Lecture 10

Banking, Liquidity, and Bank Runs

in an

In…nite Horizon Economy

Mark Gertler
NYU

Fall 2014

Based on Gertler and Kiyotaki (2014)

0
 Motivation

Banking distress and the real economy: Two complementary approaches:

1. "Macro" (e.g. Gertler and Kiyotaki, 2011)

(a) Bank balance sheets a¤ect the cost of bank credit.

(b) Losses of bank capital in a downturn raises intermediation costs

2. "Micro" (e.g Diamond and Dybvig, 1983)

(a) Maturity mismatch opens up the possibility of sunspot runs.

(b) Runs lead to ine¢ cient asset liquidation and loss of banking services.

1
 Motivation (con’t)

During the crisis both "macro" and "micro" phenomena were at work.
– (Gorton, 2010, Bernanke, 2010).

Starting point: Losses on sub-prime related assets depleted bank capital

– Forced a contraction of many …nancial institutions.

– Bank credit costs sky-rocketed

– Some of the major investment and money funds experienced runs, enhancing
the …nancial distress

August 2007 -August 2008: "slow" runs

September 2008: "fast" runs

2
 Motivation (con’t)

Macro models of banking distress:

– Emphasize balance sheet/…nancial acceletorator e¤ects

– Bank runs are excluded.

Micro models of banks

– Highly stylized; e.g. two periods

– Runs often unrelated to health of the macroeconomy.

3
 What We Do

Develop a simple macro model of banking instability that features both

– Balance sheet …nancial/accelerator e¤ects

– Banks runs (both slow runs and fast runs)

The model emphasizes the complementary nature of the mechanisms

– Balance sheet conditions a¤ect whether runs are feasible

– Two key variables:

Bank leverage ratio (a¤ects degree of maturity mismatch)

Liquidation prices

– Both depend on macroeconomics conditions

4
 Model Overview
Baseline Model: In…nite horizon endowment economy with …xed capital

– Households and Bankers

– Bankers hold imperfectly liquid long term assets and issue short term non-
contingent debt (maturity mismatch).

– Runs: panic failure of depositors to roll over short term debt

Similar to Cole/Kehoe(2001) self-ful…lling sovereign debt crisis

– Runs are unanticipated

Extended Model: Anticipated runs

– Anticipated runs can have harmful e¤ects even if the runs do not occur ex
post.

– Can distinguish between slow and fast runs

5
 Intermediated vs. Directly Held Capital

Capital Allocation

Ktb + Kth = K = 1

– Ktb intermediated capital

– Kth capital directly held by households

6
 Intermediated vs. Directly Held Capital (con’t)
Technology for intermediated capital

date t+1
date t
8
o >
< Ktb capital
Ktb capital !
>
:
Zt+1Ktb output

Rate of return on intermediated capital

b Zt+1 + Qt+1
Rt+1 =
Qt

7
 Intermediated vs. Directly Held Capital (con’t)
Technology for capital directly held by households

date t date t+1

) (
Kth capital Kth capital
!
f (Kth) goods Zt+1Kth output

f (Kth) management cost; f 0 > 0; f 00 0:

Rate of return on directly held capital

h Zt+1 + Qt+1
Rt+1 =
Qt + f 0(Kth)
Households directly hold capital due to …nancial constraints on banks.
8
 NO BANK RUN EQUILIBRIUM



       

Dt
Qt Kbt

Nt

CAPITAL HOUSEHOLDS

K
 

!"#$%

Qt Kht

BANK RUN EQUILIBRIUM

CAPITAL

K Q*t K HOUSEHOLDS
 Households

Rt+1 deposit rate; Rt+1 return on deposits

xt depositor recovery rate after run (endogenous)

Deposit contract:

– Short term (one period)

(
Rt+1 if no bank run
– Rt+1 =
xt+1Rt+1 if run occurs

– No sequential service constraint

Unanticipated runs: pt = 0 ) EtRt+1 = Rt+1

Anticipated runs: pt [0; 1] ) EtRt+1 < Rt+1

9
 Households (con’t)

Choose fCth; Dt; Kthg to max:

0 1
1
X
Ut = Et @ i ln C h A
t+i
i=0

Subject to:

Cth + Dt+1 + QtKt+1

h + f (K h ) = Z W h + R
t t
h
t+1 Dt 1 + (Zt + Qt)Kt

Fonc for Dt and Kth :

1 = Etf t;t+1Rt;t+1g
Z + Qt+1
1 = Etf t;t+1 t+1 h
g
0
Qt + f (Kt )

t;t+1 = Cth=Ct+1
h

10
 Bankers
A measure unity of bankers

Each has an i.i.d. survival probability of

– ) expected horizon is 1 1

Banker consumes wealth upon exit

Preferences are linear in "terminal" consumption cbt+i

Vt = Et [(1 )cbt+1 + Vt+1]

Each exiting banker replaced by a new banker.

– Starts with an endowment wb

1
 Bankers (con’t)
Bank balance sheet
Qtktb = dt + nt

Net worth nt for surviving bankers

nt = RtbQt 1ktb 1 Rtdt 1:

nt for new bankers

nt = wb

cbt for exiting bankers

cbt = nt
12
 Limits to Bank Arbitrage

Agency Problem:

– After the banker borrows funds at the end of period t; it may divert: a fraction
of loans

– If the bank does not honor its debt, creditors can

recover the residual funds and
shut the bank down.

Incentive constraint
Qtktb Vt

13
 Solution
"Leverage" constraint
Qtktb
t
nt
t is
– decreasing in
– increasing in t
= E [( R b Rt+1) t+1]
t t t+1

where t+1 > 1 is the banker’s expected shadow value of nt+1

Two key results

– t is countercyclical ) t is countercyclical.
– nt 0 ) bank cannot operate (key for run equilbria)

14
 Aggregation
Aggregate leverage constraint

QtKtb = tNt

Aggregate net worth

Nt = [(Rtb R t ) t 1 + R t ]N t 1 ] + W b

Volatility of Nt depends on t 1 and volatility of Rtb.

15
 Bank Runs

Baseline: Ex ante zero probability of a run.

Consider the possibility of a run ex post:

Ex post a "bank run" equilibrium" is possible if:

– A depositor believes that if other households do not roll over their deposits,
the depositor will lose money by rolling over.

– This condition is met if the bank’s net worth goes to zero if other depositors
do not roll over

16
 Conditions for a Bank Run Equilibrium (BRE)
Timing of events:
– At the beginning of period t, depositors decide whether to roll over their
deposits with the bank.
– If they choose to "run", the bank liquidates its capital and it sells it to house-
holds who hold it with their less e¢ cient technology.

A run is then possible if

(Qt + Zt)Ktb 1
xt = <1
RtDt 1
xt depositor recovery rate
Qt the liquidation price of a unit of the bank’s assets.
xt < 1 ) nt = 0 in the event of a run.

17
 Conditions for a Bank Run Equilibrium (BRE) (con’t)
We can simplify the condition for a BRE:

)
Rtb 1
xt = <1
Rt 1 1= t 1

with
(Zt + Qt )
Rtb =
Qt 1

Whether a BRE exists depends on (Qt ; t 1; Rt):

Qt is procyclical and t.is highly countercyclical ) the likliehood of a BRE is

countercyclical.

18
 Liquidation Price
After a bank run at t :
Kth = K = 1

Entry of new bankers begins at t + 1

Nt+1 = W b + W b;
Nt+i = b
[(Zt+i + Qt+i)Kt+i Rt+iDt+i 1] + W b; for all i 2:
1

Household euler equation for direct capital holding

Zt+1 + Qt+1
Etf t;t+1 g=1
Qt + f 0(1)
)
2 3
1
X
Qt = Et 4 h ))5
f 0(Kt+i :f 0(1)
t;t+i(Zt+i
i=1

19
Table 1: Parameters
Baseline Model
0.99 Discount rate
0.95 Bankers survival probability
0.19 Seizure rate
0.008 Household managerial cost
0.95 Serial correlation of productivity shock
Z 0.0126 Steady state productivity
! b 0.0011 Bankers endowment
! h 0.045 Household endowment

1
Table 2: Steady State Values
Steady State Values
Baseline
K 1
Q 1
ch 0.055
cb 0.0036
Kh 0.31
Kb 0.69
10
Rk 1.0504
Rh 1.0404
R 1.0404

2
 FIGURE 3: A Recession in the Baseline Model; No Bank Run Case
z y kb
0 0 0

-0.1
% from ss

% from ss

% from ss
-0.02 -0.02
-0.2
-0.04 -0.04
-0.3

-0.06 -0.06 -0.4

0 20 40 0 20 40 0 20 40

Q x 10
-3 ERb-R n
0 8 0
Ann.  from ss
6 -0.2
% from ss

% from ss
-0.02
4 -0.4
-0.04
2 -0.6

-0.06 0 -0.8
0 20 40 0 20 40 0 20 40

-3 R cb
x 10 ch
8 0 0
Ann. % from ss

6 % from ss -0.2
% from ss

-0.01
4 -0.4
-0.02
2 -0.6

0 -0.03 -0.8
0 20 40 0 20 40 0 20 40
Quarters Quarters Quarters
 Figure 4: Ex-Post Bank Run in the Baseline Model
y kb Q
0 0 0

-0.05 -0.05
% from ss

% from ss

% from ss
-0.1 -0.5 -0.1

-0.15 -0.15

-0.2 -1 -0.2
0 20 40 0 20 40 0 20 40

RUN Q* *
0.02 0 6

4
% from ss

% from ss
0.01 -0.02
2
0 -0.04
0

-0.01 -0.06 -2
0 20 40 0 20 40 0 20 40

ch cb ERb-R
0.1 0 0.03

0.05 Ann.  from ss

% from ss

% from ss

0.02
0 -0.5
0.01
-0.05

-0.1 -1 0
0 20 40 0 20 40 0 20 40
Quarters Quarters Quarters
No Run Recession unanticipated run
 Anticipated Bank Runs

pt = probability of run in t + 1

Deposit Rate
(
Rt+1 with probability 1 pt
Rt+1 =
xt+1Rt+1 with probability pt

with
(Qt+1 + Zt+1)Ktb b
Rt+1 1
xt+1 = =
Rt+1Dt Rt+1 1 1= t

20
 Anticipated Bank Runs (con’t)

Household FONC for deposits:

h i
1 = Rt+1Et (1 pt) t;t+1 + pt t;t+1xt+1

= C h =C h
t;t+1 t t+1

If t;t+1xt+1 < t;t+1, an increase in pt raises Rt+1:

21
 Anticipated Bank Runs (con’t)

Assume pt depends inversely on the expected depository recovery rate Et(xt+1)

(
g (Et(xt+1)) with g 0( ) < 0
pt =
0; if Et(xt+1) = 1:

– Parametric example

g( ) = 1 Et(xt+1)

Run is still a sunspot, but probability of sunspot depends on Et(xt+1):

22
 Anticipated Bank Runs (con’t)
Leverage constraint

QtKtb
= t
Nt
where t is increasing in t :

t = Et[(1 pt) t+1(Rkt+1 Rt+1)]

Evolution of net worth

Nt = [(Rtb R t ) t 1 + R t ]N t 1 ] + W b

an anticipated increase in pt is contractionary in two ways

– t declines since t falls ( slow run)
– Nt+1 declines since Rt+1 increases.
23
 Figure 5: Recession with positive probability of a run
p y kb
0.03 0 0

-0.02 -0.2

% from ss

% from ss
 from ss

0.02
-0.04 -0.4
0.01
-0.06 -0.6

0 -0.08 -0.8
0 20 40 0 20 40 0 20 40

Q  n
0 1 0

-0.2
% from ss

% from ss

% from ss
-0.05 0.5 -0.4

-0.6

-0.1 0 -0.8
0 20 40 0 20 40 0 20 40

ERb-Rd x 10
-3 Rfree
0.015 x 10
-3 Rd-Rfree 8
2

Ann. % from ss
Ann.  from ss

Ann.  from ss

6
0.01
1 4
0.005
2

0 0 0
0 20 40 20 40 0 20 0 40
Quarters Quarters Quarters
Recession with positive probability of run No Run Recession
 Figure 6: Recession with positive Run Probability and Ex-Post Run
p y kb
0.03 0 0

-0.05

% from ss

% from ss
 from ss

0.02
-0.1 -0.5
0.01
-0.15

0 -0.2 -1
0 20 40 0 20 40 0 20 40

Q  n
0 10 0

-0.05
% from ss

% from ss

% from ss
5 -0.5
-0.1
0 -1
-0.15

-0.2 -5 -1.5
0 20 40 0 20 40 0 20 40

ERb-Rd x 10
-3 Rd-Rfree Rfree
0.03 2 0.3

Ann. % from ss
Ann.  from ss

Ann.  from ss

1.5 0.2
0.02
1 0.1
0.01
0.5 0

0 0 -0.1
0 20 40 0 20 40 0 20 40
Quarters Quarters Quarters
anticipated Run No Run Recession
 Figure 7: Credit Spreads and Bank Equity: Model VS Data

2.5
Bear Sterns Lehman Brothers
2

1.5
Credit Spreads

0.5

-0.5

-1
2007 Q2 2008 Q2 2009 Q2 2010 Q2

40

35

30

25
Bank Equity

20

15

10

0
2007 Q2 2008 Q2 2009 Q2 2010 Q2

DATA MODEL

Description: The data series for Credit spreads is the Excess Bond Premium as computed by Gilchrist
and Zakrasjek (2012); Bank Equity is the S&P500 Financial Index. The model counterparts are the paths
of E(Rb-Rd) and V as depicted in Figure 6 normalized so that their steady-state values match the actual
values in 2007 Q2.
 Some Remarks About Policy

As in Diamond/Dybivg a role for deposit insurance.

– Eliminates bank run equilibrium
– But may have moral hazard e¤ects on risk-taking.

Can o¤set with capital requirements

– Reduces risk-taking
– Reduces the liklelihood of a bank run equilibrium
– But if bank equity capital costly to raise, can increase intermediation costs.

Alernative: commitment to lender-of-last resort policies

– Stabilizing liquidation prices reduces likliehood of bank runs
– Examples: lending againt good collateral
– Asset purchases of good quality securities (e.g. AMBS)
24