Bank runs

Mrinal Mishra, Alex Osberghaus

Spring semester 2023

1 Narrow banking, suspension of convertibility, and deposit in-

surance
1. Consider the simplified version of Diamond and Dybvig (1984) of the lecture. There is one good, three
dates, and a continuum of agents endowed with one good. With probability πi (i = 1, 2, with π1 +π2 = 1),
ex-ante identical consumers need to consume at date t = i. Ex-ante, all agents have the same utility:
U = π1 u(C1 )+π2 u(C2 ). There exists a storage technology that yields 1 in the next period and an illiquid
investment technology that yields R > 1 at t = 2 upon investment at t = 0. If liquidated prematurely,
at date t = 1, the investment technology only returns l ≤ 1.
Diamond and Dybvig (1984) show that the fractional reserve system is prone to bank runs. If a patient
consumer anticipates that all other patient consumers want to withdraw at t = 1, it becomes optimal for
her to also withdraw at t = 1 (this is a Nash Equilibrium). A remedy might be narrow banking, a concept
that requires banks to be able to fulfill their contractual obligations under any possible circumstance.
There are three different interpretations of narrow banking:

1. A bank with enough liquidity to guarantee repayment to all depositors even in case of a bank run.
2. A bank that obtains enough liquidity after liquidation of its long-run technology to face a bank run.
3. A bank that obtains enough liquidity after the securitization of its long-run technology to cope with
a bank run.

Two other possible solutions to bank runs are the following. (i) Suspension of convertibility, i.e. the bank
announces that it will only serve π1 C1∗ withdrawals at date t = 1. After that threshold, convertability is
suspended. (ii) Deposit insurance.

Consider the following problems:

1. Set up the optimization problem (utility maximization under respective constraints) for each of
the interpretations of narrow banking. How do solutions to the optimization problems relate to
the solution under autarky, the market solution1 , and the solution under fractional reserve banking
that we saw in class?
2. Briefly discuss the efficiency of suspension of convertibility and deposit insurance as remedies for
bank runs. What are the potential downsides?

1 Suppose a bond market is opened at t = 1, whereby p units of good at t = 1 are exchanged against the promise to receive

one unit of good at t = 2

1
Solution:
1. Let’s remember the problems under autarky, the market solution, and under optimality (fractional
reserve banking):
(i) Autarky

max π1 u(C1 ) + π2 u(C2 )

I,C1 ,C2

s.t. C1 ≤ 1 − I + lI and C2 ≤ 1 − I + IR = 1 + I(R − 1)

(ii) Market Solution

max π1 u(C1 ) + π2 u(C2 )

I,C1 ,C2
1−I
s.t. C1 ≤ 1 − I + pRI and C2 ≤ + RI
p

(iii) Optimality

max π1 u(C1 ) + π2 u(C2 )

I,C1 ,C2

s.t. π1 C1 ≤ 1 − I and π2 C2 ≤ RI

In the first case, we require the bank to invest such that it can always serve all withdrawals ⇒
reserve ratio of 100 percent. The maximum amount withdrawn is C1 ⇒ Deposit contract must
satisfy C1 ≤ 1 − I and C2 ≤ RI.
The maximization problem is therefore

max U = π1 u(C1 ) + π2 u(C2 )

I,C1 ,C2

s.t. C1 ≤ 1 − I, C2 ≤ RI

The solution to this is even dominated by the solution under autarky.

In the second case, the bank can liquidate its long-run technology at t = 1 to face a bank run.
Maximization problem is therefore

max U = π1 u(C1 ) + π2 u(C2 ) (1)

I,C1 ,C2

s.t. C1 ≤ 1 − I + lI, C2 ≤ 1 − I + RI (2)

This coincides with the maximization problem under autarky.

Page 2
In the third case, the bank can securitize its long-term technology and sell it at price p at date t = 1
in case of a bank run. But this is just like selling a claim to receive RI at t = 2. This is therefore
equivalent to a bond and therefore the market solution.

Suspension of convertibility:
The bank announces that it serves only π1 C1∗ withdrawals at date t = 1, after that, convertibility is
suspended. Patient consumers know that the bank will be able to satisfy its engagements at date
t = 2.
This gets rid of a bank run but would be difficult to implement:

• knowing π perfectly is difficult

• legal challenges

Deposit insurance
Even if the bank cannot fulfill its obligations, the difference between the deposit value and liquidation
value of banks’ assets is covered by liquidity insurance. Financed by insurance premiums or taxes.
This gets rid of a bank run. Downsides:

• moral hazard
• inattentiveness of depositors to banks’ riskiness
• how much is covered?

Page 3
2 Bank runs and moral hazard
1. Consider a Diamond-Dybvig economy with a unique good and three dates, where bank managers have a
choice of the technology they implement. This choice is unobservable and consists in investing one unit
in either project G or B, where project G yields G with probability pG and zero otherwise, and project
B yields B with probability pB and zero otherwise, where G < B, and pG G > pB B. A continuum of
agents is endowed with one unit at time t = 0. Of these agents, a nonrandom proportion π1 will prefer
to consume at time t = 1, and the complementary proportion π2 will prefer to consume at time t = 2.
The agents’ utility function is U (C1 ) for impatient consumers, and ρU (C2 ) for patient consumers; so
that the ex ante expected utility is π1 U (C1 ) + π2 ρU (C2 ).
If there are bank runs, they occur with probability α.
1. Assuming that the risk-neutral bank manager brings in equity, and the other agents have deposit
contracts, compute under what conditions the G allocation is obtained. Interpret this condition in
terms of regulation.
2. In what follows, we restrict our attention only to the particular case of risk neutral depositors,
U (C) = C. What is the optimal contract? What are the manager’s incentives to implement G?
Do they depend upon α? Could we propose a better contract by defining an equity economy?

Solution:
1. Let C̄2 be the consumption when the project is successful. The manager is risk-neutral and
therefore maximizes her equity value (value of the bank). The equity value is the expected return
minus the consumption of the depositors, i.e. it is pG (IG − π2 C̄2 ) if she invests into the project
G and pB (IB − π2 C̄2 ) if she invests into project B. For her to implement the good project, we
therefore need

pG (IG − π2 C̄2 ) ≥ pB (IB − π2 C̄2 ) (3)

This implies an upper limit on C̄2 :

I(pG G − pB B)
C̄2 ≤ (4)
π2 (pG − pB )

⇒ The payment to the depositors cannot be too generous in case of success. The manager needs
”skin in the game” ⇒ Capital requirements.

2. The maximization problem becomes

max U = π1 C1 + π2 C̄2 (5)

I,C1 ,C2

s.t. π1 C1 ≤ 1 − I, π2 C̄2 ≤ RI (6)

and pG (IG − π2 C̄2 ) ≥ pB (IB − π2 C̄2 ) (7)

Rearrange 6 and observe that the utility is increasing in I provided that ρpG > 1. The incentives
have not changed except now I = 1. There are no bank runs.
An equity contract could replicate the deposit contract but cannot do better.

Page 4
3 Bank runs
1. Consider an economy with a unique good and three dates, with a storage technology that yields a zero
net interest and a standard long-run technology that yields R units with certainty at time t = 2 but
yields only L < 1 if prematurely liquidated at time t = 1. Both technologies are available to any agent.
A continuum of agents is endowed with one unit at time t = 0. Of these agents, a nonrandom proportion
π1 will prefer to consume at time t = 1, and
√ the complementary proportion π√ 2 will prefer to consume at
time t = 2. The agents’ utility function is√ C1 for impatient
√ consumers and ρ C2 for patient consumers,
so that the ex ante expected utility is π1 C1 + π2 ρ C2 . Assume first that ρR > 1.
1. Compute the first-order condition that fully characterizes the optimal allocation. Compare it with
the market allocation that is characterized by C1 = 1 and C2 = R.
2. Consider a banking contract where a depositor’s type is private information. Are bank runs possible?
If so, for what parameter values?
3. Is the optimal contract implementable within an equity economy, where each agent has a share of
a firm that distributes dividends, and a market for ex-dividend shares opens at time t = 1?
Assume now that ρR < 1.

4. What would be the optimal banking contract? Are bank runs possible? if so, for what parameter
values?
5. Is the optimal contract implementable within an equity economy?

Solution:
1. Max. problem:

p p
max U = π1 C1 + π2 ρ C2 (8)
I,C1 ,C2

s.t. π1 C1 = 1 − I, π2 C2 = RI (9)

Solve constraints for C1 and C2 , respectively. Substitute into maximization problem and take FOC.
Rearrange and find that the FOC is

C2 = R2 ρ2 C1 (10)

To find the optimal allocation, solve the FOC for the optimal investment and substitute into con-
straints of the maximization problem to find C1 and C2 , respectively. You will find that

1
C1∗ = (11)
π1 + ρ2 Rπ2
ρ2 R 2
C2∗ = (12)
π1 + ρ2 Rπ2

We observe that C1 < C2 since ρR > 1 by assumption. Further, C1 > 1 if ρ2 R < 1 (to get this
result, remember that π1 + π2 = 1).

Page 5
2. Since C2 > C1 , we only look at the inefficient Nash Equilibrium in which it is best to withdraw
if every agent withdraws. This happens if

C1 > π C + (1 − π1 )L (13)
| 1{z }1 | {z }
reserves kept for impatient liquidation value

Rearrange, plug in for C1 , let L̄ be the threshold above which bank runs are possible and find that

1
> L̄ (14)
π1 + ρ2 Rπ2

E.g. if L = 1 and ρ2 R > 1, bank runs are never possible.

3. Let d be the dividend and p the ex-dividend share price at t = 1. Early consumers sell shares
and consume at t = 1 while late consumers buy shares and consume at t = 2. Therefore,

C1 = d
|{z} + p∗1 (15)
|{z}
consume dividend sell 1 share at price p
d R(1 − d)
C2 = R( 1−d )+ R(1 − d) = (d + p) (16)
| {z } p | {z } p
amount invested |{z} ROI
shares that d can buy

Solve for d and p to have allocation as a function of parameters and compare allocation with efficient
one.
Note that under efficiency

C2 ρ2 R 2 R(1 − d)
= 2
(π1 + ρ2 Rπ2 ) = ρ2 R2 = (17)
C1 π1 + ρ Rπ2 p

Note also that in equilibrium, supply of shares has to equal demand of shares, or π1 = π2 dp . Solve
for d and substitute into 17. Solve the resulting equation for d to find that

π2
p= (18)
π1 + ρ2 Rπ2

Substitute this into π1 = π2 dp and find that

π1
d= (19)
π1 + ρ2 Rπ2

Finally, substitute this into the expressions for C1 and C2 from the start to find that efficiency is
achieved.

Page 6
4. Remember that

1
C1∗ = (20)
π1 + ρ2 Rπ2
ρ2 R 2
C2∗ = (21)
π1 + ρ2 Rπ2

ρR < 1 implies that ρ2 R2 < 1 which is why C1∗ > C2∗ . However, if the banking contract was such
that C1 > C2 , it would be optimal for the patient consumer to withdraw her money at t = 1 after
having learned her type. Therefore, the optimal banking contract is such that C1 = C2 .
Are bank runs possible? Remember the condition for bank runs, rearrange to find that

1 1
= > 1 > L, (22)
π1 + ρ2 Rπ2 1 + π2 (ρ2 R − 1)
| {z }
<0

where the last inequality follows by assumption. Bank runs are therefore always possible.

5. It will not reach the first best as demand for ex-dividend shares will be zero.

Page 7