WORKING PAPER SERIES

NO 1495 / november 2012

Bubbles, Banks and

Financial Stability

Kosuke Aoki and Kalin Nikolov

Macroprudential
Research Network
,QDOO(&%
SXEOLFDWLRQV
IHDWXUHDPRWLI
WDNHQIURP
WKH»EDQNQRWH

NOTE: This Working Paper should not be reported

as representing the views of the European Central
Bank (ECB). The views expressed are those of the
authors and do not necessarily reflect those of the ECB.
Macroprudential Research Network
This paper presents research conducted within the Macroprudential Research Network (MaRs). The network is composed of economists
from the European System of Central Banks (ESCB), i.e. the 27 national central banks of the European Union (EU) and the European
Central Bank. The objective of MaRs is to develop core conceptual frameworks, models and/or tools supporting macro-prudential
supervision in the EU.
The research is carried out in three work streams: 1) Macro-financial models linking financial stability and the performance of the
economy; 2) Early warning systems and systemic risk indicators; 3) Assessing contagion risks.
MaRs is chaired by Philipp Hartmann (ECB). Paolo Angelini (Banca d’Italia), Laurent Clerc (Banque de France), Carsten Detken
(ECB), Cornelia Holthausen (ECB) and Katerina Šmídková (Czech National Bank) are workstream coordinators. Xavier Freixas
(Universitat Pompeu Fabra) and Hans Degryse (Katholieke Universiteit Leuven and Tilburg University) act as external consultant.
Angela Maddaloni (ECB) and Kalin Nikolov (ECB) share responsibility for the MaRs Secretariat.
The refereeing process of this paper has been coordinated by a team composed of Cornelia Holthausen, Kalin Nikolov and Bernd
Schwaab (all ECB).
The paper is released in order to make the research of MaRs generally available, in preliminary form, to encourage comments and
suggestions prior to final publication. The views expressed in the paper are the ones of the author(s) and do not necessarily reflect those
of the ECB or of the ESCB.

Acknowledgements
We are especially grateful to Jaume Ventura whose excellent comments helped us improve the paper a lot. We also thank Henrik
Borchgrevink, Wouter den Haan, Piti Disyatat, Martin Eichenbaum, Philipp Hartmann, Jonathan Heathcote, Xavier Freixas, Roman
Inderst, Peter Karadi, Nobu Kiyotaki, Alberto Martin, Tomoyuki Nakajima, Helene Rey, Victor Rios-Rull, Immo Schott, Shigenori
Shiratsuka, Harald Uhlig, Christopher Waller and Philippe Weil. The views expressed here are those of the authors and should not be
interpreted as those of the European Central Bank.

Kosuke Aoki
at University of Tokyo; e-mail: kaoki@e.u-tokyo.ac.jp

Kalin Nikolov
at European Central Bank; e-mail: kalin.nikolov@ecb.europa.eu

© European Central Bank, 2012

Address
Kaiserstrasse 29, 60311 Frankfurt am Main, Germany

Postal address
Postfach 16 03 19, 60066 Frankfurt am Main, Germany

Telephone
+49 69 1344 0

Internet
http://www.ecb.europa.eu

Fax
+49 69 1344 6000

All rights reserved.

ISSN 1725-2806 (online)

Any reproduction, publication and reprint in the form of a different publication, whether printed or produced electronically, in whole
or in part, is permitted only with the explicit written authorisation of the ECB or the authors.

This paper can be downloaded without charge from http://www.ecb.europa.eu or from the Social Science Research Network electronic
library at http://ssrn.com/abstract_id=2176147.

Information on all of the papers published in the ECB Working Paper Series can be found on the ECB’s website,
http://www.ecb.europa.eu/pub/scientific/wps/date/html/index.en.html
 Abstract

We build a model of rational bubbles in a limited commitment econ-

omy and show that the impact of the bubble on the real economy cru-

cially depends on who holds the bubble. When banks are the bubble-

holders, this ampli…es the output boom while the bubble survives but

also deepens the recession when the bubble bursts. In contrast, the real

impact of bubbles held by ordinary savers is more muted.

Keywords: Financial stability

 Non-technical summary
This paper builds a macroeconomic model with defaultable banks that can serve as a
framework for analysing systemic risk 1 in terms of the endogenous build-up and
unravelling of financial imbalances. The model is used to examine the role of asset
price bubbles in generating widespread or systemic financial instability 2.
Ours is an economy in which credit constraints create a ‘savings glut’ – too many
savings chase too few assets that savers can invest in. This pushes real interest rates to
very low levels and opens up the possibility that asset valuations can depart from
fundamentals. The main contribution of our paper is to show that the impact of such
asset price bubbles on systemic risk depends crucially on which economic agents are
exposed to these bubbles. When banks hold exposures to overvalued assets, their net
worth and solvency is at risk from an asset price correction. When an asset price
bubble collapses, it wipes out the equity of the banking system causing widespread
financial instability - bank failures and a severe contraction of lending to the real
economy. The collapse of bank credit supply is the main channel through which
financial instability is transmitted to the real economy. As a result, investment,
employment and real activity decline.
In contrast, when the exposure to asset price busts is with ordinary savers, the effect
of an asset price correction is surprisingly muted. Those exposed to the bubble suffer
a loss of net worth but there is no spillover to the rest of the economy. Our model
implies that different bank exposures to bubbles are an important reason why the ‘Dot
Com Bubble’ did not lead to a banking crisis while the ‘Subprime Bubble’ did.
Our analysis finds that bank exposures to overvalued assets (and hence systemic risk)
are highest whenever financial institutions believe that they will be bailed out in the
event of losses. This boosts the degree of asset overvaluation and concentrates bubble
exposures into the hands of banks. Systemic risk increases as a result.
Working in the opposite direction is the ‘franchise value of the bank’. Banks earn
large profits in equilibrium and this makes them less willing to take risks by
undertaking exposure to asset price bubbles. This suggests that boosting banks’
‘franchise values’ is a key channels through which macro-prudential policy acts as a
deterrent to bank risk-taking.
Our research carries a number of additional implications relevant for macro-prudential
policies. First of all, it suggests that it is not necessarily the pure size of an asset
bubble that matters most for systemic risk but the degree of banking sector exposure
to it. Second, it suggests that systemic risk will remain a potential problem as long as
certain banks remain ‘too big’ or ‘too inter-connected to fail’.

1
Systemic risk is defined as the risk of insolvency of the aggregate banking system.
2
We define financial instability as a situation in which the aggregate banking system is insolvent.
1 Introduction

The last 20 years have seen two spectacular episodes of …nancial instability.
First, stock prices experienced a truly unprecedented increase and subsequent
collapse around the turn of the millenium. Then came the dramatic rise and
fall of world-wide housing prices, culminating in the …nancial crisis and ‘Great
Contraction’ of 2008-2009. These movements in …nancial prices were large
enough to reawaken interest in asset price bubbles and the way they a¤ect the
real economy. In addition, the near failure of the global banking system during
the Lehman Crisis of 2008 has increased awareness of the importance of banks
in amplifying macroeconomic ‡uctuations.

Motivation and economic questions Our paper studies the interaction

of asset price bubbles, the banking system and the real economy in a general
equilibrium model with credit constraints. The current …nancial crisis is the
most immediate motivation for wanting to study this interaction. Bank losses
due to subprime mortgage defaults depleted bank capital positions and led to
a severe credit crunch and the deepest recession since the 1930s. This episode
is by no means unique in history. In Japan, during the 1980s, real estate
prices grew rapidly allowing property developers to borrow heavily from banks
using real estate as collateral. When property prices collapsed in the 1990s,
Japan went through a protracted banking crisis and a ’lost decade ’ of low
economic growth. A similar set of events occurred during the housing bust
and …nancial crisis that hit the Scandinavian countries in the early 1990s. In
all of these historical episodes, the banking system was exposed to a bursting
housing bubble through a higher incidence of household default. As Reinhart
and Rogo¤ (2008) show, using their rich time-series and cross-country data
set, the resulting insolvency of the …nancial system tends to lead to a deep
downturn in real activity.
The one bubble episode that stands in clear contrast to all the costly boom-
bust cycles described above is the 1998-2000 ’technology bubble’. Stock prices
rose and then collapsed dramatically without triggering a banking crisis or a
deep recession. Why was this bubble episode so di¤erent from other (mainly
housing) recent bubbles? More generally, what makes some asset price busts
more costly than others? These are the main questions we focus on in this
research. We build a model which focuses on bank exposures to bubbles as an
important reason why some lead to crisis while others do not1 .

Main results Following the insights of the growing literature on rational

bubbles with credit constraints (Caballero and Krishnamurthy (2006), Kocherkalota
(2009), Martin and Ventura (2011a), Martin and Ventura (2011b), Arce and
Lopez-Salido (2008), Farhi and Tirole (2011), Hirano and Yanagawa (2010))
we focus on an environment in which binding borrowing constraints lead to a
shortage of means of saving and the use of dynamically ine¢ cient investment
technologies in equilibrium. This creates the conditions for bubbles to circu-
late, helping to improve the supply of liquidity but also exposing the economy
to the risks of their bursting. Our innovation relative to the rest of the ratio-
nal bubbles literature is to model …nancial intermediation explicitly. In our
model credit ‡ows along a chain that starts with a saver, goes via a bank and
ends up with a …nal borrowing …rm. This realistic feature of our environment
also has the implication, overlooked by the rest of the literature, that asset
1
Commercial banks in the US tend not to hold equities and their Commercial and In-
dustrial Loans had fallen to 20% of total loans by the late 1990s. In contrast, almost 60%
of their total loans and leases were secured by real estate. This is an important reason why
housing busts tend to be more costly for the banking system compared to equity busts.
Martin and Ventura (2011a) argue that a reason why the ’dot com’bust did not lead to a
recession is that it was immediately followed by the housing bubble which helped to expand
the economy. This argument is plausible and we see it as complementary to our channel
which stresses the health of the banking system.
price bubbles can be held by a variety of agents with very di¤erent economic
roles in equilibrium. Our paper shows that the identity of the bubble-holder
is vital for understanding the e¤ect of bubbles on the real economy. Bubbles
held by banks expand output more during the boom phase and then lead to
a much more severe contraction in credit and output when they …nally burst.
In contrast, bubbles held by ordinary savers have a relatively muted e¤ect on
the real economy.
The key to understanding our main result lies in the way the identity of the
bubble holder a¤ects the evolution of the wealth distribution in equilibrium.
Bubbles in our world are risky assets which deliver a higher return compared
to safe assets in order to compensate the holder for the possibility of losing
everything in the bust. This implies that the bubble holder experiences strong
growth of net worth while the bubble survives and then a sharp fall in net worth
when the bubble bursts. These movements in the wealth distribution have large
real e¤ects when the a¤ected agents are the credit constrained intermediaries.
The gains and losses they make on bubble holdings a¤ect aggregate credit
supply and therefore the production scale of the bank’s credit constrained
borrowers. During the bubble’s run up, banks expand their size, increase credit
supply and this allows productive …rms to increase output. The re-allocation
of productive factors towards more e¢ cient …rms then boosts aggregate TFP.
When the bubble crashes, banks make large losses and in‡ict a credit crunch on
the rest of the economy. In contrast, ordinary savers are credit unconstrained
and their net worth does not play the same role in the credit intermediation
mechanism. When savers hold the bubble, their wealth also ‡uctuates but this
has a very limited impact on the credit constraints faced by other agents in
the economy.
In this paper, banks only undertake bubble exposures when they believe
that potential losses will be underwritten by the government (moral hazard).
The promise of bailouts gives banks an advantage in holding bubbly assets,
pushing the risky exposure into their hands. However, the main result of our
paper is general: the identity of the bubble holder is key to understanding the
real e¤ects of bubbles during the boom as well as the bust. Bubbles held by
banks create a bigger boom and a bigger crash by stimulating a boom-bust
cycle in credit supply.

Related Literature Our paper contributes to a large literature that stud-

ies the way asset price bubbles a¤ect economic activity. The seminal work
of Tirole (1985) showed that, in an environment without credit constraints,
bubbles have a contractionary e¤ect on economic activity because they dis-
place ine¢ cient investments in savers’portfolios. This is the famous ’crowd-
ing out’e¤ect of bubbles. Subsequent papers (Caballero and Krishnamurthy
(2006), Kocherkalota (2009), Martin and Ventura (2011a), Martin and Ventura
(2011b), Farhi and Tirole (2011)) have shown that when there are credit mar-
ket imperfections, bubbles may have an expansionary e¤ect through a variety
of mechanisms that help to reduce the severity of credit constraints.
In Martin and Ventura (2011b) the expansionary e¤ect of bubbles arises
because bubble creation increases both current and anticipated future pro…ts
for productive entrepreneurs who cannot borrow against their tangible assets
due to a moral hazard problem. Because the anticipated pro…ts from future
bubble sales are collateralisable, this allows the most productive entrepreneurs
to increase borrowing and production in the current period, thereby increas-
ing aggregate TFP. Related to the results in our paper, Martin and Ventura
(2011b) demonstrate that the e¤ects of bubble creation depend crucially on the
identity of the bubble creator. When credit constrained productive …rms cre-
ate bubbles, the bubble is strongly expansionary. When unconstrained savers
bene…t from the creation, the bubble is contractionary as in Tirole (1985). In
our paper, we abstract from bubble creation, instead focusing on the impact of
bubble holding and bubble destruction on economic outcomes. We show that
this impact is also highly dependent on the identity of the bubble holder.
Similarly to Martin and Ventura (2011b) our paper also relies on a redis-
tribution of productive resources towards more able …rms in order to generate
an increase in aggregate TFP and output as a result of the bubble. But the
way in which resources are channelled into productive hands does not rely
on bubble creation. Instead, bubble holdings turn out to be expansionary or
contractionary due to the interplay of the ’liquidity e¤ect’and the ’competi-
tion e¤ect’identi…ed in Farhi and Tirole (2011) as well as the ’credit supply
e¤ect’ which is unique to our paper. It is the ’credit supply e¤ect’ which
really depends on who holds the bubble. It is at its strongest when the bub-
ble is in the banks’ hands. There it boosts banks’ net worth and expands
credit supply. This also exerts a powerful positive externality on the net worth
of borrower entrepreneurs, helping to increase their productive activities and
the economy’s aggregate e¢ ciency. When savers hold the bubble, such credit
supply externalities are relatively weak and the bubble is less expansionary.
Our paper’s …ndings are in line with several important stylised facts uncov-
ered by the literature on the empirical regularities around …nancial crises. Our
focus on rational bubbles that can only occur in a low interest rate environment
is supported by Schularick and Taylor (2012) who report that growth adjusted
real interest rates are very low for several years in the run up to global crises.
Schularick and Taylor (2012) present evidence that credit booms are associated
with larger cumulative increases and steeper subsequent declines in GDP. Our
theoretical results are exactly in line with such empirical regularities.
Plan of the paper Section 2 introduces the economic environment, section
3 describes the equilibrium without government sector and discusses the con-
ditions for the existence of bubbles and bubble ownership in the absence of
government bailout policy. Section 4 examines the e¤ect of the …nancial safety
net on the existence and ownership of bubbles, and analyses the implications
of bubble ownership on the real e¤ects of bubbly episodes. Finally, section 5
concludes.

2 The Model

The economy is populated with three kinds of agents. There is a continuum of

in…nitely lived entrepreneurs and a continuum of in…nitely lived workers both
of measure 1. There is also a continuum of bankers who have …nite lives and
can die with probability 1 in any period, conditional on being alive in the
previous period.

2.1 Entrepreneurs

Each entrepreneur is endowed with a constant returns to scale production

function which converts labor ht into output in the next period yt+1 .

yt+1 = ait ht ; (1)

where ait is a productivity parameter which is known at time t.

In each period some …rms are productive (ait = aH ) and the others are
unproductive (ait = aL < aH ). Each entrepreneur shifts stochastically between
productive and unproductive states following a Markov process. Speci…cally, if
a productive entrepreneur in this period may become unproductive in the next
period with probability ; and an unproductive entrepreneur in this period may
become productive with probability n . This probability is independent across
entrepreneurs and over time. This Markov process implies that the fraction of
productive entrepreneurs is stationary over time and equal to n=(1 + n), given
that the economy starts with such population distribution. We assume that
the probability of the productivity shifts is not too large:

+ n < 1: (2)

This assumption implies that the productivity of each agent is persistent.

Entrepreneurs are ex-ante identical and have log utility over consumption
streams
X
1
E t
U = Et ln ct (3)
t=0

Entrepreneurs purchase consumption (ct ), bubbles (met ) at price t and

bonds bt . They also pay wages to the workers they hire wt ht in order to receive
future revenues ai ht which the government taxes at rate t after deducting debt
repayments. wt and ht denote real wage and labor respectively.

ct +wt ht +met t bt = (1 t) ai ht 1 Rti 1 bt 1 + met 1 t (1 t ) zt (4)

where zt stands for entrepreneur’s net worth. Rti is the interest rate which is
equal to the loan rate Rtl when the entrepreneur is a borrower and Rtd when
the household is a saver. The bubble asset in our economy is a durable but
intrinsically worthless asset which has no productive or consumption value. In
other words, its fundamental value is zero.
Due to limited commitment in the credit market, agents will only honour
their promises if it is in their interests to do so. We assume that only a fraction
of the value of the …rm can be seized by creditors and tax collectors. Further-
more we assume that the tax authorities have a …rst call on the …rm’s resources
while private creditors are second in line. Hence the collateral constraint is
given by:
Rtl bt + Et t+1 6 Et yt+1 ; 0< <1 (5)

where
Et t+1 = Et t+1 ai ht Rti bt + met t+1

stands for expected tax payments in the following period. Entrepreneurs max-
imize (3) subject to (4) and (5).

2.2 Banks

We assume that only banks can enforce debt repayments in our economy.
Consequently, all borrowing and lending will be bank-intermediated. Bankers
are risk neutral and live for a stochastic length of time. Once bankers receive
an “end of life”shock, they liquidate all their asset holdings and consume their
net worth before exiting. This shock hits with probability 1 :
Banks maximize the following objective function:

X
1
B
U = Et ( )t cB
t (6)
t=0

subject to the following constraints explained below.

In each period the bank has net worth (nt ). It collects deposits (dt ) from
the savers. Then it lends to the borrowers (bt ), purchases bubbles ( t mbt ), or
consumes (cbt ). We assume that intermediation is entirely costless. The bank’s
balance sheet constraint is given by

cbt + bt + b
t mt = nt + dt : (7)

The evolution of net worth is given by

nt+1 = Rtl bt + b
t+1 mt Rtd dt : (8)

when the bubble does not burst and by:

nt+1 = Rtl bt + b
t+1 mt Rtd dt : (9)

when it does burst. t+1 is the fraction of the banks’bubble investment which
is guarranteed by the government. In the event of a bubble collapse, the
government transfers these funds to the banks to compensate them for losses
made. The parameter t+1 is a simple means of capturing the explicit or
implicit guarrantees given by the government to the banking system
Following Gertler and Karadi (2009), we model banks subject to limited
commitment. More speci…cally, the banker may divert 1 fraction of de-
posits. Once he diverts, he consumes the funds and closes his bank, remaining
inactive until his ’death’. The savers can recover the remaining fraction of
deposits. Since the savers recognize the banker’s incentive to divert funds,
they will restrict the amount they deposit with the intermediary, according to
the following borrowing constraint:

(1 )dt 6 V (nt ) : (10)

The left hand side of equation (10) is the value when the banker diverts, while
the right hand side is the value when he did not (i.e., the continuation value
of the bank). We also assume that the bank cannot short mt . The bank
maximizes (6) subject to (7), (8), (9) and (10).

2.3 Workers

Unlike the entrepreneurs, the workers do not have access to the production
technology nor any collateralizable asset in order to borrow. They maximize
the following utility

X
1
h1+
t
W t
U = Et cw
t (11)
t=0
1+

subject to their ‡ow-of-funds constraint

cw w
t + mt t bw w
t = wt ht + mt 1 t Rtd 1 bw
t 1; (12)

here superscript ‘w’stands for ‘workers’.

2.4 The Government

We assume that the only role for the government in this economy is to levy
taxes on entrepreneurs and bail out the banking system when it makes losses.
We assume that the government follows a balanced budget rule and does not
issue government debt. Consequently taxes are only levied whenever bailout
spending is necessary. For the rest of the time, taxation is zero.

3 Equilibrium without government

We consider a stochastic bubble that persists with probability . With prob-

ability 1 the bubble bursts and its value reverts to zero. We assume this
probability is constant over time. Also, we assume that once bubbles burst
they never arise again.
In equilibrium, due to the di¤erence in their productivity, productive en-
trepreneurs borrow from banks and unproductive entrepreneurs make deposits
to banks. We focus on equilibria in which the productive entreprenerus borrow
up to their borrowing constraint2 . In this section, we characterise the equi-
librium without government. Therefore we set t+1 = 0 and t+1 = 0 at all
times.

3.1 Entrepreneurs’optimal behavior

The entrepreneurs’problem can be interpreted as a savings problem with un-

certain returns. Since utility function is logarithmic and there is no labour
income or transfer income, entrepreneurs consume a constant fraction of net
worth (zt ).
ct = (1 ) zt : (13)

and save the remaining fraction3 .

The entrepreneur has several possibilities for accumulating net worth. First
of all, an entrepreneur with productivity at = ah ; aL can undertake unleveraged
at
investments using his own technology. This yields a return of wt
: the cost of
hiring a worker is wt and the worker produces output at in the following period.
Secondly, the entrepreneur can deposit in the banking system, earning a return
of Rtd . Thirdly, the entrepreneur can pledge up to fraction of future output to
banks and borrow at interest rate Rtl in order to invest in her own production
2
This happens when the borrowing constraints are tight enough. See Aoki et al. (2009).
3
See, for example, Sargent (1987).
technology. In this case the leveraged rate of return is

at (1 )
: (14)
wt at =Rtl

By borrowing from banks secured by fraction of output, the entrepreneur can

…nance externally at =Rtl amount (equation (5)). Therefore the denominator
is the required downpayment for the unit labor cost. Finally, the entrepreneur
can invest in bubbles. In this case the rate of return is given by t+1 = t , where

t+1 stands for the market value of the bubble next period. When it bursts,
the return is zero.
High productivity entrepreneurs enjoy better returns on production so they
are the ones who borrow in equilibrium. When their borrowing constraints
bind the rate of return on wealth (r(aH )) is given by:

aH (1 )
r(aH ) = = Rtl : (15)
wt a =Rtl
H

From (5) and (4) the investment (employment) of a productive agent is given
by
zt
ht = : (16)
wt aH =Rtl

The entrepreneur saves a fraction of wealth zt and uses her entire savings as
a downpayment for wage payments to the workers she hires.
While the productive entrepreneurs borrow, the unproductive entrepre-
neurs make deposits. In addition, they have two other means of savings: un-
leveraged production and investing in bubbles. Notice that both deposits and
production are riskless in our environment. When

aL
Rtd > (17)
wt
low productivity agents are inactive in production. However when the credit
constraints on banks and borrowing entrepreneurs are tight enough, the pro-
ductive entrepreneurs cannot absorb all national saving. The supply of deposits
is limited and the low productivity technology may be viable in equilibrium.
In such case
aL
Rtd = (18)
wt

and the saver entrepreneurs use both bank deposits and their own production
technology to accumulate wealth.
Bubbles are risky. When savers invest in bubbles as well as deposits, the
arbitrage condition for bubbles is determined by the savers’state-contingent
wealth valuation
1 t+1 1
Et = Et Rtd ; (19)
cLt+1 t cLt+1

where 1=cLt+1 represents the shadow value of wealth at time t + 1 of the entre-
preneur who is unproductive at time t4 , where expectation operator is taken
over whether bubble survives or crashes.

b
t+1 with probability
t+1 = (20)
0 with probability 1

b
where t+1 is the market value of the bubble on survival.

3.2 Banks’optimal behaviour

Next, we characterise the optimal behaviour of a representative bank in our

economy. The problem of the bank can be represented in recursive form by
4
Namely, it is given by
1=cL
t+1 = (1
L
)Zt+1
H
where Zt+1 is given by equation (34)
using the bank’s value function representation as follows:

V (nt ) = max cbt + Et [ V (nt+1 ) + (1 ) nt+1 ] (21)

cbt ;dt ;bt ;mt

V (nt ) is the value of a bank with net worth nt which chooses current con-
sumption, deposits, bubbles and loans optimally. This value is equal to cur-
rent consumption and the expected future discounted value of bank net worth
Et [ V (nt+1 ) + (1 ) nt+1 ]. This value includes the continuation value of
being a banker - this happens only if the banker survives with probability .
With probability 1 , the banker receives the death shock and consumes his
entire net worth in the following period.
Because of risk neutrality, we can guess that the value of the bank is a
linear function of net worth nt

V (nt ) = t nt (22)

When Rtl > Rtd , the credit constraint (10) binds and consumption is postponed
until death. Then, with equation (10) binding, deposits are given by

t
dt = nt : (23)
1

Here t =(1 ) is the bank’s leverage.

Regarding the bank’s choice over bubbles and loans,

t+1
Et 1 + t+1 Et 1 + t+1 Rtl ; (24)
t

where expectation operator is again taken over the bubble surviving or not.
If equation (24) holds with strict inequality, then the bank will not invest in
bubbles (mbt = 0). When the bank invests in bubbles (24) must hold with
equality. By substituting (22), (23), and (24) into (21), t satis…es

Et (1 ) + Et t+1 Rtl
t = Rtl Rtd
: (25)
1 Et (1 ) + Et t+1 1

This expression states that the value of a unit of net worth for a banker
is equal to the value of the returns on its loan book (the numerator), suitably
boosted by leverage (the denominator). The banker issues one unit of loans
but the downpayment he has to make is only given by the denominator of (25)
because he can pledge some of the future expected excess returns from inter-
Rtl Rtd
mediation ( Et (1 ) + Et t+1 1
) to depositors who …nance a large
part of the loan outlay. Note that the above formulas show that t increases
with t+1 . This implies that the current leverage depends on the future fran-
chise value of the bank which is re‡ected by the leverage next period.5 It also
shows that t is an increasing function of the spread Rtl Rtd :

3.3 Workers’optimal behaviour

Workers are risk-neutral and consequently their consumption-savings behav-

iour is a knife-edge one. When the loan interest rate is lower than the rate
of time preference, workers want to borrow unlimited quantities. Because the
workers cannot operate the production technology, they cannot pledge collat-
eral to lenders. Hence workers cannot borrow.
Of course workers could save at the deposit rate but they only want to do
this when
Rtd > 1
: (26)

If this condition is not satis…ed, workes will consume their entire net worth
(…nancial wealth and labour income) and save nothing. Their labour supply
5
See Nikolov (2010), who considers a similar problem for …rms.
hst is given by
hst = wt : (27)

Because ours is a limited commitment economy, we guess and verify that Rtd <
1
at all times along the equilibrium paths we consider. Hence our workers
are hand-to-mouth consumers at all times.

3.4 Aggregation and market clearing

Let the total supply of the bubble asset be normalized to 1. In other words,

met + mbt = 1; (28)

where met and mbt , respectively, denote the shares of the bubble held by unpro-
ductive entrepreneurs and banks.
Let ZtH and ZtL , respectively, denote aggregate wealth of the productive and
unproductive entrepreneurs. Then we can characterise the aggregate equi-
librium as follows. From (16) the aggregate employment of the productive
entrepreneurs is given by

ZtH
HtH = : (29)
wt aH =Rtl

When (17) holds, the unproductive entrepreneurs are indi¤erent between mak-
ing deposits and producing, thus their aggregate saving is split as follows

HtL = ZtL Dt met t (30)

where Dt denotes aggregate deposit.

Let us turn to banks. Under the banks binding borrowing constraint, the
aggregate deposit is given by

t
Dt = Nt : (31)
(1 )

Notice that 1 fraction of banks exits in each period by liquidating all their
net worth. Therefore the aggregate net worth of the operating banks is given
by Nt . The aggregate balance sheet of the operating banks is given by

Dt + Nt = Bt + mbt t : (32)

Let us turn to the transition of state variables. Note that the unproductive
entrepreneurs become productive in the next period with probability n and
the productive entrepreneurs continues to be productive with probability 1 .
Their rates of return are given by (15) and (17). Therefore the net worth of
the productive entrepreneurs evolves from (15) and (13) as

H aH (1 )
Zt+1 = (1 ) ZH + n Rtd ZtL met t + met t+1 (33)
wt a =Rtl t
H

Similarly, the aggregate net worth of the unproductive entrepreneurs evolves

as

L aH (1 )
Zt+1 = H l
ZtH + (1 n ) Rtd ZtL met t + met t+1 (34)
wt a =Rt

From aggregating production function, aggregate output is given by

Yt = aH HtH 1 + aL HtL 1 : (35)

Finally, aggregate bank net worth is given by

 Nt+1 = Rtl Bt + mbt t+1 Rtd Dt (36)

The markets for goods, labour, capital, loan and deposit must clear. Goods
market clearing implies that aggregate saving must equal to aggregate invest-
ment.
(ZtH + ZtL ) + Nt = w(HtH + HtL ) + t: (37)

From (27), labour market clearing implies

wt = HtH + HtL : (38)

Now equations (18), (19), (24), (25), (28)-(38) jointly determine 15 variables
Rtd ; Rtl , wt , HtH , HtL , Yt ; t,
H
Dt , Bt , Zt+1 L
, Zt+1 , Nt+1 , t; met ; mbt with three
states ZtH , ZtL , Nt 6 . At t = 0, Z0H is given by (33).

De…nition 1 Competitive bubbly equilibrium without government is a sequence

1 1
of decision rules HtH ; HtL ; Yt ; Dt ; Bt ; met ; mbt t=0
, H
aggregate state variables Zt+1 L
; Zt+1 ; Nt+1 t=0
1
and a price sequence Rtd ; Rtl ; wt ; t; t t=0 such that: (i) entrepreneurs, banks
1
and workers optimally choose decision rules HtH ; HtL ; Yt ; Dt ; Bt ; met ; mbt t=0

taking the evolution of aggregate states, prices and idiosyncratic productivity

1
opportunities as given; (ii) the price sequence Rtd ; Rtl ; wt ; t; t t=0 clears the
goods, labor, capital, loan, bubble and deposit markets and (iii) the equilib-
H L 1
rium evolution of state variables Zt+1 ; Zt+1 ; Nt+1 t=0
is consistent with the
individual choices of entrepreneurs, banks and workers and with the exogenous
evolution of productive opportunities at the individual …rm level.
6
By Warlas law one of these equations is redundant.
3.5 Calibration

We have 8 parameters ; aH =aL ; ; n; ; ; ; we need to calibrate before

we proceed to examine the quantitative predictions of our model economy.
There is little consensus regarding , the Frisch elasticity of labour supply.
Micro-data evidence suggests a value close to zero based on the labour supply
behavior of primary earners. The real business cycles literature usually sets a
much higher value in the region of 3 or even higher. The di¤erences is justi…ed
by the presence of labour market frictions that ensure that aggregate labour
is highly elastic even though individuals are relatively unwilling to vary their
market hours over time. Gertler and Kiyotaki (2010) make this argument and
set the Frisch elasticity to 10 in their model. We pick a value of = 5, which
is within the range set in calibrating macro models.
aH =aL is an important parameter, whose value is also highly uncertain. As
studies such as Bernard et al. (2003) and Syverson (2004) have documented,
the dispersion of plant level productivity in US manufacturing is enormous,
with the most productive plants having more than 4 times more productive
compared to the least productive. But as Aoki et al. (2009) argue, it is hard
to believe that such a huge dispersion of productivity levels is entirely due to
the presence of credit constraints. More likely, inputs could be mismeasured in
a number of ways. For example, intangible assets such as managerial quality
could be an important missing input which could explain some of the huge
di¤erences in measured plant level TFP. Following Aoki et al. (2009) we set
a value for aH =aL = 1:1 implying a substantial cross-sectional dispersion in
plant level TFP in the model.
We calibrate the remaining 6 parameters in order to match the steady state
predictions of the model in the absence of bubbles to 7 moments in the US
data. These are (1) the real loan rate minus the growth rate of real GDP and
minus intermediation costs; (2) the real deposit rate minus the growth rate of
real GDP; (3) commercial bank leverage; (4) average corporate leverage; (5)
average leverage for highly leveraged corporates; (6) the rate of return on bank
equity and (7) the ratio of M2 to GDP.
Calibration targets (1) and (2) deserve further discussion. For simplicity, in
our model we assume there is no growth and no intermediation costs. In reality
these two assumptions of course do not hold. Growth in the US economy
has averaged close to 3% per annum since the second world war. Since we
are interested in the dynamic e¢ ciency of the investments of US savers and
banks, we want to know whether the real return on these investments exceeds
the economy’s growth rate. This is why we subtract the real growth rate from
the real return on deposits and loans.
In addition, when it comes to evaluating the dynamic e¢ ciency of banks’
loan investments, we need to take intermediation costs into account. FDIC
data on US commercial banks’ cash ‡ow sources reveals that there are sub-
stantial intermediation costs (80% of those are labour costs). Here we assume
that all of these costs arise due to loan issuance rather than deposit taking.
This assumption is not entirely unreasonable given the labour intensive nature
of arranging loans, monitoring them and then recovering them if they become
non-performing. It does, however, err on the side of assuming that banks’real
loan returns are more dynamically ine¢ cient. We subtract these loan costs
from banks’real loan returns to get the …nal numbers shown in Table 1 below.
Full details of data sources and construction are available in Appendix A.
 Table 1: Model and data moments
Moment (Model concept) Data Model

Real deposit rate - real GDP growth (Rd ) 0.950 0.971

Real loan rate - real GDP growth - costs/Assets (Rl ) 0.982 0.982

Ratio of M2 to GDP (D=Y ) 0.500 0.464

Bank leverage (D=N ) 10.00 10.00

Average corporate leverage (L=Z) 0.500 0.530

Leverage of indebted corporates (L=(sZ)) 2.000 2.000

(Rl Rd )
Bank rate of return on equity (Rtl + t (1 t ) t ) 1.100 1.103

Table 2 below presents the values of the parameters chosen to match the
moments.

Table 2: Baseline calibration

Parameter Value

0.167

n 0.011

aH =aL 1.100

5.000

0.622

0.788

0.907

0.958

3.6 Bubbly equilibria without government

3.6.1 Credit frictions and existence of bubbly equilibria

Now we characterise the deposit rate Rtd and loan rate Rtl in the steady state
without bubbles and discuss when bubbles can circulate. For bubbles to cir-
culate, two conditions are needed. Firstly, bubbles should be attractive. For
savers, the opportunity cost of investing in bubbles is the deposite rate, and
for banks it is the lending rate. Secondly, bubbles should be a¤ordable. This
implies that the rate of return of bubbles conditional on survival is no larger
than the rate of economic growth. In our economy the gross growth rate of
the economy is unity because there is no technological growth.
As a benchmark case here we show the condition for the existence of bubbly
equilibria when = 1. Credit frictions suppress the interest rates and those
1
rates are lower than when the credit constraints bind.7 Similarly to Farhi
and Tirole (2011), when = 1 whether a bubbly steady state exists and who
owns bubbles depend on whether the two interest rates are lower than the
growth rate in the ‘no bubbles’steady state8 .
In our economy, the severity of credit frictions is represented by two para-
meters, and : Figure 1a shows the region of and in which the deposit
rate is less than one and low productivity agents produce in equilibrium (the
red area). In this case, the savers (unproductive entrepreneurs) have incentive
to buy bubbles in order to boost the rate of return they receive on their sav-
ings. The blue parts of the graph show parts of the parameter space where the
economy is very credit constrained. At such low values of and low produc-
tivity entrepreneurs are active but wages are so low that even such ine¢ cient
projects deliver a rate of return greater than unity. As a result, savers have
no incentive to hold bubbles in such economies. The white parts of the graph
7
See Aoki et al. (2009)) for the general discussion of the relationship between the interest
rate and credit frictions.
8
In Martin and Ventura (2011a) emergence of bubbles itself can creates a “pocket of dy-
namic ine¢ ciency" so that bubbly equilibria may exist even when the interest rate is greater
than the growth rate in no-bubble equilibrium. This result depends on their model setting
that credit-constrained agents can create bubbles in subsequent periods. We abstracts from
new bubble creation. Because of this assumption, the condition for existence of bubbly
equilibria reduces to whether the interest rates in no-bubble equilibrium are smaller than
the growth rate.
(very high values of and ) shows parts of the parameter space where low
productivity entrepreneurs do not produce because the …nancial system is well
developed. Here again, the rate of return on deposits is greater than unity and
savers have no incentive to hold bubbles. So it should be clear from Figure
1b that the conditions for the existence of bubbles is satis…ed at intermediate
levels of …nancial development.
[Figure 1a here]
The red area of Figure 1b shows the region in which the loan rate is less
than one. Then the banks have an incentive to buy bubbles. It is natural
that the part of the parameter space where banks bubbles can exist is more
limited compared to the parts of the parameter space where saver bubbles
exist. Because banks’ borrowing constraints bind, this introduces a positive
spread between lending and deposit rates. Hence the parameter space where
bank bubbles exist is a subset of the space where savers have an incentive
to invest in bubbly assets. Since the deposit rate is always lower than the
loan rate, the savers also have incentive to hold bubbles at these parameter
values. In equilibrium, it turns out that only the savers have bubbles because
their arbitrage implies that the rate of return on bubbles must be equal to
the deposit rate, which is lower than the loan rate. Therefore the banks are
crowded out from the market for the bubble.
[Figure 1b here]

3.6.2 Who holds risky bubbles: stochastic steady state without

government policy

In the previous section we show that in the deterministic steady state, bank
deposits and the bubble become perfect substitute and only the savers hold
bubbles. Now we allow for risky bubbles and consider an environment in
which bubbles only survive with a certain probability ( < 1.) We focus on
a stochastic steady state in which bubbles are traded at a positive value and
all the endogenous variables including bubbles are constant. In such a steady
state, agents take the probability of bubble bursting into account when they
make their investment decision. In that case, bubbles, loans and deposits
are no longer perfect substitutes. Also, note that the savers and banks have
di¤erent preferences for risk. Then it is of interest to analyse who holds risky
bubbles.
Table 3 below shows the share of bubbles held by banks. The table looks at
banks’participation under di¤erent probabilities of the bubble’s survival ( ).
The dashes in the table show economies in which the bubble is too risky for
a stochastic steady state to exist in which such a bubble trades at a positive
price. This occurs when the survival probability is 0.965. Such a risky bubble
would need to command a very high return conditional on survival in order
to compensate its holders for the losses when it eventually collapses. But
when this ’risk premium’is too high, the bubble follows an explosive path and
violates feasibility constraints in …nite time. Consequently such a bubble is
never valued in equilibrium.
It turns out that banks mostly stay out of the bubble market. Their hold-
ings are zero for very high or very low values of and barely reach 1.6% of
the total stock of bubbly assets when = 0:985.
 Table 3: Bank bubble holdings in the stochastic steady state
= 0:965 = 0:975 = 0:985 = 0:995

Fraction of bubbles held by banks - 0.000 0.016 0.000

Ratio of bubbles to total wealth - 0.118 0.317 0.403

Expected saver loss (fraction of wealth) - 0.004 0.006 0.003

Expected bank loss (fraction of wealth) - 0.000 0.002 0.000

b s
t+1 = t+1 - 0.984 1.060 0.987

The table also shows the size of bubbles as percentage of total wealth. It
shows that bubbles are not large. As a result, the expected losses when bubbles
burst are also small for both savers and banks.
The reason why banks hold very few bubbles is straightforward. When
banks’borrowing constraints bind, Rl > Rd and banks’investment opportuni-
ties (Rl ) are superior to savers’(Rd ). It is natural therefore that savers should
be more willing to take on the risk of saving via the bubble rather than via
deposits or own production.
Here, it is worth noting that banks’linear utility does not necessarily make
them the natural risk-bearers in equilibrium. Due to binding borrowing con-
straints and time-varying loan spreads, banks behave as if they are risk-averse.
Similar to Gertler and Karadi (2009), periods when bank capital is low are
periods when lending spreads are high. Therefore banks’ marginal value of
an extra unit of net worth ( t ) is high when they choose to hold bubbles and
bubbles burst. Indeed, Table 3 shows that, when the banks choose to hold
bubbles ( = 0:9875), the marginal value when bubbles burst at time t + 1
b s
( t+1 ), is larger than the marginal value when bubbles survive ( t+1 ). Conse-
quently they have a strong pro…t motive to ensure some stability in their net
worth. Holding a very large quantity of the bubbly asset is not optimal since
the excess return (which is pinned down by savers’…rst order condition too)
is not big enough to justify the risks.

4 Equilibrium with government bailouts

4.1 Modi…cation of optimal behavior

Now we characterize the economy with an active government. Introducting

government bailout policy changes agents’ behaviour in an important way.
Firstly, consider productive entrepreneurs. We continue to focus our analysis
on the case in which they borrow up to their borrowing constraint. However,
their rate of return of production with maximum borrowing (after taxation) is
now given by
aH (1 et )
(1 t+1 ) ; (39)
wt et aH =Rtl

where et is the maximum amount that can be pledged to private creditors after
taking into account expected tax payments. By borrowing from banks secured
by
et = Et
6
t+1
1 Et t+1

fraction of output, the entrepreneur can …nance externally et at =Rtl amount

(equation (5)). The fraction of output that can be pledged to outside private
creditors (et ) is less than the total amount of pledgable returns . This is
because the government is expected to call on the resources of the …rm in the
form of taxation Et t+1 . Because the government holds the senior claim to the
…rm’s pledgable returns, private creditors can only rely on the residual fraction
Et t+1 which is boosted to some extent by the fact that debt repayments
are shielded from taxation (hence the division by 1 Et t+1 ). Also, note that
the rate of return on wealth is potentially subject to taxation which is why
(39) contains a 1 t+1 term.
 In order for the productive entrepreneurs to borrow up to limit, the follow-
ing inequalities must hold:

aH (1 e) 1 1
Et f(1 t+1 )g > Rtd Et (1 t+1 )
wt eaH =Rtl cH
t+1 cH
t+1

and

aH (1 e) 1 t+1 1
Et f(1 t+1 )g > Et (1 t+1 ) ;
wt eaH =Rtl cH
t+1 t cH
t+1

where 1=cH
t+1 is the shadow value of wealth at time t + 1 of the entrepreneur

who is productive at time t9 . The expectation operator is taken over whether

bubbles survive or not (and therefore taxes are levied or not).
The equilibrium portfolio choice of bubble owners also changes in an im-
portant way. Since they are subject to taxation, the arbitrage condition for
the savers is now given by

1 t+1 1
Et (1 t+1 ) = Et (1 t+1 ) Rtd (40)
cLt+1 t cLt+1

The arbitrage condition for the bank is also modi…ed:

~ t+1
Et 1 + t+1 = Et 1 + t+1 Rtl ; (41)
t

where
b
t+1 with probability
~ t+1 = : (42)
t+1 t with probability 1

Notice that when the bubble bursts, banks receive a fraction t+1 of their origi-
9
Namely, it is given by
1=cH
t+1 = (1
H
)Zt+1
H
where Zt+1 is given by equation (33)
nal bubble investment. This is due to a bailout payment from the government.
The portfolio balance condition for entrepreneurs di¤ers in two crucial as-
pects from (41) above. First of all, the state income valuations di¤er due
to the di¤erent preferences of bankers (assumed to be linear in consumption)
and entrepreneurs (assumed to be logarithmic). But secondly, the risks faced
by the groups of potential bubble investors are very di¤erent because of their
di¤erent access to government bailouts. Here we assume that banks may get
at least partially bailed out in the event of losses on direct bubble holdings
( t+1 > 0) whereas ordinary savers will not.

4.2 Aggregation and market clearing

The employment of the productive entrepreneurs ZtH is given by (29) but is

replaced by ~. Net worth of the productive entrepreneurs is modi…ed as
" #
H aH (1 e)
Zt+1 = (1 t+1 ) (1 ) H l
ZtH + n Rtd ZtL met t + met t+1 :
wt a =Rt
(43)
b
Notice that t+1 is equal to t+1 > 0 with probability . In this case t+1 = 0.
With probability 1 , t+1 = 0 and t+1 > 0. Similarly, the aggregate net
worth of the unproductive entrepreneurs evolves as
" #
L a (1 e)
H
Zt+1 = (1 t+1 ) H l
ZtH + (1 n ) Rtd ZtL met t + met t+1 :
wt a =Rt
(44)
Aggregate bank net worth is given by

Nt+1 = Rtl Bt + mbt ~ t+1 Rtd Dt (45)

Finally the government’s budget constraint implies that taxes will be levied
on entrepreneurs in order to bail out the banks’s bubble holdings in the event
of a bust. This implies that

c b
t+1 = t+1 mt t (46)

where mbt t is the value of the bank’s bubble purchase in period t. The tax
rate is zero whenever the bubble survives and no bailout is needed. The other
equilibrium conditions remain the same as Section 3.
Now equations (18), (25), (28)-(32), (35), (37), (38), (40)-(46) jointly de-
termine 16 variables Rtd ; Rtl , wt , HtH , HtL , Yt ; t,
H
Dt , Bt , Zt+1 L
, Zt+1 , Nt+1 , t;

met ; mbt c
t with three states ZtH , ZtL , Nt 10 . At t = 0, Z0H is given by (33).

De…nition 2 Competitive bubbly equilibrium with government is a sequence of

1 1
decision rules HtH ; HtL ; Yt ; Dt ; Bt ; met ; mbt t=0
, H
aggregate state variables Zt+1 L
; Zt+1 ; Nt+1 t=0
1
and a price sequence Rtd ; Rtl ; wt ; t; t t=0 and government policy c
t+1 such
that: (i) entrepreneurs, banks and workers optimally choose decision rules
1
HtH ; HtL ; Yt ; Dt ; Bt ; met ; mbt t=0
taking the evolution of aggregate states, prices,
government policy and idiosyncratic productivity opportunities as given; (ii)
1
the price sequence Rtd ; Rtl ; wt ; t; t t=0 clears the goods, labor, capital, loan,
c
bubble and deposit markets and (iii) governmenta taxes t satis…es the gov-
ernment budget constraint given exogenous t; (iv) the equilibrium evolution of
H L 1
state variables Zt+1 ; Zt+1 ; Nt+1 t=0
is consistent with the individual choices
of entrepreneurs, banks and workers and with the exogenous evolution of pro-
ductive opportunities at the individual …rm level.
10
By Warlas law one of these equations is redundant.
4.3 Moral Hazard, Bubble Ownership and Ampli…ca-

tion

Having described the way anticipation of bailout policy modi…es the structure
of our model economy, we proceed to examine its e¤ect on the economic allo-
cation by using numerical simulations with our calibrated model. Our focus is
on two main questions. What determines banks’ownership of bubbly assets in
equilibrium? How does bubble ownership a¤ect the size of the bubble-driven
boom-bust cycle?

4.3.1 Moral Hazard Comparative Statics

We start by considering the e¤ect of the …nancial safety net on the bubbly
equilibrium in which bubbles are only expected to burst with probability 0.5%
per annum (i.e. bursts happen once every 200 years). The results are presented
in Table 4 below.

Table 4: Bailouts and bank risk with low probability of bursting ( = 0:995)
= 0:00 = 1=3 = 2=3

Fraction of bubbles held by banks 0.000 0.000 0.000

Bubble to GDP ratio 0.674 0.674 0.674

E(Bubble Returnjbank) - Rl11 -0.004 -0.002 -0.001

E(Bubble Returnjsaver) - Rd 0.006 0.006 0.006

Bank NW/GDP (pre-crash) 0.057 0.057 0.057

Bank Loss/GDP 0.000 0.000 0.000

% fall in bank NW12 0.000 0.000 0.000

11
This is the expected return to a bank of holding the bubble. It may di¤er from the
expected return for the saver because of bailouts.
12
The percentage fall in bank net worth is computed after the receipt of government
assistance. In other words, in many of the more extreme scenarios considered, the banking
system would have negative net worth without a government bailout.
 The …rst row of the table shows that the percentage of bubbles held by
banks remains close to zero even for relatively generous …nancial safety nets
(high values of ). There, the presence of …nancing constraints ensures that the
bank enjoys a spread of lending over deposit rates, giving rise to a ’franchise
value’of the bank. When ordinary savers hold bubbles which are not expected
to burst (or are expected to burst with a very low probability), the return on
bubbles and deposits must be approximately equal in equilibrium. For banks,
in contrast, the indi¤erence condition implies an equality between the loan
rate and the return on bubbles. When the spread Rl Rd , is large enough,
banks do not …nd it pro…table to hold bubbles under this value of .
We also compute the e¤ects of the bubble’s collapse on banks’capital po-
sitions. The 6th and 7th rows of Table 4 shows banks’ loss of net worth as
percentage of GDP and the percentage fall in bank net worth when the bubbles
burst. Since banks are not exposed to bubbly assets, bank losses are zero. De-
spite the large bubble size (around 67.5% of GDP), the bursting of the bubble
does not cause a systemic banking crisis.
 Table 5: Bailouts and bank risk with medium probability of bursting
( = 0:98)
= 0:00 = 1=3 = 2=3

Fraction of bubbles held by banks 0.000 0.123 0.515

Bubble to GDP ratio 0.296 0.385 0.415

E(Bubble Returnjbank) - Rl13 -0.002 0.007 0.010

E(Bubble Returnjsaver) - Rd 0.010 0.011 0.005

Bank NW/GDP (pre-crash) 0.055 0.067 0.092

Bank Loss/GDP 0.000 0.032 0.071

% fall in bank NW14 0.000 0.469 0.779

In Table 5 above we consider the e¤ect of the …nancial safety net when the
probability of the bubble bursting is equal to 2% (i.e. the bubble’s expected
life span is 50 years). Here again banks do not hold bubbles when = 0.
Expanding the …nancial safety net (increasing towards unity) increases the
incentive for banks to hold bubbles because it shields them from an increasing
fraction of the potential losses. The share of bubbles held by banks increases
to 51.5% as rises to 2=3 and the size of the bubble grows from 30% to 41.5%
of GDP. As the banks’bubble holdings grow, the banking sector expands to
absorb these and take advantage of the government guarantee on its risky bub-
ble holdings. The net worth of the banking system relative to GDP increases
from 5.5% to 9.2% as increases from 0 to 2=3.
As banks’bubble holdings expand, bank risk grows substantially. The last
two rows of Table 5 presents some statistics measuring the impact on bank
balance sheets when the bubble bursts. We see that as the …nancial safety net
13
This is the expected return to a bank of holding the bubble. It may di¤er from the
expected return for the saver because of bailouts.
14
The percentage fall in bank net worth is computed after the receipt of government
assistance. In other words, in many of the more extreme scenarios considered, the banking
system would have negative net worth without a government bailout.
expands, bank losses relative to GDP increase from zero ( = 0) to 7.1% of
GDP ( = 2=3). Bank capital also experiences much larger falls during the
crisis when is high. For example, when = 2=3, bank capital falls by 78%.
In the event that the expected government assistance does not materialise, the
banking system would be deeply insolvent.

Table 6: Bailouts and bank risk with high probability of bursting ( = 0:965)
= 0:00 = 1=3 = 2=3

Fraction of bubbles held by banks - - 1.000

Bubble to GDP ratio - - 0.254

E(Bubble Returnjbank) - Rl15 - - -0.001

E(Bubble Returnjsaver) - Rd - - 0.015

Bank NW/GDP (pre-crash) - - 0.057

Bank Loss/GDP - - 0.109

% fall in bank NW16 - - 0.774

Finally, Table 6 presents the case of a bubble with a 3.5% probability of

bursting (expected life span of about 30 years). As is explained in Table 3,
such a bubble would not be valued in the absence of a …nancial safety net.
The dashes in the table show economies in which the bubble is too risky for
a stochastic steady state to exist in which such a bubble trades at a positive
price.
However, once rises to 2=3, the equilibrium changes. The insurance avail-
able to banks means that these bubbles are now su¢ ciently safe for them and
they no longer require such a high risk premium in order to hold them. This
ensures that for high values of a stochastic steady state with bubbles exists
15
This is the expected return to a bank of holding the bubble. It may di¤er from the
expected return for the saver because of bailouts.
16
The percentage fall in bank net worth is computed after the receipt of government
assistance. In other words, in many of the more extreme scenarios considered, the banking
system would have negative net worth without a government bailout.
even at high values of the probability of bursting ( ). The value of the bubble
(entirely held by banks) rises sharply to 25% of GDP and so does bank risk.
Potential losses reach 10% of GDP and bank capital falls by more than three
quarters in the event of a crisis.
Table 6 reveals an interesting and important result. The introduction of
a …nancial safety net widens the range of bubbles that exist in equilibrium.
Fragile bubbles now exist and trade at potentially large valuations amongst
banks who hold them due to the explicit or implicit guarantees given to them by
governments and central banks. The result that bailout anticipation increases
the valuation of risky assets is of course well-known in the literature (see for
example Kareken and Wallace (1978)). The novel contribution of our paper is
to show how risk-shifting can relax the conditions for the existence of stochastic
rational bubbles.

4.3.2 Bubble ownership and ampli…cation

In the previous subsection we showed how the …nancial safety net a¤ected bank
bubble ownership and bank losses in the event of the bubble collapsing. We
now examine what e¤ect moral hazard and bubble ownership have on the real
e¤ect of the bubble.
Figure 3 below compares the dynamics of the economy starting at the
stochastic steady state and tracking the economy’s evolution following the
bubble’s collapse. We compare two scenarios. In one, a generous …nancial
safety net ( = 2=3) ensures that the bubble asset is held only by banks. In the
other, there is no …nancial safety net ( = 0) and the bubble is optimally held
by savers while banks stay out. In each case, we set the probability that the
bubble survives into the following period in order to generate a bubble which
is exactly 20% of GDP. Therefore the di¤erence between the two scenarios
shown in Figure 3 is due to di¤erent bubble ownership rather than di¤erent
bubble size. The vertical axis of the …gure shows the percentage deviation of
each variable from its no-bubble steady state value.
[Figure 3 here]
The main feature of the simulations shown in the …gure is that the economy
with more bank guarantees experiences a more volatile path for output and net
worth during the bubbly episode. While the collapse of the saver-held bubble
actually generates an expansion in output, output falls under the bank bubble.
This di¤erence comes from the degree of banks’ exposure to bubbly assets.
When bubbles collapse, banks lose a large portion of net worth, causing a sharp
lending contraction and an increase in the lending-deposit rate spread. This
results in a credit crunch and pushes down the investment of the productive
entrepreneurs17 .
Table 7 below tries to go deeper into the underlying mechanism which
generates the large positive real e¤ects of bank bubbles during their survival.
The table decomposes the increase in output relative to the bubbleless steady
state according to the following identity

1 L
Y = aL HtL + aH HtH a wt HtL + HtH + aH aL wt HtH (47)
wt

which shows that output is determined by total investment wt HtL + HtH and
the investment of productive agents wt HtH as well as the impact of the cost of
1
employment wt
holding investment …xed18 . We can use the aggregate resouce
constraint (37) to substitute total investment wt HtL + HtH out from (47)
17
Output expands after bubbles collapse in the case of saver-held bubble because the savers
increase their own ine¢ cient production. Bursting of bubbles causes shortage of means of
savings. Then the savers increase their production because this represents another means
of savings.
18
Other things equal, higher wages reduce output and employment because entrepreneurs
can only a¤ord to operate on a smaller scale.
and the balance sheet identity of the productive agents in order to substitute
wt HtH out from (47). The resulting expression is given by (48) below. This is
what we base our output decomposition on.

1 L
Y = a ( Zt + Nt ) aL t + aH aL ( st Zt + Lt ) (48)
wt

Here st ZtH =Zt represents the share of wealth in the hands of productive
agents. The aL ( Zt + Nt ) is the ’liquidity e¤ect ’ stressed by Farhi and
Tirole (2011). Zt + Nt is the economy’s aggregate saving (comprising of
total entrepreneurial saving and bank saving). Bubbles help to increase wealth
and (other things equal) this tends to boost investment. The negative aL t

term is the traditional ’bubble crowding out’e¤ect: when the economy (both
savers and bankers) hold bubbles, they hold fewer real productive assets and
this (other things equal) tends to reduce output.
The aH aL ZtH + Lt term is the investment composition e¤ect dis-
cussed by Martin and Ventura (2011a). In a Kiyotaki-Moore environment like
ours, a greater share of wealth in the hands of productive agents (st ) and
improved access to bank credit (Lt ) lead to a re-allocation of resources from
unproductive to productive agents. Even holding total investment constant,
this would tend to increase output because aggregate TFP rises due to the fact
that factors of production are used by more productive …rms. Finally, higher
1
labour costs reduce output for given investment: this is captured by the wt

term in (48).
The di¤erent columns of the table correspond to di¤erent values of but
in each case we vary the probability of bubble survival ( ) in order to ensure
that the bubble is always equal to 20% of GDP.19 This allows us to focus on
19
When = 0, = 0:978; when = 1=3, = 0:976; when = 2=3, = 0:963.
 the real e¤ect of bubble ownership holding bubble size …xed.

Table 7: Decomposing the real e¤ects of bubbles

=0 = 1=3 = 2=3

Bank bubble holdings (fraction of tot. bubble value) 0.00 0.11 1.00

% increase in output relative to ’no bubble’SS 1.03 1.32 3.11

(1) Liquidity e¤ect, % 19.64 20.01 21.78

(2) Bubble ’crowding out’e¤ect, % - 19.54 - 19.67 - 19.99

(3) Investment composition e¤ect, % 0.95 1.04 1.65

...of which

(3.1.) Productive net worth, % 0.31 0.34 0.54

(3.2.) Bank lending, % 0.64 0.70 1.11

(4) Labour costs, % - 0.02 - 0.06 - 0.32

Note: Rows (1)-(4) show the percentage point contributions of various channels to the
total increase in output relative to the bubbleless steady state.

The table shows that the larger real e¤ect of bank-held bubbles is largely
due to the fact that it generates a bigger liquidity (wealth) e¤ect and due to
the fact that it stimulates bank lending to a greater extent.

Table 8: Cost of funds (percentage point deviation from bubbleless steady

state)
=0 = 1=3 = 2=3

Bank net worth (% increase) 14.43 25.03 113.7

Bank lending to entrepreneurs (% increase) 13.03 14.35 22.69

Lending-Deposit spread (percentage points) 0.02 - 0.01 - 0.22

Table 8 above provides some further intuition on the reasons behind the
expansion of credit in the bank bubble case. The …rst row of the table clearly
shows that the increase in bank capital relative to the bubbleless steady state
is much larger in the case when banks hold the bubble. This is because banks’
ownership of the risky bubble asset boosts the rate of return on wealth while the
bubble survives. Bubbles carry a risk premium to compensate the holder for
the risk of bursting. Conditional upon no crisis occuring, this risk premium
allows the bubble holder to expand its balance sheet substantially. When
the bubble holder is a bank, the expansion of its balance sheet also leads to
greater credit supply. This is shown in the second row of Table 8 - lending to
corporates increases by more when banks are the bubble holders (the = 2=3
column) compared to the case when savers are the holders (the …rst and second
columns).
This credit expansion lowers lending-deposit spreads (the third row of the
table) helping to boost the leverage of productive entrepreneurs and allowing
them to expand the scale of their operations. Note again, that this e¤ect is only
present when banks hold the bubble (the third column of the table). When
savers are the main (or only) bubble holders, the expansion in bank lending is
more limited and lending spreads do not change much.

5 Conclusions

In this paper we build a model with explicit …nancial intermediation in which

rational asset price bubbles arise due to credit frictions. Our aim is to model
the interaction between banks’net worth and the value of asset price bubbles
which has been a key feature in many historical episodes of boom and bust.
This interaction turns out to be very important in understanding the real
impact of asset price bubbles on the wider economy.
We show that in the baseline version of our economy, banks have better
investment opportunity than savers. As a result, savers have a stronger in-
centive to ’search for yield’ by holding bubbles. Nevertheless, we invoke the
possibility that in some cases, banks may be the only bubble holders due to
technological or informational constraints on savers’direct participation. We
then proceed to compare the way the bubble’s impact on the real economy
depends on who the main holder is. It turns out that the identity of the holder
is very important. When banks are the bubble-holders, this ampli…es both the
boom while the bubble survives and the bust when it …nally bursts. This is
because the bubble delivers a high return conditional on not bursting in order
to compensate for the tail risk of total loss. During the boom phase, this high
return leads to very high pro…ts for the bubble holder. When the banks are
the recipients of these pro…ts, this helps to expand credit and boost the net
worth of other credit constrained agents. When the savers are the recipients
of these pro…ts, no such relaxation of borrowing constraints occurs and the
impact of the bubble is relatively limited.
In the second part of the paper, we explore the possibility that government
guarantees a part of banks’ bubble investments against loss. This can make
banks the natural bubble holders even in an environment of full participation
in the bubble market by all agents. When bailouts are expected, the banking
system expands rapidly to exploit its advantage in holding the bubble, bidding
up its price very substantially. This subsidy to bank risk-taking leads to a
dramatic increase in the riskiness of bank balance sheets: bank capital falls
very sharply when the bubble eventually bursts. Finally, banks’ diminished
sensitivity to risk increases the range of parameter values for which bubbly
equilibria exist. In particular, bank-held bubbles with a very high probability
of bursting exist only when the …nancial safety net is su¢ ciently generous.
6 Appendix A: Data Sources

In this section we provide details of the sources of the data used for calibrating
the model. This is given in Table A1 below:

Table A1:

Theor. concept Data concept Source

Real bank loan rate Real prime loan rate-GDP FRB, Table H.15,
growth-costs FDIC, BEA

Real deposit rate Real M2 own rate - GDP FRED

growth

Expected in‡ation Average CPI in‡ation (All FRED

Urban Consumers)

Expected real GDP growth Average real GDP growth FRED

(chained measure)

Deposit stock M2 FRED

Nominal GDP Nominal GDP FRED

Bank leverage Bank Debt Liabilities/Bank FRB, Table H.8

Net Worth

Average corporate leverage Corporate Debt/Corporate Welch (2004)

Net Worth

Leverage of indebted corpo- Debt/Corp Net Worth for Welch (2004)

rates high leverage corporates

Bank rate of return on eq- Bank rate of return on eq- FDIC

uity uity
7 Appendix B: Computational procedure

In this Appendix, we outline the computational procedure we use to solve for

the stochastic steady state in the general case of full participation in the bubble
market for all agents.
(1) Start by solving the deterministic steady state of the model in the
absence of bubbles. This is where the economy will converge to after the bubble
bursts because we do not consider any fundamental shocks in our simulations.
(2) We have one signi…cant forward looking variable in the model t and
we need to characterise the way it evolves after the bubble bursts. This will
give us an expected value of t conditional on bursting. We start by guessing
such a path.
(3) Solve the system of equations that de…nes the stochastic steady state,
taking the future assumed path of t as given. We use Matlab’s own non-linear
equation solver fsolve.m in order to do this.
Within step (3) we need to determine which regime our economy operates
in. In particular we check whether (a) only savers hold bubbles, (b) only banks
hold bubbles or (c) both hold bubbles. In addition we need to check whether
unproductive agents are active producers in equilibrium20 . We do this by trial
and error:
(3.1) First we check who holds the bubble
(a) First we try to solve the model under the assumption that only savers
hold bubbles. This means that savers’s …rst order condition determines the
equilibrium growth rate of bubbles conditional upon survival. Once we solve
the model under this assumption we check whether banks would like to invest
in bubbly assets on the margin by examining the di¤erence in the expected
20
There are a multitude of other regimes the economy can operate in, depending on
whether the banks’ and the entrepreneurs’ borrowing constraints bind or not. We only
consider parameter values for which both of these constraints bind.
utility for bankers from investing in one unit of bubbles and loans. If we …nd
that banks stictly prefer to invest in loans on the margin, this con…rms our
assumption that only savers buy bubbles and we move on to step (4). If we
…nd that banks strictly prefer to invest in bubbles, then not only savers hold
bubbles in equilibrium and we need to try a di¤erent assumption.
(b) Only banks hold bubbles. This means that banks’s …rst order condition
determines the equilibrium growth rate of bubbles conditional upon survival.
Once we solve the model under this assumption we check whether savers would
like to invest in bubbly assets on the margin by examining the di¤erence in the
expected utility for savers from investing in one unit of bubbles and deposits.
If we …nd that savers stictly prefer to invest in deposits on the margin, this
con…rms our assumption that only banks buy bubbles and we move on to step
(4). If we …nd that savers strictly prefer to invest in bubbles, then not only
banks hold bubbles in equilibrium and we need to try a di¤erent assumption.
(c) Both hold bubbles. This means that both banks and savers hold bubbles
and each holds a strictly positive fraction of the total bubble stock.
(3.2) Check whether the low productivity agents produce in equilibrium
(a) We solve the model, assuming that the deposit rate is equal to the
unleveraged rate of return on the low productivity technology. We then check
if the investment of the low productivity …rms is positive. If not we reject this
equilibrium and assume that low productivity agents are inactive as producers.
(b) We solve the model, assuming that the employment (investment) of
low productivity entrepreneurs is zero. We check the resulting equilibrium
to ensure that the deposit rate is higher than the unleveraged rate of return
on the low productivity technology. If this condition holds, this con…rms our
initial assumption that saver entrepreneurs prefer not to produce.
(4) Use the output in (3) to compute the initial conditions of the econ-
omy when the bubble bursts. We then solve for a perfect foresight path that
converges to the no-bubble deterministic steady state, following the bubble’s
collapse. We do this by iterating on the t path after the bubble’s collapse
until it converges
(5) Return to step (1) and keep iterating until the entire path of t (includ-
ing its stochastic steady state value) has converged to within error tolerance.
References

Aoki, K., Benigno, G., Kiyotaki, N., 2009. Adjusting to capital liberalization.
Working Paper, LSE.

Arce, O., Lopez-Salido, D., 2008. Housing bubbles. Banco de Espania Working
Paper No. 0815.

Bernard, A., Eaton, J., Jensen, B., Kortum, S., 2003. Plants and productivity
in international trade. American Economic Review 93(4), 1268–1290.

Caballero, R., Krishnamurthy, A., 2006. Bubbles and capital ‡ow volatility:
Causes and risk management. Journal of Monetary Economics 53, 35–53.

Farhi, E., Tirole, J., 2011. Bubbly liquidity. Review of Economic Studies forth-
coming.

Gertler, M., Karadi, P., 2009. A model of unconventional monetary policy.

Working Paper.

Gertler, M., Kiyotaki, N., 2010. Financial intermediation and credit policy in
business cycle analysis. In: Woodford, M. (Ed.), Handbook of Monetary
Economics. Elsevier.

Hirano, T., Yanagawa, N., 2010. Asset bubbles, endogenous growth, and …-
nancial crisis. Working Paper, University of Tokyo.

Kareken, J., Wallace, N., 1978. Deposit insurance and bank regulation: A
partial equilibrium exposition. Journal of Business 51, 412–438.

Kocherkalota, N., 2009. Bursting bubbles: Consequences and cures. Working

Paper, Federal Reserve Bank of Minneapolis.
Martin, A., Ventura, J., 2011a. Economic growth with bubbles. American Eco-
nomic Review forthcoming.

Martin, A., Ventura, J., 2011b. Theoretical notes on bubbles and the current
crisis. IMF Economic Review 59 (1), 6–40.

Nikolov, K., 2010. A model of borrower reputations as intangible collateral.

Working paper, LSE.

Reinhart, C., Rogo¤, K., 2008. This time is di¤erent: A panoramic view of
eight centuries of …nancial crisis. Working paper, University of Maryland.

Sargent, T., 1987. Dynamic Macroeconomic Theory. Harvard University Press,

Cambridge, USA.

Schularick, M., Taylor, A., 2012. Credit booms gone bust: Monetary policy,
leverage cycles and …nancial crises, 1870-2008. American Economic Review
forthcoming.

Syverson, C., 2004. Market structure and productivity: A concrete example.

Journal of Political Economy 112 (6), 1181–1222.

Tirole, J., 1985. Asset bubbles and overlapping generations. Econometrica

53 (6), 1499–1528.

Welch, I., 2004. Capital structure and stock returns. Journal of Political Econ-
omy 112 (1), 106–131.
 0.9

0.8

0.7

0.6

0.5
λ

0.4

0.3

0.2

0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
θ

Figure 1a. Deposit rate less than one (red area)

 0.9

0.8

0.7

0.6

0.5
λ

0.4

0.3

0.2

0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
θ

Figure 1b. Lending rate less than one (red area)

 Output Bank capital

1.035 2.4

1.03 2.2

2
1.025

1.8
1.02

1.6
1.015
1.4
1.01
1.2
1.005
1

1
0.8

0.995 0.6

0.99 0.4
1 11 21 31 41 1 11 21 31 41

Saver bubbl e Bank bubbl e Saver bubbl e Bank bubbl e

Loans Aggregate investment

1.25 1.03

1.025
1.2

1.02

1.15
1.015

1.1 1.01

1.005
1.05

1
0.995

0.95 0.99
1 11 21 31 41 1 11 21 31 41

Saver bubbl e Bank bubbl e Saver bubbl e Bank bubbl e

Figure 3: Comparing a bank-held (solid line) and a saver-held (dashed line) bubble