Tải bản đầy đủ
4 ELA provided by other banks, coordinated by the supervisor or the central bank

4 ELA provided by other banks, coordinated by the supervisor or the central bank

Tải bản đầy đủ


Bindseil, U.

needs to be able to understand this financial risk taking, also to be sure that
it is reasonable and fair to invite banks to take it.
3.5 Moral hazard
FCM measures by central banks may be regarded as an insurance against
adverse outcomes, and, as any insurance, it would weaken incentives to
prevent bad outcomes to occur. Incentives issues are not generally seen to
prevent insurance from being useful for society, although in some cases, it
may be a reason to not insure some activity, which, in a world without any
moral hazard, would be ensured. Also, incentive issues are generally
something taken into account when designing insurance contracts (e.g. by
foreseeing that the insured entity takes a part of the possible losses). Thus
the negative impact of moral hazard on the welfare improvements from
insurances is not just a given fact, but can be minimized through intelligent
institutional design. Still, different schools on applying this to FCM can be
found in the literature. For instance de Grauwe (Financial Times, 12 August
2007, ‘ECB bail-out sows seeds of crisis’) suggests that the liquidity-injecting
open market operations by the ECB done in early August 2007 created
a moral hazard dilemma. Another school represented by e.g. Humphrey
(1986) or Goodfriend and Lacker (1999) arguing that only aggregate equal
access emergency liquidity injections (though open market operations) are
legitimate from an incentive point of view, while any individual access ELA
would be detrimental overall to welfare because of its moral hazard inducing
effects. Other authors like Flannery (1996), Freixas (1999) Goodhart (1999)
and most central bankers would in contrast argue that also ELA may be
justified (namely as explained in Section 3.3). The fact that an insurance
always has some distorting effects is not a proof that there should be no
insurance, in particular if bank failures have huge negative externalities. This
does not mean that there should always be a bail-out, but in some cases there
should. Also it does not mean that the bail-out should be complete in the
sense of avoiding losses to all stakeholders – quite the contrary (see below).
Another widespread view is that moral hazard and incentive issues imply
that crisis-related liquidity injections ‘should only be given to overcome
liquidity problems as a result of inefficient market allocation of liquidity’4,
and in particular not in the case of true solvency problems. However, as

See also Goodhart (1999, 352–3). For instance Goodfriend and Lacker (1999) seem to take this conservative view, and
also Sveriges Riksbank (2003, 64).


Central bank financial crisis management

argued above, the welfare considerations on solvency problems are not so
different from those regarding liquidity problems. In both cases, a rescue
scheme may be useful ex post, and also the incentives issues are similar after
all. So in fact, for both, one needs to accept that looking for the right design of
the ‘insurance contract’ is essential. In both cases, it seems important to not
take away incentives to monitor banks to those who are in a position to do so.
Moreover, it is difficult to distinguish insolvency and illiquidity ex ante.
Goodhard (1999, 352), for whom the claim that ‘moral hazard is everywhere and at all times a major consideration’ is one of the four myths that
have been established around ELA, distinguishes four groups of stakeholders
with regard to the incentives they need to be confronted with. The most
important decision makers and thus those most directly responsible for
liquidity or solvency problems are the senior management of the bank.
Therefore, often a bank requesting support from the central bank has to
sacrifice one or several of its responsible Board members. This was for
instance also the case in the most recent bank rescue operation, namely of
IKB in Germany in August 2007, in which both the CEO and the CFO had to
quit within two weeks. Goodhart (1999, 353) considers it a failure that the
executives of LTCM were never removed after the rescue of this hedge fund,
exactly for these incentive reasons. He notes that unfortunately ‘the current
executives have a certain monopoly of inside information, and at times of
crisis that information may have particular value’. Second, equity holders,
which have the highest incentives to ask the management to run a high-risk
strategy with negative externalities, should obviously suffer. Third, also bond
holders, and probably interbank market lenders should suffer if equity
holders cannot absorb losses, as these groups should still be made responsible
for monitoring their exposures. Fourth, it is according to Goodhart (1999,
354) today considered as socially wasteful to require ordinary small depositors to monitor their bank, and that some (though preferably not 100
per cent) deposit insurance for those would be justified.
To become more precise, consider quickly the moral-hazard relevance of
all types of FCM measures:
 A-I: OMO aggregate liquidity injection. This is supposed to be for the
benefit of all banks and the system, which has fallen or risks falling into
an inferior equilibrium, as a whole. Therefore, referring to moral hazard
would imply a collective responsibility of the system, which is difficult to
see. If banks have collectively embarked in reckless activity, then the
banking and financial market supervision, and/or the legislator have not
done their job either, and it may be difficult to hold responsible the


Bindseil, U.

competitive banking system which was led towards this by the
environment set or tolerated by the legislator. In any case, in times of
money market tensions, interbank interest rates move to levels above the
target rate set by the central banks, if the profile of the liquidity supply by
the central bank is unchanged. Therefore, it is natural for a central bank
to re-adjust the profile of liquidity supply to demand in order to stabilize
short-term interbank rates around their target. ¼> Moral hazard
considerations hardly relevant.
 A-II: lowering the penalty level associated with the borrowing facility. Same
arguments. ¼> Moral hazard considerations hardly relevant.
 A-III: widening the collateral set. Admittedly, this type of measure could
have moral hazard implications, particularly (1) if the general widening
of the collateral set in fact targets a small number or, say, even a single
bank under liquidity stress, which is rich in the specific type of additional
collateral; and (2) if the central bank offers facilities, for instance a
standing borrowing facility, to effectively refinance such collateral. This
type of action invites moral hazard as it may indeed be decisive to
establish whether the single bank fails or not, while at the same time
sparing to the banks’ management and shareholders the substantial costs
associated with resorting to real emergency liquidity assistance.
Therefore, a widening of the collateral set accepted for monetary policy
purposes should probably only be considered if this measure would
substantially help a significant number of banks and if the set of assets
were very narrow. In this case the lack of collateral obviously seems to be
more of a systemic issue, and the central bank should consider taking
action.5 ¼> Moral hazard is an issue, in particular if problems are
limited to a few banks.
 B: individual ELA. Moral hazard is in principle an issue, but it can be
addressed by ensuring that shareholders and managers are sanctioned.
Being supported individually by the central bank always comes at a
substantial cost to the bank’s management, which can expect that the
central bank / supervisor will ask for the responsible persons to quit, and
which will control closely the actions of the bank for some while. Also the
reputation damage of a liquidity problem is substantial (as soon as the
support becomes public). In so far, it seems implausible to construct a

The Institute of International Finance (2007) proposes the following guiding principle with regard to moral hazard:
As a principle, central banks should be more willing to intervene to support the market and its participants and be
more lenient as to the type of collateral they are willing to accept, if the crisis originates outside the financial industry.


Central bank financial crisis management

scenario in which the bank deliberately goes for much more liquidity risk
than would be optimal from the point of view of society. This does not
mean that the calculus of the bank is untouched by the perspective to be
bailed out. But probably the distortion remains weaker than the one that
might be caused by solvency aid – as far as the two are clearly distinct.
Also in the case of solvency aid, the authorities should ensure, to the
extent possible ex ante and ex post, that in particular shareholders and
senior managers suffer from losses. ¼> Moral hazard is an issue, but
can be addressed to a significant extent.
 C: Public authorities as catalysts for peer institution’s help. Again, the
public authorities and the helping institutions can and should ensure that
shareholders and senior management are sanctioned. ¼> Moral hazard
is an issue, but can be addressed to a significant extent.
In sum, it is wrong to speak generally about moral hazard associated to FCM
measures, since it makes a big difference what type of FCM measure is
taken. An individual ELA (and solvency aid) framework can be designed
with a perspective to preserve to the extent possible the right incentives, the
optimum having to be determined jointly with the prudential supervision
rules. In the optimum, some distortions will still occur, as they occur in
almost any insurance or agency contract. Recognizing the existence of these
distortions is not a general reason for concluding that such contracts should
not exist at all. In the case of individual ELA, the concrete issue of incentives
may be summarized as stated by Andrew Crocket (cited after Freixas et al.
1999, 161): ‘if it is clear that management will always lose their jobs, and
shareholders their capital, in the event of failure, moral hazard should be
alleviated’. For equal access widening of collateral, moral hazard issues are
potentially tricky, and would deserve to be studied further. Risk management expertise of the central bank is relevant in all this because for all
measures except A-I and A-II, asset valuation, credit quality assessment and
haircut setting are all key to determine to what extent the different measures
are pure liquidity assistance, and when they are more likely to turn out to
also consist of solvency assistance. In the latter case, moral hazard issues are
always more intense that in the former.
3.6 Constructive ambiguity
‘Constructive ambiguity’, which is a term due to Corrigan, is considered
one important means to limit moral hazard. The idea is that by not
establishing any official rules on FCM measures, banks will not count on it


Bindseil, U.

and their incentives to be prudent will not be weakened. Still, ex post, the
central bank may help. Already Bagehot (1873, chapter 7) touches on the
topic, and taking the perspective of the market, criticizes the ambiguity
surrounding the Bank of England’s FCM policies (whereby it needs to be
admitted that this refers to equal access FCM measures, and not to what
today’s debates mainly have in mind, which is individual ELA):
Theory suggests, and experience proves, that in a panic the holders of the ultimate
Bank reserve (whether one bank or many) should lend to all that bring good
securities quickly, freely, and readily. By that policy they allay a panic; by every
other policy they intensify it. The public have a right to know whether the Bank of
England the holders of our ultimate bank reserve acknowledge this duty, and are ready
to perform it. But this is now very uncertain.

Apparently central bankers remained unimpressed by this claimed ‘right’ of
the public, and still 135 years after Bagehot, the financial industry continued
expressing the wish for more explicitness by central banks (Institute of
International Finance 2007, 42):
Central banks should provide greater clarity on their roles as lenders of last resort in
both firm-specific and market-related crises . . . Central banks should be more
transparent about the process to be followed during extraordinary events, for
example, the types of additional collateral that could be pledged, haircuts that could
be applied, limits by asset type (if any), and the delivery form of such assets.

Freixas (1999) proposes an explicit model of the role of constructive
ambiguity, in which he shows that mixed strategies, in which the central
bank sometimes bails out and sometimes does not, can be optimal when
taking into account the implied incentive effects. In mixed strategies, a
player in a strategic game randomizes over different options applying some
optimal probabilities, and it can be shown that such a strategy may maximize the expected utility of the player (in this case the central bank, the
utility of which would be social welfare; see e.g. Myerson (1991, 156) for
a description of the concept of mixed equilibrium, or any other game
theory textbook). In so far, constructive ambiguity could be considered as
reflecting the optimality of mixed strategies. As it may however be difficult
to make randomization an official strategy (as this would raise legal problems), hiding behind ‘constructive ambiguity’ may appear optimal. Another
interpretation of constructive ambiguity could be that it is a doctrine to
avoid legal problems: if there would be a clear ELA policy, the central bank
would probably be forced to act accordingly, and not become exposed to


Central bank financial crisis management

legal proceedings. In addition, even if it follows in principle its policies,
cases may be so complex that legal proceedings may always be opened
against the central bank in order to blame it for losses that eventually
In particular three central banks worldwide have made attempts to be
transparent on their ELA policies, thereby rejecting in principle the doctrine
of constructive ambiguity, namely Sveriges Riksbank (2003, 58), Hong
Kong Monetary Authority (1999), and Bank of Canada (Daniel et al. 2004).6
Of course, all of these three central banks do not promise ELA under any
conditions, they only specify necessary conditions for ELA, such that they
could in principle still randomize over their actual actions. The specifications they provide are to a substantial part focused on risk management.
From a risk manager’s perspective, one would generally tend to be
suspicious of the concept of constructive ambiguity. It is the sort of concept
which triggers alarm bells in the risk manager’s view of the world, since it
could be seen to mean non-transparency, discretion and deliberate absence
of agreed procedures (see also e.g. Freixas et al. 1999, 160). In the risk
manager’s perspective in a wider sense, the following doubts could be
expressed vis-a`-vis constructive ambiguity, if it is supposed to mean that no
philosophy, principles and rules have been studied even internally in the
central bank. First, it could be interpreted to reflect a lack of thinking by
central banks, i.e. an inability to formulate clear contingent rules on when
and how to conduct FCM measures. Second, it will imply longer lead times
of actual conduct of FCM measures, and more likely wrong decisions.
Third, constructive ambiguity would concentrate power with a few senior
management decision makers, who will not be bound to policies (as such
policies would not exist, at least not in an as clear way as if they were written
down) and will have more limited accountability.
If constructive ambiguity is supposed to mean that a philosophy, principles and rules should exist, but are kept secret, then the following points
could still be made. First, it could still seem to be the opposite of transparency, a value universally recognized nowadays for central banks, in


For instance Sveriges Riksbank (2003, 58) explains: ‘Some central banks appear unwilling to even discuss the
possibility of possible LOLR operations for fear that this could have a negative effect on financial institutions’
behaviour, that is to say, that moral hazard could lead to a deterioration in risk management and to a greater risk
taking in the banking system. The Riksbank on the other hand, sees openness as a means of reducing moral
hazard . . . A well reasoned stance on the issue of ELA reduces the risk of granting assistance un-necessarily . . .
[and is] a defence against strong pressure that the Riksbank shall act as a lender of last resort in less appropriate


Bindseil, U.

particular if large amounts of public money are at stake.7 Second, it could
be argued that this approach would reiterate an old fallacy from macroeconomics, namely the idea that one can do things ex post, but as long as
they are not transparently described ex ante, they will not affect behaviour
ex ante. This fallacy was overcome in macroeconomics by the theory of
rational expectations. From the perspective of rational expectations theory,
it could be argued that having a policy but trying to be non-transparent on
it eventually does not mean that policies will not be taken into account ex
ante, but will be taken into account in a more ‘noisy’ way since market
players will estimate the ‘reaction function’ of the central bank under more
uncertainty. Noise in itself is however unlikely to be useful. Finally, constructive ambiguity could still imply delays in implementing measures, since
even if the central bank is internally well prepared and has pre-designed its
reaction function as far as possible, banks requesting FCM measures would
be likely to be much better prepared if they knew in advance the relevant
Generally, it could be argued that constructive ambiguity is the opposite
from what the regulators expect from banks, namely to have well-documented
risk taking policies in particular in crisis situations, and to do stress testing and
to develop associated procedures. Risk management becomes most important
in stress situations, and it is counter-intuitive to say that exactly for these
situations, no prior thinking should take place. Prior thinking does not mean
believing in the possibility to anticipate every detail of the next financial crisis,
but only by the belief that one will be in a much better position to react,
compared with the case of no preparation at all.
3.7 At what rate to provide special lending in a crisis situation?
This issue has been debated for a long time, as it is part of the Bagehot
legacy. Interestingly, both Thornton (1802) and Harman in 1832 were less



Full transparency in the middle of a crisis and associated rescue operations may also be harmful, and information on
banks accessed by the central bank may be confidential. Ex post, a high level of transparency appears desirable as a key
element of accountability of public authorities operation with public resources.
This is also the opinion expressed by the industry in Institute of International Finance (2007, 42): ‘there is a fear that
greater transparency on the part of central banks would lead to moral hazard. It is the Special Committee’s belief,
however, that the benefits of increased clarity on how central banks would respond to different types of crises
outweigh this risk. In times of crisis involving multiple jurisdictions and regulators, there will always be challenges in
the coordination of information collection, sharing, and decision making. To the extent possible, the more protocol
that is established prior to such an event, the better prepared both firms and supervisors will be to address a crisis.’


Central bank financial crisis management

explicit on this. Debates on it are linked to the incentive / moral hazard
issue discussed above. The key passage in Bagehot (1873, 197 – see also
Goodhart 1999) is:
The end is to stay the panic; and the advances should, if possible, stay the panic.
And for this purpose there are two rules: First. That these loans should only be
made at a very high rate of interest. This will operate as a heavy fine on unreasonable
timidity, and will prevent the greatest number of applications by persons who do
not require it. The rate should be raised early in the panic, so that the fine may be
paid early; that no one may borrow out of idle precaution without paying well for
it; that the Banking reserve may be protected as far as possible.

First, it is important to recall one more time that Bagehot referred to the
case of equal access FCM, not to what is mostly debated today, namely
individual bank ELA. Second, it may be noted that today, central banks offer
borrowing facilities, typically at þ100 basis points relative to the target rate,
i.e. at some moderate penalty level, but that even this penalty level is
apparently considered too high, since central banks e.g. in August 2007
injected equal access emergency liquidity through open market operations
almost at the target level, instead of letting banks bear a 100 basis points
penalty. So this would not at all have been in the sense of Bagehot (1873).
Without saying that it was necessary to shield banks in August 2007 from
paying a 100 basis point penalty for overnight credit, it is difficult also to
believe that a 100 basis point penalty would have been very relevant in terms
of providing incentives. For aggregate FCM measures, the topic simply does
not appear overly relevant, at least not in terms of providing or not the right
incentives for banks. A general liquidity crunch in the money market is
anyway a collective phenomenon, which may have been triggered only by
the irresponsible behaviour of a few participants, or even by completely
exogenous events. Therefore, collective punishment (anyway only by small
amounts) does not make too much sense.
The same seems to hold true for single access ELA: single access ELA
implies a lot of problems for a bank and its stakeholders, and this is how it
should be (as argued above). Also, in expected terms, ELA often means
subsidization of banks, since ELA tends to correlate with solvency problems.
The rate at which an ELA loan is made to a bank is in this context only a
relatively subordinated issue, which will not decide on future incentives. For
the sake of transparency of financial flows, it would probably make sense to
set the ELA rate either at a market rate for the respective maturity (in
particular if one is confident that there will be enough ‘punishment’ of the


Bindseil, U.

relevant stakeholders anyway), or at say þ100 basis points, which is a kind
of penalty level, but none which would compensate for the risks.9

4. Financial stability role of central bank operational framework
The setting-up of a central bank’s risk management framework for collateralized lending operations is discussed in-depth in Chapters 7 and 8. When
designing a framework and setting thereby the eventual amount of eligible
collateral and the implied borrowing potential of the banking system with
the central bank, the relevance of this for financial stability needs to be
recognized. Financial markets are inherently unstable as the prospect of a
systemic crisis may be self-fulfilling (as has been modelled in various ways
ever since Diamond and Dybvig 1983). Bank runs (both by non-banks and
by other banks) are inherently dynamic, and in principle the state of the
money market can switch from smooth and liquid to tense and dry up from
one moment to the next, as observed at least since Bagehot (1873) and as
experienced most recently in August 2007. Potential reasons (more substantial than sunspots) which could trigger a liquidity crisis are always out
there in the market, whereby the intensity of these potential triggers for a
crisis can be thought as a stochastic process across time. The central bank’s
normal credit operations framework is key in deciding what the critical level
of stress will be before an actual liquidity crisis breaks out. For instance, it is
plausible that the critical stress level will depend on the following five
(1) Availability of collateral for central bank credit operations. It will be
stabilizing that: (i) the set of collateral eligible for central bank credit
operations is wide; (ii) amounts of collateral available to the banks are
large, in comparison to average needs with regards to central bank
credit operations, and needs at high confidence levels; (iii) collateral
buffers are well-dispersed over the banking system; (iv) risk control
measures imposed by the central bank such as limits and haircuts are
not overly constraining (e.g. avoidance of limits). The collateral and
risk control framework may, or may not be differentiated across the
three different types of central bank credit operations (open market

The Hong Kong Monetary Authority (1999, 79) puts emphasis on the idea of a penalty rate: ‘The interest rate charged
on LOLR support would be at a rate which is sufficient to maintain incentives for good management but not at a level
which would defeat the purpose of the facility, i.e. to prevent illiquidity from precipitating insolvency.’


Central bank financial crisis management

operations, borrowing facility, intra-day). If they are somehow
differentiated, then also the impact on money market stability needs
to be differentiated across these three dimensions – even if ‘the more
the better’ will somehow hold across all of these dimensions.
(2) Existence of an end-of-day borrowing facility at a moderate penalty
rate and which does not stigmatize. Most leading central banks
impose a penalty level of 100 basis points on their borrowing facility.
This is not really an important cost factor for a bank if we talk about
covering exceptional short-term liquidity needs even of a considerable
size. For instance a EUR 1 billion loan overnight at a 100 basis point
penalty means penalty costs of EUR 28 thousand. If a bank takes EUR
10 billion over ten days, it would thus mean a cost of EUR 2.8 million.
For a medium- or large-sized banks, this is literally peanuts compared
with the potential damage caused by a run on the bank (or a cutting of
credit lines from other banks causing a liquidity squeeze). What will
really be relevant will be the availability of collateral and the certainty
perceived that the information on a large access to the borrowing facility
will be kept secret (unless the bank decides voluntarily on an outing).10
(3) A high number of financial institutions has direct access to the
central bank borrowing facility. Wide access to open market
operations may also be relevant in so far as it is considered that open
market operations have their own merits in terms of contributing to
financial stability. A wide range of counterparties is relevant since one
core characteristics of a money market crunch is that due to general
mistrust and scarcity, no one lends, not even at a high price. Therefore,
the central bank becomes the lender of last resort for all financial
institutions which lack liquidity, and it does not help a financial
institution to know that others could borrow from the central bank and
pass on the liquidity to itself, as exactly this will not happen.
(4) The intra-day payment system is well designed to avoid deadlocks –
for which also limits in intra-day overdrafting and possible collateral
requirements are important. There is an extensive literature on this
topic see e.g. CPSS (2000).

An important issue in this context is how close the borrowing facility is to an emergency facility. In the US, the
discount window had been understood before 2003 as being something in between a monetary policy instrument
and an automated emergency liquidity facility. In contrast, the Eurosystem’s facility had been designed from the
outset more as a monetary policy tool, as suggested by (i) the identity of the collateral set with the one for open
market operations; (ii) the absence of any quantitative limitation; (iii) the absence of any follow-up investigations by
the central bank.


Bindseil, U.

(5) Reserve requirements and averaging. At least initially, i.e. in the US
until the early 1930s, reserve requirements were considered as a means
to ensure financial stability (see e.g. Bindseil 2004 chapter 4 and the
literature quoted there). Although this is no longer seen to be an
essential motivation of reserve requirements, the existence of reserve
requirements must be relevant in some sense for financial stability.
It may be argued that obtaining central bank reserves to fulfil reserve
requirements drains available collateral, which is negative for financial
stability. On the other side, averaging allows to buffer away moderate
end-of-day liquidity shocks without recourse to a borrowing facility,
which may be positive if recourse to the borrowing facility is stigmatized.
The averaging functionality also allowed the Eurosystem to inject
massively excess reserves into the system in the second half of 2007, as it
knew it could absorb these excess amounts again in the second half of the
reserve maintenance period.
In sum, a first crucial contribution to financial stability that the central
bank can provide lies in its normal-times credit operations and collateral
framework. When designing this framework, the central bank should always
also have financial stability in mind, and not only risk mitigation of the
central bank in normal times. This will bias the framework in the directions
indicated above, and should be based on some cost–benefit analysis,
although estimating benefits is very difficult, since it is not observable how
many liquidity crises are avoided by building in one or other stabilityenhancing features into the framework.

5. The inertia principle of central bank risk management
in crisis situations
In the previous section, it was argued that the operational framework for
central bank credit operation was a first major contribution a central bank
can make and should make to financial stability. In particular, a wide range
of eligible collateral (to be made risk-equivalent through risk control
measures) is crucial in this respect. In the subsequent section, equal access
FCM measures will be discussed. In between, something very fundamental
needs to be introduced, which is called here the ‘inertia principle’ of central
bank risk management. The inertia principle says that the central bank’s risk
management should not react to a financial crisis in the same way as banks’
risk managers should, namely by restricting business such as to limit the