Liberalization, Moral Hazard in Banking, and Prudential Regulation: Are Capital

Requirements Enough?
Author(s): Thomas F. Hellmann, Kevin C. Murdock and Joseph E. Stiglitz
Source: The American Economic Review , Mar., 2000, Vol. 90, No. 1 (Mar., 2000), pp. 147-
165
Published by: American Economic Association

Stable URL: [Link]

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in 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 support@[Link].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
[Link]

is collaborating with JSTOR to digitize, preserve and extend access to The American Economic
Review

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 Liberalization, Moral Hazard in Banking, and Prudential
Regulation: Are Capital Requirements Enough?

By THOMAS F. HELLMANN, KEVIN C. MURDOCK, AND JOSEPH E. STIGLITZ*

In a dynamic model of moral hazard, competition can undermine prudent bank

behavior. While capital-requirement regulation can induce prudent behavior, the
policy yields Pareto-inefficient outcomes. Capital requirements reduce gambling
incentives by putting bank equity at risk. However, they also have a perverse effect
of harming banks' franchise values, thus encouraging gambling. Pareto-efficient
outcomes can be achieved by adding deposit-rate controls as a regulatory instru-
ment, since they facilitate prudent investment by increasing franchise values. Even
if deposit-rate ceilings are not binding on the equilibrium path, they may be useful
in deterring gambling off the equilibrium path. (JEL G2, E4, L5)

Banking crises are pervasive. In the last two (S&L) crisis in the United States, which re-
decades, the frequency of severe banking crises sulted in estimated losses of $180 billion or 3.2
has increased significantly. Banking crises are percent of GDP, and the ongoing banking crisis
important not just because of the devastation in Japan, where some estimates of nonperform-
that they bring to one particular sector of the ing loans approach 25 percent of GDP.1
economy, but because typically the shock Prudential regulation is meant to protect the
waves affect the entire economy. In the nine- banking system from these problems. Tradition-
teenth century, most of the U.S. economy's ally, it has consisted of a mixture of monitoring
economic downturns were related to financial individual transactions (ensuring, for instance,
panics. The budgetary consequences for gov- that adequate collateral was put up), regulations
ernments, which often bear a significant part of concerning self-dealing, capital requirements,
the costs of the bailout, cannot be ignored ei- and entry restrictions. In some countries, restric-
ther. A compilation of cases over the past two tions were placed on lending in particular areas:
decades by the World Bank shows costs ranging many East Asian countries, for example, used to
up to 40 percent of GDP. Probably the best have restrictions on real estate lending.2 Finally,
known examples are the savings and loan many countries imposed interest-rate restric-
tions.3 Concerns about bank runs also led many
countries to provide deposit insurance and to
* Hellmann: Graduate School of Business, Stanford Uni-
versity, Stanford, CA 94305; Murdock: Graduate School of
Business, Stanford University, Stanford, CA 94305, and
McKinsey & Company, 21 South Clark Street, Chicago, IL ' Typically, recovery rates on nonperforming loans are
60603; Stiglitz: The World Bank, 1818 H Street, NW, less than 50 percent. If this holds true for Japan, losses could
Washington, DC 20433, and on leave from the Department exceed 10 percent of GDP.
of Economics, Stanford University, Stanford, CA 94305. 2 The intent of these restrictions was only partially to
We thank Masahiko Aoki, Sudipto Bhattacharya, Jerry enhance the safety and soundness of the banking system;
Caprio, Serdar Dinm, John Roberts, Paul Romer, two anon- these restrictions were also intended to direct credit toward
ymous referees, and seminar participants at the Federal what were viewed at the time as more productive invest-
Reserve Bank of New York, the University of Mannheim, ments.
Stanford University Department of Economics and Gradu- 3 Interest-rate restrictions, like many of the other restric-
ate School of Business, the William Davidson Institute at tions, served several other objectives; low interest rates had
the University of Michigan, and the World Bank. Any significant positive effects on government budgets in peri-
remaining errors are ours. This paper represents the views of
ods in which the debt GDP ratio was high, such as after
the authors and does not necessarily represent that of any World War II. We discuss this in Hellmann and Murdock
organization with which they are or have been affiliated. (1997) and Hellmann et al. (1997, 1998a).

147

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 148 THE AMERICAN ECONOMIC REVIEW MARCH 2000

establish central banks to serve as lenders of last This paper asks whether financial liberaliza-
resort. tion itself could be part of the problem. Some
Over the past decade, several changes in the recent empirical studies suggest it may be: large
systems of prudential regulation have occurred. interest-rate increases, which are associated
First, given the increased number and complex- with financial-market liberalization, are also
ity of transactions, there has been greater em- systematically related to financial crises (see
phasis on monitoring banks' risk-management Asli Demirgiiu-Kunt and Enrica Detragiache,
systems, and less emphasis on monitoring indi- 1997, 1998). This paper approaches the ques-
vidual transactions. Second, in a wave of tion from a theoretical perspective. Financial-
financial-market liberalization, interest rates market liberalization increases competition;
have been deregulated, and restrictions on the competition erodes profits; lower profits imply
asset choices of banks have been lifted. Third, lower franchise values (i.e., the capitalized
greater emphasis has been placed on capital value of expected future profits); and lower
requirements, typically using the Bank of Inter- franchise values lower incentives for making
national Settlements (BIS) standards of the good loans, increasing the moral-hazard prob-
Basle Accord. lem. With sufficient competition banks will find
As these changes have occurred, financial it desirable to gamble. There is thus an incon-
crises have become more frequent (see Gerard sistency of interest-rate liberalization and pru-
Caprio, Jr. and Daniela Klingebiel, 1997; Klaus dential bank behavior.
P. Fischer and Martin Chenard, 1997). Most We ask whether an increase in capital re-
observers agree that moral hazard plays an im- quirements can offset the adverse effects of
portant role in these failures. Edward Kane liberalization. If banks hold sufficient capital,
(1989) and Rebel A. Cole et al. (1995) docu- they internalize the adverse consequences of
ment the problem of "gambling on resurrec- gambling and thus will choose to invest pru-
tion": banks choose a risky asset portfolio that dently. While it is possible to combat moral
pays out high profits or bonuses if the gamble hazard with capital requirements, we find that
succeeds but leaves depositors, or their insurers, banks must be forced to hold an inefficiently
with the losses if the gamble fails.4 It has been high amount of capital. It is impossible to im-
suggested that deposit insurance is the problem, plement any Pareto-efficient outcome using just
since it reduces the incentives for depositors to capital requirements as the tool of prudential
monitor.5 Others have argued that it makes little regulation. That is, freely determined deposit
difference whether countries have a formal sys- rates are inconsistent with Pareto efficiency.
tem of deposit insurance since, in the event of a We then ask whether there is any way to
financial crisis, there will be a bailout.6 More- implement outcomes along the Pareto frontier.
over, the fact that there have been financial The reason why capital requirements alone are
crises in countries with and without formal insufficient is because, with freely determined
deposit-insurance systems suggests that elimi- deposit rates, banks have excessive incentive to
nating formal deposit insurance by itself does compete for deposits by offering higher rates.
not solve the problem. Capital requirements only become effective
when they raise banks' costs sufficiently to im-
pact the banks' willingness to pay out high
4 George A. Akerlof and Paul M. Romer (1993) further deposit rates. But if capital requirements are an
elaborate on the moral hazard, arguing that banks may use indirect way of lowering deposit rates, why not
fraudulent lending practices (such as insider lending) to
control deposit rates in the first place? Indeed,
"loot" banks. In this case bank managers extract value out of
the banks even if this leads to insolvency. we show that any Pareto-efficient outcome can
5One may question, however, both the efficacy and be implemented by a combination of deposit-
desirability of depositor monitoring, given that monitoring rate controls and capital requirements.
is a public good. See Stiglitz (1985, 1992, 1994) and the
The benefit of deposit-rate controls for pro-
discussion in our working paper (Hellmann et al., 1998b).
moting stability in the banking sector seems to
6 As one commentator quipped, there are two kinds of
countries: those that have deposit insurance, and those that have been intuitively understood by regulators.
don't yet know that they have it. In discussing the "Temporary Interest Rate Ad-

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. 1 HELLMANN ET AL: MORAL HAZARD IN BANKING 149

justment Law" of 1947, which regulated Japa- cial liberalization undermined franchise value
nese deposit rates, Yoshio Suzuki (1987 p). 41) and opened up new gambling opportunities, few
notes: compensatory actions were taken to strengthen
regulatory oversight. Capital requirements were

Of course, the purpose of this law was to not increased, and other instruments of pruden-
prevent interest rate competition that was tial control were eliminated. The capabilities of
destructive to the profitability of financial the agencies in charge of regulatory oversight
institutions. That is, the law aimed more were not upgraded, and in fact, they frequently
at ensuring stable business condition for declined.9
financial institutions through external In our model we examinie the moral-hazard
controls on competition than at ensuring problem of banks in a dynamic setting. Banks
prudent management of such institutions
can either invest in a prudent asset yielding high
based on internal controls.
expected returns or in an inefficient gambling
Similarly, in the United States one of the main asset that can yield high private returns for the
regulatory changes preceding the S&L crisis bank if the gamble pays off but imposes costs
was the abolition of regulation Q, which had on depositors if the gamble fails. If markets are
imposed deposit-rate controls. Michael C. Kee- sufficiently competitive, the bank earns rela-
ley (1990) finds a direct relationship among tively little from prudent investment, but the
reforms that increased competition, the reduc- bank can always capture a one-period rent from
tion in franchise value of banks, and an increase gambling. Thus increased competition tends to
in the number of bank failures during the 1980's promote gambling in the banking sector.
in the United States. Some form of prudential regulation is then
One view of the recent financial crises in East necessary to induce banks to invest prudently.
Asia and the weakened financial system in Ja- Capital requirements force banks to have more
pan is that the problems arose at least in part as of their own capital at risk so that they internal-
a result of financial-market liberalization.7 First, ize the inefficiency of gamibling. Clearly, once
financial-market liberalization reduced barriers banks have enough of their own capital in-
to entry and increased competition: more for- vested, banks can be induced to invest in the
eign banks were allowed in the country; restric- prudent asset. This paper develops an alterna-
tions on opening branches were reduced; and tive form of prudential regulation: the use of
where there were deposit-rate ceilings, these deposit-rate ceilings to create franchise value
were either eliminated or reduced. All these for banks. Franchise value is the discounted
reduced profitability and thus franchise value of stream of future profits for the bank, a value that
existing domestic banks. Second, other aspects can only be captured if the banks stays in op-
of the liberalization agenda reduced restrictions eration. If the bank gambles and fails, it loses its
imposed on banks. A range of new activities franchise value. Franchise value acts as intan-
that had previously been precluded, such as gible capital, which can be a substitute for tan-
many derivative trades and foreign currency gible capital. If a bank has sufficient franchise
transactions, opened up many new ways for value, it will choose to invest in the prudent
banks to engage in gambling activities. More- asset.
over, restrictions on real estate lending were Given two potentially effective forms of
eliminated.8 Third, at the same time that finan-

(particularly when banks do not mark to market) because

7 For further discussion on the dynamics underlying thebanks can sell assets that have increased in value but hold
crises in East Asia in 1997, see Steven Radelet and Jeffrey on to assets that have decreased in value.
D. Sachs (1998) and James Tobin (1998). 9 For example in Thailand (arid elsewhere too) the rapid
8 Real estate lending expands the opportunity for exces- increases in salaries in the private sector, combined with
sive risk taking. This is because there may be great volatility fiscal constraints on public authorities, led to a mismatch
in the underlying land prices. To the extent that it is difficult between public and private salaries, and the central banks
to ascertain the market value of real estate assets, financial and government lost many of their most talented people to
institutions can also obfuscate the value of their capital the private sector.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
150 THE AMERICAN ECONOMIC REVIEW MARCH 2000

prudential regulation, this paper seeks to deter- the interbank elasticity will increase market-
mine the optimal form of prudential regulation. stealing incentives, and this creates the link
The Pareto frontier is given by the lowest level between liberalization and financial crises.
of capital for each level of deposit rates, which Of course, the policy we study in detail in this
is consistent with the bank's choosing to invest paper (deposit-rate ceilings) is not the only pol-
in the prudent asset. icy that could be used to generate improvements
To see why it impossible to have a Pareto- over using capital requirements alone. Other
efficient equilibrium where banks freely deter- policy instruments that the government could
mine deposit rates, consider the incentives of a apply include asset-class restrictions, entry re-
bank where all of its competitors are choosing strictions, and enhancing direct supervision.
to offer the Pareto-efficient deposit rate. Along The goal of each of these policies is either to
the Pareto frontier, the bank is exactly indiffer- limit the scope of the bank's ability to engage in
ent between gambling and the prudent asset moral-hazard behavior or to increase the posi-
(i.e., the total profits that the bank earns on a tive incentives of the bank to invest prudently.
fixed amount of deposits are identical). Then, it Determining the optimal application of all of
must be the case that the bank earns a higher these policies is beyond the scope of this paper;
expected margin on the gambling asset than on the point of the paper is first to show the inad-
the prudent asset because, when the bank gam- equacy of capital requirements alone, and sec-
bles, it will forfeit its franchise value should the ond to identify the potential value of including
gamble fail. If this bank offered the same de- deposit-rate ceilings as an instrument of pruden-
posit rate as its competitors, it would invest in tial regulation.
the prudent asset and get its market share. If, We discuss some of these broader policy issues
however, the bank were to offer a slightly in previous work (Hellmann et al., 1996, 1997,
higher deposit rate, it would capture additional 1998a; Hellmann and Murdock, 1997). In those
deposits, upon which it could earn a higher papers, we consider a framework of financial-
margin if it were to gamble. Because of the market regulation. We propose a set of policies
market-stealing effect, each individual bank has that we term "financial restraint," which we dis-
an incentive to defect from any candidate equi- tinguish from financial repression. An important
librium along the Pareto frontier. We thus find difference is that with financial restraint,
that freely determined deposit rates are incon- interest-rate controls are used to improve the effi-
sistent with Pareto efficiency. ciency of private financial markets (as shown in
The response that this paper focuses on is the the current paper), whereas with financial repres-
use of deposit-rate controls to create franchise sion, interest rates are typically a mechanism for
value. Deposit-rate ceilings effectively combat the government to extract rents from the private
this market-stealing effect by precluding banks sector.
from competing through inefficiently high de- The theoretical model builds on the work of
posit rates. We show that, with an appropriate Sudipto Bhattacharya (1982), which noted the
ceiling in place, all Pareto-efficient outcomes usefulness of deposit-rate controls in a simple
can be implemented. static model.10 Caprio and Lawrence H. Sum-
Our results suggest that there is a clear theo- mers (1996) emphasize the importance of fran-
retical connection between liberalization and chise value for prudential regulation.' l The
the degree of the moral-hazard problem. We analysis of capital requirements is related to
find that freely determined deposit rates are Jean-Claude Rochet (1992), who explains how
inconsistent with Pareto efficiency. Banks offer
inefficiently high deposit rates in an effort to
steal share from their rivals. Liberalization usu- 10 Bruce D. Smith (1984) also has a model where deposit-
ally has a stated goal of increasing competition rate controls reduce the likelihood of financial instability. In his
model deposit-rate controls eliminate an instability problem a
in the financial sector. This will have the effect
la Michael Rothschild and Stiglitz (1976).
of increasing the interbank elasticity of deposits
l Keeley (1990), Steven R. Weisbrod et al. (1992), and
while having a more modest, if any, effect on Rebecca S. Demsetz et al. (1996) provide empirical evi-
the total elasticity of deposits. This increase in dence from the United States and Japan.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. I HELLMANN ET AL.: MORAL HAZARD IN BANKING 151

capital requirements affect the incentives for rates offered.'5 In our judgmient, the assumption
gambling.f2 In the analysis, we abstract from of deposit insurance best reflects reality. Note,
some of the other criticism of capital require- however, that our results do not depend on this
ments, such as the fact that they fail to recog- assumption. In our working paper version (Hell-
nize all relevant risk (see Mathias Dewatripont mann et al., 1998b), we show that all of our
and Jean Tirole [1994] for a more comprehen- qualitative results continue to hold in an envi-
sive treatment).13 ronment without deposit insurance.
The remainder of this paper is organized as After funds have been raised, the bank allo-
follows: In Section I, the model is introduced. cates its assets, wherein the bank faces a moral-
We then examine competitive equilibria and hazard problem in choosing its loan portfolio.
develop conditions under which gambling oc- For simplicity, we assume that the bank may
curs in equilibrium in Section II. Section III choose between two assets: the prudent asset,
develops the instruments of prudential regu- yielding a return a; and the gambling asset,
lation and highlights the trade-offs between yielding a return of y with probability 0 and ,B
using capital requirements and deposit-rate with probability 1 - 0. The prudent asset has
controls. Section IV introduces the distinction higher expected return (a > Oy + (1 - O)f),
between binding and nonbinding deposit-rate but if the gamble succeeds the bank earns higher
controls and shows how nonbinding ceilings private return (,y > a). The bank invests both
can limit the equilibrium-path behavior by the deposits it mobilizes and its own capital k,
gambling banks to promote prudent out- which is expressed as a percentage of the de-
comes. In Section V, we consider how intro- posits mobilized so that the total assets invested
ducing endogenous rates of return creates equal (1 + k)D(ri, r-j).
feedback effects that strengthen our conclu- The opportunity cost of that capital is p.l6 In
sions from the basic model. Concluding re- this section we assume that p is exogenous and
marks follow in Section VI. that p > a (i.e., bank capital is costly). A simple
revealed-preference argument suggests that the
I. The Model case of p > a is the relevant one. If capital truly
had no opportunity cost, then the problem of
Consider a bank that operates for T periods. moral hazard in banking would not be so prev-
In each period, the bank offers an interest rate alent as it remains today, because regulators
on deposits of ri in competition with other would simply ensure that banks hold sufficient
banks which offer depositors interest rates r-j. capital to induce prudent investment, and banks
The total volume of deposits mobilized by the would willingly comply. In Section V we de-
bank is D(ri, r-j), with the volume of depositsvelop this argument further. We examine an
increasing in the bank's own interest rate and extension of the model where p is determined
decreasing in the competitors' rate (DI > endogenously
0, as the equilibrium rate of return
2 < o).4 that clears the market for bank equity capital.
Depositors have deposit insurance, so the Without capital requirements capital is not
volume of deposits depends only on the interest costly (i.e., p = a), but banks will typically hold
either too little capital or none at all. A binding

12 Note, however, that Chun H. Lam and Andrew H.

Chen (1985), Gerard Genotte and David Pyle (1991), and 15 We do not explicitly model the fees paid by banks for
David Besanko and George Katanas (1996) show that, in deposit insurance, although our results hold for any fixed fee
some circumstances, capital requirements may actually in- for insurance. Yuk-Shee Chan et al. (1992) and Ronald
crease the portfolio risk. Gianmarino et al. (1993) show how more sophisticated fee
13 Criticisms include that (i) they induce rigidities in the schemes can be used to reduce moral hazard.
adjustment process; (ii) they typically are implemented by 16 We can think of the cost of capital as the dilution cost
examining risk on an asset-by-asset basis, ignoring correla- to the owners. Equity investors know the expected returns of
tions; and (iii) they often focus on credit risk, ignoring the bank in every period (which may depend on whether the
market risks. bank will gamble or not, which outside equity investors can
14 The only additional assumption on the demand func- rationally anticipate). They will provide capital if the ex-
tion is that it satisfies concavity of the bank's value function. pected return on investment equals their opportunity cost p.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 152 THE AMERICAN ECONOMIC REVIEW MARCH 2000

capital requirement, however, increases the de- sponding to the infinitely repeated static Nash
mand for capital, endogenously making capital equilibrium.
costly (i.e., p > a) 17 The timing of the stage game works as fol
Banks are subject to prudential regulation by lows: Banks simultaneously choose their own
the government. At the end of each period, the level of capital and offer a deposit rate. Depos-
regulator inspects the balance sheet of all banks. itors then choose the bank in which to place
If a bank has negative equity (i.e., the bank their funds. Banks then choose their asset port-
cannot repay all its depositors in full), its fran- folio. Finally, returns are realized, and the reg-
chise is revoked. Thus, if a bank were to gamble ulator inspects the balance sheet of the bank.'9
and the gamble fails, then the bank will lose its
franchise and cease operation.18 In this econ- II. Competitive Equilibria with Gambling
omy with an (ex post) perfectly informed regu-
lator there are no "zombie" banks "gambling on Before comparing different forms of pruden-
their resurrection" (see Kane, 1989) that have tial regulation, it is worthwhile to determine
negative equity and are hoping that a successful whether any regulation is indeed necessary. In
gamble will return them to solvency. The reg- particular, the purpose of this section is to de-
ulator, however, cannot perfectly monitor the ex termine conditions under which banks would
ante investment portfolio of the bank. This is choose to gamble in equilibrium. The expected
consistent with the shift in regulator structure to return from the prudent asset is Vp(ri, r-j,
monitoring the risk-management system of the k) -= 7p(ri, r-j, k)I(1 - 8), while the ex-
bank rather than examining each individual fi- pected return from the gambling asset is VG(ri,
nancial transaction. r-j, k) = wG(ri, r-j, k)/(1 - 80). The
The per-period profits of the bank when it investment process by banks occurs in two stag-
chooses the prudent asset are rp(ri, r-i, k) es: the deposit mobilization stage and the asset-
mp(ri, k)D(ri, rJ), where mp(ri, k) = a(l +allocation stage. At the allocation stage, banks
k) - pk - ri is the effective profit margin that
have D(r, r_j) units of deposits to invest with
the bank earns on each unit of deposit, net of its an interest-rate cost of r. Banks will choose to
cost of capital. When the bank gambles, per-
invest in the prudent asset if Vp(r, r-j, k) >!
period profits are WrG(ri, ri, k) = MG(ri,
VG(r, r-, k), and they will invest in the gam-
k)D(ri, rJ), where mG(ri, k) 0(,y(l + bling asset otherwise. From this relationship,
k) - ri) - pk. The gambling margin depends we can develop a "no-gambling condition"
on whether the gamble is successful. If so, then which determines the threshold interest rate at
the bank captures a high return on assets and which gambling will occur, namely,
repays its depositors. If the gamble fails, then
the bank is closed down by the regulator. 7TG(r, r_, k) - irp(r, r_, k)
Banks maximize their expected discounted
profits, V = IT 0 8t t. Following Douglas W. : (1 - 0)8Vp(r, rj, k).
Diamond (1989) we will look at the limit as T->
oo. Banks will thus choose strategies corre- This constraint is intuitive. The one-period rent
(7TG - Tp) that the bank expects to earn from

17 Gary Gorton and Andrew Winton (1997) also derive

that bank capital is costly endogenously in a general- 19 We should note that our model has a static structure
equilibrium model. within each investment period. There is a growing literature
18 This assumes that the return earned by the bank when of papers (Charles W. Calomiris and Charles M. Kahn,
the gamble fails is insufficient to repay depositors. It is 1991; Mark J. Flannery, 1994) that has a richer dynamic
straightforward to show that this assumption is always sat- structure within each investment period. This matters most
isfied when a bank chooses to gamble in equilibrium be- in a world without deposit insurance, where one would want
cause the gambling asset has a lower expected return than to analyze the role of demand deposits and the potential
the prudent asset. It is only when the bank can impose a cost disciplinary role of the interim withdrawal by depositors.
on depositors (or, more specifically, the deposit insurer) that This literature generally finds that interim monitoring by
the private expected return to the bank is higher from demand depositors can mitigate but not fully eliminate
gambling. problems of moral hazard.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. ] HELLMANN ET AL: MORAL HAZARD IN BANKING 153

gambling must be less than the lost franchise rp(k) [a(1 + k) - pk]81(s + 1).
value (5Vp) that the bank gives up if the gamble
fails (with probability 1 - 0). From this we can Because aVplak- -(p - a)D(r, r)I(1 -
determine the critical deposit interest rate, r(k), 3) < 0, increasing a bank's capital only re-
such that for r c r(k) the bank will choose to duces the bank's expected profits, and so the
invest in the prudent asset (assuming a symmet- bank will choose to minimize its own capital
ric equilibrium in deposit rates): that it invests. Thus, in the competitive equilib-
rium, if the bank were to choose the prudent
asset, then rp(O) - a8/(s + 1). As competi-
r(k) = (1 - a)(- y (1 + k) tion for deposits becomes sufficiently intense
(i.e., as ? --> o), then the competitive deposit
+ 3[a(l + k) - pk]. rate approaches a, with the result that the fran-
chise value of the bank becomes arbitrarily
When the bank is farsighted (as 3 -- 1), the
small. But as discussed above, once the deposit
bank can pay out a deposit interest rate that rate exceeds a critical threshold [rp(k) > r(k)],
approaches the bank's net return on assets the bank earns greater expected returns from
(a[ 1 + k] - pk) and still choose to invest in gambling than from investing in the prudent
the prudent asset. This is sensible because as asset, and thus no equilibrium where all banks
3 -> 1 the bank only cares about average per- choose to invest in the prudent asset can exist.
period returns, so the bank would never engage If we turn our attention to the case where the
in a gambling activity that returns a finite pos- bank invests in the gambling asset, we can
itive one-period rent at the risk of losing all repeat the preceding logic to show that
future returns. Once the bank is less than per-
fectly farsighted, however, the bank must earn a mG( rG, k) = OD (rG, rG)I(aD(rG, rG)Ia ri)
sufficiently large positive profit each period so
that its franchise value at risk is greater than the implying that
expected returns from gambling.20
Given the bank's asset-allocation decision, rG,(k) = [,y(l + k) - (pkIO)]s1(s + 1).
we can turn to how banks compete in the market
for deposits. We will assume that, if a compet- As above,
itive equilibrium with no gambling exists, then
that is the equilibrium that will be selected by aVGcak =-(p - Oy)D(r, r_j)I(1 = 30) < 0
the banks. If a bank intends to invest in the
prudent asset, then it will choose so no bank would voluntarily hold capital.
The above argument is summarized in the
(rp, kp) = arg maxr,k{Vp(r, r-j, k)}. following proposition. Let ? - r(O)Io[a -
r^(O)] I.
For a symmetric equilibrium (i.e., r-i rp),
using the first-order condition (9VpIari - 0),PROPOSITION 1: For sufficiently competitive
we have markets (i.e., ? > ?), the only symmetric equi-
librium has banks choosing to hold no capital,
mp(rp, k) = D(rp, rp)I(aD(rp, rp)Iari) pay rG(O) to depositors, and invest in the gam-
bling asset.
which implicitly defines rp(k). Using ? (aDI
arj)(rID), we have III. Prudential Regulation

Once markets are sufficiently competitive

20 The remaining comparative statics on the no-gambling
that gambling must occur in the free-market
condition are immediate: aP/Iaa > 0, aPIay < 0, aPIaO <
competitive equilibrium, some form of pruden-
0. The no-gambling condition becomes less stringent as the
prudent asset is more attractive, and more stringent as the tial regulation is necessary. A consensus has
gambling asset is more attractive. emerged among both economists and policy

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 154 THE AMERICAN ECONOMIC REVIEW MARCH 2000

makers that a minimum capital requirement for the minimum effective capital requirement
combined with effective monitoring of banks by as well:
a central regulator forms an effective basis for
prudential regulation of banks. The underlying aklaa < 0
logic of capital requirements is that when the
bank invests its own capital that capital acts as akl/ay > O
a bond, so that the bank bears some of the
downside risk from investing in risky assets. akla0 > 0.
(We call this the capital-at-risk effect.) Provided
that the bank has sufficient capital at risk, the
bank will choose to invest in the prudent asset in Since financial liberalization grants greater free-
equilibrium. dom to banks in determining their lending port-
An alternative form of prudential regulation folio, we would expect that either y or 0 (or
is deposit-rate controls, creating a ceiling on the both) would increase. This raises the possibility
interest rate that banks may pay to depositors. that the minimum effective capital requirement
By limiting the degree of competition in the may increase as liberalization is introduced.22
deposit market, a deposit-rate control will in- Implementing an effective policy of deposit-
crease the per-period profits captured by each rate controls is relatively straightforward. Since
bank, thereby increasing the franchise value. As for all r ? r(k) it is optimal for the bank to
described in Section II, once the franchise value invest in the prudent asset, then any deposit-rate
at risk exceeds the one-period expected gain control of r(k) in combination with a minimum
from gambling, the bank will choose to invest in capital requirement of k will implement a no-
the prudent asset. gambling equilibrium.23 Furthermore, even if
To implement an effective policy of capital no capital requirements are used, a deposit-rate
requirements (i.e., one that eliminates gambling ceiling of r(O) will suffice. This leads naturally
as a competitive equilibrium), it must be the to Proposition 2.
case that no profitable deviation to gambling is
available to any bank. If banks are required to PROPOSITION 2: Any Pareto-efficient out-
hold k of capital, then the equilibrium interest come can be implemented by a combination of a
rate, assuming that all banks invest in the pru- minimum capital requirement of k* with a
dent asset, will be rp(k). If a bank were consid- deposit-rate control of r(k*).
ering a deviation, then it would choose its
deposit rate to maximize its return, conditional The income for the banks, depositors, and the
on all other banks paying rp(k) and the bank government (as the deposit insurer) are deter-
investing in the gambling asset. For this devia- mined as a function of the deposit rate, the level
tion not to be profitable, and thus ensuring that of capital held by banks, and whether the banks
a capital requirement of k will implement a gamble or invest prudently. The Pareto frontier
no-gambling equilibrium, it must be the case is defined by the set of outcomes such that no
that Maxr{VG(r, rp, k)} ? Vp(rp, rp, k).2' agent (bank, depositor, or the government) can
Define k as the minimum level of capital that
satisfies the constraint. We know that k exists 22 This is potentially an important observation. One of
because, for sufficiently large k, the bank bears the objectives of the Basle Accord was to create a 'level
enough of the cost of gambling that its returns playing field" for international competition. Yet if banks in
different countries face different gambling opportunities
are strictly greater when it invests in the prudent
and different degrees of competition, the optimal capital
asset. We can determine the comparative statics requirement will not be the same across countries.
23 Normally, we think of deposit-rate controls as deposit-
rate ceilings. If k* > k, where k- { rp(k) = r(k) }, then
21 For notational simplicity, when writing rp(k), rG(k), in a competitive equilibrium with a deposit-rate ceiling but
and r(k) as arguments of a function, we will suppress the no floor, banks would offer an interest rate that is less than
fact that each of these depends on the level of capital P(k*). To implement the constrained social optimum, the
regulator also would need to impose a deposit-rate floor of
deployed by the bank and write rp, rG, and P [i.e., Vp(rp, rp,
k) = Vp(rp(k), rp(k), k)]- P(k*) along with a capital requirement of k*.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 VOL. 90 NO. ] HELLMANN ET AL.: MORAL HAZARD IN BANKING 155

gambling. This static analysis has failed to in-

r
clude the dynamic effect of capital requirements
on a bank's franchise value. Since holding cap-
> s Z o r~~~~~~(k)
ital is costly, the per-period future profits of the
bank are lower, ceteris paribus, when bank cap-
ital increases. Thus, increasing the amount of
capital held by the bank has two effects: the
positive capital-at-risk effect and the negative
/ \ ~~~~~~~rP(k) franchise-value effect. This discussion is sum-
marized in the following proposition.

PROPOSITION 3: When banks are suffi-

ciently farsighted (8 ' 6 < 1), the negative
franchise-value effect dominates the positive
k
Pareto frontier capital-at-risk effect of capital requirements.
E -Z~ Gambling region Consequently, there always exists a policy of
No-gambling region deposit-rate controls that Pareto-dominates
any policy of capital requirements.
FIGURE 1. No-GAMBLING REGION
AND COMPETITIVE EQUILIBRIA
The Pareto-dominance result follows natu-
Notes: The line designated r(k) is the no-gambling condi-
rally from the condition that banks are far-
tion, and the line designated rp(k) is the equilibrium deposit
rate conditional on prudent investment; rp(k) above r(k) sighted because the no-gambling constraint is
cannot be implemented, because banks strictly prefer gam- downward sloping. For any positive capital
bling to prudent investment in this region. requirement, then, the interest rate realized is
less than that obtained by simply using a
deposit-rate control without a capital require-
be made better off without some other becom- ment [i.e., r(k) ' ri(O)]. A deposit-rate con-
ing worse off. Within the no-gambling region, trol of ri(k), combined with no capital
the first part of the Pareto frontier is defined by
requirement, would yield the same returns to
the vertical axis under ri(O) [i.e., all r ? r(O)]. depositors and higher profits to banks, which
(See Figure 1, where the no-gambling condition would save banks the incremental capital
is upward sloping [i.e., when drildk > 0].) costs ([p - a]k). Whenever banks are suffi-
When examining the no-gambling condition, ciently farsighted that the franchise-value ef-
we find, however, that it is not necessarily fect dominates the capital-at-risk effect, then
upward sloping: any policy of capital requirements is a Pareto-
inferior policy. Under these circumstances,
drJdk= (1 - 6)(a - Oy)/(l -0) the optimal capital requirement is always
zero. For the remainder of the paper, we will
- 5(p - a) assume that banks are sufficiently myopic that
the no-gambling condition is upward sloping
which implies that dr/dlk ' 0 for > ' - (a - (i.e., d r/dk > 0), where it is possible that the
Oy)/[a - Oy + (1 - 0)(p - a)]. This finding optimal capital requirement can be positive.
implies that, for farsighted banks, increasing the All of our results will also hold for the case
amount of capital held by the bank actually when the no-gambling condition is downward
increases the bank's incentives to gamble. This sloping.
runs counter to the intuition traditionally found We have shown that we can always implement
in the literature (see Bhattacharya, 1982). The the constrained optimum using both capital
traditional analysis has focused on the static role requirements and deposit-rate controls. We
of bank capital. The greater the amount of the now consider whether there are circumstances un-
bank's own equity at risk, the greater is the der which we can implement the optimum using
extent to which the bank internalizes the cost of just capital requirements and no deposit-rate

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
156 THE AMERICAN ECONOMIC REVIEW MARCH 2000

controls-the set of policies usually associated

with financial liberalization. r
r~~~~~~~~~~~~~~~ k /
With no deposit-rate controls, banks can rp(k)
freely choose their deposit rates. To under-
stand the effectiveness of using just a capital
requirement, it is important to understand a

(k:0), t(k) f < :

bank's incentives to deviate by offering a
different interest rate and switching its loan
portfolio to the gambling asset. Define k by
rp(k) = r'(k). Lemma 1, which is proved in
the Appendix, derives from the fact that, at k,
the bank has an incentive to deviate by offer-
ing a higher interest rate and switching to the ko k k k
gambling asset. Pareto frotiiier
Pareto improvemeni

LEMMA 1: The minimum effective capital re- Gambling regioni

quirement is strictly greater than k (i.e., k > k). ZJ No-gainbling region

FIGURE 2. PARETO IMPROVEMENTS

The intuition for Lemma I comes from the WITH DEPOSIT-RATE CONTROLS
difference between how the no-gambling con-
Notes: The line designated P(k) is the no-gambling condi-
dition is determined in comparison to how the
tion, and the line designated rp(k) is the equilibrium deposit
equilibrium interest rate is determined. The rate conditional on prudent investment and no deposit-rate
no-gambling condition is determined such ceiling. A deposit-rate ceiling or r~ = rp(k) yields the same
that the bank has no incentive to gamble at the return and allows the government to reduce the capital

asset-allocation stage. The bank thus consid- requirement from nk to ko while still inducing prudent in-
vestment by the banks.
ers its return from gambling and prudent in-
vesting, conditional on having a fixed pool of
deposits to invest. For the bank to be indif-
ferent between investing in the prudent versus icy would yield an identical return to depositors
the gambling asset (which is the definition of while increasing the profits of the banks. Prop-
r), then the bank's current-period expected osition 4 summarizes this result.
margin must be strictly greater from the gam-
bling asset; that is, PROPOSITION 4: There always exists a pol-
icy regime consisting of both a capital require-
mG= mp(1 - 80)/(1 - 6) > mp. ment and a deposit-rate control that Pareto-
dominates any policy regime that only uses a
When we consider the deposit-mobilization capital requirement.
stage, the bank therefore has a greater incentive
to mobilize deposits when it is gambling than This is a strong result. It states that the current
when it invests prudently, because it earns a policy regime practiced in most countries
higher expected margin on its incremental around the world (i.e., using just a capital re-
deposits. quirement with no deposit-rate control) is a
For any capital requirement greater than k, Pareto-inferior policy choice.
the equilibrium interest rate is strictly inside the As long as a positive capital requirement is
efficient frontier because rp(k) is downward necessary to induce the prudent outcome, the
sloping. This implies that there exists some capital-requirement-only regime is Pareto inef-
lower level of capital, ko, such that ri(ko) = This is particularly important for analyz-
ficient.
rp(k) (see Figure 2). Thus, as an alternative ing
to aan alternative proposal for creating
capital requirement of k, we could use a franchise value: the use of entry restrictions.
deposit-rate control of r(ko) combined with a
Lowering the number of banks competing with
each other has a qualitatively similar effect to
capital requirement of ko. This alternative pol-

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. ] HELLMANN ET AL: MORAL HAZARD IN BANKING 157

reducing the interbank elasticity ofaway

tition, like giving deposits.24
consumer goods ("toast-
This, in turn, will increase the equilibrium fran- ers"), opening new branches, and advertising. This
chise value of the bank. It does not, however, non-price competition is inefficient if it compen-
allow the implementation of a Pareto-efficient sates consumers with an inferior substitute, and it
outcome. As long as banks can freely determine can have the adverse consequence of dissipating
deposit rates, each bank has a positive incentive part of the franchise value induced by the deposit-
to raise deposit rates along the Pareto frontier. rate ceiling.25
Since entry restrictions do not affect the bank's It is not unambiguous, however, that non-price
deposit-rate-setting ability, this incentive prob- competition is socially wasteful. In particular,
lem precludes implementation of Pareto- when the financial sector is underdeveloped, non-
efficient outcomes. price competition can have the positive effect of
A related issue is that we have so far assumed deepening the financial sector. In Hellmann et al.
that the number of banks in the economy is (1996), we consider an economy where there are
exogenously fixed. Clearly, with endogenous underserved deposit markets. Banks need to make
entry, there exists the potential for the franchise some investments (such as building a branch net-
value to be eroded by competition from new work) to attract customers. In these circumstances
banks. Banks will only enter, however, when non-price competition is socially desirable, but
their sunk costs of entry are less than or equal to private banks only have the incentive to make
the ex post franchise value in the post-entry these investments if there are binding deposit-rate
equilibrium. If these sunk costs exceed the nec- controls. In these circumstances the gains from
essary franchise value to support prudent invest- financial deepening may exceed the inefficiencies
ment, endogenous entry has no effect on our induced by non-price competition.26
results. If the sunk costs are too low, our results Of course, those results only apply in a de-
can still be supported provided that the govern- veloping-country context. In an advanced in-
ment charges a license fee such that the total dustrial economy, where almost all households
sunk costs (entry costs plus license fee) equal have access to the financial sector, there are few
the franchise value. potential gains from further investments in ex-
panding the number of branches. In this context,
Finally, it is also interesting to consider the
case where no capital requirements are used and a deposit-rate ceiling is likely to induce some
socially wasteful non-price competition, like
regulators rely solely on deposit-rate controls. It
is immediate that deposit-rate controls can onlywhat the United States experienced in the
implement outcomes with k = 0 and r c r(0). 1960's and 1970's. Moreover, the proliferation
of financial products available to households
All these outcomes are Pareto efficient, but they
implement only a subset of the Pareto frontier. (stock, bond, and money-market mutual funds,
etc.) implies that households are more likely to
IV. Binding and Nonbinding
Deposit-Rate Ceilings
25 Non-price competition, however, will not fully dissi-
The results of the previous section develop the pate the rents induced by a deposit-rate ceiling. Banks will
use instruments of non-price competition until the marginal
advantages of using deposit-rate ceilings as a cost of attracting another unit of deposit equals the marginal
mechanism of prudential regulation. The experi- return. With increasing marginal costs, then, the bank will
ence of the United States in the 1960's and 1970's capture rents on all of its inframarginal deposits.
suggests that there are potential difficulties caused 26 Similarly, non-price competition may have value
when entry is endogenous. As discussed in the previous
by the use of deposit-rate ceilings that were not
section, new banks will enter the market as long as the sunk
formally analyzed in our stylized model. In par- costs of entry are less than the ex post franchise value
ticular, when banks are precluded from freely de- achieved in the post-entry game. If the sunk costs of entry
termining deposit rates, they may seek to capture increase when the banks engage in more non-price compe-
deposits through other forms of non-price compe- tition (i.e., building branches with heavy marble content),
then the equilibrium level of franchise value that can be
supported likewise increases. Thus non-price competition
can even have a positive effect of enhancing banks' com-
24 This can be derived explicitly in a Hotelling [Link] to invest in the prudent asset.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 158 THE AMERICAN ECONOMIC REVIEW MARCH 2000

substitute out of deposits in response to a bind- ir such that rp(k) < r < rD(k), then the policy
ing ceiling. will have no effect on the equilibrium deposit
A related concern arises about the informa- rate, while at the same time reducing the returns
tion requirements implied by a policy of using to gambling [because VG(i, rp, k) < VG(rD,
deposit-rate ceilings as a tool of prudential reg- rp, k)1.
ulation. In this policy regime, it is the govern- This is important, because (from Lemma 1),
ment that determines the rate of return on it is the gambling bank's excess incentive to
deposits. But a government regulator may have raise deposit rates that, in the absence of a
difficulty implementing the optimal deposit-rate deposit-rate ceiling, forces the government to
ceiling, particularly in response to significant set a capital requirement of k > k. If the
macroeconomic shocks. Clearly, this was the government uses a deposit-rate ceiling r such
case experienced by the United States in the that rp(k) < i ? ri(k) for any k < k ? k, then
1970's. Regulation Q set a nominal ceiling on all banks are constrained to offer deposit rates
rates. Following the oil shock in 1973, that that satisfy the no-gambling condition, and a
ceiling was below the rate of inflation, forcing deviation to gambling will not be profitable.28
banks to offer depositors a negative real rate of This leads to Proposition 5.
return and encouraging depositors to switch to
other financial assets. PROPOSITION 5: For k < k ? k, if the
Even under circumstances where a binding government uses a nonbinding deposit-rate
deposit-rate ceiling is not the appropriate pol- ceiling i E (rp(k), rf(k)] and a binding capital
icy, we do believe that deposit-rate ceilings can requirement of k, the equilibrium deposit rate
be an effective tool of prudential regulation. will be determined by market competition
This is because deposit-rate ceilings may be [rp(k)], and all banks will invest prudently.
useful even if they are not binding in equilib-
rium. Nonbinding deposit-rate ceilings do not From Propositions I and 4, we know that, in
affect banks that invest their assets prudently the absence of a deposit-rate ceiling, the gov-
but may constrain banks that want to pursue a ernment needs to impose a binding capital re-
risky gambling strategy.27 Nonbinding deposit- quirement and that constraint results in the
rate ceilings then have the attractive feature that selection of a Pareto-inferior outcome. The
they do not invite inefficient non-price compe- strongest argument for selecting this outcome is
tition. They also reduce the information prob- that, when the government just uses capital re-
lems of the regulator in the sense that they allow quirements, the deposit rate is determined by
for a greater margin of error. market forces, and (as discussed above) that has
To see this in our model, note that gambling value for reasons not captured by our stylized
banks always want to offer a higher deposit rate model. Proposition 5 responds directly to this
than prudent banks. When the other banks are argument. With a deposit-rate ceiling that does
investing prudently, a bank that deviates to not bind in equilibrium, deposit rates will still
gambling will choose its deposit rate according be determined by market forces, but we can
to rD(k) -arg maxr{ VG(r, rp, k) 1. Whenever relax the constraint that binds on bank capital.
gambling is potentially attractive, the bank This will then implement an outcome closer to
earns a higher margin on the gambling invest- the Pareto frontier than can be implemented
ment than on prudent investment [i.e., mG(k) > using just capital requirements.
mp(k)], so the deviation occurs by offering a Of course, the government still is required to
higher deposit rate [i.e., rD(k) > rp(k)]. Thus, set an appropriate deposit-rate ceiling. Set too
should the government set a deposit-rate ceiling low, it will bind, with all the implications dis-
cussed above, but set too high, it will not pre-

27 We use the term "nonbinding" to refer to the effect of

the deposit-rate ceiling on the equilibrium deposit rate. The
deposit-rate ceiling does bind, however, on any bank con-
sidering a deviation to gambling. That is from whence it 28 Note that, for all k < k ' k, rp(k) < P(k) < rD(k
derives it force. so the ceiling does not bind in equilibrium.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. I HELLMANN ET AL.: MORAL HAZARD IN BANKING 159

clude banks from gambling.29 One thing that is analyze the cost of bank capital and the return to
clear from the United States's experience with bank lending endogenously. The first main in-
Regulation Q is that any deposit-rate ceiling sight is that the assumption that bank capital is
should be set in real and not nominal terms, so costly (i.e., p > a) is actually a natural conse-
that the policy is not undermined by a large quence of the fact that a binding capital require-
change in inflation expectations. We suggest ment inflates the bank's demand for scarce
that a reasonable benchmark is the rate of return equity capital. The second insight is that the
on equivalent-duration Treasury bills. The feedback effect from an endogenous determina-
yields on government notes are determined by tion of these rates of return tends to exacerbate
the market, and they automatically incorporate the negative impact that capital requirements
inflation expectations. If the deposit-rate ceiling can have on the incentive to invest prudently.
were set at some fixed premium above the It is a well-known fact that multiple feedback
equivalent-duration Treasury yield, then the effects may occur at a general-equilibrium
ceiling would adjust in a timely way to changes level. The approach we will take here is to limit
in market conditions.30 ourselves to a few first-order effects that can be
There is an additional reason why it rnakes expected of many reasonable equilibrium mod-
sense to limit the premium over Treasury rates els. In particular, below we will introduce some
that banks may offer depositors: bank deposits aggregate demand and supply functions that
are government insured, and so banks are es- exhibit a limited number of intuitive properties.
sentially borrowing using the government's A detailed microeconomic specification of the
credit rating. Shoven et al. (1992) link the sharp technology and preferences that would lead to
and sudden increase in the real interest rates in these supply and demand functions is beyond
the 1980's to the impact of insolvent S&Ls the scope of this paper. Instead, we take a
offering high rates on certificates of deposits in "(reduced-form" approach that allows us to cap-
competition with Treasury securities. Essen- ture fairly broad equilibrium feedback effects
tially, when insured financial institutions devi- that do not depend on any specific underlying
ate to gambling, they can use the government's model.
own credit rating to offer high deposit rates in To model the endogeneity of p and a, we will
competition with the government. use a demand and supply model for bank equity
and bank lending, respectively. Consider first
V. Endogenous Rates of Return the market for bank equity. Suppose that savers
can either put their savings into deposits or
So far we have taken the cost of capital p and supply funds to a market for bank equity. With
the return to the lending portfolio a as given. In a slight abuse of notation, denote the aggregate
general, however, one may expect that, while supply of deposits by D and the aggregate sup-
any individual bank may take these as given, ply of funds for bank equity by E.3' A reason-
they are endogenously determined at an aggre- able assumption that would be predicted by
gate level. This may give rise to feedbac-k from many general-equilibrium models is that the
the policy interventions to the now endogenous supply of funds is (weakly) increasing in their
return variables. In this section we introduce a own returns and (weakly) decreasing in the re-
simple extension of the model that allows us to turns of substitutes. Moreover, it seems reason-
able that the own-price effect dominates the
cross-price effect. We therefore assume that
D(r, p) and E(r, p) satisfy Dr (-aDla r) ? 0,
29 Finding the appropriate range of deposit-rate ceilings
may be further complicated by bank heterogeneity. Obvi- Dp p 0, Dr + Dp ' 0, Ep ? O, Er F 0, and
ously, the same problems apply to capital requirements. Ep + Er : 0.
30 Certificates of deposit may offer higher yields than
equivalent Treasury securities. Much of this premium can
be explained by differences in the tax treatment of the two
assets: interest income from CDs are subject to state tax, 3' The aggregate functions are simply obtained by mul-
whereas Treasury-bill income is exempt (see John B. tiplying the individual supply functions by the number of
Shoven et al., 1992). banks, N.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 160 THE AMERICAN ECONOMIC REVIEW MARCH 2000

The implicit idea behind these supply func- ard in banking, and no prudential regulation
tions is that bank equity and deposits have dif- would be necessary.
ferent characteristics that savers might care Unfortunately the real world does not seem to
about. First, bank deposits provide liquidity on correspond to this scenario. We note that in
demand and other related services such as general the return to bank equity needs to com-
check-writing. Second there may be transaction pensate for the lack of liquidity and other in-
costs for investing funds in bank equity, and conveniences. This suggests that the supply of
investors may need to be more sophisticated to funds for bank equity capital may be somewhat
participate in this market. Finally, there may be scarce. In particular, when kv(rp(k)) < k, the
different levels of risk associated with these amount of capital that bank are willing to hold
securities. While analytical tractability prevents voluntarily is not enough to induce prudent be-
us from modeling these effects directly, our havior. Suppose first that there are no capital
reduced-form demand functions are consistent requirements. If p > a, then banks would not be
with these interpretations. willing to hold any capital. It follows that p oa
The bank's demand for equity depends on in equilibrium, and banks hold less capital than
capital requirements as well as on the relation- necessary to prevent gambling. While the equity
ship between p and a. It is convenient to ex- market is clearing and the return to bank assets
press the demand as a fraction k of deposits. Let equals their cost of capital, gambling occurs in
k be the required minimum level of capital. equilibrium.
Then the bank's demand for capital is given by It is precisely under these circumstances that
kD, where k = oc if p < a, k E [k, oc) if p = a regulator would want to impose some capital
a, and k = k if p > a. This simply says that requirements. But imposing a binding capital
banks would be willing to raise any amount of requirement will drive up demand and thus the
equity capital if p < a but only want to raise theprice of equity. For the minimum effective cap-
minimum required when p < a. At p = a, they ital requirement k, the equity market clears at
are indifferent as to the amount of capital raised. some p > a, such that E(p, r) - kD(r, p). We
Bank equity is demanded and supplied in a have thus shown that the high cost of capital is
standard competitive market where individual endogenously induced by regulation (i.e., it is
banks and savers take the price of equity capital the natural consequence of a binding capital
p as given. The equilibrium is determined by the requirement).
interaction of supply and demand. Clearly p < Taking total derivatives in E(p, r) = kD(r,
ae cannot be an equilibrium, since any individual p), we immediately obtain (in obvious notation)
bank would offer a higher return to equity- Pk - DI(kEP - DP) ? 0 and Pr (kDr
providers to attract additional capital. To see Er)(Ep - kDp) > 0. If the cost of capital is
whether p= a or p > a, define kv ? 0 such that endogenous, it is an increasing function of both
E(a, r) =kVD(r, a). kv(r) is the amount of k and r. Higher capital requirements inflate the
capital that banks are willing to hold voluntarily banks' demand for capital, thus requiring a
(hence the superscript v) even in the absence of higher rate of return. Higher deposit rates also
capital requirements.32 We now distinguish the inflate the banks' demand for capital, to match
case where this voluntary level of capital is the increased amount of deposits. In addition,
sufficiently large to support a prudent equilib- higher deposit rates make it harder to convince
rium. Suppose first that there is an abundant savers to invest in bank equity and thus also
supply of funds for bank equity. In particular, require increased returns to lure savers back into
consider the case where kv(rp(k)) > k. In thisthe bank equity market.
case, banks voluntarily maintain sufficiently We also extend the model to allow for an en-
large amounts of capital not to gamble. Such an dogenous determination of the return to lending,
economy would have no problem of moral haz- a. Again, a fully specified general-equilibrium
model is beyond the scope of this paper, but we
can use a reduced-form approach to capture the
32 It is straightforward to check that kV is decreasing in r. main dependency of price on quantity (i.e., the
This is also true when a is endogenous. dependency of the return to lending on the total

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. ] HELLMANN ET AL.: MORAL HAZARD IN BANKING 161

volume of bank lending).33 Denote the demand r

for bank loans by L(a). Most models of bank

lending would predict that loan demand is a P(kIP, a exogeneous)
(weakly) decreasing function of a (i.e., as banks
increase their lending rates and/or tighten other
terms [an increase in a], firms are likely to curtail
their demand for loans). The supply of loans is
r (kp, a endogeneous)
given by the total bank assets A = D + E. The
equilibrium in the lending market is then given by
L(a) = A(r, p(r, k)), where p(r, k) from above.
Taking total derivatives and using Lae < 0, Ar=
Dr+ Er ' 0, and AP =DP + Ep 0, we obtain
a,r = (Ar + Appr)ILU ?0 and ak ApPkILa ? 0.
Increases in the deposit rate or the capital require-
k
ment drive down the return to bank lending. EL=i Gambling region when returns are exogenous
We are now in a position to examine how the Additional gambling region when returns are endogenous

endogeneity of p and a affects the policy trade- L22IZ No-gambling region

offs. For this we revisit the no-gambling condi-

FIGURE 3. THE EFFECTS OF ENDOGENOUS
tion, which defines the set of feasible
RATES OF RETURN
combinations of r and k that a policy maker can
implement. The proof of Proposition 6 is given
in the Appendix. proposition. If p and a are exogenous, ri(k) is
a straight line. If, however, p and a are en-
PROPOSITION 6: Consider an increase in the dogenous, the slope of r(k) is always lower,
capital requirement in the extended model. In so that the endogenous r(k) locus lies below
addition to the usual capital-at-risk and the exogenous ri(k) line35 This implies that
franchise-value effects derived in Proposition 3, the regulator now faces a smaller set of r and
the endogeneity of the cost of capital p and the k that can be used to implement prudent bank-
return to bank lending a implies that there are ing equilibria. We also noted before that if the
additional "feedback" effects that further re- franchise-value effect dominates, ri(k) would
duce the incentive to invest prudently. have a negative slope. When p and a are
endogenous, this is even more likely, in the
The intuition for this result is as follows. We sense that the feedback effects provide an
have seen that an increase in capital require- additional downward force on the r(k) locus.
ments increases the cost of capital. But that This means that, if the feedback effects are
further reduces the franchise value, thus under- sufficiently strong, then an increase in capital
mining the incentive to invest prudently. More- requirements can never induce more prudent
over, the increase in capital requirements also banking behavior.
reduces the return to bank lending, which has a It is worth reexamininig our results from Section
negative effect both on the franchise value and II in the context of endogenous rates of return. In
the static capital-at-risk effect.34 general, our claims about the potential costs of
Figure 3 illustrates the implication of this capital requirements and the potential benefits of
deposit-rate controls in combating moral hazard
are strengthened, but our claims about Pareto ef-
3 An interesting point to note is that repeated play in the ficiency must be relaxed. Propositions 1 and 2
lending market itself may lead to "franchise-value" effects. follow immediately in the present context, and
See Serdar I. Ding (1997) for a model along those lines. Proposition 6 is the generalized version of
3 Note that we assumed here that the return to gambling,
as characterized by 0, ,B, and y, remains constant. This
seems the most natural assumption, given that in equilib-
rium no gambling investments are ever made. It is also 35 The exogenous r(k) line is drawn for the value of p
straightforward to relax this assumption. and a at r(O).

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
 162 THE AMERICAN ECONOMIC REVIEW MARCH 2000

Proposition 3. The main point of Proposition 5 is VL. Conclusion

that a nonbinding deposit-rate control allows the
government to reduce the minimum capital re- The aim of this paper has been to understand
quirement while still having market-determined the interaction between financial liberalization
deposit rates. With endogenous rates of return, and prudential regulation. Financial liberaliza-
relaxing the capital-requirement constraint has an tion tends to increase the intensity of competi-
additional benefit because the cost of bank equity tion between banks at the same time that banks
p will also decline. We must relax our claims, are given greater freedom to allocate assets and
however, from Proposition 4 because we can no to determine interest rates. As a consequence,
longer apply a formal Pareto analysis in this ex- the potential scope for gambling by banks also
tended model. This is not surprising: we have increases. We consider two potential instru-
introduced some additional constituencies into the ments of prudential regulation: capital require-
model, and the more distinct agents there are, the ments and deposit-rate controls. We first point
more likely any policy will adversely affect at out that, in a dynamic economy, capital require-
least one type of agent.6 For instance, a reduction ments may not always be as powerful as previ-
of capital requirements, as in Figure 2, can have a ously thought. This is because, in addition to a
negative impact on bank lending, so that borrow- one-period capital-at-risk effect that reduces the
ers may be worse off. Obviously the reduction incentive to gamble, there is a future-franchise-
also creates cost savings for the banks (since, in value effect that increases the incentive to gam-
equilibrium, p > a).7 In order to assess the full ble. We then show that, while a sufficiently
impact of any policy intervention, one may thus large capital requirement can generate an equi-
have to consider more complex welfare trade-offs librium in which banks choose to invest effi-
among all the constituencies.38 Our basic result, ciently, the equilibrium is Pareto inefficient. In
however, that deposit-rate controls give the regu- particular, one can always find a regulatory re-
lators an additional instrument that may be used to gime that uses both deposit-rate controls and
complement capital requirements, is robust in this capital requirements that Pareto-dominates
extended model. any regulatory regime that uses just capital
requirements.
Three key intuitions are developed through
our analysis of this model: freely determined
deposit rates undermine the franchise value of
36 Technically, another reason why the Pareto criterionbanks; capital requirements are costly because
cannot be applied in this extended model is that the reduced- they force banks to hold expensive capital; and
form supply and demand functions were specified without the use of either binding or nonbinding deposit-
specific references to utility functions.
rate ceilings to complement capital require-
37 It is possible that another policy may yield a Pareto
improvement for at least the set of depositors, bank owners,
ments creates additional policy flexibility that
and bank borrowers. Consider a policy (r, k) with r > rp(k) allows the government to relax a binding con-
and k < k such that bank franchise value is the same under straint on capital, reducing the total costs im-
both policies [i.e., Vp(r, k) = Vp(rp, k)]. Clearly, banks are
posed by the capital requirement.
indifferent, and depositors are better off. If the total banking
While our analysis has focused on the use of
assets are greater under the alternative policy [i.e., D(r,
r)(1 + k) > D(rp, rp)(1 + k)], then bank borrowers are deposit-rate controls as an additional tool of
better off as well. While this condition is not satisfied in prudential regulation to prevent moral hazard, it
general (e.g., when deposit savings are inelastic), it will is worth recognizing that there are other policy
hold provided that the total elasticity of savings is suffi-
instruments that could improve on capital re-
ciently large or the supply of funds for bank equity is
sufficiently inelastic. quirements. For example, regulations such as
38 Another constituency that we have not considered so asset-class restrictions and exposure rules may
far is the insurer of deposits. In the model, banks are help combat the moral-hazard problem, partic-
homogenous, and so the regulator may find a policy such
ularly if they limit banks' ability to invest in
that in equilibrium no gambling occurs at all. With hetero-
geneity and noise in the economy, some banks may fail.
assets that facilitate gambling. Risk-based de-
Capital requirements may have another useful role in pro- posit-insurance premiums could also lessen the
tecting the insurer of deposits. incentives of banks to gamble, although effec-

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. ] HELLMANN ET AL.: MORAL HAZARD IN BANKING 163

tively implementing such a policy and appro- gambling, however, is defined when profits are
priately defining risk classes presents a equal regardless of asset choice. Because aVpl
challenge for regulators. ar < 0, it must be the case that the bank pays
Another policy recommendation that could a strictly lower deposit rate (i.e., rp(k) < r(k)).
address the moral-hazard problem is the use For k k, we know that mG(rp, k) ' Omp(rp,
of "speed limits" on growth. As shown in our k), whereas mG(r, k) = mp(r, k)(1 - 60)/
analysis, gambling strategies are based on (1 - 6). Because the gambling profit margin
rapid growth by banks to take advantage of falls less rapidly in r than does the prudent
the artificially high current return on the gam- profit margin (i.e., 0 > am53ar > amplar),
bling asset. A policy that set limits on bank then rp(k) < r(k). Because rp(k) -( r(k) and
growth would lower the return from the gam- arplak < 0, we must have k > k.
bling strategy, reducing bank incentives to
make this deviation. This policy would have PROOF OF PROPOSITION 6:
trade-offs, however, as banks with either bet- Consider the maximal deposit rate r consis-
ter investment opportunities or lower costs of tent with the no-gambling equilibrium. Let
intermediation would be limited in their rate
of growth as well. r (1 - 8)(1 Jr k)[a(rP, k) - Oy]/(l - 0)

APPENDIX + 8[a(Q, k)(1 + k) - p(r, k)k]

PROOF OF LEMMA 1: --f(k, a-(r^, k), o(r^, k)).

First we show that for a sufficiently low cap-
ital requirement (k < k, where k~- tmG(rp, This defines an implicit function for r(k). To-
k) = Omp(rp, k)}), argmax,(VG(r, rp, k)} > tally differentiating, we obtain
rp(k), and also maxr{VG(r, rp, k)I > VG(rp,
rp, k). For this we need to show that a VG/ drldk

Ari r = r,(k) > 0

(fk + faak + fppk)(l faOr fpPr)
a VGb3ri= {mG(rp, k)[aD(rp, rp)Iari]
The sum of the capital-at-risk and franchise-
- OD(rp, rp)}I(l -80). value effects (derived for Proposition 3) are
given byfk = (1 - S)(a - Oy)/(l - 0) -
From the first-order condition that defines 5(p- a). Usingft =(1 - 8)(1 + k)I(1 -
rp(k), we have D(rp, rp) -mp(rp, k)[VD(rp,
0) > 0 andfp- -8k 0, we immediately
rp)Ia ri], which implies that obtain drldk ' fk when p and a( are endog-
enously determined.
a VGHari= {mG(rp, k)[JD(rp, rp)lari]
REFERENCES
- Omp(rp, k)}I(1 -0) > 0
Akerlof, George A. and Romer, Paul M. "Loot-
whenever k ? k. ing: The Economic Underworld of Bank-
From the definition of k, we have ruptcy for Profit." Brookings Papers on
maxrt VG(r, rp, k) I - Vp(rp, rp, k).Economic FromActivity, September 1993, (2), pp.
above, provided that k < k, maxr{ VG(r, rp, 1-60.
k) I > VG(rp, rp, k), which implies that Vp(rp, Besanko, David and Kanatas, George. "The Reg-
rp, k) I > VG(rp, rp, k). From the definition of ulation of Bank Capital: Do Capital Stan-
r(k), Vp(r, r, k)} VG(r, r, k). Thus, in the dards Promote Bank Safety?" Journal of
competitive equilibrium using just capital re- Financial Intermediation, April 1996, 5(4),
quirements, the bank must earn strictly greater pp. 160-83.
profits from choosing the prudent asset versus Bhattacharya, Sudipto. "Aspects of Monetary
gambling. The maximum deposit rate with no and Banking Theory and Moral Hazard."

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
164 THE AMERICAN ECONOMIC REVIEW MARCH 2000

Journal of Finance, May 1982, 37(2), pp. D*in, Serdar I. "Bank Reputation, Bank Com-
371-84. mitment, and the Effects of Competition in
Calomiris, Charles W. and Kahn, Charles M. Credit Markets." Mimeo, Stanford Univer-
"The Role of Demandable Debt in Structur- sity, 1997.
ing Optimal Banking Arrangements." Amer- Fischer, Klaus P. and Che'nard, Martin. "Financial
ican Economic Review, June 1991, 81(3), pp. Liberalization Causes Banking Systems Fragil-
497-513. ity." Centre de Recherche en Economie et Fi-
Caprio, Gerard, Jr. and Klingebiel, Daniela. nance Appliquees, Universit6 Laval (Canada),
"Bank Insolvency: Bad Luck, Bad Policy, or Working Paper No. 97-12, June 1997.
Bad Banking?" in Michael Bruno and Boris Flannery, Mark J. "Debt Maturity and the Dead-
Pleskovic, eds., Annual World Bank Confer- weight Cost of Leverage: Optimally Financ-
ence on Development Economics, 1996. ing Banking Firms." American Economic
Washington, DC: World Bank, 1997, pp. 79- Review, March 1994, 84(1), pp. 320-31.
104. Genotte, Gerard and Pyle, David. "Capital Con-
Caprio, Gerard, Jr. and Summers, Lawrence H. trols and Bank Risk." Journal of Banking and
"Finance and Its Reform: Beyond Laissez- Finance, September 1991, 15(5), pp. 805-24.
Faire," in D. Papadimitriou, ed., Stability of Gianmarino, Ronald M.; Lewis, Tracy R. and
the financial system. New York: MacMillan, Sappington, David E. M. "An Incentive Ap-
1996, pp. 400-21. proach to Banking Regulation." Journal of
Chan, Yuk-Shee; Greenbaum, Stuart I. and Tha- Finance, September 1993, 48(4), pp. 1523-
kor, Anjan V. "Is Fairly Priced Deposit Insur- 42.
ance Possible?" Journal of Finance, March Gorton, Gary and Winton, Andrew. "Bank Capital
1992, 47(1), pp. 227-45. Regulation in General Equilibrium." National
Cole, Rebel A.; McKenzie, Joseph A. and White, Bureau of Economic Research (Cambridge,
Lawrence J. "Deregulation Gone Awry: MA) Working Paper No. 5244, 1997.
Moral Hazard in the Savings and Loan Indus- Hellmann, Thomas F. and Murdock, Kevin C.
try," in A. Cottrell, M. Lawlor, and J. Woo, "Financial Sector Development Policy: The
eds., The causes and consequences of depos- Importance of Reputational Capital and Gov-
itory institutions failures. Boston, MA: Klu- ernance," in R. Sabot and I. Skekely, eds.,
wer, 1995, pp. 29-73. Development strategy and management of
Demirgiu~-Kunt, Asli and Detragiache, Enrica. the market economy, Vol. 2. Oxford: Claren-
"The Determinants of Banking Crises: Evi- don, 1997, pp. 269-323.
dence from Developed and Developing Hellmann, Thomas F.; Murdock, Kevin C. and
Countries." Mimeo, World Bank, 1997. Stiglitz, Joseph E. "Deposit Mobilization
_____. "Financial Liberalization and Financial Through Financial Restraint," in N. Hermes
Fragility," in Boris Pleskovic and Joseph E. and R. Lensink, eds., Financial development
Stiglitz, eds., Annual World Bank Conference and economic growth: Theory and experi-
on Development Economics, 1997. Washing- ences from developing economies. London:
ton, DC: World Bank, 1998, pp. 303-31. Routledge, 1996, pp. 219-46.
Demsetz, Rebecca S.; Saidenberg, Marc R. and "Financial Restraint: Toward a New
Strahan, Philip E. "Banks with Something to Paradigm," in M. Aoki, M. Okuno-Fujiwara,
Lose: The Disciplinary Role of Franchise and H. Kim, eds., The role of government in
Value." Economic Policy Review (Federal East Asian economic development: Compar-
Reserve Bank of New York), October 1996, ative institutional analysis. Oxford: Claren-
2(2), pp. 1-14. don, 1997, pp. 163-207.
Dewatripont, Mathias and Tirole, Jean. The pru- ______ "Financial Restraint and the Market
dential regulation of banks. Cambridge, MA: Enhancing View," in Y. Hayami and M.
MIT Press, 1994. Aoki, eds., The institutional foundations of
Diamond, Douglas W. "Reputation Acquisition East Asian economic development. New
in Debt Markets." Journal of Political Econ- York: MacMillan, 1998a, pp. 255-79.
omy, August 1989, 97(4), pp. 828-62. " "Liberalization, Moral Hazard in

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]
VOL. 90 NO. 1 HELLMANN ET AL.: MORAL HAZARD IN BANKING 165

Banking, and Prudential Regulation: Are Journal of Economic Perspectives, Winter

Capital Requirements Enough?" Graduate 1992, 6(1), pp. 155-67.
School of Business, Stanford University, Smith, Bruce D. "Private Information, Deposit
Working Paper No. 1466R, 1998b. Interest Rates, and the 'Stability' of the
Kane, Edward. The S&L insurance crisis: How Banking System." Journal of Monetary Eco-
did it happen? Washington, DC: Urban Insti- nomics, June 1984, 14(3), pp. 293-317.
tute Press, 1989. Stiglitz, Joseph E. "Credit Markets and the Con-
Keeley, Michael C. "Deposit Insurance, Risk, trol of Capital." Journal of Money, Credit
and Market Power in Banking." American and Banking, May 1985, 17(2), pp. 133-52.
Economic Review, December 1990, 80(5), . "Introduction-S&L Bail-Out," in
pp. 1183-200. J. R. Barth and R. D. Brumbaugh, eds., The
Lam, Chun H. and Chen, Andrew H. "Joint Ef- reform of the Federal Deposit Insurance:
fects of Interest Rate Deregulation and Cap- Disciplining the government and protecting
ital Requirements on Optimal Bank Portfolio the taxpayers. New York: Harper Collins,
Adjustments." Journal of Finance, June 1992, pp. 20-35.
1985, 45(2), pp. 563-75. _. "The Role of the State in Financial
Radelet, Steven and Sachs, Jeffrey D. "The East Markets," in Michael Bruno and Boris Ples-
Asian Financial Crisis: Diagnosis, Remedies, kovic, eds., Proceedings of the World Bank
Prospects." Brookings Papers on Economic Annual Conference on Development Eco-
Activity, March 1998, 1, pp. 1-74. nomics, 1993. Washington, DC: World Bank,
Rochet, Jean-Charles. "Capital Requirements 1994, pp. 19-52.
and the Behaviour of Commercial Banks." Suzuki, Yoshio. The Japanese financial system.
European Economic Review, June 1992, Oxford: Clarendon, 1987.
36(5), pp. 1137-70. Tobin, James. "Asian Financial Crisis." Japan
Rothschild, Michael and Stiglitz, Joseph E. "Equi- and the World Economy, July 1998, 10(3),
librium in Competitive Insurance Markets: pp. 351-53.
An Essay on the Economics of Imperfect Weisbrod, Steven R.; Lee, Howard and Rojas-
Information." Quarterly Journal of Econom- Suarez, Liliana. "Bank Risk and the Declining
ics, November 1976, 90(4), pp. 619-28. Franchise Value of the Banking Systems in
Shoven, John B.; Smart, Scott B. and Waldfogel, the United States and Japan." International
Joel. "Real Interest Rates and the Savings and Monetary Fund (Washington, DC) Working
Loan Crisis: The Moral Hazard Premium." Paper No. 92-45, June 1992.

This content downloaded from

[Link] on Tue, 27 Feb 2024 [Link] +00:00
All use subject to [Link]