LH Advanced Macroeconomics

(07 33109)

Part 4: Financial Markets

& Financial Crises

Dr Ceri Davies
[email protected]

Introduction
• We have considered how households save – as the
difference between income and consumption in any given
period. We have also considered how firms invest – they
set aside current resources for future use/benefits.

• Financial markets are where savings and investment

meet; e.g. the banking sector as a financial intermediary.

• The ‘price signals’ sent by financial markets can influence

household/firm choices about how to allocate their
scarce resources, including how much risk to take. The
price of risky assets is determined by the interaction of
demand and supply.

1
 Financial Markets & Macroeconomics

• In equilibrium, firms and households behave optimally,

taking prices as given; prices are such that markets
clear.

• Financial markets can provide another source of shocks

and/or contribute to the propagation of other, non-
financial, shocks. ‘Frictions’ in financial markets are
typically required to magnify shocks to the economy.

• As macroeconomists, we are primarily interested in real

spill-overs; Financial crises can sometimes cause
significant disruption to the real economy.

Source: Oregon Office of Economic Analysis (2014)

Spain: Betrán and Pons (2017)

Nordic countries: Honkapojha (2009)
Japan: Hirakata et al. (2016)

2
 Overview

• Baseline:
– A Model of Perfectly-Functioning Financial Markets

• With imperfections/frictions:
– Agency Costs and the Financial Accelerator

• Financial Crises:
– The Diamond-Dybvig Model
– Financial contagion

Baseline Model
Assumptions: two-period model; a representative household
has an endowment E of the economy’s single good in period 1
and no endowment in period 2; we abstract from labour
market (or leisure) decisions and focus on financial markets.

The household maximises the (expected) value of its lifetime

utility:

Consumption 𝛽>0 Discount factor

𝑈′(•) > 0
𝑉 = 𝑈 𝐶1 + 𝛽𝑈 𝐶2
𝑈 ′′ • < 0

Output in period 2 is derived from financial investments

undertaken in period 1; households are price-takers.

3
 Investment Projects
• The are N possible ‘investment projects’ undertaken by firms;
there are S possible ‘states of the world’ in period 2; firms are
price-takers too.

• Suppose that a quantity 𝐾𝑖 (≥0) of period-1 output is devoted

to investment project 𝑖; assume that in period 2, for state of
the world s, this project (𝑖) generates a future output payoff:

𝐾𝑖 𝑅𝑖𝑠 𝑅𝑖𝑠 ≥ 0 for all 𝑖 and 𝑠

• Importantly, there is uncertainty about the state of the

world in period 2; we denote the probability of state s
occurring as πs, where each probability must be ≥0 and the
sum of all probabilities, across all possible states, is 1.

Equilibrium
• It is conceptually straightforward to compute the equilibrium
in such a model with complete markets and no imperfections.
• We can describe the outcome of the model in terms of claims
on period-2 output in the various possible future states of the
world. This is known as an Arrow-Debreu-McKenzie
contingent claim (see next slide).
• Notation: qs denotes the price of a claim on one unit of
period-2 output in state s, stated in units of period-1 output
(i.e. in relative terms, with period 1 output as the numeraire).
• An equilibrium consists of: a set of prices over future states
{qs}, investment decisions {𝐾𝑖 } and consumption choices 𝐶1
and {𝐶2𝑠 }.

4
 Aside: State-Contingent Claims
• We use these as a helpful modelling device; we assume
that we can put a price on contracts which pay-out in
different possible future states of the world; we can cover
all cases, ‘complete markets’ (Arrow-Debreu-McKenzie).

• As a result of the assumptions adopted in our model,

future consumption {𝐶2𝑠 }, indexed by state s, is supported
by the payoff to investment projects.

• The assets themselves are not necessarily risky but the

future state of the world is unknown and unknowable ex
ante. Intuitively: you could think of insurance, or even a
simple bet.

• The equilibrium outcome has three properties…

1. Households must be maximising their utility subject to

the following budget constraint:

𝐶1 + ෍ 𝑞𝑠 𝐶2𝑠 = 𝐸
𝑠=1

Utility maximisation implies that reducing 𝐶1 and using the

proceeds to increase 𝐶2𝑠 will not affect lifetime utility; a
familiar argument by now. The Euler equation is (for all s):

𝑈′(𝐶1 ) = 𝑞1 𝜋𝑠 𝛽𝑈′(𝐶2𝑠 ) The price of a claim on future

𝑠 output in state s equals the
probability that the state occurs
′ 𝑠 multiplied by the MUC in that
Or: 𝑞𝑠 = 𝜋𝑠 𝛽𝑈 (𝐶2 )
𝑈 ′ (𝐶 )1 state relative to the MUC today

5
 Aside: The Stochastic Discount Factor

• Recall:
′ 𝑠
𝑞𝑠 = 𝜋𝑠 𝛽𝑈 (𝐶2 )
𝑈 ′ (𝐶 ) 1

• The highlighted part of the equation above is known as

the stochastic discount factor.

• It shows how the household values future state-

contingent consumption relative to consumption today; it
therefore determines how much the h.h. is willing to pay
for various assets which give a claim on future output in
different states of the world.

2. There must be no unexploited profit opportunities.

The (opportunity) cost of investing marginally more in

project 𝑖 is 1 unit of C1; the payoff to this investment is the
revenue that can be obtained from selling the state-
contingent output in state s, if this state materialises.

If a positive amount is invested in a particular investment

project (𝐾𝑖 > 0), in equilibrium the payoff to investing more
in that project must equal the cost. If nothing is invested in
project 𝑖 (𝐾𝑖 = 0), the payoff from the first unit of the
project must be smaller than the initial (opportunity) cost.

This can be summarised as:

𝑆
= 1 if 𝐾𝑖 > 0
෍ 𝑞𝑠 𝑅𝑖𝑠 for all 𝑖
≤ 1 if 𝐾𝑖 = 0
𝑠=1

6
3. Markets must clear.

𝑁
For period 1: 𝐶1 + ෍ 𝐾𝑖 = 𝐸
𝑖=1

For claims on period-2 output in state s:

𝐶2𝑠 = ෍ 𝐾𝑖 𝑅𝑖𝑠
𝑖=1

for all s

Summary
𝑆
= 1 if 𝐾𝑖 > 0
෍ 𝑞𝑠 𝑅𝑖𝑠
≤ 1 if 𝐾𝑖 = 0
Number of equilibrium 𝑠=1
for all 𝑖
conditions
𝑁

=1+𝑆+𝑁+1+𝑆 𝐶2𝑠 = ෍ 𝐾𝑖 𝑅𝑖𝑠

𝑖=1
for all s
𝑆
′ 𝑠 𝑁
𝐶1 + ෍ 𝑞𝑠 𝐶2𝑠 = 𝐸 𝑞𝑠 = 𝜋𝑠 𝛽𝑈 (𝐶2 )
𝑈 ′ (𝐶 )
1
𝑠=1 𝐶1 + ෍ 𝐾𝑖 = 𝐸
for all s 𝑖=1

Unknowns: 𝑆 x 𝑞𝑠 𝑆 x 𝐶2 𝑁 x 𝐾𝑖 𝐶1

7
 Interpretation
• So far, the model is conceptually similar to a standard model
of consumption or investment under uncertainty.

• All markets are perfectly competitive, full information, there

are no externalities: the equilibrium is Pareto efficient.

• As modelled, financial markets can be thought of as the

place where economic agents – i.e. households and firms –
meet to share risk and determine consumption of
investment outcome.

• ‘Unimportant’ extensions: adding additional time periods,

heterogeneity amongst households, investment adjustment
costs, adding labour market decisions; we consider ‘more
important’ extensions instead.

Plan for the Remainder of the Topic

• Add meaningful extensions/imperfections to the

baseline model

• With imperfections/frictions:
– Agency Costs and the Financial Accelerator

• Financial Crises:
– The Diamond-Dybvig Model
– Financial contagion

8
 Introducing Market Imperfections
• We have so far assumed that all economic agents are equally-
well informed; financial markets function efficiently with
symmetric (‘perfect’) information; potential investment
projects go ahead if their value exceeds the cost of capital.

• In reality, firms are better-informed about potential

investment projects than investors; financing typically comes
from ‘outsiders’ who possess less information, for example.

• Partly for this reason, intermediaries – such as banks – play a

key role in financial markets. They have the resources to
gather sufficient information about firms or investment
projects, but even this is not perfect. Example: your bank
knows less about your creditworthiness than you do!

• As Diamond (1984) points out, the presence of a financial

intermediary standing between investors and firms, though
potentially very useful, adds an additional layer to the
asymmetric information problem.

Households Financial Firms

(ultimate source Intermediary (initiate investment
of investment) (e.g. a bank) projects)

• Asymmetric information generates agency problems between

households and firms (+ banks in between, if relevant).
• Investors take risk, e.g. the firm could go bankrupt. This gives
the potential for moral hazard, where firms are more reckless
with investors’ capital than they would be with their own.
• Asymmetric information can create adverse selection, generate
‘monitoring costs’ and ultimately deter investment.

9
 A Model with Asymmetric Information
• This theoretical framework belongs to the class of costly
state verification models (Townsend, 1979).
• An entrepreneur has the opportunity to undertake an
investment project; the project requires 1 unit of resources.
• The entrepreneur has wealth W<1; (s)he therefore requires
outside financing of (1–W) for the project to proceed.
• The project has an expected output of 𝛾 (>0) ex ante;
however, actual project output is distributed uniformly
along the interval [0,2𝛾].
• Outside investors must bear some risk since the
entrepreneur will put all personal wealth into the project.

• To simplify our analysis, assume that the entrepreneur and

the investors are risk-neutral.
• There is a ‘risk-free’ alternative asset that pays a rate of
return r for certain; could be a government bond that pays
2%, say, so r=0.02.
• Outside investors compete for investment projects.

This implies:
1. The investment project is only ‘socially desirable’ from a
social planner’s perspective if expected output 𝛾 > 1 + 𝑟
2. The entrepreneur will only choose to proceed with the
project if the difference between 𝛾 and expected payments
to outside investors exceeds 1 + 𝑟 𝑊; otherwise buy bonds.
3. Due to competitive bidding, in equilibrium investors can
expect a return of 𝑟 on any particular project.

10
 Information and Monitoring Costs
• Entrepreneurs are better-informed than outside investors
about investment projects and, therefore, the actual output
that a particular project produces.

• An entrepreneur observes project output without cost; an

outside investor must incur a cost c to observe output, could
be time and/or monetary, where 𝑐 > 0 and 𝑐 < γ

• Preview: asymmetric information affects investment

outcomes and the structure of the optimal financial contract.

• Other types of information asymmetries – such as information

about project risk or about entrepreneurs’ behaviour – may
also be important but they have broadly similar implications
for economic distortions and the amplification of shocks.

Equilibrium under Symmetric Information

• Simple: entrepreneurs who present projects which have an
expected return >(1+r) obtain financing; those that cannot
meet this threshold don’t and the project cannot proceed.

• For those projects which win financing, the contract

between entrepreneur and outside investors rewards the
latter with an expected payment of:

(1 − 𝑊)(1 + 𝑟)

• The entrepreneur’s expected reward is then:

𝛾− 1−𝑊 1+𝑟 = 𝑊 1+𝑟 +𝛾− 1+𝑟 > 0

The entrepreneur is better off too >0

11
 Equilibrium under Asymmetric Information
• It is costly for outside investors to observe a project’s
output. To simplify matters, assume that there is just one
outside investor per investment project, i.e. the investor
is wealthy enough to fund the project alone.

• Under these assumptions, in equilibrium an entrepreneur

expects to receive pay out (1–W)(1+r) plus the monitoring
costs; i.e. the project’s return needs to cover the newly-
introduced costs as well, otherwise the risk-free asset is
more attractive.

• Again, the entrepreneur can expect to receive the

project’s expected output, which we take to be
exogenous, minus the expected payment to the investor.

The ‘Optimal’ Financial Contract

This will provide the outside investor with the required rate
of return while minimising time/money spent on monitoring.

Optimal contract design (in this case): if the declared project

payoff exceeds some set critical level D, the entrepreneur
pays the investor D and the latter does not bother to verify
output. n.b. the entrepreneur makes the output declaration.
However, if the declared payoff is less than D, the investor
verifies, pays the monitoring cost and takes all of the output.

This is structured like debt. The entrepreneur borrows (1−W)

and agrees to pay back an amount D, if possible; otherwise,
hand over everything (s)he can to the investor.

12
 The Optimal Payment Function

If output exceeds the

amount that is owed, the
entrepreneur pays off
the loan and keeps the
surplus…

Source: Romer (2019, p.468)

… but if the entrepreneur cannot make the required payment,

D, all output goes to the investor (45 degree line).

Reasoning (in place of a formal proof)

1. When the investor does not verify output, the payment does
not depend on actual output. Suppose that the payment should
be Q1 when output is Y1 and Q2 when output is Y2 (where
Q2>Q1). Since the investor does not know output, when output
is Y2 the entrepreneur pretends that it is Y1 and pays Q1. Thus
the contract has no mechanism to force the payment when
output is Y2 to exceed the payment when it is Y1 when there is
no verification.

2. The payment to the investor with verification can never exceed

the payment without verification, D. If it did, the entrepreneur
would pretend that output is not equal to the values of output
that yield a payment greater than D; the payment with
verification cannot exactly equal D either (wouldn’t verify).

13
3. The payment is D whenever output exceeds D. If the
payment is ever less than D when output is greater than
D, it is possible to increase the investor’s expected
receipts and/or reduce expected verification costs for
these levels of output; it would therefore be possible to
construct a more efficient contract for the investor.

4. The entrepreneur is unable to pay D if output is less than

D. In these cases the investor is forced to verify.

5. If the payment is less than total output when output is

less than D, increasing the payment in these situations
raises the investor’s expected receipts without changing
expected verification costs; not an efficient contract.

The Optimal Payment Function

X (1)
(2)
(1)
This reasoning leads
us to the optimal
contract structure, as
(4)
shown by the solid line
(3)
(5)

Source: Romer (2019, p.468)

Result: the optimal contract is structured just like debt

14
 The Equilibrium Value for D

• This needs to be set in the contract.

• We must determine how the investor’s expected receipts

net of verification costs vary with D; then find the value of
D that provides the investor with the required expected net
receipts.

• We need to calculate net receipts; remember, actual output

of the project is distributed uniformly along the interval
[0,2𝛾]; we need to recall the statistical properties of this
distribution. If we assume some other distribution for
actual output, our results would potentially be different.

Suppose that 𝐷 ≤ 2𝛾 (verification threshold ≤max. output):

– Actual (declared) output could be more than or less than D.
– If output > D, don’t verify and the investor receives a
payment of D. The probability of this occurring is, under a
uniform distribution: (2𝛾 − 𝐷)/(2𝛾 − 0)
– If output <D, the investor pays the verification cost and
receives all output. Under a uniform distribution, the
probability of this occurring is (𝐷 − 0)/(2𝛾 − 0) and the
average level of output conditional on output <D is:
(0 + 𝐷)/2

Suppose that 𝐷 > 2𝛾 (‘high’ verification threshold):

– Then output is always less than D. The investor always pays
the verification cost and takes all output; the expected
payment in this case is 0 + 2𝛾 /2 = 𝛾

15
 Summary
• From the results on the previous slide, the investor’s
expected returns minus verification costs, c, are:

Probability Probability

2𝛾−𝐷 𝐷 𝐷
2𝛾
𝐷 + 2𝛾 2
−𝑐 if 𝐷 ≤ 2𝛾
𝑅(𝐷) = 0
𝛾−𝑐 if 𝐷 > 2𝛾

Pay-off net
of cost

• There is non-linearity here so let’s plot this function…

R(D) against D in Diagram Form

When output exceeds

2γ−c, setting D=2γ−c and
accepting 2γ−c without
verification is better than
setting D>2γ−c

Source: Romer (2019, p.470)

16
 Suppose the investor has some ‘target’ return
(V1, V2 or V3) in mind:

The investor will not

want to participate
(‘credit rationing’) For V3: the contract cannot ‘solve’
the asymmetric information problem

Two equilibrium
values for D

A unique value for D

Source: Romer (2019, p.470)

Equilibrium Investment
• Both parties must be happy to proceed with the project;
this incentive compatibility constraint is needed because
investor and entrepreneur both have the opportunity to
invest in the ‘safe asset’.

• The contract is better than nothing at all, where

asymmetric information could potentially shut down all
project financing and lead to investment only in the safe
asset only; not good at the macro level.

• However, the contract is not perfect; it cannot cover all

scenarios. Some ‘credit rationing’ still occurs, which is
worse than a stylised world with full/perfect information;
i.e. some ‘good’ investment projects might not go ahead
because of asymmetric information.

17
 Interpretation and Application
• Our specific results hold for the particular model presented;
however, the concepts covered have broader applicability.

• Moral hazard: the entrepreneur might take more risk than

he/she would with external funds – where the investor
shares risk – than he/she would with own funds only.
• Adverse selection: suppose that there are many
entrepreneurs and that their investment projects have
different unobservable risk profiles; the projects which seek
external finance are likely to be the most risky.

• The ‘state verification’ costs that we have discussed can be

applied more generally.

The Financial Accelerator Model

• Imperfections in financial markets, such as the asymmetric
information problem we have covered, can amplify shocks to
the economy.

• Suppose that agency costs (e.g. monitoring costs) are lower for
wealthier entrepreneurs who can pledge more ‘skin in the
game’. When wealth falls generally – e.g. a recession – agency
costs rise, investment projects become more costly and some
projects are no longer viable; investment declines and wealth
falls again… this triggers a vicious circle.

• This idea can be applied to firms and households more

generally; an initial shock to ‘net worth’ leads to a downward
spiral via investment (or consumption; think credit cards).

18
 Credit Spreads
• The assumptions of the model give rise to an inverse
relationship between credit spreads – which is the cost of
external borrowing over the risk-free rate – and the net worth
of entrepreneurs, or firms or households more generally.

• When a borrower’s net worth is low, the incentive to truthfully

report returns/output or give maximum effort is also low. In
other words, agency costs are higher when economic interests
are less-closely aligned; credit spreads then increase.

• Assuming that net worth is procyclical, credit spreads will be

countercyclical; e.g. borrowing costs increase during
downturns and amplify fluctuations in investment, and
therefore output.

The Role of Collateral

• Suppose that the entrepreneur needs to post collateral to obtain
investment. If a recession, say, reduces the value of this
collateral, this only strengthens the financial accelerator effect
described previously.

• This works through expectations – for example, a drop in output

is expected to persist – but also by reallocating capital from
credit constrained to credit unconstrained firms, leading to a
lower MPK for the economy overall (Kiyotaki and Moore, 1997).

• The idea that financial-market imperfections magnify the effects

of shocks to economy-wide output is known as the financial
accelerator: Bernanke and Gertler (1989), Bernanke, Gertler and
Gilchrist (1999); see the summaries by Bernanke (2007) and de
Groot (2016).

19
 Modelling a ‘Financial Crisis’
with/without F.A. Effects

Source: de Groot (2016)

--- no agency costs --- quick recovery
= no credit spread in capital stock

Key Implication
• The financial system is important to investment; for
example, an increase in c – verification costs – will reduce
investment in our model.

• More broadly, the efficiency with which the financial sector

can ‘solve’ the asymmetric information problem and monitor
borrowers is important for investment at the macro level.

• As discussed previously, a banking sector intermediary

would add an extra layer to the asymmetric information
problem.

• For business cycles: disruptions to the financial system can

affect investment, and thus aggregate output and other key
macro variables, such as unemployment.

20
 A Model with F.A. Effects
Calibrated to the UK Economy

A negative
shock to
output is
amplified
with F.A.
effects

Source: Hall (2002)

Financial Crises

• We have seen how financial markets can be important

under asymmetric information. For example, they can
amplify the effects of real shocks to the economy.

• But we also know, from historical experience, that

financial markets themselves can be subject to sudden,
abrupt fluctuations; i.e. an independent source of shocks.

• For example, a bank that appears to be operating normally

one day can get into severe difficulty the next day (e.g.
Lehman Brothers); if this leads to knock-on effects then it
can pose a severe threat to macroeconomic stability.

21
 The Diamond-Dybvig Model
• This model considers the possibility of a bank run.
Depositors wish to be able to withdraw their savings
whenever they want, e.g. an instant-access account.

• This is problematic because of maturity mismatch: banks

borrow short-term (from depositors) and lend long-term
(they issue loans to firms or other households).

• In other words, the bank holds short-term liabilities and

long-term assets.

• Implication: if we all decide to withdraw our savings

tomorrow morning the banking system will collapse – the
money is simply not there for immediate withdrawal!

When might we all wish to withdraw our savings at the

same time? Example: during times of crisis or panic.

Friday 14
September
2007
The first run
on a British
Bank in 150
Years

Source: Reuters (2007)

22
 The Model
• There are three time periods: 0, 1, and 2.

• The economy consists of a continuum of agents, each of

whom is endowed with 1 unit of the economy’s single good
in period 0.

• If the good is invested, it yields R>1 units of the good if it is

held until period 2, but only 1 unit if the investment is
halted (i.e. liquidated) ‘early’, in period 1.

• The fact that the two-period return exceeds the one-period

return gives investment, an asset to the bank, a long-term
character; deposits will be short-term liabilities to the
bank, which is the maturity mismatch we wish to capture.

Individuals
• To begin with, all individuals are the same. However, in
period 1 a fraction 𝜃 of individuals learn that they only like
period 1 consumption (C1) while others discover that they
value period 2 consumption (C2) as well; patient vs.
impatient? We assume that one’s type cannot be observed
by other individuals (an information friction):

For ‘Type a’ individuals: 𝑈 𝑎 = ln𝑐1𝑎

For ‘Type b’ individuals: 𝑈 𝑏 = 𝜌ln(𝑐1𝑏 + 𝑐2𝑏 )

where: 0<𝜌<1 𝜌𝑅 > 1

23
 Autarchy
• Rule out any sort of trade, insurance or risk sharing among
individuals; this is just a starting point.
• Period 0 consumption isn’t valued by anybody so all agents
invest their endowment in period 0.
• Given the assumptions imposed: Type a’s liquidate their
investment projects early to enjoy period 1 consumption
(=1) only. Type b’s hang onto their investment and enjoy
period 2 consumption (=R) only; note ρ and ρR>1.
• Overall, expected utility is:

𝑈 𝐴𝑈𝑇𝐴𝑅𝐶𝐻𝑌 = 𝜃ln1 + (1 − 𝜃)𝜌ln𝑅

= (1 − 𝜃)𝜌ln𝑅

The Social Planner

• Now imagine a social planner who is benevolent and
omniscient, i.e. can observe individuals’ types at all times.

• Type a’s get no utility from period 2 consumption so for

them the planner will choose: 𝑐2𝑎 = 0. Similarly, for Type
b’s, the planner will choose: 𝑐1𝑏 = 0

What about 𝑐1𝑎 and 𝑐2𝑏 ?

• The fraction of investment projects liquidated early = 𝜃𝑐1𝑎

• Then the fraction of investment projects held to maturity
= (1 − 𝜃𝑐1𝑎 ), which will return R and be shared out among
Type b’s.

24
• We can state that the social planner’s resource constraint is
(for 𝑐2𝑎 = 𝑐1𝑏 = 0):

(1 − 𝜃𝑐1𝑎 )𝑅 Shares this out equally

𝑐2𝑏 = among Type b’s
1−𝜃

• Therefore, expected (weighted average) utility is:

𝐸 𝑈 = 𝜃ln𝑐1𝑎 + (1 − 𝜃)𝜌ln𝑐2𝑏

= 𝜃ln𝑐1𝑎 + 1 − 𝜃 𝜌 ln(1 − 𝜃𝑐1𝑎 + ln𝑅 − ln(1 − 𝜃)]

∗
• Take the FOC with respect to 𝑐1𝑎 then solve for 𝑐1𝑎 and
∗
𝑐2𝑏 (where * denotes optimal; see seminar class):

∗ 1 ∗ 𝜌𝑅
𝑐1𝑎 = >1 𝑐2𝑏 = <𝑅
𝜃+ 1−𝜃 𝜌 𝜃+ 1−𝜃 𝜌

Introduce a Bank
• Diamond and Dybvig show that we do not need a social planner
to achieve the efficient outcome; instead, we could allow one
individual in the economy to set up a bank.
• The bank takes deposits, makes investments and allows
customers to withdraw their funds whenever they wish, so long
as the bank has funds available; those who don’t withdraw in
period 1 share the investment proceeds in period 2.
• There is an equilibrium in the model in which Type a’s can
∗ ∗
achieve 𝑐1𝑎 and Type b’s can achieve 𝑐2𝑏 , i.e. the socially
optimal outcome is a Nash equilibrium in the model.
• Therefore, the bank provides liquidity to the economy; it makes
long-term investments and satisfies individuals’ consumption
timing preferences; Type b’s wait until period 2, Type a’s don’t.

25
 An Alternative Equilibrium
• However, Diamond and Dybvig show that there is another
Nash equilibrium in the model. It happens if Type b’s
believe that all agents – not just Type a’s – will try to
withdraw their deposits in period 1.

• If every agent tries to withdraw in period 1, the bank will

have to liquidate all investment projects early and there
will be nothing left to share in amongst Type b’s in period 2.
What would you do if you were a Type b individual?
Understandably, this causes a rush to withdraw funds as
soon as possible, i.e. in period 1, for both Types.

• Therefore, a bank run is also a Nash equilibrium in the

model; the asymmetric information is at the root of this
problem.

Intuition

Romer (2019, p.498): “The possibility of a run is inherent

when a bank has illiquid assets and liquid liabilities.
Liquid liabilities give depositors the option of withdrawing
early. But since the bank’s assets are illiquid, if all
depositors try to withdraw early, the bank will not be able
to satisfy them. As a result, if each agent believes that all
others are trying to withdraw early, they believe the bank
cannot meet its obligations, and so they too will try to
withdraw early.”

Notice that self-fulfilling prophecies are possible!

26
 Financial Contagion
If one bank (or financial institution) gets into financial difficulty,
this is likely to cause worries for clients at other banks. The
initial concern – whether warranted or not – can quickly
spread. Why?

1. Counterparty contagion: banks hold various types of claims

on one another – e.g. overnight lending – so the collapse of
one bank can trigger a domino effect.
2. Confidence contagion: fear based on asymmetric
information about which other banks might be dragged in.
3. Fire sale contagion: forced liquidations which bring down
market prices (i.e. reduce asset valuations for all).
4. Macroeconomic contagion: negative impact on economic
activity more generally; i.e. spill-over to the real economy.

Summary/Conclusions
• We have introduced financial markets into our macroeconomic
modelling framework, beginning with a baseline case under
complete markets and perfect information.

• However, in the real-world financial markets do not function

this smoothly. Various imperfections/frictions can give rise to
inefficiencies which affect the level of investment at the macro
level; we have considered asymmetric information, which
gives rise to agency costs.

• Policymakers need to consider how to respond appropriately

to these inefficiencies and the possibility of extreme volatility
in financial markets, i.e. financial crises which could include
bank runs and contagion effects.

27