ECONOMIC REVIEW

Theories of Bank Loan Commitments 2by 0. Emre Ergungor

A Simple Model of Money and Bankingby David Andolfatto and Ed Nosal 20

Theories of Bank Loan Commitmentsby O. Emre Ergungor

A loan commitment is an agreement by which a bank promises to lend toa customer at prespecified terms while retaining the right to renege on its promise if the borrower’s creditworthiness deteriorates. The contractalso specifies the various fees that must be paid over the life of thecommitment. Loan commitments are widely used in the economy. Parallelto their widespread use, a rich literature has evolved to explain why theyexist, how they are priced, and how they affect the risk of the bank andthe deposit insurer. This article summarizes what we have learned onthese issues. Its main insight is that loan commitments are an optimaltool for risk sharing and for resolving informational problems. The author also points out some issues that the current literature leavesunexplained.

A Simple Model of Money and Bankingby David Andolfatto and Ed Nosal

This article presents a simple environment that has banks creating andlending out money. The authors define money to be any object thatcirculates widely as a means of payment and a bank to be an agency thatsimultaneously issues money and monitors investments. While theirframework allows private nonbank liabilities to serve as the economy’smedium of exchange, they demonstrate that the cost-minimizing structurehas a bank creating liquid funds. In practice, the vast bulk of the moneysupply consists of private debt instruments issued by banks. Thus, theirmodel goes some way in addressing the questions of why private moneytakes the form it does, and why private money is typically supplied by banks.

Economic Review is published quarterlyby the Research Department of theFederal Reserve Bank of Cleveland. To receive copies or to be placed on themailing list, e-mail your request to:4d.subscriptions@clev.frb.orgor fax it to 216/579/3050.

Economic Review is available electronically through the ClevelandFed’s site on the World Wide Web: http://www.clev.frb.org.

Research Director: Mark SnidermanEditor: Paul GommeManaging Editor: Monica Crabtree-

ReusserArticle Editors: Michele Lachman,

Deborah ZorskaDesign: Michael GalkaTypography: MAC Services

Opinions stated in Economic Review arethose of the authors and not necessarilythose of the Federal Reserve Bank ofCleveland or of the Board of Governorsof the Federal Reserve System.

Material may be reprinted provided that the source is credited. Please sendcopies of reprinted material to the managing editor.

We welcome your comments, questions,and suggestions. E-mail us ateditor@cleve.frb.org.

ISSN 0013-0281

1

E C O N O M I C R E V I E W

2001 Quarter 3Vol. 37, No.3

2

20

http://clevelandfed.org/research/review/

2

Theories of Bank LoanCommitmentsby O. Emre Ergungor

O. Emre Ergungor is an economist at the Federal Reserve Bank of Cleveland.This article was originally an essay in his doctoral dissertation. He is grateful toProfessor Anjan Thakor and AssistantProfessor Sugato Bhattacharyya of theUniversity of Michigan for their guidanceand thanks Joseph Haubrich, JamesThomson, and an anonymous referee for many valuable suggestions.

Introduction

Bank loan commitments—contractual promisesto lend to a specific borrower up to a certainamount at prespecified terms—are widely usedin the economy. A recent Federal Reserve sur-vey shows that 79 percent of all commercialand industrial lending is made under commit-ment contracts.1 Moreover, as of March 2001,outstanding (unused) loan commitments ofU.S. corporations exceeded $1.6 trillion, upfrom $743 billion in 1990.2

As the use of loan commitments has grown,so has the literature on them. This article seeksto summarize what we have learned after yearsof research and to determine whether we havea reasonable idea of what value loan commit-ments provide to borrowers and lenders.

Two features of loan commitment contracts—various fees, which must be paid over the lifeof the commitment, and the material adversechange (MAC) clause—turn out to be particu-larly important in theoretical models.3

The fee structure may include a commitmentfee, which is an up-front fee paid when thecommitment is made, an annual (service) fee,which is paid on the borrowed amount, and ausage fee, which is levied on the available

unused credit. A loan commitment contract sel-dom includes all three kinds of fees together.Booth and Chua (1995) study a sample of1,347 loans and find that only 46 percent had a commitment fee, 38 percent had an annualfee, and 69 percent had a usage fee. A loancommitment without a fee structure is rarebut possible.

The second important feature, the MACclause, grants the bank some measure of dis-cretion over whether to honor the contract.A typical MAC clause reads: “Prior to [loan]closing, there shall not have occurred, in theopinion of the Bank, any material adversechange in the Borrower’s financial condition

■ 1 Board of Governors of the Federal Reserve System, “Survey ofTerms of Business Lending,” Federal Reserve Board Statistical Releases, E.2 (June 2000) <http://www.federalreserve.gov/releases/E2/200008/e2.pdf>.

■ 2 Federal Deposit Insurance Corporation, Statistics on Banking,table RC-6 (August 2001)<http://www.fdic.gov/bank/statistical/statistics/0103/cbrc06.html>.

■ 3 Because banks do not disclose the features, such as the feestructure, of the commitments they are selling, the loan commitmentliterature contains few empirical papers. Consequently, this article focuseson theoretical models.

3

from that reflected in its annual report for itsfiscal year ending December 31, ____, or in theBorrower’s business operations or prospects.”Note that a bank may repudiate the contractbased solely on its own opinion about a bor-rower’s financial condition. That is, the clauseallows the bank to use its private informationabout the borrower, which outsiders may beunable to verify.

To show what a typical loan commitmentlooks like, table 1 provides a sample of 15 loancommitments. To give a more informativepicture, Shockley and Thakor (1997) study asample of 2,513 variable-rate loan commitments,offering funds at a fixed markup over a market

interest rate, purchased by very large firms.Table 2 presents their summary statistics.Shockley and Thakor find that the marketinterest rate used is usually prime or LIBOR (theLondon interbank offered rate). The borrowingfirm is offered takedown alternatives of prede-termined markups over several different indexes,such as Treasury, federal funds, CD, and A1/P1commercial paper rates. These authors alsoobserve that although few commitments carryall three fees, many combine a usage fee witheither a commitment fee or an annual fee.

I organize this review around three mainquestions. Because these are very general,however, the literature has divided them into

Sample of 15 Loan Commitment Contracts

T A B L E 1

Fees (basis points)

Credit limit AnnualCommitment buyer (millions of dollars) Stated use Commitment servicing Usage Take-down alternatives

Turner Broadcasting 200 Commercial paper backup 0 0 62.5 Prime + 75, LIBOR +175, CD + 187.5

Levi Strauss 500 Debt repayment/consolidation 12.5 0 0 Prime, LIBOR + 100, CD + 112.5

Safeway Stores 480 Debt repayment/consolidation 0 0 0 Prime

Seagull Energy 60 Debt repayment/consolidation 0 12.5 17.5 Prime, LIBOR + 87.5, CD + 87.5

Blockbuster Entertainment 200 General corporate purposes 0 12.5 12.5 Prime, LIBOR + 50, CD + 62.5

J.C. Penney 750 General corporate purposes 0 0 18.75 Prime, LIBOR + 37.5

AT&T 6,000 Takeover 79.17 13 0 Prime, LIBOR + 37.5, CD + 50

Union Pacific 550 General corporate purposes 0 0 15 Prime, LIBOR + 25, CD + 37.5

UAL Corporation 1,300 Leveraged buyout 157.64 0.69 50 Prime + 100, LIBOR + 200

John Fluke Manufacturing 37.5 Stock buyback 0 0 0 Prime, LIBOR + 50, CD + 50

Universal Corporation 150 Working capital 0 14.17 0 Prime, LIBOR + 37.5

Dunkin’ Donuts 35 Working capital 28.57 0 37.5 Prime, LIBOR + 100

L.A. Gear 150 Working capital 0 0 50 Prime + 100

R.H. Macy & Co. 600 Working capital 150 4.84 50 Prime + 150, LIBOR + 250

American Oil and Gas 20 Working capital 0 0 50 Prime, LIBOR + 300

SOURCE: Greenbaum and Thakor (1995).

4

smaller, related questions:1) Why do loan commitments exist and howare they priced?

a) Why do borrowers demand them?b) Why do banks offer them?c) Why are loan commitments sold by

banks and not by individuals or otherfinancial intermediaries?

d) Are loan commitments put options?e) Why are loan commitments not exer-

cised up to the credit limit?2) How do loan commitments affect the

bank’s risk exposure?a) How should the bank’s risk exposure

be managed?b) Do loan commitments affect the

bank’s risk exposure?c) Should loan commitments be

regulated?3) How do loan commitments affect the

interest rate and rationing channels of monetary policy?

On the first main question, presented in thisarticle’s section I, the essence of what we knowis that loan commitments are a contractualmechanism for optimal risk sharing when bor-rowers are risk averse and future interest ratesare random. Even under universal risk neutral-ity, loan commitments may still be used toattenuate moral hazard or resolve precontractinformational asymmetry. On the valuationquestion, the principal insight is that loan com-mitments can be priced as put options wherethe borrower’s debt is the underlying deliver-able. The main findings are summarized below:

• Borrowers demand loan commitmentsbecause– Loan commitments prevent banks from

exploiting borrowers and extractingrents by threatening to withhold credit;

– Loan commitments can prevent marketfailure by attenuating moral hazardand resolving precontract informationalasymmetry.

Summary Statistics of a Sample of 2,513 Loan Commitments

T A B L E 2

Mean, (Standared deviation), [Minimum–Maximum]

Interest rate mark-up(basis points) Fees (basis points)

N Size DurationStated use (percent of total) (millions of dollars) (months) Prime + LIBOR + Upfront Annual Usage

Commercial 42 557.5 39 45.8 47.8 3.1 6.2 11.4paper backup (1.7) (800.5) (16.3) (26.0) (37.8) (8.7) (7.9) (17.9)

[30–4,300] [11–84] [25–75] [12.5–175] [0–50] [0–25] [0–62.5]

Liquidity 857 56.9 28.4 115.6 135.4 24.2 6.1 22.8(34.1) (148.7) (22.3) (73.2) (81.4) (52.0) (18.7) (25.5)

[0.1–2,000] [1–126] [–75–500] [9–350] [0–366] [0–200] [0–400]

Capital 470 142.6 39.7 115.6 148.4 28.6 3.6 27.8structure (18.7) (352.0) (26.8) (64.4) (82.8) (56.7) (10.6) (21.3)

[.2–5,500] [3–121] [–50–450] [15–425] [0–550] [0–100] [0–125]

General 931 179.1 38 105.6 90.6 18.6 4.5 19.6corporate (37.0) (449.2) (27.7) (72.9) (77.2) (49.3) (11.0) (19.8)purposes [.1–6,000] [1–198] [–50–500] [15–425] [0–550] [0–135] [0–100]

Takeover 65 74.6 36.2 111.3 125.1 13.8 3.2 29(2.6) (136.6) (25.2) (82.3) (80.3) (26.5) (8.5) (20.1)

[0.3–845] [3–120] [12.5–450] [12.5–325] [0–100] [0–40] [0–50]

Leveraged 137 139.3 65.2 149 244.6 89.8 4.2 40.3buyout (5.5) (288.4) (26.4) (32.6) (43.5) (88.3) (9.6) (19.0)

[1.5–1,848] [11–122] [75–400] [80–475] [0–302] [0–54] [0–62.5]

Debtor-in- 11 120 14.2 188.6 293.7 112.8 16.7 43.18possession (0.4) (100.9) (8.6) (30.3) (31.5) (106.8) (44.5) (16.2)

[3.3–250] [1–30] [150–250] [250–325] [0–235] [0–150] [0–50]

SOURCE: Shockley and Thakor (1997), table 1.

5

• Banks sell loan commitments because– Loan commitments facilitate forecasting

future loan demand;– By honoring discretionary loan commit-

ments, banks may enhance their repu-tation for keeping their promises andcharge higher fees for future promises;

– Lenders may use the fee structure ofloan commitments as a screeningmechanism for distinguishing amongborrowers with a priori unobservablecharacteristics.

• Loan commitments are sold by banksalone because– It is more costly for an organization not

to honor its contractual commitments thanit is for an individual;

– Reserves that the bank keeps to fundunexpected demand deposit withdrawalscan also be used to fund unexpected loancommitment takedowns. Therefore,deposit-taking institutions have a costadvantage over other financial inter-mediaries in issuing loan commitments.

• Loan commitments can be priced as putoptions where the underlying deliverable isthe debt instrument of the commitmentbuyer.

• Borrowers limit their loan takedown becausebanks penalize borrowers that fully exploittheir put options with higher future fees.The second main question, presented in

section II, asks about the effect of loan com-mitments on the bank’s risk exposure. Inselling fixed-interest-rate loan commitments,banks assume the risks associated with threeuncertain quantities: the future level of interestrates, the borrower’s uncertain credit needs,and the borrower’s future creditworthiness.The issues are how a bank can manage theserisks and whether loan commitments shouldbe regulated to protect the deposit insurer. Themain conclusion is that banks have the toolsthey need (for example, the MAC clause) toprotect themselves against the risks involved inselling loan commitments. There is little theo-retical or empirical support to justify regula-tion. The important findings are:• The bank cannot fully hedge against interest

rate and takedown-quantity risks throughfinancial futures contracts.

• Loan commitments reduce the bank’s riskexposure by inducing it to manage its creditportfolio better.

• Capital requirements, imposed on loan com-mitments by regulators to protect the depositinsurer, are not needed because loan com-mitments with a MAC clause do not imposeany additional credit risk on the bank.The third main question, presented in

section III, deals with loan commitments’effects on the transmission of monetary policy.Monetary policy is conducted through quantityrationing and interest rate channels by alteringthe quantity of credit and its price, the interestrate. Loan commitments help attenuate rationingby providing a guarantied source of funds,and thus reduce monetary policy’s ability toaffect bank lending. The main finding is:• Loan commitments introduce significant lags

in the effect of monetary policy.While the current literature improves our

understanding of loan commitments consider-ably, some stylized facts remain unexplained.First, courts limit banks’ use of discretionarypowers, often ruling that a bank’s use of theMAC clause is an abuse of power and lack ofgood faith. (See Goldberg [1988], Mannino[1994], and Budnitz and Chaitman [1998]). Ifthe MAC clause is so difficult and costly toexercise, then why do banks continue to incor-porate it into contracts? Second, moral hazardin spot lending, which can be resolved by loancommitments, can also be resolved throughrelationship (repeated) lending (Sharpe [1990],Rajan [1992], Petersen and Rajan [1995], andBoot [2000]); why, then, do we have loan com-mitments? I discuss these and other unresolvedissues briefly in the final section of this article.

I. The Purpose andPricing of LoanCommitments

I will investigate the existence and pricing literature in four subsections. First, I will dis-cuss why borrowers demand loan commit-ments (demand-side explanations). Then, Iwill explain why banks sell loan commitments(supply-side explanations). Next, I will focuson the question of why banks alone sell loancommitments. Finally, I will recapitulate whatwe know about the similarities between loancommitments and put options.

6

Demand-Side Explanations

The literature has suggested five benefits thatloan commitments offer purchasers.4

Loan CommitmentsImprove Risk Sharingbetween the Bank andthe Borrower

When a bank sells a fixed-rate loan commit-ment, it accepts the interest rate and quantityrisk that the borrower would bear if he were toborrow in the spot market. Borrowers who aremore risk-averse than the bank are willing topay the bank a premium for taking the interestrate risk on their behalf. In Campbell (1978),the premium is the usage fee. In Thakor andUdell (1987), it is the commitment fee. With afixed-rate commitment, the bank bears the riskof changes in the index rate as well as ofchanges in the borrower’s credit risk premium.With a variable-rate commitment, the bankbears only the latter risk. I will further investi-gate this issue in section II, where I discuss thebank’s risk exposure.5

Loan CommitmentsHelp Attenuate Moral Hazard

With risky debt and limited liability, the higherthe loan interest rate, the lower the borrower’snet return from a project and the greater hisincentive to switch to a riskier project (Boot,Greenbaum, and Thakor [1993]) or to under-supply effort (Boot, Thakor, and Udell [1987,1991]). To illustrate this concept, consider thefollowing example.

There are two periods and three points intime {0,1,2}. At time 0, the borrower knows thathe needs funds next period (t =1) to invest inone of two mutually exclusive projects {h, l }.Each project requires a $1 investment, which isassumed to be financed by a bank loan. Theprojects have the following characteristics: Ifthe project is successful with probability µi,it generates a cash flow Xi, i�{h,l } and zerootherwise. It is also assumed that Xh > Xl andµh <µl . Hence, l is a low-risk project and h isa high-risk project. It is further assumed thatXl µl >Xh µh. That is, the low-risk project issocially optimal. At time 0, the market interestrate at time 1 is random. It can be shown thatwhen the market interest rate at time 1 is

greater than (Xl µl – Xh µh) (µl – µh)–1, the

borrower prefers the risky project as a con-sequence of limited liability.6

Boot, Thakor, and Udell (1987, 1991) andBoot, Greenbaum, and Thakor (1993) proposethe following solution to this problem. Attime 0, the bank sells the borrower a loancommitment with a fixed interest rate of

R = (Xl µl –Xh µh) (µl – µh)–1.

If the market interest rate is less than R, theborrower is free to use the market. Otherwise,he exercises his option and takes down theloan from the commitment contract. Hence,the loan commitment guarantees that the bor-rower always chooses the safe project. Notethat the bank suffers a loss when the borrowerexercises the option. To break even, the bankcharges a commitment fee at time 0 equal tothe expected loss to the borrower at time 1.Also note that because the commitment feebecomes a sunk cost at time 1, when the bor-rower makes the investment decision, it doesnot affect the borrower’s incentives.

Boot, Thakor, and Udell also show that loancommitments are more effective than equityinvestment in attenuating moral hazard. Theintuition is as follows: When a borrowerinvests in equity, he reduces his interest bur-den for all realizations of future interest rates.This is clearly inefficient because low interestrates are not distortionary, yet the equity stillreduces the payment burden in those states.The effect of a loan commitment, on the otherhand, is selective across interest rates. Whenmarket rates are low, the borrower can stillbenefit from them. The loan commitmentreduces the interest burden only when marketrates are high. Therefore, the commitment feerequired to mitigate moral hazard is less thanthe equity investment needed to create thesame effect.

■ 4 Some of these benefits arise from the possibility of solvinginformation problems by using the multiple-fee structure. Clearly, thesepapers could be reviewed as part of the pricing literature discussed at theend of section I, but I prefer to group all the demand-side explanations in a single section.

■ 5 Also see Hawkins (1982) and James (1982).

■ 6 (Xl �l –Xh �h) (�l – �h)–1 is the rate at which the borrower’sexpected profit from the safe project equals its expected profit from therisky project if the lender believes that the borrower will invest in the safeproject and prices the loan accordingly. In other words, if the spot rate isgreater than this critical value, the lender must believe that the borrowerwill invest in the risky project if it borrows from the spot market.

7

Boot, Thakor, and Udell’s model explainsthe commitment fee and the interest rate guar-anty of loan commitments. Other importantaspects of the contract, such as the multiple feestructure or the MAC clause, are assumed awayby simplifying modeling choices, which aresummarized below.

One-period simple projects: A sure invest-ment is made at time t and the outcome isrealized at time t +1. After the loan commit-ment is purchased, no new information aboutthe project is revealed to the borrower or thebank, which may induce parties to renegotiateor walk away from the deal. This assumptionwill be relaxed in the next section.

Homogeneous investors: Every investor hasthe same project choice at the time the loancommitment is negotiated, so problems likeadverse selection are not at issue. This assump-tion will be relaxed when we discuss theinformational role of loan commitments.

Credible precommitment: In the model ofBoot, Thakor, and Udell, the bank commitsitself to provide a subsidy at time 1 and is com-pensated for the expected subsidy at time 0.Note that at time 1, when the market rate ishigh, the actual subsidy is greater than theexpected subsidy. Despite the obvious loss,the bank still honors the commitment. We willdiscuss this issue further in the “Loan Commit-ments Help Banks to Balance Reputational andFinancial Capital Optimally” section, below.

A final caveat: The results of Boot, Thakor,and Udell apply only to fixed-rate commit-ments. Within their sample of 2,526 loancommitments, Shockley and Thakor (1997)found only 13 (0.5 percent) that had fixedrates. Therefore, although preventing moralhazard is a plausible reason for the existenceof loan commitments, it does not seem to bethe driving force behind them.

Loan CommitmentsHelp Reduce OtherInvestment Distortions

Moral hazard created by debt financing is not limited to the asset-substitution problemdescribed above. Loan commitments alsoaddress overinvestment, underinvestment, andsuboptimal liquidation problems. From amodeling point of view, papers in this categoryuse Boot, Thakor, and Udell’s “tax now, subsi-dize later” idea but relax the “simple-project”assumption.

Consider a project with two investmentperiods, 0 and 1.7 At time 0, the time-1 invest-

ment is random. A risk-neutral borrower con-siders only the expected time-1 investment andtakes the project at time 0 if the expected netpresent value (NPV) is positive. With equityfinancing, the time-0 investment is a sunkcost at time 1, so the borrower continues theproject if the expected terminal cash flowsexceed the second-period investment. Withdebt financing, however, the borrower pro-ceeds differently. He repays the initial loanwhen cash flows are realized at the end, sothe initial investment is not sunk and causesunderinvestment if the repayment obliga-tion is sufficiently large.

A loan commitment with a usage feereduces the borrower’s payment burden fromthe first-period loan without negative profitimplications for the bank. The usage fee, paidon the available unused credit, compensatesthe bank for the interest rate concession, butits incidence is selective across borrowers.More fortunate investors, with lower second-stage requirements, pay more because of thegap between their borrowing and the creditlimit of the loan commitment. Investors withhigher second-stage requirements pay smaller fees because their borrowing is closer to the credit limit. That is, borrowers with lowfunding needs subsidize the less fortunate borrowers, giving not-so-lucky—but stillprofitable—investors an incentive to proceedwith their projects.

In a similar setting, Houston andVenkataraman (1996) further relax the “simple-project” assumption and analyze the firm’sliquidation decision. This time, the equity-financed firm compares its liquidation valueat time 1 to future cash flows and liquidates ifthe liquidation value is greater. With short-term debt, the initial investment is not sunk,but is a liability to be covered by the expectedpayoff. The firm liquidates when the debtobligation (not the liquidation value) is greaterthan the expected payoff. As a result, bond-holders receive the liquidation value, which isless than the expected payoff. Thus, short-termdebt leads to too-frequent liquidations. Withlong-term debt, firms never liquidate when thefirm’s liquidation value is less than the initialborrowing, because the liquidation value goesto bondholders. Thus, long-term debt causestoo-infrequent liquidations.

A short-term loan with a loan commitmentfor future funding alleviates the problem. The

■ 7 This example is from Berkovitch and Greenbaum (1991).

8

bank gives an interest rate subsidy and reducesthe borrower’s debt burden. This solves thetoo-frequent liquidation problem. The bor-rower compensates the bank with a commit-ment fee. However, because the fee must alsobe financed ex ante, the amount of debt thatthe firm must issue at the outset increases aswell. This offsetting effect limits commitments’ability to reduce the costs of suboptimalliquidations.

One problem with this explanation is thatthe subsidized interest rate on the initial loanmay cause overborrowing. Shockley (1995)points out that a loan commitment that includesa MAC clause mitigates this distortion; the com-mitment interest rate can be set low enough toprevent debt overhang, while the MAC clauseallows the bank to prohibit excessive reinvest-ment. As usual, the bank breaks even with thecommitment fee.8 Shockley provides evidencethat loan commitments reduce the cost of debt.Therefore, the capital structure of firms that useloan commitments is tilted in favor of more debt.

In all the papers discussed above, the bankprovides a sufficiently low interest rate andthe borrower always takes the right action.However, these papers do not consider animportant question: If the bank commits itselfto provide a subsidy, can the borrower exploitthat commitment and extract rents fromthe bank?

Houston and Venkataraman (1994) addressthis question.9 Banks acquire private infor-mation about their borrowers, which enablesthem to extract rents from successful firms bythreatening to withhold further credit.10 Thisreduces the borrower’s effort input, whichdetermines the probability that the borrower’sproject will turn out to be good or bad ; that is,the project will have safe and positive NPV orrisky and negative NPV. By providing a pre-arranged source of funds, loan commitmentslimit the lender’s ability to extract rents fromsuccessful projects. However, when the bankcommits itself to lend, two problems arise. Theborrower may exploit the commitment andextract rents from the bank by threatening toliquidate when the project is good and contin-uing when liquidation is more advantageous.More specifically, when the project is bad, theborrower refuses to liquidate unless the bankis willing to share the liquidation value withthe borrower. With a loan commitment, thebank charges a sufficiently high interest rateto induce liquidation. Note that this argumentcontradicts previous papers that found thatbanks reduced the interest rate by using a loancommitment to prevent debt overhang. How-

ever, as I explain next, loan commitmentscreate a selective debt overhang problem inthis model. Houston and Venkataramanassume that in a competitive banking mar-ket, the borrower’s project quality may berevealed to other lenders with positive proba-bility. So, although the high interest rate alsohurts the good project, a borrower with agood project can borrow from another bankand avoid commitment financing altogetherif his type is revealed. Therefore, the loancommitment’s high interest rate hurts borrowerswith bad projects that cannot find an alterna-tive funding source more than it hurts borrowerswith good projects. Selective debt overhangresolves the moral hazard problem because theborrower increases his effort supply to avoidthe high interest rate and the bad project.

The literature shows that loan commitmentsalso solve precontract information problems.This is what I discuss next.

Informational Role ofLoan Commitments

In this section, I relax the “homogeneousinvestors” assumption and introduce borrowerswith unobservable characteristics.

James (1981) is one of the early papersshowing that loan commitment parameters canbe designed to reveal a borrower’s unobserv-able characteristics. By demonstrating that thecost of maintaining compensating balancesdiffers among customers of different creditquality, James proved that the customer’schoice of payment option can be an effectivetool in separating borrowers with differentcredit qualities.

The observation that loan commitments canbe used as a screening or signaling mechanismhelps clarify a puzzle in the loan commitmentmarket. Borrowers often purchase loan com-mitments in order to back up commercialpaper issues. The argument is that loan com-mitments provide insurance to commercialpaper lenders. If the borrower’s cash flows arenot sufficient to cover its repayment obligation,it can always take down a loan under the com-mitment to meet its obligation. The problemwith this argument is the MAC clause. The fact

■ 8 Also see Morgan (1993).

■ 9 I will present a simplified version of the intuition here.

■ 10 See, for example, Rajan (1992).

9

that the borrower cannot repay its commercialpaper loan is sufficient reason for the bank tovoid the commitment. Then why do borrowerspurchase back-up loan commitments? Kanatas(1987) solved this puzzle.11 He showed that aloan commitment reduces a corporation’s bor-rowing cost in the commercial paper market,not because it provides a guaranty to commer-cial paper investors but because the purchaseof the loan commitment, along with its associ-ated price and future borrowing rate, commu-nicates payoff-relevant information to the com-mercial paper market.

The intuition is as follows: As of time 0,there are three possible states at time 1. If thefirm realizes the “good” state, it will be viewedas an improved credit risk and be able to rollover its first-period commercial paper at alower cost than it would have by exercisingthe commitment. Alternatively, the firm may bein one of the unobservable states in which itsdefault risk has increased. In the “impaired”state, the firm’s default probability hasincreased in such a way that the commitment-borrowing rate is lower than the new commer-cial paper rate and the commitment is exer-cised to repay the first-period commercialpaper debt. In the “very bad” state, the firm’sdefault risk has increased to such an extentthat the commercial paper market denies thefirm further credit and the bank refuses tohonor the commitment. The firm is thus forcedto default. Firms with a greater probability ofexercising the commitment (a higher prob-ability of being in the impaired state, given thatthe default risk has increased) are induced topurchase a larger commitment. An increase inthe commitment fee (expressed as a percent-age of the credit line) and a decrease in theinterest rate with increasing probability of theimpaired state is incentive compatible. Firmswith a high probability of being in the impairedstate recognize their greater likelihood ofbeing able to exercise the commitment advan-tageously and are therefore willing to pay ahigher fee. Firms with a low probability ofimpairment (higher likelihood of the very badstate) pay a lower fee in exchange for a highercommitment-borrowing rate in the unlikelyevent that they are able to exercise the commitment.

Deterioration in the borrower’s credit qualityis not the only risk a bank faces. Whether theborrower will actually take down the loan isanother uncertainty. Thakor and Udell (1987)show that when the bank does not know bor-rowers’ takedown probabilities, commitmentand service fees12 induce borrowers to sepa-

rate themselves through contract choice. Onecontract will have a high commitment fee anda low service fee, whereas the other will havea low commitment fee and a high service fee.A borrower with a high takedown probabilitywill want to avoid a large service fee becausethe likelihood of actually paying it is greater.On the other hand, a borrower with a lowtakedown probability is less averse to accept-ing a high service fee because the likelihood of actually paying that fee is lower. Such aborrower would like to minimize the commit-ment fee because it is a sunk cost that isincurred regardless of whether he exerciseshis commitment option. The borrower with ahigh takedown probability finds the largecommitment fee less burdensome because itrepresents the price of an option that he isvery likely to exercise. Thus, the difference intakedown probabilities fundamentally altersthe appeal of varying combinations of commit-ment and service fees to different borrowers,inducing each borrower to reveal his type.

The commitment and service fee combina-tion is not the only screening mechanism.Shockley and Thakor (1997) develop a ratio-nale for using commitment and usage feesjointly. In their model, there are three types ofborrowers: good (G ), medium (M ), and bad(B ). G is more likely than M to have a profit-able project and therefore more likely to takedown the loan. B does not have a project toinvest in. The bank wants to lend to G and Mbut not to B. In this case, the commitment feealone is not enough to separate the typesbecause if the fee is set to a level at which Mcan invest and B does not wish to invest, Gwill mimic M although he can pay a higher fee.Note that the commitment is more valuable toG than to M because G is more likely to exer-cise its option. Solving this problem requiresmaking M ’s contract less attractive to G ’s man-ager. This is achieved by reducing the payoffto firm M in the state in which the loan istaken down, by increasing the interest rate.This increase diminishes the value of the com-mitment less for M than for G because M has alower probability of taking down the loan.Because the bank operates in a competitivemarket, it reduces the commitment fee tocompensate M for the higher interest rate. The

■ 11 Also see Calomiris (1989).

■ 12 The paper refers to usage fees, but it is more accurate to callfees levied on the borrowed amount service fees.

10

problem is that this simultaneous reduction inthe commitment fee makes the contract attrac-tive to B . A usage fee makes the contractexpensive for B because he never takes downthe loan. On the basis of their model, Shockleyand Thakor make the following four predic-tions and provide evidence to support them.First, if the fee structure helps reveal the bor-rower’s type, loan commitments should con-tain a pricing structure with multiple fees whenthe firm has assets that are hard to value orthe firm’s credit quality is poor. Second,there must be a negative correlation betweeninterest rate markups and usage fees. Third,announcing a loan commitment purchaseshould generate an abnormal positive pricereaction. Fourth, the price reaction must begreater if the commitment has a multiple feestructure because the commitment revealsinformation about a firm that is hard to value.

Although it is possible to obtain a full sepa-ration of types by using the multiple fee struc-ture, this method is limited to two—or at mostthree—types. If there are several unobservabletypes, the multiple fee structure alone may notbe enough to separate all of them. Thakor(1989) analyzes this case, deriving the condi-tions under which a forward contract is moreeffective than a spot contract in separatingtypes. The intuition is that in the forward mar-ket, the future state of the world is still uncer-tain. If the relationship among types is suchthat, for each type, there is at least one state ofthe world where that agent type is the mostlikely to attain that state, state-specific subsi-dies can be used as an additional contractingvariable. For example, at some point in time,the bank promises an agent of a given type asubsidized contract in a particular state at thenext point in time. In exchange, the bankdemands a fee at the first point in time. Typesthat are less likely to attain that state find thesubsidy too expensive. A separate fee-subsidycombination can be designed to be the mostattractive for each agent type.

Finally, Duan and Yoon (1993) explain howloan commitments can be used as a signalingdevice. Like Shockley (1995) and Morgan(1993), Duan and Yoon recognize that the sub-sidized funds provided by a loan commitmentlead to overinvestment. So the larger a bor-rower’s credit line is, the higher is the cost ofoverinvestment. Note that borrowers with highsuccess probabilities (high expected profits)can operate at higher costs than borrowers

with low success probabilities. Therefore, aborrower with a high probability of successcan use overinvestment to distinguish itselffrom other borrowers, anticipating that it willbe treated favorably in terms of loan pricing.That is, the credit limit can be used to signala borrower’s quality. Once the credit lines arein place, the firms with higher success proba-bilities will engage in suboptimal investmentswhen future spot rates are higher than loancommitment rates. Thus, the signalingequilibrium destroys value.

Loan Commitments GiveBorrowers a StrategicAdvantage

Maksimovic (1990) shows that the structure ofthe borrower’s industry determines the termsof loan commitments. In industries with im-perfect competition, the option to acquirefinancing at predetermined rates enhances theborrower’s strategic position and creates valuefor the borrower. A firm that has access toresources at a lower marginal cost than itscompetitors has a strategic advantage that itcan exploit to gain a larger market share andhigher profits. A firm can create such an advan-tage by purchasing, for a fixed initial fee, anoption to acquire financing on favorable terms.The ability to exercise the commitment makesthe firm a strategic threat to its rivals and movesthe industry to an equilibrium more favorableto that firm. Therefore, it is optimal for allfirms to acquire bank loan commitments,altering the industry equilibrium in the process.

All the models that attempt to explain whyloan commitments exist have two major short-comings. First, as I noted earlier, the modelsthat assume a fixed interest rate can justify onlya small fraction of the outstanding loan com-mitments. Second, models that rationalize themultiple fee structure as a screening mecha-nism are applicable only to situations in whichthere are at most three unobservable types ofborrowers. Although Thakor (1989) allows forseveral types, his model imposes very strongrestrictions on the attributes of types. Theconclusion is that we still have a lot to learnabout the significance of loan commitments’fee structure.

11

demand information. The usage fee will behigher for firms that report higher expectedloan demand, whereas the loan rate offered tosuch firms will be lower. The intuition is thatfirms with high loan demand are insensitive tohigh usage fees because they will most likelyuse the entire credit line and not pay the usagefee. Firms with low loan demand will reporttheir information truthfully despite the lowinterest rate offered to investors with highdemand because they wish to avoid the highusage fee.

Loan CommitmentsHelp Banks BalanceReputational andFinancial Capital Optimally

Loan commitments are discretionary contractsbecause the MAC clause gives the bank theright to refuse a loan when the borrowerrequests it. However, if a bank honors itscommitment even when it is costly to do so, it can enhance its reputation for keeping itspromises. A good reputation makes its futurecommitments more valuable because borrow-ers are willing to pay a premium for a crediblecommitment. Thus, a bank may use the loancommitment to enhance its reputation.

Boot, Greenbaum, and Thakor (1993) for-malize this idea. The party that has discretiongains the option of taking a costly action. If thecost is sufficiently high, only agents that canafford to pay the cost can take the action, sig-nal their types, and improve their reputations.In Boot, Greenbaum, and Thakor’s model,future spot rates are uncertain. The bankpromises to give the borrower an interest ratesubsidy if the future spot rate is too high. Thebank is compensated beforehand with a feethat represents the expected cost of the sub-sidy.14 The cost of honoring a discretionaryloan commitment is that when the borrowertakes down the loan, the actual subsidy isgreater than the expected subsidy that the

Supply-Side Explanations

Loan Commitments HelpLower Regulatory Taxesfor Banks

Regulatory taxes are defined as the costs ofthe federal deposit insurance premium, theconstraints placed on increased financial inter-mediation by regulators’ capital requirements,and the opportunity cost of maintaining legallyrequired reserves. It has been argued that off-balance-sheet activities allow banks to generatefee income and bypass regulatory taxes. Forexample, until the commitment is taken down,there is no loan, which means that the bankdoes not have to collect deposits, keepreserves, or pay deposit insurance premiums.Actually, the bank can sell the commitment,collect the fee, and avoid regulatory taxes al-together by selling the loan to another bank assoon as it is originated.13 However, Kareken(1987) reports that there was no change inbank regulatory policy of the sort that wouldhave prompted banks to start issuing loancommitments suddenly. From April 1969through mid-1973, the Federal Reserve System’sreserve requirement schedule was changedonly once, in November 1972, when the aver-age reserve requirement was decreased. Theeffective per dollar deposit insurance premiumwas not changed in that period either. In 1971,there were no minimum capital-asset ratios.Thus, regulatory taxes fail to explain the exis-tence of loan commitments.

Loan CommitmentsImprove Banks’ Forecasts of Future Loan Demand

Greenbaum, Kanatas, and Venezia (1991)suggest that loan commitments reduce banks’uncertainty about future loan demand and itsattendant costs. In their setting, banks canborrow after the loan demand is known or byprearrangement. Prearranged funds can beobtained at a lower interest rate. Recognizingtheir informational disadvantage, banks offer toshare the benefit of their lower funding costs,provided that clients disclose private informa-tion regarding prospective credit demand. Aloan commitment contract incorporating ausage fee and a forward interest rate motivateshonest disclosure of the borrower’s loan

■ 13 Greenbaum (1986) argues that banks became high-cost lendersbecause the Federal Reserve and the FDIC ceased to set limits on the ratesthat banks could pay to creditors; as a result, banks are burdened by higherborrowing costs as well as regulatory taxes. However, they still maintaintheir cost advantage as raters of borrowers. Hence, they offer loan commit-ments and then sell the loans they originate.

■ 14 I discussed the same model in the section titled “Loan Commitments Help Attenuate Moral Hazard.”

12

bank was compensated for when the spot ratewas still uncertain. Therefore, it is costly for abank to honor a commitment. Then, high-quality banks with more economic power thanlow-quality banks can signal their type andimprove their reputational capital by honoringthe discretionary contract and reducing theircurrent financial capital, while low-qualitybanks repudiate their commitments, preservetheir financial capital, and forgo the futurebenefits of a better reputation. In other words,a loan commitment helps the bank to manageits portfolio of financial and reputationalcapital optimally.

The idea that banks can use loan commit-ments as a signaling mechanism has beenempirically verified by Mosebach (1999). Hisargument is based on a paper by Billett,Flannery, and Garfinkel (1995) reporting that“more reputable” lenders give the market morenew information than “less reputable” lendersdo. Billett, Flannery, and Garfinkel also pro-pose that firms endeavor to send the strongestsignal possible to the market by using the bestlender. Mosebach argues that large companies,wishing to send the strongest possible signal tothe market, use the best bank and largest lineof credit available. So the purchase of a loancommitment transmits the following informa-tion: First, by selecting a particular bank, theborrower signals his belief that this is the bestlender available to him. Second, the purchasecommunicates to the market new, positiveinformation about the bank’s current andfuture financial position. Mosebach’s findingsshow a positive and significant market reactionto the bank’s stock when the bank grants aline of credit.

Banks’ Advantagesover Individuals andOther Institutions in Providing Liquidity throughCommitments

If loan commitments have the benefits describedin the previous section, then why do not otherfinancial intermediaries offer them? The litera-ture on this question builds on literature deal-ing with the emergence of organizations. So Ifirst explain why institutions’ commitments aremore credible than individuals’ and thendescribe banks’ advantage over other financialintermediaries in selling loan commitments.

Banks Can CommitThemselves Credibly butIndividuals Cannot

When individuals sign up for the futuredelivery of a product or service, they prefer tocontract with a firm or organization rather thananother individual. Thus, individuals buy in-surance from insurance companies and rarelyfrom other individuals; loan commitments aresold by banks and not by individuals. Whycan firms—but not individuals—crediblycommit themselves to supply a product orservice in the future in exchange for currentcompensation?

Boot, Thakor, and Udell (1991) offer theintuition that it is more costly for an organiza-tion not to honor its contractual commitmentsthan it is for an individual. In a setting whererisky debt and limited liability create moralhazard at sufficiently high interest rates, thelender gives the borrower a subsidized rate toprevent moral hazard and recovers that sub-sidy with an up-front fee paid when the com-mitment is sold. The problem with an individ-ual lender offering a commitment is that hecan collect the commitment fee, consume hisentire wealth, and repudiate the commitment.No penalty or other legal enforcement mecha-nism can remedy the situation. To prevent theindividual lender from consuming his wealth,an individual banker with a nonconsumableproject endowment can collect this wealth asa deposit and sell a commitment to the bor-rower. If the banker repudiates the contract, a court can seize the banker’s project endow-ment. The trouble with this setting is thatbecause the subsidy is provided only wheninterest rates are high, the commitment feereflects only the subsidy’s expected cost andtherefore is less than the ex post amount of thesubsidy. Therefore, the banker will repudiatethe contract if the loss from honoring it (thedifference between the commitment fee andthe subsidy) is greater than the cost of losingits project endowment. In contrast, a bank ismade of a countable infinity of individualbankers (equity holders), each with a projectendowment that will be seized if the commit-ment is repudiated. Note that in this case, theloss incurred by each banker from honoringthe commitment is zero because a finite loss isdivided among an infinity of bankers, whilerepudiation entails the loss of each banker’sproject endowment. Clearly, the bank alwayshonors the commitment. Hence, the emer-gence of organizations prevents market failure

13

that might be caused by individuals not honor-ing contracts. Although this result is quiteintuitive, it is unclear why individuals cannotplace a fraction of their wealth in an escrowfund that the courts may seize if the individualfails to honor his commitment. Such an escrowfund would easily make individuals’ commit-ments credible.

Finally, note that the courts play an impor-tant role in Boot, Thakor, and Udell (1991) bypenalizing bank shareholders when commit-ments are not honored. Boot, Greenbaum, andThakor (1993) ignore the judiciary and showthat reputational concerns may be enough toinduce the banks to keep their promises. How-ever, reputational concerns are not enough toexplain why banks alone sell loan commit-ments; an individual might have similar con-cerns and honor his commitments.

Banks Have a CostAdvantage over OtherInstitutions

It is clear from the previous discussion thatloan commitments will be sold by institutionallenders. The question is, why must this institu-tion be a bank and not another form of finan-cial intermediary? Kareken (1987) argues thattechnological advances decreased the cost ofacquiring and processing information, whichopened the direct credit market to a largenumber of borrowers. These borrowers, how-ever, have to be rated and monitored by mar-ket participants. Kareken assumes that techno-logical advances created a larger decrease inbanks’ information acquisition costs than inthose of other lenders. Then, purchasing abank loan commitment results in lower directcosts for lending, rating, and monitoringbecause the bank assumes the default risk anddoes the monitoring. Two objections may beraised against this argument. First, it is not clearwhy technological advances benefit banksmore than they benefit other intermediaries.Second, Kareken ignores the MAC clause thatrelieves the bank of its commitment when theborrower’s financial condition deteriorates.Therefore, Kareken’s argument does not explainwhy banks alone offer loan commitments.

Kanatas (1987) provides an informal solu-tion to this puzzle, arguing that only banks sellloan commitments because they have access tothe discount window. Their ability to meetunexpectedly high commitment loan demandwith relatively low-cost funds from the dis-

count facility makes their expected cost offunding commitments lower than that of non-bank competitors. If this subsidy more thanoffsets the cost of the reserve requirement,only banks will sell commitments.

Similarly, Kashyap, Rajan, and Stein (forth-coming) allude to the cost of the reserverequirement to formalize the cost advantageissue and explain why the intermediary sellingthe loan commitment must be a bank. Theyshow why deposit taking and lending activitiesare carried out by a single institution (commer-cial bank) rather than separate institutions.They argue that loan commitments let a banktake advantage of economic synergies betweenits deposit-taking and lending activities.Demand deposits and loan commitments bothprovide liquidity on demand to bank cus-tomers who have unpredictable liquidityneeds. If these contracts require costly over-head in the form of cash and security holdings,a synergy will exist to the extent that the twoactivities can share some of the costly over-head. A bank that offers both deposits andloan commitments can get by with a smallertotal volume of cash and securities on its bal-ance sheet than would two separate institu-tions, each specializing in only one of the twofunctions. Hence, efficiency is enhanced.

Pricing Loan Commitments as Put Options

Loan commitments have several similarities toput options. The commitment buyer pays acommitment fee for the right to sell a securityto the bank at a prespecified price over somepreviously established time interval. The secu-rity is the commitment owner’s debt, and thestrike price is the dollar amount of the borrow-ing. The buyer will exercise the put option andtake down the loan if the value of his debt onthe exercise date is less than the committedloan amount. Clearly, this description excludesthe two most important features of loancommitments that the literature attempts toexplain: the multiple fee structure and the MACclause. Yet, these and other simplifying assump-tions, which I review next, are needed to applythe option pricing theory to loan commitments.

14

Thakor, Hong, and Greenbaum (1981)made the first attempt to rationalize and priceloan commitments as put options.15 Theirpaper develops a model for valuing variable-rate bank loan commitments within the frame-work of the Black and Scholes methodology.16

It is a preliminary step in the valuation of loancommitments and therefore ignores several keyfactors in order to obtain a valuation formula.

There are four key differences between loancommitments and exchange-traded put options.

1) Exchange-traded options are binding,while loan commitments are discretionarybecause of the MAC clause.

2) Loan commitments are not transferable.3) Loan commitments have a different

pricing structure (usage and service fees).4) A put option is either exercised in full or

not at all. Loan takedowns, however, areusually only a fraction of the commit-ment’s face value.

The literature has not addressed the ques-tion of how the first three points affect thevaluation of loan commitments as put options.Attempts have been made, however, to explainthe partial takedown phenomenon.

Thakor, Hong, and Greenbaum (1981)provide the first explanation of the partial take-down phenomenon. They argue that the futurepricing and availability of bank services areinfluenced by the degree to which a customerexercises his loan commitment, because a gainfor the customer is a loss to the bank. In estab-lishing the price of the commitment and thesize of the fixed mark-up, the bank considersexpected borrower behavior under alternatestates of the world. If the borrower surprisesthe lender by borrowing more than expected,the lender revises his expectations and adjustsupward the price and/or the mark-up appli-cable to future commitment transactions.Therefore, when the firm chooses the take-down fraction, it minimizes the expected costof the next loan plus the opportunity loss fromnot taking down the current loan fully.

In Thakor, Hong, and Greenbaum, the bankuses an exogenous process for updating thecommitment fee and the fixed mark-up basedon take-down behavior. Greenbaum andVenezia (1985) endogenize this process byassuming that the loan amount taken down bythe borrower depends on his productivity,unobservable to the bank. The borrower’sproductivity is subject to random mean-zerochanges. The bank infers the borrower’s pro-ductivity from the takedown. High takedownsignals high productivity; this means that highfuture takedowns are associated with higher

net costs to the bank since they imply that theborrower is exercising his put option more.As the bank obtains new estimates of the bor-rower’s productivity, the average of those esti-mates yields a less noisy signal of productivity,so price adjustments to unexpectedly hightakedowns become less significant over time.The interest rate smoothing that results fromthe bank–borrower relationship prevents theborrower from switching to other banks. Thislast result, however, depends on the strong, ifnot unrealistic, assumption that the new bankknows nothing about the client’s takedownhistory and that new customers are indistin-guishable from switching customers.

II. The Effects ofLoan Commitmentson the Bank’s RiskExposure

I have already mentioned that when a banksells a loan commitment, it accepts the interestrate and quantity risk that the customer wouldbear if he were to borrow in the spot market.Although the commitment fee is expected tocompensate the bank for its risk exposure,regulators believe that loan commitmentsincrease the risk exposure of banks and thedeposit insurer. Regulators argue that becausethe potential liability of a loan commitment isnot quantified and reflected in the depositinsurance premium, a bank may be tempted totake on excessive risk by expanding its loancommitments, which may result in an under-estimation of the deposit insurer’s risk expo-sure. Therefore, regulators have imposedcapital requirements against bank loan com-mitments to control their growth. Some of theliterature on loan commitments providesinsight on the merit of these arguments.

■ 15 Hawkins (1982) argues that revolving credit agreements (loancommitments with infinite maturity) are similar to callable bonds. He baseshis argument on transaction costs to rationalize loan commitments.

■ 16 See Thakor (1982) for the valuation of fixed-rate loan commitments.

15

Early papers (Ho and Saunders [1983] andKoppenhaver [1985]) asked whether the bankcould use financial futures contracts to hedgeagainst interest and quantity risks. The mainfinding is that unless the spot loan price andthe expected quantity of loan takedowns areperfectly correlated, the bank cannot hedge itsrisks fully. That is, it can hedge against one ofthe two variables by buying or selling futurescontracts, but if the two variables are not per-fectly correlated, a single type of contract isinsufficient to hedge against both types of risk.Clearly, these early papers took the increase inbanks’ risk exposure as a given and did notinvestigate whether loan commitments actuallyincrease the bank’s risk exposure. Avery andBerger (1991) and Boot and Thakor (1991)addressed this issue.17

Avery and Berger argue that selling a loancommitment is risky because the bank islocked into lending to a borrower who mightsuffer a decline in creditworthiness that wouldotherwise dictate a higher interest rate or noloan at all. To make this argument, they assumethat invoking the MAC clause is costly and thebank bears the legal costs. However, they donot clarify why the bank cannot recoverthe costs ex ante with the commitment fee.Because the borrower’s creditworthiness maychange over time, the bank has less informa-tion about the borrower when the loan com-mitment is sold than when spot contracts aresigned. This leads to moral hazard. Now, sup-pose there are borrowers with and withoutmoral hazard problems. Due to informationaldifficulties, the bank may ration moral hazardborrowers. If those who are rationed and waitfor the spot market are safe borrowers (infor-mation is revealed in the spot market andthese borrowers can borrow there), the bank’sloan commitment portfolio consists of riskier-than-average borrowers and the bank’s riskexposure is augmented. Otherwise, if moralhazard borrowers are the risky ones, thebank’s risk exposure is reduced. Avery andBerger empirically find that fewer problemloans and higher bank income are associatedwith loan commitments. Therefore, commit-ments reduce the bank’s risk exposure.

Boot and Thakor (1991) find that loan com-mitments lower bank asset portfolio risk fortwo reasons: First, the loan commitment con-tract can be designed to resolve the asset sub-stitution problem between the bank and theborrower.18 Second, if the bank’s existing spotloan portfolio in a given period is observableto its loan commitment customers in thatperiod, then optimally the bank will choose to

make spot loans to less risky borrowers. Theintuition is that the bank’s current loan com-mitment revenue is an increasing function ofthe likelihood that the bank will be solvent inthe future when it will honor the commitment.Hence, an increase in the riskiness of its spotloan portfolio causes a reduction in its loancommitment revenue. From this result, Bootand Thakor draw the following importantpolicy implication: The deposit insurer shouldinsist that all of the bank’s outstanding commit-ments be voided if the bank cannot pay off itsdepositors and is bailed out by the insurer.That is, the deposit insurer should transfersome of the risk to loan commitment cus-tomers to give them an incentive to monitorthe bank’s spot loan portfolio.

Clearly, the capital requirements that regula-tors impose on loan commitments to protectthe deposit insurer are not needed because,unlike other off-balance-sheet liabilities, suchas standby letters of credit for which the bankacts as a guarantor, loan commitments with aMAC clause do not impose any credit risk on the bank. In fact, as Boot and Thakor show,they lower the bank’s asset risk when the bank’sloan portfolio is observable to customers.

III. Loan Commitmentsand Monetary Policy

Regulators conduct monetary policy throughquantity rationing and interest rate channels byaltering the quantity of credit and its price, theinterest rate. Tighter monetary policy creates areserve shortage that raises the cost of funds tobanks. When their cost of funds rise, banksraise loan rates, which causes businesses andconsumers to cut down expenditures. Theinterest rate channel implies a relationshipbetween monetary policy and bank loan rates,loan volume, and economic activity. The quan-tity rationing channel refers to the possibilitythat when banks’ funds costs rise, they chooseto reduce the volume of loans above anyreduction caused by an increase in interestrates on loan demand. This channel implies adirect link between monetary policy and thequantity of bank loans.

■ 17 Hassan and Sackley (1994) showed empirically that loan commitments reduce a bank’s risk exposure.

■ 18 See the discussion on moral hazard in section I.

16

Since loan commitments protect borrowersfrom quantity rationing, in the short run, mon-etary policy changes will affect loans undercommitment only through the interest ratechannel (Duca and Vanhoose [1990], Morgan[1994], and Woodford [1996]). If monetarypolicy tightens, banks resort to rationing cus-tomers without commitment agreements(Sofianos, Wachtel, and Melnik [1990] andGlick and Plaut [1989]). In the long run, as loancommitments expire, quantity rationing appearsin the form of refusing to renew a commitmentor reducing its size. Therefore, loan commit-ments introduce significant lags in the effect ofmonetary policy (Deshmukh, Greenbaum, andKanatas [1982] and Morgan [1998]).

IV. ConcludingRemarks

I have provided a summary of what we knowabout loan commitments after years of research.Although we have gained some understandingof what value loan commitments provide, ourknowledge has clear limitations. For example,in many instances, economists’ conclusionsdepend on the use of fixed-rate commitments,which are rather uncommon in the market. Inpapers where loan commitments can be usedto distinguish between borrowers with a prioriunobservable characteristics, the results arelimited to settings where there are at mostthree unobservable types, which is far toorestrictive.

There is still much to be learned about loancommitments. I conclude by briefly reviewingthree of the major unresolved issues.

First, the courts have often obstructedbanks’ right to invoke the MAC clause anddeny credit to a loan commitment owner,arguing that the banks had not acted in goodfaith (Edelstein [1991] and Budnitz and Chaitman[1998]). That is, the courts have often inter-preted banks’ use of the clause as an abuse ofpower. This is at odds with the current litera-ture, which views the MAC clause in loan com-mitments as providing the bank discretion thathas economic value (Boot, Greenbaum, andThakor [1993]). Courts’ reasons for interveningand the welfare effects of their interventionremain to be understood.

Second, the analysis of loan commitmentshas been limited to models where, in mostinstances, the bank collects a fee at time 0 andin return provides a subsidy at time 1. Unfor-tunately, all these models ignore the fact thatbank loans are relationship loans. That is,

banks acquire private, firm-specific informationduring their relationship with borrowers andexploit their informational advantage relativeto other lenders to earn positive profits in thefuture.19 An important implication is that banksare willing to take losses early if they expect torecover them in the future. In a setting likethis, a bank can give the borrower a subsidywith a standard debt contract and recover thesubsidy from future transactions rather thanwith the commitment fee. So it is not clear whya borrower would choose a loan commitmentover a spot loan or vice versa. Therefore, weneed a model that rationalizes loan commit-ments in a relationship setting.

Finally, a loan commitment is an incompletecontract. Important issues, such as loan maturityand debt covenants, are left open to negotia-tion and are finalized before the loan closes.Although loan commitments have been metic-ulously scrutinized, we know nothing aboutthe properties of loans made under commit-ment. How much the final loan agreementdiffers from the terms specified in the loancommitment deserves further investigation.

■ 19 Boot and Thakor (1994) show that long-term relationships arefeasible even without the learning component. In their model, the bankinitially lends with a secured contract (collateral is costly) at a highinterest rate. Once the borrower succeeds, future loans are unsecured andsubsidized. This feature induces the borrower to work hard at the outset tosucceed as soon as possible. In contrast to Rajan’s (1992) relationshipsetting, in which the borrower is subsidized initially and taxed later, inBoot and Thakor taxation occurs before the subsidy.

17

References

Avery, R.B., and A. Berger. “Loan Commit-ments and Bank Risk Exposure,” Journal of Banking and Finance, vol. 15 (1991), pp. 173–92.

Berkovitch, E., and S.I. Greenbaum. “TheLoan Commitment as an Optimal FinancingContract,” Journal of Financial and Quanti-tative Analysis, vol. 26 (1991), pp. 83–95.

Billett, M.T., M.J. Flannery, and J.A.Garfinkel. “The Effect of Lender Identity ona Borrowing Firm’s Equity Return,” Journalof Finance, vol. 50 (1995), pp. 699–718.

Boot, A.W.A. “Relationship Banking: What DoWe Know?,” Journal of Financial Interme-diation, vol. 9 (2000), pp. 7–25.

________, S.I. Greenbaum, and A.V. Thakor.“Reputation and Discretion in FinancialContracting,” American Economic Review,vol. 83 (1993), pp. 1165–83.

________, and A.V. Thakor. “Off-balanceSheet Liabilities, Deposit Insurance and Capital Regulation,” Journal of Bankingand Finance, vol. 15 (1991), pp. 825–46.

________, and ________. “Moral Hazard andSecured Lending in an Infinitely RepeatedCredit Market Game,” International Eco-nomic Review, vol. 35 (1994), pp. 899–920.

________, ________, and G.F. Udell. “Compe-tition, Risk Neutrality and Loan Commit-ments,” Journal of Banking and Finance,vol. 11 (1987), pp. 449–71.

________, ________, and ________. “CredibleCommitments, Contract Enforcement Prob-lems and Banks: Intermediation as Credibil-ity Assurance,” Journal of Banking andFinance, vol. 15 (1991), pp. 605–32.

Booth, J.R., and L. Chua. “Structure andPricing of Large Bank Loans,” FederalReserve Bank of San Francisco EconomicReview, no. 3 (1995), pp. 52–62.

Budnitz, M.E., and H.D. Chaitman. Law ofLender Liability. Arlington, Va.: A.S. Pratt &Sons, 1998.

Calomiris, C.W. “The Motivations for LoanCommitments Backing Commercial Paper,”Journal of Banking and Finance, vol. 13(1989), pp. 271–77.

Campbell, T.S. “A Model of the Market forLines of Credit,” Journal of Finance, vol. 33(1978), pp. 231–44.

Deshmukh, S. D., S.I. Greenbaum, andG. Kanatas. “Bank Forward Lending inAlternative Funding Environments,” Journalof Finance, vol. 37 (1982), pp. 925–40.

Duan, J., and S.H. Yoon. “Loan Commit-ments, Investment Decisions and the Signaling Equilibrium” Journal of Bankingand Finance, vol. 17 (1993), pp. 645–61.

Duca, J.V., and D.D. Vanhoose. “Loan Com-mitments and Optimal Monetary Policy,”Journal of Money, Credit, and Banking,vol. 22 (1990), pp. 178–94.

Edelstein, P.M. “Effective Letters of Intent andCommitment Letters,” Journal of Commer-cial Bank Lending, vol. 73 (1991), pp. 6–16.

Glick, R., and S.E. Plaut. “Money Demandand Off-Balance Sheet Liquidity: EmpiricalAnalysis and Implications for MonetaryPolicy,” Journal of Accounting, Auditing,and Finance, vol. 4 (1989), pp. 147–59.

Goldberg, K.J. “Lender Liability and GoodFaith,” Boston University Law Review, vol. 68(1988), p. 653.

Greenbaum, S.I. “Securitization, RegulatoryReform, and Bank Asset Quality,” Unpub-lished manuscript, 1986.

_________, and A.V. Thakor. ContemporaryFinancial Intermediation. Fort Worth, Texas,Dryden Press, 1995.

_________, and I. Venezia. “Partial Exercise ofLoan Commitments under Adaptive Pricing,”Journal of Financial Research, vol. 8 (1985),pp. 251–63.

________, G. Kanatas, and I. Venezia. “LoanCommitments and the Management ofUncertain Demand,” Journal of Real EstateFinance and Economics, vol. 4 (1991), pp. 351–66.

18

Hassan, M.K., and W.H. Sackley. “A Method-ological Investigation of Risk Exposure ofBank Off-Balance Sheet Loan CommitmentActivities,” Quarterly Review of Economicsand Finance, vol. 34 (1994), pp. 283–99.

Hawkins, G.D. “An Analysis of RevolvingCredit Agreements,” Journal of FinancialEconomics, vol. 10 (1982), pp. 59–81.

Ho, T.S.Y., and A. Saunders. “Fixed RateLoan Commitments, Take-Down Risk andDynamics of Hedging with Futures,”Journal of Financial and QuantitativeAnalysis, vol. 18 (1983), pp. 499–516.

Houston, J.F., and S. Venkataraman. “Infor-mation Revelation, Lock-In, and Bank LoanCommitments,” Journal of Financial Inter-mediation, vol. 3 (1994), pp. 355–78.

________, and ________. “Liquidation UnderMoral Hazard: Optimal Debt Maturity andLoan Commitments,” Journal of Bankingand Finance, vol. 20 (1996), pp. 115–33.

James, C. “Self-selection and the Pricing ofBank Services: an Analysis of the Market forLoan Commitments and the Role of Com-pensating Balance Requirements,” Journalof Financial and Quantitative Analysis,vol. 16 (1981), pp. 725–46.

________. “An Analysis of Bank Loan RateIndexation,” Journal of Finance, vol. 37(1982), pp. 809–25.

Kanatas, G. “Commercial Paper, Bank ReserveRequirements, and the Informational Role ofLoan Commitments,” Journal of Bankingand Finance, vol. 11 (1987), pp. 425–48.

Kareken, J.H. “The Emergence and Regu-lation of Contingent Commitment Banking,”Journal of Banking and Finance, vol. 11(1987), pp. 359–77.

Kashyap, A.K., R. Rajan, and J.C. Stein.“Banks as Liquidity Providers: An Expla-nation for the Co-Existence of Lending andDeposit-Taking,” Journal of Finance,(forthcoming).

Koppenhaver, G. D. “Variable-Rate LoanCommitments, Deposit Withdrawal Risk andAnticipatory Hedging,” Journal of FuturesMarkets, vol. 5 (1985), pp. 317–30.

Maksimovic, V. “Product Market Imperfec-tions and Loan Commitments,” Journal ofFinance, vol. 45 (1990), pp. 1641–54.

Manino, E.F. “Lender Liability and Other Com-plex Litigation Involving Financial Institu-tions,” study materials, ALI-ABA Conference,February 17–18, 1994, Palm Beach, Florida.

Morgan, D.P. “Financial Contracts when Costsand Returns are Private,” Journal of Mone-tary Economics, vol. 31 (1993), pp. 129–46.

_________. “Bank Credit Commitments, CreditRationing and Monetary Policy,” Journal ofMoney, Credit, and Banking, vol. 26 (1994),pp. 87–101.

_________. “The Credit Effects of MonetaryPolicy: Evidence Using Loan Commitments,”Journal of Money, Credit, and Banking,vol. 30 (1998), pp. 102–18.

Mosebach, M. “Market Response to BanksGranting Lines of Credit,” Journal of Banking and Finance, vol. 23 (1999), pp. 1707–23.

Petersen, M.A., and R. Rajan. “The Effect ofCredit Market Competition on Lending Rela-tionships,” Quarterly Journal of Economics,vol. 110 (1995), pp. 407–42.

Rajan, R. “Insiders and Outsiders: The ChoiceBetween Informed and Arm’s-Length Debt,”Journal of Finance, vol. 47 (1992),pp. 1367–1400.

Sharpe, S.A. “Asymmetric Information, BankLending, and Implicit Contracts: A StylizedModel of Customer Relationships,” Journalof Finance, vol. 45 (1990), pp. 1069–87.

Shockley, R.L. “Bank Loan Commitments andCorporate Leverage,” Journal of FinancialIntermediation, vol. 4 (1995), pp. 272–301.

19

_________, and A.V. Thakor. “Bank LoanCommitment Contracts: Data, Theory andTests,” Journal of Money, Credit, andBanking, vol. 29 (1997), pp. 517–34.

Sofianos, G., P. Wachtel, and A. Melnik.“Loan Commitments and Monetary Policy,”Journal of Banking and Finance, vol. 14(1990), pp. 677–89.

Thakor, A.V. “Toward a Theory of Bank LoanCommitments,” Journal of Banking andFinance, vol. 6 (1982), pp. 55–83.

_________. “Competitive Equilibrium withType Convergence in an AsymmetricallyInformed Market,” Review of FinancialStudies, vol. 21 (1989), pp. 49–71.

________, H. Hong, and S.I. Greenbaum.“Bank Loan Commitments and Interest RateVolatility,” Journal of Banking and Finance,vol. 5 (1981), pp. 497–510.

________, and G.F. Udell. “An EconomicRationale for the Pricing Structure of BankLoan Commitments,” Journal of Bankingand Finance, vol. 11 (1987), pp. 271–89.

Woodford, M. “Loan Commitments and Opti-mal Monetary Policy,” Journal of MonetaryEconomics, vol. 37 (1996), pp. 573–605.

20

A Simple Model ofMoney and Bankingby David Andolfatto and Ed Nosal

David Andolfatto is an associate professorof economics at Simon Fraser University.Ed Nosal is an economist at the FederalReserve Bank of Cleveland. This articlerepresents a simplified version of theirwork in progress, “Money and Bankingwith Costly State Verification.” Theauthors thank Peter Rupert andBruce Smith for their helpful commentsand discussion.

Introduction

This article presents a simple environmentgiving rise to banks that create and lend outmoney. We define money to be any objectthat circulates widely as a means of payment.In our model, this object takes the form of afully secured and redeemable bearer bond.This monetary instrument is issued by an agentthat can credibly commit to monitoring a poolof real investments; that is, this capital formsthe requisite backing for a circulating privatedebt instrument. While direct trade in securitiesis feasible without money, we find that moneycan economize on monitoring costs, whichenhances the efficiency of the exchange process.

We define a bank as an agency that simul-taneously issues money and monitors invest-ments. In reality, banks also accept deposits ofmoney, which are then redirected to borrowers.In our model, banks do not accept deposits;we do not view this function as a definingcharacteristic of a bank.1 In particular, financialmarkets also accept deposits of money inexchange for marketable liabilities (equity anddebt instruments). We think that banks differfrom financial markets in two ways. First, bankliabilities are designed to be high-velocity

payment instruments (money). Second, banksspecialize in screening and monitoring theirinvestments. Banks in our model perform bothof these functions.

While our framework allows privatenon-bank liabilities to serve as the economy’smedium of exchange (as mentioned earlier,exchange is even possible without any moneyat all), we demonstrate that the cost-minimizingstructure has a bank creating liquid funds,which are then lent to borrowers (for example,entrepreneurs) with suitable collateral (contin-gent claims against future output). These liquidfunds constitute real bills of exchange; that is,they are backed by the issuing bank withenforceable claims against real assets (thecollateral supplied by borrowers).

■ 1 Needless to say, there are those who would disagree with this pointof view. To our knowledge, Bullard and Smith (2001) have the only modelfeaturing intermediaries that simultaneously take deposits, make loans, andissue circulating liabilities.

21

In reality, the vast bulk of the moneysupply consists of private debt instruments withcontractual features similar to those embeddedin the debt instruments that circulate in ourmodel. In addition, the bulk of this money iscreated by banks; that is, institutions that spendconsiderable resources monitoring their invest-ment portfolio. Thus, our model goes someway in addressing the questions of why privatemoney takes the contractual form it does, aswell as why private money is typically suppliedby banks (as opposed to other types of privateagencies). In our model, money and bankingare inextricably linked.

Of course, we are not the first to explicitlymodel money and banking together. Someimportant recent contributions include Kiyotakiand Moore (2000), Bullard and Smith (2001),and Cavalcanti (2001). We view our work ascomplementary to these papers. In Kiyotakiand Moore (2000), banks are agents endowedwith some sort of commitment technology thatallows their liabilities to circulate. In Cavalcanti(2001), banks have verifiable histories butnon-banks have not; as in Kiyotaki and Moore(2000), this special feature of banks allows theirliabilities to circulate. In Bullard and Smith(2001), the pattern in which agents meetendows bank liabilities with relatively lowtransactions costs, making these instruments thepreferred medium of exchange. Our setup issimilar to that of Diamond (1984), whichemphasizes the role of monitoring in the busi-ness of banking; banks are endowed with nospecial characteristics relative to other agentsin the economy. Under some circumstances, itmakes sense to have the economy’s monitoringagencies (banks) issue the medium of exchange.

I. A Simple Model of Money

The Physical Environment

Consider an economy with four periods,indexed by t =0, 1, 2, 3. Period 0 is interpretedas a “contracting period” (no consumption orproduction takes place) where individualsmay (or may not) meet to trade in securities. Insubsequent periods, goods are produced andconsumed, and spot markets may (or may not)open. The economy is populated by a largenumber (3N ) of individuals who have prefer-ences defined over deterministic, time-dated

consumption profiles (c1, c2, c3). Individuals arespecialized in the production of a nonstorabletime-period good (y1, y2, y3). In particular, weassume that there are three types of individualsand N individuals of each type: Person A pro-duces y3; person B produces y1; and person Cproduces y2. One interpretation of this setup isthat type A (C ) individuals are endowed with along- (short-) term capital project. Let us alsoassume (for simplicity only) that people havespecialized preferences: Person A wants c1;person B wants c2; and person C wants c3(assume also that each person values his owngood just a little bit). The chart below describes,for each individual type, the goods that hedesires, ci , and the good that he produces (isendowed with), yj .

A B CGood 1 c1 y1

Good 2 c2 y2

Good 3 y3 c3

Note the complete lack of double-coincidence of wants: Any bilateral pairing ofindividuals will result in no exchange of goods.

Trade occurs at a centralized location that isaccessible by all agents in each period. Inperiod 0, the only objects available forexchange are claims (contracts) against y1, y2,and y3. In an environment where such claimscan be costlessly exchanged and enforced, amarket needs to open only once (in period 0).In this period, A sells a claim to y3 and pur-chases a claim to y1, B sells a claim on y1 andpurchases a claim on y2, and C sells a claim ony2 and purchases a claim on y3.

IOUAy3

IOUCy2

A C

B

IOUBy1

IOUCy2

IOUAy3

y1

IOUBy1

y2

y3

Figure 1

22

Figure 1 characterizes the various trades thatoccur in equilibrium. Period-0 trades are denotedby the inner straight-line arrows and the vari-ous time-dated trades are given by the outercurved-line arrows. We denote a claim issuedby agent j for output produced at time t asIOU j yt . As time unfolds, previously agreed-tocontracts are simply executed, that is, no furthertrades occur. This “Arrow–Debreu” marketdelivers an equilibrium allocation that is Paretooptimal without the aid of anything that onemight identify as “money” in the model.

Limited Commitment and MonetaryExchange

Consider now an environment in which notall individuals can commit to keeping theirpromises. In particular, suppose that only typeA agents can credibly commit to honoringclaims against their anticipated earningsstream, y3. In this case, the market value ofboth B ’s and C ’s securities as of period 0equals zero (since these securities representunenforceable claims against y1 and y2, respec-tively). At first blush, one might be inclined tothink that financial markets could break downcompletely. After all, B (C ) can acquire claimsto y2 (y3) only by selling his claims to y1 (y2);but if these latter claims are worthless, then B (C ) will be unable to purchase any claimsto y2 (y3).

In fact, if spot markets open up after period0, then the Pareto optimal allocation can beimplemented with the following sequence of

trades: In period 0, “nothing” happens. Inperiod 1, agent A sells his claim y3 to agent Bin exchange for y1. While agent B does notvalue y3 directly, he is nevertheless willing toaccept the security as payment, anticipatingthat he will be able to resell it in the future forsomething he does value. When the marketreopens in period 2, agent B is in a position topurchase y2 directly (instead of trying to col-lect on a previously negotiated claim to y2; thispurchase can be made with the security issuedby agent A. Agent C is willing to accept theclaim against agent A’s output because C valuesy3 and the claim against y3 can (by assumption)be enforced. Figure 2 summarizes the varioustrades.

Notice what has happened here. In effect,agent A has issued a debt instrument entitlingthe bearer to the output generated in period 3by agent A. Since agent A can commit to keep-ing his promises, this bearer-bond will circulateas a medium of exchange; in other words,agent A’s security can be properly identified as“money” in this model economy.

The money that arises in this model takesthe form of a circulating private debt instru-ment, redeemable in some form of good orservice. As such, it may appear somewhatremoved from most modern-day monies, whichprimarily take the form of either government-issued fiat currencies or privately-issued debtinstruments that are redeemable in governmentfiat (such as demand deposits). However, thereare, in fact, several instances of privately-issuedmonies that are redeemable in goods or ser-vices. For example, the Canadian Tire Corpora-tion has for many years issued small-denomina-tion paper notes (referred to as Canadian Tiremoney) redeemable in a wide array of storeproducts; in small communities, these noteshave been known to circulate as a medium ofexchange. The Ithaca hour is a privately-issuedmonetary instrument that circulates quitewidely in Ithaca, New York; these notes, invarious denominations, are meant to beredeemable in the labor services of local resi-dents. As well, if one were to interpret y3 asgold, then history offers innumerable examplesof “gold-backed” monetary instruments.2

■ 2 Some of the earliest forms of paper money possessed this feature.In the sixteenth century, for example, merchants would deposit their gold invaults rented from goldsmiths. Apparently, the receipts issued by the gold-smiths (representing claims against the gold in the vault) began to circulateas a means of payment (primarily among merchants); see Smith (1936).}

IOUAy3

IOUAy3

A C

B

y1

IOUAy3

y2

y3

Figure 2

23

The basic lesson here is that money isneeded to facilitate some trades because not allindividuals have the ability or willingness tocommit to their promises (as John Moore hascleverly remarked, evil is the root of all money).The institutions that do arise to supply moneywill be those that have an ability, eitherendowed or manufactured, to make crediblecommitments. By issuing a debt instrumentdesigned to circulate as a means of payment,the supplier of money is, in a sense, renting hiscommitment power to those who lack it.Specifically, even though individual B lackscommitment, he is able to purchase good y2by virtue of the fact that individual A cancommit to promises.

One might legitimately ask where agent A’scommitment power comes from and whyothers seem to lack it. The model of Kiyotakiand Moore (2000) provides some foundationsto the structure of commitment power thatdepend on: 1) the existence of multiperiodinvestment projects; 2) the ability of initialcreditors and debtors, if given the opportunity,to conspire against a third party who purchasesexisting debt; and 3) an institution, called abank, that by design cannot conspire againstanyone. In that environment, it is the bank’sliability that circulates in the economy. Below,we provide a different foundation that is basedon asymmetric information and monitoringactivities, instead of asymmetric distribution ofcommitment power.

II. Money and Banking

In this section, we modify the physical envi-ronment described above in a few simple ways.To begin, assume that all individuals are identi-cal in terms of their willingness and/or abilityto commit to their promises and that commit-ment is feasible only up to what is verifiable(for example, through observation by a third-party enforcement agency).

Assume that there is now some risk associ-ated with the endowment of each agent; inparticular, for t = 1, 2, 3,

y with probability 1 – �yt = { 0 with probability � .

In addition, suppose that there is noaggregate risk, so that (1 – �)Ny represents theaggregate output in each period. We will con-tinue to assume that individuals are risk-neutral.

The structure of information is as follows:Each person has the ability to costlesslyobserve the return realized on his own “project.”Other agents are also in a position to observethis return, but only at a utility cost equal to �;think of this cost as representing the effortexerted in monitoring project returns. Thissetup is similar to that of Diamond (1984),except that we will assume that if an agent ismonitored, the information revealed becomesa matter of public record.3

Arrow–Debreu Securities

The type of securities that will be exchangedon this market are contracts that promisedelivery of a good in the event that returns arereported to be positive. Since it will always bein the interest of the person who issues asecurity to report zero output (we assumethat people cannot commit to tell the truth),it has to be understood that the holder of anysuch security will, in equilibrium, monitorproject returns.

Clearly, for any kind of trade to occur,agents must have an incentive to purchaseclaims from other agents and then to monitorthem. If the marginal utility of state-contingentconsumption is constant (and equal to unity),then the parameter restriction (1– �)y > � issufficient to guarantee that trade and monitor-ing will occur. As in the previous section,period 0 trades are as follows: A sells a contin-gent claim to y3 and purchases one on y1;B sells a contingent claim on y1 and purchasesone on y2; and C sells a contingent claim on y2and purchases one on y3. In period 1, agent Awill monitor agent B ; in period 2, agent B willmonitor agent C ; and in period 3, agent C willmonitor agent A. Note that the total (economy-wide) monitoring costs are 3N � . Figure 3summarizes the various trades and monitoring.

■ 3 This assumption is made primarily to simplify the exposition; itdoes not affect our main conclusions.

Monetary Exchange

Instead of trades in Arrow–Debreu securities,imagine that trades occur in a sequence of spotmarkets with the help of a circulating privatedebt instrument issued by type A agents. Theprivate debt instrument issued by A is a contin-gent claim, as described earlier. Assume thatproject returns are realized at the beginningof each period and that a spot trade of moneyfor goods occurs after the period’s risk hasbeen resolved.

The sequence of trades is as follows: Inperiod 1, just after project returns are realized,type A agents offer their security to anyonewho actually has y1 to sell (as opposed to acontingent claim against y1 that they wouldhave purchased before realizing projectreturns). In the context of this environment,the people who are in a position to approachtype A agents are the “successful” type Bagents. It is important to note here that underthis scenario, successful type B agents cancostlessly reveal their success by the very actof displaying the goods they have to trade; inother words, there is no need for type A agentsto monitor.

Successful type B agents are willing toaccept a type A security as payment becausethey anticipate being able to use this securityas payment for future goods that they desire.In particular, following the resolution of riskin period 2, a type B agent can purchase y2directly from a successful type C agent. Again,there is no need for monitoring. Type C agentswillingly accept type A securities as paymentbecause they represent direct claims against

24

the goods that they desire. In period 3, eachtype C agent with a claim against y3 will pre-sent the claim for redemption.

In order for any claim against A to beenforced, a monitoring expense must beincurred. Notice that while all N type A agentshave issued securities, these securities end upbeing held by only (1–�)N type C agents.Consequently, each type C agent will holdclaims for y3 that were issued by different typeA agents. To avoid coordination and monitoringproblems, it makes sense here to appoint a“designated monitor,” that is, to let one (arbi-trarily chosen) type C agent set up a “monitoringbusiness” that agrees to monitor a type A agentin exchange for some fraction � of the projectreturns.4 Since there are N projects that requiremonitoring (assume that projects cannot bemonitored sequentially), the monitor incurs atotal cost N�. Assuming free entry into themonitoring business, the equilibrium monitoringfee � must adjust to ensure zero net returns tomonitoring; thus (1–�)N�y =N� or

� �*

=(1– �)y

Under this “monetary regime,” the expectedutility payoff for agents A and B is equal to(1 – �)y , which clearly exceeds the payoff theywould have generated under the Arrow–Debreumarket structure: (1 – �)y –�>0. For each typeC agent (including the monitor), the expectedpayoff is identical to what he would have gen-erated under the Arrow–Debreu market struc-ture. Consequently, we see that monetaryexchange dominates trade in state-contingentsecurities by economizing on aggregate moni-toring costs; that is, N� < 3N � . Figure 4 sum-marizes the various trades and monitoring.The C agent who does all of the monitoringis denoted by C M. Note that in Figure 4, C M

receives �* from another C agent only if that C

■ 4 For values of λ less than half, the typical successful type Cagent will hold one plus some fraction of type A IOUs. A type C agentwill not have an incentive to monitor the type A agent for which he holdsa fraction of an IOU if the fraction is sufficiently small: In this situation,the monitoring cost exceeds its expected benefit. If all type C agentswho hold a fraction of an IOU issued by the same type A agent cannotsomehow co-ordinate their monitoring activities, then this type A agentwill not be monitored. But if a type A agent is not monitored, there willbe a misallocation of resources. A designated monitor can overcomethese coordination problems.

A C

B

IOUBy1

IOUCy2

IOUAy3

y1

y2

y3

Figure 3

IOUAy3

+ monitor

IOUCy2+

monitor

IOUBy1+

monitor

25

agent’s claim pays off. In an attempt to reduceits complexity, figure 4 does not depict thepotential exchange of y3 for a claim against y3between agent C M and an A agent.

Banking

The type of monitoring activity describedabove is commonly regarded as an importantpart of the business of banking. But banks arealso associated with the business of creatingliquidity (nowadays in the form of transactiondeposits, but historically also in the form ofpaper money), which is injected into the econ-omy by way of money loans (as well as wageand dividend payments). In the model described above, the money creation andmonitoring activities are undertaken by sepa-rate sets of agents; in reality, these activitiesappear to be bundled. What might account forthis bundling?

The first thing to note is that our model isnot necessarily inconsistent with the fact thatmoney creation and monitoring activities arebundled (although the model does not neces-sarily point to bundling as the unique organi-zational form either). We could, for example,imagine that the monitoring agent also decidesto take on the responsibility of creating theeconomy’s monetary instrument. In this case,such an agent would more closely resemblewhat is commonly called a bank. So let usconsider what happens when the monitoringagent also issues money (for example,banknotes).

Trading activity proceeds as follows: Inperiod 0, A agents approach the bank for amoney loan. Suppose that each type A agentborrows (M/N) banknotes, which he promisesto pay back in period 3 (if his project return ispositive) with interest equal to R ; that is,(1+R)(M/N) represents the principal and inter-est that is to be paid back in the form of ban-knotes. Agent A then takes these banknotesand uses them to purchase y1 from type Bagents with output to sell. Since M “dollars”are exchanged for (1–�)Ny units of output, thefirst-period price level is

1 MP *1 = [(1–�)y ] N

.

In period 2, each B agent with moneypurchases the output displayed for sale bysuccessful type C agents; the equilibrium pricelevel remains the same, that is, P *2 = P *1 .

Now, in period 3, type C agents who holdbanknotes will want to purchase the outputproduced by successful type A agents. Thequestion here is whether a type A agent is will-ing to give up a good that he values (slightly)in exchange for paper that he does not valueat all. To give an A agent the incentive tobehave “properly,” the initial money-loan con-tract must contain a clause that transfers prop-erty rights over project returns from A to thebank in the event of default. In effect, themoney loan is collateralized with securities thatconstitute contingent claims against y3. Asbefore, all type A agents must be monitored.Consequently, it will do no good for a success-ful type A agent to claim failure. At this stage,the agent has the choice of either selling hisoutput for the banknotes that he needs to payback the money loan; or of having his output“seized” by the bank (which owns an enforce-able contingent claim against it). Either ofthese options leaves a type A agent with thesame payoff, so the agent might rationallychoose either one. A third possibility is that anA agent might renegotiate the terms of the loancontract, leaving both the bank and himselfbetter off (at the expense of type C agents). Toprevent either outright default or renegotiation(avoiding both would be necessary for thebanknote to circulate in the first place), themonetary instrument must be include aredemption clause: The note-bearer must havethe right to redeem the banknote for output.5

■ 5 A similar redemption clause appears to be embedded withinmodern-day private monetary instruments (for example, deposits inchecking accounts are typically redeemable in government cash).

IOUAy3

IOUAy3

A C

B

y1

IOUAy3

y2

y3

Figure 4

CMmonitor �*

26

Finally, if R >0 (as must be the case), then itappears that there are not enough banknotesin circulation: How can A agents acquire themoney needed to pay off a debt equal to(1+R)M when there are only M dollars in cir-culation? It turns out that some “new money”(RM dollars) must be injected into the econ-omy in period 3 by the bank itself. That is, thebank simply prints up RM dollars of newmoney, which can be used to purchase someperiod-3 output to compensate the bank for itsmonitoring services. With free entry in thebanking business, the interest rate charged bythe bank must result in zero profits; so

R *M = P *3N µ .

What prevents a bank from “overissuing”money at this stage? We have to assume thatthe supply of banknotes is verifiable (that is,the bank’s balance sheet can be observed by acourt and cannot be falsified). Consequently,the bank will be bound to charge a maximuminterest of R * and will not be in a legal posi-tion to inject more than R*M of new moneyinto the economy. Now, in order to derive R *,we need an expression for the price level inperiod 3. Since (1+R )M dollars are exchangedfor (1–�)Ny units of output, the price levelmust satisfy the condition

(1+R *)MP *3 =

(1 – �)Ny .

Combining these latter two expressions, wecan solve for the equilibrium interest rate andprice level:

µR* =

(1–�)y – µ> 0 ;

1 MP *3 = [(1–�)y – µ ] N

> P *1

= P *2.

Notice that period-3 inflation (which isfully expected) has served to diminish thepurchasing power of C’’s money holdings; butin equilibrium, this loss is exactly equal to theamount of purchasing power the type C agentwould willingly have transferred to a profes-sional monitor (as demonstrated in the earlierscenario). The various trades and monitoringare depicted in figure 5. The C agent who isthe bank is denoted as C B.

Transaction Costs

The equilibrium allocation associated with this“banking regime” corresponds to the allocationthat resulted when money creation and moni-toring activities were performed by separatesets of agents. Strictly speaking, the modelhere is unable to pin down the banking regimeas a unique organizational form.

However, suppose that we follow Bullardand Smith (2001) and extend the modelslightly by assuming that every time a good“changes hands,” a small fraction of it, 0 < ε < 1, disappears. One can think of ε asbeing a small transaction cost. Let us nowcompare the total transactions costs whenmoney creation and monitoring are bundledrelative to when they are not.

A C

B

y1

$y

2

y3

Figure 5

$

$

CB

$

(1+R) $

R $ + monitor

y3

27

When money creation and monitoring areunbundled, each type A agent issues a securitythat circulates which is ultimately monitored byone of the type C agents. Goods produced inperiods 1 and 2 change hands once, so thetotal transactions costs in these periods is 2(1–�)N εy. At date 3, however, some of thegoods produced will change hands twice:There is a set of transactions between C agentswho possess A’s security and A agents whoproduced output; there is also a set of transac-tions between C agents who now possessgoods and the monitor who requires paymentfor his services. The transactions costs incurredin period 3 are (1–�)N (1+�*) εy if the monitordid not possess any securities issued by A agentsand (1 –�)N (1+�*) εy – �* εy if he did.Hence, total transactions costs for the economyare 3 (1–�)N εy +N (1+�*) εy –� �* εy, where� = 0 if the monitor did not possess anysecurities issued by A agents and � =1 if hedid (recall that whether the monitor ends upholding A’s security depends on whether hisproject is successful).

In contrast, when money creation and mon-itoring are bundled, goods change hands onlyonce at each date. In particular, at date 3 thebank’s compensation for its monitoring coststakes the form of printing money and directlypurchasing goods from type A agents. Hence,under this regime, total transaction costs are3 (1–�)N εy, which is strictly less than the totaltransaction costs associated with the latterarrangement. Hence, when there are transac-tion costs associated with the exchange ofgoods, a banking regime—an institutionalsetup in which monitoring and money creationare undertaken by the same agent—will Paretodominate an environment in which moneycreation and monitoring activities are per-formed by separate agents.

Conclusions

We have constructed an environment in whichsomething that looks like a bank emerges asan efficient exchange mechanism. A bankmonitors projects and issues money that circulates as a medium of exchange. Whenindividual transactions are costly, the bankinginstitution turns out to be an efficient tradingmechanism because goods are only exchangedbetween the initial seller and the final con-sumer. An economy that has private (nonbank)securities circulating will have some fraction offinal output exchanging hands more thanonce; as a result, its transaction costs will behigher than those of a banking economy.

28

References

Bullard, J., and B. Smith. “Intermediariesand Payments Instruments,” unpublishedmanuscript, 2001.

Cavalcanti, R. “A Monetary Mechanism forSharing Capital: Diamond and Dybvig MeetKiyotaki and Wright,” unpublished manu-script, 2001.

Diamond, D. “Financial Intermediation andDelegated Monitoring,” Review of EconomicStudies, vol. 51 (1984), pp. 393–414.

Kiyotaki, N., and J. Moore. “Inside Moneyand Liquidity,” unpublished manuscript,London School of Economics, 2000.

Smith, V.C. The Rationale of Central Banking.Westminster, England: P.S. King and Son,Ltd., 1936.