See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/5143103

A Risk Assessment Model for Banks

Article in Annals of Finance · February 2005

DOI: 10.1007/s10436-004-0006-3 · Source: RePEc

CITATIONS READS

106 6,387

3 authors, including:

Dimitrios P. Tsomocos
University of Oxford
126 PUBLICATIONS 2,092 CITATIONS

SEE PROFILE

All content following this page was uploaded by Dimitrios P. Tsomocos on 21 May 2014.

The user has requested enhancement of the downloaded file.

 A Risk Assessment Model for Banks
Charles A.E. Goodhart Pojanart Sunirand
Bank of England, London School of Economics, Bank of England and
and Financial Markets Group London School of Economics
Dimitrios P. Tsomocos
Bank of England, Said Business School
and St. Edmund Hall, University of Oxford, and
Financial Markets Group

First version: 8 March 2004

This version: 28 June 2004

Abstract
The objective of this paper is to propose a model to assess risk for banks. Its main
innovation is to incorporate endogenous interaction between banks, recognising that the
actual risk to which an individual bank is exposed also depends on its interaction with
other banks and other private sector agents. To this end, we develop a two-period gen-
eral equilibrium model with three active heterogeneous banks, incomplete markets, and
endogenous default. The setting of three heterogeneous banks allows us to study not only
interaction between any two individual banks, but also their interaction with the rest of
the banks in the banking system. We show that the model is analytically tractable and
can be calibrated against real UK banking data and therefore can be implemented as a
risk assessment tool for financial regulators and central banks. We address the impact of
monetary and regulatory policy as well as credit and capital shocks in the real and financial
sectors.
JEL Classification: C68; E4; E5; G11; G21
Keywords: Financial Fragility; Financial Contagion; Systemic Risk; Banks; Monetary
Policy; Regulatory Policy; Equilibrium Analysis

1 Introduction
The objective of this paper is to propose a model to assess risk for banks. Existing models
for this purpose, e.g. stress-testing models, focus almost entirely on individual institutions.1
The major flaw of these models is that they fail to recognise that the actual risk to which
an individual bank is exposed also depends on its interaction with other banks and other
private sector agents and therefore is endogenous. This endogeneity aspect of risk may matter
enormously in times of financial crises; a negative shock which is ex ante specific to a particular
bank can produce a serious strain on others through a series of interactive contagion effects. The
main innovation of our approach, therefore, is to take into account such endogenous interaction
between banks, particularly through their mutual exposures in the interbank market. To this
end, we require a set-up which can incorporate heterogeneous banks, each with a unique
1
An exception is, for example, an approach developed in Elsinger, Lehar and Summer (2003).

1
risk/return portfolio; if they were identical, there would be no interbank market by definition.
We also need default to exist, since if there were no default, there would be no crises. Moreover,
financial markets cannot be complete. Otherwise banks can always hedge themselves against
all kinds of shocks, in which case, there would, again, be no crises. We have constructed a
two-period general equilibrium model along these lines in Goodhart, Sunirand, and Tsomocos
(2003).
We show in Goodhart et al. (2003) that an equilibrium exists in such model and that
financial fragility emerges naturally as an equilibrium phenomenon. However, given the scale of
the model which contains B heterogeneous banks, H private sector agents, S possible states, a
variety of financial assets and default, it is impossible to find either a closed-form or a numerical
solution to this general model. In Goodhart, Sunirand, and Tsomocos (2004), we therefore
present a simplified version of our general model and show that it can be solved numerically.
This implies that the model, in some sense, ‘works’ and can, in principle, be used to assess
various policies for crisis management. However, the model was solved on the basis of an
arbitrarily chosen set of initial conditions. The outcome, therefore, is a somewhat artificial
construct of our own assumed inputs. In this paper, we take a step further by attempting to
calibrate an alternative version of our general model against real UK banking data. Thus, we
argue that our model is not only rich enough to incorporate endogenous interaction between
banks, but it is also sufficiently flexible to be implemented as a risk assessment tool for financial
regulators and central banks.
The model presented in Goodhart et al. (2004) has two heterogeneous banks. A banking
system, however, generally comprises multiple banks. Thus, although the model can be used to
study interaction between any two individual banks, its level of complexity is not sufficient to
incorporate contagion effects arising from their interaction with the rest of the banking sector.
In this paper, we therefore introduce an additional bank, which can be thought of as the
aggregation of the remaining banks in the system. Given the lack of disaggregated household
and investors’ portfolio data, we model household behaviour via reduced-form equations which
relate their actions to a variety of economic variables such as GDP, interest rates, and aggregate
credit supply etc. In this sense, our model is a partially-microfounded general equilibrium
model.2 However, the main aspects of equilibrium analysis which are market clearing, rational
expectation, and agent optimisation are maintained. Moreover, contagion effects between the
banking sector and the real economy still operate actively in equilibrium via the reduced-form
equations. Thus, we adhere to the general equilibrium spirit of Goodhart et al. (2003 and
2004). The upshot of our modelling framework is that financial decisions generate real effects
in the rest of the macroeconomy. Our model is therefore amenable to welfare analysis.
The rest of the paper is structured as follows. The next section presents the model. Section
3 then explains how the model is calibrated against UK banking data. Section 4 provides a
stress-testing analysis for the UK banking sector. Section 5 presents a robustness check of the
result obtained in section 4. The final section provides concluding remarks.

2 The Model
The model has three heterogeneous banks, b ∈ B = {γ, δ, τ }, four private sector agents,
h ∈ H = {α, β, θ, φ}, a Central Bank and a regulator. The time horizon extends over two
periods, t ∈ T = {1, 2} and two possible states in the second period, s ∈ S = {i, ii}. We
2
As is shown in Goodhart et al. (2004), household and investor optimisation problem can be easily introduced.
However, owing to the limited availability of disaggregated household data, we chose not to follow that route.

2
 1. Borrow and deposit in the interbank markets (B)
2. OMOs (CB)
3. Borrow and deposit in the commercial bank loan
t=1 and deposit markets (B)

Nature decides which of the s∈S occurs

1. Settlement of loans and deposits (H and B)

t=2 2. Settlement of interbank loans and deposits (CB and B)
3. Default and capital requirements’ violation settlement

All banks are wound up.

CB= Central Bank

B = Commercial Banks

Figure 1: The time structure of the model

assume that state i is a normal/good state whereas state ii represents an extreme/crisis event.
The probability that state i will occur is denoted by p.
As in Goodhart et al. (2003 and 2004), we assume that individual bank borrowers are
assigned during the two periods, by history or by informational constraint, to borrow from a
single bank (i.e. a limited participation assumption).3 Given this assumption, together with
our set-up of a system of three heterogeneous banks, we need at least three borrowers. We
therefore assume that agents α, β, and θ borrow from banks γ, δ, and τ , respectively. The
remaining agent, Mr. φ, represents the pool of depositors in this economy which supplies
funds to every bank. This implies that we have multiple active markets for deposits (by
separate bank) and for loans (by borrower and bank). In addition, we also assume a single,
undifferentiated, interbank market where deficit banks are allowed to borrow from surplus
banks, and wherein the Central Bank conducts open market operations (OMOs).
The time structure of the model is presented in figure 1. At t = 1, loan, deposit and
interbank markets open. Banks decide how much to lend/borrow in each market, expecting
rationally any one of the two possible future scenarios to be realised. Moreover, the Central
Bank also conducts OMOs in the interbank market. At t = 2, depending on the state which
actually occurs, all financial contracts are settled, subject to any defaults and/or capital re-
quirements’ violations, which are then penalised. At the end of the second period, all banks
are wound up.
3
In Bhattacharya, Goodhart, Sunirand and Tsomocos (2003) we show that restricted participation in the
loan market can also arise as an equilibrium outcome given that the objective functions of banks also include a
relative performance criterion, i.e. a preference to outperform their competitors.

3
2.1 Banking Sector: UK banks
Without loss of generality, our specification of the banking sector is based on the UK bank-
ing sector, which we assume to comprise of seven largest UK banks; Lloyds, HSBC, Abbey
National, HBOS, Barclays, Royal Bank of Scotland, and Standard Chartered. Banks γ and δ
can represent any two of these individual banks, whereas bank τ represents the aggregation
of the remaining banks. As will be explained more below, in our calibration exercise, banks
γ and δ are chosen specifically to represent two of these actual UK banks. However, for data
confidentiality reason, we do not reveal their identities.
All banks in the model, b ∈ B = {γ, δ, τ }, are assumed to operate under a perfectly
competitive environment (i.e. they take all interest rates as exogenously given when making
their optimal portfolio decisions). The structure of their balance sheets is given below;

Assets Liabilities
Loans to agents Deposits from Mr.φ
Interbank deposits Interbank borrowing
Market book Equity
Others

We assume that all banks endogenise their decisions in the loan, deposit and interbank
markets.4 The remaining variables are treated as exogenous.5 We further assume that banks
in our model can default on their financial obligations, subject to default penalties set by
the regulator. Thus, by varying the penalties imposed on default from 0 to infinity, we can
model 100% default, no default or an equilibrium level of default between 0 and 100%.6 At
first sight, this ‘continuous’ default rate approach may seem problematic since in reality banks
either repay in full at the due date or are forced to close down. However, we interpret a bank’s
default rate in our model as a probability that such bank chooses to shut down, and hence in
the short run to default completely on its financial obligations. For example, a default rate of
4 percent implies that there is roughly a 4 percent chance of a shut down and a 96 percent
chance that the bank will repay in full and continue its normal operation. Therefore, a bank’s
decision to increase its default rates is isomorphic to its decision to adopt a riskier position in
pursuit of higher expected profitability.7 Finally, as in Bhattacharya et al. (2003), we make
a simplifying assumption by assuming that banks’ default rates in the deposit and interbank
markets are the same, i.e. banks are restricted to repay all their creditors similarly.
Analogous to the modelling of default, banks can violate their capital adequacy requirement,
subject to capital requirement violation penalties set by the regulator. In principle, each
bank’s effective capital to asset ratios may not be binding, (i.e. their values may be above
the regulator’s requirement), in which case they are not subject to any capital requirement
4
The modelling of the banking sector follows Shubik and Tsomocos (1992) and Tsomocos (2003a and b).
5
Given the present set-up, we cannot endogenise banks’ decisions on market book or equity. This is because
the model has two states in the second period and one unconstrained asset (i.e. an asset that banks can either
go infinitely short or long), which is the interbank market investment. By adding another unconstrained asset,
markets would be complete. In principle, our model can be extended to incorporate additional states in the
second period and therefore can be used to study the economic effects on the market (trading) book. For
example, we can disaggregate the market book into two components according to their riskiness (or rating) and
endogenise banks’ decisions on these variables. This would allow us to study the endogenous response of risk
premia on corporate debt to a series of shocks. However, we face a practical problem on this front since there
are insufficient data on the composition of the market book by category, e.g. rating maturity, and currency.
6
This modelling of default follows Shubik and Wilson (1977).
7
For more on this issue, see work in progress by Tsomocos and Zicchino (2004).

4
penalty. However, in our calibration exercise, we assume for simplicity that each bank wants
to keep a buffer above the required minimum, so that there is a non-pecuniary loss of comfort
and reputation as capital declines; in this sense the ratios are always binding. Put differently,
we assume that banks’ self-imposed ideal capital holdings are always above the actual values of
all banks’ capital to asset ratios. Given this assumption, we can rule out corner equilibria and
therefore focus our analysis entirely on well-defined interior solutions whereby banks violate
their enhanced capital requirements. We assume that penalties are linear as capital declines
from its ideal level. In practice, there will be some non-linearity as capital falls below its
required minimum, but this is just too complex to model at this stage.
As will be elaborated in section 3, our calibration exercise is based on the data of UK banks
at the end of 2002. At that point in time, bank δ is a net lender whereas banks γ and τ are net
borrowers in the interbank market.8 Given this setting, we describe the optimisation problems
of these banks below.

2.1.1 The interbank net borrowers’ (banks γ and τ ) optimisation problems

Bank b ∈ {γ, τ } maximises its payoff, which is a quadratic function of its expected profitability
in the second period minus non-pecuniary penalties that it has to incur if it defaults on its
deposit and interbank obligations. It also suffers a capital violation penalty proportional to
its capital requirement violation. Formally, the optimisation problem of bank b ∈ {γ, τ } is as
follows;

µ b ¶2 " b
#
X π bs πs X λbks max[0, k − ksb ]+
b b
max Π = ps [ 10 − cs ] − ps λbs λbs
mb ,µb ,µbd ,vsb ,s∈S 10 1010 [ µb − vsb µb ] + [ µbd − vsb µbd ]
s∈S s∈S 1010 1010

subject to

µb µbd
mb + Ab = + + eb0 + Othersb (1)
(1 + ρ) (1 + rdb )
b
π bs = vsb
h
(1 + rb )mb + (1 + rA )Ab − (vsb µb + vsb µbd + Othersb + eb0 ), s ∈ S (2)
ebs = eb0 + π bs , s∈S (3)
ebs
ksb = hb (1 + r b )mb + ω
, s∈S (4)
ωvsb e (1 + rA )Ab

where,

ps ≡ probability that state s ∈ S will occur,

cbs ≡ coefficient of risk aversion in the utility function of bank b ∈ B,
λbks ≡ capital requirements’ violation penalties imposed on bank b ∈ B in state s ∈ S ,
b
k ≡ capital adequacy requirement for bank b ∈ B,
λbs ≡ default penalties on bank b ∈ B,
µb ≡ amount of money that bank b ∈ {γ, τ } owes in the interbank market,
µbd ≡ amount of money that bank b ∈ B owes in the deposit market,
8
As noted earlier, we have chosen banks γ and δ to represent specifically two of the seven largest UK banks
in our calibration exercise. Bank τ then represents the aggregation of the remaining five banks.

5
 vsb ≡ repayment rates of bank b ∈ B to all its creditors in state s ∈ S,
mb ≡ amount of credit that bank b ∈ B extends in the loan market,
Ab ≡ the value of market book held by bank b ∈ B,
ebs ≡ amount of capital that bank b ∈ B holds in state s ∈ {0} ∪ S,
Othersb ≡ the ‘others’ item in the balance sheet of bank b ∈ B,
rb ≡ lending rate offered by bank b ∈ B,
rdb ≡ deposit rate offered by bank b ∈ B,
ρ ≡ interbank rate,
rA ≡ the rate of return on market book,
hb ≡ repayment rates of agent hb ∈ H b = {αγ , β δ , θ τ } to his nature-selected bank b ∈ B
vsb
in the consumer loan market,
ω ≡ risk weight on consumer loans, and
e ≡ risk weight on market book.
ω

Equation (1) implies that, at t = 1, the assets of bank b ∈ {γ, τ }, which consist of its
credit extension and market book investment, must be equal to its liabilities obtained from
interbank and deposit borrowing and its initial equity endowment, where ‘Othersb ’ represents
the residual. Equation (2) then shows that, dependent on which of the s ∈ S actually occurs,
the profit that bank b incurs in the second period is equal to the difference between the amount
of money that it receives from its asset investment and the amount that it has to repay on its
liabilities, adjusted appropriately for default in each market. As shown in equation (3), this
profit earned is then added to its initial capital, which in turn becomes its capital in the second
period. Finally, equation (4) implies that the capital to asset ratio of bank b in state s ∈ S is
equal to its capital in state s divided by its risk-weighted assets in the corresponding state.

2.1.2 The interbank net lender’s (bank δ) optimisation problem

Bank δ, unlike the other two banks, is a net lender in the interbank market. Thus it suffers
only a default penalty in the deposit market. Formally, bank δ’s optimisation problem is as
follows;

µ δ ¶2 " #
X π δ π X δ λ δ
s s
max Πδ = ps [ 10 − cδs ]− ps λδks max[0, k − ksδ ] + s10 [ µδd − vsδ µδd ]
mδ ,dδ ,µδd ,vsδ ,s∈S 10 1010 10
s∈S s∈S

subject to

µδd
Aδ + dδ + mδ = eδ0 + + Othersδ (5)
(1 + rdδ )
δ
β
π δs = vsδ es dδ (1 + ρ) − (vsδ µδ + Othersδ + eδ0 ) (6)
mδ (1 + rδ ) + Aδ (1 + rA ) + R d
eδs = eδ0 + π δs (7)
eδs
ksδ = δ (8)
β
ωvsδ es dδ (1 + ρ) + ω
(1 + rδ )mδ + ωR e (1 + rA )Aδ

where,

dδ ≡ bank δ’s investment in the interbank market,

6
 es ≡ the rate of repayment that bank δ expects to get from its interbank investment, and
R
ω ≡ risk weight on interbank investment.

The budget set of bank δ is similar to those of the other two banks except that it invests
in, instead of borrows from, the interbank market. Moreover, its risk-weighted assets in the
second period, as shown in equation (8), also includes bank δ’s expected return on its interbank
investment.

2.2 Central Bank and Regulator

The Central Bank’s and the regulator’s decisions are exogenous. The Central Bank and the
regulator may, but need not, be a single institution. The Central Bank conducts monetary pol-
icy by engaging in open market operations in the interbank market. We assume as our baseline
specification that the Central Bank sets its base money (M ) as its monetary policy instrument,
allowing the interbank rate to be determined endogenously. However, as will be seen in section
3, we also consider an alternative instrument targeting regime whereby the Central Bank fixes
the interbank rate and lets its base money adjust endogenously to clear the interbank market.
As will be explained below, the simulation results depend crucially on which monetary policy
instrument the Central Bank employs. Therefore, whether monetary authorities target base
money or the interbank rates in response to shocks has different implications with respect to
financial stability.
b
The regulator sets capital adequacy requirements for all banks (k , b ∈ B) as well as imposes
penalties on their failures to meet such requirements (λbks , b ∈ B, s ∈ S) and on default on
their financial obligations in the deposit and interbank markets (λbs , b ∈ B, s ∈ S). Finally, he
also sets the risk weights on consumer loan, interbank and market book investment (ω, ω, ω e ).

2.3 Private agent sector

Each household borrower, hb ∈ H b = {αγ , β δ , θτ }, demands consumer loans from his nature-
selected bank b and chooses whether to default on his loans in state s ∈ S. The remaining agent,
Mr. φ, supplies his deposits to each bank b ∈ B. As mentioned, we do not explicitly model the
optimisation problems of households. The reason is that it is very difficult, if at all possible,
to find real data for the (heterogeneous) private agent sector, e.g. the monetary and good
endowment of each banks’s borrowers and depositors. This latter is particularly important
since one of the key objectives of this paper is take our model to real data. So, instead
of explicitly providing microfoundations for households’ decisions, we ‘artificially’ endogenise
them by assuming the following reduced-form equations.

2.3.1 Household Borrowers’ Demand for Loans

Because of the limited participation assumption in every consumer loan market, each house-
hold’s demand for loans is a negative function of the lending rate offered by his nature-selected
bank. In addition, his demand for loans also depends positively on the expected GDP in the
subsequent period. Put differently, we implicitly assume that household borrowers rationally
anticipate GDP in both states of the next period, which then determines their expected fu-
ture income, and adjust their loan demand in this period accordingly in order to smooth their
consumption over time. As in Goodhart et al. (2003, 2004), our money demand function man-
ifests the standard Hicksian elements whereby it responds positively to current and expected

7
income and negatively to interest rates. In particular, we assume the following functional form
of household hb ’s loan demand from his nature-selected bank b, ∀hb ∈ H b , and b ∈ B.
b
ln(µh ) = ahb ,1 + ahb ,2 ln[p(GDPi ) + (1 − p)GDPii ] + ahb ,3 rb (9)

where,
b
µh ≡ amount of money that agent hb ∈ H b chooses to owe in the loan market of bank
b ∈ B, and
GDPs ≡ Gross Domestic Product in state s ∈ S of the second period.

2.3.2 Mr. φ’s Supply of Deposits

Unlike the loan markets, we do not assume limited participation in the deposit markets. This
implies that Mr. φ can choose to diversify his deposits with every bank. Thus, Mr. φ’s deposit
supply with bank b depends not only on the deposit rate offered by bank b but also on the rates
offered by the other banks. Moreover, since banks in our model can default on their deposit
obligations, the expected rate of return on deposit investment of Mr. φ with bank b has to
be adjusted appropriately for its corresponding expected default rate. Finally, Mr. φ’s deposit
supply is a positive function of the expected GDP in the subsequent period.
In sum, since his deposit decisions determine his investment portfolio, given the expected
rates of return, he diversifies among the existing deposit markets. Mr. φ’s deposit supply
function with bank b, ∀b ∈ B, is as follows;

ln(dφb ) = zb,1 + zb,2 ln[p(GDPi ) + (1 − p)GDPii ] + zb,3 [rdb (pvib + (1 − p)viib )]

X
+zb,4 [rdb́ (pvib́ + (1 − p)viib́ )] (10)
b́6=b∈B

where,

dφb ≡ amount of money that agent φ chooses to deposit with bank b ∈ B.

2.3.3 Households’ Loan Repayment Rates

We assume that each household’s repayment rate on his loan obligation to his nature-selected
bank in state s ∈ S is a positive function of the corresponding GDP level as well as the aggregate
credit supply in the economy. The latter variable captures the effect of ‘credit crunch’ in the
economy whereby a fall in the overall credit supply in the economy aggravates the default
probability of every household.9 Specifically, the functional form of the repayment rate of
household hb , ∀hb ∈ H b , to his nature-selected bank b ∈ B, in state s ∈ S is as follows;

h b
ln(vsb ) = ghb ,s,1 + ghb ,s,2 ln(GDPs ) + ghb ,s,3 [ln(m̄γ ) + ln(m̄δ ) + ln(m̄τ )] (11)
9
Higher interest rates, given that households are liquidity constrained, ultimately increase their debt obliga-
tions in the future. Hence, defaults rise.

8
2.4 GDP
As can be seen from equations (9) to (11), we have assumed that households’ actions depend on
their expected GDP in the second period. So, in this section we endogenise GDP in both states
of the second period. We assume that GDP in each state is a positive function of the aggregate
credit supply available in the previous period. Since the Modigliani-Miller proposition does
not hold in our model10 , higher credit extension as a result of loosening monetary policy, or
any other shocks, generates a positive real balance effect that raises consumption demand and
ultimately GDP. In particular, the following functional form for GDP in state s ∈ S of the
second period (GDPs ) holds.

ln(GDPs ) = us,1 + us,2 [ln(m̄γ ) + ln(m̄δ ) + ln(m̄τ )] (12)

2.5 Market Clearing Conditions

There are seven active markets in the model (three consumer loan, three deposit and one
interbank markets). Each of these markets determines an interest rate that equilibrates demand
and supply in equilibrium.11

b
b µh
1+r = , hb ∈ H b , ∀b ∈ B (i.e. bank b’s loan market clears) (13)
mb
µbd
1 + rdb = , ∀b ∈ B (i.e. bank b’s deposit market clears) (14)
dφb
µγ + µτ
1+ρ = (i.e. interbank market clears) (15)
M + dδ

We note that these interest rates, i.e. rb , rdb , and ρ, b ∈ B, are the ex ante nominal interest rates
that incorporate default premium since default is permitted in equilibrium. Their effective (ex
post) interest rates have to be suitably adjusted to account for default in their corresponding
markets.12

2.6 Equilibrium
Let σ b = {mb , µb , µbd , vsb , π bs , ebs , ksb } ∈ R+ × R+ × R+ × R+ 2 × R2 × R2 × R2 for b ∈ {γ, τ };

σ δ = {mδ , dδ , µδd , vsδ , π δs , eδs , ksδ } ∈ R+ × R+ × R+ × R+ 2 × R2 × R2 × R2 ; σ hb ≡ (µhb , v hb ) ∈

sb
R+ × R+ 2 for hb ∈ H b and ; σ φ ≡ (dφ ) ∈ R for b ∈ B; and GDP ∈ R2 . Also, let η ∈
b + s
{rγ , rτ , rδ , rdγ , rdτ , rdδ , ρ}, B b (η) = { σ b : (1) − (4) hold }, and B δ (η) = { σ δ : (5) − (8) hold
b
}. We say that ((σ b )b∈{γ,τ } , σ δ , η, (σ h )hb ∈H b , σ φ , (GDPs )s∈S ) is a monetary equilibrium with
commercial banks and default for the economy
b
E{(eb0 , Othersb , Ab )b∈B ; M ; (k , λbs , λbks , ω, ω, ω
e )b∈B,s∈S ; rA ; p}
iff:
10
See Goodhart et al. (2003) for an extensive discussion.
11
The interest rate formation mechanism is identical to the offer-for-sale mechanism in Dubey and Shubik
(1978). The denominator of each of the expressions (13-15) represents the supply side whereas the numerator
divided by (1 + r), r ∈ {rb , rdb , ρ}, b ∈ B corresponds to the demand. Note that this interest rate formation
mechanism is well-defined both in, and out of, equilibrium.
12
For more on the method of calculating the ex post interest rates, see Shubik and Tsomocos (1992).

9
 (i) (a) σ b ∈ Argmax Πb (π b ), b ∈ {γ, τ }
σb ∈B b (η)
(b) σδ ∈ ArgmaxΠδ (π δ )
σ δ ∈B δ (η)
(i.e. all banks optimise.)

(ii) All markets (13)-(15) clear.

P
vsb µb
b∈{γ,τ }
es =
(iii) R P ,s∈S
µb
b∈{γ,τ }
(i.e. bank δ is correct in its expectation about the repayment rates that it gets from its
interbank investment.)

We emphasise here that the equilibrium conditions (i)−(iii) are consistent with the defining
properties of a competitive equilibrium with rational expectations.
b
(iv) σ h , σ φ and GDPs , for h ∈ H and s ∈ S satisfy the reduced-form equations (9)-(12).
(i.e. loan demand, deposit supply, repayment rates, and GDP in both states satisfy the
reduced-form equations (9)-(12))

3 Calibration
Excluding the Lagrange multipliers, conditions (i) − (iv) in the previous section imply that
we have a system of 56 simultaneous equations in 135 unknown variables, 79 of which are
exogenous variables/parameters in the model. This implies that there are 79 variables whose
values have to be chosen in order to obtain a numerical solution to the model. Thus, they
represent the degrees of freedom in the system and can either be set appropriately or calibrated
against the real data. It is important to note that these variables, which are exogenous when
solving the system of simultaneous equations, do not necessarily have to be those which are
exogenous in the model.13 We report the values of exogenous parameters/variables in the
model and the resulting initial equilibrium in table I. The table also summarises whether the
value of each variable reported is (1) calibrated against real data, (2) arbitrarily selected, or
(3) endogenously solved. We note, however, that, owing to the data confidentiality reason, we
suppress those numbers which are based on the calibrated balance sheet data of UK banks
and replace them by ‘xxx’ in table I. Unless stated otherwise, the values of all the nominal
variables reported therein, e.g. all bank balance sheet items, are normalised by 1010 .
The values of all banks’ balance sheet items in the initial period, i.e. {mb , µbd , Othersb , Ab }b∈B ,
{µ }b∈{γ,τ } , and {dδ }, are calibrated using the 2002 annual account data for seven largest UK
b

banks. Based on this source of data, we also calibrate the values of private agents’ loan re-
payment rates to their nature-selected banks in the good/normal state, i.e. (vib hb )
hb ∈H b } , using
each bank b’s ratio of provision at the end of the year to total customer loans. However, since
there are no data available for crisis/extreme events, the default rates of all private agents in
the bad state (state ii) are arbitrarily set to 0.1.
13
For example, the Central Bank in our model fixes its base money and lets the interbank interest rate adjust
endogenously, i.e. base money is exogenous in the model. However, in solving for a numerical solution, we can
first choose the value of the interbank rate and let the system of simultaneous equations determine endogenously
the value of base money that supports the preset value of the interbank rate.

10
 The probability that state ii will occur, 1 − p, is chosen to be 0.05, given that it reflects
an extreme event. Since banks rarely default on their debt obligations in the good state, the
corresponding repayment rates in the deposit and interbank markets for all banks, i.e. vib , b ∈ B,
are set to 0.999. In state ii, the bad state, we arbitrarily set the analogous repayment rates to
0.95 for banks γ and τ and 0.955 for bank δ. These values are selected to be relatively higher
compared with households’ repayment rates in state ii (0.9) since in reality the probability
that banks would default on their financial obligations is smaller than that of households.
Note also that the chosen value for bank δ’s repayment rate is slightly greater than those of
the other two banks because its deposit rate, whose values are determined endogenously, is
slightly smaller in equilibrium. This may suggest at first glance that we are assuming what we
need to estimate, i.e. a bank’s willingness to run a risky position, which could lead to enforced
shut down. Not quite so, since each chosen value for a bank’s chosen default rate relates to
an equivalent subjective default penalty. If you give us, the model builders, some guidance on
banks’ aversion to default penalties, i.e. the size of the λbs , b ∈ B, s ∈ S, we can adjust the
default probabilities accordingly.
We choose the value of the interbank interest rate, ρ, to be 4 percent to match with the
actual value of UK RP rate in December 2002. The value of risk weight for loans is set to 1
whereas the corresponding values for market book and interbank lending are 0.2. The value
b
of capital to asset requirement set by the regulator for each bank (k , b ∈ B) is chosen to be
slightly higher, but almost equal to, its corresponding value in state i so that all banks always
violate their capital requirement.14
The values of default and capital violation penalties (λbs and λbks , b ∈ B, s ∈ S) reflect
both the tightness of the regulator’s regulatory policy and the (subjective) aversion of banks’
managements to putting themselves at risk of default and/or regulatory violations, and can, in
principle, be treated as inputs given by the practitioner users of this model. Their values are,
however, unobservable and therefore have to be chosen somehow. We have chosen them in this
example to be consistent with the following outcomes. First, the resulting endogenously-solved
banks’ lending rates are such that all banks earn positive profit in state i, whereas they suffer
a loss in state ii. This in turn implies that banks’ capital at t = 2 deteriorates if the bad
state (ii) occurs. Second, all banks’ coefficients of risk aversion (cbs , b ∈ B, s ∈ S) are positive,
implying that banks’ utility functions are well-behaved, i.e. concave. Lastly, the rate of return
on market book is arbitrarily chosen to be 4.5 percent.
We calibrate the value of GDP in the good state (GDPi ) to match with the actual UK
(annual) GDP in 2002. We set the value of GDP in the bad state (GDPii ) to represent a 4%
fall from its corresponding value in the good state. The values of coefficients ahb ,2 and ahb ,3 ,
∀hb ∈ H b , in the reduced-form equation (9) are calibrated, respectively, using the values of the
long-run income and interest rate elasticities of UK household sector estimated by Chrystal
and Mizen (2001).
To our knowledge, we do not know any empirical study which estimates deposit supply
and default probability functions for UK household/private sectors. Although this can, in
principle, be done, such an exercise is beyond the scope of this paper. So, we arbitrarily choose
the appropriate values of zb,2 , zb,3 , zb,4 , ∀b ∈ B, in equation (10), and the values of ghb ,s,2 and
ghb ,s,3 , ∀hb ∈ H b , s ∈ S, in equation (11). As can be seen from table I, the values of ghb ,i,3
14
As mentioned in section 2.1, this is a simplifying assumption. Recall that capital requirements’ violation
b
penalty enters banks’ objective functions as ‘max[0, k − ksb ]’. However, given our assumption that banks always
b
violate their capital requirement, we can restrict the optimisation problem to k − ksb > 0, thus avoiding
‘corner’equilibria.

11
is chosen to be greater than the corresponding values of ghb ,ii,3 , ∀hb ∈ H b , implying that the
effect of a ‘credit crunch’ is assumed to be stronger in the bad state.
The remaining parameters for which their values have to be chosen are the coefficients us,2 ,
∀s ∈ S, in the reduced-form equation (12). We set them to be equal to 0.1. Because the
value of these coefficients capture the inter-relationship between real and nominal variables in
the economy, they are therefore important in determining the strength of the ‘amplification’
effect when a shock hits the economic system. For this reason, as will be seen in section 5, we
conduct a robustness check by considering alternative initial values of these two parameters
in our simulation exercise. At this point we note that the above specifications rely on the
monetary and regulatory non-neutrality properties of our model. We formally prove these
propositions in Goodhart et al. (2003).
Given the chosen values of the variables mentioned above, we are left with the system of
56 simultaneous equations in 56 unknown variables. By solving such system, the values of all
the remaining variables are specified and a numerical solution to the model is obtained.

Table I: Exogenous variables and the resulting initial equilibrium

Initial Equilibrium Exogenous variables in the model
rγ = 0.0601 kiδ = 0.155 eγii = 0.818 Otherγ = 2.31 aαγ ,1 = −3.85
δ
r = 0.069 δ
kii = 0.147 δ
ei = 3.894 δ
Other = −1.142 aβ δ ,1 = −3.35
rτ = 0.0615 kiτ = 0.119 eδii = 3.431 Otherτ = 27.064 aθτ ,1 = −2.08
γ τ τ
rd = 0.04 kii = 0.089 ei = 10.58 gαγ ,i,1 = −0.75 zγ,1 = 1.56
Endogenously rdδ = 0.0357 π γi = 0.25 eτii = 7.307 gαγ ,ii,1 = −1.04 zδ,1 = 2.53
solved τ
rd = 0.04 γ e
π ii = −0.357 Ri = 0.999 gβ δ ,i,1 = −0.76 zτ ,1 = 3.3
γ
µd = xxx δ
π i = 0.327 e
Rii = 0.95 gβ δ ,ii,1 = −1.04 cγi = 0.214
γ
µδd = xxx π δii = −0.136 µα = xxx gθτ ,i,1 = −0.75 cγii = 0.129
δ
µτd = xxx π τi = 2.1 µβ = xxx gθτ ,ii,1 = −1.04 cδi = 0.159
γ τ
ki = 0.098 π τii = −1.173 µθ = xxx ui,1 = 3.64 cδii = 0.351
γ γ
kii = 0.061 ei = 1.425 uii,1 = 3.6 cτi = 0.024
M = 6.928 cτii = 0.042
dφδ = xxx
γ
ρ = 0.04 α = xxx
viγ eγ0 = xxx ω=1
δ
Calibrated mγ = xxx dφτ = xxx β
viδ = xxx eδ0 = xxx ω(eω ) = 0.2
τ
mδ = xxx µγ = xxx θ
viτ = xxx eτ0 = xxx γ
A = xxx
mτ = xxx dδ = xxx GDPi = 104.3 ahb ,2 = 1.41, ∀h ∈ H b Aδ = xxx
dφγ = xxx µτ = xxx ahb ,3 = −0.68, ∀h ∈ H b Aτ = xxx
γ γ
viγ = 0.999 viiδ = 0.955 α = 0.9
viiγ λbks = 0.1, ∀b ∈ B, s ∈ S k = 0.11
δ δ
viiγ = 0.95 viτ = 0.999 β
viiδ = 0.9 λbi = 0.9, ∀b ∈ B k = 0.16
τ τ
viδ = 0.999 viiτ = 0.95 θ
viiτ = 0.9 λbii = 1.1, ∀b ∈ B k = 0.13
GDPii = 100.2 us,2 = 0.1, ∀s ∈ S rA = 0.045
zb,2 = 0.19, ∀b ∈ B p = 0.95
Arbitrarily zb,3 = 0.5, ∀b ∈ B
selected zb,4 = −0.1, ∀b ∈ B
gh,i,2 = 0.05, ∀h ∈ H b
gh,ii,2 = 0.05, ∀h ∈ H b
gh,i,3 = 0.05, ∀h ∈ H b
gh,ii,3 = 0.1, ∀h ∈ H b

12
4 Comparative Static Analysis: Stress Testing UK Banks
In this section we show how the model can be used as a risk assessment tool for UK banks.
Given that the initial equilibrium has been found, we conduct a series of comparative statics
by perturbing each of the variables which are exogenous in the model and studying how the
initial equilibrium changes.15 Tables A1 in the Appendix reports the directional responses of
all endogenous variables in the model given that we increase the values of the variables listed in
the first column one at a time. Recall that these comparative statics are based on the baseline
specification, which assumes that the Central Bank’s monetary policy instrument is its base
money. In what follows, we also conduct an alternative set of comparative statics whereby the
Central Bank is assumed to set the interbank rate as its instrument and let the interbank rate
adjusts endogenously.16 Table A2 in the Appendix reports the results of this alternative set of
comparative statics in an analogous format to table A1.

4.1 An expansionary monetary policy

We first analyse the case when the Bank of England engages in an expansionary monetary
policy. As expected, we found that the result is exactly the same regardless of whether the
Bank uses its base money or the interbank rate as its monetary policy instrument (i.e. the
Bank increases its base money and allows the interbank rate to be determined endogenously,
produces the same result as the case when the Bank decreases the interbank rate and lets
its base money adjust endogenously). We summarise the percentage changes in the values of
certain key variables in response to an increase in the Bank’s base money from 69.28 to 71.28
trillion pounds (approximately 2.9%) in table II. As can be seen from the table, the interbank
rate decreases by 0.8%. Given a lower rate of return on interbank market investment, other
things constant, bank δ invests less in this market (recall that bank δ is the net lender in this
market). It demands less funds from the deposit market and it increases its loan supply to
its nature-selected customer, Mr. β. This portfolio adjustment of bank δ produces a negative
pressure on both its deposit and lending rates. Unlike bank δ, banks γ and τ are the net
borrowers in the interbank market. Thus they respond to a lower cost of interbank borrowing
by reducing their demand for deposits, borrowing more from the interbank market, and lending
more to their nature-selected customers, Mr. α and θ, respectively. This, in turn, causes a
negative pressure on these two banks’ deposit and lending rates.

Table II: % changes in key variables given a positive shock in base money
(Central Bank sets its base money)
Interest rates π bi π bii ebi ebii kib kiib vib viib GDPi GDPii
rdb r b ρ
−4 −5 −3 −2
Bank δ -0.9 -0.9 103 103 104 104 -0.1 -0.1 1063 0.01
−9 −2 −5 −2 3
Bank γ -0.8 -1 -0.8 10 3 -0.01 10 3 103 -0.1 -0.2 10 3 103 0.05 0.05
−2 −2 −3 −3 −1 4
Bank τ -0.8 -1 103 103 104 104
-0.1 -0.1 103 103

All banks rationally anticipate that their greater credit extension would increase the overall
supply of credit in the economy, thus causing the probability of household default to decline.
15
The calculation is carried out using a version of Newton’s method in Mathematica.
16
More specifically, the interbank rate becomes an exogenous parameter in the model under this alternative
specification.

13
This is because greater aggregate credit supply not only directly increases households’ liquidity
but also increases their income in both states of the subsequent period. As can be seen from
table II, GDP increases by 0.05% in both states. Thus, the expected rate of return from
extending loans increases for all banks, implying that their willingness to supply more credit
rises even further.
Given higher expected GDP in both states, every household borrower (i.e. Mr. α, β, and
θ) demands more loans, imposing a positive pressure on the lending rates offered by their
respective nature-selected banks. However, this ‘crowding-out’ effect is dominated by the
corresponding negative pressure from greater credit supply by all banks. Thus, we observe
that their lending rates decline (0.9% for bank δ, and 1% for banks γ and τ ). We also find that
the deposit rates offered by all banks decrease (i.e. 0.9% for bank δ and 0.8% for banks γ and
τ ). This is not only because all banks demand less funds from the deposit markets but also
because Mr. φ responds to higher expected GDP by supplying more deposits to every bank.
Banks in our model choose their optimal expected level of profitability by equating the
derived marginal benefit with the corresponding marginal cost. On the one hand, higher prof-
itability not only directly increases their utility but also raises their capital to asset ratios,
allowing them to suffer less capital violation penalties. This latter source of marginal benefit
is lower the higher the value of banks’ risk-weighted assets.17 On the other hand, in order
to achieve higher profitability, other things constant, they take more risk and therefore suf-
fer higher cost in the form of higher expected default penalties.18 In this comparative static
exercise, since the default probability of all household borrowers decreases in both states, the
corresponding values of every bank’s risk-weighted assets increase. This leads all banks to
revise the trade off between the relative marginal benefit and cost in such as way that they are
willing to achieve a marginally lower level of profitability in both states, compared with the
corresponding initial equilibrium value, in pursuit of suffering less default penalties.19 Con-
sequently, their capital declines slightly in both states compared with the initial equilibrium.
This, together with the fact that the values of their risk-weighted assets increase, causes all
banks to suffer greater capital violation penalties.
Given that banks operate under a perfectly competitive environment, they treat all interest
rates and the households’ default probability as given when making their optimal decisions.
However, in response to shocks, these variables have to adjust endogenously to satisfy market
clearing conditions, where the direction and extent of the adjustment depend on how banks
and households adjust their portfolios. This ‘portfolio reallocation’ effect produces pressures
on banks’ profitability and their willingness to take risk.20 As for bank δ in both states and
banks γ and τ in the bad state, this portfolio reallocation effect is relatively weak, causing
17
For example, if the profit of a bank increases by 1 dollar, given that its risk-weighted asset is 100 dollar, its
1
capital to asset ratio would increase by 100 . However, if the value of the risk-weighted assets is 1000 dollar, the
1
bank’s capital to asset ratio would increase by much less, i.e. by 1000 .
18
Recall that we interpret a bank’s continuous default rate as isomorphic to the probability that the bank
will shut down, and therefore, at least in the very short run, completely default on its obligations. Thus, when
banks choose higher default rates, this implies that it adopts a riskier position.
19
More precisely, higher risk-weighted assets in both states for all banks cause their marginal benefit of
achieving higher profitability in terms of suffering less capital violation penalties to decrease. Given that the
marginal benefit of achieving higher profitability is now lower than the corresponding marginal cost, they reduce
their optimal desired level of profits.
20
For example, a bank may respond to a positive shock by supplying more loans, thereby imposing a negative
pressure on the lending rate. Similarly, the shock may cause household borrowers to demand more loans from
such a bank, causing a positive crowding-out pressure on the lending rate. The relative strength of the pressures
caused by the bank’s and the borrower’s portfolio adjustment depends in general on the relative elasticities of
demand and supply in such a market.

14
them to simply adopt a more conservative position in response to their lower targeted level of
profitability. However, for banks γ and τ in the good state, the portfolio reallocation effect
produces a relatively strong negative pressure on their profitability so that they end up adopting
a slightly riskier position in order to achieve their targeted level of profits.

4.2 A positive deposit supply shock to bank δ in the initial period

We next turn to the scenario where there is a positive deposit supply shock to bank δ in the
initial period. In particular, we increase the (log) autonomous deposit supply of Mr. φ with
bank δ (zδ,1 ) by approximately 0.8%. We first assume that the Bank fixes base money as
its monetary policy instrument, allowing the interbank rate to be determined endogenously.
We report the percentage changes of certain key variables in response to the shock in table
III. Not surprisingly, the results are almost qualitatively identical to that of a change in the
Bank’s injection of funds, since both represent a change in the overall broad money supply.
The difference is that here we assume that the shock is concentrated in bank δ.

Table III: % changes in key variables given a positive deposit supply

shock on bank δ (Central Bank sets base money)
Interest rates π bi π bii ebi ebii kib kiib vib vib GDPi GDPii
rdb r b ρ
−3 −2
Bank δ -2.7 -2.9 -0.03 -0.04 10 3 103
-0.8 -0.9 0.03 0.03
−5 −6 9
Bank γ -2.3 -3.2 -2.3 -0.03 -0.04 103 -0.02 -0.4 -0.5 10 3 103
0.16 0.16
−5 −6 −9 −9 −3
Bank τ -2.3 -3.1 103 103 104 104
-0.4 -0.4 103
0.01

Major differences arise, however, when we analyse the effects of the shock in the context
whereby the Bank fixes the interbank rate as its monetary policy instrument. Key results are
summarised in table IV. In the previous case, even though the effect of the shock is initially
concentrated in bank δ, we observe that the other two banks also benefit from increased overall
liquidity through their interactions in the interbank market. In particular, a positive deposit
supply shock in bank δ causes the bank to supply more liquidity in the interbank market. Such
higher liquidity is then passed on to banks γ and τ in the form of a lower cost of interbank
borrowing. However, in this case, the Bank’s intervention to maintain the interbank rate at its
original level precisely shuts down this interbank rate channel, causing the dominant channel
of contagion to become the one which operates via changes in the household sector’s default
probability in the loan market instead. The chain of contagion of this ‘consumer loan default’
channel begins with bank δ’s decision to supply more credit. This in turn causes the overall
credit supply in the economy to increase. This implies that every household benefits directly
from greater liquidity as well as from higher income (GDP) in both states i and ii of the
subsequent period. Thus, the default probability of every household in the consumer loan
market decreases, causing the expected rate of return from extending more loans to increase
not only for bank δ but also for banks γ and τ . Consequently, their respective lending rates
fall. Since the cost of interbank borrowing is fixed regardless of the amount demanded, the two
banks finance their greater credit extension by borrowing more from the interbank market.

15
 Table IV: % changes in key variables given a positive deposit supply shock
on bank δ (Central Bank sets the interbank rate)
Interest rates π bi π bii ebi ebii kib kiib vib viib GDPi GDPii
rdb r b ρ
4 −2 −1
Bank δ 103 -0.01 -0.02 -0.02 10 3 103
-0.5 -0.5 1083 10
−3
3
−3 −8 −9 3 3
Bank γ 0 103
0 0 0 0 0 104 104
0 0 104 104
−3 −7 −8
Bank τ 0 103
0 0 0 0 104 104
0 0

As can be seen clearly by comparing the results presented in tables III and IV, the contagion
effects when the Bank’s monetary instrument is the interbank rate are much weaker than those
observed when the Bank fixes its base money. This is because fixing the interbank rate not
only directly shuts down the interbank rate contagion channel but also weakens the extent of
contagion effects which operate through the consumer loan default channel. As mentioned, the
contagion effects which operate via the latter channel arise from bank δ’s decision to extend
more credit. However, the Bank’s intervention to fix the interbank rate implicitly increases
the attractiveness of the interbank investment for bank δ since its rate of return does not
diminish as it invests more. Thus, even though we observe that bank δ increases its credit
extension when the Bank fixes the interbank rate, the extent of such increase is not as strong
compared to the case when the rate is allowed to adjust endogenously. Put differently, the
money supply multiplier is larger when the Bank does not target interest rates since besides
their direct effect due to increased deposits, a second-order effect from allowing the interbank
rate to change enhances credit supply of the entire banking sector.

4.3 A positive bank capital shock to bank δ in the initial period

In this comparative static exercise, we increase the capital endowment of bank δ in the initial
period by approximately 5.6%. As before, we first consider the case when the Bank sets its
base money as its instrument. The percentage changes in the values of certain key variables in
response to such shock is summarised in table V. We observe that the directional responses of
most of the variables are the same as the case when we assume that there is a positive deposit
supply shock to bank δ. This is because these two shocks both result in more available funds
for bank δ to invest. However, the exception is the response of bank δ’s capital, and capital to
asset ratios, in the second period. Unlike the case of the positive deposit supply shock, here we
observe that bank δ’s capital position and its capital to adequacy ratios improve in both states
of the world. The reason for this lies in the fact that a positive endowment shock imposes a
direct positive effect on the capital of bank δ in the second period, thus reducing its capital
requirements’ violation penalty.

Table V: % changes in key variables given a positive shock on

bank δ’s capital at t = 1 (Central Bank sets base money)
Interest rates π bi π bi ebi ebii kib kiib vib viib GDPi GDPii
rdb rb ρ
Bank δ -0.9 -0.9 −0.01 -0.01 5.1 5.8 4.9 5.5 0.03 1053
−9 −2 −5 −2 3
Bank γ -0.8 -1 -0.8 103 -0.01 10 3 103 -0.1 -0.2 10 3 103 0.05 0.05
−2 −2 −3 −3 −1 4
Bank τ -0.8 -1 103 103 104 104 -0.1 -0.1 103 103

16
 We now turn to analyse the effects of the same shock but this time under the assumption
that the Bank fixes the interbank rate. Key results are reported in table VI. As in the case
when there is a positive deposit supply shock to bank δ in state ii, we found in this case that
the main contagion effects operate via the consumer loan default channel. However, the major
difference is that here the effects of liquidity injection in bank δ produces negative contagion
effects onto the rest of the banks in the banking sector. Recall that the Bank’s sterilisation
policy in the interbank market increases the relative attractiveness of interbank investment,
implying that bank δ responds to higher capital by increasing its investment in the interbank
market. Unlike the results found in section 1.2, however, the extent of such increase is so large
that bank δ has to switch part of its investment away from the loan market. This in turn puts
a downward pressure on the overall credit supply in the economy, aggravating the probability
of default in the consumer loan markets. This negative contagion effect depresses the other two
banks’ expected return on their credit extension. Thus, they extend less credit, causing the
aggregate output to fall in both states. This in turn worsens the severity of credit crunch in
the economic system even further. Moreover, unlike before, we observe that banks γ and τ now
violate ‘less’ capital adequacy requirements and therefore suffer less capital violation penalties.
This is because a higher default probability of every household causes the values of banks γ’s
and τ ’s risk-weighted assets to decrease. In sum, we see that emergency liquidity assistance in
the form of bank capital injection, when implemented under interest rate targeting regime, may
engender adverse effects to financial stability. Banks may radically restructure their portfolios
in a way that depresses credit extension in the loan markets.

Table VI: % changes in key variables given a positive shock on

bank δ’s capital at t = 1 (Central Bank sets the interbank rate)
Interest rates π bi π bii ebi ebii kib kiib vib viib GDPi GDPii
b
rd r b ρ
−6 −7 −5
Bank δ -0.02 0.05 103 103 5.1 5.8 5 5.7 0.03 10 3
4 4 −2 −2
Bank γ 0 0.02 0 0 0 0 0 103 103 0 0 103 103
3 4
Bank τ 0 0.02 0 0 0 0 103 103
0 0

One of the key implications from our results presented thus far is that, as far as financial
stability is concerned, there is no clear-cut answer as to whether a Central Bank should set the
interbank rate or base money as its monetary policy instrument. Although all the contagion
effects which operate through the interbank rate channel are completely sterilised away when
the Central Bank’s instrument is the interbank rate, such a sterilisation policy may mitigate the
extent of contagion effects which operate via other channels of contagion. As the results in this
section show, the direction of contagion effects can even be reversed, thereby producing negative
rather than positive contagion effects to the rest of the economic system. Disentangling the
issue is clearly beyond the scope of this paper. Since our main objective here is to study financial
contagion and banks’ inter-linkages, we have chosen to focus our analysis in the remainder of
this paper on our baseline specification (i.e. the Bank’s monetary policy instrument is base
money). This allows all contagion channels to operate actively in equilibrium.

4.4 An increase in default penalties imposed on all banks in the bad state
We now turn to analyse the case where the regulator engages in a restrictive regulatory policy
by increasing the default penalties imposed on all banks in the bad state from 1.1 to 1.12

17
(approximately 1.8%). This implies that adopting a riskier position in the bad state is more
costly for all banks since by doing so they have to suffer greater default penalties in this
particular state. Thus, as reported in table VII, all banks optimally choose a significantly lower
desired level of profitability in the bad state, allowing them to adopt a much more conservative
position and therefore to mitigate the extent of default penalties that they have to face in this
particular state. Moreover, bank δ increases its investment in the interbank market which is
relatively ‘safer’ in the bad state.21 In doing so, it borrows more from the deposit market,
and invests less in the loan market. This portfolio adjustment of bank δ imposes a negative
pressure on the interbank rate and produces a positive pressure on its lending and deposit
rates. In response, Mr. φ supplies more deposits with bank δ and less with the other two
banks.

Table VII: % changes in key variables given a rise in default penalty for
all banks in state ii (Central Bank sets base money)
Interest rates π bi π bii ebi ebii kib kiib vib viib GDPi GDPii
b
rd r b ρ
3
Bank δ 0.3 0.3 104
-20.9 1035 -0.8 1093 -0.8 3
103
0.1
Bank γ -0.4 −0.08 -0.4 104 -21.6 105 -9.4 103 −9.4 1083 0.6
1 2 2 −3
103
−3
103
2
Bank τ -0.4 −0.08 105
-20.2 1036 -3.3 1023 −3.3 1093 0.3

Unlike bank δ, banks γ’s and τ ’s initial positions in the interbank market are net borrowers.
Thus, even though interbank investment is now relatively safer, switching its ‘net’ position in
such market would result in a relatively more severe negative portfolio reallocation effect on
these banks’ overall payoff. Anticipating this, they ‘gamble to resurrect’ by extending more
loans, expecting that such an action would increase their profitability in the good state, which
is not subject to higher default penalties, and therefore to reduce the extent of the decline in
their overall payoff which is due mainly to the significant fall in their profitability in the bad
state. This causes their lending rates to decrease both by 0.08%. Given a lower interbank rate,
they also adjust their portfolios by switching away from the deposit market and borrowing
more from the interbank market, causing their deposit rates to decrease by 0.4%.
We observe that the extent of decrease in bank δ’s credit extension is larger than the extent
of increase in banks γ’s and τ ’s credit supply combined, causing the overall supply of credit
in the economy to fall. This directly decreases households’ liquidity, which in turn imposes
a downward pressure on the probability that households will repay their loan obligations in
full. This pressure is further exacerbated since lower aggregate credit supply depresses GDP
in both states of the second period (i.e. by 0.003% in both states), causing the corresponding
income of every household to fall.
As mentioned, because of the direct first-order effect of the initial shock (i.e. higher default
penalty in state ii), all banks are willing to obtain lower profitability in the bad state, compared
with the corresponding initial equilibrium level, and therefore to adopt a more conservative
position in the bad state, causing their corresponding profitability and capital to decrease
so much that they suffer more capital violation penalties in this particular state. In state
i, although households’ default probability is higher, the resulting positive pressure on all
21
More specifically, the probability with which banks (γ, τ ) choose to default completely on their interbank
obligations is less than the corresponding probability of Mr. β on his loans.

18
banks’ risk-weighted assets is relatively weak and therefore outweighed by the corresponding
negative portfolio adjustment effect, causing the values of their risk-weighted assets in this
particular state to decrease. This changing condition alters the trade off between the benefit
derived from higher profitability and the corresponding cost in terms of suffering more default
penalties in such a way that it favours the former. Thus, all banks optimally choose to target a
slightly higher level of profitability in the good state, causing the corresponding value of their
capital to increase marginally. Even though the targeted level of profits is higher, owing to
the relatively strong and positive portfolio reallocation effect, all banks can adopt a slightly
more conservative position in the good state. Finally, since the values of risk-weighted assets
decrease for all banks in the good state and that the corresponding values of their capital
increase, they suffer slightly less capital violation penalties in this particular state.

4.5 An increase in capital violation penalties for all banks in the bad state
We next turn to another restrictive regulatory policy. Let the regulator increase the capital
violation penalties for all banks in the bad state from 0.1 to 0.12 (approximately 20%). We
summarise the percentage changes of some of the key variables in response to such shock in
table VIII.22

Table VIII: % changes in key variables given a rise in CAR penalty for
all banks in state ii (Central Bank sets base money)
Interest rates π bi π bii ebi ebii kib kiib vib viib GDPi GDPii
rdb r b ρ
6 1 2 6 −4
Bank δ −0.02 103 105
0.9 0 0.04 104
0.04 104 103
−6 5 −6 2 4 5 −8 −8
Bank γ 10 3 10 3 103 10 5 1.6 0 0.7 104 0.7 104 -0.05 104 104
−6 −1 1 1 −4
Bank τ 10 3 103
0 0.2 0 0.04 104
0.04 104 103

Since violating capital requirement in state ii is now more costly for all banks, in response,
they engage in the following actions in an attempt to increase the values of their capital to
asset ratios in the bad state. First, they choose to increase their optimal profitability level
in this particular state by adopting a riskier position. The higher profitability, other things
constant, improves their capital position in the bad state, allowing them ultimately to suffer
fewer capital violation penalties. Second, banks γ and δ adjust their portfolios such that the
size of their risk-weighted assets in the bad state decreases, thereby alleviating the extent of
their capital requirements’ violation even further. For bank γ, it reduces its overall investment
by demanding less funds, both from the deposit and interbank markets, and invests less in
the loan market. This causes both the interbank rate and its deposit rate to decrease and its
lending rate to increase. Consequently, Mr. φ supplies less deposits with bank γ and Mr. α
22
As can be seen from the results reported in tables VII and VIII, the effects of an increase in default penalty
are much stronger than the case when we increase the capital requirements violation penalty by a comparable
magnitude. This is owing to our choice of parameterisation which assumes that banks care relatively more
about them being penalised from defaulting on their financial obligations, rather than violating more their
capital requirements. This is a reasonable assumption since banks in our model compare their actual level of
capital with their self-imposed ideal capital holding, not the (Pillar 1) minimum capital requirements. Thus,
as mentioned earlier, the cost of violating capital requirements in our model largely represents a non-pecuniary
loss of comfort and reputation. Such cost should be of less concern to the banks as contrasted with the cost
that they bear from breaching their financial obligations.

19
demands less funds from bank γ. Bank δ also demands less funds from the deposit market,
causing its deposit rate to fall and Mr. φ to supply less deposits with bank δ. Moreover, it
reduces its investment both in the loan and interbank markets. The extent of decline in bank
δ’s interbank investment is further aggravated since the interbank rate is now lower. Because
bank δ’s action produces an upward pressure on its lending rate, Mr. β demands less consumer
loans.
Even though a higher capital violation penalty in state ii produces an upward pressure on
the cost of extending more credit for bank τ , such a pressure is marginally outweighed by the
negative pressure arising from a cheaper cost of interbank borrowing. Thus, bank τ borrows
less from the deposit market, switches to borrow more from the interbank market, and extends
slightly more loans to Mr. θ. Thus, both its deposit and lending rates decrease by 0.006% and
0.001%, respectively.
Because the extent of decrease in credit extension by banks γ and δ over-compensates the
extent of increase in the credit supply of bank τ , the aggregate supply of credit in the economy
decreases. This causes GDP to decline in both states (i.e. by 0.0008% in both states), which
in turn results in a higher default probability of every household.

4.6 A positive shock in GDP in the bad state

Our final comparative static exercise studies the effect of a positive shock in the autonomous
component of GDP in the bad state. In particular, we increase the value of u1 by 0.6%.23
Table IX summarises certain key results.

Table IX: % changes in key variables given a rise GDP in the bad state
(Central Bank sets base money)
Interest rates π bi π bii ebi ebii kib kiib vib viib GDPi GDPii
b
rd r b ρ
−4 −9 −4 −4
Bank δ 2.4 1.1 103 103 103 104 -0.1 -0.2 -0.02 0.06
−7
Bank γ 2 1.1 1 103
-0.02 103 103 -0.1 -0.2 1013
−1 −7 1
103
0.01 2.03
−1 −2 −2 −4 −1
Bank τ 2 1.1 10 3 103 10 3 10 4 -0.1 -0.2 103 0.1

The shock directly increases the expected aggregate output in the second period. This
simultaneously raises all individual borrowers’ demand for loans, and the probability that they
will repay their loans in full, as well as increases Mr. φ’s supply of deposits with every bank.
Given a higher loan demand by Mr. α and θ, the lending rates offered by their respective nature-
selected banks (γ and τ ) increase by 1.1%. Moreover, a decrease in the default probability of
Mr. α and θ further raises the expected rates of return on credit extension for banks γ and
τ . Thus, these banks supply more credit, and demand more funds from both the deposit and
interbank markets. Even though Mr. φ’s decision to deposit more with these banks imposes a
negative pressure on banks γ’s and τ ’s deposit rates, such a pressure is relatively weak when
compared to the positive effect from these banks’ greater demand for deposits. Thus, we
observe that the interbank rate and the deposit rates offered by the two banks increase.
For the same reason as banks γ and τ , bank δ supplies more loans to Mr. β. Unlike the
other two banks, however, bank δ is the net lender in the interbank market. So, it responds
to a higher interbank rate by investing more in the interbank market. To finance its greater
23
This is tantamount to an increase in the autonomous GDP in the bad state by 2.02%.

20
investment, bank δ demands more funds from the deposit market, pushing up its deposit rate
by 2.4%.
As mentioned, the initial shock directly increases the probability that an individual house-
hold will repay his loans in full. The extent of such increase is further magnified since the
overall credit supply in the economy and the aggregate output in both states increase.24 Con-
sequently, the values of all banks’ risk-weighted assets rise. This in turn implies that the
marginal benefit of higher profitability in terms of suffering less capital violation penalties for
all banks is lower. Thus, banks revise their optimal level of profits downward in order to suffer
lower default penalties. This directly worsens their capital position. Given lower profitability,
bank γ are able to adopt a more conservative position in both states. The same is true for
banks δ and τ in the bad state. However, the negative portfolio reallocation effect on banks
δ’s and τ ’s profitability in the good state is so strong that they have to adopt a riskier position
to be able to achieve the targeted profitability level.

5 Robustness Check: The effect of an expansionary monetary

policy under alternative initial conditions
In section 4, we analysed the effect of an expansionary monetary policy on the economic
system based on a given value of the parameters us,2 , ∀s ∈ S, in the reduced-form equation
(12). We have already noted that these parameters are important because they capture the
degree of responsiveness of aggregate economic activities, as measured by GDP, to changes
in financial variables of the model. Put differently, they represent the multiplier effect of
changes in the overall credit supply in the economy with respect to aggregate output in the
subsequent period. Thus, in this section, we provide a robustness check for the result of one
of the comparative statics that we obtained in the previous section, i.e. a positive monetary
policy shock, by studying the sensitivity of the impact of the shock on the economy as we
simultaneously vary the value of the parameters us,2 , ∀s ∈ S. In doing so, we re-do another
two independent comparative static exercises analogous to that shown in section 4.1, assuming
instead that the value of the two parameters is 0 in the first and 0.05 in the second. Moreover,
in order to maintain the same ‘initial’ equilibrium values of all endogenous variables across the
these comparative statics, the values of autonomous part of (log) GDP (i.e. us,1 , ∀s ∈ S) are
adjusted accordingly. In this way, the initial equilibrium values of GDP in both states are the
same across all comparative static exercises. Figure 2 reports the percentage changes in key
variables in response to an expansionary monetary policy (i.e. an increase in base money from
69.28 to 71.28 trillion pounds) under different values of us,2 , ∀s ∈ S, i.e. 0, 0.05, and 0.1.
Assuming the same positive monetary policy shock, the remainder of this section analyses
how the response of key endogenous variables changes as we simultaneously increase the value
of us,2 , ∀s ∈ S. Given a higher value of these parameters, other things constant, an increase in
same amount of aggregate credit supply results in a larger increase in GDP in both states of
the subsequent period, causing, in turn, households’ default probability in the loan market to
decrease more precipitously. This implies that, a higher value of these two parameters, other
things constant, results in a higher expected rate of return from extending consumer loans for
all banks. Thus, all banks increase their credit supply even more, causing the extent of increase
in the aggregate credit supply in the economy to rise. This implies that extent of increase in
GDP and the probability that households will repay in full in both states increases.
24
As can be seen from table IX, GDP in the bad state rise by 2.03%, 2.02% of which is from the direct effect
and 0.01% of which is from the indirect effect via higher aggregate supply of credit in the economy.

21
 Banks' deposit rates Banks' lending rates interbank rate
0 0 0

0.05

0.1

0.05

0.1
0

0.05

0.1
-0.006 -0.006

-0.015

-0.012 -0.012
delta delta
gamma gamma
tao tao -0.018
-0.03 -0.018

Banks' loan supply Banks' deposit demand Delta's interbank investment

0.0018 0
deta -0.014

0.05

0.1
gamma

0.05

0.1
0.0017
tao
-0.0004
delta
0.0016
gamma
-0.015
tao
0.0015
-0.0008

0.0014
0

0.05

0.1

-0.0012 -0.016

Banks' interbank borrowing Banks' profitability in the good Banks' profitability in the bad
0.04 state state
gamma
0 0
tao
0

0.05

0.1
0.03

0.05

0.1
0.02
-0.00005 -0.0001
0.01
delta deta
0 gamma gamma
tao tao
0

0.05

0.1

-0.0001 -0.0002

Banks' CAR ratio in the good Banks' CAR ratio in the bad state Banks' repayment rate in the
state good state
0
0 0.0002
0

0.05

0.1

delta
0

0.05

0.1

gamma
tao
0.0001
-0.001
-0.001

delta 0
delta
0

0.05

0.1
gamma gamma
tao tao
-0.002 -0.002 -0.0001

Banks' repayment rate in the bad GDP Households' loan demand

state 0.0006 0.0015
alpha
0.0002 good state
delta beta
gamma bad state theta
0.0004 0.001
tao
0.0001
0.0002 0.0005

0 0 0
0

0.05

0.1

0.05

0.1
0

0.05

0.1

Mr. Phi's deposit supply with Households' repayment rate in Households' repayment rate in
banks the good state the bad state
0.0001 0.0004 0.0006
alpha alpha
beta beta
0 0.0003
theta 0.0004 theta
0

0.05

0.1

0.0002
-0.0001
0.0002
0.0001
-0.0002 delta
gamma 0 0
tao
0

0.05

0.1

0.05

0.1

-0.0003

Figure 2: % changes in key variables in response to an expansionary monetary policy under

different values of us,2 , ∀s ∈ S (i.e. 0, 0.05, 0.1).
22
 Since all banks supply more credit, they need more funds to finance their higher investment.
Thus, although, as explained in section 4.1, all banks respond to a positive monetary policy
shock by decreasing their demand for deposits, the rate of such decrease falls. This in turn
implies that the extent of decline in every bank’s deposit rates has to be smaller as well.
Moreover, banks γ and τ , which are net borrowers in the interbank market, demand an even
greater amount of interbank borrowing. This causes a smaller decline in the interbank rate.
Because of the bigger multiplier effect of GDP with respect to aggregate credit extension
in both states, the increase in all household borrowers’ loan demand is greater. This effect is
strong enough to reduce the rate of decrease in every banks’ lending rates. Moreover, Mr. φ
increases his deposit supply to banks, thus increasing their available investment funds. This
causes not only the percentage increase in bank δ’s investment in the loan market to be higher
but also the percentage decrease in its interbank investment to be lower.
Because the extent of decrease in households’ default probability in both states is bigger,
the values of every bank’s risk-weighted assets grow by an even larger amount. This reduces
the incentive for banks to achieve higher profitability even further. Consequently, all banks
choose to achieve an even lower optimal level of profits, exacerbating the extent of decline in
the capital violation penalties (recall that profits enter the numerator of the capital adequacy
requirements). Given that the targeted level of profits decreases more substantially, banks
γ and τ choose a relatively more conservative portfolio, though such position is still riskier
compared to their initial equilibrium positions. We observe, however, that the percentage
increase in bank δ’s probability of default declines. This is because the negative pressure from
the portfolio reallocation effect is relatively stronger for bank δ.
In sum, we observe that the effect of an expansionary monetary policy on the economic
system is, in general, directionally the same under alternative values of us,2 , ∀s ∈ S. However,
the magnitude of its effect on the banking sector and the real economy is found to be stronger
as we increase the value of the parameters. This is because, in such a case, the ‘multiplier’
effect from changing aggregate credit condition to default and ultimately to aggregate output
is stronger.

6 Conclusion
This is the first attempt at calibration, to bring the model to real (UK) data, that we have
made; the programme of work that we are following is just beginning to span the chasm between
pure theory and practical empirical modelling. Nevertheless there is a long way yet to go. For
example, this remains a two-period model only, in which banks take decisions in period one, e.g.
determining their holdings of loans and deposits, on the basis of expectations of the potential
states in period two; they then are stuck with their decisions in period two. In a longer period
model, the outcome in period two would cause banks to revise their expectations of states in
period three (e.g. as in a Markov switching model), thereby causing them dynamically to revise
their loans/deposits in period two, and so on. Also, for a variety of reasons, mostly connected
with data availability, we took the value of each bank’s trading books (its investments) as a
constant throughout.
Although we have based this exercise on real UK data, it remains a simulation. It does not
provide an independent check whether our model can capture the main time series properties
of the major UK banks. Trying to do this latter, and also to work out some way to lengthen the
number of time periods in the model, without making the model far too unwieldy to solve are
priorities for future research. In particular, this analysis reveals how complex, and complicated,

23
default is as an institution, which is one reason why it rarely figures in other model.
Given these constraints, the focus of this exercise was on adjustments in the interbank
market, and in the relative interest rates on deposits and loans; hence, in part via changes
in bank margins, this fed back into changes in bank profits, capital and CARs. Like most
other empirical research in this field, we do not find much serious contagion occurring via
the interbank market with our arbitrarily chosen set of banks. Note that we could re-do this
exercise for any other pair of UK banks (with a third residual banking sector). One result
that practitioners will have expected, but not perhaps academics, is that contagion is much
diminished if the Central Bank targets interest rates rather than a fixed time path for base
money. We intend to write this latter up as a separate academic article.
Perhaps the most striking result is the sensitivity of banking profitability (in a bad state)
to changes in the ‘default penalty’ on banks, see Table VII. We think of this as measuring
the general risk aversion of banks, which is a function of the banks’ own conservatism, and
concern for regulation, interacting with externally-imposed discipline from markets and from
the regulatory/supervisory regime. The greater the risks that bank managers are prepared to
run, in pursuit of profit, the greater the probability of default in a bad state.
No doubt a glimpse of the obvious, but a problem is that such risk aversion is not objec-
tively measurable. Indeed, but it is crucial, at least in our models. What this means is that
to run models where default matters, the model builders will have to depend on the regula-
tors/supervisors to give them some input, e.g. in the form of rankings, on the relative appetite
for risk of the various banks involved.
It is not, however, clear whether this is a ‘good’ result, since it would require all concerned
to focus on the really important issues, or a ‘bad’ result since it reveals just how difficult
quantification and modelling continues to be in this field.

7 Appendix
Table A1:
(Central Bank sets base money)
rδ rγ rτ rdδ rdγ rdτ ρ mδ mγ mτ µδd µγd µτd dδ µγ µτ ei
R eii
R
M − − − − − − − + + + − − − − + + − +
≈ ≈
zδ1 − − − − − − − + + + + − − + + + − +
≈
zγ1 − − − − − − − + + + − + − − − + − +
≈ ≈
eδ0 − − − − − − − + + + − − − + + + − +
≈ ≈ ≈ ≈ ≈
λbii + − − + − − − − + + + − − + + + + +
≈ ≈ ≈ ≈
λbkii + + − − − − − − − + − − − − − + + −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω + − + + + + + − + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
u1 + + + + + + + + + + + + + + + + − +
≈
Note: +(−) : substantial increase (decrease),
+(−) :weak increase (decrease), ≈: approximately equal
≈ ≈
An increase in λbii (λbkii ) denotes an increase in the default (CAR) penalty for all banks

24
 Table A1 (continue):
(Central Bank sets base money)
viδ viiδ vδ viγ viiγ v γ viτ viiτ v τ Πδ Πγ Πτ ΠB π δi π δii π γi π γii
M + + + − + − − + − + − − + − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
zδ1 + + + − + − − + − + − − + − − − −
≈ ≈ ≈ ≈
zγ1 + + + − + − − + − + − − + − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
eδ0 + + + − + − − + − + − − + − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
λbii + + + + + + + + + − + + + + − + −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
λbkii + − + + − − + − ≈ + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω − − − + + + + + + − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
u1 − + − + + + − + + − + + − − − − −
≈ ≈ ≈ ≈ ≈ ≈
Note: v b ≡ pvib + (1 − p)viib , b ∈ {γ, δ, τ } and
ΠB ≡ (Πδ + Πγ + Πτ )/3
Table A1 (continue):
(Central Bank sets base money)
π τi π τii eδi eδii eγi eγii eτi eτii kiδ kiiδ k δ kiγ kiiγ kγ kiτ kiiτ kτ k
M − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
zδ1 − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈
zγ1 − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
eδ0 − − + + − − − − + + + − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈
λbii + − + − + − + − + − − + − − + − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈
λbkii ≈ + ≈ + ≈ + ≈ + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈
u1 − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈
Note: kb≡ pkib + (1 − p)kiib , b ∈ {γ, δ, τ } and
k ≡ (k δ + kγ + k τ )/3
Table A1 (continue):
(Central Bank sets base money)
δ γ τ
µβ µα µθ dφδ dφγ dφτ viγ
α α
viiγ β
viδ β
viiδ θ
viτ θ
viiτ GDPi GDPii
M + + + + + + + + + + + + + +
≈ ≈ ≈
zδ1 + + + + + + + + + + + + + +
≈ ≈
zγ1 + + + + + + + + + + + + + +
≈ ≈
eδ0 + + + + + + + + + + + + + +
≈ ≈ ≈
λbii − − − + − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
λbkii − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω + + + + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
u1 + + + + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈

25
 Table A2:
(Central Bank sets the interbank rate)
rδ rγ rτ rdδ rdγ rdτ M mδ mγ mτ µδd µγd µτd dδ µγ µτ ei
R eii
R
zδ1 − − − + ≈ ≈ − + + + + + + + + + ≈ +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
zγ1 ≈ ≈ ≈ ≈ ≈ ≈ − ≈ ≈ ≈ ≈ + ≈ ≈ − ≈ ≈ ≈
eδ0 + + + − ≈ ≈ − − − − − − − + − − ≈ −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
λbii + + + + ≈ ≈ − − − − + − − + − − + +
≈ ≈ ≈
λbkii + + + − ≈ ≈ − − − − − ≈ ≈ + − − + −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω − − − + ≈ ≈ + + + + + + + − + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
u1 − − − + ≈ ≈ + + + + + + + − + + − +
≈
Note: +(−) : substantial increase (decrease),
+(−) :weak increase (decrease), ≈: approximately equal
≈ ≈
Table A2 (continue):
(Central Bank sets the interbank rate)
viδ viiδ vδ viγ viiγ v γ viτ viiτ v τ Πδ Πγ Πτ ΠB π δi π δii π γi π γii
zδ1 + − + ≈ + ≈ ≈ + ≈ − − ≈ − − − ≈ ≈
≈ ≈ ≈ ≈
zγ1 ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
eδ0 + − + ≈ ≈ ≈ ≈ ≈ ≈ + + + + − − ≈ ≈
≈ ≈ ≈ ≈ ≈ ≈ ≈
λbii − + + + + + + + + − + + + + − + −
≈ ≈ ≈ ≈ ≈ ≈
λbkii + − + ≈ − − + − ≈ + + + + ≈ + ≈ +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω − − − + + + + + + − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
u1 − + − − + + − + + − − − − − − − −
≈ ≈ ≈ ≈ ≈
Note: vb≡ pvib + (1 − p)viib , b ∈ {γ, δ, τ } and
ΠB ≡ (Πδ + Πγ + Πτ )/3
Table A2 (continue):
(Central Bank sets the interbank rate)
π τi π τii eδi eδii eγi eγii eτi eτii kiδ kiiδ k δ kiγ kiiγ kγ kiτ kiiτ kτ k
zδ1 ≈ ≈ − − ≈ ≈ ≈ ≈ − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
zγ1 ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
eδ0 ≈ ≈ + + ≈ ≈ ≈ ≈ + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈
λbii + − ≈ − ≈ − ≈ − + − + + − − + − − −
≈
λbkii ≈ + ≈ + ≈ + ≈ + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈
u1 − − − − − − − − − − − − − − − − − −
≈ ≈ ≈ ≈
Note: kb ≡ pkib + (1 − p)kiib , b ∈ {γ, δ, τ } and
k ≡ (k δ + kγ + k τ )/3

26
 Table A2 (continue):
(Central Bank sets the interbank rate)
δ γ τ
µβ µα µθ dφδ dφγ dφτ viγ
α α
viiγ β
viδ β
viiδ θ
viτ θ
viiτ GDPi GDPii
zδ1 + + + + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
zγ1 ≈ ≈ ≈ ≈ + ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
eδ0 − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
λbii − − − + − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈
λbkii − − − − − − − − − − − − − −
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
ω + + + + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈ ≈
u1 + + + + + + + + + + + + + +
≈ ≈ ≈ ≈ ≈ ≈ ≈

References
[1] Bhattacharya, S., Goodhart, C.A.E., Sunirand, P., Tsomocos, D.P., 2003, Relative per-
formance, banks and sequential contagion. mimeo, University of Oxford.

[2] Catarineu-Rabell, E., Jackson, P., Tsomocos, D.P., 2003, Procyclicality and the new Basel
Accord-banks’ choice of loan rating system. Economic Theory (forthcoming).

[3] Chrystal, A., Mizen, P., 2003, Consumption, money, and lending: a joint model for the
UK household sector. Journal of Money Credit and Banking (forthcoming).

[4] Dubey, P., Shubik, M., 1978, The Non-cooperative equilibria of a closed trading economy
with market supply and bidding strategies. Journal of Economic Theory, 17:1-20.

[5] Elsinger, H., Lehar, A., Summer, M., 2003, Risk assessment for banking systems. Oester-
reichische Nationalbank (Austrian National Bank) Working paper no. 79.

[6] Goodhart, C.A.E, 1989, Money, Information and Uncertainty, 2nd Edition (Macmillan:
London).

[7] Goodhart, C.A.E., Sunirand, P., Tsomocos, D.P., 2003, A model to analyse financial
fragility. Oxford Financial Research Centre Working Paper No. 2003fe13.

[8] Goodhart, C.A.E., Sunirand, P., Tsomocos, D.P., 2004, A model to analyse financial
fragility: Applications. Journal of Financial Stability (forthcoming).

[9] Shubik, M., 1999, The Theory of Money and Financial Institutions (Cambridge, Massa-
chusetts: The MIT Press).

[10] Shubik, M., Tsomocos, D.P., 1992, A strategic market game with a mutual bank with
fractional reserves and redemption in gold. Journal of Economics, 55(2):123-150.

[11] Shubik, M., Wilson, C., 1977, The optimal bankruptcy rule in a trading economy using
fiat money. Journal of Economics, 37:337-354.

[12] Tsomocos, D.P., 2003a, Equilibrium analysis, banking and financial instability. Journal of
Mathematical Economics, 39(5-6):619-655.

27
 [13] Tsomocos, D.P., 2003b, Equilibrium analysis, banking, contagion and financial fragility.
Greek Economic Review 23.

[14] Tsomocos, D.P., Zicchino, L., 2004, On modelling endogenous probability of default.
mimeo, Bank of England.

28

View publication stats