-
Ohio State University Press is collaborating with JSTOR to
digitize, preserve and extend access to Journal of Money, Creditand
Banking.
http://www.jstor.org
Liquidity and Risk Management Author(s): Bengt Holmstrm and Jean
Tirole Source: Journal of Money, Credit and Banking, Vol. 32, No.
3, Part 1 (Aug., 2000), pp. 295-319Published by: Ohio State
University PressStable URL:
http://www.jstor.org/stable/2601167Accessed: 20-03-2015 12:33
UTC
REFERENCESLinked references are available on JSTOR for this
article:
http://www.jstor.org/stable/2601167?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked
references.
Your use of the JSTOR archive indicates your acceptance of the
Terms & Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars,
researchers, and students discover, use, and build upon a wide
range of contentin a trusted digital archive. We use information
technology and tools to increase productivity and facilitate new
forms of scholarship.For more information about JSTOR, please
contact [email protected].
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
MONEY, CREDIT, AND BANKING T,E,CTURE
Liquidity and Risk Management
BENGT HOLMSTROM JEAN TIROLE
Firms and financial institutions are best viewed as ongoing
entities, whose project completion may require renewed injections
of liq- uidity. This paper proposes a contract-theoretic framework
inte- grating three dimensions of corporate financing and
prudential regulation: (1) liquidity management, (b) risk
management, and (c) capital structure. It concludes with a
preliminary assessment of re- cent regulatory approaches to the
treatment of market risk.
THIS PAPER IS CONCERNED with the corporate demand for liquidity
and the various ways in which firms in the real and the financial
sectors manage their liquidity needs so as to be able to carry out
production and investment plans effectively without being held back
by temporary liquidity shortages. Several key decisions impact a
corporation's future ability to avail itself of financial
funds.
First, the corporation's capital structure sets, among other
things, a timetable for reimbursing investors. Short-term debt
forces the firm to pay out cash, drying up liq- uidity. Long-term
debt allows the Elrm more room to adjust to liquidity shocks,
exerting pressure mainly by the constraints it places on the amount
of new debt that can be raised. Preferred stock explicitly embodies
a liquidity option (a form of line of credit) by allowing the firm
to delay reimbursement. Equity is, of course, the most
accommodating claim with no precise timetable for the payment of
dividends.
Second, corporations do not invest all their resources in
profitable, long-term pro- jects. They also invest in less
profitable liquid assets that are held on their balance sheets as
buffers against shocks. We define a liquid asset as one that the
firm can quickly resell or pledge as collateral at its true value
and whose market value is un- likely to be depressed when the firm
needs resources. Looking at this dual condition,
This lecture was delivered April 16, 1999, at Ohio State
University by the second author. The authors are grateful to the
participants for helpful comments.
BENGT HOLMSTROM is Paul A. Samuelson Professor of Economics at
MIT. E-mail: [email protected]. JEAN TIROLE is professor of economics
at IDEI and GREMAQ, Toulouse, CERAS, Paris, and atMIT. E-mail:
[email protected] Journal of Money, Credit, and Banking, Vol. 32, No.
3 (August 2000, Part 1) Copyright 2000 by The Ohio State
University
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
296 : MONEY, CREDIT, AND BANKING
we observe that the first part is just the notion of liquidity
emphasized in the litera- ture on market microstructure. The second
part is the analog of the covariance con- dition in the consumption
CAPM model stemming from the producers' demand for liquidity [see
Holmstrom-Tirole (1998a) for the derivation of liquidity premia in
this context and a discussion of how they differ from risk premia
in a consumption-based asset pricing model]. The firm's demand for
a liquid asset depends on whether and how much the asset will
deliver when the firm needs cash. In this respect, corporate equity
or commercial real estate may be poor instruments for securing
liquidity even when they are liquid in the sense of market
microstructure theory. Short-term trea- sury bonds better satisfy
our two conditions as do cash instruments, of course.
Rather than hoarding liquidity themselves, corporations may
secure lines of credit from finaricial institutions.l For example,
they can contract with a bank or an insur- ance company for the
right to draw a specified amount of cash at a given rate of in-
terest by a given date in exchange for an upfront commitment fee.
The associated liability for the financial institution must be
backed up by an increase in its liquid as- sets, sufficient to make
it likely that it can deliver on its promise. Liquidity provision
is an important activity of banks. For example, roughly 80 percent
of commercial and industrial loans at large U.S. banks are
take-downs under loan commitments (Greenbaum and Takor 1995).
Third, corporations engage in risk management. They can use
derivatives to hedge specific risks (interest rate, currency, raw
materials, etc.). For example, a corporation with substantial
exports may quickly become short of cash if the exchange rate sud-
denly turns unfavorable. Foreign exchange swaps allow the firm to
insure against this type of liquidity shortage. Using derivatives
and forward and futures markets is only one of many ways in which
firms can cover themselves against specific risks: other ways
include securitization, insurance against theft, fire or the death
of a key employee, trade credit insurance and diversification of
various sorts (geographic, product mix, etc.).
Last, corporations also attempt to measure their global risk
exposure. Sophisti- cated tools, such as RAROC or Risk Metrics,
give an imperfect, but useful picture of a firm's or bank's
exposure to various macroeconomic factors. Such Value at Risk (VAR)
models, extensively used by banks to control their dealers' and
traders' risk taking, and by prudential regulators to monitor
banks, estimate the extreme lower tail risk of a portfolio. The
value at risk is the level of loss that will be exceeded with some
prespecified probability (1 percent, 5 percent, ...) over some time
horizon (1 day, 10 days, a year, ....)2 The purpose of such models
is to assess the probability that an entity runs into a serious
liquidity problem and is forced into a costly liquidation of
assets. These models have become popular in the 90s in the wake of
recent scan- dals at Barings, Procter & Gamble,
Metallgesellschaft, and the Orange County.
1. For more on lines of credit and loan commitments, see Crane
(1973) and Greenbaum and Thakor (1995).
2. See Duffie and Pan (1997) and Gordy (1998) for a review of
the techniques employed in such models.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 297
The Arrow-Debreu model, the cornerstone of modern finance,
offers few clues for understanding the above-mentioned three
practices, let alone how they are related. In the Arrow-Debreu
model capital structure is irrelevant (Modigliani-Miller). Nor do
firms need to hoard liquid assets, since they can issue claims
against the full value of the new investments (see below). Finally,
claimholders cannot gain by having firms engage in risk management,
since reshuffling state-contingent resources in a com- plete market
does not affect the market portfolio.
Two arguments that have been put forward to explain the value of
liquidity man- agement are taxes and managerial incentives. Taxes
are specific to locality and time and do not appear very helpful in
explaining the observed patterns of liquidity man- agement (Stultz
1996). And while risk management techniques could be used to fil-
ter out some of the exogenous noise in managerial compensation
(Fite-Pfleiderer 1995; Stulz 1984, 1996), Froot, Scharfstein, and
Stein (1993) rightly observe that this argument does not make a
strong case for risk management, since the same could be
accomplished by building the filter directly into the manager's
contract. It has also been suggested that corporate risk management
reduces the risk of bank- ruptcy, but without an explanation of why
bankruptcy leads to inefficient liquidation and reallocation of
assets,3 this point is subsumed in the earlier observation that
issu- ing new claims in a complete market will not improve the
claimholders' lot.
In our analysis, liquidity management derives its rationale from
the corporations' concern for refinancing, a concern emphasized in
various contexts by Thakor, Hong, and Greenbaum (1981) and Froot,
Scharfstein, and Stein (1993) among others. In the Arrow-Debreu
world, refinancing is not a concern because the entire benefit from
reinvestment can be pledged to outside investors. Consequently, any
positive net pre- sent value project can be funded at the time the
opportunity for investment arises. Even when there is a
debt-overhang problem, the initial debtholders can be per- suaded
to exchange their claims for more valuable ones in order to raise
the needed capital. The situation is very different when claims on
the full value of the firm can- not be issued. If part of the
corporate cake is nonpledgable, because insiders (man- agers,
workers, owner-monitors, etc.) have to hold a share of that cake in
order to behave properly, or if insiders enjoy significant private
benefits for some other rea- son, it is possible that an (ex ante)
socially beneficial liquidity need cannot be met unless the firm
has secured sufficient liquidity in advance (see, for instance,
Holm- strom and Tirole 1998b). Long-term financing is also of value
if assets can become temporarily illiquid (nonpledgable) because of
adverse selection problems (see, for instance, von Thadden 1995).
Information about the true value of a firm's assets is typically
limited to a smaller set of experts. If these experts have
difficulties raising funds at the same time that the firm needs
liquidity, the firm's assets cannot be sold at full value. While
adverse selection may be a realistic reason for illiquidity, we
will rely on our own moral hazard model to study the three
earlier-mentioned aspects of liquidity management because of its
greater ease of use.
3. See, however, Caillaud, Dionne, and Jullien (2000) for a
model of hedging in the presence of en- dogenous bankruptcy
costs.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
298 : MONEY, CREDIT, AND BANKING
The paper is organized as follows: Section 1 models corporate
liquidity demand as in Holmstrom and Tirole (1998b) and applies the
framework to the analysis of finan- cial structure. Sections 2
through 4 contain the more novel material. Section 2 ap- plies this
model to the free cash flow problem and undesired refinancing.
Section 3 derives the optimal risk management strategy. Last,
section 4 provides a preliminary analysis of the regulatory
treatment of market risk in prudential regulation. In 1996 the
Basle Committee amended the 1988 international accord on prudential
regula- tion to incorporate market risk, in reaction to the concern
that banks' trading book losses might jeopardize their ability to
manage their banking book. Section 4 as- sesses recent regulatory
approaches to market risk in the light of section 3.
While this paper focuses on microeconomic issues, liquidity
considerations are central to a number of other topics such as
asset pricing and country-wide liquidity crises, which will not be
pursued here (see Caballero and Krishnamurthy 1999).
1. MODELING CORPORATE LIQUIDITY DEMAND
1.1 Roadmap In order to study corporate demand for liquidity we
need to extend the standard
two-period model of investment to include an intermediate, third
stage at which the firm might need additional funding (see Figure 1
for a representation of the timing of the model). Initially, we
assume that the firm does not produce any income at the in-
termediate stage; rather, it may be hit by an adverse shock and
required to plow in some extra cash in order to be able to continue
the project. There are two approaches
'tCash poor firtrl'l
/ overruns /reinvestment Cash need
/ \shortfall in earnings
O 1 Continue 2 * / E E
Financing Outcome
Liquidation downsuing
FIG. 1
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROMAND JEAN TIROLE : 299
the firm can take to deal with urgent liquidity needs. The first
is to secure some source of liquidity before the shock occurs. For
example, the firm may keep liquid assets such as treasury bills on
its balance sheet or it can secure a line of credit from a bank. In
the second approach the firm waits for the shock to occur before
its starts raising funds.
We will show that the wait-and-see approach does not sufElce
when liquidity shocks are exogenous. There will be situations where
the firm would have been res- cued under an optimal ex ante
contract, but will fail without such a contract. Neither initial
lenders nor new lenders want to rescue the firm in certain states
unless the firm secured liquidity services in advance. This is due
to the fact that the borrower must always keep a stake in the firm
and hence the firm's full value cannot be pledged to the outsiders.
Consequently, the lenders do not internalize the loss incurred by
the borrower when the project is stopped, resulting in excessive
liquidation.
1.2 Optimal Liquidity Management At date O an entrepreneur (also
called the "insider" or the "borrower") can invest
in a project with constant returns to scale. The scale of the
project, I, is a continuous variable that can be selected
freely.
The entrepreneur initially has "assets" or "net worth" A. These
assets could be cash or liquid securities that can be used to cover
the cost of investment. To imple- ment a project of scale I > A
the entrepreneur must borrow I-A.
A project started at date O and continued at date 1 (see below)
either succeeds, that is, yields verifiable income RI > O at
date 2, or fails, that is, yields no income at date 2. The
probability of success is denoted by p. The project is subject to
moral hazard between dates 1 and 2. The entrepreneur can "behave"
("work," "exert effort"), or "misbehave" ("shirk"); or,
equivalently, the entrepreneur chooses between a project with a
high probability of success and another project which ceteris
paribus he prefers (is easier to implement, is more fun, has
greater spin-offs in the future for the entrepreneur, benefits a
friend, etc.), but has a lower probability of success. Behaving
yields probability P = PH Of success and no private benefit to the
entrepreneur, and misbehaving results in probability P = PL < PH
of success and private benefit BI > O (measured in units of
account) to the entrepreneur. Let Ap--PH - PL. In the "effort
interpretation," BI can also be interpreted as a disutility of
effort saved by the entre- preneur when shirking. Note that the
private benefit is proportional to investment. For notational
simplicity the rate of interest is taken to be zero. Both the
borrower and the potential lenders (or "investors") are risk
neutral. The borrower is protected by limited liability. Lenders
behave competitively so that loans make zero profit.
After investment I is sunk at date O but before the entrepreneur
works on the pro- ject, an exogenous liquidity shock p E [O, )
occurs at date 1. A cash infusion equal to pI is needed to cover
"cost overruns" and allow the project to continue. If pI is not
invested, the project is abandoned and thus yields no income. The
fraction p is dis- tributed according to the continuous
distribution F(p) on [O, oo), with densityt(p).
Regardless of the required fraction of the cash infusion, the
project, if pursued, is still a project of size I, in that the
income in case of success is RI and the borrower's
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
300 : MONEY, CREDIT, AND BANKING
private benefit from misbehaving is BI. One cannot increase the
size of the project after the initial stage. The timing is
summarized in Figure 2.
We assume that the investment has a positive net present value.
That is, under a rule that specifies that the project is abandoned
if and only if p ' p for some thresh- old p, the expected payoff
per unit of investment is strictly positive. The positive NPV
condition under liquidity shocks is
m-ax{F(p)pHR-1-Jo pf(p)dp} > 0 * (1)
We first look for the optimal loan agreement and later discuss
its implementation. It is easy to show that it is optimal to have a
"cutoff rule" for the date-l reinvestment. There exists a threshold
p* such that it is optimal to continue if and only if
P ' P* (2) The incentive constraint in case of continuation
requires that the borrower's stake in case of success, Rb, times
the reduction in the probability of success due to shirking be
larger than the private benefit from shirking (due to risk
neutrality the entrepre- neur optimally receives 0 in case of
failure):
(P)Rb ' BI. (ICb)
The break-even condition for the investors is
F(p*)[pH(RI-Rb)]-I-A + IP pIt(p)dp (IRe)
The lenders receive a return only if the project is continued,
which occurs with prob- ability F(p*). The left-hand side of (IR)
is the expected pledgeable income. The right-hand side is the
investors' date-0 outlay, I-A, plus the expected liquidity need.
From these two constraints, we deduce the "debt capacity" (or more
precisely the maximal investment that allows the lenders to break
even):
Date 0 Date 1 Date 2 Disbursement
)< )< 3< x )< )( b Loan Investment Need for Moral
Outcome agreement I cash infusion hazard
pI realized
No disbursement
V Project is abandoned FIG. 2
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 301
I = k(p*)A,
where
1 + JP p+(p)dp-F(P*)[PHR-PH AP I 1 (3)
1 + JP p+(p)dp-F(p*)po Note that the borrower's debt capacity is
maximal when the threshold p* is equal to
/ B \ the unit expected pledgeable income Po-PH VR--J .
Given that lenders make no profits the borrower's net utility is
the social surplus of the project, namely,
Ub = m(p *)I = m(p *)k( p *)A
where
m(p*)--F(p*)pHR-1-JP p+(p)dp (4)
is the margin per unit of investment. What is the optimal
continuation rule? Intuition might first suggest that, given
that
liquidity shocks are exogenous, one would want to continue if
and only if this is ex post efficient, that is, if and only if p '
pHR. Indeed, p* = pHR = P1, maximizes the margin m(p*). However, at
p* = pHR, the multiplier k is decreasing in p*. So one ac- tually
ought to choose a lower threshold than the ex post efficient one.
It is easily seen from (3) and (4) that
1 + JP* pf(p)dp ub= F(p*) A,
1+JP pf(p)dp / B\ F(p*) \ AP J
and so the optimal threshold minimizes the expected unit cost
c(p*) of effective in- vestment:
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
302 : MONEY, CREDIT, AND BANKING
p * minimizes c(p*) = 1 + lo Pf(p)dp
or
JoP F(p)dp = 1. (6) Condition (6) can be obtained by integrating
by parts and rewriting the expected unit cost, of effective
investment as
1-JP* F(p)dp
This expression also shows that at the optimum,4 the threshold
liquidity shock is equal to the expected unit cost of egffective
investment:
c(p*)= p*.
This in turn implies that the borrower's net utility is
P1 -p* Ub=p*_p A. (7)
Next, we observe that this optimal threshold lies between the
pledgeable income po and the expected return P1
( /\P ) (8) This follows from the fact that the margin m(p*) and
the multiplier k(p*) are both decreasing above P1 and both
increasing below pO: see Figure 3.5 Condition (8) is consistent
with (7): If p* were to exceed P1, the project could not be
financed prof- itably. If p* were lower than pO, the debt capacity
and the borrower's utility would be infinite.6
4. It is easy to show that c(-) is quasi-convex (c"(p*) > O
if c'(p*) = O). 5. Indeed, m(-) is quasi-concave with a maximum at
Pl and k(-) is quasi-concave with a maximum at
Po 6. Note that p* does not depend on pO and Pl as long as it
falls between the two. The investment scale
only depends on pO.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 303
investment multiplier >(P)
0 Po p* P1 P pledgeable NPV: liquidity income: maxirnizes shock
maximizes profit debt capacity per uIiit of
investment
FIG. 3
We conclude that the pervasive logic of credit rationing applies
not only to the choice of initial investment, but also to the
continuation decision. In order to be able to invest more ex ante,
the borrower accepts a level of reinvestment below the ex post
efficient level (p* < pHR). The logic is important. Because the
entrepreneur is credit constrained, his return on internal funds
(A) exceeds the market rate (O). Therefore, he does not want to buy
full insurance against the liquidity shock p (that is, set p* = P1)
At full insurance the marginal return for his money is zero, while
the marginal re- turn from expanding scale is strictly
positive.
Equation (8) implies that a wait-and-see policy, under which the
borrower tries to raise funds from the lenders after observing the
liquidity shock, is suboptimal. Even under perfect coordination
(there is no "debt-overhang" phenomenon), lenders will provide new
credit at date 1 only if the pledgeable income exceeds the amount
of reinvestment, that is, only if p < pO. Because pO < p*, it
is optimal for the borrower to secure in advance more funds than
can be raised by a wait-and-see policy. This creates a corporate
demand for liquidity. See Figure 4 for a summary of our
findings.
Condition (6) has an interesting implication. An increase in the
riskiness of the liquidity shock in the sense of a mean-preserving
spread of F7 raises the left-hand side of (6) and thus reduces the
threshold p*. So, the borrower should hoard more liquidity when the
liquidity shock incurs a mean-preserving reduction in risk, as may
occur, for example, when a new market opens that allows the firm to
insure against a
7. See, for example, Rothschild and Stiglitz (1970, 1971). The
distribution G(p) (with density g(p), say) is a mean-preserving
spread of distribution F(p) if
lUDO lUDO lUDO lUDO
(i) Jo G(p)dp = Jo F(p)dp ( Jo pg(p)dp = jo p0(p)dp, so the
means are the same), and rP rP (ii) J 0 G(p)dp 2 J 0 F(p)dp for all
p.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
304 : MONEY, CREDIT, AND BANKING
capital inefficient liquidadon market will do under wait-and-see
policy
Po P P1 l l I I ,p ' s I ' v '
positirre positiveNPV, negative net pledgeable negative NPV
income net pledgeable continuation continuation mcome
continuation FIG. 4
previously uninsurable risk. Furthermore, (7) shows that welfare
increases with an increase in the riskiness of p. The reason is the
option value of discontinuing the project. Ex ante uncertainty
about the liquidity shock p is a key ingredient in the de-
mand for liquidity. Suppose p were deterministic. If p 2 po =
pH(R _ B ), then
investors do not want to lend at date 0, since they know that
they will have to cover at date 1 a liquidity shock that exceeds
the income that can be pledged to them at date 2. There will never
be anything to distribute to investors. If p < pO, then the firm
is always solvent at date 1, and new claims can be issued at date 1
(that partially di- lute existing ones) in order to meet the
liquidity shock and continue; hence there is no need to hoard
reserves.
A good way of thinking about this issue is in terms of
insurance. A high liquidity shock is similar to an illness or an
accident, and a low liquidity shock is similar to an absence of
such a mishap. There is no scope for insurance if it is known in
advance whether there will be an illness or an accident.
1.3 Implications The first implication of our analysis has
already been stated: because the firm may
not be able to raise funds to pursue certain worthwhile projects
at date 1, it ought to make sure at date O that it can avail itself
of some liquidity. If negotiations run fric- tionessly at date 1,
the firm can raise up to the pledgeable income (poI) in the capital
market at that date. Thus the minimum amount of liquidity that the
firm must secure at date O is the difference between the targeted
reinvestment and the pledgeable in- come, that is, (p*-po)I.
At, this stage, the theory says nothing about the nature of this
liquidity buffer. The firm can hold liquid securities on a balance
sheet, or contract with an intermediary for a credit line. It is
only when liquidity is scarce at the economy level and liquid
as-
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 305
sets therefore sell at a premium that economic agents must
organize the dispatching of liquidity so as not to waste it. Waste
occurs when in some state of nature at date 1 assets are held by an
economic agent who does not need them while another agent is short
of liquidity. The transfer of liquidity between these two agents
cannot be arranged at date 1 because the latter agent is unable to
pledge enough to make it worthwhile for the former agent to lend.
We refer the reader to Holmstrom and Tirole (1998b) for a
discussion of wasted liquidity and the role of intermediaries as
liquid- ity pools.
For the moment the theory also says nothing about the maturity
of claims on the firm. Since the firm does not produce any cash
flow at date 1, all reimbursements/ dividends must be paid at date
2. We now relax this assumption by allowing the firm to generate a
profit at the intermediate stage.
2. INTERMEDIATE INCOME, FREE CASH FLOW, AND THE SOFT BUDGET
CONSTRAINT
Let us first extend the model by assuming that the firm
generates an exogenous (for simplicity), deterministic, and
verifiable cash flow, rI, at date 1, where r ' 0: see Figure 5.
To solve for the optimal financing contract, we do not need to
redo the analysis of section 1.2. Intuitively, since the pledgeable
income as well as the NPV are increased by the amount rI, the unit
cost of investment is no longer 1, but 1-r. The formulae above
remain unchanged except that the cutoff p* is determined by
tP F(p)dp = 1-r . (9)
Even though the results closely track those of the liquidity
shortage model of sec- tion 1, it is important to note that the
short-term income, while deterministic and
Date 0 . Date 1 . Date 2 *. Disbursement
x x * )< X )( . )< > Loan Investment . * Accrual of
Moral Date-2 agreement short-term hazard . income
income rI (RI or 0) * Realization pI No disbursement
of investment need
V .. Project is abandoned
FIG. S
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
306 : MONEY, CREDIT, AND BANKING
fully pledgeable, is not equivalent to an increase in the
borrower's equity A. Such an increase in equity would result in a
larger investment (as is the case here), but not in a modification
of the continuation rule. By contrast, condition (9) shows that the
larger the short-term profit, the lower the optimal threshold p*.
To understand this point, recall the scale expansion-insurance
trade-off between increasing debt capac- ity (by reducing p*) and
increasing the probability of continuation (by increasing p*). The
short-term revenue makes investment more attractive and therefore
makes it worth sacrificing continuation more in order to boost debt
capacity.
While young firms or firms with substantial investment needs are
well depicted by the liquidity-shortage model of section 1,
Easterbrook (1984) and Jensen (1986, 1989) considered the opposite
situation of "cash-rich firms" that generate large cash inflows
exceeding their efficient reinvestments needs. Such firms have
excess liquid- ity that must be "pumped out" in order not to be
wasted on poor projects, unwar- ranted diversification, perks, and
so forth. Jensen's (1989) list of industries with huge
free-cash-flow problems in the 1980s includes oil, steel, chemical,
television and radio broadcasting, brewing, tobacco, and wood and
paper products.
The liquidity-shortage and free-cash-flow problems are opposite
sides of the same coin. The key challenge in liquidity management
is to ensure that, at intermediate dates, just the right amount of
money is available for payment of operating expenses and
reinvestments. Whether this results in a net inflow (the
liquidity-shortage case) or outflow (the free-cash-flow case) is
important for corporate finance, but from an economic point of view
there is no conceptual distinction. Indeed we can merely
reinterpret the liquidity-shortage model as a free-cash-flow model
without changing the analysis.8
More formally, let us make the following free-cash-flow
assumption: r > p*. Under the free-cash-flow assumption, and
given that the entrepreneur cannot steal the intermediate income,
the entrepreneur would reinvest excessively. He would con- tinue as
long as p ' r.
To obtain the optimal reinvestment rule, an amount
P1-(r-p*)I,
must be pumped out of the firm. The payment P1 can be
interpreted either as repayment of short-term debt as in
Jensen or as a dividend payment as in Easterbrook. Note, though,
that the dividend must be capped by a covenant. Otherwise,
investors would want to pay dividends up to (r-po)I > P1 in
order to prevent the entrepreneur from reinvesting whenever the
liquidity shock exceeds the date- 1 pledgeable income p0. With this
interpretation, we see that covenants specifying a maximum of
dividend payment serve to protect the entrepreneur against
excessive liquidation.9
8. For a related study of free cash flow and liquidity, see
Krishnamurthy (1998). 9. This insight complements the standard and
important explanation for the existence of such
covenants. They are usually viewed as protecting creditors
against expropriation by the equityholders, who could use dividend
distributions and share repurchases to leave creditors with an
"empty shell."
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
. .
BENGT HOLMSTROM AND JEAN TIROLE : 307
Until now, we have assumed that the date-l income is exogenously
determined. Let us briefly consider the case in which income is
influenced by the entrepreneur's date-O effort eO. It is natural to
assume that effort shifts the distribution G(r|eO) of the date-l
income higher (in the sense of first-order stochastic
dominance).l
What is now the optimal liquidity management? Intuitively, there
are two ways in which the entrepreneur can be induced to exert
effort at date O.The first is standard and consists of increasing
the entrepreneur's stake (giving new stock options) in case of a
high date- 1 income. Alternatively, the entrepreneur may be given
more liquidity when the income is high. This means that the date-l
income is not mechanistically pumped out (redistributed to
claimholders), but is in part kept by the firm to reinvest if
needed. This yields a state-contingent continuation rule
p ' p*(r), (10) where p8( ) is an increasing function of the
intermediate income.
Figure 6 depicts the optimal continuation rule as a function of
the intermediate in- come for two cases. When date-O moral hazard
is light (the private gain from misbe- having is small and the
date-l income is a good indicator of managerial effort), the
continuation rule is strictly increasing in date-l income, and it
is inefficient to in- crease the entrepreneur's stake in date-2
income beyond what is needed to induce good date- 1 behavior. Two
new features appear when the date-O moral hazard is sub- stantial.
First, the continuation function p* (r) is steeper. Second, since a
cutoff p* (r) > P1 is inefficient even in first best, it will be
optimal to award extra stock op- tions to the entrepreneur after
p*(r) hits the ceiling P
p* (r) p* (r)
pl P1 /
/.- pO pO/Sa@-bb
O r r O r r (a) Light moral hazard (b) Substantial moral
hazard
at date O at date O FIG. 6
10. The following draws from Rochet and Tirole (1996), which
applies the model in Holmstrom and Tirole (1998b) to peer
monitoring and systemic risk.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
308 : MONEY, CREDIT, AND BANKING
On the other hand, for r small, a prescribed continuation rule
p8(r) < pO is not credible. The investors and the entrepreneur
are better off renegotiating the closure of the firm if p < pO.
The entrepreneur always prefers continuation, while the in- vestors
gain (pO-p)I. This explains the flat part of the dotted line in
Figure 6(b).ll
This is just the standard soft budget constraint problem: the
entrepreneur is opti- mally threatened with closure in case of bad
performance; however, the firm may be illiquid but solvent, which
gives the investors an incentive to renege on their com- mitment
and to rescue the firm. This, naturally, impacts negatively the
entrepreneur's initial incentive.
3. OPTIMAL RISK MANAGEMENT
We have seen that, even in a world of universal risk neutrality,
firms ought to ob- tain some insurance against liquidity shocks as
long as capital market imperfections prevent them from pledging the
entire value of their activity to new investors. Fol- lowing Froot,
Scharfstein, and Stein (1993), we can use this idea to derive an
ele- mentary explanation of corporate hedging.l2
In this section, we focus on optimal risk management when the
investors can ob- serve that the firm is indeed hedging and they
understand the correlation between the derivative instrument and
the rest of the firm's portfolio. In practice, investors may have
trouble knowing whether the entrepreneur uses the derivatives for
hedging or gambling purposes; we will study the corresponding
issues in section 4.
3.1 Costless Hedging To build intuition, let us begin with the
case in which the firm can at no cost hedge
against a shock on its date- 1 income. We extend the model of
section 2 by adding an exogenous shock E to the date-l income,
which becomes r-, with E(E|p) = O. For example, E may represent the
uncertainty in the exchange rate of a country in which the firm
produces or sells. An FX derivative allows the firm to stabilize
its income to r which, for simplicity, we assume deterministic.
It can be shown (and this will be a special case of the result
obtained in section 3.2) that it is optimal for the entrepreneur
and the investors to agree to fully hedge the risk. Intuitively,
the extra noise imposes undue variation in the liquidity available
to the firm and there is in this model no reason to create
ambiguity as to the continua-
11. In Figure 6(b), the dark line represents the optimal
continuation policy when renegotiation can be ruled out. The
optimal policy under renegotiation is depicted by the dotted
line.
12. Other explanations have been offered in the literature.
Stultz (1984) argues that corporate hedging allows managers to
obtain some insurance for their risky portfolio (stock options,
....) against shocks that they have no control over. While this
point is well taken Froot, Scharfstein, and Stein (1993) note that
managers could obtain such diversification by going to the
corresponding markets themselves, and so Stulz' argument relies on
a transaction cost differential. Tax reasons have also been
discussed in the liter- ature. See Smith and Stulz (1985) and Stulz
(1996) for more complete discussions.
Froot, Scharfstein, and Stein (1993) assume that capital markets
are imperfect (firms are unable to bor- row or issue equity in case
of a shortfall), so that "internal finance" is cheaper than
"external finance." They argue that the hedging activity shifts
internal funds from excess cash scenarios toward deficit sce-
narios. Our treatment follows their idea while endogenizing the
capital market imperfection.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 309
tion rule. For a given amount of liquid assets, the continuation
rule is p ' p* under hedging, and p + E < pt in the absence of
hedging. The absence of hedging will lead to reinvestment in some
states of nature with liquidity shock above pt and clo- sure in
other states with liquidity shock under p* (but higher than p0).
This is ineffi- cient because the marginal value of liquidity
decreases with p. The entrepreneur should not be held accountable
for a shock she does not control, provided it can be hedged at a
fair rate.
Two important remarks are in order. First, the investors may
need to check that the entrepreneur indeed hedges. For small FX
shocks E the entrepreneur wants to hedge only if the distribution
of liquidity shocks F(p) is concave. For large FX shocks, the
umbrella provided by the soft budget constraint (the willingness of
investors to fi- nance p < p0) makes it less likely that the
entrepreneur will want to abide by her promise to hedge. Similarly,
if information accrued to the entrepreneur between dates 0 and 1
concerning the date- 1 liquidity shock p, hedging could become
prob- lematic. In case of bad news and knowing that the project is
unlikely to be continued, the entrepreneur would want to undo the
hedge in order to "gamble for resurrection."
Second, the reader may be puzzled by the fact that the
entrepreneur is risk loving with respect to reinvestment needs p
see section 1-but risk averse with respect to cash-flow shocks e.
The reason is that reinvestment embodies an option value. The firm
can forgo reinvestment for bad realizations p, which makes
variation in p valu- able. In contrast, there is no scope for
opting out of a bad realization of the cash-flow shock e.
To sum up, we have argued that firms should be insulated from
all shocks that can be costlessly hedged in capital markets. Even
firms owned and financed by risk neu- tral investors should
purchase insurance against fire, theft, and other indiosyncratic
events in order to reduce the variability in the reinvestment rule.
Note that the inter- est in this fair-rate case is limited in that
it does not make any prediction as to which firms should hedge:
They should all fully hedge indiosyncratic shocks outside the
entrepreneur's control! We now extend the risk management model to
allow for par- tial hedging.
3.2 Costly Hedging While the death of a key employee is an
idiosyncratic shock, a number of other
contingencies, such as interest rate fluctuatioIls, against
which the firm can hedge in well-organized markets, are
macroeconomic in nature. Thus even with perfect mar- kets, such
risk cannot be covered at a fair rate. To the extent that the
firm's risk is pos- itively correlated with aggregate risk, the
firm must pay a premium in order to hedge. It is then optimal for
firms to engage in partial hedging. Intuitively, entrepreneurs/in-
siders should want their fair share of aggregate uncertainty. In
practice, hedging by the corporate sector is indeed quite
incomplete (Culp, Miller, and Neves 1998).
What kinds of firms should hedge more?l3 One potential
determinant is the firm's leverage. Stulz (1996) argues that firms
with little debt or highly rated debt have no
13. We abstract from the fact that some types of hedging are
subject to increasing returns (due in par- ticular to the need for
expertise), and so are best performed by large banks or firms.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
310 : MONEY, CREDIT, AND BANKING
need to hedge because such firms are better able to raise new
funds on the capital market in case of a liquidity shock. Scholesl4
in contrast argues that such Elrms should hedge and borrow more.
One goal of this section is to revisit this debate in the light of
our model of optimal liquidity management. Importantly, a firm's
leverage is endogenous, and may interact with factors that impact
its liquidity and hedging strat- egy. The same point applies to
other determinants of hedging. For example, Tufano (1996), studying
mining corporations, finds that firms in which management has a
bigger stake hedge more. Rationalizing this observation requires
looking at the fac- tors that impact the share of management.
Let us return to our canonical model, and denote by u ' O the
unit cost of hedging (in section 3.1 we assumed u = 0). That is, to
eliminate risk eI, the firm must pay Iu at date 0. We will still be
able to write the investors' break-even condition in terms of an
expected rate of return.l5 Let X denote the hedging ratio (0 < X
c 1). The initial investment then costs (1 + Bu)I at date 0, and
the firm has enough liquidity to con- tinue at date 1 if and only
if pI < [pt-e(1 -B)]I, where pt is, as earlier, the planned
level of available liquidity.
As in section 2, we allow for a (deterministic for simplicity)
date-l income rI. As we noted there, this date- 1 income in effect
reduces the per unit cost of investment at date 0, here to 1 +
Bu-r. The aggregate shock E can be viewed as a shock on date- 1
income, for example. The investors' break-even constraint states
that, in expectation the investors' total outlay is equal to their
benefit:
(1 + Bu-r)I-A + E[|P ( ) pf(p)dp] I = E [F(p 8-8(1-))]poI * (1
1)
As usual, the entrepreneur's utility is equal to the firm's
NPV:
[ ( ))]plI-[1 + Xu-r+ E [lP8-(l-4P+( )d ]
Let
1 + Xu - r + E [I0 p+(p)dp] ' E [F(p 8-8(1 ))]
14. Cited on page 16 of "A Survey of Corporate Risk Management,"
The Economist, February 10, 1996.
15. One may think of this model as being embedded in a model of
aggregate liquidity premia, as in Holmstrom and Tirole (1998a). But
it can also be embedded in a CAPM-type model in which premia are
associated with investors' risk aversion: The investors' break-even
condition can then be written in the equivalent risk-neutral form,
with u denoting the price of aggregate risk and the other
uncertainty being firm specific and diversified away.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 311
denote the unit cost of eXective investment, that is, the cost
of obtaining (in expecta- tion) one unit of unliquidated
investment. It is equal to the net expected investment cost divided
by the probability of continuation.
Using (11 ) and (12), the entrepreneur's objective filnction
is
b C(p8,)_po (13)
The extent of hedging X and the hoarding of (other) liquidity pt
are determined by a cost-minimization program:
min {c(pt, ) } . {p8 k}
This immediately yields the following separability result: The
extent, of hedging (X) is invariant to changes in variables that
affect only date-2 total benefit (Pl) and pledgeable income
(p0).
Thus, suppose that moral hazard increases (B increases, and
therefore p0 de- creases). The entrepreneur's share in date-2
profit goes up, investment is reduced and presumably leverage also
goes down.l6 Similarly, if we identify the rating of the long-term
(date-2) debt with the probability, PH, Of reimbursement, then the
extent of hedging is unrelated to the rating of the firm's
debt.
We now turn to factors that potentially affect the firm's
hedging behavior. Note first that for u = 0, the derivative of the
cost function c with respect to the hedging ratio X is equal to 0
at X = 1. We thus verify the intuition given in section 3.1: If
hedging is costless, the firm should fully hedge. On the other
hand, for u > 0, the de- rivative of c at X = 1 is positive, and
so partial hedging is optimal.
Rather than attempting a general analysis, we specialize the
model to a uniform distribution: F(p) = p on [0, 1], say. Then
1 + Bu-r +-[(pe)2 + (1-)2a2 ] c(pt, >,) = 2
Pt
where o2 is the variance of e. In the uniform case, the optimal
hedging ratio de- creases with the ratio of the cost of hedging
over the variance of the noise, that is, with the unit cost of
hedging:
16. "Presumably" refers to the fact that there are several
possible definitions of leverage (in this model as in reality),
depending on what is on and off the balance sheet (is the extra
liquidity (p*-po)I invested in securities or is it a credit line
secured from a bank?) and on how the maturity composition of
outsiders' claims is accounted for. We looked (in a non-exhaustive
way) at a couple of accounting conventions, for which the leverage
ratio was indeed increasing in the pledgeable income pO.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
312 : MONEY, CREDIT, AND BANKING
=1- 2.
T
In contrast, pt depends also on the date-l (expected) income
r:
(p8)2 = 2(1-r + u) _ u
An increase in the cost of hedging u reduces the hedging ratio
and (in the relevant range X ' O) decreases the hoarding of
liquidity, (pt-r)I, if positive. It also raises the cost c.
Last, consider the impact of an increase in short-term income r.
This increase does not affect the hedging ratio, but increases the
amount of short-term debt per unit of investment (r-p8), if
positive.l7
To conclude, hedging may or may not depend on factors that
affect other dimen- sions of liquidity management including short-
and long-term leverage. Any covari- ation between these endogenous
variables requires a detailed analysis of the factors that create
within-sample heterogeneity.
4. BANKING REGULATION AND RISK MANAGEMENT
4.1 Brief Overview of the Debate A famous accord designed by the
Basle committee and passed in 1988 provides an
international harmonization of prudential rules. This accord
focuses on credit risk. Its main objective is to define minimum
capital requirements for the banks' on- and off-balance-sheet
activities that depend on the identity and on the value of the
assets of the borrower (government, other banks, private sector, .
. .). This regulation of the banks' banking book raises a number of
practical as well as conceptual issues re- garding the nature of
the risk, the lack of correlation measures, the choice of histori-
cal cost versus market value accounting, the incentive to
securitize, and the definition of equity. [See Dewatripont and
Tirole (1994) for a description of the 1988 regulations and a
theoretical analysis of these issues.]
We will here focus on a specific but crucial issue: the
treatment of the bank's trad- ing book and market risk. The trading
book "means the bank's proprietary positions in financial
instruments (including positions in derivative products and
off-bal- ance-sheet instruments) that are intentionally held for
short-term resale and/or that are taken on by the bank with the
intention of benefitting in the short term from ac- tual and
expected differences between their buying and selling prices, or
from other price or interest rate variations, and positions in
financial instruments arising from
17. The short-term debt can alternatively be measured as
[r-(p*-pO)] if dilution is permitted, but this does not affect the
conclusion.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 313
matched principal brokering and market making, or positions
taken in order to hedge other elements of the trading book.''l8
Derivative products (for example, swaps) that are clearly
intended to hedge posi- tions in the banking book can be classified
by the banks as belonging to the banking book; they are then
excluded from the market risk measure, but become subject to the
credit risk (counterparty default risk) capital requirements.
When the initial accord was passed, it was already clear to the
regulators that the absence of a coherent treatment of market risk
was a serious shortcoming. The 1996 amendment to the 1988 Basle
accord imposed a second capital adequacy require- ment (CAR) for
the trading book. Banks were now subject to two separate CARs, one
on the banking book based on credit risk, and the other on the
trading book based on market risk.
The 1996 amendment has drawn criticism, and is viewed by its
designers only as an intermediate step toward a better framework.
There are several problems.
First, it is sometimes hard to distinguish between credit risk
and market risk. Sup- pose a borrower pledges real estate or equity
as collateral for a loan. The value of this collateral is subject
to significant market risk. Yet, for the bank, the risk is a credit
risk (how much will the bank receive if the borrower defaults?).
Similarly, a loan to a highly leveraged institution, such as Long
Term Capital Management or to an emerging country, involve a fair
amount, of market risk, even though formally the risk is one of
counterparty default. As a last example, a loan take-down is not a
ran- dom event in the business cycle, since borrowers are much more
likely to draw on their credit facilities in bad times. Some of
these examples also show how conceptu- ally difficult it can be to
distinguish between the trading and the banking books (even though
in practice the division is usually quite easy since the two
activities are car- ried out in separate units within the bank).
This raises a concern about using a piece- meal approach, a concern
we will come back to.
Second, there is an issue regarding the measurement of the
trading portfolio's overall risk. Unlike the regulation of the
banking book, which by and large ignores any correlation between
its component risks, the capital adequacy requirement for the
trading book is meant to be based on an aggregate view of the
trading portfolio. Measuring such risk requires sophisticated
econometric techniques and good data concerning the volatility of
the elements in the portfolio (bonds of various maturi- ties,
currencies, derivative instruments, etc.), of the covariance
matrix, and of the ef- fectiveness of the state-contingent
portfolio reallocations. There are also problems with using
historical simulations to predict future movements, relying on a
small amount of data about extreme lower tail events (Gaussian
approximations have thin- ner tails than the empirical returns) and
choosing the time horizon.
The 1996 Amendment specifies a ten-day Value-at-Risk (VaR) with
a 99 percent confidence level. That is, the capital requirement for
the trading book (to be added to that on the banking book) is
proportional to the maximum loss in the bank's portfo-
18. Basle Committee on Banking Supervision (1996), Amendment to
the Capital Accord to Incorpo- rate Market Risks.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
314 : MONEY, CREDIT, AND BANKING
lio that can occur within ten days with probability more than 1
percent. The bank must meet, on a daily basis, a capital
requirement on the trading book expressed as the higher of (i) its
previous day's VaR number and (ii) an average of the daily VaR
numbers on each of the preceding sixty business days, multiplied by
a factor that is at least 3 and can be brought up to 4 depending on
the outcome of backtesting.
How is this VaR computed? The Basle committee allows banks to
use their own internal model, provided this model is approved by
the regulators. The approval process is based on a number of
criteria: skill of the staff (in risk control and back of- fice
management), track record on accuracy in measuring risk, and stress
testing to cover a range of factors than can create extraordinary
losses. The use of an internal model is therefore a priori limited
to sophisticated (and often large) banks. Other banks use the
default option, the so-called "standardized measurement method" de-
scribed in the amendment, which relies on a more prescriptive and
mechanical treat- ment of risk.
It is also worth noting that another approach, the Precommitment
Approach (PCA), originating at the U.S. Federal Reserve (see Kupiec
and O'Brien l995a and b),was proposed as an alternative to this
Internal Model Approach (IMA). The PCA proponents argued that the
IMA requires substantial regulatory expertise and inde- pendence,
and that the disclosure of the internal model is useful only if the
risk struc- ture is highly correlated over time. They argued in
favor of a more flexible and lighter regulatory approach in which
the bank itself would assess its maximum pos- sible loss, which in
turn would determine the capital requirement. Incentive compat-
ibility would then be ensured by ex post penalties. The PCA
proposal has been criticized on grounds that ex post penalties are
particularly limited in situations of undercapitalization .
We refer to Rochet (1999) for a theoretical perspective on the
relationship be- tween IMA and PCA (as Rochet notes, the PCA is an
"indirect mechanism" while the IMA is "direct mechanism" in the
terminology of mechanism design. The two ought to be equivalent if
the risk structure changes quickly over time and the regula- tors
lack expertise to see through internal models). The criticism
leveled at PCA con- cerning the difficulty of implementing ex post
punishments applies as well to IMA to the extent that the
inspection of the model does not suffice to ensure truth telling.
The benefit of the PCA is its greater flexibility in letting the
bank announce private infor- mation about future innovations in the
risk structure and in relying less on regulatory expertise and
benevolence. The benefit of the IMA is that if either the risk
structure is highly correlated over time or/and the regulators are
sufficiently staffed and com- petent and are able to see through a
complex model, backtesting and inspection allow early action by the
supervisory body. There is then a lower reliance on a diffi- cult
ex post intervention.
19. See Daripa, Jackson, and Varotto (1997). For general
discussions of the dangers of and limit to ex post penalties, see
the Federal Deposit Insurance Corporation Improvement Act (1991),
Goodhart et al. (1998), and Dewatripont and Tirole (1994). Monetary
penalties, public disclosures of financial conditions, and
increases in the deposit insurance premium are likely to trigger
gambling for resurrection or runs on the interbank market. Capital
charges, inspections, and line-of-business restrictions are likely
to be more effective punishments.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 315
4.2 What Is It About? The introduction of a CAR on the trading
book is motivated by a concern that
large losses on the trading book may dry up the bank's liquidity
and thereby jeopar- dize the banking book. This spillover from the
trading book onto the banking book raises the first question: Why
is there a trading book in a first place? After all a bank's
proprietary trading could be performed through a separate affiliate
with no possibility for cross-subsidization between the two arms.
The trading arm of the bank would then be just another securities
firm. This question is rarely asked in the debate on the prudential
regulation of market risk, but seems important when think- ing
about desirable rules.
There are two possible responses to the separation point. The
first is that there are returns to scale in trading activities, in
particular, due to the need for expertise. Thus, the knowledge used
to hedge the banking book can be used for other purposes as well.
It would be interesting to measure the extent of increasing returns
of this kind.
Second, the trading book may really be about hedging the banking
book, or, more precisely, about providing non-obvious hedges that
insure against risk of the overall banking portfolio (we have seen
that financial transactions that are clearly meant to hedge a
specific risk can be switched to the trading book). In our view,
this is an im- portant argument in favor of an integrated or "whole
bank" approach. A weaker ver- sion of this argument is that, even
if the two books are stochastically independent rather than
negatively related, the bank might still reduce its capital
requirements by cross-pledging their outcomes (as in Diamond
1984).
Either way, it is hard to see the rationale for a piecemeal
approach: Either the two activities are unrelated and then they are
best separated into independent entities, which cannot
cross-subsidize each other; or they are linked and then a
whole-bank (integrated) approach would be preferable.
Let us pursue this whole-bank approach using the second
rationale for integration: The trading book serves to hedge the
banking book. In the framework of this paper, assume that the bank
is subject to a dual income shock at date 1: rb on the banking book
and rt on the trading book. We can entertain three different
hypotheses:
(H1) Both rb and rt are observed. (H2) Only the sum rb + rt is
observed. (H3) rt is observed, rb is not. In all cases, we will
assume that the regulators are unable to verify that the bank is
hedging. Let us look at the implications of each hypothesis:
(H1) If both income shocks are separately observed by the
regulators, then the bank should be, ceteris paribus, punished when
the realizations of the shocks are positively related and rewarded
when they are negatively related. An implication of this is that
the bank should see some of its surplus expropriated when both
shocks to the banking book and to the trading book are favorable.
The amendment in contrast focuses on the extreme lower tail and is
unable to yield discipline through action in states of nature (very
favorable ones) in which discipline is easiest to enforce (as we
have seen, punishments are hard to implement when the bank is
undercapitalized).
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
316 : MONEY, CREDIT, AND BANKING
(H2) In the polar case, the regulator cannot tell the two income
shocks apart. The bank is able to carry out transfers between the
two books, perhaps through arrange- ments with a third party such
as a highly levered institution. The bank may have an incentive to
carry out such transfers either to avoid punishment or to minimize
capi- tal charges. The scope for this fungibility depends on what
counts as hedging. In the most narrow definition, there is little
scope for switching income from one book to the other. But then, of
course, there is little recognition of the role of the trading book
as a hedge against uncertainty in the banking book. Conversely, a
broad definition of a hedge recognizes this role, but is subject to
manipulations.
Under (H2), the regulators should again adopt a whole bank
approach. The diffi- culty is that the regulators may not know
whether the bank is using the financial in- struments (the trading
book) for hedging or gambling purposes.
Formally, the issue is one of "double moral hazard." The bank
affects its income in two ways: through effort it shifts the
distribution of income toward higher values (in the sense of
first-order stochastic dominance); through risk-taking or hedging,
it af- fects the riskiness of the distribution (in the sense of
second-order stochastic domi- nance). The literature on this type
of agency problem is unfortunately quite small.20
Alger (1999) considers a simple example with a two-date (no
liquidity shock) structure. A risk-averse banking entrepreneur with
capital/assets A, invests an amount I borrowing I-A at date 0, and
then selects some effort e, together, possi- bly, with a choice of
riskiness. The date- 1 final income is given by
y = (e + 8 + 68)I,
where E is a shock that can be insured in derivatives markets.
Regulators lack exper- tise to control the bank's hedging behavior.
If allowed to play with derivatives, the banking entrepreneur can
select to hedge (6 =-1) or to gamble (6 = + 1). So 6 is a second
dimension of moral hazard. Alternatively, the entrepreneur may be
prohibited from playing the derivatives game, which automatically
results in 6 = O (this is equivalent to a full separation between
the banking and trading books in this model).
The entrepreneur has expected utility E[U(Rf)]-g(e), where Ut )
is a concave function of the banking firm's profit Rf, and g( ) is
the disutility of effort. In Alger's example, e E {O, 1}, and E E
{-1,0, 1}. Entrepreneurial risk aversion argues in favor of
removing the noise E through hedging. However, the entrepreneur may
then shirk and gamble so as to benefit from the upper tail of the
distribution. Alger character- izes the minimum capital requirement
when derivatives are allowed and when they are not.
Intuition suggests that a higher capital requirement should be
imposed when the bank wants to use derivatives. Alger shows that
this intuition may or may not be cor- rect. The possibility of
hedging may increase the signal-to-noise ratio and allow a better
control of the first form of moral hazard. So introducing the risk
dimension of
20. See, however, Bester and Hellwig (1987) and Bolton and
Harris (1999) for structured examples.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 317
moral hazard may actually help control the effort dimension of
moral hazard and lower the CAR.21
While the effect unveiled by Alger is quite robust the
conclusion that a lower CAR may be associated with the use of
derivatives still seems a bit counteractive. The development of
more general frameworks combining both types of moral hazard is
eagerly awaited.
(H3) Last let us consider the case in which only rt is observed.
To be certain, this case can only be a metaphor. Regulators do
measure capital ratios for banking books and therefore are able to
assess their evolution.
What (H3) is a metaphor for is a situation in which information
about the banking and trading books accrue at different
frequencies. Information about the quality of loans for example
accrues much more slowly than information about the market value of
the trading book. A more plausible (but still highly stylized)
representation of (H3) in the context of our model would then be
one in which rb is learned at date 1 while there are two
realizations of rt (shocks on the trading book value) at date 1/2
and 1, say. The exploration of models in which different portfolios
(or at least their measurability by the regulator) move at
different frequencies seems a fruitful avenue for research.
5. CONCLUDING REMARKS
The paper has developed a unified and optimal contracting
approach to the choice of capital structure, liquidity, and risk
management, and their relationship to the soft budget constraint
and free-cash-flow theories. It has investigated the determinants
of hedging and analyzed the incorporation of the market risk in
banking regulation. Much remains to be done, though. As discussed
in section 47 little is known about risk management when the risk
structure is unknown to outsiders (investors for a firm, regulators
for a bank). This important policy issue actually raises a much
more general question of how to monitor a corporation's use of
liquidity. This and other important microeconomic issues concerning
liquidity and risk management await fu- ture research.
LITERATURE CITED
Alger, G. "The Beauty in the Beast: Derivatives, Double Moral
Hazard, and Regulation." Mimeo, GREMAQ, University of Toulouse,
1999.
Basle Committee on Banking Supervision. "Amendment to the
Capital Accord to Incorporate Market Risks." January 1996.
21. To take an extreme, if unpalatable case, suppose that e E {
0, l }, but, unlike in Alger's paper, is a continuous variable.
Then, y = 1 for sure under {e = 1, hedging} and with probability 0
under any other policy. Hence, any deviation from {e = 1, hedging}
is detected with probability 1, and so the use of de- rivatives
completely eliminates moral hazard (the first best obtains).
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
318 : MONEY, CREDIT, AND BANKING
Bester, H., and M. Hellwig. "Moral Hazard and Equilibrium Credit
Rationing: An Overview of the Issues." In Agency Theory,
Information and Incentives, edited by G. Bambers and K. Spremann.
Heidelberg: Spririger Verlag, 1987.
Bolton, P., and C. Harris. "The Continuous-Time Principal-Agent
Problem: Frequent Moni- toring Contracts." Mimeo, Princeton
University and University of Cambridge, 1999.
Caballero, R., and A. Krishnamurthy. "Emerging Markets Crises:
An Asset Markets Perspec- tives." Mimeo, Department of Economics,
MIT, 1999.
Caillaud, B., G. Dionne, and B. Jullien. "Corporate Insurance
with Optimal Financial Con- tracting." Economic Theory, 2000.
Crane, D. "Managing Credit Lines and Commitments." Study
prepared for the Trustees of the Banking Research Fund Association
of Reserve City Bankers, Graduate School of Business
Administration, Harvard University, 1973.
Culp, C., M. Miller, and A. Neves. "Value at Risk: Uses and
Abuses." Journal of Applied Cor- porate Finance 10(4) (1998),
26-38.
Daripa, A., P. Jackson, and S. Varotto. "The Pre-Commitment
Approach to Setting Capital Re- quirements." Bank of England and
SIB, Financial Stability Review (1997 Autumn), 42-50.
Dewatripont, M., and E. Maskin. "Credit and Efficiency in
Centralized and Decentralized Economics." Review of Economic
Studies 62 (1995), 541-55.
Dewatripont, M., and J. Tirole. The Prudential Regulation of
Banks. MlTPress, 1994. Diamond, D. "Financial Intermediation and
Delegated Monitoring." Review of Economic
Studies 51 (1984), 393 -414. Duffie, D., and J. Pan. "An
Overview of Value at Risk." Journal of Derivatives 4(3) (1997),
749. Dybvig, P., and W. Marshall. "The New Risk Management: The
Good, the Bad, and the Ugly."
Federal Reserve Bank of St Louis Review 79 (6) (1997), 9-22
Easterbrook, F. "Two-Agency-Cost Explanations of Dividends."
American Economic Review
74 (1984), 650-59. Fite, D., and P. Pfleiderer. "Should Firms
Use Derivatives to Manage Risk?" In Risk Manage-
ment: Problems and Solutions, edited by W. Beaver and G. Parker.
Stanford, 1995. Froot, K., D. Scharfstein, and J. Stein. "Risk
Management: Coordinating Corporate Invest-
ment and Financing Policies." Journal of F-inance 48 (1993),
1629-58. Goodhart, C., P. Hartman, D. Lewellyn, L. Rojas-Suarez,
and S. Weisbrod. "Financial Regula-
tion: Why, How, and Where Now?" Mimeo, Bank of England, 1998.
Gordy, M. "A Comparative Anatomy of Credit Risk Models." FEDS 47,
Federal Reserve
Board, Washington, 1998. Greenbaum, S., and A. Thakor.
Contemporary Financial Intermediation. Dryden Press, Har-
court Brace College Publishers, 1995. Harrington, R. "Asset and
Liability Management by Banks." OECD, 1987. Holmstrom, B. "Moral
Hazard and Observability." Bell Journal of Economics 10 (1979),
74-91. Holmstrom, B., and J. Tirole. "LAPM: A Liquidity-Based
Asset Pricing Model." mimeo,
1998a. . "Private and Public Supply of Liquidity." Journal of
Political Economy 106 (1998b),
1-40. Jensen, M. "Agency Costs of Free Cash Flow, Corporate
Finance and Takeovers." American
Economic Review 76 (1986), 323 -29. . "Eclipse of the Public
Corporation," Harvard Business Review (1989), 61-74.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
-
BENGT HOLMSTROM AND JEAN TIROLE : 319
Krishnamurthy, A. "Essays in Corporate Finance and
Macroeconomics." Chapter 3, Ph.D. the- sis, Department of Finance,
MIT, 1998.
Kupiec, P., and J. O'Brien. "Recent Developments in Bank Capital
Regulation of Market Risks." Federal Reserve Board, Finance and
Economics discussion paper, 1995a.
. "A Precommitment Approach to Capital Requirements for Market
Risk." Federal Re- serve Board, Finance and Economics discussion
paper, 1995a.
Rochet, J. C. "Solvency Regulations and the Management of
Banking Risks." European Eco- nomic Review, Papers, and Proceedings
43(4-6) (1999), 981-90.
Rochet, J. C., and J. Tirole. "Interbank Lending and Systemic
Risk." Journal of Money, Credit, and Banking 28 (1996), 733-62.
Rothschild, M., and J. Stiglitz. "Increasing Risk I: A
Definition." Journal of Economic Theory 2:3 (1970), 225-43.
. "Increasing Risk II: Its Economic Consequences." Journal of
Economic Theory 3:1 (1971), 66-84.
Smith, C., and R. Stulz. "The Determinants of Firms' Hedging
Policies," Journal of Financial and Quantitative Analysis 20
(1985), 391-405.
Stulz, R. "Optimal Hedging Policies." Journal of Financial and
Quantitative Analysis 19 (1984), 127-40.
. "Rethinking Risk Management." Journal of Applied Corporate
Finance 9 (1996), 8-24.
Thakor, A., H. Hong, and S. Greenbaum. "Bank Loan Commitments
and Interest Rate Volatil- ity." Journal of Banking and Finance 5
(1981), 497-510.
Tufano, P. "Who Manages Risk? An Empirical Examination of the
Risk Management Prac- tices of the Gold Mining Industry." Journal
of Finance 51 (1996), 1097-1138.
von Thadden. "Long-Term Contracts, Short-Term Investment and
Monitoring." Review of Economic Studies 62 (1995), 557-75.
This content downloaded from 210.160.35.61 on Fri, 20 Mar 2015
12:33:55 UTCAll use subject to JSTOR Terms and Conditions
Article Contentsp. [295]p. 296p. 297p. 298p. 299p. 300p. 301p.
302p. 303p. 304p. 305p. 306p. 307p. 308p. 309p. 310p. 311p. 312p.
313p. 314p. 315p. 316p. 317p. 318p. 319
Issue Table of ContentsJournal of Money, Credit and Banking,
Vol. 32, No. 3, Part 1 (Aug., 2000), pp. 295-442Front MatterMoney,
Credit, and Banking LectureLiquidity and Risk Management [pp.
295-319]
Political Regime Change and the Real Interest Rate [pp.
320-334]Household Credit and the Monetary Transmission Mechanism
[pp. 335-356]International Lending by U.S. Banks [pp. 357-381]The
Cyclical Relationship between Output and Prices: An Analysis in the
Frequency Domain [pp. 382-399]Perfect Competition and the Effects
of Energy Price Increases on Economic Activity [pp.
400-416]Interest Rates, Inflation, and Federal Reserve Policy Since
1980 [pp. 417-434]Does a Bias in FOMC Policy Directives Help
Predict Intermeeting Policy Changes? [pp. 435-441]Erratum: Are
Banks Risk Averse? Intraday Timing of Operations in the Interbank
Market [p. 442]Back Matter