Risk Management Strategies for Banks∗

Wolfgang Bauer†‡ Marc Ryser§

November 12, 2002

Abstract
We analyze optimal risk management strategies of a bank financed
with deposits and equity in a one period model. The bank’s motivation
for risk management comes from deposits which can lead to bank runs. In
the event of such a run, liquidation costs arise. The hedging strategy that
maximizes the value of equity is derived. We identify conditions under
which well known results such as complete hedging, maximal speculation
or irrelevance of the hedging decision are obtained. The initial debt ratio,
the size of the liquidation costs, regulatory restrictions, the volatility of
the risky asset and the spread between the riskless interest rate and the
deposit rate are shown to be the important parameters that drive the
bank’s hedging decision. We further extend this basic model to include
counterparty risk constraints on the forward contract used for hedging.

JEL Classification: G1; G21; G28

Keywords: Bank; Bank risk management; Corporate hedging

1 Introduction
The focus of this paper is to study the rationale for banks’ risk management
strategies where risk management is defined as set of hedging strategies to alter
the probability distribution of the future value of the banks’ assets.
There is a broad literature on these decisions for firms in general, beginning
with Modigliani and Miller (1959): Their famous theorem states that in a world
of perfect and complete markets, financial decisions are irrelevant as they do not
alter the value of the shareholder’s stake in the firm. The only way to increase
∗ Wewish to thank the RiskLab, ETH Zürich, for the financial support of the first author
and ECOFIN Research and Consulting for the financial support of the second author. This
work is part of the RiskLab project titled ‘Banks’ Optimal Hedging Decisions Under the
Threat of Bank Runs’. We gratefully acknowledge helpful comments from Rajna Gibson,
Rene Stulz, Giorgio P. Szegö as well as an anonymous referee.
† Swiss Banking Institute at University of Zürich and RiskLab at ETH Zürich.
‡ Corresponding author: Swiss Banking Institute, Plattenstr. 14, 8032 Zürich, Switzerland,

Phone: +41 1 634 27 22, Fax: +41 1 634 49 03, E-mail: wobauer@isb.unizh.ch, web page:
http://www.math.ethz.ch/~bauer.
§ ECOFIN Research and Consulting (Zürich), E-mail: marc.ryser@ecofin.ch.

1
shareholder’s wealth is to increase value of the firm’s assets. Neither the capital
structure nor the risk management decisions have an impact on shareholder’s
wealth.
Some important deviations from the perfect capital markets in the Modigliani
Miller setting have been identified, giving motivations for firms to care about
risk management, such as taxes, bankruptcy costs, agency costs and others
(Froot et al., 1993; Froot and Stein, 1998; Smith and Stulz, 1985; DeMarzo and
Duffie, 1995; Stulz, 1996; Shapiro and Titman, 1986). When these reasons for
risk management are incorporated into the firm’s objective function, one finds
the following basic result: When all risks are perfectly tradeable the firm max-
imizes shareholder value by hedging completely (Froot and Stein, 1998; Broll
and Jaenicke, 2000; Mozumdar, 2001).1
However, the Modigliani-Miller-theorem as well as the aforementioned hedg-
ing motives are ex ante propositions: Once debt is in place, ex post financial
decisions can alter the equity value by expropriating debt holders. This strategy
is known as asset substitution (Jensen and Meckling, 1976). Because of limited
liability, the position of equity holders can be considered as a call option on
the firm value (Black and Scholes, 1973). This implies that taking on as much
risk as possible is the optimal ex post risk management strategy. In summary,
theory is inconclusive regarding the question of the optimal hedging strategy of
firms.
Turning to the question of optimal hedging and capital structure decisions of
banks, a first finding is that the analysis within the neoclassical context of the
Modigliani-Miller-theorem would be logically inconsistent. Banks are redundant
institutions in this case and would simply not exist (Freixas and Rochet, 1998,
p. 8). The keys to the understanding of the role of banks and their financial
decisions are transaction costs and asymmetric information. These features
have been dealt with extensively in the banking literature, departing from the
neoclassical framework (Baltensperger and Milde, 1987; Freixas and Rochet,
1998; Merton, 1995; Schrand and Unal, 1998; Bhattacharya and Thakor, 1993;
Diamond, 1984, 1996; Kashyap et al., 2002; Allen and Santomero, 1998, 2001):

• Banks have illiquid or even nontradeable long term assets because of the
transformation services they provide.

• Part of the illiquidity of banks’ assets can be explained by their infor-

mation sensitivity; banks can have comparative informational advantages
due to their role as delegated monitors. Examples include information
about bankruptcy probabilities and recovery rates in their credit portfo-
lio. This proprietary information can be further improved through long
term relationships with creditors (Boot, 2000; Diamond and Rajan, 2000).

• In contrast to other firms, banks’ liabilities are not only a source of financ-
ing but rather an essential part of their business: Depositors pay implicit
1 This result is a consequence of the payoff-function’s concavity induced by the risk man-

agement motives and the application of Jensen’s inequality.

2
 or explicit fees for deposit-related services (i.e. liquidity insurance, pay-
ment services, storage). The leverage in banks’ balance sheets is thus
many times higher.
• Bank deposits can be withdrawn at any time. The sequential service
constraint on these contracts and uncertainty about the bank’s ability to
repay can lead to a “bank run” situation: All depositors rush to the bank
at the same time to withdraw their money, trying to avoid being the last
one in the waiting queue. This threat of bank runs creates an inherent
instability for the bank’s business (Diamond and Dybvig, 1983; Jacklin
and Bhattacharya, 1988).
These characteristics highlight the major differences between banks and other
firms: Banks, in contrast to other corporations, are financed by deposits. Their
ongoing operating value would be lost to a large extent in case of bankruptcy;
depositors can immediately call their claims and run whereas illiquid and infor-
mation sensitive assets have to be liquidated by fire sales at significant costs (Di-
amond and Rajan (2001, 2000); Shrieves and Dahl (1992); the size of bankruptcy
costs of banks was estimated in James (1991)). However, these features of a bank
are ignored by most of the literature on capital structure and hedging decisions,
which usually deals with nonfinancial firms.
In a recent contribution, Froot and Stein (1998) developed a framework to
analyze a bank’s optimal capital allocation, capital budgeting and risk manage-
ment decisions. Their motivation for the bank to care about risk management
stems from convex costs of external financing for a follow-up investment oppor-
tunity. This induces the bank’s objective function to be concave (the authors
call this internal risk aversion): The more difficult it is for the bank to raise
external funds, the more risk averse it behaves. A publicly traded bank in an
efficient and complete market does not reduce shareholder value by sacrificing
return for a reduction in risk. Thus, risk reduction is always desirable for the
risk-averse bank in the Froot and Stein (1998)-setting. Hence, the resulting
optimal strategy is to hedge completely. However, the authors omit equity’s
feature the limited liability and the corresponding agency problems between
shareholders and debtholders. Furthermore, since in their model, there is no
depository debt and thus no bank run possibility, potential effects of defaults
on capital structure and risk management decisions are ignored.
In this paper, we model the hedging decision of a bank with the aforemen-
tioned characteristics. We assume the capital budgeting decision to be fixed. In
a one-period-two-states-model, the bank has a given amount of depository debt.
The deposit rate contains a discount due to deposit-related services. The present
value of this discount constitutes the bank’s franchise value. On the other hand,
bank runs can force the bank to sell all of its assets at once, incurring signif-
icant liquidation costs. This creates an incentive for not having extraordinary
high levels of depository debt. Further, we assume that the bank is restricted
in its risk taking behavior by a regulator. We also incorporate limited liability
for equity. We assume that the bank’s management acts in the shareholder’s
interest and maximizes the present value of the equity. It faces thus conflicting

3
incentives for risk management: Regulatory restrictions and liquidation costs in
case of bank runs limit the risk taking on one hand. On the other hand, limited
liability creates incentives for risk taking. This setting allows us to identify
situations in which well known results from the corporate finance literature are
found: We show that for some banks, it is optimal to hedge completely as in
Froot and Stein (1998). Other banks will take on as much risk as possible to
augment shareholder value by expropriating wealth from depositors, a strategy
known as asset substitution (Jensen and Meckling, 1976). For still other banks,
the risk management decision is shown to be irrelevant as in Modigliani and
Miller (1959).
The remainder of this paper is organized as follows. In section 2, we present
the model, discuss the bank’s objective function and derive the optimal hedging
strategy. In section 3, we discuss the impact of forward counterparty restrictions
on the hedging positions of the bank: Since depositors have absolute priority
because of their possibility to withdraw at any time, the forward counterparty
can face additional default risk. It may therefore limit its contract size with the
bank. Section 4 concludes the analysis and gives an outlook on further research
possibilities.

2 The general model

2.1 The market
Let a probability space (Ω, F, P) be given, where we define Ω := {U, D}, F :=
{∅, {U}, {D}, Ω} and P(U) = p. The model has one period, between time t = 1
and t = 2 and T ≡ {1, 2} denotes the set of time indices.
The market consists of two assets: A riskless asset has at time t = 1 a value
normalized to 1, B1 = 1, and B2 = B1 R at time t = 2 where R > 1 is fixed and
given; further, a risky asset with value P1 > 0 at time t = 1 and a value P2 (ω)
at time t = 2 where (
Pu ≡ P1 u, ω = U,
P2 (ω) =
Pd ≡ P1 d, ω = D,
where we assume that
u>R>d (1)
For hedging purposes, we further introduce a redundant forward contract on the
risky asset: It is entered at time t = 1 at no cost and the buyer of the contract
has to buy one unit of the risky asset at time t = 2 at the forward price RP1 .
Hence, the value ft of the forward contract is

f1 = 0,
(
fu ≡ Pu − RP1 , ω = U,
f2 (ω) =
fd ≡ Pd − RP1 , ω = D.

4
Since we have two assets with linearly independent payoffs and two states of the
world, the market is complete.
h Wei define the unique risk neutral probability Q
Q P2
by Q(U) =: q such that E B2 = P1 , where q = R−d u−d .

2.2 The bank

To derive the bank’s objective function, we make the following two assumptions
that deal with agency problems: We explicitly exclude agency problems between
shareholders and bank managers as their decision-taking agents. However, since
banks empirically have very high debt levels, we take asset substitution as an
agency problem between shareholders and depositors into account. Therefore,
the problem of choosing risk after the choice of the initial capital structure is
especially pronounced (Leland, 1998):

Assumption 1. Management’s compensation is structured to align the man-

ager’s interests with those of the shareholders. Therefore the firm’s objective is
to maximize the value of equity.

This objective is based on the completeness of the financial market. It is

therefore possible to achieve any distribution of wealth across states. The Fisher
separation theorem then states the following: All utility maximizing sharehold-
ers agree on the maximization of firm value as the appropriate objective function
for the firm, notwithstanding the differing preferences and endowments (Eich-
berger and Harper, 1997, p. 150). However, as Jensen and Meckling (1976)
pointed out, shareholders in levered firms can do better behaving strategically.
They will prefer investment or hedging policies that maximize the value of only
their claim, if they are not forced to a precommitment on the investment and
hedging strategy.

Assumption 2. When setting its capital structure, the bank can not precontract
or precommit its hedging strategy. It will choose the hedging strategy ex post,
after deposits have been raised.
At time t = 1, the bank has a loan portfolio, which has the same dynamics
as the risky asset. Its value at t = 1 equals αP1 , we will thus say it has a
prior position of α > 0 units of the risky asset.2 The bank has two sources
of capital: Depository debt and equity where the latter has limited liability.
The initial amount of depository debt D1 is given. While it would also be
interesting to analyze the bank’s capital structure decision, we limit our analysis
to the hedging policy, assuming that the bank has already set its target capital
structure.
2 Through their monitoring activity, banks may be able to generate additional rents on the

asset side as well. These proprietary assets are however often not tradeable. In this situation,
the market is incomplete. This incompleteness creates problems for the determination of a
unique objective function for the bank and we leave the analysis of the case with nontradeable
proprietary assets for further research.

5
2.3 The deposits and the run-threat to equity
In most papers dealing with the capital structure of firms in general, the tax-
advantage of debt is a main incentive for firms to carry debt. For banks, however,
there is a more important motivation for carrying depository debt. Depository
debt in banks can be regarded as a real production element (Bhattacharya and
Thakor, 1993). Due to deposit-related services (liquidity provision, payment
services), the deposit rate will be lower than the rate that fully reflects the risk.
We assume that the bank gets a discount of τ > 0 on the deposit rate, resulting
in
D2 = D 1 R D ,
where RD > 1 is the deposit rate net of the discount that the bank receives. We
call the net present value of these discounts from future periods the franchise
value of the deposits, denoted by F Vt , t ∈ T :
τ D1
F V1 =
B2
F V2 = 0,
h i
where τBD21 = EQ τBD21 .
Because of its significant influence on the bank’s equity payoff, we should
highlight another important difference between bank deposits and traded debt:
Asset-return shocks affect market prices of traded debt equally over all debt
holders, whereas the nominal amount of deposits can be withdrawn at any
time. However, as Diamond and Dybvig (1983) pointed out, the sequential ser-
vice constraint on these fixed-commitment contracts along with sudden shocks
in the liquidity needs of depositors can lead to a situation in which all depositors
withdraw their money at the same time. This is because the amount received
by a individual depositor solely depends on his relative position in the waiting
queue. Such a bank run can happen as a “sunspot phenomenon”, whenever
there is a liquidity shock and even in the absence of risky bank assets. 3 When
uncertain asset returns are introduced into the analysis, there is another reason
why bank runs can occur: Whenever the value of the bank’s assets is not suffi-
cient to repay every depositor’s full claim, all fully informed rational depositors
would run to the bank at the same time and cause a so-called information based
bank run (Jacklin and Bhattacharya, 1988).
Let us assume that there are n depositors with equal amounts of D2 /n of
deposits. We denote by VL the critical asset value below which there will be a
bank run. Without liquidation costs, we find that VL = D2 . Indeed, whenever
the value V2 of the bank’s assets at time t = 2 exceeds the nominal deposits
D2 , all depositors will receive their nominal claim. But as soon as the value V 2
of the bank’s assets falls below the nominal value D2 of the deposits, not all
3 The right to withdraw at any time is an essential prerequisite for the efficiency of the

deposit contract. In accordance with Diamond and Rajan (2000), we therefore exclude the
possibility of suspension of convertibility for the bank in our model; the bank cannot deny
redemption of deposits as long as there are any assets left.

6
depositors can withdraw their full nominal amount anymore. In the latter case,
each depositor faces the problem of choosing between two compound lotteries:
By running, he chooses the lottery LR with payoffs depicted in Figure 1. By
not running, he chooses the lottery LN R with payoffs also depicted in Figure 1:

INSERT FIGURE 1 ABOUT HERE

• When there is a bank run, the first V2 /D2 percent of the depositors in
the waiting queue receive their full nominal deposit D2 /n. Thus, if the
individual depositor runs, the likelihood of arriving early at the queue
(denoted ‘early’) is V2 /D2 . The payoff in this case is Dn2 . If he joins the
queue in a later position (denoted ‘late’), his payoff is 0. When there is
no bank run, the individual depositor is the only to run and he receives
his nominal deposit D2 /n or all of the assets remaining.

• When the individual does not run he either receives 0 if there is a bank
run or Vn2 if there is no bank run because in this case the value of the
remaining assets is distributed equally among the depositors.

For V2 < D2 , the payoffs of the run-strategy LR are higher or equal to those
of the no-run-strategy in all states of the world. Equivalently, the distribution
of the run-strategy LR first-order stochastically dominates the distribution of
the no-run-strategy LN R . Hence, every expected utility maximizing depositor
with positive marginal utility will prefer the run-strategy LR (see e.g. Mas-
Colell et al. (1995)). This leads to an equilibrium situation which is called
information-based bank run.
In run situations, fire sales of assets necessary to pay out the depositors
may create significant liquidation costs (indirect bankruptcy costs) on the other
hand (Diamond and Rajan, 2001): Asset market prices can drastically decline
if big blocks of assets have to be sold immediately. If the bank has to sell all
of its assets at once during a run, we assume that there are liquidation costs of
γV2 , 0 < γ < 1. The fraction γ of firm value lost in case of bank runs creates
a major incentive for the bank to hedge its risk: Averaging 30 percent of the
bank’s assets, these losses are substantial in bank failures as James (1991) found
in his empirical work.

INSERT FIGURE 2 ABOUT HERE

Since there is always a possibility of “sunspot”-bank runs due to unexpected

liquidity shocks, the individual depositor is uncertain whether there will be a
bank run at time t = 2 (Diamond and Dybvig, 1983).4 Because of this uncer-
tainty and the liquidation costs γV2 , VL , the value of the assets below which an
4 We assume that the bank can only raise deposits of the size D ≤ (1 − γ)αP such
1 1
that bank runs at time t = 1 are excluded. This condition can equivalently be written as
D1
αP1
≤ (1 − γ) and interpreted in the following way: Banks can only raise deposits up to the
point where the debt ratio equals the recovery rate in case of a run.

7
information based bank run will be triggered, shifts to
D2
VL = ; (2)
1−γ
D2
for D2 < V2 < 1−γ we now have the payoffs given in Figure 2. They resemble
the payoffs shown in Figure 1 without liquidation costs, but now total value of
the assets is reduced to (1 − γ)V2 instead of V2 in situations of bank runs:
• When there is a bank run, the first V2 (1 − γ)/D2 percent of the deposi-
tors in the waiting queue receive their full nominal deposit D2 /n. If the
individual depositor runs, the likelihood of arriving ‘early’ is V2 (1 − γ)/D2
and the payoff in that case is Dn2 . If he joins the queue in a later position
(denoted’late’), his payoff is 0. When there is no bank run, the individual
depositor is the only one to run and he receives his nominal deposit D2 /n
or all of the assets remaining.
• When the individual does not run, he receives 0 if there is a bank run.
Otherwise he either receives his full nominal amount Dn2 (if D2 ≤ V2 ≤ VL ),
or a fraction Vn2 of the remaining assets, which are distributed equally
among the depositors.
D2
Again, for V2 < 1−γ , the distribution of the run-strategy LR first-order stochas-
tically dominates the distribution of the no-run-strategy LN R , causing an infor-
mation-based bank run equilibrium.
Without this bank run-threat, the payoff function for the bank’s equity at
time t = 2 would be
(
V2 − D 2 , V 2 ≥ D 2 ,
S(V2 , D2 ) ≡
0, 0 ≤ V 2 < D2 .

This is the payoff of an ordinary call option on the firm value with strike D 2 .
However, in the presence of liquidation costs, a bank run will always take place
if V2 < VL . thus the residual payoff to shareholders drops to zero below VL .
Since VL > D2 , the equity payoff changes to
(
V2 − D 2 , V 2 ≥ V L ,
S(V2 , D2 ) ≡ (3)
0, 0 ≤ V 2 < VL ,

as shown in Figure 3.
INSERT FIGURE 3 ABOUT HERE

2.4 The optimization problem

At time t = 1, the bank chooses a hedging position consisting of h units of
the forward contract on the risky asset. As a function of the chosen hedging
position h, the value of the bank’s assets at time t = 1 hence is
V1 (h) = αP1 + hf1 + F V1 , h ∈ R, (4)

8
and the value of the bank’s assets in state U and D respectively at time t = 2
for a given hedging position h is denoted by

Vu (h) ≡ αPu + hfu , h ∈ R, (5)

Vd (h) ≡ αPd + hfd , h ∈ R. (6)

To study the impact of regulatory or other restrictions on the risk management,

we introduce lower and upper bounds5 on the hedging position h,

−α ≤ h ≤ a1 . (7)

The lower bound −α is equivalent to a no net-shortsales-constraint.6

The payoff S to shareholders at liquidation at time t = 2 is a function of
firm value and deposits, S(V2 , D2 ). Since the financial market is arbitrage-free
and complete, the present value of equity at time t = 1 for a given future value
V2 of the assets equals  
S(V2 , D2 )
EQ .
B2
The bank’s management’s goal is to maximize the present value of equity at
time t = 1, by choosing a hedging position h. Let
 
Q S(V2 (h), D2 )
I(h) ≡ E (8)
B2

denote the objective function. I(h) is the value of equity at time t = 1 as a

function of the hedging portfolio h. Then, the bank’s optimization problem at
time t = 1 is
 
S(V2 (h), D2 )
max I(h) = max EQ
−α≤h≤a1 −α≤h≤a1 B2
  (9)
S(αP 2 + hf2 , D2 )
= max EQ ,
−α≤h≤a1 B2
5 Liquidation costs γV are expressed as a fraction of the final firm value V . Thus, by intro-
2 2
ducing the following (merely technical) restriction on the bank’s hedging decision, admitting
only hedging strategies for which the firm value is always positive, we guarantee nonnega-
u
tive liquidation costs. This amounts to the restriction h ∈ [Z u , Z d ] where Z u ≡ −α u−R
d
and Z d ≡ −α d−R . Indeed, these constants follow from solving the inequalities Vu (h) ≥ 0
and Vd (h) ≥ 0 using the definitions (5) and (6) of Vu (h) and Vd (h). V u (h) ≥ 0 holds for
h ∈ [Z u , ∞) and V d (h) ≥ 0 holds for h ∈ (−∞, Z d ], thus nonnegative firm value is the
outcome for hedging strategies in (−∞, Z d ] ∩ [Z u , ∞)). It follows from the definition of Z u
that Z u < −α, thus −α as the lower bound of the set of feasible hedging positions guarantees
nonnegative firm value in state U . Throughout the following we assume that the upper bound
is more restrictive than Z d , a1 < Z d .
6 The net position in the risky asset is restricted to be nonnegative and thus, the cases

where Vu (h) < Vd (h) can be omitted.

9
2.5 Optimal Hedging Strategy
The present value of equity as a function of the hedging position h, I(h), has
the following form, depending on whether for a given hedging position h there
will be a positive payoff to shareholders in both states U and D or only in state
U:

1. For hedging positions h such that the assets’ total value exceeds the bank
run trigger VL in both states U and D, Vu (h) > VL and Vd (h) > VL ,
amounting to a positive payoff to shareholders in both states (we call
these hedging portfolios portfolios of type 1), the objective function is
   
q 1−q
I(h) = αPu + h(Pu − RP1 ) − D2 + αPd + h(Pd − RP1 ) − D2 .
B2 B2

2. For hedging positions h such that the assets’ total value exceeds only
in state U the bank run trigger VL , Vu (h) > VL , but is smaller than
VL in state D, Vd (h) < VL (that is, there is a bank run in state D and
shareholders receive only in state U a positive payoff; we call these hedging
positions portfolios of type 2), the objective function is
 
q
I(h) = αPu + h(Pu − RP1 ) − D2 .
B2

Thus, in order to solve the optimization problem, we need to find conditions

which guarantee the existence of hedging positions h of either type 1 or type 2
for a given market and bank structure.

Lemma 1. The payoff to shareholders is positive in state U for hedging positions

h
VL − αPu
h≥ =: K u . (10)
Pu − RP1
The payoff to shareholders is positive in state D for hedging positions h
VL − αPd
h≤ =: K d . (11)
Pd − RP1
K u is the minimal hedging position for which shareholders receive a positive
payoff in state U. K d is the maximal hedging position for which shareholders
receive a positive payoff in state D. Hence, the portfolios of type 1 are in the
set [K u , K d ]. Therefore, the relationship among the terms K u and K d will
determine whether this set is empty and whether there are hedging positions of
type 1 or only of type 2.

Corollary 1. Hedging positions h of type 1 exist if

VL ≤ αP1 R =: V̄ . (12)

10
 V̄ is the value of the assets of the ‘fully hedged bank’ at time t = 2, i.e. the
value attained if the bank sells forward its whole position α in the risky asset
and the future value becomes certain, Vu (−α) = Vd (−α) = αP1 R = V̄ . Thus, if
the bank run trigger VL is smaller or equal to the forward price V̄ of the bank’s
prior position, there exist hedging positions for which shareholders receive a
positive payoff in both states U and D. It is obvious that the fully hedged
position −α would be such a position. And if the payoff to shareholders with
this forward position is strictly positive, there will be other forward positions
close to −α which also yield a positive payoff to shareholders. Otherwise, if the
forward price V̄ of the bank’s prior position is smaller than the bank run trigger
VL then there are only hedging positions of type 2. Shareholders then receive
a positive payoff only in state U and a zero payoff in state D, Vu (h) ≥ VL and
Vd (h) < VL .
The shape of the objective function is further clarified by the following

Lemma 2. If VL ≥ V̄ , then the inequality

K d ≤ −α ≤ K u (13)

holds. Otherwise, if VL ≤ V̄ , then the inequality

K u ≤ −α ≤ K d (14)

holds.

The intuition for (13) and (14) is as follows:

• If VL > V̄ , then the fully hedged position −α, leading to a value of V̄
in both states of the world, results in a payoff of zero to shareholders,
since the firm value is in this case below the bank run trigger VL . Thus,
the minimal (maximal) hedging position at which shareholders receive a
positive payoff in state U (D), i.e. K u (K d ), would be higher (smaller)
than −α. This corresponds to (13).

• The converse holds if V̄ ≥ VL . Then the fully hedged position −α yields

a positive payoff to shareholders. Then the minimal (maximal) hedging
position at which shareholders receive a positive payoff in state U (D), i.e.
K u (K d ), would be smaller (higher) than −α. This corresponds to (14).

INSERT FIGURE 4 ABOUT HERE

Figure 4 displays the objective function I(h) with the no net short sales
restriction in the three cases where VL > V̄ (Figure 4(a)), VL < V̄ (Figure 4(b))
and VL = V̄ (Figure 4(c)). The bold line is the feasible part of the objective
function.

Proposition 1. 1. If the forward price of the bank’s prior position is less

than the bank run trigger, αP1 R < VL , and if there is a positive payoff

11
 to shareholders in state U at the maximal admissible hedging position,
a1 ≥ K u , then the optimal hedging position is

h∗ = a 1 .

2. Otherwise, if αP1 R > VL , we find the following optimal hedging strategies:

(a) If a1 ≥ K d and a1 > Ju , then h∗ = a1 , where Ju is defined by (15)
below.
(b) If Ju > a1 > K d , then −α ≤ h∗ ≤ K d .
(c) If a1 < K d , then −α ≤ h∗ ≤ a1 .
3. If αP1 R = VL , we find the following optimal hedging strategies:
(a) If a1 < Ju , then h∗ = −α.
(b) If a1 > Ju , then h∗ = a1 .
Ju can be characterized as follows (see Figure 4(b)): The position h =
Ju belongs to the set of hedging positions for which the objective function is
increasing; further, Ju is the position for which the value function equals the
value that is attained on the set where the objective function is constant,
D2 (αPu −D2 )
αP1 − R −q R
Ju ≡ . (15)
q P1 (u−R)
R

The three optimal hedging decisions in Proposition 1 have the following

economic interpretation:
• h = a1 is the strategy of maximal speculation;
• h = −α is the case of complete hedging where the bank sells forward its
whole initial position;
• In the case where K u < K d , the bank is indifferent between the hedging
strategies in the range K u ≤ h ≤ K d .
Part 1 of Proposition 1 covers the case in which the payoff to shareholders would
be zero if the bank hedged completely. It is the case in which the forward price
of the prior position is less than the bank run trigger, V̄ < VL . If a positive
payoff to shareholders in state U is attainable by taking on more risk, a1 ≥ K u ,
we have a “gamble for resurrection”-situation: It is always optimal to take as
much risk as possible, h∗ = a1 . The condition VL > V̄ can equivalently be
written as
D1 /αP1 > (R/RD )(1 − γ), (16)
which says that the initial debt ratio is higher than the recovery rate (in case of
a run) multiplied by the spread between the deposit rate and the riskless interest

12
rate. Hence, for banks with high initial debt ratio and/or high liquidation costs,
it is always optimal to gamble for resurrection.
Parts 2a until 2c of Proposition 1 cover the cases in which shareholders
would still receive a positive payoff if the bank hedged completely, VL < V̄ ,
resp. D1 /αP1 < (R/RD )(1 − γ).

• In 2a, the maximal admissible hedging position a1 yields a higher expected

payoff than the ‘fully hedged’ position −α. Since shareholders could lock
in a sure positive payoff by hedging completely, this is not a “gamble for
resurrection”, although the optimal hedging strategy is the same. Due
to equity’s nonlinear payoff, they can expropriate wealth from depositors
by taking on more risk: The increase of the payoff in state U overcom-
pensates the liquidation costs in state D. This strategy is known as asset
substitution (Jensen and Meckling, 1976). Hence, in the current model,
banks in this situation have loose regulatory or other restrictions (a large
a1 ). They can take that much risk that the bank run threat does not have
a disciplinary effect anymore.

• In 2b, the risk management restriction is so constraining, that at the

maximal admissible position a1 , the gain in expected return does not
outweigh the expected liquidation costs of this portfolio. The expected
payoff to shareholders for this hedging position is smaller than for the
‘fully hedged’ position −α. However, there is no unique optimal hedging
strategy: Shareholders are indifferent with respect to the hedging strate-
D1
gies in the whole range between −α and K d . If the initial debt ratio αP 1
is higher than RdD (1 − γ), then K d < 0 and the optimal hedging strategy
is risk reducing, h∗ < 0. Risk reducing banks in this case are those with a
high initial debt ratio, high asset volatility and/or high liquidation costs.

• In 2c, the maximal admissible hedging position a1 belongs to the portfolios

for which shareholders receive a positive payment in both states U and D.
The expected payoff is the same as the one of the ’fully hedged’ position
−α. In this case, the Modigliani-Miller-result of hedging-irrelevance also
holds ex post, after the determination of the capital structure: Sharehold-
ers are indifferent with respect to all admissible hedging strategies. Banks
in this case are, however, forced towards a safe behavior: The risk man-
agement restrictions prevent asset substitution since they guarantee that
the value of banks’ assets can never fall below the bank run trigger.

In part 3a, the ‘fully hedged’ position h∗ = −α is optimal. Any risk taken
by the bank induces liquidation costs. But the expected return cannot be in-
creased sufficiently such that the shareholders would receive a higher expected
payment at least in one state since Ju > a1 > K d . Banks in this situation
don’t have any risk tolerance. They cannot improve the shareholders’ position
by asset substitution. In our model, only for this special situation, the Froot
and Stein (1998)-result of complete hedging is derived as the unique optimal
hedging strategy.

13
 In part 3b, the regulatory constraint a1 is loose enough to allow the bank to
take on enough risk such that the expected return again outweighs the expected
liquidation costs.
Overall, for a regulatory restricted bank financed with deposits that is sub-
ject to liquidation costs in the event of bank runs, the common interpretation
of equity as a call option does not necessarily apply: Equity value is not always
increased by an increase in asset-risk. Further, higher liquidation costs lead to
an increase of the bank run trigger. This creates larger downside risk for share-
holders that cannot always be outweighed by a higher expected return, because
regulatory restrictions place an upper limit on risk taking.
On the other hand, depending on how much risk taking regulatory or other
restrictions allow, hedging completely as in Froot and Stein (1998) is almost
never the unique optimal hedging strategy: Over a wide range, all hedging
positions can be equally optimal. Risk shifting to depositors is optimal as long
as the higher expected return outweighs the possible downside loss. If risk
management restrictions are set to prevent asset substitution, the value of the
bank’s assets can not fall below the bankrun trigger. The result then coincides
with the Modigliani-Miller-result of hedging irrelevance.

3 Impact of counter party risk constraints

We extend the analysis of the previous section by introducing counter party
restrictions on the attainable forward contract size used for hedging. The for-
ward price RP1 is set such that expected profit from the forward contract is
zero under the risk neutral probability measure Q. Yet, if the bank can default
on the forward contract, the counter party will demand a higher forward price
to get compensated for the additional risk. If we leave the forward price fixed,
the bank will not be able to enter every desired forward position any more. The
counter party restricts the hedging decision by offering only forward contracts
for which the probability of default does not exceed some threshold. In the
current binomial setting, statements on probabilities correspond to conditions
on states U and D:
• Zero probability of default is equivalent to no default in both states U and
D.

• If the probability of default can be positive, then the bank is not allowed
to default either in state U or in state D.

Proposition 2. The bank will not default on the forward contract in state U
for contracts of size h ≥ K u . Further, the bank will not default on the forward
contract in state D for contracts of size h ≤ K d . Under the requirement that
the bank should not default in any state of the world on its obligations from the
forward contract, it will not be able to enter a forward contract unless K u ≤ K d .
It will only be offered contracts h such that K u ≤ h ≤ K d .

14
The question when the bank is offered both long and short or only long or only
short positions is answered by the following
Lemma 3. If the bank is not allowed to default in state U (state D), it will be
offered short (long) positions if and only if VL < αP1 u (VL < αP1 d).
Hence, the restriction not to default in state D may prevent the bank to enter
long positions, namely if D1 /αP1 > (1 − γ)(d/RD ). These banks either have
a high debt ratio, high liquidation costs and/or a high asset-volatility. They
would face a bank run in state D without hedging and the costs would be borne
by the counter party. On the other hand, the restriction not to default in state
U may prevent the bank to enter short positions (if D1 /αP1 > (1 − γ)(u/RD )).
The debt ratio, the liquidation costs and/or the asset-volatility of these banks
is that high that they would face a bank run already in the ‘good’ state of the
world U and the counter party enforces the asset substitution in this case. The
most important type of restriction7 is the one which does not allow default in
any state. In the case where VL > V̄ ,8 the bank cannot enter a forward contract.
With its combination of deposits, initial position and liquidation costs, it will
not be offered forward contracts due to default risk. Thus, the gamble for
resurrection is not possible any more. When VL ≤ V̄ , the bank is prevented
from taking on any risk which would trigger a bank-run. It can only enter
positions in the forward in the range −α ≤ h ≤ K d . Thus, the bank will always
have the possibility to reduce risk by entering short positions. In the subcase
where VL > αP1 d, it will not be able to obtain long positions (Lemma 3). That
is, when the bank’s prior position is sufficient to prevent a bank run only in
state U, but not in state D, the bank will only be offered contracts that reduce
the risk sufficiently to ensure that there will be no bank run in state D. Without
hedging, the bank would face a bank run in state D. But with the positive cash
flow −(Pd − RP1 ) from the short position in the forward contract in state D, the
bank’s assets are sufficient to prevent a run in state D. The following Lemma
tells when the bank will choose to hedge.
Lemma 4. In the case where VL ≤ V̄ and αP1 d < VL , the bank will choose to
hedge if D2 ≤ αP1 d < VL ; if αP1 d < D2 , then it is optimal for the bank not to
7 The constraint that the bank should not default in state U but is allowed to default in

state D is only meaningful if the risk neutral probability of state D is very low. The forward
contract price is then approximately not affected by the additional default risk. The following
results then apply: If VL > V̄ , the bank can obtain only positions h ≥ K u , since it would
default in state U on all other positions. Therefore, the bank can still follow a strategy of
asset substitution by holding long positions in the forward contract. If even V L ≥ αP1 u,
that is, if the bank faced a bank run without hedging in state U , it would only be offered
long positions to hedge and thus be forced to ‘gamble for resurrection’: For sufficiently large
hedging positions, the value of the bank’s asset is above the bank run trigger in state U
(whereas in state D, the bank will default on its obligation from the forward contract). In the
case where VL ≤ V̄ , the constraint that the bank is not allowed to default in state U is not
binding: It will be offered any contracts of size h ≥ K u but the lower bound on its hedging
position h already is −α where −α ≤ K u ((14) in Lemma 2).

8 V > V̄ implies K d < K u , and if follows from Proposition 2 that the bank will not be
L
offered forward contracts.

15
hedge.
The reason for this hedging-strategy is the following: By hedging when
D2 < αP1 d < VL , the bank can preserve asset value in the down state D, that
otherwise would be completely lost for the shareholders as liquidation costs.
On the other hand, if αP1 d < D2 , all the remaining asset value up to D2 goes
to the depositors anyway. If the bank hedges, it thus sacrifices some payoff to
shareholders in state U in exchange for securing payoffs for depositors in state
D. The bank can do better for the shareholders by not hedging at all, that is,
by keeping the higher expected return of the unhedged position while letting
the depositors bear the downside loss in state D.
Overall, the introduction of counterparty-restrictions mitigates risk taking
incentives for a bank, since it is not possible to gamble for resurrection anymore.

4 Conclusion
We have presented a one-period model in which we analyze the bank’s risk
management decision. The bank is regulatory restricted, financed by deposits
and is subject to liquidation costs in the event of a bank run.
We find that the common interpretation of equity as an ordinary call op-
tion does not apply: Equity value is not always increased by increasing the
asset’s volatility, since this also raises the likelihood of a bank run. Whenever
the expected costs of such a run for shareholders can not be outweighed by an
increase of the expected return (because regulatory restrictions limit the maxi-
mum achievable risk), it is not optimal to take as much risk as possible. In these
cases, safe banks with low debt ratios and asset volatility can still augment their
risk exposure to the point where downside loss comes into play. However, for
banks with a high debt ratio and a high asset volatility, risk reduction is the
optimal strategy.
This deterrence of asset substitution however vanishes in the absence of reg-
ulatory constraints or with a complete deposit insurance (Calomiris and Kahn,
1991): Without the possible downside loss, the equity payoff would be that of
an ordinary call option and it would always be optimal for the bank to take as
much risk as possible. Also, without regulatory restrictions, the possible down-
side loss could always be outweighed by higher expected return through higher
risk-exposures.
On the other hand, depending on how much risk taking regulatory or other
restrictions allow, it may not be optimal for the bank to hedge completely as in
Froot and Stein (1998): Because equity features limited liability, risk shifting to
depositors is still preferred as long as the higher expected return outweighs the
possible downside loss. The less restrictive regulatory restrictions are, the more
relevant becomes this strategy of asset substitution. Without any restrictions
of regulators or counter parties, asset substitution would always be the optimal
strategy.
Further, there is one constellation for which the hedging decision is shown to
be irrelevant, which coincides with the result of the Modigliani-Miller-theorem.

16
This, however, is only a special situation, where the risk management restric-
tions, the size of the liquidation costs in case of a bank run and the initial debt
ratio are all set such that risk shifting to depositors is impossible and no bank
run takes place.
Among the open questions remains the analysis of the hedging decision in
a multiperiod setting. Bauer and Ryser (2002) have looked at the effect that
the bank’s franchise value of deposits then has. It gives an incentive to reduce
risk taking since the whole stream of future income from deposit services would
be lost in a run situation. Furthermore, it would be interesting to analyze the
hedging decision in the presence of a non-tradeable proprietary bank asset that
generates an extra rent as in Diamond and Rajan (2000). The market com-
pleteness breaks down in this case and the determination of a unique objective
function for the bank is not trivial anymore.

Proofs
Proof of Lemma 1. Using the definition (5) of Vu (h) we find that for a given h
VL − αPu
Vu (h) > VL ⇔ h > = K u;
Pu − RP1
similarly for a given h
VL − αPd
Vd (h) > VL ⇔ h < = Kd.
Pd − RP1

Proof of Corollary 1. From Lemma 1 we know that Vu (h) > VL for h ∈ [K u , ∞) and
Vd (h) > VL for h ∈ (∞, K d ]. Hence, hedging positions h of type 1 (that is h for which
both Vu (h) > VL and Vd (h) > VL ) are h ∈ [K u , K d ]; this interval is not empty if

Ku ≤ Kd. (17)

Using the definitions (10) and (11) of K u and K d this can be written equivalently as
VL −αPu −αPd
Pu −RP1
≤ VPLd −RP 1
. Solving for VL yields VL ≤ αP1 R = V̄ .

Proof of Lemma 2. Consider first the case VL > V̄ . From the definition (12) follows
VL ≥ V̄ = αP1 R. Subtracting αPu yields VL − αPu ≥ −α(Pu − P1 R), dividing by
(Pu − P1 R) yields K u ≥ −α. The inequalities for K d and for the case where VL ≤ V̄
follow in the same way.

The following Lemma will be useful to prove Proposition 1.

Lemma 5. The sets of candidates for the optimal hedging strategy h ∗ are {a1 }, {−α}
and (−∞, K d ] ∩ [K u , ∞). The values of the objective function evaluated at these can-

17
didate points are
  
q
αP + a (P − RP ) − D , K d ≤ K u ≤ a1 or K u ≤ K d < a1 ,

u 1 u 1 2

 B2

I(a1 ) = αP − D2 , K u ≤ a1 ≤ K d ,
 1 B2

K d < a1 < K u ,

0,
(18)
(
D2 u d
αP1 − B2
, K <K ,
I(−α) = (19)
0, K d < K u,
(
D2
αP1 − B2
, Ku ≤ h ≤ K d,
I(h) = (20)
0, K d < h < K u.

Proof of Lemma 5. For convenient notation, we write the constraints (7) −α ≤ h ≤ a 1

in the form
Ah ≤ a (21)

0
0
0
where A ≡ A1 A2 = 1 −1 and a ≡ a1 −α .
We consider first the case VL ≤ V̄ = αP1 R. From VL < V̄ = αP1 R and (17)
follows that K u ≤ K d . Hence, for any h ∈ [K u , K d ] holds that Vu (h) ≥ VL and
Vd (h) ≥ VL . On this interval the Lagrangian is as follows:
   
q 1−q
L(h, µ1 , µ2 ) = αPu + h(Pu − RP1 ) − D2 + αPd + h(Pd − RP1 ) − D2
B2 B2
2 2
X hP1 D2 X
+ µk (Ak h − ak ) = (α + h)P1 − − + µk (Ak h − ak ),
B1 B2
k=1 k=1

yielding the first order conditions

2
∂L P1 X
= P1 − + µk Ak = 0, (22)
∂h B1
k=1
∂L
= Ak h − ak = 0, k = 1, 2.
∂µk
We have the following candidate points for an optimum:
P1
1. µ1 = µ2 = 0: (22) reduces to the condition P1 − B 1
= 0 where the last equality is
due to B1 = 1. Thus, in this case any h such that Vu (h) ≥ VL and Vd (h) ≥ VL
is optimal, that is the set of candidates is [K u , K d ]. From B1 = 1 follows
I(h) = αP1 − DB2
2
, Ku ≤ h ≤ K d.
2. If µ1 6= 0, µ2 = 0 then h∗ = a1 . Then, if K u < K d < a1 I(a1 ) = Bq2 [αPu +
a1 (Pu − RP1 ) − D2 ] since the payoff to shareholders is zero in state D for h = a1
due to the fact that K d < a1 . If K u ≤ a1 ≤ K d
   
q 1−q D2
I(a1 ) = αPu +a1 (Pu −RP1 )−D2 + αPd +a1 (Pd −RP1 )−D2 = αP1 − .
B2 B2 B2

6 0 then h∗ = −α. Then, since K u ≤ −α ≤ K d

3. If µ1 = 0, µ2 =
   
q 1−q D2
I(−α) = αPu −α(Pu −RP1 )−D2 + αPd −α(Pd −RP1 )−D2 = αP1 − .
B2 B2 B2

18
 Consider now the case VL > αP1 R when there are only h such that Vu (h) > VL
and Vd (h) < VL holds for fixed h. The Lagrangian is then
  X2
q
L(h, µ1 , µ2 ) = αPu + h(Pu − RP1 ) − D2 + µk (Ak h − ak )
B2
k=1

yielding the first order conditions

  X2
∂L q
= Pu − RP1 + µk A k = 0 (23)
∂h B2
k=1

We have the following candidate points for an optimum:

Pu
1. µ1 = µ2 = 0 could hold only if P1
= R which was excluded in (1).
2. µ1 6= 0, µ2 = 0 and h = a1 . Then, if K d < K u ≤ a1 I(a1 ) = Bq2 [αPu + a1 (Pu −
∗

RP1 ) − D2 ] since the payoff to shareholders is zero in state D due to the fact
that K d < a1 .
If K d < a1 < K u we have I(a1 ) = 0 since the payoff to equity holders is both
zero in state D (from K d < a1 ) and in state U (from a1 < K u ).
3. µ1 = 0, µ2 6= 0 and h∗ = −α. Then it follows that I(−α) = 0 since from (13)
follows that K d < −α and hence at the position −α equity holders receive a
zero payoff in state D.
As for the last equality, it is obvious that for h > K d , the pay off to shareholders
is zero in state D and for h < K u the pay off to shareholders is zero in state U,
hence for K d < h < K u follows I(h) = 0.

Proof of Proposition 1. We consider first part 1 of the proposition, that is, the case
where VL > αP1 R and a1 ≥ K u . As this is the case when K d < K u . It follows
from (20) that the objective function equals zero for all h ∈ (K d , K u ) including −α
(since from Lemma 2 K d < −α < K u ). Definition (10) of K u yields for a1 ≥ K u
−αPu
a1 ≥ VPLu −RP 1
or αPu + a1 (Pu − RP1 ) ≥ VL . From (18) I(a1 ) = Bq2 [αPu + a1 (Pu −
RP1 ) − D2 ] > 0 since VL > D2 , hence a1 is the optimum.
We now turn to parts 2 and 3 of the proposition; both cases are covered by the
inequality VL ≤ αP1 R or equivalently K u ≤ K d , 2 being the case of strict inequality
and 3 the case of equality. Consider first 2a and 3b respectively. From (20) follows
that for h ∈ [K u , K d ] (and hence also for h = −α) I(h) = αP1 − D 2
B2
. From Ju < a1
D2 q ∗
follows αP1 − B2 < B2 [αPu + a1 (Pu − RP1 ) − D2 ] = I(a1 ), hence h = a1 .
Similarly follows for 2b and 3a when Ju > a1 > K d that αP1 − D 2
B2
> Bq2 [αPu +
a1 (Pu − RP1 ) − D2 ] = I(a1 ), hence h∗ ∈ [−α, K d ].
In part 2c, feasible portfolios h are h ∈ [−α, a1 ] ⊂ (K u , K d ). Hence, from (20)
all feasible portfolios have the same value of the objective function, I(h) = αP1 −
D2
B2
, K u < h < K d , h ∈ (K u , K d ).

Proof of Proposition 2. If the bank should not default in state U on the forward con-
tract, then the bank’s assets net of the value of debt in state U must be positive,
αPu + hfu − D2 ≥ 0. Solving for h yields the required inequality h ≥ K u . In the
case where the bank should not default in state D on the forward contract, the bank’s
assets net of the value of debt in state D must be positive, αPd + hfd − D2 ≥ 0. Solving

19
for h yields again the required inequality h ≤ K d . If the counter party of the forward
contract requires that the bank does not default in any state, then h needs to be in the
intersection of the intervals [K u , ∞) and (−∞, K d ]. If K d < K u then this intersection
is empty, hence the bank will not be able to enter a forward contract. If K d > K u
then the intersection is exactly [K u , K d ].

Proof of Lemma 3. By (1) Pu = P1 u > RP1 holds and thus P1 u − RP1 > 0. Therefore
−αP1 u
K u = VPL1 u−RP 1
< 0 ⇔ VL − αP1 u < 0 ⇔ VL < αP1 u. If the bank is not allowed
to default in state U, it will by Proposition 2 be constrained to hedging strategies
h ≥ K u . Thus, short positions (i.e. h < 0) will only be available if K u < 0, hence the
statement follows. The inequality for K d follows by the same arguments.

Proof of Lemma 4. In the case where VL ≤ V̄ and αP1 d < VL , it follows from Lemma
2 and Lemma 3 that K u ≤ K d < 0. Since VL ≤ V̄ , there is a positive payoff to
shareholders without hedging and the value of the objective function for the decision
not to hedge is I(0) = Bq2 (αP1 u − D2 ). The bank is offered only hedging positions
in the interval [K u , K d ] for which (by (20) of Lemma 5) the value of the objective
function is αP1 − DB2
2
which can be written as Bq2 αP1 u + 1−q
B2
αP1 d − qD
B2
2
− (1−q)D
B2
2
.
For h ∈ [K u , K d ], I(h) > I(0) ⇔ Bq2 αP1 u+ 1−q
B2
αP1 d− qD
B2
2
− (1−q)D
B2
2
> Bq2 (αP1 u−D2 )
which is, due to the fact that both (1 − q) > 0 and B2 > 0, equivalent to αP1 d ≥ D2 .
The inequality αP1 d < VL follows from the fact that we look at the case where K d < 0
and Lemma 3.

References
Allen, F., Santomero, A. M., 1998. The theory of financial intermediation.
Journal of Banking & Finance 21, 1461 – 1485.
Allen, F., Santomero, A. M., 2001. What do financial intermediaries do? Journal
of Banking & Finance 25, 271 – 294.
Baltensperger, E., Milde, H., 1987. Theorie des Bankverhaltens. Springer Ver-
lag, Berlin, Heidelberg.
Bauer, W., Ryser, M., 2002. Bank risk management and the franchise value of
deposits. Working paper.
Bhattacharya, S., Thakor, A. V., 1993. Contemporary banking theory. Journal
of Financial Intermediation 3, 2– 50.
Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities.
Journal of Political Economy 81, 637 – 659.
Boot, A. W. A., 2000. Relationship banking: What do we know? Journal of
Financial Intermediation 9, 7 – 25.
Broll, U., Jaenicke, J., 2000. Bankrisiko, Zinsmargen und flexibles Futures-
Hedging. Schweizerische Zeitschrift für Volkswirtschaft und Statistik 136(2),
147 –160.

20
Calomiris, W. C., Kahn, C. M., 1991. The role of demandable debt in structuring
optimal banking arrangements. American Economic Review 81(3), 497 – 513.
DeMarzo, P., Duffie, D., 1995. Corporate incentives for hedging and hedge
accounting. Review of Financial Studies 8, 743 – 772.
Diamond, D. W., 1984. Financial intermediation and delegated monitoring.
Review of Economic Studies 51, 393 – 414.
Diamond, D. W., 1996. Financial intermediation as delegated monitoring: A
simple expample. Federal Reserve Bank of Richmond Economic Quarterly
82(3), 51 – 66.
Diamond, D. W., Dybvig, P. H., 1983. Bank runs, deposit insurance and liq-
uidity. Journal of Political Economy 91, 401 – 419.
Diamond, D. W., Rajan, R. G., 2000. A theory of bank capital. Journal of
Finance LV(6), 2431 – 2465.
Diamond, D. W., Rajan, R. G., 2001. Liquidity risk, liquidity creation and
financial fragility: A theory of banking. Journal of Political Economy 109(2),
287 – 327.
Eichberger, J., Harper, I. R., 1997. Financial Economics. Oxford Press, New
York.
Freixas, X., Rochet, J. C., 1998. Microeconomics of banking. MIT Press,
Cambridge, Massachusetts.
Froot, K., Stein, J., 1998. Risk management, capital budgeting and capital
structure policy for financial institutions: An integrated approach. Journal
of Financial Economics 47, 55 – 82.
Froot, K. A., Scharfstein, D. S., Stein, J. C., 1993. Risk management: Co-
ordinating corporate investment and financing policies. Journal of Finance
XLVIII(5), 1629 – 1658.
Jacklin, C. J., Bhattacharya, S., 1988. Distinguishing panics and information-
based bank runs: Welfare and policy implications. Journal of Political Econ-
omy 96(3), 568 – 592.
James, C., 1991. The losses realized in bank failures. Journal of Finance 46(4),
1223–1242.
Jensen, M. C., Meckling, W., 1976. Theory of the firm: managerial behavior,
agency costs and ownership structure. Journal of Financial Economics 3, 305
– 360.
Kashyap, A. K., Rajan, R., Stein, J. C., 2002. Banks as liquidity providers:
An explanation for the coexistence of lending and deposit-taking. Journal of
Finance 57(1), 33 – 73.

21
Leland, H. E., 1998. Agency costs, risk management and capital structure.
Journal of Finance 53, 1213 – 1243.

Mas-Colell, A., Whinston, M. D., Green, J. R., 1995. Microeconomic Theory.

Oxford University Press.

Merton, R. C., 1995. Financial innovation and the management and regulation
of financial institutions. Journal of Banking and Finance 19, 461 – 481.

Modigliani, F., Miller, M. H., 1959. The cost of capital, corporation finance,
and the theory of investment. American Economic Review 48, 261 – 297.

Mozumdar, A., 2001. Corporate hedging and speculative incentives: Implica-

tions for swap market default risk. Journal of Financial and Quantitative
Analysis 36(2), 221 – 250.

Schrand, C., Unal, H., 1998. Hedging and coordinated risk management: Evi-
dence from thrift conversions. Journal of Finance LIII(3), 979 – 1013.

Shapiro, A., Titman, S., 1986. An Integrated Approach to Corporate Risk

Management, in: J. Stern and D. Chew (Eds.): The Revolution in Corporate
Finance. Basil Blackwell Inc., Cambridge, Massachusetts.
Shrieves, R. E., Dahl, D., 1992. The relationship between risk and capital in
commercial banks. Journal of Banking & Finance 16(2), 439 – 457.

Smith, C. W., Stulz, R. M., 1985. The determinants of firms’ hedging policies.
Journal of Financial and Quantitative Analysis 20(4), 391 – 405.

Stulz, R. M., 1996. Rethinking risk management. Journal of Applied Corporate

Finance 9(3), 8 – 24.

22
 Early D2
n

0
0
Bank run Late Bank run

LR LN R

No bank run No bank run

¡ D2 ¢ V2
min n , V2 n

Figure 1: Payoffs to a depositor in absence of liquidation costs

23
 Early D2
n

0
0
Bank run Late Bank run

LR LN R

No bank run No bank run

¡ D2 ¡ D2
, Vn2
¢ ¢
min n , V2 min n

Figure 2: Payoffs to a depositor in the presence of liquidation costs

24
S(V2 , D2 )

0 D2 VL V2

Figure 3: Payoff function of equity

25
 I(h)

−α Ku h

(a) VL > V̄ (K d < K u )

I(h)

(1 − q)S(Vd (h), D2 ) qS(Vu (h), D2 )

K u −α Kd Ju h

(b) VL < V̄ (K u < K d )

I(h)

(1 − q)S(Vd (h), D2 ) qS(Vu (h), D2 )

K u = K d = −α Ju h

(c) VL = V̄ (K u = K d )

Figure 4: Three types of objective functions with no net short sales restrictions

26