Banks, Shadow Banking, and Fragility
Banks, Shadow Banking, and Fragility
Banks, Shadow Banking, and Fragility
(0) = , and u
() = 0.
In each period t, there are two dierent assets (investment technologies): a short
asset (storage technology), and a long asset (production technology). The short asset
transforms one unit of the good at time t into one unit of the good at t+1, thus eectively
storing the good. The long asset is represented by a continuum of investment projects.
An investment project is a metaphor for an agent who is endowed with a project (e.g.,
an entrepreneur with a production technology or a consumer who desires to nance a
house), but has no funds she can invest.
There is no aggregate, but only idiosyncratic return risk: each investment project
requires one unit of investment in t and yields a stochastic return of R
i
units in t +
2. The return R
i
is the realization of an independently and identically distributed
random variable
R, characterized by a probability distribution F. F is continuous and
strictly increasing on some interval [R, R] R
+
, with E[R
i
] = R > 1. We assume that
the realization of an investment projects long-term return, R
i
, is privately revealed to
whoever nances the project.
The idiosyncratic return risk of the long asset implies that nancial intermediaries
dominate a nancial markets solution in terms of welfare because of adverse selection
ECBWorkingPaper 1726, August 2014 8
in the nancial market.
4
In turn, unlike participants of a nancial market, a nancial
intermediary will not be subject to these problems as he is able to diversify and create
assets that are not subject to asymmetric information.
5
Finally, an investment project may be physically liquidated prematurely in t + 1,
yielding a liquidation return of R
i
/R, where (0, 1/R). The liquidation return
of a project thus depends on the projects stochastic long-term return. The average
liquidation return of a project is equal to .
Intergenerational Banking
In the following, we describe the mechanics of intergenerational banking and derive
steady state equilibria, closely following Qi (1994). We assume that there is a banking
sector operating in the economy, consisting of identical innitely lived banks that take
deposits and make investments. It is assumed that the law of large numbers applies at
the bank level, i.e., a bank neither faces uncertainty regarding the fraction of impatient
investors nor regarding the aggregate return of the long asset.
In each period t Z, banks receive new deposits D
t
. They sign a demand-deposit
contract with investors which species a short and a long interest rate. Per unit of
deposit, an investor is allowed either to withdraw r
t,1
units after one period, or r
t,2
units
after two periods. In period t, banks yield the returns from the last periods investment
in storage, S
t1
, and the returns from investment in the production technology in the
second but last period, I
t2
. They can use these funds to pay out withdrawing investors
and to make new investment in the production and in the storage technology.
We are interested in steady states of this intergenerational banking. A steady state is
given by a collection of payos, i.e., a short and a long interest rate, (r
1
, r
2
), a deposit
decision D, and an investment decisions I and S. We are only interested in those steady
states in which investors deposit all their funds in the banks, D = 1, and the total
investment in the storage and production technology does not exceed new deposits, i.e.,
S + I D.
6
This yields the investment constraint
S + I 1. (1)
4
Because asset quality is not observable, there is only one market price. Impatient consumers with
high-return assets have an incentive to liquidate them instead of selling them, and patient consumers
with low-return assets have an incentive to sell. This drives the market price below average return
and inhibits the implementation of the rst-best.
5
Critiques on the coexistence of nancial markets and intermediaries, as by Jacklin (1987) and Thadden
(1998), therefore do not apply to our model.
6
There also exist steady states with S + I > D, but this implies that banks have some wealth which
is kept constant over time and the net returns of which are paid out to investors each period. This
scenario does not appear particularly plausible or interesting.
ECBWorkingPaper 1726, August 2014 9
Moreover, we restrict attention to those steady states in which only impatient consumers
withdraw early. We will show later that these withdrawal decisions as well as the deposit
decision are actually optimal choices in a steady state equilibrium. In such a steady
state, banks have to pay r
1
units to impatient investors and (1 )r
2
units to patient
consumers in every period. Since payos and investments are limited by returns and
new deposits, the following resource constraint must hold:
r
1
+ (1 )r
2
+ S + I RI + S + 1. (2)
This constraint can be simplied to obtain a simple feasibility condition for steady-state
payos:
Denition 1 (Steady-state Payo). A steady-state payo (r
1
, r
2
) is budget feasible if
r
1
+ (1 )r
2
(R 1)I + 1. (3)
In a next step, we want to select the optimal steady state among the set of budget
feasible steady states. Our objective is to choose the steady state that maximizes the
welfare of a representative generation of investors, or equivalently, the expected utility
of one representative investor. We can partition this analysis by deriving the optimal
investment behavior of banks in a rst step, and then addressing the optimal interest
rates. We see that the budget constraint (3) is not inuenced by S. Thus, the banks
optimal investment behavior follows directly:
Lemma 1 (Optimal Investment). The optimal investment behavior of banks is given by
I = 1 and S = 0, i.e., there is no investment in storage. The budget constraint reduces
to
r
1
+ (1 )r
2
R. (4)
The intergenerational feature of banking implies that storage is not needed for the
optimal provision of liquidity. Any investment in storage would be inecient and would
hence imply a deterioration.
We can now derive the optimal steady-state payos (r
1
, r
2
), i.e., the optimal division
between long and short interest rate. It is straightforward to see that the rst-best
steady-state payo is given by perfect consumption smoothing, (r
FB
1
, r
FB
2
) = (R, R).
However, the rst-best cannot be implemented as it is not incentive compatible. The
incentive-compatibility and participation constraints are given by
r
1
r
2
, (5)
r
2
1
r
2
, (6)
and r
2
R. (7)
ECBWorkingPaper 1726, August 2014 10
Constraint (5) ensures that patient investors wait until the last period of their life-
time instead of withdrawing early and storing their funds. Constraint (6) ensures that
patient investors do not withdraw early and re-deposit their funds. By this type of
re-investment, investors can earn the short interest rate twice. As long as net returns
are positive, the latter condition is stronger, implying that the yield curve must not
be decreasing. Finally, constraint (7) ensures that investors do not engage in private
investment and side-trading. In fact, this condition is the upper bound to the side-
trading constraint. The adverse selection problem induced by the idiosyncratic return
risk relaxed this constraint, but the constraint will turn out not to be binding anyhow.
Obviously, constraint (6) is violated in the rst-best, inducing patient investors to
withdraw early and to deposit their funds in the banks a second time. In the second-
best, constraints (4) and (6) are binding, resulting in a at yield curve, r
2
= r
2
1
. Following
Equation (4), the interest rate is such that
r
1
+ (1 )r
2
1
= R. (8)
Proposition 1 (Qi 1994). In the second-best steady state, the intergenerational banking
sector collects the complete endowment, D = 1, and exclusively invests in the long-asset,
I = 1. In exchange, banks oer demand-deposit contracts with a one-period interest rate
given by
r
1
=
2
+ 4(1 )R
2(1 )
, (9)
and a two-period interest rate given by
r
2
= r
1
2
. (10)
It holds that r
2
> R > r
1
> 1. Unlike in the Diamond and Dybvig model, the rst-
best and the second-best do not coincide. The intergenerational structure introduces the
new IC constraint that the long interest rate must be suciently larger than the short
one in order to keep patient investors from withdrawal and reinvestment.
7
Steady-State Equilibrium
Until now, we have not formally specied the game in a game-theoretic sense. Consider
the innite game where in each period t Z, investors born in period t decide whether to
7
However, the intergenerational structure also relaxes the feasibility constraint. Although the yield
curve is allowed to be decreasing in the model of Diamond and Dybvig, the second-best of intergen-
erational banking dominates the rst-best of Diamond and Dybvig for a large set of utility functions
because banks do not have to rely on inecient storage.
ECBWorkingPaper 1726, August 2014 11
deposit, and investors born in t 1 decide whether to withdraw or to wait for one more
period. We do not engage in a full game-theoretic analysis. In particular, we do not
characterize all equilibria of this game, but only focus on the equilibrium characterized
by the above steady state, and analyze potential deviations. Banks are assumed to
behave mechanically according to this steady state.
Lemma 2. The second-best steady state constitutes an equilibrium of the innite game.
If all investors deposit their funds in the banks, and if only impatient consumers with-
draw early, it is in fact individually optimal for each investor to do the same. The
second-best problem already incorporates the incentive compatibility constraints as well
as the participation constraint. Patient investors have no incentive to withdraw early,
given that all other patient investors behave in the same way and given that new in-
vestors deposit in the bank. Nor do investors have an incentive to invest privately in
the production or storage technology, as the bank oers a weakly higher long-run return
that R.
Fragility
We will now study the stability of intergenerational banking in the absence of a deposit
insurance. Models of maturity transformation such as Diamond and Dybvig (1983) and
Qi (1994) may exhibit multiple equilibria in their subgames. Strategic complementarity
between the investors may give rise to equilibria in which all investors withdraw early,
i.e., bank run equilibria.
In the following, we analyze the subgame starting in period t under the assumption
that behavior until date t 1 is as in the second-best steady-state equilibrium. We
derive the condition under which banks might experience a run by investors, i.e., the
condition for the existence of a run equilibrium in the period-t subgame. In the case
of intergenerational banking, we consider a run in period t to be an event in which
all investors born in t 1 withdraw their funds, and none of the newly-born investors
deposit their endowment. In case of such a run, the bank has to liquidate funds in order
to serve withdrawing investors. In addition to the expected withdrawal of impatient
consumers, the bank now also has to serve one additional generation of patient investors
withdrawing early. Thus, it needs an additional amount of liquid funds (1 )r
1
=
1/2
2
+ 4(1 )R
2
+ 4(1 )R
.
Assumption 1 implies that, if in some period t all depositors withdraw their funds and
newborn investors do not deposit their endowment, the liquidation return that the bank
ECBWorkingPaper 1726, August 2014 12
can realize does not suce to serve all withdrawing consumers. Therefore, the bank is
illiquid and insolvent.
Proposition 2. Assume that the economy is in the second-best steady state. In the
subgame starting in period t, a run of investors on banks constitutes an equilibrium.
This proposition states that the steady state is fragile in the sense that there is scope
for a run. Assumption 1 implies that it is optimal for a patient investor to withdraw
early if all other patient investors do so and if new investors do not deposit. Note
that Proposition 2 only states that a run is an equilibrium of a subgame, but does not
say anything about equilibria of the whole game. However, our emphasis lies on the
stability/fragility of the steady-state equilibrium.
An important insight from Diamond and Dybvig (1983) and Qi (1994) is that a credible
deposit insurance may actually eliminate the adverse equilibrium at no cost. If the
insurance is credible, it eliminates the strategic complementarity and is thus never tested.
In fact, this is also true in the setup described above. Assume that there is a regulator
that can cover the liquidity shortfall in any contingency, including a full-blown bank run.
In the context of our model, this amounts to assuming that the regulator has funds of
(1 )r
1
at its disposal in any period. Whenever patient investors are guaranteed
an amount r
1
by the regulator, they do not have an incentive to withdraw early.
8
In
contrast, this does not hold in the presence of regulatory arbitrage, as we will show in
the following sections.
3. A Model of Banks and Shadow Banking
We now extend the model described above by three elements: First, we make the as-
sumption that commercial banks are covered by a safety net, but are also subject to
regulation and therefore have to bear regulatory costs. Second, there is an unregulated
shadow banking sector that competes with banks by also oering maturity transfor-
mation services. Investors can choose whether to deposit their funds in a bank or in
the shadow banking sector. Depositing in the shadow banking sector is associated with
some opportunity cost that varies across investors. Third, there is a secondary market
in which securitized assets can be sold to arbitrageurs. The amount of liquidity in this
market is assumed to be exogenous.
8
We ignore the possibility for suspension of convertibility. Diamond and Dybvig (1983) already indicate
that suspension of convertibility is critical if there is uncertainty about the fraction of early and late
consumers. Moreover, as Qi (1994) shows, suspension of convertibility is also ineective if withdrawing
depositors are paid out by new depositors.
ECBWorkingPaper 1726, August 2014 13
In the following, we describe the extended setup in detail and derive the steady-
state equilibrium, before analyzing whether the economy is stable or whether it features
multiple equilibria and panic-based runs may occur.
Commercial Banking and Regulatory Costs
From now on, we assume that commercial banks are covered by a safety net that is pro-
vided by some unspecied regulator, ruling out runs in the commercial banking sector.
9
Because of this safety net, banks are not disciplined by their depositors, such that in a
richer model moral hazard could arise. We therefore assume that banks are regulated
(e.g., they are subject to a minimum capital requirement). This is assumed to be costly
for the bank. In what follows, we will not model the moral hazard explicitly and assume
that regulatory costs are exogenous. However, in Appendix A we provide an extension
of our model in which we illustrate how moral hazard may arise from the existence of
the safety net, and why costly regulation is necessary to prevent moral hazard.
We assume that banks have to pay a regulatory cost per unit invested in the long
asset, resulting in a gross return of R . We assume that regulatory costs are not
too high, i.e., even after subtracting the regulatory costs, the long asset is still more
attractive than storage.
Assumption 2. R > 1 + .
Because of the lower gross return, banks can now only oer a per-period interest rate
r
b
such that
r
b
+ (1 )r
2
b
= R .
Under this regulation, the interest rate on bank deposits is explicitly given by
r
b
=
2
+ 4(1 )(R )
2(1 )
. (11)
The banking sector thus functions like the banking sector in the previous section. The
only dierence is that banks cannot transfer the gross return R to investors, but only
the return net of regulatory cost, R .
Shadow Banking Sector
We now introduce a shadow banking sector that also oers credit, liquidity, and maturity
transformation to investors. The structure of the shadow banking sector (compare Fig-
ure 1) is exogenous in our model. We selectively follow and simplify the descriptions by
9
The regulator is assumed to have sucient funds to provide a safety net. Moreover, he can commit
to actually applying the safety net in case it is necessary, i.e., in case of a run.
ECBWorkingPaper 1726, August 2014 14
Investors / Ultimate Lenders
Projects / Ultimate Borrowers
Commercial Banks
Insured
Deposits
Loans
MMF
ABCP conduit
SPV
NAV
1 Dollar
ABCP
ABS
Loans
Secondary
Market
for ABS
Liquidity Guarantees
Service provided in
shadow banking sector
Liquidity transfor-
mation by MMF
Maturity transformation
by shadow bank (ABCP
conduit such as SIV)
Risk transformation
(securitization) by
investment bank via SPV
Loan origination by
commercial bank
or mortgage broker
Figure 1: Structure of the nancial system: The structure of the shadow banking sector is mostly exogenous in our model.
We selectively follow and simplify the descriptions by Poszar et al. (2013). In our setup, shadow banking consists
of investment banks, shadow banks (ABCP conduits such as structured investment vehicles (SIVs)), and money
market mutual funds (MMFs). Investment banks securitize assets via special purpose vehicle (SPVs) in order to
make them tradable, i.e., they conduct risk and liquidity transformation. Once the projects are securitized, they
are purchased by shadow banks that nance their long-term assets by borrowing short-term from money market
mutual funds (MMFs) via, e.g., ABCP, i.e., they conduct maturity transformation. MMMFs are the door to the
shadow banking sector by oering deposit-like claims to investors such as shares with a stable net assets value
(NAV), conducting another form of liquidity transformation. Finally, there is a secondary market in which ABS
can be sold to arbitrageurs.
ECBWorkingPaper 1726, August 2014 15
Pozsar et al. (2013). Altogether, the actors of the shadow banking system invest in long
assets and transform these investments into short-term claims. However, we distinguish
between dierent actors in the shadow banking sector. This structure is exogenous and
empirically motivated.
In our setup, shadow banking consists of investment banks, shadow banks, and money
market mutual funds (MMFs). Investment banks securitize assets such as loans (i.e., the
long assets in our model) via special purpose vehicles (SPVs), thereby transforming them
into asset-backed securities (ABS). Through diversied investment, they eliminate the
idiosyncratic risk of loans and conduct risk transformation. Note that SPVs typically do
not lend to rms or consumers directly, but rather purchase loans from loan originators
such as mortgage agencies or commercial banks.
Shadow banks purchase securitized long assets and nance their business by issuing
short-term claims that they sell to MMFs. To put it more technically, ABCP conduits
such as structured investment vehicles (SIVs) purchase ABS and nance themselves
through ABCPs which they sell to MMFs.
10
Shadow banks hence conduct maturity
transformation. Maturity transformation is the central element and the key service of
banking in our model, and it is the main source of fragility.
For investors, MMFs are the door to the shadow banking sector as they transform
short-term debt (such as ABCP) into claims that are essentially equivalent to demand
deposits, such as equity shares with a stable net assets value (stable NAV). MMFs
thus conduct liquidity transformation. For tractability, we will assume that MMFs are
literally taking demand deposits.
Investment banks use their SPVs to invest in a continuum of long assets with idiosyn-
cratic returns R
i
. As the law of large numbers is assumed to apply, the return of their
portfolio is R. Securitization is assumed to come with a per-unit cost of . Therefore,
the per-unit return of securitized loans is R. Investment banks sell these securitized
loans to shadow banks. Similar to the regulatory cost , we also assume that the securi-
tization cost is not too high, i.e., even after subtracting the securitization cost, the long
asset is still more attractive than storage:
Assumption 3. R > 1 + .
The empirically motivated narrative is that investment banks purchase loans from
loan originators such as mortgage brokers or commercial banks. They bundle the claims
into securitized loans (ABS) via SPVs, successfully diversifying the idiosyncratic return
risk. Securitization costs accrue and can be thought of as the costs of creating an ABS
10
Note that we ignore other securities that shadow banks also use to nance their activities, such as
medium term notes (MTNs).
ECBWorkingPaper 1726, August 2014 16
out of many small loans, e.g., the costs of hiring a rating agency and a lawyer and setting
up the information technology to process payments.
11
Securitization ultimately makes
the long assets tradable by eliminating the adverse selection problem that is associated
with idiosyncratic return risk. Ultimately, securitized loans (ABS) are sold to shadow
banks.
At the heart of the shadow banking sector is the maturity transformation by shadow
banks (ABCP conduits). Shadow banks purchase securitized assets (ABS) from in-
vestment banks SPVs. As described above, these assets have a return of R and
a maturity of two periods. Shadow banks can nance themselves by borrowing from
MMFs via ABCPs. Moreover, they can also sell ABS to arbitrageurs in the secondary
market which is specied below.
Shadow banks oer a per-period interest rate r
abcp
such that
r
abcp
+ (1 )r
2
abcp
= R ,
implying a return of
r
abcp
=
2
+ 4(1 )(R )
2(1 )
.
We assume that there exists a secondary market for securitized assets (ABS). There
are arbitrageurs who are willing to buy ABS at a price that equals expected revenue.
Arbitrageurs can be thought of as experts (pension funds, hedge funds) that do not
necessarily hold ABS in normal times, but purchase them if they are available at small
discounts and promise gains from arbitrage.
The secondary market is assumed to be such that there is no market power on any
side of the market. Moreover, there is a xed amount of cash in this market. We assume
that arbitrageurs have a total budget of A which they are willing to spend for buying
ABS which can lead to cash-in-the-market pricing. The equilibrium supply and price of
ABS on the secondary market will be derived below.
The idea behind this assumption is that not every individual or institutions has the
expertise to purchase nancial products such as ABS. Moreover, the equity and collateral
of these arbitrageurs is limited, so they cannot borrow and invest innite amounts.
12
Investors can access the services of the shadow banking sector vian MMFs, which
assumed to intermediate between investors and shadow banks.
13
MMFs oer demand-
11
Securitization costs could also be understood as the regulatory costs that accrue in shadow banking.
While shadow banking activities are outside the regulatory perimeter of banking regulation, there
are nonetheless existing regulations.
12
See theories on the limits to arbitrage Shleifer and Vishny (1997).
13
This is like assuming that investors face large transaction costs or do not have the expertise to deal
with shadow banks directly.
ECBWorkingPaper 1726, August 2014 17
deposit contracts to investors while purchasing short-term claims on shadow banks.
14
MMFs oer a per-period interest rate r
mmf
to investors and purchase ABCP (short-
term debt) with a per-period return r
abcp
from shadow banks. Competition among
MMFs implies that r
mmf
= r
abcp
.
Upon birth, investors can choose whether to deposit their endowment in a regulated
bank or in an MMF. Depositing at MMFs comes at some opportunity cost. We assume
that investors are initially located at a regulated bank. Switching to an MMF comes
at a cost of s
i
, where s
i
is independently and identically distributed according to the
distribution function G. We assume that G is a continuous function that is strictly
increasing on its support R
+
, and that G(0) = 0. The switching cost is assumed to enter
into the investors utility additively separable from the consumption utility.
This switching cost should not be taken literally. One can think of these costs as
monitoring or screening costs for investors that become necessary when choosing an
MMF as these are not protected by a deposit insurance (see Appendix A for more
details). For simplicity, we have assumed that all depositors have the same size. However,
we could alternatively write down a model where investors have dierent endowments
(see Appendix B). It is very plausible that the ratio of switching costs to the endowment
is lower for larger investors (e.g., for corporations that need to store liquid funds of
several millions for a few days). Another interpretation is the forgone service benets
that depositors lose when leaving commercial banks, such as payment services and ATMs.
Investors Behavior
Given the interest rates of commercial banks, r
b
, of MMFs, r
abcp
, and given the switching
cost distribution G, we can pin down the size of the shadow banking sector.
Lemma 3. Assume that banks oer an interest rate r
b
and MMFs oer an interest rate
of r
abcp
, as specied above. Then there exists a unique threshold s
). It holds that s
= f(, ), where f
> 0 and
f
< 0.
Proof. Take r
b
and r
abcp
as described above. We know r
b
decreases in , and r
abcp
decreases . Staying at a commercial bank provides an investor with an expected con-
sumption utility of EU
b
= u(r
b
) + (1 )u(r
2
b
). Switching to an MMF is associated
14
an MMF typically sells shares to investors, and the funds sponsor guarantees a stable NAV, i.e., it
guarantees to buy back shares at a price of one at any time. As mentioned above, the stable NAV
implies that an MMF share is a claim that is equivalent to a demand-deposit contract. For simplicity,
we will assume that MMFs are literally taking demand deposits.
ECBWorkingPaper 1726, August 2014 18
with an expected consumption utility of EU
sb
= u(r
abcp
) + (1 )u(r
2
abcp
). Observe
that EU
b
decreases in and EU
sb
decreases in .
An investor with switching cost s
i
switches to the shadow banking sector if EU
b
<
EU
sb
s
i
. This implies that all investors with s
i
EU
sb
EU
b
switch to MMFs. We
dene s
f(, ) = EU
sb
() EU
b
(). A mass G(s
> 0 and f
< 0.
An investor with s
i
= s
). The
size of the shadow banking sector increases in the regulatory cost and decreases in
the cost of securitization . For example, if investors had linear consumption utility, it
would hold that s
= .
We are now equipped to characterize the economys steady state equilibrium:
Proposition 3. In the second-best steady-state equilibrium, the intergenerational bank-
ing sector collects an amount of deposits D
b
= 1 G(s
2
+ 4(1 )(R )
2(1 )
. (12)
MMFs collect an amount of deposits D
sb
= G(s
2
+ 4(1 )(R )
2(1 )
. (13)
It holds that s
= f(, ), where f
> 0 and f
) investors
deposit their funds in the shadow banking sector each period, shadow banks have to
serve MMFs with a total amount of G(s
)[(1 )r
2
abcp
+ r
abcp
].
However, shadow banks only have an amount G(s
)[(1 )r
2
abcp
+ r
abcp
(R )].
In order to cover this shortfall, shadow banks can either sell the ABS that they bought
in t 1 to the arbitrageurs, or they can liquidate these assets.
15
We assume that
15
Liquidating ABS might not be straightforward, as all tranches would have to be collected and the un-
ECBWorkingPaper 1726, August 2014 20
liquidation of ABS will never be enough to cover the shortfall. Similar to Assumption 1,
this is equivalent to making the following assumption:
Assumption 4. < 1/2
2
+ 4(1 )(R )
Observe that in case of a run, the supply of shadow banks is partially inelastic: they
have to cover their complete liquidity shortfall. There are two cases to be considered:
In the rst case, the arbitrageurs funds are sucient to purchase all funds the shadow
banks sell at face value, while in the second case, the arbitrageurs budget is not sucient
and the price is determined by cash-in-the-market pricing. Runs of MMFs on shadow
banks become possible in this second case.
Proposition 4. Assume that the economy is in the second-best steady state. A run
of MMFs on shadow banks (ABCP conduits) constitutes an equilibrium of the period-t
subgame if and only if
G(s
) >
A
1/2
2
+ 4(1 )(R )
,
where s
= f(, ), with f
> 0 and f
< 0.
Proof of Proposition 4. Assume that MMFs collectively withdraw funds from shadow
banks and deposit no new funds in period t. It will be optimal for a single MMF to also
withdraw if the shadow banks become illiquid and insolvent in t.
We calculate the liquidity shortfall if shadow banks in case of a run as
G(s
)[(1 )r
2
abcp
+ r
abcp
(R )].
Recall from Proposition 3 that r
abcp
+ (1 )r
2
abcp
= R . Making use of this by
substituting for (R ), we know that the shortfall is given by G(s
)[(1 )r
abcp
].
Recalling Equation (13), the liquidity shortfall is given by
1/2
2
+ 4(1 )(R )
G(s
).
Liquidating the ABS portfolio would yield G(s
2
+ 4(1 )(R )
G(s
2
+ 4(1 )(R )
G(s
), where shadow
banks cannot sell all their assets at face value. If all MMFs stop rolling over ABCP,
shadow banks cannot raise the required funds to fulll their obligations by selling their
ABS because the amount of assets on the secondary market exceeds the budget of ar-
bitrageurs. The price of ABS drops below face value and shadow banks are forced to
sell their complete ABS portfolio. Still, shadow banks can only raise a total amount A
of liquidity, which is insucient to serve withdrawing MMFs. It follows that it is not
optimal for an MMF to roll over ABCP if no other MMF does so. A self-fullling run
thus constitutes an equilibrium whenever
G(s
) >
A
1/2
2
+ 4(1 )(R )
.
The key mechanism giving rise to multiple equilibria is cash-in-the-market pricing (see,
e.g., Allen and Gale, 1994) in the secondary market for long-term securities that results
from limited arbitrage capital and is related to the notion of limits to arbitrage (see,
e.g., Shleifer and Vishny (1997)). The fact that there are not enough arbitrageurs (and
that these arbitrageurs cannot raise enough funds) to purchase all assets of the shadow
banking system possibly induces the price of ABS to fall short of their face value. This
implies that shadow banks may in fact be unable to serve their obligations once they sell
all their long-term securities prematurely. This in turn makes it optimal for an MMF to
run on shadow banks once all other MMFs run.
In order to illustrate the role of limited availability of arbitrage capital we examine the
hypothetical re-sale price in the market for ABS. Cash-in-the-market pricing describes
a situation where the buyers budget constraint is binding and the supply is xed. The
price adjusts such that demand balances the xed supply. In our case, the price p is
ECBWorkingPaper 1726, August 2014 22
0
0
R
(1 )r
abcp
p
G(s
) A/
Stability Fragility
Unique equilibrium
Multiple equilibria
Figure 2: This graph depicts the potential re-sale price of ABS. Whenever G(s
) =
A/(1 )r
abcp
, the funds of arbitrageurs are sucient to purchase all assets
of shadow banks at face value. There is a unique equilibrium of the period t
subgame in which there are no panic-based withdrawals of MMFs. In turn, if
G(s
) > , the funds of arbitrageurs are insucient, and the period t subgame
has multiple equilibria. If all MMFs withdraw from shadow banks, the price
of ABS in the secondary market drops to the red line.
ECBWorkingPaper 1726, August 2014 23
such that
pG(s
) = A.
The re-sale price is a function of the amount of assets that are on the market in case
of a run on the shadow banking sector, which is given by the size of the shadow banking
sector G(s
) =
R if G(s
) ,
A/G(s
) if G(s
) (, A/] ,
if G(s
) > A/.
The equilibrium re-sale prices as a function of the size of the shadow banking sector
is illustrated in Figure 2.
Whether the period t subgame has multiple equilibria ultimately depends on the
parameters and , as they determine the size of the shadow banking sector. This
is depicted in Figure 3. Whenever the regulatory costs exceed the costs of securitizing
assets (i.e., if we are above the 45 degree line), the shadow banking sector has positive
size in equilibrium, i.e. G(s
R 1
No Shadow Banking
G(s
) = 0
Stable
Shadow Banking
0 < G(s
) <
Fragility
G(s
) >
Figure 3: This gure visualizes the equilibrium characteristics of the nancial system for
dierent values of and . For < , shadow banking is not made use of in
equilibrium, as it is dominated by commercial banking. If > , the shadow
banking sector has positive size. As long as the dierence is small, shadow
banking is stable. If the dierence increases, the size of the shadow banking
sector also increases and nally introduces fragility into the nancial system.
do not experience runs by investors. Patient investors who are located at a commercial
bank will thus never withdraw their funds early.
Liquidity guarantees imply that in case of a run on shadow banks, commercial banks
supply liquid funds to shadow banks. This increases the critical size up to which the
shadow banking sector is stable. However, this comes with an unfavorable side eect:
once this critical size is exceeded an shadow banks experience a run, the crisis spreads
to the commercial banking sector and makes the safety net costly.
Proposition 5. Assume that the economy is in the second-best steady state described in
Proposition 3 and all shadow banks (ABCP conduits) are granted liquidity guarantees by
commercial banks. A run of MMFs on shadow banks constitutes an equilibrium of the
ECBWorkingPaper 1726, August 2014 25
subgame starting in period t if and only if
G(s
) >
max[A, ] + 1
1/2
2
+ 4(1 )(R )
+ 1
,
where s
2
+ 4(1 )(R )
G(s
).
Banks can sell their loans on the same secondary market in case of a crisis. Still, the
total endowment of arbitrageurs in this market is given by A. Therefore, either banks
and shadow banks sell their assets in the secondary market, or both types of institutions
liquidate their assets. They jointly still only raise an amount A from selling long-term
securities on the secondary market or units from liquidating all long assets. The
maximum amount they can raise is thus max[A, ]. On top, commercial banks also have
an additionally amount 1G(s
)) 1/2
2
+ 4(1 )(R )
G(s
),
which is equivalent to
G(s
)
max[A, ] + 1
1/2
2
+ 4(1 )(R )
+ 1
= .
If G(s
)
R
(1 )r
abcp
A/
Stability Fragility
Unique equilibrium
Multiple equilibria
Liquidity
guarantees
Figure 4: This graph depicts the potential re-sale price of ABS for the case that reg-
ulated commercial banks provide liquidity guarantees to shadow banks. The
critical size above which multiple equilibria exist moves from to .
ECBWorkingPaper 1726, August 2014 27
In traditional banking models, policy tools like a deposit insurance eliminate self-
fullling adverse equilibria at no cost. This is not necessarily true in our model: once
the shadow banking sector exceeds the size , a run in the shadow banking sector
constitutes an equilibrium despite the safety net for commercial banks, and despite the
liquidity guarantees of banks. Shadow banks by circumventing the existing regulation
place themselves outside the safety net and are thus prone to runs. If the regulated
commercial banks oer liquidity guarantees, a crisis in the shadow banking sector also
spreads to the regulated banking sector. Ultimately, self-fullling adverse equilibria are
not necessarily eliminated by the safety net and may become costly.
Corollary 1. Assume that G(s
2
+ 4(1 )(R )
G(s
) max[A, ] (1 G(s
)) > 0.
If the regulated commercial banking and the shadow banking sector are intertwined, a
crisis may not be limited to the shadow banking sector, but also spread to the commercial
banks, thus testing the safety net. Ultimately, the regulator has to step in and cover
the commercial banks liabilities. Therefore, the model challenges the view that policy
measures like a deposit insurance necessarily are an ecient mechanism for preventing
self-fullling crises. Historically, safety nets such as a deposit insurance schemes were
perceived as an eective measure to prevent panic-based banking crises. The view is
supported by traditional banking models of maturity transformation such as Diamond
and Dybvig (1983) and Qi (1994). In the classic models of self-fullling bank runs, a
credible deposit insurance can break the strategic complementarity in the withdrawal
decision of bank customers at no cost. We show that this may not be the case when
regulatory arbitrage is possible and regulated and unregulated banking activities are
intertwined.
5. Runs on MMFs
In the previous sections, we ruled out runs on MMFs by assuming that they have credible
support by a sponsor. Credible sponsor support means that even if all investors withdraw
their funds from an MMF, the sponsor is able to provide sucient liquidity to the MMF
such that it can serve all investors. Recall that we use the narrative that MMFs are
literally oering demand-deposit contracts. In practice, an MMF issues equity shares,
ECBWorkingPaper 1726, August 2014 28
and its sponsor guarantees stable NAV for theses shares, i.e., it promises to buy these
shares at face value in case of liquidity problems.
We now relax the assumption that the guarantee is always credible. We explicitly
model the credibility of the guarantee by assuming that the sponsors have m units
of liquidity per unit of investment in the MMF that they can provide in case of a
crisis. Moreover, we keep the assumption of existing liquidity guarantees. We show that
providing mG(s
) >
max[A, ] + 1
1/2
2
+ 4(1 )R
+ 1 m
= > .
If the > G(s
1/2
2
+ 4(1 )R
+ 1
G(s
) max[A, ] 1.
Therefore, a run of investors on MMFs constitutes an equilibrium only if
mG(s
) <
(1/2
2
+ 4(1 )R
+ 1
G(s
) max[A, ] 1.
The result builds on the fact that sponsor support is like a liquidity backstop. If there
is a run by MMFs on shadow banks, MMFs will make losses. This additionally triggers
ECBWorkingPaper 1726, August 2014 29
a run of investors on MMFs if the sponsor is not able to cover these losses. Again, losses
depend on the re-sale price. The re-sale price in turn depends on the amount of assets
sold in case of a run by MMFs on shadow banks, which is determined by the size of the
shadow banking sector. If the shadow banking sector is so large that runs by MMFs
on shadow banks occur, but not so large that losses cannot be covered by the sponsors,
investors do not run. This is the case for > G(s