indicator
from the perspective of the complexity of a bank’s liquidation, as non-performing loans
(NPLs) are typically harder to evaluate and sell than standard loans (for empirical
support, see Klingebiel, 2000).
Taking the perspective of individual banks, investors or academia, that is, those
observers who do not have access to confdential regulatory data on all the banks,
Brämer and Gischer (2011) and Masciantonio (2013) suggest modifcations to the list of
indicators used such that the D-SIBscores are based purely on those data that, while still
refecting a bank’s systemic importance, are publicly available.
Another issue as regards the selection of indicators is the problem of high
correlations: should we rely on empirics, that is, discard all indicators that are highly
correlated with others, or should we rely on theory/intuition, that is, include all
indicators that are available and seem relevant? An argument in favour of the latter
option is that the high correlation based on historical observations might decrease in the
future. An argument in favour of the former option is that if two indicators are highly
correlated because they convey the same aspect of systemic importance, including both
such indicators in the calculation of the SIBscore will increase the overall weight of that
particular aspect, possibly above the level that would be appropriate. At this early stage
of the SIB score debate, however, little is known about the appropriate weights of the
various aspects of systemic importance, and so it seems prudent to include all relevant
indicators, even if some of the correlations are very high. Moreover, the degree of
correlation may happen to be high over the recent past but may fall in the future, so that
a choice of indicators on the basis of their cross-correlations may imply undesirable
instability in the list of indicators.
2.2 More fundamental departures from the BCBS guidance
Besides the new indicators mentioned above, there are also several other dimensions in
which the D-SIB framework could (and in some cases should) depart from the G-SIB
methodology more fundamentally.
First, as already hinted above, in the G-SIB framework, systemic importance is
measured as purely relative. This means that G-SIB scores express importance relative
to the rest of the banking sector alone, rather than importance as such, that is, the
absoluteimportance to the economy as a whole. In the G-SIB framework, the G-SIB
scores are calculated for the single purpose of being used as a basis for determining the
appropriate G-SIB buffer rate. In that case, their relative nature is not a problem as long
as the subsequent calculation of the buffer rate somehow also manages to incorporate
the importance of the banking sector to the economy, so that even if G-SIBscores as such
JFEP
7,3
210
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
are relative, the ultimate buffer rates based on them are absolute. The issue of whether
the D-SIB procedure for setting the buffer rates does manage to incorporate the
importance of the banking sector to the economy will be tackled in the next section.
If a bank’s D-SIB score itself were to be used as a proxy for actual systemic
importance (to determine, for instance, the set of banks which should develop their own
resolution plans), then the calculation of the score should obviously be somehow
extended, compared to the G-SIBframework, such that it becomes absolute in the above
sense. The ambition to measure actual systemic importance in this absolute sense, i.e. as
the absolute volume of costs of a given bank’s failure to the economy, raises a number of
issues though. For example, we would need to specify exactly what state of affairs we
mean by “failure of the bank”, i.e. at what stage of a bank’s process of failing (before
insolvency or after it) do we measure the costs to the economy. The relative approach
simply assumes that the stage is the same for all banks; it need not be specifed
explicitly.
The relative nature of SIB scores (whether in the G-SIB context or the D-SIB context)
may seemto be a problemnot only because the scores fail to showthe absolute systemic
importance, but also for another reason (raised, for instance, in IIF, 2011): the
relativeness implies that if Bank A changes its business model to reduce its SIB score
(motivation for such changes is one of the intended consequences of SIB frameworks),
the ultimate effect on its SIB score will be reduced (or neutral or even opposite) if the
remaining banks take similar steps to a smaller (or comparable or bigger) extent. The
reward, in the form of a fall in the D-SIB buffer rate, that Bank A will get for its effort to
become less systemic may then end up being lower than proportional (or the rate may
stay the same or it may rise). But here again, all that is needed is the incorporation of
absolute importance via the buffer rates. If absolute importance does get incorporated,
then Bank A’s effort to reduce its SIBscore will be rewarded: even if, for example, its SIB
score as such does not change at all because the other banks take the same steps towards
reducing their SIB scores, the systemic importance of the banking sector as a whole will
fall and the SIB buffer rates of all the banks, including Bank A, will fall. The sensitivity
of a bank’s SIB score to what other banks do may also be removed, for instance, if we
calculate the SIB score from “standardised” values of the indicators. By “standardised”
we mean that the value of a bank on a given individual indicator is measured as the
bank’s percentile position within a range whose lower and upper bounds are fxed and
the same for all banks. Under this approach, however, our choice of the bounds would
have a clear impact on the effective weights of the individual indicators in the
determination of the overall SIB score.
A second direction in which we may want the D-SIB framework to depart from the
G-SIB one is to give up the cardinal approach of calculating a bank’s D-SIB score as a
weighted average of the relevant indicators and to content ourselves with the ordinal
approach of concluding that a bank is highly systemically important if it exceeds a
certain threshold for at least one of the indicators. If we opt for this departure, however,
either we end up imposing the same buffer rate on all those banks which we have found
to be highly systemically important or, if we want the buffer to be commensurate with
the systemic importance of banks in that “highly systemically important” group, we still
need something akin to D-SIB scores and, thus, may be forced to return to a cardinal
approach.
211
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
A third direction in which departure from the G-SIB framework may be considered
concerns the question of whether systemic importance is judged at the level of individual
banks or their groups consolidated for regulatory (rather than accounting) purposes.
The BCBS suggests the consolidated approach (for both the G-SIB and D-SIB context).
In principle, this suggestion is reasonable: if we use solo data for setting the D-SIB
buffer, the bank may successfully reduce its D-SIB score artifcially as regards, for
instance, interconnectedness, by lending money to another bank via a non-bank special
purpose vehicle (SPV) established within the lending bank’s group as a subsidiary or an
affliate. In this way, the lending bank’s resulting solo claim on the SPV does not get
included in the lending bank’s value score for interconnectedness (as long as it is
bank-to-bank interconnectedness). Using the consolidated approach, and assuming that
it also covers the SPV, the claim of the group on the borrowing bank does get recorded
as a claim of the group on a bank and, thus, does increase the lending bank (group)’s
interconnectedness score.
In reality, however, consolidated data (in the regulatory sense) need not – and often
will not – be a good basis for assessing the impact of the group’s distress on the domestic
economy, that is, of the group’s domestic systemic importance. First, consolidated data
usually exclude certain domestically operating entities (especially non-fnancial entities)
whose distress may have sizeable impact on the domestic economy; in this regard, the
use of consolidated data has no value added compared to solo data. Consolidated data
will typically also exclude insurers. This exclusion will cease to be a major problemonly
after the framework for regulation of domestic systemically important insurers is
activated in a given jurisdiction. Second, consolidated data often include entities
operating abroad, whose distress may have a negligible effective impact on the domestic
economy, while foreign branches – which may often make the bank’s resolution
perceptibly more diffcult – are already included in unconsolidated data.
Therefore, consolidated data, just like solo data, should be taken as a kind of proxy
for the actual type of data needed to assess domestic systemic importance correctly. The
choice between using solo or consolidated data should be made after considering
relevant particular features of the relevant banking and fnancial system and the
resulting risk of psychological or other contagion among members of the respective
groups. Data availability and quality (especially as regards more distant history) may
play a role, too. Indeed, the European Union’s Capital Requirements Directive allows the
systemic importance-based buffers to be set on an individual as well as
(sub-)consolidated level. For a treatment of the related but separate issue of the
interaction of a G-SIB buffer on the group and a D-SIB buffer on a subsidiary, see
Skorepa and Seidler (2014).
One issue where the BCBS guidance leaves the decision fully to national authorities
is the extent to which the national D-SIB regime covers not only banks and subsidiaries
of foreign banks, but also foreign banks’ branches. If the D-SIB scores are calculated
solely for the purpose of determining the appropriate rate of the D-SIB buffers for
individual banks, then inclusion of foreign banks’ branches is not needed because the
host regulator is not able to impose a capital buffer on them anyway. On the other hand,
as long as the branches are not highly systemically important, their inclusion is not a
problem, as it will not change the banks’ D-SIBscores much. Conversely, if at least some
of the foreign banks’ branches are likely to be highly systemically important for the
domestic economy, then the calculation of D-SIB scores for the purpose of setting the
JFEP
7,3
212
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
D-SIBbuffer rates should not include them; in that case, however, a separate calculation
with the branches included may make sense to fnd out their systemic importance and to
determine which of them – given that the host regulator does not regulate them – may
require at least more intensive host supervision (see, e.g. MAS, 2014). When
communicating this type of calculation, the host authorities must, of course, avoid
giving the impression to markets and the public that they assume any responsibility for
the stability of these branches.
3. Calculation of the D-SIB buffer rate
In this section, we describe in detail the central procedure suggested by the BCBS for
calculating the SIB buffer rate based on the principle of “equal expected impact” and we
comment on selected aspects of this procedure.
3.1 Assumptions
In line with the BCBS publications, we start with the following assumptions set out in
Basel III when determining the capital buffer based on the SIB score:
• Each bank must meet a minimum capital requirement for Common Equity Tier 1
capital (CET1) of k
min
?4.5 per cent of risk-weighted assets. CET1 is composed of
common stock and retained earnings, i.e. capital that can be used immediately and
unconditionally to cover any losses of the bank.
• In normal circumstances, each bank additionally holds the full basic component of
the CET1 capital conservation buffer[2] of k
basic
? 2.5 per cent of risk-weighted
assets.
• In normal circumstances, a bank with an SIB score denoted by sib should comply
not only with k
min
and k
basic
, but also with the SIB buffer, i.e. the full SIB
component, k(sib), of the CET1 conservation buffer.
Consequently, of the three components of the total CET1 capital requirements listed
above, only k(sib) is sensitive to the bank’s SIB score.
If the capital of the bankfalls belowk
min
?k
basic
?k(sib), the bankmust take remedial
action whose intensity (and thus also the costs to the economy arising fromthe situation)
is proportional to the decline in capital. In what follows, the situation where, as a result
of a large negative proft in the quarter, the bank’s CET1 falls below the regulatory
minimumk
min
, i.e. it records a negative quarterly proft of ?[k
basic
?k(sib)] or lower, will
be referred to as distress (this need not mean a straightforward fall in the sense of a loss
of licence)[3]. The probability P(sib) of distress for a bank with an SIB score of sib is
obviously lower for a higher SIB capital buffer k(sib), i.e. for a higher level of sib. The
costs to the economy arising from the distress of a bank with an SIB score equal to sib
will be denoted C(sib).
The capital buffer is then determined on the basis of the “equal expected impact”
principle (for an early suggestion of this principle, see Squam Lake Working Group,
2009). This principle can be generally expressed as follows: the expected costs to the
economy resulting fromdistress of any bank that is systemically more important than a
certain reference bank chosen by the regulator should be the same as the expected costs
to the economy, resulting fromdistress of the reference bank. Of course, one loose end of
this principle is how the regulator selects the reference bank; we return to this question
shortly.
213
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Besides the expected impact approach, BCBS (2013) uses the results of other
approaches (models created by the Long-term Economic Impact Group and a method
based on the implicit subsidies that some highly systemically important banks receive
because the market considers them to be too big to fail (TBTF), i.e. it expects public
money to be spent on bailing them out if they get into diffculty). In many jurisdictions,
however, these approaches will be diffcult to apply due to a lack of relevant data and/or
models.
According to the expected impact principle, the point of the SIBbuffer is to reduce the
probability P(sib) of distress of the bank such that the expected costs of this situation, i.e.
C(sib)*P(sib), are equal to the expected costs of distress of the reference bank, i.e.
C(sib
R
)*P(sib
R
). It is obvious that the SIBbuffer will be zero for the reference bankandfor
every systemically less important bank.
3.2 Calculation
BCBS (2013) uses two methods to determine the SIB buffer according to the expected
impact principle. The frst method uses a Merton model to estimate a bank’s
market-perceived probability of failure from the market prices of its equity. The second
method is based on the historical frequency distribution of the return on risk-weighted
assets (RORWA – see Kuritzkes and Schuermann, 2010). Nevertheless, in many
countries, many of their domestic banks’ shares are not traded on public markets (or not
traded intensively enough to make the data suffciently reliable), and so the Merton
model has limited application. The same problem arises if we want to use for this
purpose market prices or yield spreads of banks’ bonds or other liabilities (such as
uninsured deposits). For this reason, we focus in this article on the RORWAmethod. For
an example of actual use of the method in regulatory practice, see Skorepa and Seidler
(2013) who describe the analytical basis of the approach selected by the Czech National
Bank.
The expected impact principle can be expressed formally as follows: for all sib ?sib
R
,
P(sib) should satisfy:
P
(
sib
)
* C
(
sib
)
? P
(
sib
R
)
* C
(
sib
R
)
, i.e.
P
(
sib
)
? P
(
sib
R
)
/
?
C
(
sib
)
/C
(
sib
R
)?
. (1)
To derive the values of P(sib) and subsequently also the capital buffer k(sib) based on the
bank’s sib, we frst need to determine the value of P(sib
R
). The frst step is to choose the
level of sib
R
itself. While the sib for each bank is given by the empirically observed levels
of the various indicators for that bank, there is no natural, obvious, objective way to set
sib
R
; it thus has to be determined on the basis of regulatory considerations.
One method which seems fairly acceptable and transparent is to set sib
R
equal to q
times sib
aver
, where the latter is the average sib for the entire domestic banking sector
and values of q ? 1 are considered. The value of q is chosen at the discretion of the
regulator, depending on how weak and narrow it wants the effect of the SIB buffer
regime to be: the higher q is, the lower the buffers will be; moreover, increasing q may
reduce the set of banks to which the buffers will apply. For example, the G-SIB regime
appears to work with q ? 1; the Czech National Bank opted for q ? 2 (Skorepa and
Seidler, 2013). Alternatively, sib
R
can be set equal to the q-th percentile of the distribution
JFEP
7,3
214
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
of all SIB scores observed in the banking sector. Here, again, q measures how weak and
narrow the effect of the SIB regime is to be.
Not surprisingly, the different methods for setting sib
R
have somewhat different
implications. For example, if the dispersion of SIB scores across the relevant
banking sector falls below a certain level, then the q-times-the-average method will
imply no SIB buffers, while the q-th percentile-based method will, by construction,
always imply SIB buffers for the 1-q per cent of banks with the highest SIB scores (in
regulatory practice, some or all of these buffers may be so small as to be rounded to
zero). The intuitive appeal of the two methods may thus depend on the particular
circumstances.
Assumptions (i)-(iii) listed above and the assumption k(sib
R
) ? 0 imply that P(sib
R
)
corresponds to the probability that the bank will make a negative proft of:
??k
basic
? k
(
sib
)
? ? ?
(
2.5 ? 0
)
? ?2.5% of RWA
or lower. With the historical RORWA distribution, this is therefore the relative
frequency of cases where RORWA ??2.5 per cent. If we simultaneously interpret the
historical RORWA distribution as being the RORWA probability distribution in the
future, then:
P
(
sib
R
)
? p(RORWA ? ?2.5%).
To calculate P(sib) from equation (1), we now need to determine the value of C(sib)/
C(sib
R
). In accordance with intuition and with a proposal contained in BCBS (2013), we
can assume for simplicity that this ratio can be approximated as sib/sib
R
. Using the
historical RORWA distribution, we can then derive the minimum capital loss for each
level of P(sib).
The capital requirement k
basic
?k(sib) ?2.5 ?k(sib) should be of an amount covering
this loss. This gives us the SIB capital buffer k(sib) based on the degree of systemic
importance of the bank.
As a practical matter, the level of the buffer is likely to be rounded by the
regulator before it is imposed. This rounding may bring with it some welcome
stabilisation in the sense that it reduces the frequency of change in the D-SIB buffer
level. Naturally, the resilience of buffers to excessive volatility can be further
enhanced by calculating the D-SIB scores not from the values of source indicators as of
a single date, but from longer-term averages. On the other hand, the stability of D-SIB
buffers must not be so strong as to limit their “motivational” effect: it must not lead to a
situation where a bank’s efforts to reduce its D-SIB buffer by reducing its systemic
importance take too long to bear fruit.
To close this section, a short note on determining the overall capital requirement for
a bank may be in order. The SIBbuffer is motivated by the potential losses that the bank
may impose on the rest of the economy. In contrast, the existing capital requirements on
banks, whether within Pillar 1 or Pillar 2, have all been motivated by the potential losses
that the rest of the economy may impose on the bank. Given this fundamental difference
in the motivation of the existing capital requirements and the SIB buffer, it seems
natural to add the latter buffer on top of any existing Pillar 1 or Pillar 2 capital
requirements.
215
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
3.3 Does the banking sector’s importance to the economy get refected in the buffer
rate?
Now, we return to the issue raised earlier of whether the above RORWA-based
procedure for setting the D-SIB buffer rates ensures that while based on relative D-SIB
scores, the buffer rates also encompass the importance of the banking sector to the
economy. Alas, the procedure does not meet this criterion: the relative frequency of
various levels of RORWAsays nothing about the implications that these levels have for
the economy.
As regards specifcally the G-SIBbuffers proposed in BCBS (2013), a cross-check that
the buffer rates suggested by the RORWA method are also absolutely adequate is
provided by the confrmation (conducted in Appendix II of BCBS, 2013) that capital
buffers of similar rates seem roughly optimal according to the macroeconomic models
used by the Long-term Macroeconomic Impact Working Group.
In principle, this “absoluteness cross-check” is possible also at the D-SIB level: after
the national regulator has calculated the D-SIB buffers using the relative approach, she
may verify that similar buffer rates are roughly optimal according to an appropriate
macroeconomic model. The problem, especially for smaller or less advanced economies,
is that signifcant parts of the data set required to estimate or calibrate such a model
reliably may not be available. At the same time, direct application of the models used in
BCBS (2013) may not give a reliable answer either, as they have been developed to ft a
handful of specifc economies, most of them large and/or advanced (see BCBS, 2010a).
3.4 Dynamics of the SIB regime
Aspecial class of issues arises once we take a longer-termperspective and consider how
the SIB scores and buffers may evolve over time. First, the shorter is the RORWA
history on which the buffers are based, the more procyclical the buffers are bound to be:
in boomtimes, the level of proft occurring with any given probability will rise and, thus,
the implied buffer rate will fall, and vice versa. The obvious way to avoid this
procyclicality is to work with an RORWAtime series long enough to contain at least one
full fnancial cycle – if such a long time series is available.
Second, assume that the introduction of the SIB buffers and other regulatory
measures aimed at reducing the probability of distress of individual banks does achieve
this aim. Then, the histogram of the new RORWA data (recorded after the buffers were
imposed) can be expected to lie on average to the right of the old data histogram, that is,
the frequency of losses of a given extent should be lower than it used to be before the
buffers were introduced. The reason for this effect is that higher capital buffers held by
the bank’s counterparties in the interbank market imply lower incidence of distress
among those counterparties and, thus, fewer credit or similar losses for the bank in
question. This consideration leads to the expectation that at the frst point of SIB buffer
rate re-assessment, the RORWA data are more likely to suggest that the buffers should
be lowered by some amount. When re-assessing the buffer rates for the second time, the
RORWA data are more likely to suggest that the buffers should be increased but by a
smaller amount, and so on.
3.5 Data issues
On a practical level, in this second stage of the D-SIB procedure, in which the D-SIB
buffer rates are determined, the issue of including or excluding branches resurfaces. The
JFEP
7,3
216
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
host regulator cannot impose this buffer on a branch of a foreign bank and so it makes
no sense to bother calculating the buffer for such branches. Nevertheless, there is still the
question of whether we want to include branches in the pool of historical RORWA data
used to determine the D-SIB buffer rates for domestic banks which are not branches. At
frst sight, it seems preferable not to include branches on the grounds that their fnancial
results may be tilted by the fact that they are not “complete banks” as they share or
outsource some of their operations within the bank they are part of. This train of
thought, however, loses steamas soon as we realise that even subsidiaries often share or
outsource some of their operations within their groups such that even many subsidiaries
are not “complete banks” (Fiechter et al., 2011). More pragmatically, it may be necessary
to collect RORWA data also from branches for the simple reason that domestic banks
that are not branches are too few or exist for too short a time.
A similar issue is whether to include in the pool of RORWA data also the starting
periods of newentrants in the domestic banking sector. Indeed, the frst several quarters
or even years of their existence may be viewed as a special “start-up” regime with
higher-than-usual investment expenditures on brand recognition and fght for a
satisfactory market share (DeYoung, 1999). But again, this fne consideration may have
to be disregarded and most of the newentrant data (except, perhaps, for a fewperiods at
the very start) may have to be used to obtain a suffciently rich historical RORWA data
set.
Nevertheless, the pool of historical RORWA data provided by domestic banks may
be too small even after including acceptable data for branches and new entrants. Or we
may fear that a major part of the past time period fromwhich the pool originates was too
non-representative of the way the domestic banking sector is expected to work in the
future (for which we want the D-SIB buffers to be adequate) – perhaps the sector or its
economic and/or legal environment underwent a transition period or a substantial
structural break. In these cases, data pertaining to the domestic banking history may
actually be a poorer guide for assessing the future than data pertaining to foreign
banking history. Under such circumstances, it may be preferable to determine the
overall level of D-SIB buffers based on foreign data. Of course, various adjustments are
possible to refect, for example, the difference in business models (and thus in riskiness)
of domestic banks relative to foreign ones (determinants of banks’ riskiness are studied,
for example, by Altunbas et al., 2011) or the difference in the amplitude of fnancial
cycles in emerging markets relative to developed economies (as documented by, for
example, Terrones et al., 2011).
4. Summary and conclusions
This article discusses some of the issues that arise when a regulatory authority intends
to impose capital buffers on domestic banks based on their domestic systemic
importance. In the current post-crisis environment, implementation of this type of
buffers is a highly topical issue around the globe.
We consider several specifc indicators of systemic importance that, while not used in
the BCBS guidance for capital buffers based on banks’ global systemic importance,
might be relevant on the domestic level. Furthermore, the usual assumption that
systemic importance should be calculated at the sub-consolidated (rather than solo)
basis is pointed out as being debatable whenever domestic banks control large
non-fnancial undertakings or their foreign subsidiaries operate mostly in very distant
217
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
economies. We then explain the advantages of calculating an overall SIB score (rather
than judging whether a bank exceeds thresholds for several indicators individually). We
also document the fact that the BCBS guidance results in systemic importance scores
which, in themselves, are only relative, in the sense that they capture the internal
structure of the banking sector but do not refect the importance of the sector to the
whole economy.
As regards the calculation of the buffer rate, we focus on the equal expected impact
approach and discuss the pros and cons of several distinct methods of arriving at the
cut-off point above which the buffer should be imposed on a bank – methods such as
clustering, a fxed multiple of the average systemic importance score and a fxed
percentile of the systemic score distribution. We also point out certain longer-term
implications of the SIB buffer regime. For example, the introduction of buffers may
gradually shift the observed frequency distribution of RORWA fgures and, thus, have
an impact on the future reassessment of the appropriate buffer rates.
Concerning public communication of the buffers, we would like to emphasise the
crucial signalling effects of public communication of the SIB regulation: the
macroprudential authority should make every effort to explain that the imposition of an
SIB capital buffer on a bank does not signal a guarantee of government bail-out. If this
communication fails (or if it actually equates TBTF banks with banks with an SIB
buffer, as is often the case in the media and some fnancial publications), then the
introduction of the SIB regime might make the TBTF problem even more pressing (for
a more detailed exposition of this problem, see Skorepa and Seidler, 2014).
Notes
1. The procedure for G-SIB scores is that for each indicator, the value for a given bank is
normalised by the value for the whole sample of banks under consideration and the G-SIB
score of each bank is obtained as a weighted average of the bank’s normalised values for all
indicators. By implication, the G-SIB score of each bank is a score relative to the rest of the
sample, and the sum of the G-SIB scores across all banks in the sample is 1.
2. We use the descriptor “basic” here because we take the “total” conservation buffer to include an
SIB buffer and a countercyclical buffer (where introduced): as regards the capital-conserving
implications for the bank of not meeting the buffers, both the countercyclical buffer and the
bank’s systemic importance-based buffer that Basel III sets the stage for should work the
same way as the capital conservation buffer.
3. In determining whether the return (income) is positive or negative, we take zero proft as the
benchmark, which seems in line with the approach used in BCBS (2010b, 2013). In contrast,
Kuritzkes and Schuermann (2010) argue that capital is to cover unexpected downside risk,
that is, any negative deviation from expected proft, and so they take as the benchmark the
bank’s expected proft (proxied by its average historical proft). This alternative might be
justifed from a long-term point of view, where avoiding negative values of net income is not
enough for the bank to survive, as the bank must actually meet its shareholders’ expectations
regarding dividends. These expected dividends can be viewed as an additional, special type
of expected expense. This train of thought would imply that, for the purposes of the SIBbuffer
size assessment, net income should be calculated net of this special type of expected expense,
that is, net of expected income. Here, however, we are concerned with the standard one-year
horizon for the assessment of capital adequacy.
JFEP
7,3
218
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
References
Altunbas, Y., Manganelli, S. and Marques-Ibanes, D. (2011), “Bank risk during the fnancial crisis:
do business models matter?”, ECB Working Paper No. 1394, European Central Bank,
Frankfurt am Main.
BCBS (2002), “Supervisory guidance on dealing with weak banks”, Report of the Task Force on
Dealing with Weak Banks, Bank for International Settlements, Basel, March.
BCBS (2010a), “An assessment of the long-term economic impact of stronger capital and liquidity
requirements”, Basel Committee on Banking Supervision, Bank for International
Settlements, Basel, August 2010.
BCBS (2010b), “Calibrating regulatory minimum capital requirements and capital buffers: a
top-down approach”, Basel Committee on Banking Supervision, Bank for International
Settlements, Basel, October 2010.
BCBS (2012), “A framework for dealing with domestic systemically important banks”, Basel
Committee on Banking Supervision, Bank for International Settlements, Basel, October
2012.
BCBS (2013), “Global systemically important banks: updated assessment methodology and the
additional loss absorbency requirement”, Basel Committee on Banking Supervision, Bank
for International Settlements, Basel, July 2013.
Berg, S.A. (2011), “Systemic surcharges and measures of systemic importance”, Journal of
Financial Regulation and Compliance, Vol. 19 No. 4, pp. 383-395.
Bisias, D., Flood, M., Lo, A.W. and Valavanis, S. (2012), “A survey of systemic risk analytics”,
Annual Review of Financial Economics, Vol. 4 No. 1, pp. 255-296.
Brämer, P. and Gischer, H. (2011), “Domestic systemically important banks: an indicator-based
measurement approach for the Australian banking system”, FEMM Working Paper No.
3/2012, Otto-von-Guericke University of Magdeburg.
Craig, B.R., Fecht, F. and Tümer-Alkan, G. (2013), “The role of interbank relationships and
liquidity needs”, Bundesbank Discussion Paper No. 54/2013, Deutsche Bundesbank,
Frankfurt am Main.
De Bandt, O., Héam, J.-C., Labonne, C. and Tavolaro, S. (2013), “Measuring systemic risk in a
post-crisis world”, Débats économiques et fnanciers No. 6, Autorité de Contrôle Prudentiel,
Banque de France.
DeYoung, R. (1999), “Birth, growth, and life or death of newly chartered banks”, Federal Reserve
Bank of Chicago, Economic Perspectives, Vol. 23 No. 3, 18-34.
Fiechter, J., Ötker-Robe, I., Ilyina, A., Hsu, M., Santos, A. and Surti, J. (2011), “Subsidiaries or
branches: does one size ft all?”, International Monetary Fund Staff Discussion Notes No.
11/04, International Monetary Fund, Washington, DC.
IIF (2011), “IIF views on the Basel committee’s consultative document on global systemically
important banks: assessment methodology and the additional loss absorbency
requirement”, Institute of International Finance, Washington, DC.
Klingebiel, D. (2000), “The use of asset management companies in the resolution of banking crises:
cross-country experience”, World Bank Policy Research Paper No. 2284, World Bank,
Washington, DC.
Kuritzkes, A. and Schuermann, T. (2010), “What we know, don’t knowand can’t knowabout bank
risk: a view from the trenches”, in Diebold, F.X., Doherty, N.A., Herring, R.J. (Eds), The
Known, the Unknown, and the Unknowable in Financial Risk Management: Measurement
and Theory Advancing Practice. Chapter 6, Princeton University Press, Princeton, NJ.
219
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Löffer, G. and Raupach, P. (2013), “Robustness and informativeness of systemic risk measures”,
Deutsche Bundesbank Discussion Paper No. 04/2013, Deutsche Bundesbank, Frankfurt am
Main.
MAS (2014), “Proposed framework for systemically important banks in Singapore”, Consultation
Paper, P008-2014, Monetary Authority of Singapore, Singapore.
Masciantonio, S. (2013), “Identifying and tracking systemically important fnancial institutions
(SIFIs) with public data”, MPRA Paper No. 46867, University Library of Munich, Munich.
Parker, D.C. (2011), Closing a Failed Bank: Resolution Practices and Procedures, International
Monetary Fund, Washington, DC.
Shleifer, A. and Vishny, R. (2011), “Fire sales in fnance and macroeconomics”, Journal of
Economic Perspectives, Vol. 25 No. 1, pp. 29-48.
Skorepa, M. and Seidler, J. (2013), “An additional capital requirement based on the domestic
systemic importance of a bank”, Financial Stability Report 2012/2013, Czech National
Bank, pp. 96-102.
Skorepa, M. and Seidler, J. (2014), “Capital buffers based on banks’ domestic systemic importance:
selected issues”, Czech National Bank Research and Policy Notes No. 1/2014, Czech
National Bank, Prague.
Squam Lake Working Group (2009), Reforming Capital Requirements for Financial Institutions,
Council on Foreign Relations, New York, NY.
Terrones, M., Kose, M.A. and Claessens, S. (2011), “How do business and fnancial cycles
interact?”, IMF Working Paper No. 11/88, International Monetary Fund, Washington, DC.
Further reading
FSB (2012), “Thematic review on deposit insurance systems: peer review report”, Financial
Stability Board, Basel, February 2012.
Skorepa, M. (2014), “Interaction of capital buffers in a banking group”, Financial Stability Report
2013/2014, Czech National Bank, pp. 128-136.
Corresponding author
Michal Skorepa can be contacted at: [email protected]
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
JFEP
7,3
220
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_570481447.pdf
The purpose of this paper is to assist the numerous regulators around the globe who are
currently considering ways to impose domestic systemic importance-based capital requirements on
banks
Journal of Financial Economic Policy
Capital buffers based on banks’ domestic systemic importance: selected issues
Michal Skorepa J akub Seidler
Article information:
To cite this document:
Michal Skorepa J akub Seidler , (2015),"Capital buffers based on banks’ domestic systemic
importance: selected issues", J ournal of Financial Economic Policy, Vol. 7 Iss 3 pp. 207 - 220
Permanent link to this document:
http://dx.doi.org/10.1108/J FEP-07-2014-0040
Downloaded on: 24 January 2016, At: 21:52 (PT)
References: this document contains references to 27 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 50 times since 2015*
Users who downloaded this article also downloaded:
Kangbok Lee, Wenling Lu, (2015),"Do bank regulation and supervision matter?: International
evidence from the recent financial crisis", J ournal of Financial Economic Policy, Vol. 7 Iss 3 pp.
275-288 http://dx.doi.org/10.1108/J FEP-03-2015-0019
Gurbachan Singh, (2015),"Thinking afresh about central bank’s interest rate policy", J ournal of
Financial Economic Policy, Vol. 7 Iss 3 pp. 221-232 http://dx.doi.org/10.1108/J FEP-08-2014-0051
Abdullah Noman, Mohammad Nakibur Rahman, Atsuyuki Naka, (2015),"Portfolio investment outflow
and the complementary role of direct investment", J ournal of Financial Economic Policy, Vol. 7 Iss 3
pp. 190-206 http://dx.doi.org/10.1108/J FEP-03-2014-0024
Access to this document was granted through an Emerald subscription provided by emerald-
srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.
*Related content and download information correct at time of
download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Capital buffers based on banks’
domestic systemic importance:
selected issues
Michal Skorepa
Czech National Bank, Prague, Czech Republic, and
Jakub Seidler
Financial Stability Department, Czech National Bank, Prague,
Czech Republic
Abstract
Purpose – The purpose of this paper is to assist the numerous regulators around the globe who are
currently considering ways to impose domestic systemic importance-based capital requirements on
banks.
Design/methodology/approach – The article discusses in some detail a number of issues from the
viewpoint of regulatory practice, mentioning relevant literature where available. Comments partly
refect the experience that the Czech National Bank gathered over the past two years while preparing its
own regime of domestic systemic importance-based capital requirements on banks.
Findings – The authors stress, among other points, one weakness of the (otherwise well-designed)
method suggested by the Basel Committee for Banking Supervision (BCBS) for assessment of banks’
systemic importance: the method is “relative” in that it does not refect the absolute importance of the
banking sector for the economy. The paper also explains that in some cases, use of individual-level
rather than consolidated-level data may be preferable, in contrast to what the BCBS guidance suggests.
Further, implications of the buffers over a longer term are pointed out.
Originality/value – As far as the authors are aware, this article is the frst to comprehensively discuss
the main issues surrounding both key steps (systemic importance assessment and determination of
buffer level) in the process of introducing buffers based on domestic systemic importance. A number of
questions related to these two steps are raised which regulators may appreciate to be reminded of, even
if some of the questions are such that it is not possible to give a generally applicable answer to them.
Keywords Banks, Regulatory change, Capital, Financial risk and risk management
Paper type General review
1. Introduction
The fact that bank regulation may need to be sensitive to a bank’s systemic importance
has been acknowledged for a number of years (BCBS, 2002). The debate on appropriate
ways to measure systemic importance and to adapt regulation to it intensifed in light of
JEL classifcation – G21, G28
The authors would like to thank Christian Castro, Jan Frait, Václav Hausenblas, Jan Kubíc?ek,
Sergio Masciantonio, Gor Sahakian, Jan Sobotka, Boøek Vašíc?ek and an anonymous referee for
helpful comments and useful recommendations. However, all errors and omissions are those of the
authors. The views expressedinthis paper are those of the authors andnot necessarilythose of the
Czech National Bank.
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1757-6385.htm
Banks’
domestic
systemic
importance
207
Received1 July2014
Revised5 November 2014
6 January2015
Accepted6 January2015
Journal of Financial Economic
Policy
Vol. 7 No. 3, 2015
pp. 207-220
©Emerald Group Publishing Limited
1757-6385
DOI 10.1108/JFEP-07-2014-0040
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
the dramatic repercussions caused by the failure of Lehman Brothers in September of
2008. The result of this debate is the Basel Committee for Banking Supervision’s
(BCBS’s) guidance on the setting of buffers for globally systemically important banks
(G-SIBs) and domestically systemically important banks (D-SIBs) – see BCBS (2013,
2012), respectively. As a result, implementation of systemic importance-based buffers is
currently a highly topical regulatory issue across the globe.
The BCBS guidance suggests measuring systemic importance based on a set of
indicators which are likely to correlate with the bank’s systemic importance; none of
the model-based approaches suggested in the academic literature (surveyed by
Bisias et al., 2012; de Bandt et al., 2013) is viewed as robust enough (yet) to be used
as a basis for actual regulatory measures (Berg, 2011; Löffer and Raupach, 2013).
Once individual banks’ systemic importance is assessed – expressed quantitatively as
their “SIB scores” – the guidance relies primarily on the principle of “equal expected
impact” (see BCBS, 2013, Appendix II) to determine the appropriate level of the buffer.
In this article, we discuss some of the issues that arise when a regulator intends to
introduce, more or less along the lines of the above BCBS guidance, capital buffers based
on banks’ domestic systemic importance. Appropriate measurement of the degree of a
fnancial institution’s systemic importance and the question of whether and how to
refect this degree in regulation and supervision are issues whose study faces a number
of conceptual as well as empirical challenges. For this reason, our comments, while
partly refecting the hands-on experience that the Czech National Bank gathered over
the past two years during the process of preparing its own regime of domestic systemic
importance-based capital requirements on banks, will at multiple points have to refer to
common sense or intuition, rather than robust conclusions of theoretical or empirical
literature. Indeed, literature focusing on this topic is still relatively scarce as the D-SIB
regulation is a rather recent addition to the regulatory framework. Thus, our
step-by-step discussion of establishing D-SIB regulation based on the experience of a
particular regulatory authority is the main value added of this article and contribution to
the on-going policy debate.
The article is structured as follows. Section 2 discusses some issues concerning the
measurement of systemic importance. Section 3 then focuses on estimation of
appropriate capital buffers for individual banks based on their systemic importance.
Section 4 summarises the main conclusions.
2. Calculation of systemic importance
2.1 Some possible indicators beyond the BCBS guidance
In contrast to the G-SIB methodology, the BCBS does not go into detail on the indicators
that belong to each category or on the method for calculating the D-SIB score itself (for
more on the G-SIB guidance and the D-SIB guidance, see BCBS, 2013; 2012). Given the
general tone that the BCBS guidance on D-SIBindicators has, one possibility is to use the
indicator list applied within the G-SIB methodology, summarised in Table I, as a basis
and modify it by dropping or, conversely, taking on board various indicators and
adjusting the weights of the indicators and/or indicator categories (such as by moving
cross-jurisdictional claims and liabilities under complexity) as deemed appropriate from
the perspective of the domestic fnancial system or for other relevant reasons[1].
For example, the list of indicators within the interconnectedness category might be
augmented with a measure of concentration of interbank connections (measured, for
JFEP
7,3
208
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
example, as the ratio of the bank’s top three interbank claims to its total interbank
claims, and similarly for liabilities). The idea – which seems relevant even at the G-SIB
level – is that a higher concentration of assets or liabilities implies stronger contagion:
the higher is the proportion of Bank A’s assets that have been lent to Bank B, the lower
is the probability that Bank A will survive Bank B’s failure to repay the loan; an
analogous proportionality holds for concentration of liabilities (for some empirical
evidence pointing in this direction, see Craig et al., 2013).
In countries where the banking sector is the dominant source of funding for the
corporate sector, the substitutability category may, in turn, be augmented with the share
of banks’ total loans provided to non-fnancial corporations. This indicator refects the
importance of the particular bank from the perspective of providing funds to the real
sector. The higher share the bank has, the more signifcant would be the problemcaused
to the real economy in the event of distress of the bank. Conversely, in some countries,
the banking sector might be exposed to some sectors to such an extent that any bank
distress and resulting fre sale of exposures might lead to a signifcant drop in prices of
the instruments concerned, adversely affecting the whole sector (Shleifer and Vishny,
2011). If this is the case, the list of indicators might be augmented to refect this
specifcity of the domestic banking sector as well, e.g. a high concentration of the
domestic banking sector towards sovereign bonds of domestic government.
In highly concentrated banking sectors, another possible substitutability indicator is
the share in total insured deposits (or a proxy for this share where the data are not
available) because bank failure and reimbursement of insured depositors may entail
high costs for public fnance. This risk is high in a number of situations: when there is no
deposit guarantee fund, when it is small relative to the deposits paid back, or when it is
of the ex post type and there are doubts about the possibility to collect enough
contributions from banks suffciently quickly ex post. The share-of-insured-deposits
Table I.
List of indicators
applied within the
G-SIB methodology
Category (and weighting) Individual indicator Indicator weighting (%)
Cross-jurisdictional activity (20%) Cross-jurisdictional claims 10
Cross-jurisdictional liabilities 10
Size (20%) Total exposures as defned for use
in the Basel III leverage ratio
20
Interconnectedness (20%) Intra-fnancial system assets 6.67
Intra-fnancial system liabilities 6.67
Securities outstanding 6.67
Substitutability/fnancial
institution infrastructure (20%)
Assets under custody 6.67
Payment activity 6.67
Underwritten transactions in debt
and equity markets
6.67
Complexity (20%) Notional amount of over-the-
counter derivatives
6.67
Level 3 assets 6.67
Trading and available-for-sale
securities
6.67
Source: BCBS
209
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
indicator might therefore refect the possible costs incurred by taxpayers if banks
experience distress.
If it is decided that systemic importance will be measured on an individual basis (an
issue to which we return below), then information on the relative size of the bank’s
sub-consolidated group can still be included via a new complexity indicator which will
express the sub-consolidated group’s assets as a multiple of the bank’s solo assets. Our
approximation of the level of complexity of a given bank’s liquidation may be improved
by means of data on its number of branches and number of employees, as this indirect
information refects the fragmentation of the bank’s business and the increased
diffculty of resolving the bank in the event of the bank’s liquidation (Parker, 2011).
Furthermore, data on the share of non-performing loans might also be a useful indicator
from the perspective of the complexity of a bank’s liquidation, as non-performing loans
(NPLs) are typically harder to evaluate and sell than standard loans (for empirical
support, see Klingebiel, 2000).
Taking the perspective of individual banks, investors or academia, that is, those
observers who do not have access to confdential regulatory data on all the banks,
Brämer and Gischer (2011) and Masciantonio (2013) suggest modifcations to the list of
indicators used such that the D-SIBscores are based purely on those data that, while still
refecting a bank’s systemic importance, are publicly available.
Another issue as regards the selection of indicators is the problem of high
correlations: should we rely on empirics, that is, discard all indicators that are highly
correlated with others, or should we rely on theory/intuition, that is, include all
indicators that are available and seem relevant? An argument in favour of the latter
option is that the high correlation based on historical observations might decrease in the
future. An argument in favour of the former option is that if two indicators are highly
correlated because they convey the same aspect of systemic importance, including both
such indicators in the calculation of the SIBscore will increase the overall weight of that
particular aspect, possibly above the level that would be appropriate. At this early stage
of the SIB score debate, however, little is known about the appropriate weights of the
various aspects of systemic importance, and so it seems prudent to include all relevant
indicators, even if some of the correlations are very high. Moreover, the degree of
correlation may happen to be high over the recent past but may fall in the future, so that
a choice of indicators on the basis of their cross-correlations may imply undesirable
instability in the list of indicators.
2.2 More fundamental departures from the BCBS guidance
Besides the new indicators mentioned above, there are also several other dimensions in
which the D-SIB framework could (and in some cases should) depart from the G-SIB
methodology more fundamentally.
First, as already hinted above, in the G-SIB framework, systemic importance is
measured as purely relative. This means that G-SIB scores express importance relative
to the rest of the banking sector alone, rather than importance as such, that is, the
absoluteimportance to the economy as a whole. In the G-SIB framework, the G-SIB
scores are calculated for the single purpose of being used as a basis for determining the
appropriate G-SIB buffer rate. In that case, their relative nature is not a problem as long
as the subsequent calculation of the buffer rate somehow also manages to incorporate
the importance of the banking sector to the economy, so that even if G-SIBscores as such
JFEP
7,3
210
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
are relative, the ultimate buffer rates based on them are absolute. The issue of whether
the D-SIB procedure for setting the buffer rates does manage to incorporate the
importance of the banking sector to the economy will be tackled in the next section.
If a bank’s D-SIB score itself were to be used as a proxy for actual systemic
importance (to determine, for instance, the set of banks which should develop their own
resolution plans), then the calculation of the score should obviously be somehow
extended, compared to the G-SIBframework, such that it becomes absolute in the above
sense. The ambition to measure actual systemic importance in this absolute sense, i.e. as
the absolute volume of costs of a given bank’s failure to the economy, raises a number of
issues though. For example, we would need to specify exactly what state of affairs we
mean by “failure of the bank”, i.e. at what stage of a bank’s process of failing (before
insolvency or after it) do we measure the costs to the economy. The relative approach
simply assumes that the stage is the same for all banks; it need not be specifed
explicitly.
The relative nature of SIB scores (whether in the G-SIB context or the D-SIB context)
may seemto be a problemnot only because the scores fail to showthe absolute systemic
importance, but also for another reason (raised, for instance, in IIF, 2011): the
relativeness implies that if Bank A changes its business model to reduce its SIB score
(motivation for such changes is one of the intended consequences of SIB frameworks),
the ultimate effect on its SIB score will be reduced (or neutral or even opposite) if the
remaining banks take similar steps to a smaller (or comparable or bigger) extent. The
reward, in the form of a fall in the D-SIB buffer rate, that Bank A will get for its effort to
become less systemic may then end up being lower than proportional (or the rate may
stay the same or it may rise). But here again, all that is needed is the incorporation of
absolute importance via the buffer rates. If absolute importance does get incorporated,
then Bank A’s effort to reduce its SIBscore will be rewarded: even if, for example, its SIB
score as such does not change at all because the other banks take the same steps towards
reducing their SIB scores, the systemic importance of the banking sector as a whole will
fall and the SIB buffer rates of all the banks, including Bank A, will fall. The sensitivity
of a bank’s SIB score to what other banks do may also be removed, for instance, if we
calculate the SIB score from “standardised” values of the indicators. By “standardised”
we mean that the value of a bank on a given individual indicator is measured as the
bank’s percentile position within a range whose lower and upper bounds are fxed and
the same for all banks. Under this approach, however, our choice of the bounds would
have a clear impact on the effective weights of the individual indicators in the
determination of the overall SIB score.
A second direction in which we may want the D-SIB framework to depart from the
G-SIB one is to give up the cardinal approach of calculating a bank’s D-SIB score as a
weighted average of the relevant indicators and to content ourselves with the ordinal
approach of concluding that a bank is highly systemically important if it exceeds a
certain threshold for at least one of the indicators. If we opt for this departure, however,
either we end up imposing the same buffer rate on all those banks which we have found
to be highly systemically important or, if we want the buffer to be commensurate with
the systemic importance of banks in that “highly systemically important” group, we still
need something akin to D-SIB scores and, thus, may be forced to return to a cardinal
approach.
211
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
A third direction in which departure from the G-SIB framework may be considered
concerns the question of whether systemic importance is judged at the level of individual
banks or their groups consolidated for regulatory (rather than accounting) purposes.
The BCBS suggests the consolidated approach (for both the G-SIB and D-SIB context).
In principle, this suggestion is reasonable: if we use solo data for setting the D-SIB
buffer, the bank may successfully reduce its D-SIB score artifcially as regards, for
instance, interconnectedness, by lending money to another bank via a non-bank special
purpose vehicle (SPV) established within the lending bank’s group as a subsidiary or an
affliate. In this way, the lending bank’s resulting solo claim on the SPV does not get
included in the lending bank’s value score for interconnectedness (as long as it is
bank-to-bank interconnectedness). Using the consolidated approach, and assuming that
it also covers the SPV, the claim of the group on the borrowing bank does get recorded
as a claim of the group on a bank and, thus, does increase the lending bank (group)’s
interconnectedness score.
In reality, however, consolidated data (in the regulatory sense) need not – and often
will not – be a good basis for assessing the impact of the group’s distress on the domestic
economy, that is, of the group’s domestic systemic importance. First, consolidated data
usually exclude certain domestically operating entities (especially non-fnancial entities)
whose distress may have sizeable impact on the domestic economy; in this regard, the
use of consolidated data has no value added compared to solo data. Consolidated data
will typically also exclude insurers. This exclusion will cease to be a major problemonly
after the framework for regulation of domestic systemically important insurers is
activated in a given jurisdiction. Second, consolidated data often include entities
operating abroad, whose distress may have a negligible effective impact on the domestic
economy, while foreign branches – which may often make the bank’s resolution
perceptibly more diffcult – are already included in unconsolidated data.
Therefore, consolidated data, just like solo data, should be taken as a kind of proxy
for the actual type of data needed to assess domestic systemic importance correctly. The
choice between using solo or consolidated data should be made after considering
relevant particular features of the relevant banking and fnancial system and the
resulting risk of psychological or other contagion among members of the respective
groups. Data availability and quality (especially as regards more distant history) may
play a role, too. Indeed, the European Union’s Capital Requirements Directive allows the
systemic importance-based buffers to be set on an individual as well as
(sub-)consolidated level. For a treatment of the related but separate issue of the
interaction of a G-SIB buffer on the group and a D-SIB buffer on a subsidiary, see
Skorepa and Seidler (2014).
One issue where the BCBS guidance leaves the decision fully to national authorities
is the extent to which the national D-SIB regime covers not only banks and subsidiaries
of foreign banks, but also foreign banks’ branches. If the D-SIB scores are calculated
solely for the purpose of determining the appropriate rate of the D-SIB buffers for
individual banks, then inclusion of foreign banks’ branches is not needed because the
host regulator is not able to impose a capital buffer on them anyway. On the other hand,
as long as the branches are not highly systemically important, their inclusion is not a
problem, as it will not change the banks’ D-SIBscores much. Conversely, if at least some
of the foreign banks’ branches are likely to be highly systemically important for the
domestic economy, then the calculation of D-SIB scores for the purpose of setting the
JFEP
7,3
212
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
D-SIBbuffer rates should not include them; in that case, however, a separate calculation
with the branches included may make sense to fnd out their systemic importance and to
determine which of them – given that the host regulator does not regulate them – may
require at least more intensive host supervision (see, e.g. MAS, 2014). When
communicating this type of calculation, the host authorities must, of course, avoid
giving the impression to markets and the public that they assume any responsibility for
the stability of these branches.
3. Calculation of the D-SIB buffer rate
In this section, we describe in detail the central procedure suggested by the BCBS for
calculating the SIB buffer rate based on the principle of “equal expected impact” and we
comment on selected aspects of this procedure.
3.1 Assumptions
In line with the BCBS publications, we start with the following assumptions set out in
Basel III when determining the capital buffer based on the SIB score:
• Each bank must meet a minimum capital requirement for Common Equity Tier 1
capital (CET1) of k
min
?4.5 per cent of risk-weighted assets. CET1 is composed of
common stock and retained earnings, i.e. capital that can be used immediately and
unconditionally to cover any losses of the bank.
• In normal circumstances, each bank additionally holds the full basic component of
the CET1 capital conservation buffer[2] of k
basic
? 2.5 per cent of risk-weighted
assets.
• In normal circumstances, a bank with an SIB score denoted by sib should comply
not only with k
min
and k
basic
, but also with the SIB buffer, i.e. the full SIB
component, k(sib), of the CET1 conservation buffer.
Consequently, of the three components of the total CET1 capital requirements listed
above, only k(sib) is sensitive to the bank’s SIB score.
If the capital of the bankfalls belowk
min
?k
basic
?k(sib), the bankmust take remedial
action whose intensity (and thus also the costs to the economy arising fromthe situation)
is proportional to the decline in capital. In what follows, the situation where, as a result
of a large negative proft in the quarter, the bank’s CET1 falls below the regulatory
minimumk
min
, i.e. it records a negative quarterly proft of ?[k
basic
?k(sib)] or lower, will
be referred to as distress (this need not mean a straightforward fall in the sense of a loss
of licence)[3]. The probability P(sib) of distress for a bank with an SIB score of sib is
obviously lower for a higher SIB capital buffer k(sib), i.e. for a higher level of sib. The
costs to the economy arising from the distress of a bank with an SIB score equal to sib
will be denoted C(sib).
The capital buffer is then determined on the basis of the “equal expected impact”
principle (for an early suggestion of this principle, see Squam Lake Working Group,
2009). This principle can be generally expressed as follows: the expected costs to the
economy resulting fromdistress of any bank that is systemically more important than a
certain reference bank chosen by the regulator should be the same as the expected costs
to the economy, resulting fromdistress of the reference bank. Of course, one loose end of
this principle is how the regulator selects the reference bank; we return to this question
shortly.
213
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Besides the expected impact approach, BCBS (2013) uses the results of other
approaches (models created by the Long-term Economic Impact Group and a method
based on the implicit subsidies that some highly systemically important banks receive
because the market considers them to be too big to fail (TBTF), i.e. it expects public
money to be spent on bailing them out if they get into diffculty). In many jurisdictions,
however, these approaches will be diffcult to apply due to a lack of relevant data and/or
models.
According to the expected impact principle, the point of the SIBbuffer is to reduce the
probability P(sib) of distress of the bank such that the expected costs of this situation, i.e.
C(sib)*P(sib), are equal to the expected costs of distress of the reference bank, i.e.
C(sib
R
)*P(sib
R
). It is obvious that the SIBbuffer will be zero for the reference bankandfor
every systemically less important bank.
3.2 Calculation
BCBS (2013) uses two methods to determine the SIB buffer according to the expected
impact principle. The frst method uses a Merton model to estimate a bank’s
market-perceived probability of failure from the market prices of its equity. The second
method is based on the historical frequency distribution of the return on risk-weighted
assets (RORWA – see Kuritzkes and Schuermann, 2010). Nevertheless, in many
countries, many of their domestic banks’ shares are not traded on public markets (or not
traded intensively enough to make the data suffciently reliable), and so the Merton
model has limited application. The same problem arises if we want to use for this
purpose market prices or yield spreads of banks’ bonds or other liabilities (such as
uninsured deposits). For this reason, we focus in this article on the RORWAmethod. For
an example of actual use of the method in regulatory practice, see Skorepa and Seidler
(2013) who describe the analytical basis of the approach selected by the Czech National
Bank.
The expected impact principle can be expressed formally as follows: for all sib ?sib
R
,
P(sib) should satisfy:
P
(
sib
)
* C
(
sib
)
? P
(
sib
R
)
* C
(
sib
R
)
, i.e.
P
(
sib
)
? P
(
sib
R
)
/
?
C
(
sib
)
/C
(
sib
R
)?
. (1)
To derive the values of P(sib) and subsequently also the capital buffer k(sib) based on the
bank’s sib, we frst need to determine the value of P(sib
R
). The frst step is to choose the
level of sib
R
itself. While the sib for each bank is given by the empirically observed levels
of the various indicators for that bank, there is no natural, obvious, objective way to set
sib
R
; it thus has to be determined on the basis of regulatory considerations.
One method which seems fairly acceptable and transparent is to set sib
R
equal to q
times sib
aver
, where the latter is the average sib for the entire domestic banking sector
and values of q ? 1 are considered. The value of q is chosen at the discretion of the
regulator, depending on how weak and narrow it wants the effect of the SIB buffer
regime to be: the higher q is, the lower the buffers will be; moreover, increasing q may
reduce the set of banks to which the buffers will apply. For example, the G-SIB regime
appears to work with q ? 1; the Czech National Bank opted for q ? 2 (Skorepa and
Seidler, 2013). Alternatively, sib
R
can be set equal to the q-th percentile of the distribution
JFEP
7,3
214
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
of all SIB scores observed in the banking sector. Here, again, q measures how weak and
narrow the effect of the SIB regime is to be.
Not surprisingly, the different methods for setting sib
R
have somewhat different
implications. For example, if the dispersion of SIB scores across the relevant
banking sector falls below a certain level, then the q-times-the-average method will
imply no SIB buffers, while the q-th percentile-based method will, by construction,
always imply SIB buffers for the 1-q per cent of banks with the highest SIB scores (in
regulatory practice, some or all of these buffers may be so small as to be rounded to
zero). The intuitive appeal of the two methods may thus depend on the particular
circumstances.
Assumptions (i)-(iii) listed above and the assumption k(sib
R
) ? 0 imply that P(sib
R
)
corresponds to the probability that the bank will make a negative proft of:
??k
basic
? k
(
sib
)
? ? ?
(
2.5 ? 0
)
? ?2.5% of RWA
or lower. With the historical RORWA distribution, this is therefore the relative
frequency of cases where RORWA ??2.5 per cent. If we simultaneously interpret the
historical RORWA distribution as being the RORWA probability distribution in the
future, then:
P
(
sib
R
)
? p(RORWA ? ?2.5%).
To calculate P(sib) from equation (1), we now need to determine the value of C(sib)/
C(sib
R
). In accordance with intuition and with a proposal contained in BCBS (2013), we
can assume for simplicity that this ratio can be approximated as sib/sib
R
. Using the
historical RORWA distribution, we can then derive the minimum capital loss for each
level of P(sib).
The capital requirement k
basic
?k(sib) ?2.5 ?k(sib) should be of an amount covering
this loss. This gives us the SIB capital buffer k(sib) based on the degree of systemic
importance of the bank.
As a practical matter, the level of the buffer is likely to be rounded by the
regulator before it is imposed. This rounding may bring with it some welcome
stabilisation in the sense that it reduces the frequency of change in the D-SIB buffer
level. Naturally, the resilience of buffers to excessive volatility can be further
enhanced by calculating the D-SIB scores not from the values of source indicators as of
a single date, but from longer-term averages. On the other hand, the stability of D-SIB
buffers must not be so strong as to limit their “motivational” effect: it must not lead to a
situation where a bank’s efforts to reduce its D-SIB buffer by reducing its systemic
importance take too long to bear fruit.
To close this section, a short note on determining the overall capital requirement for
a bank may be in order. The SIBbuffer is motivated by the potential losses that the bank
may impose on the rest of the economy. In contrast, the existing capital requirements on
banks, whether within Pillar 1 or Pillar 2, have all been motivated by the potential losses
that the rest of the economy may impose on the bank. Given this fundamental difference
in the motivation of the existing capital requirements and the SIB buffer, it seems
natural to add the latter buffer on top of any existing Pillar 1 or Pillar 2 capital
requirements.
215
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
3.3 Does the banking sector’s importance to the economy get refected in the buffer
rate?
Now, we return to the issue raised earlier of whether the above RORWA-based
procedure for setting the D-SIB buffer rates ensures that while based on relative D-SIB
scores, the buffer rates also encompass the importance of the banking sector to the
economy. Alas, the procedure does not meet this criterion: the relative frequency of
various levels of RORWAsays nothing about the implications that these levels have for
the economy.
As regards specifcally the G-SIBbuffers proposed in BCBS (2013), a cross-check that
the buffer rates suggested by the RORWA method are also absolutely adequate is
provided by the confrmation (conducted in Appendix II of BCBS, 2013) that capital
buffers of similar rates seem roughly optimal according to the macroeconomic models
used by the Long-term Macroeconomic Impact Working Group.
In principle, this “absoluteness cross-check” is possible also at the D-SIB level: after
the national regulator has calculated the D-SIB buffers using the relative approach, she
may verify that similar buffer rates are roughly optimal according to an appropriate
macroeconomic model. The problem, especially for smaller or less advanced economies,
is that signifcant parts of the data set required to estimate or calibrate such a model
reliably may not be available. At the same time, direct application of the models used in
BCBS (2013) may not give a reliable answer either, as they have been developed to ft a
handful of specifc economies, most of them large and/or advanced (see BCBS, 2010a).
3.4 Dynamics of the SIB regime
Aspecial class of issues arises once we take a longer-termperspective and consider how
the SIB scores and buffers may evolve over time. First, the shorter is the RORWA
history on which the buffers are based, the more procyclical the buffers are bound to be:
in boomtimes, the level of proft occurring with any given probability will rise and, thus,
the implied buffer rate will fall, and vice versa. The obvious way to avoid this
procyclicality is to work with an RORWAtime series long enough to contain at least one
full fnancial cycle – if such a long time series is available.
Second, assume that the introduction of the SIB buffers and other regulatory
measures aimed at reducing the probability of distress of individual banks does achieve
this aim. Then, the histogram of the new RORWA data (recorded after the buffers were
imposed) can be expected to lie on average to the right of the old data histogram, that is,
the frequency of losses of a given extent should be lower than it used to be before the
buffers were introduced. The reason for this effect is that higher capital buffers held by
the bank’s counterparties in the interbank market imply lower incidence of distress
among those counterparties and, thus, fewer credit or similar losses for the bank in
question. This consideration leads to the expectation that at the frst point of SIB buffer
rate re-assessment, the RORWA data are more likely to suggest that the buffers should
be lowered by some amount. When re-assessing the buffer rates for the second time, the
RORWA data are more likely to suggest that the buffers should be increased but by a
smaller amount, and so on.
3.5 Data issues
On a practical level, in this second stage of the D-SIB procedure, in which the D-SIB
buffer rates are determined, the issue of including or excluding branches resurfaces. The
JFEP
7,3
216
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
host regulator cannot impose this buffer on a branch of a foreign bank and so it makes
no sense to bother calculating the buffer for such branches. Nevertheless, there is still the
question of whether we want to include branches in the pool of historical RORWA data
used to determine the D-SIB buffer rates for domestic banks which are not branches. At
frst sight, it seems preferable not to include branches on the grounds that their fnancial
results may be tilted by the fact that they are not “complete banks” as they share or
outsource some of their operations within the bank they are part of. This train of
thought, however, loses steamas soon as we realise that even subsidiaries often share or
outsource some of their operations within their groups such that even many subsidiaries
are not “complete banks” (Fiechter et al., 2011). More pragmatically, it may be necessary
to collect RORWA data also from branches for the simple reason that domestic banks
that are not branches are too few or exist for too short a time.
A similar issue is whether to include in the pool of RORWA data also the starting
periods of newentrants in the domestic banking sector. Indeed, the frst several quarters
or even years of their existence may be viewed as a special “start-up” regime with
higher-than-usual investment expenditures on brand recognition and fght for a
satisfactory market share (DeYoung, 1999). But again, this fne consideration may have
to be disregarded and most of the newentrant data (except, perhaps, for a fewperiods at
the very start) may have to be used to obtain a suffciently rich historical RORWA data
set.
Nevertheless, the pool of historical RORWA data provided by domestic banks may
be too small even after including acceptable data for branches and new entrants. Or we
may fear that a major part of the past time period fromwhich the pool originates was too
non-representative of the way the domestic banking sector is expected to work in the
future (for which we want the D-SIB buffers to be adequate) – perhaps the sector or its
economic and/or legal environment underwent a transition period or a substantial
structural break. In these cases, data pertaining to the domestic banking history may
actually be a poorer guide for assessing the future than data pertaining to foreign
banking history. Under such circumstances, it may be preferable to determine the
overall level of D-SIB buffers based on foreign data. Of course, various adjustments are
possible to refect, for example, the difference in business models (and thus in riskiness)
of domestic banks relative to foreign ones (determinants of banks’ riskiness are studied,
for example, by Altunbas et al., 2011) or the difference in the amplitude of fnancial
cycles in emerging markets relative to developed economies (as documented by, for
example, Terrones et al., 2011).
4. Summary and conclusions
This article discusses some of the issues that arise when a regulatory authority intends
to impose capital buffers on domestic banks based on their domestic systemic
importance. In the current post-crisis environment, implementation of this type of
buffers is a highly topical issue around the globe.
We consider several specifc indicators of systemic importance that, while not used in
the BCBS guidance for capital buffers based on banks’ global systemic importance,
might be relevant on the domestic level. Furthermore, the usual assumption that
systemic importance should be calculated at the sub-consolidated (rather than solo)
basis is pointed out as being debatable whenever domestic banks control large
non-fnancial undertakings or their foreign subsidiaries operate mostly in very distant
217
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
economies. We then explain the advantages of calculating an overall SIB score (rather
than judging whether a bank exceeds thresholds for several indicators individually). We
also document the fact that the BCBS guidance results in systemic importance scores
which, in themselves, are only relative, in the sense that they capture the internal
structure of the banking sector but do not refect the importance of the sector to the
whole economy.
As regards the calculation of the buffer rate, we focus on the equal expected impact
approach and discuss the pros and cons of several distinct methods of arriving at the
cut-off point above which the buffer should be imposed on a bank – methods such as
clustering, a fxed multiple of the average systemic importance score and a fxed
percentile of the systemic score distribution. We also point out certain longer-term
implications of the SIB buffer regime. For example, the introduction of buffers may
gradually shift the observed frequency distribution of RORWA fgures and, thus, have
an impact on the future reassessment of the appropriate buffer rates.
Concerning public communication of the buffers, we would like to emphasise the
crucial signalling effects of public communication of the SIB regulation: the
macroprudential authority should make every effort to explain that the imposition of an
SIB capital buffer on a bank does not signal a guarantee of government bail-out. If this
communication fails (or if it actually equates TBTF banks with banks with an SIB
buffer, as is often the case in the media and some fnancial publications), then the
introduction of the SIB regime might make the TBTF problem even more pressing (for
a more detailed exposition of this problem, see Skorepa and Seidler, 2014).
Notes
1. The procedure for G-SIB scores is that for each indicator, the value for a given bank is
normalised by the value for the whole sample of banks under consideration and the G-SIB
score of each bank is obtained as a weighted average of the bank’s normalised values for all
indicators. By implication, the G-SIB score of each bank is a score relative to the rest of the
sample, and the sum of the G-SIB scores across all banks in the sample is 1.
2. We use the descriptor “basic” here because we take the “total” conservation buffer to include an
SIB buffer and a countercyclical buffer (where introduced): as regards the capital-conserving
implications for the bank of not meeting the buffers, both the countercyclical buffer and the
bank’s systemic importance-based buffer that Basel III sets the stage for should work the
same way as the capital conservation buffer.
3. In determining whether the return (income) is positive or negative, we take zero proft as the
benchmark, which seems in line with the approach used in BCBS (2010b, 2013). In contrast,
Kuritzkes and Schuermann (2010) argue that capital is to cover unexpected downside risk,
that is, any negative deviation from expected proft, and so they take as the benchmark the
bank’s expected proft (proxied by its average historical proft). This alternative might be
justifed from a long-term point of view, where avoiding negative values of net income is not
enough for the bank to survive, as the bank must actually meet its shareholders’ expectations
regarding dividends. These expected dividends can be viewed as an additional, special type
of expected expense. This train of thought would imply that, for the purposes of the SIBbuffer
size assessment, net income should be calculated net of this special type of expected expense,
that is, net of expected income. Here, however, we are concerned with the standard one-year
horizon for the assessment of capital adequacy.
JFEP
7,3
218
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
References
Altunbas, Y., Manganelli, S. and Marques-Ibanes, D. (2011), “Bank risk during the fnancial crisis:
do business models matter?”, ECB Working Paper No. 1394, European Central Bank,
Frankfurt am Main.
BCBS (2002), “Supervisory guidance on dealing with weak banks”, Report of the Task Force on
Dealing with Weak Banks, Bank for International Settlements, Basel, March.
BCBS (2010a), “An assessment of the long-term economic impact of stronger capital and liquidity
requirements”, Basel Committee on Banking Supervision, Bank for International
Settlements, Basel, August 2010.
BCBS (2010b), “Calibrating regulatory minimum capital requirements and capital buffers: a
top-down approach”, Basel Committee on Banking Supervision, Bank for International
Settlements, Basel, October 2010.
BCBS (2012), “A framework for dealing with domestic systemically important banks”, Basel
Committee on Banking Supervision, Bank for International Settlements, Basel, October
2012.
BCBS (2013), “Global systemically important banks: updated assessment methodology and the
additional loss absorbency requirement”, Basel Committee on Banking Supervision, Bank
for International Settlements, Basel, July 2013.
Berg, S.A. (2011), “Systemic surcharges and measures of systemic importance”, Journal of
Financial Regulation and Compliance, Vol. 19 No. 4, pp. 383-395.
Bisias, D., Flood, M., Lo, A.W. and Valavanis, S. (2012), “A survey of systemic risk analytics”,
Annual Review of Financial Economics, Vol. 4 No. 1, pp. 255-296.
Brämer, P. and Gischer, H. (2011), “Domestic systemically important banks: an indicator-based
measurement approach for the Australian banking system”, FEMM Working Paper No.
3/2012, Otto-von-Guericke University of Magdeburg.
Craig, B.R., Fecht, F. and Tümer-Alkan, G. (2013), “The role of interbank relationships and
liquidity needs”, Bundesbank Discussion Paper No. 54/2013, Deutsche Bundesbank,
Frankfurt am Main.
De Bandt, O., Héam, J.-C., Labonne, C. and Tavolaro, S. (2013), “Measuring systemic risk in a
post-crisis world”, Débats économiques et fnanciers No. 6, Autorité de Contrôle Prudentiel,
Banque de France.
DeYoung, R. (1999), “Birth, growth, and life or death of newly chartered banks”, Federal Reserve
Bank of Chicago, Economic Perspectives, Vol. 23 No. 3, 18-34.
Fiechter, J., Ötker-Robe, I., Ilyina, A., Hsu, M., Santos, A. and Surti, J. (2011), “Subsidiaries or
branches: does one size ft all?”, International Monetary Fund Staff Discussion Notes No.
11/04, International Monetary Fund, Washington, DC.
IIF (2011), “IIF views on the Basel committee’s consultative document on global systemically
important banks: assessment methodology and the additional loss absorbency
requirement”, Institute of International Finance, Washington, DC.
Klingebiel, D. (2000), “The use of asset management companies in the resolution of banking crises:
cross-country experience”, World Bank Policy Research Paper No. 2284, World Bank,
Washington, DC.
Kuritzkes, A. and Schuermann, T. (2010), “What we know, don’t knowand can’t knowabout bank
risk: a view from the trenches”, in Diebold, F.X., Doherty, N.A., Herring, R.J. (Eds), The
Known, the Unknown, and the Unknowable in Financial Risk Management: Measurement
and Theory Advancing Practice. Chapter 6, Princeton University Press, Princeton, NJ.
219
Banks’
domestic
systemic
importance
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Löffer, G. and Raupach, P. (2013), “Robustness and informativeness of systemic risk measures”,
Deutsche Bundesbank Discussion Paper No. 04/2013, Deutsche Bundesbank, Frankfurt am
Main.
MAS (2014), “Proposed framework for systemically important banks in Singapore”, Consultation
Paper, P008-2014, Monetary Authority of Singapore, Singapore.
Masciantonio, S. (2013), “Identifying and tracking systemically important fnancial institutions
(SIFIs) with public data”, MPRA Paper No. 46867, University Library of Munich, Munich.
Parker, D.C. (2011), Closing a Failed Bank: Resolution Practices and Procedures, International
Monetary Fund, Washington, DC.
Shleifer, A. and Vishny, R. (2011), “Fire sales in fnance and macroeconomics”, Journal of
Economic Perspectives, Vol. 25 No. 1, pp. 29-48.
Skorepa, M. and Seidler, J. (2013), “An additional capital requirement based on the domestic
systemic importance of a bank”, Financial Stability Report 2012/2013, Czech National
Bank, pp. 96-102.
Skorepa, M. and Seidler, J. (2014), “Capital buffers based on banks’ domestic systemic importance:
selected issues”, Czech National Bank Research and Policy Notes No. 1/2014, Czech
National Bank, Prague.
Squam Lake Working Group (2009), Reforming Capital Requirements for Financial Institutions,
Council on Foreign Relations, New York, NY.
Terrones, M., Kose, M.A. and Claessens, S. (2011), “How do business and fnancial cycles
interact?”, IMF Working Paper No. 11/88, International Monetary Fund, Washington, DC.
Further reading
FSB (2012), “Thematic review on deposit insurance systems: peer review report”, Financial
Stability Board, Basel, February 2012.
Skorepa, M. (2014), “Interaction of capital buffers in a banking group”, Financial Stability Report
2013/2014, Czech National Bank, pp. 128-136.
Corresponding author
Michal Skorepa can be contacted at: [email protected]
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
JFEP
7,3
220
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
5
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_570481447.pdf