Risk management Derivatives Regulation Investing Quantum Cutting Edge Books SearchAll

risk.net
sections

Events Awards White papers Analytics Hub Futures & Options Hub Webinars Subscribe Follow us My account

Search risk.net

Acknowledging the SEARCH BOOKS

Search Risk Books

Elephant in the Room: The Type to search

Mismatch Centre
David Green

15. Acknowledging the Elephant in the Room: The Contents

Chapter first published in: Mismatch Centre
A Guide to
Behavioural This chapter will explain how interest rate risk (IRR) and liquidity risk
Modelling (LR), which are innate to every levered financial institution, create a
Edited by Matteo Formenti profitability management problem that can only be solved within a
and Umberto Crespi
comprehensive and well-functioning funds transfer pricing (FTP)
First published: framework. It will be demonstrated how the introduction of FTP
04 JUN 2019
produces a new business unit, the mismatch centre, which is a true
ISBN: 978-1-78272-404-9
profit centre that must be analysed and managed as rigorously as any
BUY NOW other lending or deposit-gathering business unit if granular earnings
Subscriber discount i
attributions are to have economic integrity. The chapter will discuss the
implications for risk and profitability management when the mismatch
centre is ignored or arbitrarily manipulated, eg, when its earnings are
forced to zero. Implications for risk governance, the development and
use of behavioural models for non-maturity deposits (NMDs), risk and
profitability management systems and regulatory considerations are
also addressed.

At most banks, credit unions and levered financial institutions (FIs) in

general, FTP is not an integral component of risk and profitability
management frameworks; risk and profitability management are
therefore treated as separate and distinct exercises. This necessarily
means that the impact of IRR and LR on business segment- and
product-level measures of profitability is ignored or assumed to be
inconsequential. This flawed understanding of risk necessarily leads
organisations to develop and use granular measures of profitability
that are inaccurate (usually upwardly biased) and misleading, and
which subsequently result in a mis-allocation of capital resources and
incentive schema that are suboptimal – ie, they are not profit-
maximising or risk-minimising. Such firms are also ill-prepared for the
impact of changes in market interest rates and the price of liquidity on
earnings and the economic value of equity.

When the managers of an institution do not use FTP, instead of

making balance-sheet management decisions that are informed by
economically robust1 measures of profitability, they will tend to assume
that the customer rates on their loans and deposits will necessarily
guarantee an economic profit and an acceptable return on capital, and
that profitability can only be maximised through above-average
balance-sheet growth.2 For evidence of this phenomenon, look for
incentives and compensation schema (especially at the executive
level) that are simply volume-based; ie, they make no mention of
margin or risk-adjusted return on capital (RAROC).

Even at institutions that do utilise FTP in their risk and profitability

frameworks, FTP rates and spreads are often computed and assigned
arbitrarily such that granular measures of profitability comport with pre-
conceived levels.3 In either case, no FTP or arbitrary FTP, there is a
mismatch centre that is almost certainly being ignored; for the
asset/liability committee (ALCO) and finance, this is the proverbial
elephant in the room.4

Whether out of ignorance or willful distortion of the truth, neglect of the

mismatch centre will necessarily preclude an economically robust
understanding of the level and volatility of business segment- and
product-level earnings. While a sound understanding of profitability is
important for all members of executive management, this is especially
true for the chief financial officer (CFO) and the head of finance as
they are ultimately responsible for crafting and communicating to
shareholders, employees and regulators the story of how the FI makes
money. Critical business processes, including capital management,
risk management, strategic balance-sheet management and
compensation management, will be mis-informed when granular
earnings attributions are incorrect. The need for FTP and an effectively
managed mismatch centre speaks directly to the requirements for
maximising shareholder value. If a firm cannot manage its capital in a
blue-chip fashion, it is manifestly a poorly run institution. Equity
analysts, shareholders and financial markets as well will increasingly
begin to penalise those firms that are shown to be “swimming naked”.

Inaccurate earnings attributions are problematic in any rate

environment, even stable ones, but they become especially acute
when market interest rates change. This occurred in the US where the
Fed funds rate was increased by 225bp from the end of 2015 after
having remained static for approximately seven years (see Figure
15.1).

The increase in interest rates impacted the earnings of every FI in the

market; relative to a flat rate scenario, liability-sensitive banks
underperformed while asset-sensitive banks outperformed.5 What
most banks failed to acknowledge was that their IRR exposure not
only explained the change in earnings that occurred after interest rates
increased, but it also explained a portion of the level of their earnings
before the rate change. Liability-sensitive banks were able to increase
earnings before rates rose by funding long-term assets with short-term
funding, trading this benefit for an earnings reduction that would occur
when rates would eventually rise. Similarly, asset-sensitive banks
accepted a reduction in earnings knowing that rising rates would
increase future earnings. When the mismatch centre is ignored or the
earnings that should accrue to it are mis-stated, a portion of overall
bank earnings are incorrectly attributed to lending and deposit
gathering.6 The mathematics behind this phenomenon is explained in
detail in this chapter.

In addition to inaccurate earnings attributions, if FTP is not used (or is

not used correctly) earnings attributions will not be robust to changes
in interest rates or the price of liquidity; granular measures of product
and business unit profitability will be volatile, but for reasons that have
nothing to do with the characteristics or performance of the loans and
deposits. They will be volatile because the margin attributed to loans
and deposits remain exposed to IRR and LR. When this is the case,
winners and losers, and the rewards and punishments bestowed upon
them, will be arbitrary.7 In addition, flaws in earnings attribution
processes are almost always biased to overstate measures of loan
and deposit profitability and understate the amount of IRR- and LR-
related earnings. This occurs because there is always someone who is
accountable for lending and deposit-gathering earnings, but often no
one to answer for IRR- and LR-related earnings. This bias can lead to
the flawed perception that bank-level earnings are not materially
exposed to IRR and LR, so weaknesses in FTP frameworks should be
of concern to regulators.8

The requirement for a robust earnings attribution process also has

implications for the design and management of behavioural models
that are used to quantify the re-pricing and liquidity cash flow dynamics
of many financial instruments. Behavioural models are essential to
effective risk and profitability management since many financial
instruments behave differently than their contractual terms; examples
include the early repayment of a fixed rate loan and the early
redemption of a fixed rate time deposit. Even more challenging than
contractual deviations are the need to specify the cashflow dynamics
of non-maturity instruments such as credit cards, lines of credit and
NMDs. FTP methodologies must acknowledge these behavioural
dynamics in the same way as risk models if risk and profitability
management are to be in alignment. Behavioural models should be
constructed to not only produce re-pricing and liquidity cashflows for
risk models, but they should also be designed to either calculate or
clearly inform the calculation of FTP rates associated with these
cashflows. My preference is for behavioural models that calculate FTP
rates directly;9 this eliminates potential discrepancies that are likely to
occur when FTP rates are calculated elsewhere. In this latter case,
FTP rate calculations may be based upon different or simplifying
behavioural assumptions, or they may even use alternative data
sources and computational systems; mis-alignment is a virtual
certainty. It should be the case that any changes in behaviour that alter
the expectation of future cashflow dynamics should induce a
simultaneous change in the level and behaviour of future FTP rates.
This topic is explored in more detail later in the chapter.

RISK AND PROFITABILITY MANAGEMENT

ARE THE SAME PROBLEM
After more than 20 years helping banks navigate interest rate cycles
and liquidity shocks, my most profound recognition is that risk and
profitability management are one and the same problem; efforts to
cleave the two into separate and distinct problems inevitably lead to
multiple and conflicting stories of how a bank makes money.
Conflicting stories only serve to confuse decision-makers, and this will
almost certainly lead to a suboptimal allocation of capital and other
resources.

Think of risk and profitability as two sides of a coin: if they show

different denominations, not only is the value of the coin unclear, but
the two sides cannot both be correct! Risk and profitability
management each tell a story of how the bank and its business units
make money. When measures of risk and profitability are not aligned,
the two stories they tell will contradict one another, but when they are
in sync there is a single, economically robust story of how the bank
makes money. The best way to avoid this contradiction is to design
and leverage a comprehensive FTP framework, and recognise that a
key component of the framework is the governance process around
the mismatch centre and the earnings that should accrue to it.

Mismatch centre earnings are just as real and just as important as

those resulting from the bank’s primary lending and deposit-gathering
activities. Like the earnings in these other business units, the level and
volatility of earnings in the mismatch centre tell a story, and this story
must make sense before bank earnings can be understood and
managed effectively. To understand the role of the mismatch centre,
we need to acknowledge the fundamental tendency of a bank.

THE FUNDAMENTAL TENDENCY OF A

BANK: MATURITY TRANSFORMATION
The typical bank takes in short-term deposits and funding and invests
the proceeds into long-term loans and securities. This process is
referred to as “maturity transformation” (or the carry trade). It is the
natural tendency of banks because this strategy almost always
increases near-term earnings while deferring the problem of future
earnings volatility. The upfront earnings boost occurs as yield curves
are generally upward-sloping. Caution is warranted, however, as this
balance sheet structure correlates to a bet that pays a known benefit
upfront but has an unknown cost that can only be known in the future.

Maturity transformation creates interest

rate and liquidity risk
When a bank engages in maturity transformation, it is said to be
“liability-sensitive”; if market interest rates or liquidity costs increase,
the expense of its short-term deposits and funding will rise faster than
the income coming from its long-term, fixed rate assets. This reduces
earnings relative to a flat rate or stable liquidity-cost scenario.

A liability-sensitive balance sheet is not mandatory, however, it is

possible to construct a balance sheet that has an “asset-sensitive” risk
profile. To do this, the bank must originate short-term, floating rate
assets and raise long-term, fixed rate funding.10 In this case, a
subsequent rate or liquidity cost increase will result in an increase in
earnings. When it comes to risk positioning, banks tend to be liability-
sensitive for the simple reason that it pays upfront.11 While public
companies tend to emphasise near-term earnings growth, private
companies are not immune to the rush of a near-term earnings boost.

Regardless of the risk profile, the magnitude (and possibly even the
direction) of the exposure to changes in interest rates and liquidity
costs is inherently unstable. The make-up of the balance sheet and
customer behaviours are constantly changing, or management might,
from time to time, deliberately manipulate the risk profile. For this
reason, IRR and LR, and the earnings related to these risks, must be
constantly monitored if the understanding of risk and profitability is to
be correct.

We now compare the earnings risk profile of a bank with no IRR (bank
earnings, πBank, do not vary across rate scenarios) with two banks that
have IRR: the first is liability-sensitive, and the second is asset-
sensitive; these are shown in Figure 15.2. Next we analyse business
unit earnings dynamics for each of these banks with and without the
benefit of FTP.12

Most banks analyse and manage IRR and LR by focusing solely on

the volatility of aggregate bank earnings, with little or no regard for how
IRR and LR affect business segment- or product-level earnings. The
allocation of earnings to the individual business units is left to finance
and financial reporting, where the role of risk in the earnings-
generation process is generally ignored. It is worth noting that there is
nothing in US Generally Accepted Accounting Principles (GAAP) or
International Financial Reporting Standards (IFRS) that compels a
financial institution to quantify and report how much IRR and LR, or
even credit risk, contribute to earnings, either at a bank level or a
business segment level.

INTEREST RATE AND LIQUIDITY RISK

CREATE AN EARNINGS MANAGEMENT
PROBLEM
Once we acknowledge that IRR and LR exist at a consolidated level, it
is not difficult to see that these risks create earnings volatility for the
individual business units within the bank. To demonstrate this, consider
a bank with only two business units, a lending unit and a deposit-
gathering unit. The earnings of the bank can be allocated as shown in
Equation 15.1:

πBank = πLoans + πDeposits, (15.1)

where πLoans and πDeposits represent the earnings in the lending and
deposit-gathering business units respectively.

The analysis of IRR involves an investigation of how bank earnings

change in response to a change in market interest rates:

∂πBank
=?, (15.2)
∂r

where r represents market interest rates. Similarly, the analysis of LR

requires an investigation of how bank earnings change in response to
a change in the cost of liquidity:

∂πBank
=?, (15.3)
∂l

where l represents liquidity spreads for the bank. Within the context of
FTP, liquidity spreads represent the credit risk of the FI as measured
by the difference between the cost of issuing senior unsecured debt
and an interest rate swap rate with a corresponding term.

If the earnings of the bank change in response to a change in market

∂πBank
interest rates, ie, ∂r
≠ 0, then the earnings attributed to the two
business units must change by the same amount for Equation 15.1 to
hold:

∂πBank ∂πLoans ∂πDeposits

= + . (15.4)
∂r ∂r ∂r
13

Similarly, the earnings change associated with a change in liquidity

∂πLoans
spreads, must be reflected in the business units as follows:
∂l

∂πBank ∂πLoans ∂πDeposits

= + . (15.5)
∂l ∂l ∂l

Equations 15.4 and 15.5 make very important points: if a bank has IRR
or LR, then one or more of the business units must have IRR or LR.
Put differently, IRR and LR are not just bank-level problems, they are
business unit-level problems as well. These risks make granular
earnings management particularly problematic as individual business
units within the bank do not generally have a mandate to take or
manage IRR or LR,14 much less the ability to even quantify these
risks. The existence of these risks also means that the analysis and
reconciliation of business unit earnings in any budgeting and
forecasting exercises will be complicated by differences between
forecasted and realised interest rates and liquidity spreads. In volatile
economic environments, these rate and liquidity spread differences
can be material; regardless of their magnitude, such rate and liquidity
spread driven differences must be accurately quantified, else the
attribution of earnings and earnings variances will necessarily be
incorrect.15 However, finance and the business units need not throw
their hands up in frustration, as this is a problem can be solved; in fact,
this is exactly the problem FTP is intended to solve.

THE PURPOSE OF FTP

The purpose of FTP is to acknowledge all of the IRR and LR in the
bank and aggregate these risks into a single location where they can
be effectively managed. Through a well-functioning FTP process, the
IRR and LR in the business units is transferred away from them (this is
what the word “transfer” refers to in FTP). In effect, business units are
immunised from IRR- and LR-related earnings volatility. Of course, risk
transfer is not done for free. In the same way that a property and
casualty insurance company will accept the risk of your home in Miami
being destroyed by a hurricane (which creates an obligation to
compensate you for the loss should one occur), they expect to be paid
a fee (the insurance premium). The fee corresponds to the targeted
loss-adjusted return on the capital they choose to hold against the
hurricane risk, or, alternatively, what they will have to pay a re-
insurance company to take the hurricane risk from the company. In
either case, the insurance premium is not arbitrarily determined; it
follows logically from the cost of hedging the hurricane risk. Regarding
IRR and LR, the FTP rate for a particular transaction represents the
cost (value) of hedging the IRR and LR created by the transaction (in
the case of an asset, the FTP rate is the cost to hedge the risk, while
for a liability, the FTP rate is the value of the hedging power of the
instrument). For such calculations, a bank must look to external market
information to determine these transfer prices. The process for doing
this will be described later in the chapter.

A well-functioning FTP process should therefore achieve the following

for each business unit (BUi) in the bank:

∂πBUi
= 0, (15.6)
∂r

FTP accomplishes this by match-funding each transaction in the

business unit’s portfolio: fixed rate assets (liabilities) receive a fixed
rate charge (credit) and floating rate assets (liabilities) receive a
floating rate charge (credit). Each transaction, therefore, has a fixed
spread over its life. If the FTP rate for every transaction has been
calculated and applied properly, then all product and business
segment roll-ups will be similarly immunised from IRR and LR. We will
come back to the calculation of FTP rates in a later section, but at this
point just know that the calculation of an FTP rate follows directly from
the contractual or behaviourally-adjusted re-pricing and liquidity
cashflows of a transaction.

ALONG WITH FTP COMES A MISMATCH

CENTRE
The introduction of FTP produces a new business unit, which we refer
to here as the mismatch centre.16 When FTP functions correctly, the
lending and deposit-gathering business units are immunised against
IRR and LR, but not because the risks to which they are exposed are
simply made to vanish; rather, these risks are transferred to the
mismatch centre (think of the mismatch centre as an in-house
insurance company). Just as importantly, FTP simultaneously transfers
risk-related earnings (think of these earnings as the insurance
premium) between the business units and the mismatch centre. The
introduction of FTP, therefore, changes the profit allocation introduced
in Equation 15.1 as follows:

πBank = πLoans + πDeposits + πMM , (15.7)

where πMM represents the earnings in the mismatch centre.

The introduction of FTP will not add to or subtract from the overall level
of earnings of the bank, but it will almost certainly change the
allocation of earnings to the lending and deposit-gathering business
units.17 Relative to the allocation where there was no mismatch centre
(Equation 15.1), for a given level of πBank, if πMM is positive then either
or both of πLoans and πDeposits will necessarily be lower – ie, there will
be a net transfer of earnings out of the lending and deposit-gathering
business units. Similarly, if πMM is negative then either or both of
πLoans and πDeposits will necessarily be higher – ie, there will be a net
transfer of earnings into the lending and deposit-gathering business
units.

Once we understand that the FTP charges (credits) have meaning, we

are compelled to acknowledge that the mismatch centre must be a
true profit centre. Despite the widely held view, when positive earnings
accrue to the mismatch centre this does not mean that FTP rates were
an arbitrary tax on lending and deposit-gathering earnings. Before we
explain why positive mismatch centre earnings may make perfect
sense, it is helpful to revisit the discussion of IRR and LR.

When FTP is introduced (and the mismatch centre is acknowledged)

the decompositions of IRR and LR are expanded as follows:

∂πBank ∂πLoans ∂πDeposits ∂πMM

= + + , and (15.8)
∂r ∂r ∂r ∂r

∂πBank ∂π ∂πDeposits ∂πMM

= Loans + + . (15.9)
∂l ∂l ∂l ∂l

As we saw previously, the IRR and LR in the bank must be borne by

one or more of the business units, but because we have introduced
FTP, we are able to immunise the lending and deposit-gathering
business units from IRR and LR by transferring these risks to the
mismatch centre as follows:

The arrows indicate the direction of risk transfer. Eliminating the terms
that have a zero value, we get the following:

∂πBank ∂π
= MM , and (15.12)
∂r ∂r

∂πBank ∂π
= MM . (15.13)
∂l ∂l

These simple expressions clarify the primary purpose of FTP:

Through a well-functioning FTP process, all the IRR

and LR in the bank is captured within the mismatch
centre.

If FTP is being used in any way such that Equations 15.12 and 15.13
do not hold, the lending and deposit-gathering business units will not
be immunised from IRR and LR.

Let’s return to the bank earnings risk profiles that were introduced
earlier and consider the earnings risk profiles of the business units and
the mismatch centre across rate scenarios.

Business unit and mismatch centre

earnings: a bank with no IRR
In the remainder of this section, we only consider the business units in
the aggregate (lending plus deposit gathering) to highlight the role of
the mismatch centre. (Even if the consolidated balance sheet has no
IRR, this does not mean that individual products or business units
have no IRR, so there is still a need for FTP.) Figure 15.3
demonstrates that where there is no IRR in the bank, not only are bank
earnings invariant across rate scenarios, but aggregate business unit
earnings are invariant as well. Bank and aggregate business unit
earnings must also be one and the same, ie, πBank = πBU; as there is
no net transfer of risk to the mismatch centre, there should be no net
earnings transferred to the mismatch centre (in any scenario),
therefore πMM = 0. This shows that it is possible that the earnings in
the mismatch centre can be zero, but this generally occurs only in the
special case where there is no IRR in the bank. Given that most banks
are either asset-sensitive or liability-sensitive, it is usually the case that
some non-zero earnings will accrue to the mismatch centre.

Now let’s see what the allocation of earnings looks like in banks that
have non-zero IRR. This will begin to help us answer a very important
question: “What is the correct level of earnings in the mismatch
centre?”.

Business unit and mismatch centre

earnings: a liability-sensitive bank
If the bank introduces the carry trade by borrowing short and lending
long, it will increase the level of bank earnings (relative to an IRR-free
bank), but there will be a reduction in bank earnings if interest rates
increase (all else constant). In Figure 15.4, this is depicted by the
upward shift in the level of bank earnings in the BaseCase scenario
(by the height of the upper dotted bracket) and the change in the
earnings profile from a zero slope, πBank(IRR-free), to a negative slope,
πBank(liab. sens.).

Note the two dotted brackets in Figure 15.4. The increase in bank
earnings in the BaseCase scenario (upper dotted bracket) resulting
from the shift to a liability-sensitive position is the same as the
increase in mismatch centre earnings (lower dotted bracket): the
increase in bank earnings associated with the change in the IRR
profile of the bank accrues entirely to the mismatch centre. This must
be the case if aggregate business unit earnings are to be unchanged
relative to the IRR-free bank. Business unit immunisation will also only
hold across all rate scenarios if the bank earnings risk profile and the
mismatch centre earnings risk profile have the same slope – ie, the
two earnings risk profiles are parallel. This is exactly what Equation
15.12 says and what is shown in Figure 15.4; all the brackets between
the two earnings profiles (which represent πBU) are of equal height.

There is also nothing that logically constrains mismatch centre

earnings to being positive. Note the dashed bracket in Figure 15.4 that
indicates that mismatch centre earnings are negative in the Up500
scenario. When bank earnings have decreased by a significant
amount because of an increase in market interest rates, to keep the
same amount of earnings accruing to the lending and deposit-
gathering business units as in the BaseCase scenario, mismatch
centre earnings must be negative.

Business unit and mismatch centre

earnings: an asset-sensitive bank
If the bank introduces the opposite of the carry trade by borrowing long
and lending short, it will decrease the absolute level of bank earnings
(relative to a risk-free bank), but there will be an increase in bank
earnings if interest rates increase (all else constant). In Figure 15.5,
this is depicted by the downward shift in the level of bank earnings in
the BaseCase scenario (by the height of the upper dotted bracket)and
the change in the earnings profile from a zero slope πBank(risk-free) to a
positive slope πBank(asset sens.).
Note the two dotted brackets in Figure 15.5. The decrease in bank
earnings in the BaseCase scenario (upper dotted bracket) resulting
from the shift to an asset-sensitive position is the same as the
decrease in mismatch centre earnings (lower dotted bracket): the
decrease in bank earnings associated with the change in the IRR
profile accrues entirely to the mismatch centre. This must be the case
if aggregate business unit earnings are to be unchanged from the IRR-
free bank. Business unit immunisation will only hold across all rate
scenarios if the bank earnings risk profile and the mismatch centre
earnings risk profile have the same slope – ie, the two earnings
profiles must be parallel. This is also exactly what Equation 15.12 says
and what is shown in Figure 15.5; all the brackets between the two
earnings profiles (which represent πBU) are of equal height.

There is also nothing that logically constrains mismatch centre

earnings to being negative or zero. Note the dashed bracket in Figure
15.5 that indicates that mismatch centre earnings are very large in the
Up500 scenario. When bank earnings have increased by a significant
amount because of an increase in market interest rates, to keep the
same amount of earnings in the lending and deposit-gathering
business units as in the BaseCase scenario, mismatch centre earnings
must absorb the entire increase in earnings.

These prior figures illustrate business unit and mismatch centre

earnings dynamics under well-functioning FTP processes, but let’s see
what happens when FTP does not function correctly – ie, when it does
not immunise the lending and deposit-gathering business units from
IRR and LR.

Business unit and mismatch centre

earnings: when FTP is not well-
functioning
Figure 15.6 illustrates the decomposition of bank earnings for a
liability-sensitive bank when FTP only removes a portion of the liability-
sensitivity from the business units. When the reduction in bank
earnings resulting from an increase in interest rates is not absorbed
entirely by the mismatch centre, some of the earnings decline must be
borne by the business units. This is illustrated by the shrinking
brackets (from left to right) between the bank earnings risk profile and
the mismatch centre earnings risk profile.

No doubt, if you are a business unit head, this is not an optimal (or fair)
allocation of earnings. ALCO, the committee responsible for IRR
management, decided to bet against rising rates, but the business
units bore a portion of the earnings reduction when interest rates rose.
Conversely, although the business units are not likely to complain
about it, if rates decline and bank earnings increase, the business
units will be credited with some portion of the increase.

Figure 15.7 illustrates an extreme case of poorly-functioning FTP

where, for a liability-sensitive bank, FTP serves to transfer an
increasing amount of earnings from the business units to the mismatch
centre as market interest rates increase. As a result, business unit
earnings are not immunised from IRR; in fact, the exposure to IRR is
exacerbated by FTP – the level of business unit earnings is highly
dependent on market interest rates. In such a construct, the heads of
the business units have virtually no control over their earnings; further,
any margin-based profitability measures (including RAROC, at a
business unit or product level) will have little, if anything, to do with the
actual performance of the underlying loans and deposits.18

In a rising rate environment, if no one suspects that poorly functioning

FTP is the primary cause of declining business unit earnings, any
number of other causal factors will be mis-identified. It is worth
considering how executives, lenders and other key decision-makers
might attempt to get their earnings back on budget. It is quite possible
that their actions will be good for them, at least in the short run, but
detrimental to the bank in the long run – eg, credit standards could be
loosened, or non-transfer priced options could be given away.
Regardless, some unfortunate managers are likely to be blamed for
earnings shortfalls, and many people are not going to get bonuses
who are otherwise deserving. Meanwhile, ALCO is unjustly rewarded;
it has structured the balance sheet for a bet against rising rates, but
when rates go up mismatch centre earnings actually increase!

The problem of asynchronous earnings

risk profiles
You might be wondering whether it is possible that FTP could be so
poorly constructed that the mismatch centre earnings risk profile is
completely different from the business unit earnings risk profile. It is
possible, because I have worked at a bank where this was the case. It
wasn’t obvious that we had this problem (although the confusion
around the reason for the decline in earnings was painfully obvious)
until we took the time to model the mismatch centre. When we did, we
got a picture that looked something like Figure 15.7. It was suddenly
clear why the business units were so frustrated with FTP: FTP was
exacerbating earnings volatility for them, not mitigating it. If the
mismatch earnings risk profile is not parallel to the bank’s earnings risk
profile, then the business units must have IRR or LR.19

Most banks that employ FTP will deny the possibility of such a
discrepancy, but they cannot actually say to what extent the earnings
risk profiles of the bank and the mismatch centre are aligned because
most banks don’t bother to model the mismatch centre. While this
omission indicates a failure to model risk and profitability at a
sufficiently granular level, it ultimately reflects a governance problem
which will be addressed in more detail in the next section.

Potential problems with risk transfer could also be revealed if the

earnings risk profiles of the individual business units were to be
modelled, but most banks do not do this either. To the extent that this
showed residual IRR and LR in the business units, then it would be
safe to conclude that all the IRR and LR in the bank has not been
transferred to the mismatch centre. One must be careful, however; if
an analysis of the earnings risk profiles of the business units reveals
that they have been effectively immunised against IRR and LR, it is not
safe to conclude that all the IRR and LR in the bank has been isolated
in the mismatch centre. It is possible that FTP methodologies have
been constructed in a way that IRR and LR have not been properly
transferred to the mismatch centre. This can occur, for instance, when
equity held by the business unit against a loan (eg, credit risk capital)
is brought into the calculation of FTP rates. If equity is (arbitrarily)
assigned a duration that matches that of the asset being hedged, this
will create a hedge for the business unit, but it will fail to transfer all the
IRR in the instrument to the mismatch centre as some of the IRR gets
lost in transit.20

For this reason, in addition to the earnings risk profile of the mismatch
centre, the earnings risk profiles of each of the business units should
also be modelled and analysed. Although FTP may be used to
estimate margin contributions at a very granular level, without an
analysis of the earnings risk profiles of the individual business units
and the mismatch centre, there is no proof that FTP has immunised
each of the business units against IRR and LR, and that all IRR and
LR in the bank has been transferred to the mismatch centre. FTP
cannot be assumed to be functioning properly.

OWNERSHIP OF THE MISMATCH CENTRE

AND GOVERNANCE CONSIDERATIONS
John Abrams said “If the people who make the decisions are the
people who will also bear the consequences of those decisions,
perhaps better decisions will result”. The primary reason behind the
failure to analyse the earnings risk profile of the mismatch centre is
that ownership of the mismatch centre is not clear,21 and no one is
held accountable for the level and volatility of IRR- and LR-related
earnings. This should not occur at any bank, especially as every bank
has an ALCO policy that states ALCO is responsible for managing IRR
and LR. If ALCO doesn’t own the earnings related to these risks, then
who does? Just as it does with the owner of any other business unit of
the bank, finance should hold ALCO accountable for the level and
volatility of earnings in its business unit. This accountability becomes
especially important in volatile rate and liquidity environments as any
unhedged exposures will result in earnings volatility for the bank.
Despite this, a review of ALCO packages at most banks will reveal that
there is no analysis of the earnings in the mismatch centre. It is the
proverbial elephant in the room.

The problem of accountability exists for several reasons. First, it is

often the case that no one knows what the earnings in the mismatch
centre represent or how to determine if the level of earnings is correct.
We saw in the prior figures that illustrated what well-functioning FTP
looks like that the mismatch centre earnings risk profile must be
parallel to the earnings risk profile of the bank if the business units are
to be immunised from IRR and LR. When this occurs, mismatch centre
earnings could be positive, zero or negative, small or large in
magnitude. In the next section, we will see what drives the level of
these earnings.

Another challenge to accountability relates to responsibility for IRR-

and LR-related earnings being vested in a committee of conflicted
interests. The typical ALCO is chaired by the CFO and has as
members the chief executive officer (CEO), the head of finance, the
chief risk officer (CRO), the treasurer, the asset and liability
management (ALM) manager, the chief investment officer (CIO) and
the head of the funding desks. In addition, the executive head of retail,
commercial, mortgage, wealth management, etc, are also members.

When it comes to FTP and defence of the mismatch centre, the latter
members have an innate conflict of interest: they are members of
ALCO and, as such, should be accountable for the earnings and risk in
the mismatch centre, but they also represent the interests of their
respective business units. When it comes to decisions about the price
of transferring risks from their business units to the mismatch centre,
they surely want the benefits of risk immunisation, but they may not
want to pay full price for it.

The conflict of interest does not require that the business unit heads
be excluded from ALCO. Quite the contrary, their input and knowledge
of product behaviours, not to mention their ability to influence product
behaviours, is essential knowledge for ALCO to be effective at risk
identification, risk measurement and risk management. To address the
conflict of interest, banks should consider establishing a sub-
committee of ALCO that has the authority to develop and manage FTP
methodologies and oversee the calculation of mismatch centre
earnings. In addition to managing conflicts of interest, it is up to
finance to hold ALCO accountable for these risk-related earnings,
especially considering that banks are required to hold capital against
IRR and LR. If ALCO is unable to generate consistent and stable
earnings on this capital, it is worth considering limiting the amount of
risk ALCO can take. Without a reasonable quantification of IRR- and
LR-related earnings, this issue cannot be addressed in any meaningful
way. Everyone just assumes that ALCO is a good steward of capital.

Finally, in the US, outside of the systemically important financial

institutions (SIFIs, approximately the top 10 banks), banks are not
explicitly required by their regulators to create a mismatch centre and
analyse its earnings. Even for the SIFIs, a requirement to create and
manage a mismatch centre has only been in place since 2016,22
although this guidance is largely principals-based, leaving banks a lot
of leeway to determine how FTP is done. No wonder most bank
executives argue that if it’s not so important to the regulators, FTP
must not be necessary to run a bank.

THE MEANING OF MISMATCH CENTRE

EARNINGS
We have seen that FTP involves the transfer of earnings between the
business units and the mismatch centre. Whether the bank’s finance
department holds ALCO accountable for these earnings or the heads
of the lending and deposit-gathering business units demand (as they
should) an explanation for these earnings, ALCO should have a logical
and coherent answer the following question: “What is the correct level
of earnings in the mismatch centre?”

Answering such a question ends up being a major challenge for most

organisations, since, as Figure 15.8 shows, most people think about
the mismatch only in terms of the level of earnings it contains. Should
it be a high positive amount? Should it be a low positive amount?
Should it be zero? Should it be negative? The answer is important,
because it will often determine how FTP methodologies are designed,
calibrated and managed.

If mismatch centre earnings are positive, especially if they are

material, the business units will argue that they have been unfairly
taxed, although they may not recognise (deliberately perhaps) that
they may have originated a significant amount of IRR and LR that was
transferred to the mismatch centre. (A quote from Upton Sinclair
comes to mind: “It is difficult to get a man to understand something,
when his salary depends upon his not understanding it”.) It is not
uncommon to see a bank go through all the motions of FTP, arrive at a
positive level of mismatch centre earnings at the end of the year and
then transfer these earnings back to the lending and deposit-gathering
business units (this would be like the insurance company refunding
your premiums at the end of the year in the event there were no
storms, and it also begs the question of what happens when
mismatch-centre earnings are negative – does this loss get distributed
back to the business units as well?). Alternatively, when mismatch
centre earnings are negative, this may indicate that the lending and
deposit-gathering business units have been able to influence the
design of the FTP methodologies in their favour. Whether before or
after the fact, if a bank is not careful, mismatch earnings will be
skewed to the benefit of the business units.23 As depicted in Figure
15.9, the battle for mismatch-centre earnings rages on; well-run banks
will not let this be one-sided!

Therefore, how does a bank avoid having mismatch centre earnings

be skewed or arbitrarily determined? It must recognise that a narrow
focus on the level ignores the more important question around the risk
that should be in the mismatch centre, and the risk is represented not
by the level of earnings but by the slope of the earnings risk profile. To
determine if the risk profile is correct, it is necessary to analyse how
mismatch centre earnings will change when interest rates or the price
of liquidity change.

Recall Figures 15.3–15.7, all of which showed the earnings in the

mismatch centre across multiple rate scenarios. For a given risk profile
(other than risk-neutral), mismatch centre earnings could be extremely
positive, extremely negative or any level in between; what matters for
effective immunisation of IRR and LR at a business unit level is that
the earnings risk profile of the mismatch centre must be parallel to the
earnings risk profile of the bank. This is a binding constraint, and the
level of mismatch centre earnings will be subsequently determined by
the impact of changes in rates and the price of liquidity. The level of
earnings in the mismatch centre, therefore, cannot be meaningfully
considered without an understanding of the risk profile of the mismatch
centre. The following have to be true.

If ALCO structures the balance sheet to be free of IRR and LR, the
earnings in the mismatch centre should be zero.

If ALCO structures the balance sheet to be liability-sensitive, and if

interest rates or the price of liquidity increase, the earnings reduction
experienced by the bank must be borne by the mismatch centre.

If ALCO structures the balance sheet to be liability-sensitive, and if

interest rates or the price of liquidity stay the same or decrease, the
earnings increase experienced by the bank must accrue to the
mismatch centre.

If ALCO structures the balance sheet to be asset-sensitive, and if

interest rates or the price of liquidity stay the same or decrease the
earnings reduction experienced by the bank must be borne by the
mismatch centre.

If ALCO structures the balance sheet to be asset-sensitive, and if

interest rates or the price of liquidity increase, the earnings increase
experienced by the bank must accrue to the mismatch centre.

These bullet points can be summarised as follows: Mismatch centre

earnings represent the return to the bank from taking IRR and LR.

IRR and LR are distinct risks that are quantified and managed
separately, and FTP rates, as we will see in the next section, have
separate components which correspond to duration and liquidity.
When these components are properly maintained, the earnings in the
mismatch centre can be disaggregated into the return to taking interest
rate risk πIRR and the return to taking liquidity risk πLR:

πMM = πIRR + πLR . (15.14)

More importantly, when we consider the earnings of a bank from a

risk-attribution perspective, we get the following equation:

πBank = πCR + πIRR + πLR + πDF , (15.15)

where πCR equals the earnings derived from credit risk, πIRR equals
the earnings derived from IRR, πLR equals the earnings derived from
LR, and πFV equals the earnings derived from the deposit franchise
(which is immunised, in large part, by access to deposit insurance).
This is essentially the same as Equation 15.7 that was described
earlier, where lending was substituted for credit risk and deposit-
gathering was substituted for the deposit franchise. Equation 15.15
should make it clear that if the returns to taking IRR and LR are not
computed correctly then either the return to taking credit risk or the
return to the deposit franchise, or both, will not be correctly stated. We
contend that the ability to solve this equation correctly is fundamental
to being able to claim that an FI is well-managed.

FTP CURVES AND FTP METHODOLOGIES

To this point, there has been an extensive discussion of the mismatch
centre that has yet to include details around the construction of FTP
curves or the calculation of FTP rates. This prioritisation was
intentional, because if the purpose of the mismatch centre is not
clearly understood (by bank management in particular) the choice of
FTP curve is not likely to be correct and FTP methodologies are not
likely to be properly designed and calibrated.

Over the years, I have known many FTP managers, often highly-
skilled “quants”, who either selected the wrong funding curve (or did
not argue against the selection of an inappropriate curve) or who
developed complex FTP methodologies which failed to acknowledge
the re-pricing and liquidity characteristics of the products to which they
were applied. Worse yet, they never bothered to consider the earnings
risk profiles of their mismatch centres. One look at the level of
mismatch centre earnings told me that there was a critical error
somewhere in their FTP framework.24 Absent consideration of the role
of the funding curve and how all the pieces come together dynamically,
it is easy to end up with a framework that is not effective at immunising
the business units against IRR and LR or at transferring these risks to
the mismatch centre. Remember the expression: “Be careful that you
can’t see the forest (or wood) for the trees”.

A fulsome discussion on FTP curve selection, FTP methodologies and

FTP rate calculations is beyond the scope of this chapter, but there are
a few important points to consider as they relate to our understanding
of the mismatch centre. Before product-specific FTP methodologies
are defined, the bank must first identify the FTP curve; the FTP (or all-
in funding) curve should reflect the cost of senior unsecured debt. For
large banks with such debt outstanding, this curve is easy to develop;
market rates on outstanding debt with various terms to maturity are
observable on major trading systems. For smaller institutions with no
outstanding debt (or for credit unions that are unable to issue debt),
this is a slightly more difficult task; in such cases, the FTP curve must
be imputed. This requires the firm to approximate its debt rating and
then develop a funding curve using the yields on debt issued by similar
firms with that same rating. For very small firms, this likely means that
the FTP curve will reflect something close to a BBB rating, perhaps
worse depending on its risk profile and the market environment in
which it finds itself.

The specific mix of funding an FI has in place or expects to have in the

future – some combination of NMDs, time deposits, Federal Home
Loan Bank (FHLB) advances, senior debt, etc – is irrelevant to
determining the FTP curve. The point of the FTP curve is to quantify
how much it would cost the bank to raise unsecured liquidity with a
specified duration; the all-in funding curve is therefore a benchmark
against which all other sources of funding are compared. If the FTP
curve is established using instruments other than senior unsecured
debt – for instance, by using the average cost of all outstanding
sources of non-equity funding – then the bank will be forced to contend
with the nonsensical result that the FTP spread on senior unsecured
debt is non-zero. Alternatively, consider the extreme case where a
bank funds itself entirely with zero-rate transaction accounts; such a
blended curve approach would have a zero rate at all term points. This
would result in an FTP spread of zero on the transaction accounts
(and very large spreads on all the loans). The deposit gatherers would
have been successful at funding the entire bank (excluding equity) at a
zero rate, yet they would end up with no margin attribution; this will be
well-received for sure! Despite the obvious logical flaws associated
with using a blend of actual funding costs to construct the funding
curve, this approach is not uncommon, and it almost always results in
arbitrary adjustments to deposit FTP rates to increase their spreads.
Of course, this is just compounding bad logic with more bad logic.

It may help to recognise that when a bank issues senior unsecured

debt, no positive or negative economic value has been created; think
of it as a neutral transaction. Conversely, when a bank raises funding –
eg, in the form of unsecured savings or money market deposits – at a
lower yield than the senior unsecured debt, economic value is created.
The rate on senior unsecured debt is used to determine how much
value is being created relative to what the market would charge the
bank for funding (given that it would have to accept the credit risk that
is inherent in that funding). In contrast, when a bank raises secured or
collateralised funding – eg, FHLB advances in the case of a US bank –
it does not raise any marginal liquidity. Cash does come onto the
balance sheet, but collateral must be pledged, and for FHLB advances
(and many types of municipal deposits) the funding must actually be
over-collateralised; putting such funds on the balance sheet, in fact,
reduces overall bank liquidity. The rates on such collateralised funding
are therefore red herrings when estimating the cost of liquidity.

After an appropriate funding curve is selected, the bank can then

determine the cost of liquidity at each term point, but it must employ a
basis-adjusted interest rate swap curve to back into the cost of
liquidity, because the all-in funding rate includes the value of duration
(assuming we are looking at fixed rate debt). Liquidity premiums (or
spreads) are computed by subtracting the swap rates from the all-in
funding rates (see Figure 15.10). These spreads have their own term
structure which is referred to as the liquidity premium (spread) curve.

Note the use of the term “basis-adjusted” to describe the swap curve.
Swap rates are quoted with reference to a floating rate of some
specified term.25 The FI must decide on the basis for the swap curve it
will use in the calculation of FTP rates. My preference is to swap
everything back to the overnight rate as this brings all IRR into the
mismatch centre at a zero-duration equivalent. When adjustable-rate
instruments (here defined to mean that they reset less frequently than
overnight) are then transfer priced using such a curve, a basis
adjustment is required.

Once the all-in funding curve and swap curve have been defined, a
historical database of daily all-in funding rates and interest rate swap
rates must be developed; the FTP rate calculations for all existing
business will reference this database. The simulation engine that is
used in risk and profitability modelling work should also be able to
evolve all-in funding rates and swap rates into the future across
multiple scenarios to enable computation of the earnings risk profile of
the mismatch centre, and to compute FTP rates and spreads on new
business; the latter functionality is necessary to carry out budgeting
and forecasting exercises in a manner that is consistent with risk
quantification.

Once the curves are set, we can then consider FTP methodologies.
For FTP to immunise a transaction from IRR and LR, the methodology
must acknowledge all the re-pricing and liquidity cashflows of the
transaction. Approximations, such as the weighted average life (WAL)
or estimated duration, will not be accurate in many circumstances and
should be avoided as a general practice.

The general formula for calculating FTP rates, referred to as matched-

maturity FTP (MMFTP), is as follows:

∑(Repr cashflowτ ×Spot swap rateτ ×τ)

MMFTP ratet = ∑(Repr cashflowτ ×τ)
(15.15)
∑t=0(Cashflowτ ×Spot LPτ ×τ)
+ ∑t=0(Cashflowτ ×τ)
,

where Repr cashflowτ is the repricing cashflow, Spot swap rateτ is the
corresponding point on the basis-adjusted swap curve, τ is the term
point, Cashflowτ is the liquidity cashflow, and Spot LPτ is the
corresponding point on the liquidity premium curve (recall that this is
the difference between the all-in funding rate and the basis-adjusted
interest rate swap rate).

The first term reflects the cost to hedge (or hedging value of) the
duration of the instrument and the second term reflects the cost to
hedge (or hedging value of) the liquidity of the instrument. For fixed
rate instruments, these two terms are calculated at the origination date
of the transaction, and they do not change over the life of the
instrument. For floating and adjustable rate instruments, the first term
is recalculated at each rate reset date, while the second term is
calculated only at the time of origination and is held constant until
maturity.

There are a few special cases to consider. For fixed rate instruments,
because the re-pricing and liquidity cashflows are one and the same,
the two terms can be collapsed, and the all-in FTP rates can be
applied to the liquidity cashflow schedule. Where instruments have
embedded optionality, such as a loan prepayment option, an interest
rate cap or floor, an additional option charge can be added or
subtracted as appropriate.26 If an instrument should not receive a
liquidity charge, eg, a bond that is a high-quality liquid asset (HQLA),
 or an instrument has no liquidity value, eg, a deposit that must be fully
collateralised, then the liquidity premium should be zero.

It is also worth considering whether certain assets such mortgages or

other securitisable loans, eg, credit card receivables, should face a full
liquidity charge. The answer depends on how robust FTP is intended
to be. In business-as-usual contexts, some assets may be very liquid,
but in times of economic stress, they may not be. If such assets are
not charged a liquidity premium, the presumption is that they will
always be liquid. If at some later date they are determined to be
illiquid, the mismatch centre will have borne that risk, but it will not
have been compensated for the risk by the originator of the
transaction. I generally recommend that if an instrument qualifies as an
HQLA, it should not face a liquidity charge, but all other instruments
should.

Contingent liquidity is another important consideration, but a fulsome

discussion of this topic is beyond the scope of the chapter. The
challenges here are myriad: the level/type of contingent stress event
must be clearly identified (liquidity coverage ratio (LCR)equivalent or
some other specified idiosyncratic or systemic scenario), the
corresponding contingent use of liquidity must be quantified (for
example, the increase in the usage of guaranteed lines of credit), and
a mechanism for computing and passing a charge to the business
units must be devised (this is operationally challenging as the charge
does not correspond to an outstanding balance).

At banks with large capital markets functions, trading portfolios may

have historically been deemed to be “self-funding”. The flaw in this
perspective is revealed when market liquidity dries up and the trading
desk must rely on the bank to fund its positions. The global financial
crisis showed that when this risk was not recognised, traders were
incentivised to take excessive risks because they stood to receive all
the upside and the bank was stuck with the downside. A more
accurate measure of profitability requires that the cost of contingent
liquidity be factored into their funding costs. Oftentimes, the only
effective way to hedge contingent liquidity risk is to hold cash or
HQLAs. Of course, this is very expensive and such an
acknowledgment may severely challenge the efficacy of many trading
strategies.

IMPLICATIONS FOR BEHAVIOURAL

MODELLING
To effectively immunise transactions from IRR and LR, FTP
methodologies must acknowledge the expected re-pricing and liquidity
cashflows of a transaction, but when these cashflows do not follow
contractually defined behaviours, this complicates the calculation of
FTP rates. To estimate non-contractual cashflows, organisations must
develop behavioural models. In practice, these models do not typically
include the corresponding calculation of FTP rates. When
organisations turn to third parties for behavioural models or option-
adjusted cashflows, they may not be privy to the actual cashflow
engine and are left to estimate corresponding FTP rates after the fact.
Even when behavioural models are developed internally, they are
generally silent on FTP. In either case, it is almost certain that the
associated FTP rates will not follow logically from the re-pricing and
liquidity cashflow estimates derived from the behavioural model.

This creates a discrepancy between risk and profitability measures;

the resulting margin attributions are not consistent with the measures
of IRR and LR (think non-parallel earnings risk profiles). As this
problem is almost certainly impossible to resolve after the behavioural
model is developed, it is essential that models be developed with the
calculation of FTP rates in mind. The preferred approach to developing
a behavioural model is to have the model produce not only the
prepayment or curtailment speeds that will be used by the ALM model
(to adjust the contractual re-pricing and liquidity cashflows), but to also
have the model produce the estimated re-pricing and liquidity
cashflows for all scenarios in which the speeds will be used. In
addition, the behavioural model should produce the FTP rates for the
instruments to which the model applies;27 the FTP rates should use
the same base funding curve and liquidity spreads used to calculate
FTP rates for other transactions on the balance sheet. This approach
to model development compels the model developer to solve the
problems of risk and profitability simultaneously and consistently. More
importantly, because the behavioural model will be used to calculate
the FTP rate directly, the business units facing the FTP rates will have
an interest in ensuring that the model properly reflects the behavioural
characteristics of their products; not only will FTP rates be more
accurate, but so will measures of IRR and LR.

When developing a behavioural model that can calculate FTP rates, it

is also valuable if the model can produce FTP rates that reach back in
time (preferably to a period that encompassed a material change in
interest rates and liquidity spreads) and into the future for a broad
range of hypothetical rate and liquidity scenarios. Of course, this will
depend on the availability of historical transactional data, but assuming
it is available, these calculations can, and should, be done. This allows
the model developer to demonstrate how a particular FTP
methodology is effective at immunising transactions and portfolios
against IRR and LR.

Let’s see what this looks like in practice using the output of a
behavioural model for NMDs that I developed to align the treatment of
deposit behaviours in risk and profitability management exercises. The
model combines vintage-level decay functions, rate beta functions and
processes for dynamically bifurcating balances into stable and non-
stable portions to produce a time series of re-pricing and liquidity
cashflows. For each behaviourally-distinct NMD product on the
balance sheet, the model produces cashflow schedules that can feed
any ALM model, and it produces FTP rates that can be loaded into
product pricing and budgeting/forecasting systems. Without having to
describe the cashflow logic within the model, or even the formulas
used to calculate the FTP rates, Figures 15.11 to 15.14 demonstrate a
logical evolution of FTP spreads for two different products over a span
of time where interest rate swaps and liquidity spreads changed
significantly and are anticipated to continue changing. One product is a
low rate, small balance savings account and the other is a high rate,
large balance money market deposit account (MMDA).28

The model was used to generate FTP rates using the estimated re-
pricing and liquidity cashflows for monthly vintages of deposits that go
back to early 2006 and reach to the end of 2018. The deposit
behaviours were analysed at the end of 2017, and were forecasted
through 2018 to populate the budget for the bank as well as one-year
IRR and LR simulations. Within the behavioural model, each vintage of
deposits was assigned an FTP rate that reflected swap rates and
liquidity premiums in effect at the time of its origination.29

Figure 15.11 shows the evolution of the FTP rate and FTP spread for
the savings product. One-month Libor (the reference rate to which the
beta function is indexed), the customer rate (coupon) and the one-
month liquidity premium are also shown for reference. Note how the
FTP rate increases in late 2008 even as the Federal Reserve was
reducing the Fed funds rate. The increase occurred because the
liquidity value of the balances rose dramatically when funding spreads
widened by several hundred basis points. While the value of the
deposit’s duration decreased, its liquidity value increased substantially.
The one-month liquidity premium (LP) is shown as a proxy for the
market price of the liquidity value of the deposit cashflows. In the
calculation of the FTP rate shown, each individual cashflow receives
an LP corresponding to its term to maturity; the first liquidity cashflow
receives the one-month LP, the second cashflow receives the two-
month LP, and so on, with the final cashflow, at month 120, receiving
the 10-year LP.30

When the liquidity crisis subsided in late 2009 and LPs contracted,
while marginal FTP rates decreased quickly, the FTP rate on the
product (which is a balance weighted average of the FTP rates on
each vintage) did not adjust down instantaneously.31 Each vintage has
a long life, and the LP for a vintage is set at the time of origination, so
the quantified value (hedging power) of balances originated in 2008
and 2009 persists until the decay functions controlling these vintages
terminate. The decline in the FTP rate on the product therefore
occurred over many years, as these very valuable vintages amortised
down and were slowly replaced with newer vintages that had both
lower duration values and lower liquidity values.

The corresponding decline in the FTP spread occurred because the

customer rate was always very low (only 35bp at the cyclical peak of
rates in 2006–07), and it floored out just a few basis points above zero
near the end of 2012.32 Margin compression was bound to occur the
longer market rates remained at historical lows. What this FTP spread
analysis shows is consistent with the logical explanation that when
wholesale funding rates (reflected in the all-in funding rates faced by
the bank) reached historical lows, (near-) zero-rate deposits became
relatively less valuable. As deposit margins are material to the
earnings story, the bank’s margin declined as well. Relief for US banks
in general would only come when the Federal Reserve began to raise
interest rates in late 2015; for banks in Europe, the long-running ECB
experiment with negative interest rates continues to deliver little relief.

FTP rates and spreads on the savings product did not increase until
after the Federal Reserve began to increase the Fed funds rate, and
the yield curves re-steepened, and even then the increase has been
very slow (for the same reason that the downward adjustment in FTP
rates was slow). Note also that the FTP rate and spread extends on
through the 2018 budget horizon, a period over which the bank had
anticipated three more increases in the federal funds rate. Despite the
change in direction, the FTP spread on the product is anticipated to
remain relatively stable. The critical assumption is that its behaviour
will remain consistent with that observed over the calibration period. If
the product behaves as anticipated, the deposit-gathering business
units will be largely immunised from IRR- and LR-related volatility, and
the hedging value of the deposit will be transferred to the mismatch
centre.33 This analysis demonstrates a well-functioning FTP
methodology for this product. There are myriad other ways to produce
FTP rates for NMDs, but if the FTP rates behave materially different
than what has been shown here, it’s not clear that they have any
economic rationale and will not be effective at immunising deposit
gatherers against IRR and LR.

In contrast to the dynamic described above, Figure 15.12 shows the

evolution of FTP rates and spreads for the same savings product
where its liquidity value is ignored – ie, LPs are set at zero. During the
liquidity crisis in 2008–09, such an FTP rate would have decreased, in
effect sending a message to the deposit-gathering business units that
customer rates needed to be reduced if FTP spread compression was
to be mitigated. This would not have been a smart thing to do, yet FTP
methodologies in place at most banks at the time did exactly this
because they ascribed no value to liquidity. For most banks, low-rate
savings accounts are not a growth product, so next we will look at the
behaviour of a high rate, large balance money market account.

For many years prior to the financial crisis, LPs for highly-rated banks
were close to zero, even negative for some. Unfortunately, many of
these institutions assumed this would always be the case, and they did
not design internal transfer pricing processes that could handle non-
zero LPs. When the liquidity crisis hit, lenders kept lending as if
liquidity was free and deposit gatherers priced deposits as if they had
no liquidity value relative to wholesale funding. Seeing how banks
ignored this important component of economic value, it is clear why
the liquidity crisis was so severe.34 Despite the rapid recognition that
failures of FTP exacerbated the severity of the crisis, it was almost a
decade before US bank regulators introduced guidance mandating
that FTP methodologies should explicitly acknowledge the cost/value
of liquidity.35 Even so, the regulatory requirements are only applicable
to the SIFIs, as if smaller institutions should not bother to consider that
the value of liquidity is no longer zero and can change rapidly. In fact,
smaller institutions face higher absolute liquidity costs and are
exposed to great volatility in their funding spreads, and should
therefore be even more compelled to consider that the value of
liquidity is properly acknowledged in FTP rates for loans and deposits.
What about the following observation in the guidance is not applicable
to all levered FIs?

Failure to consistently and effectively apply FTP can

misalign the risk-taking incentives of individual
business lines with the firm’s risk appetite, resulting in
a misallocation of financial resources. This
misallocation can arise in new business and ongoing
portfolio composition where the business metrics do
not reflect risks taken, thereby undermining the
business model. Examples include entering into
excessive off-balance sheet commitments and on-
balance sheet asset growth because of mispriced
funding and contingent liquidity risks.

Figures 15.13 and 15.14 provide a similar analysis of FTP rates and
spreads for the MMDA product. Here, we see considerably more FTP
spread volatility, but this volatility does not indicate a failure of the
behavioural model. Rather, it is an indication that the behaviour of the
product was not stable or consistent with static behavioural
assumptions, and this is okay. When the Federal Reserve began to
lower the Fed funds rates in September 2007, the liquidity crisis was
still a year away. With a modelled rate beta of one (which feels risk-
averse, except, of course, when we consider behaviours in declining
rate environments), the bank would have been expected to decrease
the customer rate on the MMDA in line with the decrease in market
interest rates, but this is not what happened. As this product
represented a material source of funding, the bank was compelled to
pay above-market rates to ensure that it did not lose these deposits.
Initially, the economic value of the deposits declined, and therefore the
lag in the customer rate pushed the FTP spread below zero. It
returned to a positive value once the value of liquidity increased
dramatically (when the bank’s wholesale funding spreads blew out),
but the subsequent volatility in the value of liquidity continued to be
transmitted through the FTP rate, resulting in a volatile FTP spread.
Since this product is also modelled with a relatively high non-stable
haircut (~45%), changes in the month-to-month value of overnight
duration and liquidity flow instantly through the FTP rates. Again, this
is not a failure of the model to immunise against any spread volatility;
this product simply has a large portion of its balance that is no more
(or less) valuable than overnight funding. During the market turbulence
of 2008–09, this value was changing quickly.

Before turning to an analysis of the FTP rate and FTP spread where
the value of liquidity is ignored, it’s worth noting what happened with
the customer rate on this product during and after the crisis. When the
Federal Reserve began to lower the Fed funds rate in September
2007, the markets had yet to signal concerns about the price and
availability of liquidity. Even so, this bank understood the critical value
of these high balance deposits and deliberately chose to acknowledge
that value by keeping the customer rate high. By early to mid-2008, the
market began to express widespread concern about the value and
availability of liquidity, and it was at this point that the FTP rate
increased quickly. This is a great lesson of how a pricing model, even
one that is explicitly designed to acknowledge the contemporaneous
value of duration and liquidity, may not be sufficiently forward-looking
to anticipate dramatic shifts in economic value. It should be clear that
efforts to smooth spread volatility – for instance, through the use of
moving averages – will only serve to delay the recognition of such
market shifts; for institutions that were not otherwise self-aware, these
delays proved fatal.

Figure 15.14 shows the evolution of the FTP rate and FTP spread
where the liquidity spreads are set to zero. Given the relatively high
customer rate that persisted in the near-zero federal funds
environment, the FTP spreads would have remained negative for
almost a decade! Clearly, there is a compelling need to acknowledge
the value of liquidity when computing the hedging value of any deposit,
even ones with substantial balance volatility and rate sensitivity;
overnight liquidity may not seem like much, but at times it can be the
difference between survival and failure. It should also be clear that
when FTP rates are not economically rational, they get determined
arbitrarily and require constant adjustment to make sense in whatever
post-event narrative the company is trying to sell.

Earlier, we touched on some of the governance considerations around

the earnings that accrue to the mismatch centre. The modelling of
NMDs is closely related to this problem, as NMDs often represent the
largest source of funding for a bank and therefore the largest source of
transfers between the business units and the mismatch centre. To the
extent that NMDs are not modelled correctly, whether from a risk or
profitability perspective, the more likely it is that mismatch centre
earnings will not be at the correct level and will not exhibit the correct
sensitivity to changes in rates or the cost of liquidity.

Further, we recognise that when a behavioural model of NMDs is

developed solely to support the analysis of IRR and LR in the
consolidated balance sheet, outside of the ALM manager and possibly
a few members of ALCO, not many other people in the bank concern
themselves with it. In contrast, when the behavioural model for NMDs
is used to support the calculation of FTP rates, many more people are
concerned with its construct and the behavioural assumptions
embedded within it. To the extent that the model is compelled to
acknowledge the best understanding of NMD behaviours, if it is also
used to support the analysis of IRR and LR then it is more likely that
the understanding of IRR and LR will be accurate, or at least be
informed by the best understanding of deposit behaviours available.

COMPUTATIONAL SYSTEMS
A comprehensive analysis of earnings and risk in a bank requires that
one is able to determine the slope of the earnings risk profile of the
bank and the mismatch centre, and be able to compare them. To do
this, one needs the ability to simulate bank and mismatch centre
earnings across a multiplicity of rate scenarios. ALM systems by their
nature are forward-looking, and have the capability to generate
earnings across multiple future business scenarios.36 With this
capability, they can produce an estimate of the earnings sensitivity of
the bank.

Although FTP should acknowledge the IRR and LR characteristics of

every transaction on the balance sheet, most stand-alone FTP
systems do not have sufficiently robust simulation capabilities to
compute the earnings sensitivity of the mismatch centre. FTP systems
are often only able to compute and assign an FTP rate to existing
balance-sheet positions; there is no ability to evolve transfer prices
and spreads into the future, or to compute FTP rates for new business
or analyse the dynamics of existing or proposed FTP processes
across multiple interest rate and liquidity spread scenarios. FTP
managers are therefore left with very little in the way of pro forma
analytical capabilities; as a result, the risk profile of the mismatch
centre is never analysed. To better understand what this means, think
of an ALM manager being tasked with measuring IRR and LR, yet
having only the ability to quantify earnings in the current period. This is
laughable, but this is the position in which many FTP managers find
themselves.

The optimal solution is a system that has the ability to compute IRR
and LR in the bank and the mismatch centre at the same time. If this
sounds like a super-capable ALM system, your instincts would be
correct. An ALM system that can simultaneously compute FTP rates
on existing and new business can be used to analyse the earnings risk
profile of the bank and the mismatch centre. To tackle the comparison
using two separate systems, one for the bank and one for the
mismatch centre, is to accept the unnecessary task of ensuring that
transactional and market data, cashflow rules, behavioural functions
and output are all fully synchronised. This is a tall order. It should be
evident that consideration of the mismatch and how it must be
modelled is an essential consideration in the selection of the
computational system.

CONCLUSION
We have covered an extraordinary amount of information about the
mismatch centre. If you made it this far, congratulations! In a nutshell,
this is what you should have learned:

banks and credit unions (and levered FIs in general) play the game of
maturity transformation;

maturity transformation creates IRR and LR;

IRR and LR create earnings and earnings volatility;

the purpose of FTP is to immunise the lending and deposit-gathering

business units from IRR- and LR-related earnings volatility;

when FTP is introduced, a mismatch centre is created;

the IRR and LR profile of the mismatch centre should be parallel to that
of the bank;

a comprehensive and effective governance framework is critical to

ensuring that the mismatch centre contains all IRR, LR and IRR- and
LR-related earnings;

the earnings in the mismatch centre represent the returns to taking and
managing IRR and LR;

the choice of funding curve is important to properly calculating the

cost/value of liquidity;

behavioural models should be designed to compute FTP rates directly;

the choice of computational system for FTP is critical to ensuring the

ability to fully analyse mismatch centre earnings dynamics;

a comprehensive focus on the earnings and earnings risk profile of the

mismatch centre is required to properly align the treatment of risk and
profitability of the firm; alignment does not happen by accident;

challenges to effective FTP management are abundant; and

the financial industry and FI regulators still have a long way to go in

terms of appreciating the critical importance of a comprehensive and
well-functioning FTP framework, a central component of which is the
mismatch centre.

For the mismatch centre to serve its purpose, FTP methodologies

must be explicitly designed and applied correctly at a transaction level
to immunise product and business unit earnings against IRR and LR.
The process of immunisation transfers the risks and the returns
associated with these risks from the lending and deposit-gathering
business units to the mismatch centre. Through time, unhedged IRR
and LR positions result in gains and losses to the bank, and FTP
ensures that the correct level of earnings accrue to the mismatch
centre. Done correctly, FTP provides a solid foundation for
performance and profitability management; done poorly, every
decision which relies upon well-calculated and meaningful margin
attributions will be based on erroneous quantifications of profitability.

FTP methodologies are frequently manipulated to support

preconceived notions of product profitability;37 when the FTP
methodologies have not been specifically designed to transfer (at the
correct price) all the IRR and LR in loan and deposit products to the
mismatch centre, these risks and the earnings and earnings volatility
associated with them remain outside of the mismatch centre. This
means several things: (i) product and business unit profitability is mis-
stated; (ii) the lending and deposit-gathering business units are
exposed to risks over which they have no control; and (iii) the
computed return to taking IRR and LR is incorrect (and almost
certainly understated).

To speak intelligently about a bank’s margin, whether its level over a

certain time frame or how it has changed from one period to the next, it
is necessary to acknowledge the impact of IRR and LR. Ignoring the
contribution of these risks will result in a misattribution of earnings, and
the allocation of capital and other resources will be suboptimal,
rewards and punishments for under- and over-performance will be
arbitrary, and the perception of interest rate and liquidity risk will be
incorrect. In the aggregate, these errors will lead to ineffective
management of the bank.

prev chapter next chapter

Contents

Introduction

1. Insights on Banks’ Recourse to Behavioural Models from a Focused

IRRBB Stress Test

2. Implementing Regulatory Guidance on IRRBB Behavioural Models:

Challenges and Opportunities

3. The Stakeholders of Interest Rate Risk Behavioural Models

4. Governance of Behavioural Models

5. The Nature of IRRBB and Typical Metrics Employed

6. A Framework for Developing NMD Behavioural Models

7. The Literature on NMD Behavioural Models

8. Interest Rate Risk of Non-maturity Bank Accounts: From Marketing to

Hedging Strategy

9. NMDs and IRRBB: A Methodological Proposal for a Behavioural

Model

10. NMD Modelling: A Financial Wealth Allocation Approach

11. A Benchmark Framework for NMDs: An Application

12. NMD Behavioural Models Used in Marketing

13. The Validation of NMD Behavioural Models

14. The Choice of Maturity Profile in NMD Behavioural Models

15. Acknowledging the Elephant in the Room: The Mismatch Centre

16. Prepayment Risk Modelling for ALM, Finance and FTP: A Survival
Model

17. Modelling of Prepayment on Fixed Rate Residential Mortgages: A

Logistic Regression Approach

18. A Simple Approach to Modelling Prepayment Events

19. Integrating Credit Risk within the ALM Framework

20. Modelling Committed Credit Lines

21. Accounting of the Sight Deposit and Hedging

RECOMMENDED BOOKS

Risk Model Operational Risk Behavioural Risk Navigating Internal Models Systematic
Validation (3rd Capital Models Management European Energy and Solvency II Trading in
edition) (2nd edition) and Commodity Energy Markets
Markets
Regulation

About us Contact us Advertising Help Centre Terms and conditions Privacy policy

California Privacy Rights – Do not sell my information

© Infopro Digital Risk (IP) Limited (2020). All rights reserved. Published by Infopro Digital Services Limited, 133
Houndsditch, London, EC3A 7BX. Companies are registered in England and Wales with company registration numbers
09232733 & 04699701.

Best Digital B2B Publishing Company