The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
January/February 1990 Vol. 72, N o .l 3 M a r k e t D iscipline o f Bank Risk: T h e o r y and E viden ce 19 O n the Use o f O p tion P ric in g M o d els to A n a ly z e Deposit In su ran ce 35 W h a t D o W e K n o w A b o u t the Long-R un Real E xch an ge Rate? t iii: iin u u i A RESERVE RANK of A r S T .I/H IS 1 F e d e ra l R e s e rv e B a n k o f St. L o u is R e v ie w January/February 1990 In This Issue . . . Policymakers in the federal governm ent are considering various changes in deposit insurance in response to the large number o f bank and thrift failures in recent years and the associated large losses by the insurance funds. Some proposed reform s w ould reduce the govern m ent’s insurance coverage to increase the effectiveness o f market forces in limiting the risk that banks assume. Other reform s w ould base in surance premiums on the risk that banks assume. In the first article in this Review, "M arket Discipline o f Bank Risk: Th eory and Evidence,” R. Alton Gilbert investigates the implications o f deposit insurance reform proposals based on market discipline. Gilbert uses a simple theoretical exercise to illustrate how market forces could limit the risk assumed by banks under different approaches to reform ing deposit insurance. He also summarizes the empirical studies o f the effectiveness o f market discipline to determ ine w hether market forces actually could be expected to limit such risks. Under the current arrangements, the cost o f resolving bank and thrift failures is borne largely by the taxpayer through the federal deposit in surance agencies. In the policy debate, which considers the flaws and potential alternatives to the present system, a num ber o f economists have utilized a set o f theoretical tools — called option pricing models — fo r analytic purposes. In the second article o f this issue, "On the Use o f Option Pricing Models to Analyze Deposit Insurance,” Mark D. Flood outlines the basic theory o f option pricing, which was originally developed to assign dollar values to the option contracts traded on financial exchanges. The author then illustrates how an option model can be usefully em ployed to analyze the claims o f bankers, depositors and insurers on the assets o f a bank or thrift by applying the model to several insurance arrangements. Finally, Flood considers some o f the limitations o f this approach. * * * Exchange rates have been at the center o f numerous economic policy discussions in the 1980s. In the third article in this Review, “ W hat Do W e Know About the Long-Run Real Exchange Rate?” Cletus C. Coughlin and Kees Koedijk review what is known about movements in one type o f exchange rate, the long-run real exchange rate. Despite much research, there is no consensus on which variables cause changes in the real ex change rate. Coughlin and Koedijk review the literature to provide an elem entary understanding o f the three prim ary approaches and the variables thought to influence the long-run real exchange rate. Using a data set covering the current floating-rate period, the authors compare the three approaches empirically and conclude that none provides an adequate explanation o f movements in the long-run real exchange rate. JANUARY/FEBRUARY 1990 3 R. Alton Gilbert R. Alton Gilbert is an assistant vice president at the Federal Reserve Bank of St. Louis. David H. Kelly provided research assistance. Market Discipline of Bank Risk: Theory and Evidence D ECAUSE o f the many failures o f banks and thrift institutions in recent years and the high cost o f liquidating or reorganizing the bankrupt savings and loan associations, policymakers are now considering major changes in the w ay they supervise and regulate depository institutions in the United States. The Financial Institutions Re form, Recovery and Enforcem ent Act o f 1989, which provides the funds fo r closing bankrupt savings and loan associations (S&Ls), calls fo r several governm ent agencies to study the issues involved.1 The federal budget document fo r fis cal 1991 discusses the basis fo r reform o f deposit insurance and the advantages o f various reform s.2 To some extent, the unusually high failure rate o f depository institutions (hereafter called banks) can be attributed to developments in the econom y such as declines in the prices o f oil and farmland in the early 1980s. Some studies conclude that fraud and mismanagement ac count fo r many o f the bank failures.3 The gen eral consensus, how ever, is that deposit insur ance creates an incentive fo r banks to assume higher risk than they w ould without it. Such risk may be gauged in terms o f the variance o f 'Title X of the act directs the Secretary of the Treasury and the Comptroller General, in consultation with various federal government agencies and individuals from the private sector, to prepare reports on issues related to the reform of deposit insurance, including the implications of policies that would enhance the effectiveness of market discipline. a bank’s return on assets as a percentage o f its capital. The logic that underlies this consensus is that without deposit insurance, banks that choose portfolios o f assets with higher variance in their rates o f return, or low er ratios o f capi tal to total assets, w ould have to pay higher in terest rates on deposits. Deposit insurance blunts this penalty. The relatively high failure rate and losses o f the deposit insurance funds reflect, to some extent, the banks' response to the incentives to assume risk created by deposit insurance. Thus, a major issue in the debates over financial reform is the future role o f de posit insurance. Some recent proposals to reform deposit in surance are designed to increase the effective ness o f market forces in reducing the risk assumed by banks. Under these proposals, bank owners and creditors w ould be exposed to larger losses if their banks fail. The theory is that if bank ow ners and creditors have greater exposure to losses, they w ill limit the risk as sumed by their banks. In some proposals, this influence w ould complement the efforts o f bank supervisors. In others, market discipline would replace governm ent supervision. 2Budget (1990), pp. 246-53. 3Graham and Horner (1988) and Office of the Comptroller of the Currency (1988). JANUARY/FEBRUARY 1990 4 This paper describes some o f these proposals fo r enhancing the effectiveness o f market disci pline and illustrates how they w ould affect the banks’ incentive to assume risk. The paper also examines the empirical evidence on the effective ness o f market discipline. Proposals fo r the re form o f deposit insurance that rely on market discipline assume that market participants can differentiate among banks on the basis o f risk, and that market yields on bank debt reflect that risk. Th e paper lists the results o f several em pirical studies and draws conclusions about the potential effectiveness o f these proposals in re form ing deposit insurance. THE OBJECTIVES OF D EPO SIT INSURANCE Various approaches to enhancing the e ffe c tiveness o f market discipline o f bank risk are presented in table 1. Choosing one approach over another depends in part on which basic objective o f deposit insurance is considered to be most important. The follow ing are the prim ary objectives o f deposit insurance: 1. T o protect depositors with small accounts, 2. T o prevent widespread runs by depositors on banks, and 3. T o protect the insurance fund from losses that w ould bankrupt it.4 Th ere are tradeoffs among these objectives. The policy that provides the greatest protection against runs by depositors is complete coverage o f all deposit accounts. That policy, how ever, eliminates any incentive fo r depositors to exert their discipline over the risk assumed by their banks, leading perhaps to an increase in the in surance fund's losses. The dollar limit on the amount in each ac count that is insured, currently $100,000, reflects an attempt to balance these objectives. Total coverage o f accounts less than $100,000 protects small depositors. The limit on the in surance coverage per account is designed to in duce the depositors w ith large accounts to m onitor their banks and require that riskier banks pay higher interest rates on their de posits. Those w ith relatively large accounts are assumed to be better able to impose such m ar ket discipline. Th e limit on insurance coverage “ Federal Deposit Insurance Corporation (1983), pp. viii-xiii. Digitized forFEDERAL FRASER RESERVE BANK OF ST. LOUIS Table 1 Proposals to Increase the Effectiveness of Market Discipline of Bank Risk______________________ (1) Phase out federal deposit insurance to facilitate the development and use of private deposit insurance. Short and O ’Driscoll (1983), Ely(1985), England (1985) and Smith (1988). (2) Lower the ceiling on federal insurance coverage per account. Council of Economic Advisers (1989), pp. 203-4. (3) Co-insurance: limit federal deposit insurance to some fraction of each account. Boyd and Rolnick (1988). (4) Place a ceiling on federal deposit insurance per in dividual at all depository institutions. England (1988). (5) All institutions must maintain subordinated debt liabilities that are some fraction of their total assets. Cooper and Fraser (1988), Keehn (1989) and Wall (1989). (6) Early closure: close or reorganize depository institu tions when their capital ratios, reflecting the market value of assets and liabilities, are low but still positive. This proposal is designed to enhance the effectiveness of market discipline by closing or reorganizing banks whose shareholders have weak incentive to limit risks. Benston and Kaufman (1988). per account, how ever, tends to undermine the objective o f preventing runs by depositors on the banking system. O f course, a run by depositors on an in dividual bank does not create a serious problem fo r the banking system, because these depositors simply rem ove their cash from one bank and deposit it in another in which they have m ore confidence. If the bank subject to the run can not meet its depositors’ demand fo r currency, it will have to close. Its depositors w ill be paid as the assets o f the failed bank are liquidated. A run on a bank can serve a useful purpose—a mechanism fo r closing a bank in which deposi tors have lost confidence. Runs become a problem fo r the banking system, however, w hen depositors w ithdraw currency and, hence, reserves from banks as a group. Banking history in the United States and 5 the United Kingdom prior to their central banks acting as lenders o f last resort indicates that runs on banking systems have occurred, al though they tended to be separated by many years.5 Some argue that deposit insurance is not necessary to avoid the adverse social effects o f banking system runs. They maintain that, as long as the central bank acts as an effective lender o f last resort, the liquidity it provides would limit any damage that runs on individual banks could do.6 An alternative vie w emphasizes the dangers o f relying on the central bank to operate as the lender o f last resort to a banking system without deposit insurance. A central bank might respond inappropriately in a finan cial crisis, as the Federal Reserve did in the ear ly 1930s, leading to rapid declines in the assets o f the banking system and widespread bank failures. Deposit insurance reduces the role o f the central bank in maintaining stability in the operation o f the banking system. Thus, the choice among the potential reform s o f deposit insurance rests on view s about the vulnerability o f the banking system to runs and the effe c tiveness o f a lender o f last resort in dealing w ith runs. I f the prim ary objectives o f deposit insurance are to protect small depositors and to protect the insurance funds from large losses, a logical change w ould be to reduce the insurance cover age per account. This was proposed by the Pre sident’s Council o f Economic Advisers in 1989. Those w h o consider the possibility o f banking system runs a serious threat to the stability o f the banking system w ould oppose a large reduc tion in the insurance coverage on bank deposits. THE EFFECTS OF REFORM P R O PO SALS ON BA N K IN G RISK The reform proposals are designed to reduce the incentives fo r banks to assume risk. In evaluating their effectiveness, it is useful to con sider three indicators o f the banking system’s perform ance that reflect this risk: the expected loss by depositors due to the bank failure, the 5Gilbert and Wood (1986) and Dwyer and Gilbert (1989). 6See Kaufman (1988) and Schwartz (1988). 7See Bernanke (1983), Calomiris, Hubbard and Stock (1986), Grossman (1989), and Gilbert and Kochin (1989). 8To illustrate the need for such a model, consider a basic reform of eliminating deposit insurance. That change would increase the interest expense of a bank with a given expected loss o f the Federal Deposit Insurance Corporation (FDIC), and the probability that a bank w ill fail. The expected loss by depositors and the FDIC are considered separately since proposals that reduce the FDIC’s expected loss tend to increase the expected loss by depositors. Focusing on only one o f these measures o f per form ance misses some o f the reform proposals’ implications. The third measure, the probability o f bank failures, is o f interest because o f evi dence that bank failures have adverse effects on economic activity in addition to the wealth losses b y depositors and ow ners.7 The studies that find adverse effects o f bank failures on economic activity attribute those effects to the constraints on the availability o f credit created by bank failures. Th e proposals’ implications fo r the effective ness o f market discipline can best be derived by using a m odel o f the behavior o f banks and their creditors.8 Nature o f the M od el The implications o f the various reform pro posals are derived by examining the effects o f proposed changes in deposit insurance on the optimal choice o f risk by a representative bank er. Several assumptions are made to simplify the model. R a te o f R e t u r n o n A ssets — The only ran dom variable in the model is the rate o f return on assets o f the representative bank, which has the same probability distribution under all as sumptions about the nature o f deposit in surance. Bank regulators are assumed to deter mine the probability distribution o f the rate o f return by restricting the types o f assets the bank may hold. Th e only choice fo r the rep re sentative banker in this m odel is the level o f the bank’s total assets. The capital o f the bank is held constant at $100 in each case. W ith a given level o f capital and a given probability distribu tion o f the rate o f return on assets, the proba bility o f failure (losses exceeding capital) is posi tively related to the total assets o f the bank. Management is assumed to choose the level o f portfolio of assets, thereby tending to increase its pro bability of failure. The penalty of higher interest expense imposed by depositors on those banks that assume more risk would induce banks to assume less risk in their choice of assets. The net effect of eliminating deposit in surance on the probability of bank failure must be derived from a theoretical model that specifies the risk preferences of depositors and bank managers. JANUARY/FEBRUARY 1990 6 total assets that maximizes the expected profits o f the bank. W ith a given level o f bank capital, the condi tions under w hich the bank fails can be derived only with a specific probability distribution o f return on assets. This paper uses the discrete probability distribution presented in table 2.9 For each o f the seven possible outcomes, the rate o f return on assets is net o f the operating cost o f servicing the assets. The rate o f return associated with each out come is assumed to be inversely related to the size o f the bank’s total assets. One reason fo r this assumption o f an inverse relationship is that, as the bank increases its total assets, it must lend to b orrow ers beyond the local area in which it has some market power. Another reason is diseconomies o f scale in the operating cost o f servicing assets. For each outcome with a positive return on assets, therefore, the rate o f return falls as total assets increase. This feature o f the model yields a maximum ex pected profit fo r the bank under each assump tion about deposit insurance.10 B a n k C osts — For a given level o f total assets, the bank's cost depends on the insurance coverage on its liabilities. This paper considers the four cases described below. If, as in case A, all deposits are fully insured, the bank can at tract an unlimited supply o f deposits by paying the risk-free rate o f interest. Under each o f the 9The use of a discrete probability distribution, with a limited number of outcomes, makes the presentation simpler than if a continuous probability distribution was used. In using a discrete probability distribution, there is a trade-off be tween simplicity and continuity of the probability of failure with respect to leverage. The smaller the number of possi ble outcomes, the larger the jumps in the probability of failure at certain asset levels. Increasing the number of possible outcomes, however, increases the difficulty of il lustrating the calculations. Thus, the probability distribution in table 2 is arbitrary. 10The model abstracts from possible losses by our represen tative bank on the deposits it holds at other banks. If a reform proposal increases the probability of losses on in terbank deposits, the effects of such reform proposals on the probability of failure at our representative bank would be understated. This point about possible loss on interbank deposits is most relevant in comparing the case with no deposit in surance to the other cases examined below. The model in this paper is not modified to reflect directly the effects of possible losses on interbank deposits. This model also ignores losses from runs on the bank by its depositors in reaction to the failure of other banks. If deposit insurance coverage is reduced or eliminated, the FEDERAL RESERVE BANK OF ST. LOUIS four assumptions about deposit insurance, the costs o f servicing deposit accounts are offset by fees charged to depositors. For a given level o f total assets, the highest expense occurs in case B, w ith no deposit in surance. In this case, the interest rate that the bank must pay on deposits is positively related to its total assets. Depositors are assumed to be risk-neutral and to know the probability distri bution o f the bank’s return on assets. Hence, the bank must pay the rate to depositors that makes their expected return on deposits equal to the risk-free rate.11 The interest rate that the bank pays deposi tors is above the risk-free rate if the bank fails in at least one o f the seven possible outcomes. If it fails, the depositors receive the liquidation value o f the bank’s assets. Liquidation value in those outcomes reflects the probability distribu tion o f the bank’s return on assets. Th ere is no additional loss to depositors resulting from the elimination o f the bank as a going concern.12 Th e equation fo r calculating the rate paid to depositors in case B is presented in table 2. Cases C and D reflect tw o methods o f enhanc ing the effectiveness o f market discipline o f bank risk, w hile retaining some form o f deposit insurance. Co-insurance in case C limits deposit insurance coverage to 90 percent o f each de posit account.13 In case D, deposits are fully in sured, but each bank is required to keep its failure of some banks may induce depositors to run on other banks to receive currency in exchange for their deposits. Several such episodes occurred in the United States prior to the establishment of deposit insurance in the 1930s. See Dwyer and Gilbert (1989). To incorporate the possible effects of runs, the probability distribution of the return on assets at the representative bank would have to be specified as a function of the number of bank failures. 11The general nature of the comparisons among the four cases would not be changed if depositors were assumed to be risk averse. 12The assumption that a bank’s assets lose no value when they are liquidated results in an understatement of the ex pected loss of depositors and bank creditors in the various cases. A study by the Federal Deposit Insurance Corpora tion, Bovenzi and Murton (1988), reports that when the FDIC liquidates the assets of failed banks, their liquidation value averages about 70 percent of the book value of the assets of failed banks. 13Boyd and Rolnick (1988) suggest 90 percent coverage of deposits in their proposal for a form of co-insurance in deposit insurance. 7 Table 2 A Model of Bank Profits NET REVENUE Revenue of the bank, net of the operating cost of servicing assets, is a random variable with a discrete probability distribution. Let A be the assets of the bank. The probability distribution is as follows. Outcome number Net revenue 1 2 3 4 5 6 7 0.4(1 0.3(1 0.2(1 0.1(1 0.0(1 -0 .1 (1 -0 .2 (1 - A/10,000)A A /10,000)A A/10,000)A A /10,000)A A /10,000)A A/10,000)A A /10,000)A Probability 0.01 0.04 0.10 0.70 0.10 0.04 0.01 COST The cost of the bank is not a random variable. It is the same in each of the seven outcomes. Ex pected profits are calculated for four cases, each involving a different assumption about the in terest expense of the bank. In those outcomes in which the bank has a loss, the maximum loss to the shareholders is their investment of $100. Case A: All Liabilities Fully Insured The bank pays the risk-free rate of 8 percent on deposits, which equal A - $100. Case B: No Deposit Insurance At each level of total assets of the bank, the interest rate on deposits is set at the level that makes the expected return to holders of uninsured deposits equal to the risk-free rate of 8 percent. The interest rate on deposits in case B, for a given level of assets, can be derived by solving the following equation for R, the interest rate on deposits. To economize on notation, let A * = (1 - A/10,000)A. Then, 1.08(A - 100) = 0.01(1 +R)(A-100), or 0.01(1.4A*) if 0.4A* - R(A-100) < -1 0 0 + 0.04(1 + R)(A - 100), or 0.04(1,3A*) if 0.3A* - R (A -1 0 0 ) < -1 0 0 + 0.1(1 + R ) ( A - 100), or 0.1(1,2A*) if 0.2A* - R ( A - 100) < -1 0 0 + 0.7(1 + R ) ( A - 100), or 0.7(1.1A*) if 0.1A* - R(A -1 0 0 ) < -1 0 0 + 0.1(1 + R)(A - 100), or 0.1(A*) if - R(A - 100) < -1 0 0 + 0.04(1 + R ) ( A - 100), or 0.04(0.9A*) if -0 .1 A * - R (A -1 0 0 ) < -1 0 0 + 0.01(1 + R ) ( A - 100), or 0.01(0.8A*) if -0 .2 A * - R ( A - 100) < -1 0 0 JANUARY/FEBRUARY 1990 8 Table 2 continued A Model of Bank Profits_________________________________ Case C: Co-insurance The calculation of the stated interest rate on deposits as in case B is modified by setting the minimum return to depositors in each outcome at 90 percent of their principal plus stated interest. The equation for the interest rate on deposits is the same as that presented for case B, except that in each of the seven possible outcomes, the return to the depositors can be no less than: 0.9(1+R)(A - 100). Case D: Subordinated Debt Requirement The bank is required to have liabilities that are uninsured and subordinated to deposits, equal to at least 10 percent of its total assets. Deposits are fully insured. The interest rate on deposits is 8 percent. The interest rate on subordinated debt, for a given level of assets, can be derived by solving the following equation for R. 1.08(0.1)(A) = 0.01(1 +R)0.1A, or if 0.4A* - R(0.1A) - 0.08(0.9A -1 0 0 ) < -1 0 0 , 0.01 (the greater of 1.4A* - 1.08 (0 .9 A - 100) or zero) + 0.04(1 + R)0.1A, or if 0.3A* - R(0.1A) - 0.08(0.9A -1 0 0 ) < -1 0 0 , 0.04 (the greater of 1.3A* - 1.08(0.9A-100) or zero) + 0.1(1 +R)0.1A, or if 0.2A* - R(0.1A) - 0.08(0.9A - 100) < -1 0 0 , 0.1 (the greater of 1.2A* - 1 .0 8 (0 .9 A -100) or zero) + 0.7(1 +R)0.1A, or if 0.1A* - R(0.1A) - 0.08(0.9A - 100) < -1 0 0 , 0.7 (the greater of 1.1A* - 1.08(0.9 A - 100) or zero) + 0.1(1 +R)0.1A, or if - R(0.1A) - 0.08(0.9A - 100) < -1 0 0 , 0.1 (the greater of A* - 1.08(0.9A - 100) or zero) + 0.04(1 +R)0.1 A, or if -0 .1 A * - R(0.1 A) - 0.08(0.9A-1 0 0 ) < -1 0 0 , 0.04 (the greater of 0.9A* - 1.08(0.9A- 100) or zero) + 0.01(1 +R)0.1A, or if - 0.2A* - R(0.1A) - 0 .0 8 (0 .9 A -100) < -1 0 0 , 0.01 (the greater of 0.8A* - Digitized forFEDERAL FRASER RESERVE BANK OF ST. LOUIS 1.08(0.9A- 100) or zero) 9 subordinated debt liabilities equal to 10 percent or m ore o f its total assets.14 In cases C and D, the interest rates on bank liabilities also are set at levels that make ex pected returns to bank creditors equal to the risk-free rate. Equations fo r calculating the in terest rates on bank liabilities are specified in table 2. Other reform proposals are o f interest but are more difficult to incorporate into this simple model. For instance, m ore detail w ould be ne cessary to model the effects o f changing the deposit insurance limit per account or limiting FDIC coverage fo r each depositor to a given amount at all insured institutions. Case A: All Liabilities Fully Insured The bank maximizes expected profits in this case with total assets around $1800 (see figure 1). At this level, the bank fails (losses exceed the $100 o f capital) in outcomes 5, 6 and 7. Thus, the probability that the bank w ill fail is 15 per cent, based on the probability distribution fo r the return on assets in table 2. The FDIC’s ex pected loss, $14.26, is about 0.84 percent o f in sured deposits. The expected loss o f depositors, o f course, is zero. Case B: N o Deposit Insurance Several reform proposals call fo r phasing out deposit insurance (see table 1). W ith no deposit insurance, depositors lose part o f their principal plus interest if the bank fails (if losses exceed the $100 o f capital). The interest rate on de posits charged b y risk-neutral depositors is positively related to the bank’s total assets (see figure 2). The FDIC’s expected loss in this case is zero. The expected loss o f depositors w ith total assets equal to $1000 is $4.21, which is about one-half o f one percent o f deposits. Case C: Co-insurance Under the co-insurance option, federal deposit insurance coverage w ould be limited to a frac tion o f each deposit, w ith some low level o f each account fully covered to protect small de positors. Those w h o advocate co-insurance argue that the depositors subject to fractional coverage at the margin w ould monitor the risk assumed by their banks and demand relatively high interest rates on deposits at the banks that assume relatively high risk. T o simplify the illustration, all deposit ac counts are subject to the same percentage o f in surance coverage. In those outcomes in which the bank fails, payments to depositors under the co-insurance option w ould be the larger of: (1) the liquidation value o f the bank’s assets, or (2) 90 percent o f the principal plus interest on their deposits. Th e FDIC incurs a loss only if the bank fails and the liquidation value o f the bank is less than 90 percent o f the principal plus interest on deposits. As in case B, the market interest rate on deposits is set at the level that makes the ex pected return on deposits equal to 8 percent. The difference in this case is that depositors have the option o f receiving 90 percent o f their principal plus interest from the FDIC if their banks fail. Figure 2 indicates that fo r a given level o f total assets, the market interest rate on deposits is lo w er in case C than in case B, because in case C the losses o f depositors are limited by deposit insurance. The bank maximizes its expected profits with total assets equal to $1,000. It fails only in out comes 6 and 7. Thus, the probability that the bank w ill fail is only 5 percent, compared with a 15 percent probability o f failure associated w ith maximum profits in case A. Case B illus trates how a bank that maximizes expected pro fits can be induced to limit its probability o f failure through market discipline imposed by its creditors. Under the assumptions o f case C, the bank maximizes expected profits w ith total assets o f $1100. The bank must pay 8.44 percent to at tract the $1000 in deposits. W ith assets o f $1100, the probability o f the bank failing is 5 percent. The FDIC’s expected loss is only 0.072 percent o f insured deposits, about 9 percent o f 14Case D involves a higher percentage of subordinated debt to total assets than some of the proposals that call for subordinated debt requirements. For instance, the recent proposal of the Federal Reserve Bank of Chicago recom mends that banks be required to maintain a 4 percent ratio of subordinated debt to total assets. See Keehn (1989). The 10 percent requirement in case D is chosen to indicate that the degree of market discipline that can be imposed through a co-insurance proposal can be matched with a subordinated debt proposal. JANUARY/FEBRUARY 1990 10 Figure 1 Expected Profits Dollars Dollars Assets Figure 2 Expected Loss to FDIC Dollars Dollars Digitized forFEDERAL FRASER RESERVE BANK OF ST. LOUIS Assets 11 the loss rate fo r case A, w ith total assets at $1800. The expected loss to depositors is $5.09, which is about one-half o f one percent o f total deposits. From the FDIC’s perspective, there are tw o advantages o f co-insurance (case C) over full deposit insurance (case A). First, the bank chooses a level o f assets associated w ith a low er probability o f failure. Second, fo r a given level o f total assets o f the bank, the FDIC’s expected loss is lo w er under case C.15 Case D: Subordinated D ebt Requirem ent Some proposals fo r deposit insurance reform w ould require banks to issue subordinated debt liabilities that are not federally insured. The term "subordinated” refers to the status o f creditors o f a firm in bankruptcy. I f a failed bank is liquidated, those w h o hold subordinated debt w ould receive payments only if all deposi tors are paid in full. In case D, all deposits are fully insured by the FDIC. The bank, how ever, must have uninsured liabilities, which are subordinated to deposits, that equal at least 10 percent o f its total assets. The bank would choose to keep subordinated debt liabilities at the 10 percent minimum since, except at relatively low levels o f total assets, the interest rate on subordinated debt exceeds the risk-free rate paid on insured deposits. As in cases B and C, those w h o invest in subordinated debt are assumed to be riskneutral and know the probability distribution o f the net return on assets. Figure 3 presents the interest rate on subordinated debt as a function o f the total assets o f the bank.16 For most levels 15Co-insurance, however, has one disadvantage. A change from full coverage of insured deposits to co-insurance creates an incentive for depositors to run on banks in response to information (or rumors) about problems at banks. Even with the FDIC insuring 90 percent of the prin cipal and interest of deposit accounts, depositors have an incentive to avoid the 10 percent loss by withdrawing their deposits from a failing bank. Thus, in comparing cases A and C, co-insurance reduces the significance of deposit in surance in preventing runs on the banking system, placing greater responsibility on the role of the Federal Reserve in stabilizing the banking system in a financial crisis, as it functions as the lender of last resort. If there is some doubt that the Federal Reserve will execute its role as lender of last resort, co-insurance may be less advan tageous than full insurance of deposits. 16The humped pattern of the interest rate on subordinated debt for case D in figure 3 reflects the particular discrete probability distribution of returns on the assets used in this o f total assets, the interest rates on subor dinated debt is higher than the rates on deposits in the cases analyzed earlier because the expected loss is higher fo r those holding subordinated debt. I f the bank’s losses exceed the $100 investment o f the shareholders, holders o f the subordinated debt receive some payment only if the liquidation value o f the bank exceeds total deposits. The bank maximizes expected profits with total assets equal to $1100. A t that level, the bank must pay 12.12 percent on its subordi nated debt liabilities. The FDIC incurs losses only if the loss o f the bank exceeds the $100 capital o f the shareholders plus the subordi nated debt. Th e bank has a 5 percent probabili ty o f failure, and the expected loss o f the FDIC w ith total assets equal to $1100 is 0.06 percent o f insured deposits. Depositor losses are zero. Comparison o f Cases B, C and D A comparison o f risk in the operation o f the banking system under various assumptions depends on one’s assumption about the p ro bability o f runs on the banking system. I f this probability is assumed to be zero, the elimina tion o f deposit insurance (case B) induces banks to assume minimum risk. The FDIC’s expected loss is zero in this case, and the bank is induced by market forces to choose the lowest level o f total assets. One advantage o f the subordinated debt requirem ent over the other options is that, w hile the bank is induced to choose a level o f total assets below that in case A, the subor dinated debt is not subject to runs. Thus, the comparison o f risk betw een cases A and D does not depend on assumptions about runs on the banking system. paper. With a continuous probability distribution, or a discrete distribution with more possible outcomes, the plot of the interest rate as a function of total assets would have a less humped pattern. The fact that the interest rate on subordinated debt is higher at higher levels of total assets of the bank might in dicate a way in which the management of the bank could take advantage of those who invest in subordinated debt. The bank could issue some subordinated debt at a low level of total assets, at a relatively low interest rate, and then increase total assets and issue more subordinated debt at a higher rate. Investors in subordinated debt can protect themselves from such actions by insisting on covenants in the subordinated debt agreements that limit additional debt. If management of the bank violates such a covenant, the holders of the subordinated debt could go to court to make their debt instruments payable on demand. Restrictions on the issuance of additional debt are com mon in bond covenants. See Smith and Warner (1979). JANUARY/FEBRUARY 1990 12 Figure 3 Market Interest Rate on Bank Liabilities Rate Rate Assets Th e co-insurance option is not superior under any combination o f assumptions. I f the possibili ty o f runs on the banking system can be ruled out, there is a subordinated debt requirem ent that induces the same degree o f market disci pline o f banking risk as co-insurance. EM PIR ICAL STUDIES OF M AR KET DISCIPLINE OF THE RISK ASSUMED RY HANKS Market forces w ill be effective in constraining the risk assumed by banks only if investors can assess the relative degrees o f risk assumed by individual banks, and then set differential prices on the stock and debt instruments issued by 17The studies described in this section include only those based on data for individual banking organizations. Some studies cited in the literature estimate indices of returns on share prices or interest rates on bank liabilities for groups of banks as functions of aggregate data on banking risk. Such results are not relevant in determining whether par ticipants in the equity and debt markets can distinguish among the banking organizations, which would be necessary if market discipline of bank risk were to be effective. FEDERAL RESERVE BANK OF ST. LOUIS banks that reflect their inform ation about risk. The results o f the studies described in table 3 are relevant in evaluating the effectiveness o f market discipline. These studies estimate the in fluence o f measures o f risk assumed by banks on the stock prices o f banks and on the market interest rates on uninsured deposits and the subordinated debt o f banks.17 These studies do not test the hypothesis that banks adjust their risk in response to signals from the markets fo r bank stocks and debt.18 The Market f o r Bank Equity All but one o f these studies report evidence that is consistent w ith the hypothesis that stock prices are inversely related to the risk assumed 18Gendreau and Humphrey (1980) claim to have developed a model in which there is feedback from adverse signals in the bank equity market to bank leverage. It is difficult to see a feedback relationship between the stock price and leverage in this study, since the relationships among stock prices, leverage and other variables are estimated using contemporaneous observations. Estimating a feedback relationship from market signals to variables under the control of bank management would require dynamic relationships. 13 Table 3 Implications of Empirical Studies for the Effectiveness of Market Discipline of Bank Risk Authors Relationships estimated Results consistent with the effectiveness of market discipline Results MARKET FOR BANK EQUITY Beighley, Boyd and Jacobs (1975) Share prices of bank stocks estimated as a function of (1) capital ratios, (2) earnings and growth of earnings, (3) asset size, and (4) loss rates. Holding constant the influence of earn ings banks with higher capital ratios and lower loss rates tend to have higher share prices. Yes Pettway (1976) Betas for individual banks (a measure of risk derived from stock prices) esti mated as a function of the capital ratios of individual banks. The coefficient on the capital ratio is negative for one year but insignificant for other years. The negative coefficient on the capital ratio indicates that investors consider banks with higher capital ratios to be less risky. Yes Pettway (1980) For several large banks that failed, returns to shareholders are simulated for several years prior to their failure. Simulations are based on returns from holding stocks of large banks that did not fail. On average, returns on the stocks of banks that failed declined relative to simulated returns two years before failure. Yes Brewer and Lee (1986) Betas for individual banks are estimated as functions of ratios from balance sheets and income statements used by bank supervisors to reflect risk. Some of the measures chosen to reflect risk have positive, significant regres sion coefficients. Yes Cornell and Shapiro (1986) Returns to shareholders of 43 large banks are estimated as functions of the composition of their assets and liabilities in the years 1982-83. The percentage that Latin American loans was of total assets had a signi ficant, negative impact on returns in 1982. Energy loans had a negative impact in 1982-83. Loans purchased from Penn Square Bank had a negative impact on returns in the month in which that bank failed. Yes Shome, Smith and Heggestad (1986) Prices of bank stocks are estimated as a function of its earnings and capital ratios. The coefficient on the capital ratio is positive and significant for some years, insignificant for other years. Yes Smirlock and Kaufold (1987) Changes in stock prices of large banks at the time of the announce ment by Mexico in 1982 of its mora torium on debt payments as a function of the ratio of Mexican debt to equity capital at individual banks. Coefficient on the ratio of Mexican debt to equity capital is negative and sig nificant. Banks were not required to disclose their Mexican debt at the time of the 1982 moratorium. Yes JANUARY/FEBRUARY 1990 14 Table 3 continued Implications of Empirical Studies for the Effectiveness of Market Discipline of Bank Risk Authors Relationships estimated Results Results consistent with the effectiveness of market discipline MARKET FOR BANK EQUITY continued James (1989) and Cargill (1989) Returns on holding the stock of BHCs estimated as a function of the change in the market value of the BHCs’ loans to less-developed countries and dum my variables for individual banks and individual time periods. The change in the market value of loans to less-developed countries has a positive, significant coefficient which is not significantly different from unity. Yes Randall (1989) This is a case study of 40 BHCs that reported relatively large losses in the 1980s. For each BHC, a time period is designated when it began assuming relatively high risk and a time period when problems became public know ledge. Stock prices are compared to market averages before and after the problems became public knowledge. Stocks prices of the BHCs that re ported relatively large losses declined relative to market average stock prices only after the problems became public knowledge, not during the periods which the banks began assuming relatively high risk. No MARKET FOR UNINSURED DEPOSITS The interest rate on large denomination certificates of deposit is the dependent variable in each study. Crane (1976) Identifies the determinants of the CD rate using factor analysis. The factor that reflects profit rates and capital ratios is not a significant vari able in explaining the CD rate. No Herzig-Marx and Weaver (1979) Estimates CD rates as a function of variables used by bank supervisors to reflect risk. Of bank risk variables, only the liquidi ty measure has a significant coefficient. Capital and loss ratios have insignifi cant coefficients. No Baer and Brewer (1986) CD rate estimated as a function of variables used by bank supervisors to reflect risk, and separately, as func tions of level and variability of the prices of bank stocks. Coefficients on risk measures used by bank supervisors are not significant. Measures of the level and variability of stock prices help explain CD rates. No James (1987) The average interest rates paid by 58 large banks on their large denomina tion deposits are estimated as func tions of leverage, loan loss provision divided by total loans and the variance of stock returns. Each of these measures of risk have positive, significant coefficients. Yes These three variables have significant coefficients. CD rates tend to be higher at banks with more variable income and lower capital ratios, holding constant the influence of total assets. Yes Hannan and CD rate is estimated as a function of (1) Hanweck (1988) the variability of the ratio of income to assets, (2) the capital ratio and (3) bank assets. Digitized forFEDERAL FRASER RESERVE BANK OF ST. LOUIS 15 Table 3 continued Implications of Empirical Studies for the Effectiveness of Market Discipline of Bank Risk Authors Relationships estimated Results consistent with the effectiveness of market discipline Results MARKET FOR UNINSURED DEPOSITS continued James (1989) Interest cost on large CDs estimated as a function of risk measures: domestic loans/capital, foreign loans/capital and the loan loss provision/total loans. Interest cost positively related to the ratio of domestic loans to capital and the loan loss provision. The negative relation between interest cost and the ratio of foreign loans to capital is inter preted as evidence of an implicit government guarantee of foreign loans. Yes MARKET FOR SUBORDINATED DEBT: In each study the measure of the interest rate on the subordinated debt of banks is the rate on the subordinated debt minus the rate on long-term U.S. Treasury securities, called the rate premium. Pettway (1976) The rate premium is estimated as a function of the capital ratio of banks and other independent variables. The coefficient on the capital ratio is not significant. No Beighley (1977) The rate premium is estimated as a function of several measures of risk, including a loss ratio and a leverage ratio. The coefficients on the loss and lever age ratios are positive and significant. Yes Fraser and McCormack (1978) The rate premium is estimated as a function of the capital ratio and the variability of profits divided by total assets. Neither independent variable has a significant coefficient. No Herzig-Marx (1979) The rate premium is estimated as a function of several measures of risk assumed by banks. None of the risk measures have signi ficant coefficients. No Avery, Belton and Goldberg (1988) The rate premium is estimated as a function of risk measures derived from balance sheets and income statements and of the asset size of banks. Coefficients on the risk measures de rived from balance sheets and income statements are not significant. No Gorton and Santomero (1988) Use data in Avery, Belton and Goldberg (1988) to derive a measure of the vari ance of assets of banks implied by a contingent claims valuation model. The measure of the variance of assets is estimated as a function of the risk measures derived from balance sheets and income statements. Some of the risk measures derived from the balance sheets and income statements have significant coefficients. Yes JANUARY/FEBRUARY 1990 16 by banks, holding constant other determinants o f stock prices. Th e one study that concludes that stock prices do not reflect the risk assumed by banks, by Randall (1989), examines m ove ments in the stock prices o f bank holding com panies that reported relatively large losses in the 1980s. Randall concludes that these stock prices fell relative to the stock prices at other banks after their problems became common knowledge; how ever, they did not decline dur ing the periods w hen the banks w e re assuming the relatively high risk that led to losses. Ran dall concludes that the stock market does not discipline the risk assumed by banks, since the relative declines in bank stock prices did not precede public information on the consequences o f risk assumed by these banks. Randall’s study, how ever, has several w eak nesses. It is a case study, not a statistical study o f the determinants o f stock prices. The dating o f points at which problems became common knowledge is arbitrary; the choice o f such dates, how ever, determines the results. About half o f the cases involve banks in the Southwest W e w ould not expect relative declines in the stock prices o f these banks before the large decline in oil prices. W e cannot expect the par ticipants in the market fo r bank stocks to have greater foresight in predicting the decline in the price o f oil than the participants in the market fo r oil. T w o studies are particularly interesting in terms o f investors’ ability to differentiate among banks on the basis o f risk. Pettway (1980) com pares stock prices o f large banks that failed w ith simulated stock prices that w ere based on data from banks o f comparable size that did not fail. Returns to stockholders o f the failed banks declined relative to their simulated returns about tw o years b efore the banks failed. Rela tive returns o f the failed banks also declined before the bank supervisors put them on the problem bank list. Smirlock and Kaufold (1987) find that, w hen Mexico announced the m orato rium on its debt payments in 1982, the declines in the stock prices w ere proportional to the Mexican debt held by banks relative to the book 19This contrast can be illustrated using some recent studies and bank failure cases. Avery, Belton and Goldberg (1988) use observations for the 100 largest BHCs, which had total assets above $3 billion in 1985 and 1986. The total assets of the banks in the sample used by Hannan and Hanweck (1988) average $4 billion. In 1985 and 1986, 69 failed banks did not have their liabilities assumed by surviving FEDERAL RESERVE BANK OF ST. LOUIS value o f their equity capital. At the time o f the moratorium, banks w ere not required to dis close their loans to other nations. Nevertheless, investors appeared to have sufficient inform a tion, without such requirements, to make the appropriate adjustments to the prices o f bank stocks. The Markets f o r Uninsured Deposits and Subordinated D ebt The findings about the relationship betw een risk and interest rates on uninsured deposits and on subordinated debt are m ore mixed. Th ree o f the six studies o f bank CD rates report no evidence that higher CD rates are paid by banks that assume m ore risk. Four o f the six studies o f the determinants o f rates on the sub ordinated debt o f banks find no significant e f fects o f risk measures on interest rates. Implications f o r the Effectiveness o f Market Discipline In evaluating these results, it is important to note that, under the procedures follow ed by federal bank regulators in recent years, risk has a m ore certain implication fo r bank profits than fo r the returns to the holders o f uninsured de posits or subordinated debt. Losses on bank assets reduce profits, and if losses fo rce a bank to fail, the bank shareholders are likely to receive nothing after the liquidation or sale o f the bank. Uninsured depositors and holders o f subor dinated debt, in contrast, receive less than the principal plus contracted interest only if a bank fails. In most cases, the failed bank is m erged w ith another bank, and the surviving banks assume all liabilities o f the failed banks, in cluding those in the form o f uninsured deposits and subordinated debt. Most o f the cases in which uninsured depositors and holders o f subordinated debt absorb some losses involve banks smaller than those included in the studies described in this paper.19 These observations are consistent w ith the conclusion that interest rates on bank liabilities w ould be m ore sensitive banks. Of these 69 failed banks, 66 had total assets less than $100 million, while the remaining three had total assets less than $200 million. The failure of some large banking organizations in the Southwest, in which the BHC’s bondholders absorbed losses, occurred after the periods covered by these studies. 17 to the risk assumed by banks if bank creditors lost at least part o f their principal plus interest in each bank failure. The empirical results cannot be used to indicate the degree o f risk that banks would assume if bank supervisors eliminated various forms of supervision and regulation, relying instead on market forces to limit bank risk. To illustrate such a change in policies, suppose bank super visors eliminate capital requirements and restric tions on the types o f assets that banks may ac quire, substituting a requirement that banks issue subordinated debt. The empirical results do not tell us whether the probability o f bank failures would increase or decrease under such a change in policies. The only useful information from the empirical studies is that investors in bank stocks, w ho have the strongest incentives to be sensitive to the risk assumed by banks, are able to dif ferentiate among banks on the basis o f risk. CONCLUSIONS This theoretical exercise illustrates how market forces could limit the incentives fo r banks to assume risk. The incentives fo r banks to assume relatively high risk are reduced if the insurance coverage o f bank creditors drops from full to partial coverage. One o f the im por tant differences among the various approaches to prom oting market discipline o f banking risk involves the vulnerability o f banks to runs. Banks are m ore vulnerable to runs if depositors are at risk than if the risks are borne by those holding long-term bank debt that is subor dinated to deposits. Empirical studies o f the effectiveness o f market discipline report mixed results. The most consistent result is that the stock prices o f individual banks reflect the risk assumed by banks. Market discipline o f such risk w ould tend to be m ore effective if bank creditors w ere forced to absorb losses in a m ore consistent fashion in bank failure cases. The empirical studies do not indicate the degree o f risk that banks w ould assume if deposit insurance w ere reform ed to enhance the effectiveness o f market discipline. Thus, the empirical studies do not perm it us to determine w hether the probability o f bank failures would rise or fall if the current form s o f bank regula tion w ere eliminated in favor o f market disci pline by bank shareholders and creditors. REFERENCES Avery, Robert B., Terrence M. Belton, and Michael A. Goldberg. “ Market Discipline in Regulating Bank Risk: New Evidence from the Capital Markets,” Journal of Money, Credit and Banking (November 1988), pp. 597-610. Baer, Herbert, and Elijah Brewer. “ Uninsured Deposits as a Source of Market Discipline: Some New Evidence,” Federal Reserve Bank of Chicago Economic Perspectives (September/October 1986), pp. 23-31. Beighley, H. Prescott. “ The Risk Perceptions of Bank Holding Company Debtholders,” Journal of Bank Research (Summer 1977), pp. 85-93. Beighley, H. Prescott, John H. Boyd, and Donald P. Jacobs. “ Bank Equities and Investor Risk Perceptions: Some Entailments to Capital Adequacy Regulation,” Journal of Bank Research (Autumn 1975), pp. 190-201. Benston, George J., and George G. Kaufman. Risk and Solvency Regulation of Depository Institutions: Past Policies and Current Options. Monograph 1988-1, Monograph Series in Finance and Economics, Salomon Brothers Center for the Study of Financial Institutions, Graduate School of Business Administration, New York University. Bernanke, Ben S. “ Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression,” American Economic Review (June 1983), pp. 257-76. Bovenzi, John F., and Arthur J. Murton. “ Resolution Costs of Bank Failures,” FDIC Banking Review (Fall 1988), pp. 1-13. Boyd, John H., and Arthur J. Rolnick. “A Case for Reforming Federal Deposit Insurance,” Annual Report, 1988, Federal Reserve Bank of Minneapolis. Brewer, Elijah III, and Cheng Few Lee. “ How the Market Judges Bank Risk,” Federal Reserve Bank of Chicago Economic Perspectives (November/December 1986), pp. 25-31. Budget of the United States Government, Fiscal Year 1991 (U.S. Government Printing Office, 1990). Calomiris, Charles W., R. Glenn Hubbard, and James H. Stock. “ The Farm Debt Crisis and Public Policy,” Brookings Papers on Economic Activity (2:1986), pp. 441-79. Cargill, Thomas F. ’’CAMEL Ratings and the CD Market,” Journal of Financial Services Research (December 1989), pp. 347-58. Cooper, Kerry, and Donald R. Fraser. “ The Rising Cost of Bank Failures: A Proposed Solution,” Journal of Retail Banking (Fall 1988), pp. 5-12. Cornell, Bradford, and Alan C. Shapiro. “ The Reaction of Bank Stock Prices to the International Debt Crisis,” Journal of Banking and Finance (March 1986), pp. 55-73. Council of Economic Advisers. Annual Report, 1989 (U.S. Government Printing Office, 1989). Crane, Dwight B. “A Study of Interest Rate Spreads in the 1974 CD Market,” Journal of Bank Research (Autumn 1976), pp. 213-24. Dwyer, Gerald P., and R. Alton Gilbert. “ Bank Runs and Private Remedies,” this Review (May/June 1989), pp. 43-61. Ely, Bert. “ Yes — Private Sector Depositor Protection is a Viable Alternative to Federal Deposit Insurance!” Pro ceedings of a Conference on Bank Structure and Competi tion, Federal Reserve Bank of Chicago, May 1985, pp. 338-53. JANUARY/FEBRUARY 1990 18 England, Catherine. “A Proposal for Introducing Private Deposit Insurance,” Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 1985, pp. 316-37. James, Christopher. “An Analysis of the Use of Loan Sales and Standby Letters of Credit by Commercial Banks,” Federal Reserve Bank of San Francisco Working Paper 87-09, October 1987. ________“ First: Shut Insolvent Thrifts,” Washington Post, October 9, 1988. ________“ Empirical Evidence on the Implicit Government Guarantees of Bank Foreign Loan Exposure,” CarnegieRochester Conference Series on Public Policy (Spring 1989), pp. 129-61. Federal Deposit Insurance Corporation. Deposit Insurance in a Changing Environment, April 15, 1983. Fraser, Donald R., and J. Patrick McCormack. “ Large Bank Failures and Investor Risk Perceptions: Evidence from the Debt Market,” Journal of Financial and Quantitative Analysis (September 1978), pp. 527-32. Gendreau, Brian C., and David B. Humphrey. “ Feedback Ef fects in the Market Regulation of Bank Leverage: A TimeSeries and Cross-Section Analysis,” Review of Economics and Statistics, 1980, pp. 276-80. Gilbert, R. Alton, and Levis A. Kochin. “ Local Economic Ef fects of Bank Failures,” Journal of Financial Services Research (December 1989), pp. 333-45. Gilbert, R. Alton, and Geoffrey E. Wood. “ Coping with Bank Failures: Some Lessons from the United States and the United Kingdom,” this Review (December 1986), pp. 5-14. Gorton, Gary, and Anthony M. Santomero. “ The Market’s Evaluation of Bank Risk: A Methodological Approach,” Pro ceedings of a Conference on Bank Structure and Competi tion, Federal Reserve Bank of Chicago, May 1988, pp. 202-18. Graham, Fred C., and James E. Horner. “ Bank Failure: An Evaluation of the Factors Contributing to the Failure of Na tional Banks,” Proceedings of a Conference on Bank Struc ture and Competition, Federal Reserve Bank of Chicago, 1988, pp. 405-35. Grossman, Richard S. “ The Macroeconomic Consequences of Bank Failures Under the National Banking System,” PAS Working Paper 14, Bureau of Economic and Business Af fairs, U.S. Department of State (April 1989). Hannan, Timothy H., and Gerald A. Hanweck. “ Bank In solvency Risk and the Market for Large Certificates of Deposit,” Journal of Money, Credit and Banking (May 1988), pp. 203-11. Herzig-Marx, Chayim. “ Modeling the Market for Bank Debt Capital,” Federal Reserve Bank of Chicago Staff Memoran da 79-5, 1979. Herzig-Marx, Chayim, and Anne S. Weaver. “ Bank Sound ness and the Market for Large Negotiable Certificates of Deposit,” Federal Reserve Bank of Chicago Staff Memoranda 79-1, 1979. RESERVE BANK OF ST. LOUIS FEDERAL Keehn, Silas. Banking on the Balance: Powers and the Safe ty Net, Federal Reserve Bank of Chicago, 1989. Kaufman, George G. “ Bank Runs: Causes, Benefits and Costs,” Cato Journal (Winter 1988), pp. 559-87. Office of the Comptroller of the Currency. “An Evaluation of the Factors Contributing to the Failure of National Banks: Phase II,” Quarterly Journal, Office of the Comptroller of the Currency (Vol. 7, No. 3, 1988), pp. 9-20. Pettway, Richard H. “ Market Tests of Capital Adequacy of Large Commercial Banks,” Journal of Finance (June 1976), pp. 865-75. ________“ Potential Insolvency, Market Efficiency, and the Bank Regulation of Large Commercial Banks,” Journal of Financial and Quantitative Analysis (March 1980), pp. 219-36. Randall, Richard E. “ Can the Market Evaluate Asset Quality Exposure in Banks?” New England Economic Review (Ju ly/August 1989), pp. 3-24. Schwartz, Anna J. “ The Effects of Regulation on Systemic Risk,” Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago (May 1988), pp. 28-34. Shome, D.K., S.D. Smith, and A.A. Heggestad. “ Capital Ade quacy and the Valuation of Large Commercial Banking Organizations,” Journal of Financial Research (Winter 1986), pp. 331-41. Short, Eugenie D., and Gerald P. O ’Driscoll. “ Deregulation and Deposit Insurance,” Federal Reserve Bank of Dallas Economic Review (September 1983), pp. 11-22. Smirlock, Michael, and Howard Kaufold. “ Bank Foreign Lending, Mandatory Disclosure Rules, and the Reaction of Bank Stock Prices to the Mexican Debt Crisis,” Journal of Business (July 1987), pp. 347-64. Smith, Clifford W., and Jerold B. Warner. “ On Financial Con tracting: An Analysis of Bond Covenants,” Journal of Finan cial Economics (June 1979), pp. 117-61. Smith, Fred. “ Cap the Financial Black Holes,” Wall Street Journal, September 29, 1988. Wall, Larry D. “A Plan for Reducing Future Deposit In surance Losses: Puttable Subordinated Debt,” Federal Reserve Bank of Atlanta Economic Review (July/August 1989), pp. 2-17. 19 Mark D. Flood Mark D. Flood is a visiting scholar at the Federal Reserve Bank of St. Louis. David H. Kelly provided research assistance. On the Use of Option Pricing Models to Analyze Deposit Insurance T„e FAILURE rate o f banks and thrifts has exploded over the past decade, making reform o f the deposit insurance system a topic o f con siderable interest to regulators, bankers, and economists. As illustrated by figure 1, which shows the total number o f failed commercial banks (excluding thrifts) fo r each year since the chartering o f the Federal Deposit Insurance Cor poration (FDIC), the annual number o f com m er cial bank failures in each o f the last several years has exceeded its previous peak, attained during the Great Depression. The status o f the thrift industry is even m ore grim, with losses to the Federal Savings and Loan Insurance Cor poration (FSLIC) estimated at $160 billion or more. The prim ary consequence o f these fail ures fo r public policy is the enormous losses, especially to the FSLIC, as depositors in these failed institutions are reimbursed. This article considers a particular set o f eco nomic tools used to evaluate deposit insurance.1 Option pricing models are among the techniques available fo r analyzing the deposit insurance system. These models can be used to assign 1This article does not address the issue of why we should have deposit insurance. The rationale for the current system of bank regulation and for deposit insurance in particular is based on two related principles: protection of the depositor and the mitigation of contagious bank runs. See FDIC (1984) and U. S. Treasury (1985). Benston and specific values to the claims o f each o f the in terested parties involved in the deposit insur ance system — the insurer, financial institutions, and depositors. Such valuations can then be used, fo r example, to estimate the net value o f the governm ent’s insurance fund or to deter mine a fair price that a bank should pay fo r its insurance. M ore generally, by comparing in surance valuations with different m odel parame ters, one can investigate the system o f incen tives under a given regulatory scheme, such as the risk incentives fo r bank shareholders and depositors under the present system. Finally, comparisons o f insurance values and incentives can be made across various proposed regulatory schemes. These applications are illustrated be low w ith some examples. The usefulness o f option pricing models fo r evaluating deposit insurance is o f special in terest fo r tw o reasons. First, the consensus among the interested parties is that the present deposit insurance system has contributed to the current crisis. Second, in the context o f this debate, a number o f economists have used the Kaufman (1988) identify three reasons for bank regulation in addition to the two traditional rationales, namely disrup tion to communities from localized bank failures, moral hazard induced by deposit insurance and restrictions on competition. JANUARY/FEBRUARY 1990 20 Figure 1 Bank Failures (Insured and Uninsured) Number of Failed Banks 200 Number of Failed Banks 200 0 1-------------- ii a a s r . 1934 39 44 49 — r ........ —..........J 54 59 64 69 74 79 84 o 1988 Source: FDIC (1989) m odern theory o f option pricing to explain the incentives and measure the costs both o f the current system o f deposit insurance and of some suggested alternatives. This paper p re sents the basic theory o f option pricing, explains how it can be applied to deposit insurance, and analyzes some o f the issues involved in its use. A PR IM ER ON O P T IO N PR ICING This paper presumes no knowledge o f options or o f the various economic models that have been used in the academic literature to assign values to options. Thus, it begins with a brief description o f options and some o f the major contributions to the theory o f pricing options that have been made in the past tw o decades. Contingent Claims, or Options A call option is a legal contract that gives its ow n er the right to buy a specified asset at a fix ed price on a specified date.2 Similarly, a put op tion gives its ow n er the right to sell a specified asset at a fixed price on a specified date. Option contracts are usually sold b y one party to an other.3 The person w ho owns an option con tract is called the holder o f the option. The per 2This definition is a paraphrase of the definition given by Cox and Rubinstein (1985), p. 1. It describes a “ Euro pean” option, which is distinguished from an “ American” option. An American option gives its owner the right to buy at any time on or before the specified date. 3They are sold, because options have a non-negative value; because they are a right to buy (or sell) the asset, they do Digitized forFEDERAL FRASER RESERVE BANK OF ST. LOUIS son w ho sells an option contract — that is, the person w ho w ill be compelled to perform if the option holder invokes her right as specified in the contract — is called the writer o f the option. The act o f invoking the contract is called exer cising the option. The fixed price identified by the option contract is called the striking price. The date at which the option can be exercised is called the expiration date o f the option. These legal contracts are probably best known by the stock options that are bought and sold by brokers in the trading pits o f organized op tions exchanges in Chicago, N ew York and elsewhere. In addition to options on common stock, there are active markets fo r options on agricultural comm odity futures, foreign curren cies, stock index portfolios, and governm ent securities, to name only a few . The definition o f an option, how ever, does not limit the term to those contracts actively traded on the floors o f organized financial exchanges. By definition, an option is any appropriately constructed legal contract betw een the w riter and the holder, regardless o f w hether it is ever traded. Expiration-date Values o f Options Consider now the value to the holder o f an expiring put option, as illustrated in figure 2. The value o f the underlying asset specified by the contract is given on the horizontal axis, w hile the value o f the option itself is given on the vertical axis. The point K on the horizontal axis is the specified striking price fo r the asset. If the value o f the underlying asset is above the striking price on the expiration date, then the put option w ill not be exercised; anyone w ho truly wanted to sell the asset w ould do so outright at the going price, rather than using the option and receiving the striking price. In this case, the option expires worthless, and the option holder experiences no gain or loss on the expiration date. On the other hand, if the value o f the asset is below the striking price, then the holder w ill exercise her option and receive the striking not compel the owner of the contract to do anything. Although they are valuable, nothing in the definition of an option requires that they be offered for sale; that is, their value does not depend on how they were obtained. 21 Figure 2 Value of Put Option to Holder Expiration Value of Option Figure 3 Value of Put Option to Writer Expiration Value of Option (1) State Action Option value AT < K At > K Exercise No exercise P = K - AT P = 0. For this reason, options are also referred to as "contingent claims” on the underlying assets. The corresponding net payoffs to the w riter o f the put option are given in figure 3. Notice that his payoffs are exactly the inverse o f those fo r the option holder. Also note that the payoff at expiration to the w riter o f an option is never positive; at best it is zero. It is fo r this reason that options are sold to the holder, rather than being given away fre e o f charge. The price in itially paid fo r the option — the option price or option premium — could be incorporated into the figures by simply shifting the holder's payoffs dow n and the w riter’s payoffs up by the appropriate amount. The payoffs at expiration to the holder and w riter o f a call option are given in figures 4 and 5, respectively. Th e corresponding analysis fo r call options is precisely analogous to the analysis just given fo r put options. The Black-Scholes Option Valua tion M od el ----------------------------- |-----------------0 K Value of Underlying Asset at Expiration price fo r the asset. In this case, her net gain on the expiration date w ill be (K - A t ), the d iffe r ence betw een the striking price and the current price, since she can turn around and replace the asset immediately, if she wants to. Thus, the expiration value o f the option and the decision about w hether to exercise are contingent upon the value o f the underlying asset at that time: 4For an exposition of the arbitrage bounds on option prices, see Merton (1973), or Cox and Rubinstein, ch. 4. 5Almost all derivations of option pricing models, including that of Black and Scholes, are stated in terms of call rather than put options. As it happens, this distinction is largely irrelevant, because call option valuations are readi ly converted to put option valuations, and vice-versa, via an arbitrage relationship known as “ put-call parity” . Put- Having described the value o f an option at ex piration leaves the question o f its value prior to expiration unanswered. Instead o f being a sim ple function o f A t and K, the value o f an option before maturity depends on several additional factors. Although a num ber o f bounds had been placed on the value o f an unexpired option by using relatively simple arbitrage arguments, an important advance in the valuation o f unexpired options was made by Black and Scholes (1973).4 They obtained an exact equation fo r the value o f a put option under an unrestrictive set o f assumptions.5 Th eir result has since been elab orated and generalized by others.6 In their model, the value o f an unexpired op tion depends on five things: call parity is an exact relationship for European options and an approximate one for American options (see Cox and Rubinstein, pp. 150-52); throughout this paper it is treated as exact. Put-call parity is first presented by Stoll (1969). 6One such generalization is found in the shaded insert. For a partial survey of option pricing models, see Cox and Rubinstein, ch. 7. JANUARY/FEBRUARY 1990 22 K A T a R = = = = = the the the the the striking price current asset price time remaining to expiration volatility o f the asset price risk-free interest rate. Figure 4 Value of Call Option to Holder Expiration Value of Option Almost as notable is what the option’s value does not depend on: any characteristic o f the holder or the w riter.7 Under their assumptions, Black and Scholes are able to include an option in a riskless portfolio. Such a portfolio must earn the risk-free interest rate, and they are able to use this result, along with an assumption about the probability distribution o f the asset price, to identify an exact value, P, fo r a put option: P = (K*e - RT)*N(X + o\/t ) - A-N(X), where: X - 1 - [ln (K -e -RT)- ln (A )] - VzcrJ T <rJT N (») = the Vstandard normal cumulative probability function Figure 5 Value of Call Option to Writer Expiration Value of Option ln («) = the natural logarithm function e H the base o f the natural logarithm Although this form ula may at first appear complicated, a rough intuition can be provided relatively painlessly.8 First, e ~ RT is just the p re sent value discount factor fo r T periods at in terest rate R w ith continuous compounding, so that K-e_RT is the present value o f the striking price. Keeping in mind that N(X + oV t ) is a pro bability, the first term is the expected present value o f the striking price at expiration, given that A t < K. Similarly, the second term, A*N(X), is the expected present value o f the expirationday asset price, again given that A t < K.9 Thus, the value o f the option is the expected present value o f its value at expiration, given by condi tion 1 above. Unfortunately, no easy, correct interpretation can be attached to the specific probabilities, N(X + oV t ) and N(X), in the tw o terms. These 7For example, one might suspect that the holder’s attitudes toward risk or her beliefs about the asset price at expira tion should influence the option’s value to her. This is not the case, however. Also note that four of the five factors, at least theoretically, are well-defined and directly obser vable at the time of valuation. The exception is asset volatility, which must be estimated from observable fac tors; see Cox and Rubinstein, pp. 280-87, for an example of an estimation technique. 8A full derivation of the formula is fairly involved and will not be presented here. Interested readers are referred to FEDERAL RESERVE BANK OF ST. LOUIS probabilities are closely related to the probabili ty that A t < K, but they are not quite the same, because the present value o f the striking price is known w ith certainty, w hereas the p re sent value o f the asset’s price on the expiration day, A T*e~RT, is not; the current asset price, A, appears instead. Malliaris (1983) for a mathematically advanced approach or to Cox and Rubinstein, ch. 5, for a longer but less technical derivation. 9The corresponding expected present values for the case when A t is greater than K are both zero, because then the expiring option is worthless and will not be exercised; hence, this possibility adds nothing to the current value of the option. 23 In spite o f its complexity, the option pricing equation is still a useful tool. In one sense, the formula can be treated as a black box in which the five parameters (K, A, T, o and R) enter at one end, and the value o f the put option, P, comes out at the other; a computer spreadsheet or calculator can be program m ed to perform the intervening calculations defined by the fo r mula. For example, if the current asset value is A = $985, the standard deviation o f asset returns is o = 0.3 percent, the striking price o f the option is K = $1000, the time to expiration is one year, and the riskless interest rate is 8 percent per year, then the Black-Scholes equa tion tells us that the put option is w orth $85.45. Figure 6 graphs the Black-Scholes value o f a put option fo r a range o f current asset values from zero to $1500, w h ere the values o f the other four parameters are the ones just given. The Brownian M otion Assumption Figure 6 Value of Put Option to Holder Option Value Prior to Maturity (K =1000, T=1, o = .3 , Rf = .08) 1000 932.12 o Asset Value Figure 7 Value of Net Asset Position Net Position Value Not surprisingly, the distribution o f asset prices is a crucial factor in determining the ex act form o f the option pricing equation. In their derivation, Black and Scholes assumed that the price o f the underlying asset progressed ran domly through time according to geom etric Brownian motion. This is the assumption that leads to the specific normal probability func tions in their pricing equation. Brownian motion was first used to describe the random progress o f a single molecule through a gas from a given starting point.10 It is a mathematical model o f motion that identifies the w ay the particle can move. Three restric tions are implied by Brownian motion: 1)The path follow ed must be continuous;11 2)A11 future movements are independent o f all past m ovem ents;12 3)The change in position betw een time s and time t is normally distributed w ith mean equal to zero and a standard deviation equal to o V (t-s ). Note that standard deviation is directly p ro portional to the amount o f time that has passed. 10We are here concerned with only a single dimension of motion, for example, the East-West coordinate of the molecule or the price of an asset. "A lth ou gh this may be true of molecules, it need not be the case for asset prices, as is considered in the shaded in sert on Merton’s jump-diffusion model. 12This implies, for example, that the molecule cannot build up momentum or that prices do not have a predictable Thus, the longer one waits, the less certain one is about the location o f the molecule, or, in our case, the price o f the asset. Simple Brownian motion is not completely satisfactory fo r describing asset prices, h ow ever. W hile a norm ally distributed random vari able can take on negative values, an asset price cannot. Th erefore, geom etric Brownian motion, a variant, is assumed fo r the Black-Scholes model.13 Under geom etric Brownian motion, the third restriction is m odified, so that the loga rithm o f the change in position, rather than the trend. It does not mean that the future location is indepen dent of the past location. 13By way of terminology, simple Brownian motion (also known as arithmetic Brownian motion) and geometric Brownian motion are examples of the Wiener process (also known as the Gauss-Wiener process), which, in turn, is a special case of the Ito process. JANUARY/FEBRUARY 1990 24 change in position itself, is norm ally distributed w ith mean zero and standard deviation o V (t-s ). This distributional assumption gives us the spe cific functional form which appears in the BlackScholes equation. Thus, this assumption is im portant: a different distribution w ould generally yield a different pricing equation, as illustrated by M erton’s (1976) jump-diffusion option pricing model, which is presented in the shaded insert on the opposite page. Risk and Hedging in Options The option pricing equation has the paradox ical property that, although risk (as measured by volatility in the asset price) is itself a factor in the option’s value, the attitudes tow ard risk o f the holder and the w riter (and anyone else) are not. The option’s value is a function o f five variables, none o f which depends on the charac teristics o f the individuals involved. Black and Scholes achieved this by showing that the op tion can be made part o f a com pletely hedged (that is, riskless) portfolio. Any option w riter w ho offered a risk discount w hen selling an op tion w ould find him self selling many option con tracts to investors who, in turn, could hedge the risk com pletely and pocket the risk discount as an arbitrage profit. It is fo r this reason that the discount rate which appears in the pricing equation is the risk-free rate o f interest, and at titudes tow ard risk are irrelevant to the value o f the option. T o see how the hedged portfolio works, con sider the value o f a put option to the holder before expiration, depicted in figure 6, and the value o f the underlying asset purchased fo r the amount A, depicted in figure 7. The value o f the net asset investment increases one fo r one as the price o f the asset increases, and the val ue o f the option decreases, although not in a constant proportion. The key to the hedged portfolio is to buy put options and underlying assets in the appropriate ratio, so that, w hen the asset price increases, the increase in value o f the net asset investment w ill be precisely offset by the decrease in value o f the option position, and vice-versa. This im plies a riskless total portfolio. O f course, the ap propriate ratio (called the "hedge ratio” or "op tion delta”) also changes as the asset price changes, because the value o f a put option does 14This connection was first made by Black and Scholes and first applied to deposit insurance by Merton (1977). FEDERAL RESERVE BANK OF ST. LOUIS not decrease as a constant proportion o f the asset value (the put option’s value is represented by a curved line). This implies that the holder o f a com pletely hedged portfolio must con tinuously adjust the relative proportions o f op tions to assets if the hedged portfolio is to re main riskless. Black and Scholes presume that at least some investors are large and sophisticated enough to do this. Because the risk o f an option can be com pletely diversified, the risk-free rate is the ap propriate interest rate to use fo r discounting the option’s uncertain p a yoff at expiration. Nevertheless, the risk (defined as price volatility) o f the underlying asset is a factor in the op tion’s value, because asset risk affects the ex pected value o f the option's payoff. This is due to the limited liability nature o f the option. Although increasing the volatility o f the asset price increases both the chance o f getting a very high expiration-day asset price and the chance o f getting a ve ry low expiration-day asset price, the bad (high price) outcomes all have a w eight o f zero in the put option valua tion, w hile the good (low price) outcomes have a w eight o f (K - A t ). The volatility o f the op tion’s value also increases w ith that o f the asset price, but the volatility o f the option’s value is irrelevant, because it can be com pletely hedged. D EPO SIT INSURANCE AN D O PT IO N S The analysis o f deposit insurance is a natural, albeit not obvious, extension o f option pricing models. The connection betw een the tw o comes through the limited liability property common to both options and common stock.14 This p ro perty implies an "expiration-day” p a yoff fo r deposit insurance that can be m odeled as an or dinary put option. Similarly, other claims on a financial interm ediary’s assets can be modeled as options or combinations o f options. The benefit is that, given such a model, option pric ing theory allows us to assign values to each o f the claims. These values are the key to option pricing’s usefulness in this context, because they allow tw o sorts o f comparisons to be made. First, variations in the parameters o f the op tion pricing equation can be considered.15 Such variations are o f special interest, because, in the 15For example, in Black and Scholes’ model, the five parameters: K, A, T, o and R would be varied. 25 Merton's Jum p-Diffusion Model The Black and Scholes (1973) derivation o f an option’s value was based, in part, on a particular assumption about the random be havior o f the price o f the underlying asset. Th eir assumption o f geom etric Brownian m o tion as a description o f the movements o f asset prices is by no means the only possibili ty. In general, different assumptions about the statistical properties o f asset price behav ior produce different valuation equations fo r an option on that asset. M erton’s jumpdiffusion model is one o f several alternative formulations that have been developed.1 Just as arithmetic Brownian motion was not an apt model o f asset price movements, em pirical research suggests that geom etric Brownian motion, at least fo r some assets, is similarly inappropriate. An alternative, p ro posed by M erton (1976), is a combination o f geom etric Brownian motion with random, discontinuous jumps in the asset price, such as might occur at the announcement o f some news event.2 This arrangem ent violates the first condition fo r Brownian motion and modifies the third condition again.3 W hen a jump occurs, the asset price is abruptly shifted by a random amount; the logarithm o f this shift is normally distributed, analo gously to geom etric Brownian motion. This new process has tw o important im plications fo r the option pricing model. First, the option pricing equation is different from the Black and Scholes formula, since it must account fo r the new jumps. I f w e relabel the Black-Scholes value as Pbs = Pbs(K,A,T,o,R), 'Some others are McCulloch’s (1981, 1985) Paretianstable process, and Cox and Ross’s (1976) alternative jump processes and constant elasticity of variance (CEV) Ito process. 2Merton’s derivation is only one of several alternatives to the Black and Scholes model. It was chosen to il lustrate some of the issues involved in selecting an ap propriate pricing model, not because it outperforms the others in some sense. See Rubinstein (1985) for a per formance comparison of several models. context o f deposit insurance, some o f the pa rameters can be controlled or influenced by the parties to the option contract. Thus, each party has a clear interest in influencing the parame ters to his ow n benefit and th erefore to the then w e can w rite the M erton formula as a function o f the Black-Scholes value: oo Pm = . I J f t - P bl(K ,A,T,0I,R) i = 0*+ (1 -/?,) * (K *e-RT - A )], where: pi - £ ^ 1M U i! ©i - V (o2+ hzi/t) g - the Poisson frequency o f jumps h2 ■ the variance o f the shift distribution. Not surprisingly, the M erton equation is m ore complicated than the Black and Scholes equation. Nevertheless, it is still a function o f variables that are at least hypothetically observable.4 Second, the presence o f the jumps com plicates the diversification problem. The sim ple hedging portfolio used fo r the BlackScholes model w ill not do, because even con tinuous rebalancing cannot insure the port folio’s value at the jump points. If, how ever, the jumps are idiosyncratic (that is, firmspecific), then their risk can still be elimi nated by holding a well-diversified portfolio that includes the assets o f many firms in many industries. If the jump disturbances are idiosyncratic, then the risk-free interest rate is still the appropriate discount rate fo r the expected option payoff, and individual at titudes tow ard risk are not a factor in the op tion’s value before expiration. 3The discontinuous jumps arrive according to a Poisson process, which conforms to the second condition. 4ln practice, the statistical parameters: o, h and g would have to be estimated, either from previous observations or via some other technique. The equation given here is a special case of a more general formulation given by Merton (1976). His derivation is of a call option price, which has been re-arranged here using put-call parity. detriment o f the other. The risk-incentive pro blem presented below exemplifies this sort o f application. Comparisons based on option pric ing models indicate not only the direction, but also the magnitude, o f such incentives. JANUARY/FEBRUARY 1990 26 Second, various deposit insurance structures can be compared. The structure o f deposit in surance is defined here by the number and type o f options pertaining to each o f the in terested parties. Changes in deposit insurance structure are different from the parameter changes w ithin an insurance structure, con sidered in the preceding paragraph. Thus, fo r example, the FDIC could use option models to estimate the net increase or decrease in the p re sent value o f the insurance fund caused by a switch from one structure to another; or it could examine the change in risk incentives oc casioned by the same switch. Th ree different structures, illustrating some o f the issues involv ed, are presented in the follow ing examples.16 100 Percent Deposit Insurance Coverage T o see how deposit insurance and options are related, consider the follow ing simplified bank ing scenario. A single banker both owns and runs a bank, a single large depositor provides the entire liability portfolio o f the bank, and a single insurer, the FDIC, insures deposits and w ill liquidate the bank in the event o f insolven cy.17 The liability portfolio consists o f a single deposit due at year-end. Also at year-end, the FDIC examines the bank to determine the value o f assets, which will, in turn, determine w heth er liquidation occurs. If the bank is econom i cally insolvent, it is closed by the FDIC, which liquidates the assets at market values and pays o ff the depositor in full.18 If the bank is eco nomically solvent, control remains with the banker, w h o can either renegotiate the deposit or liquidate the bank. 16Full coverage is considered first, because it is the simplest insurance structure possible, and because it approximates the current system of extensive coverage combined with the FDIC’s tendency to arrange purchase and assumption transactions, rather than deposit payouts, for failed banks. 17The assumptions of a single owner-manager for the bank and a lone depositor are clearly broad abstractions from reality. The owner-manager assumption allows us to ignore principal-agent incentives; see Barnea, Haugen and Senbet (1985). Similarly, the assumption of a single depositor effectively precludes the ability of depositors to withdraw their funds individually without forcing an im mediate closure of the bank. Although these are both im portant issues, the purpose of the present analysis is to il lustrate the general principles involved in the application of option models to deposit insurance, rather than to model a bank in its full complexity. 18The bank is defined as economically insolvent when the market value of its assets is less than the present value of its liability to the depositor. This terminology is meant to FEDERAL RESERVE BANK OF ST. LOUIS N ow consider the payoffs to the three in terested parties — banker, FDIC and depositor — when the year-end audit is perform ed. These payoffs are illustrated in figures 8-10. Each par ty’s year-end payoff is plotted as a function o f the year-end value o f the bank’s assets. Note that the sum o f the payoffs to all o f the parties (obtained by adding the graphs vertically) equals the value o f the bank's assets. These functions show how the bank's assets w ill be distributed after the audit is perform ed. Also note the shape o f the payoff functions fo r the banker and the FDIC; in effect, the banker's portfolio consists o f the bank’s assets, whose value is uncertain before the audit but known after ward, the bank’s deposits, w hose value is known to be L, and a put option w ith striking price L w ritten by the FDIC.19 The FDIC, on the other hand, has effectively w ritten the put option on the assets o f the bank and sold that option to the banker fo r the price o f the deposit insur ance premium.20 The depositor has issued the bank a risk-free loan, which pays o ff the amount L, including accrued interest. W ith this in mind, the usefulness o f an option pricing model to evaluate deposit insurance b e comes m ore apparent. An option pricing model provides an estimate o f the actuarial dollar value o f deposit insurance, as w ell as a tool w ith which to analyze the economic incentives that deposit insurance creates. The depositor, fo r example, has a portfolio, D, that is worth, at the beginning o f the year, simply the present value o f the deposit liability discounted at the riskless rate, L*e_RT; if his year-end payoff, L, is $1000, and the riskless rate, R, is 8 percent, then the value o f this portfolio at the beginning contrast with an illiquidity, or legal insolvency, which is brought on by an inability to meet maturing short-term liabilities with liquid assets. The legal profession has a separate terminology for these two concepts: the “ balance sheet test” is used to determine economic insolvency, and the “ equity test” is used to determine legal insolvency. See Symons and White (1984), pp. 603-16, for an exposi tion. Since there are no short-term liabilities in the simplified world here, legal insolvency is not germane. 19The bank’s deposits represent a “ short” position, or bor rowing, for the banker. The net payoff shown in figure 8 can be gotten by drawing the individual payoff graphs for the components of the portfolio and adding these together vertically as before. This portfolio is also equivalent, via put-call parity, to a simple call option on the assets of the bank. “ Compare figure 9 with figure 3. The insurance premium is considered a sunk cost at the time of the audit and hence is not included in the graphs. 27 o f the year is $1000e 08 = $923.12. The FDIC, on the other hand, has w ritten a put option with striking price L = $1000; if, fo r example, the standard deviation o f the bank’s asset re turns is o = 0.3 percent, and the current value o f the bank’s assets is $985, then the value o f the FDIC’s portfolio is given by the BlackScholes equation as - P = -$85.45. Figures 8-10 Payoff to Banker 0 .. .............. f ----------------------------L Asset Value Payoff to Depositor 0 --------------------- 1-----------------------------L Asset Value 21Recall that the value of an option to the holder increases with the volatility of the underlying asset. For the Black and Scholes model, risk is defined as the standard devia tion of the logarithm of the asset’s value. 22This is the function of bond rating services, such as Moody’s and Standard and Poor’s. See Barnea, Haugen, and Senbet, especially pp. 33-35, for an exposition of the risk-incentive problem. 23For this reason, risk-taking is restricted by extensive regulation of commercial bank activities. In practice, the banker’s incentive may also be mitigated by the potential loss of a valuable bank charter or by nonpecuniary factors, for example, the potential loss of a bank manager’s pro fessional reputation in the event of a failure. These factors are beyond the scope of the option model. O f particular interest are the incentives cre ated by deposit insurance. Under 100 percent insurance, the depositor does not care about the value o f the bank's assets, since he receives his deposit back with interest, regardless o f the bank’s condition. The banker, how ever, receives the positive equity capital, if the bank is solvent; if the bank is insolvent, the loss is charged to the FDIC. This "heads I win, tails you lose" ar rangement is certainly not peculiar to banks; it applies to any corporate entity w ith limited stockholder liability. In the absence o f other in centives, the banker will make the corporation as risky as possible.21 W hat is peculiar to banks under 100 percent flat-rate deposit insurance is the absence o f such other incentives fo r the depositor and the banker to limit risk. Normally, creditors impose a risk premium on corporations, based on the riskiness o f the firm ’s assets.22 By definition, flat-rate insurance implies that the FDIC charges no risk premium. Similarly, the depositors charge no risk premium, because they are fully insured. The result, in our simplified model, is that the banker has an unmitigated incentive to increase the riskiness o f the bank's assets, while the FDIC has the inverse incentive to reduce the bank’s risk-taking.23 The risk incentive implied by extensive, flat-rate deposit insurance is the impetus fo r most o f the current proposals fo r deposit insurance reform .24 In analyzing both the current system and proposed reforms, many authors have used option pricing models.25 24Reform proposals include: risk-based insurance premia, risk-based capital requirements, larger capital re quirements, reduced insurance coverage, depositor coinsurance, subordinated debt requirements, increased supervision and more stringent asset regulation. See White (1989) for a survey of current proposals. 25See, for example, Merton (1977, 1978), McCulloch (1981, 1985), Sinkey and Miles (1982), Pyle (1983, 1984, 1986), Brumbaugh and Hemel (1984), Marcus (1984), Marcus and Shaked (1984), Ronn and Verma (1986, 1987), Thomson (1987), Furlong and Keeley (1987), Pennacchi (1987a, 1987b), Osterberg and Thomson (1988), Flannery (1989a, 1989b), and Allen and Saunders (1990). JANUARY/FEBRUARY 1990 28 A P P L IC A T IO N S TO OTHER D EPO SIT INSURANCE ARRANGEM ENTS W e can now extend the options model to other arrangements fo r deposit insurance, to evaluate their relative impacts. T w o illustrative cases will be presented: a coverage ceiling clause and a deductible clause. A significant character istic o f both these cases is that they impose a portion o f the bank's asset risk on the depos itor. The FDIC benefits directly from such pro visions, because they shift some potential losses directly to the depositors. In addition, imposi tion o f a possible loss on the depositor mitigates the risk-incentive problem that exists under 100 percent, flat-rate deposit insurance. In general, the depositor w ill m onitor the bank m ore close ly and w ill require a higher interest rate to compensate fo r the possibility o f default. Maximum Insurance Coverage Limit Although 100 percent coverage is often treated as the status quo de fa cto o f federal deposit in surance, coverage extends legally only to the first $100,000 per depositor per institution.26 A maximum coverage limit is a form o f co-insur ance, a technique used by insurers to reduce the moral hazard problem (the tendency o f in surance to alter the behavior o f the insured). Other basic forms o f co-insurance are the deduct ible and fixed proportional sharing o f losses.27 Under a deposit payout closure, the FDIC itself takes all o f the bank’s assets and liabilities into receivership. It then sells the assets and pays the depositors up to the maximum cover age limit plus any excess o f asset sales over in surance claims, distributed on a pro rata basis. As a result, this method is best modelled by the deductible considered below. The upshot is that the payoffs under the FDIC’s maximum cover age limit do not conform to the familiar (from, say, automobile or health insurance) maximum coverage arrangement illustrated here. Consider a maximum coverage limit o f M dollars fo r the depositor (w here M < L), illus trated in figures 11— 13. Under this arrange ment, the depositor receives the full deposit amount, L, in the event o f any insolvency or shortfall, up to the amount, M, o f the coverage limit. Thus, the depositor’s portfolio contains the deposit amount, L, and he has w ritten a put op tion on the bank's assets with striking price (L - M). This put option is the result o f the max imum coverage limit. The FDIC holds the put option w ith striking price (L - M), but has w rit ten a second put option on the bank’s assets with striking price L. As before, this put option (with striking price L) is held by the banker, w ho also owns the assets and ow es the amount L to the depositor. The applicability o f the maximum coverage limit considered here is complicated by the FDIC’s current closure protocol. Bank closures by the FDIC can take one o f tw o forms: pur chase and assumption or deposit payout. Under a purchase and assumption closure, healthy assets and all deposits are transferred to an other healthy bank, with the FDIC absorbing the problem assets and any net loss.28 This sort o f transaction is best modelled by the 100 per cent coverage considered above. Because o f the put option w ritten by the depositor and held by the FDIC, the depositor now shares in the risk o f the bank’s assets. His deposit is now w orth less, and he w ill discount the promised p a yoff m ore steeply. Extending the example given above fo r the case o f 100 percent coverage, the depositor’s portfolio, which contained only the riskless deposit, worth $932.12 w hen discounted, is now augmented by the put option w ritten w ith a striking price o f (L - M). If, fo r example, the coverage limit is set at M = $100, so that the striking price is $900, then with o = 0.3 percent, R = 8 percent and A = $985 as before, the Black-Scholes value o f this put option to the depositor is - P = -$47.96, and the total value o f his portfolio is D = 26The original limit was $2500 under the Banking Act of 1933; see FDIC (1984), pp. 44, 69. The impact of the coverage ceiling is limited by the availability of brokered deposits and by the tendency of the insurer to arrange purchase and assumption solutions to bank failures. in the United States was to have sharing in staggered pro portions [see FDIC (1984), p. 44]. Analyses of co-insurance tend to focus on proportional sharing arrangements. See Boyd and Rolnick (1989), and Benston and Kaufman (1988), ch. 3. 27There are other possibilities. For example, deposit in surance in the United Kingdom involves fixed proportional sharing combined with a coverage ceiling [see Llewellyn (1986), p. 20], and a temporary deposit insurance program FEDERAL RESERVE BANK OF ST. LOUIS 28Defining a "healthy” asset is a difficult chore. The task is generally accomplished by individual evaluation of assets, rather than the application of a generic rule. 29 Figures 11-13 the risk premium that the depositor w ould charge under this risk-sharing arrangement. On the one hand, w e can think o f the deposit as a riskless deposit combined w ith a put option. On the other hand, w e can think o f it as a single risky promise o f repaym ent from the banker, to be discounted at some risk-adjusted interest rate, r, so that the value o f the deposit is: D = L*e_rT. W e can equate these tw o inter pretations thus: D = L - e 'RT - P(L - M,A,T,o,R) = L-e_rT, w h ere P (») is the value o f the put option before expiration, as defined, fo r example, by the Black-Scholes model, and r is the risk-adjusted discount rate implied by the presence o f the coverage limit. Given the other variables, w e can rearrange this to find the risk premium: (2) (r - R) = - - i ln [ e ~ RT - P(*) ] - R. Applying this to the numerical example, the stated risk-adjusted interest yield on the deposit is: r = - ln fe" 08(,) - J ? l 796 1 L $ 1 0 00 .0 0 J = - ln(0.87515) = 13.3%, implying a risk premium, r - R, o f 5.3 percent. In practical terms, such an estimate o f the magnitude o f the risk premium implied by a given coverage ceiling might be useful in calibra ting the degree o f market discipline in a reform o f the insurance system. If bank risk-taking is to be curbed by limiting deposit insurance, forcing riskier banks to pay higher risk premia as some have suggested, then the insurance limitations must be such that the risk premium implied by equation 2 is large enough to make bankers alter their behavior.29 $932.12 - $47.96 = $884.16. In other words, given deposit insurance w ith a $100 coverage limit, $884.16 is the amount deposited at the beginning o f the year in exchange fo r a p rom ised year-end payoff o f $1000. In general, w e can use an option pricing m odel to solve algebraically fo r a measure o f 29For some examples of market discipline proposals, see Boyd and Rolnick (1989), Gilbert (1990), Gorton and Santomero (1988, 1989), or Thomson (1987). The magnitude o f the risk premium might also serve as a readily observable vital sign, registering the financial health o f the bank’s assets, and aiding the regulator in scheduling audits. This presumes that depositors have some advantage over regulators in assessing the bank’s risk bet w een audits.30 Such applications, how- the amount lent to a bank based on that bank’s risk, in addition to pricing that risk, 30Note that uninsured or co-insured depositors might find it more practical to engage in capital rationing, i. e., limiting JANUARY/FEBRUARY 1990 30 ever, are subject to some limitations which are illustrated by the next example. Figures 14-16 Deductible Another form o f co-insurance is a deductible. The case o f a deductible on insurance coverage introduces a twist to the problem. Now the depositor’s portfolio effectively consists o f tw o put options, one w ritten and one held, in addi tion to the promised repayment o f the deposit with interest. This case is o f special interest, because it applies to a deposit payout closure as considered above and because it can also be ap plied to subordinated debt, which is the object o f a recent debate on sources o f market disci pline o f bank risk-taking. In both cases, the payoffs to one o f the bank's creditors can be modeled as a pair o f put options with different striking prices.31 In this example, the depositor is promised the return o f his deposit amount, w ith accrued in terest, fo r a total o f L dollars. Because o f the deductible provision, how ever, this promised repayment is not certain; in the event o f the bank’s insolvency, the depositor w ill be the first to share in the shortfall. For year-end asset levels below L, the shortfall is deducted from the depositor's payoff until the deductible amount, U, is exhausted. Any shortfall beyond that is ab sorbed by the FDIC. Thus, the depositor effe c tively holds the deposit amount L and has w rit ten a put option w ith striking price L, that is held by the banker; in addition, he holds a put option w ith striking price (L - U), which is w ritten by the FDIC. These payoffs are illus trated in figures 14-16. The deductible provides a cushion fo r the FDIC, which, in the preceding examples, had written a put option w ith striking price L rather than (L - U). The year-end payoffs fo r the banker are the same as before. The real dif ference applies to the depositor's incentives and the resulting impact on the price he charges the banker fo r the deposit. Although he always prefers a higher asset value, as before, his at titude tow ard the riskiness o f the bank’s assets is now ambiguous, because he is long one put option and short another, with tw o different striking prices. Volatility in the bank’s asset 31The relevant creditors in each case are the depositor and the subordinated debt-holder, respectively. For an analysis of option models in the case of subordinated debt, see Black and Cox (1976) and Gorton and Santomero (1988, 1989). For an analysis of subordinated debt and bank FEDERAL RESERVE BANK OF ST. LOUIS Payoff to Depositor L-U Asset Value returns increases the value o f the long position and decreases the value o f the short position. As Black and Cox (1976) point out, the net im pact o f these countervailing forces w ill depend on the current asset value relative to the strik ing prices. Specifically, there is an inflection regulation, see Gilbert (1990). The approach here is at odds with that of Ronn and Verma (1986), who calculate the value for a single put on total debt and then scale that value down by the proportion of insured to total liabilities. 31 point equal to the discounted geom etric mean o f the tw o striking prices.32 For an asset value above the inflection point, w hich includes all cases in which the bank is solvent (i. e., A > L ’e _RT), the effect o f the short position out weighs that o f the long position, and the de positor w ill p re fer less risk. Conversely, w hen the current market value o f assets falls below the inflection point, the long position outweighs the short, and the depositor w ould p refer a riskier asset portfolio, given the low asset value. Thus, a decrease in the "risk prem ium ” charged by the depositor no longer necessarily implies that the bank's assets are less risky; fo r exam ple, such a decrease could instead be the result o f an increase in the current asset value and an increase in the volatility o f those assets. Under such circumstances, it is a reasonable taxonomic question w hether the interest rate markup over the riskless rate should be called a risk premium at all. The current value o f the depositor’s claim and the implicit risk premium can be calculated as before: value o f the deposit resembles m ore and m ore the staggered year-end p a yoff function o f figure 16. In fact, if the year-end p a y o ff o f figure 16 is scaled dow n by the risk-free present value dis count factor, e -RT, it becomes identical to the extreme case o f figure 17 w here o = 0. Figure 17 also illustrates graphically the Black and Cox argument that fo r some (higher) asset levels, de positors will charge a risk premium, while for other (lower) levels, they will offer a risk discount. It is also clear from the picture, however, that the asset level has a much more significant effect on the value o f the claim than does the volatili ty.34 All o f this suggests that bankers, depositors and policymakers should give considerable care to an appropriate definition o f risk in this context, and that similar care should be given to designing a practical measure o f that risk. Risk defined as volatility in bank asset returns and measured by the risk premium charged on equity, subor dinated debt or uninsured deposits may not be apt for the tasks to which it has been applied. D = L-e“ RT- P(L,A,T,o,R) + P (L-U ,A ,T ,o,R ) = L-e_rT SOME CAVEATS -* (r - R) = - -1 lnj^e“ RT- -1 [P (L ,«) - P(L - U ,«)] j - R, but the risk premium so defined is a measure o f the expected difference betw een cash pro mised, L, and cash ultimately received. It is a poor measure o f the volatility o f the returns on the bank’s assets, because the expected dif ference betw een cash promised and cash re ceived depends on several factors and is no longer a simple direct relation o f the volatility o f assets. Figure 17 graphs the value o f the depositor’s claim fo r a range o f asset levels and volatili ties.33 In interpreting this graph, note the con nection between it and figure 16. In particular, as the volatility, o, goes to zero in figure 17, the 32The discount rate used to calculate this inflection point in cludes a risk premium. Since all other discounting in this context is at the riskless rate, this means that the inflec tion point can fall below the present value of the lower dis count rate, (L - u)-e~RT, if the risk premium, o2/2, is large enough. For the same reason, the inflection point is always smaller than the solvency point, L-e . 33The riskless rate was set at 8 percent, time to maturity was one year, promised repayment was $1000, and the deductible amount was $200. Similar graphs with other maturities and deductible amounts reveal no surprises. The preceding analysis has illustrated some uses o f option pricing models in evaluating deposit in surance. There are some limitations, however, on the use o f options models in this context. Most of these limitations derive from the assumptions that form the basis for the option pricing equation, and the extent to which these assumptions are valid for the case at hand. Perhaps the most basic problem is the ques tion, w ho truly holds the option.35 Until now, it has been presumed, based on the end-of-period payoffs, that deposit insurance represents a put option w ritten by the FDIC and held by the banker. In fact, how ever, the FDIC decides w hether a bank is insolvent, and, m ore im por tantly, w hether to close a bank that is already insolvent (or one that is not quite insolvent).36 In the face o f a large-scale bank failure or run, 34This fact is noted by Pyle (1983), p. 13. 35This issue is addressed by Brumbaugh and Hemel (1984) and Allen and Saunders (1990). 36Recall that the definition of insolvency used in this paper ignores the possibility that the bank might be deemed in solvent on the basis of its current ratio — i. e., its inability to meet maturing liabilities with liquid assets. This problem relates to the maturity structure of the bank’s portfolio, which will be considered briefly below. JANUARY/FEBRUARY 1990 32 Figure 17 Value of Depositor’s Claim z Le RT =923.12 (L-U ) e 07 =738.49 1200 923.12 738.49 short-term political considerations may over whelm any prior prescriptions on closure poli cy. Th e FSLIC’s actions in the thrift crisis in dicate that this is not idle speculation; numerous thrifts w ere left open long after their insolvency had been discovered. Conversely, as Benston and Kaufman (1988) suggest, the FDIC could close solvent banks, if they have come close enough to insolvency. If the insurer w ere to follow scrupulously a well-defined rule one w ay or the other, there w ould be little at issue, since the striking price and year-end payoffs could then be easily adjusted within the context o f the current model. As matters stand, how ever, bankers and depositors effectively face a ran dom striking price, because the insurer decides if the option w ill be exercised. As a practical 37A rule is well-defined if it leaves no doubt about the circumstances which imply closure, with no room for FDIC discretion. Note that defining a closure rule, in general, is FEDERAL RESERVE BANK OF ST. LOUIS matter, it is difficult to envision how such a well-defined closure rule might be im plem ented.37 A related issue is the measurement o f bank asset values. The option pricing m odel p re sented above presumes that the current asset price can be readily observed. For stock op tions, this is an uncontentious assumption, because stock prices can be observed on the floor o f the exchange or in the over-the-counter market. For an option on the assets o f the bank, how ever, the relevant price is not readily obser vable. Indeed, one o f the prim ary functions o f bank credit analysis is to assign values to assets fo r which there is no active market. Similarly, not sufficient for our purposes; the closure rule must be defined so that the values of the resultant claims conform to the values given by the option pricing equation. 33 the insurer must invest significant effort, in the form o f an audit, to determine the year-end asset value. that this choice is a salient factor in the option’s value, because it determines the probability o f each o f the possible year-end payoffs. The inherent inaccessibility o f bank asset values has tw o implications fo r option models. First, it is no longer possible fo r an option holder to construct the appropriately hedged portfolio described in the Black and Scholes derivation, because the hedge ratio depends on the value o f the underlying asset. This casts doubt upon the appropriateness o f the riskless rate in discoun ting the expected end-of-period payoffs.38 Se cond, the current asset value is important, be cause it partly determines the probabilities fo r the various end-of-period payoffs. Ignorance o f the current asset value adds another layer o f uncertainty, and this additional uncertainty significantly affects the value o f the option.39 A closely related issue is the measurement o f asset risk. The option pricing models presented here use the variance o f the asset’s returns as a measure o f risk.40 Producing an accurate assess ment o f the variance is problematic, even fo r stock options, because the volatility that matters is the variance o f the process over the fu tu re life o f the option. For bank assets, the measure ment problem is compounded, because even past values are generally unavailable. Pyle (1983) and Flannery (1989b) consider some o f the im plications o f this problem in using the option models to price deposit insurance. The empirical evidence to date testifies to the sensitivity o f the results to the specification employed. Marcus and Shaked (1984) use the basic Black-Scholes model, adjusted fo r divi dends, and find that federal deposit insurance is currently substantially overpriced relative to the "actuarially fair” estimates provided by their option model.41 They note, how ever, that "McCulloch’s (1981, 1983) estimates o f insurance values derived from the Paretian-stable distribu tion greatly exceed” their ow n.42 Pennacchi (1987b) uses a m ore complicated model which includes the degree o f regulatory control w ield ed by the insurer. He finds that deposit insur ance may be either overpriced or underpriced, depending on the level o f regulation assumed. McCulloch’s (1985) study assumes non-normal, Paretian-stable asset returns and non-stationary random interest rates. He finds that insurance values are highly sensitive to the level and volatility o f interest rates. Ronn and Verma (1986), how ever, using a variant o f M erton’s (1977) model, conclude that neither random in terest rates nor non-stationary equity returns significantly affect the insurance valuations. In brief, the empirical evidence suggests that a w ide range o f insurance valuations can be reached by varying the returns process em ployed in the model. Just as asset values and the volatility o f re turns are not observable directly, there is the m ore general moot question o f which stochastic returns-generating process should be incor porated in the option pricing model. As w e ’ve seen above, the difference in the assumed re turns process betw een Black and Scholes’s model and M erton’s model resulted in a substantially different pricing equation. Although the choice o f an appropriate returns process fo r modeling a bank’s assets is beyond the scope o f this paper, it is sufficient to note Finally, it has been assumed in the preceding examples that there is a single deposit that does not mature until the end o f the year. In fact, o f course, banks maintain many deposit accounts w ith a w ide range o f maturities, starting with the instant maturity o f demand deposits. This is significant, because it gives many depositors another type o f insurance. A depositor w ho can w ithdraw his funds from a failing bank sooner than the FDIC can close it has 100 percent in surance, regardless o f the balance in the ac count or the insurance scheme in effect.43 The 38One might resort to the argument that the riskless rate is appropriate if the asset has no idiosyncratic (that is, firmspecific) risk, as noted in the context of Merton’s (1976) pricing model (see the shaded insert), but such an assumption is not particularly credible for bank asset port folios, prima facie. 39See Pyle (1983) for an analysis of asset value uncertainty. Figlewski (1989) and Babbel (1989) consider the im possibility of hedging, along with many other difficulties in the application of stock option pricing models. 41For an exposition of the dividend adjustment to the BlackScholes model, see Merton (1973). 42Marcus and Shaked (1984), p. 449. Their reference to McCulloch (1983) was a working paper, later published as McCulloch (1985). 43lt need only be the case that the bank funds the deposit outflow somehow, for example, through a fire sale of its assets. Such a run on Continental Illinois by institutional depositors prompted the FDIC to extend 100 percent in surance to uninsured depositors. 40Merton’s (1976) approach uses that variance together with the parameters of the jump distributions. JANUARY/FEBRUARY 1990 34 result is that the simple application o f an option pricing model does not provide an accurate evaluation o f such deposits, because it ignores certain relevant strategies. CONCLUSIONS The usefulness o f option models in the study o f deposit insurance results from tw o important characteristics. First, these models distill the host o f economic factors involved down to a handful o f relevant parameters whose interac tion is well-defined by the option-pricing equa tion. Second, they are able to evaluate deposit insurance claims under a w ide variety o f in surance structures. Although only three such structures w ere elaborated here, they can be generalized to other applications. Thus, option models provide a unified context fo r analyzing incentives w ithin an insurance structure, as w ell as fo r comparing alternative insurance schemes. Unfortunately, option pricing models, like most economic models, are an im perfect tool when directly applied to the complexities o f the real w orld. Beyond certain fundamental qualita tive results, there are theoretical and empirical reasons to believe that the insurance valuations given by any particular option pricing model w ill be incorrect, highly sensitive to changes in their specification, or both. As a result, the ab solute dollar magnitudes provided by options models o f the value o f deposit insurance are suspect. The contradictory empirical evidence on fair pricing is indicative o f this problem. In defense o f these models, how ever, there is no reason to believe that option models are any w orse in this regard than any alternative economic model. Indeed, there is some reason to believe that, although the absolute magnitude o f the valuations provided by option models may be unstable, the rankings they provide fo r a sample o f banks are not.44 Similarly, inac curacies in determining the scale o f insurance values do not deny the ability o f option models to identify the direction o f incentives or the im pact o f marginal changes in the structure o f deposit insurance. Th erefore, used judiciously, option pricing models can be an effective ana lytical tool in the study o f deposit insurance. "T h e rank-order correlation of risk measures for a sample of banks over time [Marcus and Shaked, (1984), p. 455] and over alternative risk metrics [Ronn and Verma, (1987), FEDERAL RESERVE BANK OF ST. LOUIS REFERENCES Allen, Linda, and Anthony Saunders. "Forbearance and Deposit Insurance Valuation,” Working Paper, Baruch Col lege, City University of New York, January 1990. Babbel, David F. “ Insuring Banks against Systematic Credit Risk,” Journal of Futures Markets (December 1989), pp. 487-505. Barnea, Amir, Robert A. Haugen, and Lemma W. Senbet. Agency Problems and Financial Contracting (Prentice-Hall, 1985). Barnett, Robert E., Paul M. Horvitz, and Stanley C. Silverberg. “ Deposit Insurance: The Present System and Some Alternatives,” The Banking Law Journal (April 1977), pp. 304-32. Benston, George J., and George G. Kaufman. “ Risk and Solvency Regulation of Depository Institutions: Past Policies and Current Options,” Monograph Series in Finance and Economics, Salomon Brothers Center for the Study of Financial Institutions, New York, 1988. Black, Fischer, and John C. Cox. “ Valuing Corporate Securities: Some Effects of Bond Indenture Provisions,” Journal of Finance (May, 1976), pp. 351-67. Black, Fischer, and Myron Scholes. “ The Pricing of Options and Corporate Liabilities,” Journal of Political Economy (May/June, 1973), pp. 637-54. Boyd, John H., and Arthur J. Rolnick. “A Case for Reforming Federal Deposit Insurance,” Federal Reserve Bank of Min neapolis 1988 Annual Report (1989). Brumbaugh, Dan, and Eric Hemel. “ Federal Deposit In surance as a Call Option: Implications for Depository In stitution and Insurer Behavior,” Working Paper No. 116, Federal Home Loan Bank Board, Washington, D.C., Oc tober 1984. Cox, John C., and Stephen A. Ross. “ The Valuation of Op tions for Alternative Stochastic Processes,” Journal of Financial Economics (January/March 1976), pp. 145-66. Cox, John C., and Mark Rubinstein. Options Markets (Prentice-Hall, 1985). Federal Deposit Insurance Corporation. Deposit Insurance in a Changing Environment, (FDIC, Washington, D.C., 1983). _______ . Federal Deposit Insurance Corporation: The First Fifty Years, (FDIC, Washington, D.C., 1984). _______ . 1988 Annual Report, (FDIC, Washington, D.C., 1989). Figlewski, Stephen. “ What Does an Option Pricing Model Tell Us About Option Prices?” Financial Analysts Journal (September/October 1989), pp. 12-15. Flannery, Mark J. “ Capital Regulation and Insured Banks’ Choice of Individual Loan Default Risks,” Journal of Monetary Economics (September 1989a), pp. 235-58. _______ . “ Pricing Deposit Insurance When the Insurer Measures Bank Risk with Error,” Working Paper, School of Business, University of North Carolina, May 1989b. Furlong, Frederick T., and Michael C. Keeley. “ Does Capital Regulation Affect Bank Risk-taking,” Working Paper No. 87-08, Federal Reserve Bank of San Francisco, 1987. Gilbert, R. Alton. “ Market Discipline of Bank Risk: Theory and Evidence,” this Review (January/February 1990), pp. 3-18. p. 511] indicate that the stability of the rankings is signifi cant but imperfect. Further research along these lines would be welcome. 35 Gorton, Gary, and Anthony M. Santomero. “ The Market’s Evaluation of Bank Risk: A Methodological Approach,” in: Proceedings of a Conference on Bank Structure and Com petition, (Federal Reserve Bank of Chicago, May 1988), pp. 202-18. _______ . “ Market Discipline and Bank Subordinated Debt,” Working Paper, Wharton School, University of Penn sylvania, August 1989. Huertas, Thomas F., and Rachel Strauber. “An Analysis of Alternative Proposals for Deposit Insurance Reform,” Ap pendix E to Hans H. Angermueller’s testimony in: Structure and Regulation of Financial Firms and Holding Companies (Part 3): Hearings before a Subcommittee of the Committee on Government Operations, U.S. House of Representatives, 99th Cong., 2 Sess., December 17-18, 1986 (GPO, 1987), pp. 390-463. LLewellyn, D. T. The Regulation and Supervision of Financial Institutions, Gilbart Lectures on Banking, (Institute of Bankers, London, 1986). McCulloch, J. Huston. “ Interest Rate Risk and Capital Ade quacy for Traditional Banks and Financial Intermediaries,” in S. J. Maisel, ed., Risk and Capital Adequacy in Commer cial Banks, National Bureau of Economic Research, 1981, pp. 223-48. _______ . “ Interest-risk Sensitive Deposit Insurance Premia: Stable ACH estimates,” Journal of Banking and Finance (March 1985), pp. 137-56. Malliaris, A. G. “ Ito’s Calculus in Financial Decision Mak ing,” SIAM Review (October 1983), pp. 481-96. Marcus, Alan J. “ Deregulation and Bank Financial Policy,” Journal of Banking and Finance (December 1984), pp. 557-65. Marcus, Alan J., and Israel Shaked. "The Valuation of FDIC Deposit Insurance Using Option-pricing Estimates,” Journal of Money, Credit and Banking (November 1984), pp. 446-60. Merton, Robert C. “ Theory of Rational Option Pricing,” Bell Journal of Economics and Management Science (Spring, 1973), pp. 141-83. Osterberg, William P., and James B. Thomson. “ Capital Re quirements and Optimal Bank Portfolios: A Reexamina tion,” Working Paper No. 8806, Federal Reserve Bank of Cleveland, 1988. Pennacchi, George G. “Alternative Forms of Deposit In surance,” Journal of Banking and Finance (June 1987a), pp. 291-312. _______ .“A Reexamination of the Over- (or Under-) Pricing of Deposit Insurance,” Journal of Money, Credit and Bank ing (August 1987b), pp. 340-60. Pyle, David H. “ Pricing Deposit Insurance: The Effects of Mismeasurement,” Working Paper No. 83-05, Federal Reserve Bank of San Francisco, October 1983. _______ . “ Deregulation and Deposit Insurance Reform,” Economic Review, Federal Reserve Bank of San Francisco (Spring 1984), pp. 5-15. _______ . “ Capital Regulation and Deposit Insurance,” Jour nal of Banking and Finance (June 1986), pp. 189-201. Ronn, Ehud I., and Avinash K. Verma. “ Pricing RiskAdjusted Deposit Insurance: An Option-Based Model,” Journal of Finance (September 1986), pp. 871-95. _ _ _ _ _ . “A Multi-Attribute Comparative Evaluation of Relative Risk for a Sample of Banks,” Journal of Banking and Finance (September 1987), pp. 499-523. Rubinstein, Mark. “ Nonparametric Tests of Alternative Option Pricing Models Using All Reported Trades and Quotes on the 30 Most Active CBOE Option Classes from August 23, 1976 Through August 31, 1978,” Journal of Finance (June 1985), pp. 455-80. Sinkey, Joseph F., Jr., and James A. Miles. “ The Use of Warrants in the Bail Out of First Pennsylvania Bank: An Application of Option Pricing,” Financial Management (Autumn 1982), pp. 27-32. Stoll, Hans R. “ The Relationship Between Put and Call Op tion Prices,” Journal of Finance (December 1969), pp. 802-24. Symons, Edward L., Jr., and James J. White. Banking Law (West Publishing, 1984). _______ . “ Option Pricing When Underlying Stock Returns Are Discontinuous,” Journal of Financial Economics (January/March 1976), pp. 125-44. Thomson, James B. “ The Use of Market Information in Pric ing Deposit Insurance,” Journal of Money, Credit and Bank ing (November 1987), pp. 528-37. _______ . “An Analytic Derivation of the Cost of Deposit In surance and Loan Guarantees: An Application of Modern Option Pricing Theory,” Journal of Banking and Finance (June 1977), pp. 3-11. U.S. Treasury Department, The Working Group of the Cabinet Council on Economic Affairs. “ Recommendations for Change in the Federal Deposit Insurance System” (GPO, 1985). _______ . “ On the Cost of Deposit Insurance When There Are Surveillance Costs,” Journal of Business (July 1978), pp. 439-52. White, Lawrence J. “ The Reform of Federal Deposit In surance,” Journal of Economic Perspectives (Fall 1989), pp. 11-29. JANUARY/FEBRUARY 1990 36 Cletus C. Coughlin and Kees Koedijk Cletus C. Coughlin is a research officer at the Federal Reserve Bank of St. Louis. Kees Koedijk is a professor of economics at Erasmus University, Rotterdam. Thomas A. Pollmann provided research assistance. What Do We K now About the Long-Run Real Exchange Rate? A REAL EXCHANGE rate is defined as the foreign currency price o f a unit o f domestic currency (that is, the nominal exchange rate) multiplied by the ratio o f the domestic to the foreign price level. The real exchange rate has been at the center o f economic policy discus sions in the 1980s fo r at least tw o reasons. First, this relative price has been m ore variable in the floating-rate period than in the preceding era o f fixed (nominal) exchange rates.1 Second, this price is related to international trade pat terns because the competitive position o f an in dividual exporting (import-competing) firm in a 'See Frankel and Meese (1987) and Dornbusch (1989) for surveys of this literature. As noted by Dornbusch, the in creased variability and lack of knowledge have contributed to divergent policy recommendations, which include a return to some form of a managed exchange-rate system, taxes on foreign exchange transactions as well as doing nothing. 2The U.S. dollar has been at the center of the controversy, with the dollar allegedly being undervalued in the late 1970s/early 1980s and overvalued in the mid-1980s. Dur ing the period of undervaluation, U.S. tradeable goods in dustries were stimulated and induced to overexpand. The costs of this alleged overexpansion were exacerbated by the subsequent overvaluation, which resulted in layoffs, plant closings and bankruptcies in these same industries. 3This elementary principle is ignored when the exchange rate in macroeconomic settings is treated as an ex FEDERAL RESERVE BANK OF ST. LOUIS country is affected adversely by an appreciating (depreciating) real exchange rate.2 Despite much research, h ow ever, there is no consensus on w hich variables cause changes in the real exchange rate. Like any asset price, real exchange rates are related to the determ i nants o f the relevant supply and demand curves now and in the future.3 W ith real exchange rates, the relevant determinants are those affec ting the relative supplies and demands fo r the currencies o f tw o countries. Claims have been made, how ever, that the real exchange rates ogenous rather than endogenous variable. For example, a standard assertion is that a depreciating dollar boosts U.S. manufacturing output. A declining dollar is expected to raise the dollar prices of U.S. imports and lower the foreign currency prices of U.S. exports. Consequently, consumption and production of U.S. exports and importcompeting goods would rise. This analysis is faulty because changes in the value of the dollar are not in dependent of U.S. industrial developments and, in fact, can be the direct result of industrial developments. For ex ample, an economic policy that boosts productive capacity can generate a positive relationship between the value of the dollar and U.S. manufacturing output. Details on this argument can be found in Tatom (1988). 37 often d iffer substantially from levels consistent w ith the underlying econom ic fundamentals and that these differences persist fo r long periods. A prim ary goal o f our research is to provide an elem entary understanding o f the major theo retical approaches to the determination o f longrun real exchange rates. These approaches iden tify numerous variables that have been tested fo r their relationships to the changing values o f the real exchange rate. Empirically, w e examine the six bilateral real exchange rates among the United States, W est Germany, Japan and the United Kingdom.4 Using a data set covering ap proxim ately the same time period, w e make a straightforw ard comparison o f the three p ri mary approaches and present a clear picture o f what can be said about the determinants o f real exchange rates. Research to explain movements in the longrun real exchange rate is unnecessary if pur chasing p ow er parity (PPP) holds in the long run. Thus, w e begin by review in g the literature on PPP in the long run. This provides a natural starting point from which to examine the dif ferent theoretical approaches to real exchange rate determination and the major empirical find ings. Next, w e undertake unit root and cointe gration tests to examine w hether long-run rela tionships exist betw een the real exchange rate and some o f its potential determinants. IS THE R EAL EXCHANGE RATE A R A N D O M W ALK? As a point o f departure, it is useful to define the real exchange as it is used throughout this paper. A standard representation expresses all variables in logarithms, so that a real exchange 4Our selection of countries is based on research by Koedijk and Schotman (1989), which indicates that the movements of real exchange rates for 15 industrial countries can be partitioned into four groups led by the United States, West Germany, Japan and the United Kingdom. 5Wholesale price indexes are also frequently used in the calculation of real exchange rates. The use of wholesale rather than consumer prices can generate different results. For an example, see McNown and Wallace (1989). For a brief discussion of why a broad-based measure of prices such as consumer prices is more appropriate than one of wholesale prices in calculating real exchange rates, see Cox (1987). rate, q, is defined as follows: (1) q = e + p - p * , w h ere e is the foreign currency price o f a unit o f domestic currency, p is the domestic price level as measured by the consumer price index and p* is the similarly measured foreign price level.5 Since the advent o f flexible exchange rates in 1973, real exchange rates have been m ore vari able than they w ere previously. This point is il lustrated in figure 1 over 1957 to 1988 fo r the pound/dollar, mark/dollar and yen/dollar real ex change rates. The increased variability has in duced many researchers to focus on the fun damental relationships that determine real ex change rates. The concept o f purchasing p ow er parity has been one o f the most important building blocks fo r nominal, as w ell as real, exchange rate modeling during the 1970s and 1980s. In its absolute version, PPP states that the equilibrium value o f the nominal exchange rate betw een the currencies o f tw o countries w ill equal the ratio o f the countries’ price levels.6 Thus, a deviation o f the nominal exchange rate from PPP has been view ed as a measure o f a curren cy’s over/undervaluation. In its relative version, PPP states that the equilibrium value o f the nominal exchange rate w ill change according to the relative change o f the countries’ price levels. A notew orthy implication o f both versions o f PPP is that the real exchange rate w ill remain constant over time. Economists have debated w hether PPP applies in the short run, long run or neither. By the end o f the 1970s, PPP, at least in the short run, was rejected convincingly by the data.7 W hether PPP in the long run can be rejected is less clear. A standard theoretical argument in support o f rate (domestic currency value per unit of foreign currency), P is an index of domestic prices, P* is an index of foreign prices and K is a scalar. In this view, the PPP hypothesis is a homogeneity postulate of monetary theory rather than an arbitrage condition. Thus, a monetary disturbance causes an equiproportionate change in money, commodity prices and the price of foreign exchange, while relative prices are unchanged. The influence of real factors on the relationship between exchange rates and national price levels is captured by K, which is a function of structural factors that can alter the relative prices of goods. 7See Adler and Lehmann (1983) for the references underly ing this consensus. 6More generally, PPP has been stated in Edison and Klovland (1987) as E = K(P/P*), where E is the exchange JANUARY/FEBRUARY 1990 38 Figure 1 Real Bilateral Exchange Rates in the Fixed and Floating Rate Periods 1957 59 61 63 65 67 FEDERAL RESERVE BANK OF ST. LOUIS 69 71 73 75 77 79 81 83 85 87 1989 39 PPP is that deviations from parity, assuming zero transportation costs and no trade barriers, indicate profitable opportunities fo r commodity arbitrage. Deviations from PPP imply that the same good, after adjusting fo r the exchange rate, w ill sell at different prices in tw o loca tions. Simultaneously buying the good in the low-price country and selling the good in the high-price country w ill force the nominal ex change rate to PPP and the real exchange rate to some constant value. Th e chief issue is w hether the real exchange rate returns over time to a fixed value, the long-run equilibrium real exchange rate. On the other hand, it is possible that the equilibrium value o f the real exchange rate is not constant over time, but instead changes in response to changes in some fundamental economic vari ables. For example, an increase in a country’s real interest rate, ceteris paribus, could cause an appreciation o f the country’s real exchange rate. One conclusion, how ever, is clear: if the real exchange rate follow s a random walk, long-run PPP does not hold. A variable is said to follow a random walk if its value in the next period equals its value in the current period plus a random error that cannot be forecast using available information. I f the real exchange rate follows a random walk, then it w ill not return to some average value associated with PPP over time. In fact, its deviation from the PPP value becomes unbounded in the long run. The unit root test is a common procedure to use in determining w hether a variable follow s a random walk.8 I f the existence o f a unit root eThe issue is whether the real exchange rate is stationary. If the real exchange rate is stationary, then random distur bances have no permanent effects on this rate. If the real exchange rate is nonstationary, then there is no tendency for this rate to return to an “ average” value over time. To determine whether the real exchange rate is stationary, a standard procedure is to use the Dickey-Fuller test for unit roots. This procedure is described later in the text. 9Examples include Cumby and Obstfeld (1984) and Frankel (1986). Using monthly data between September 1975 and May 1981, Cumby and Obstfeld rejected the random-walk hypothesis for the real exchange rate between the United States and Canada. On the other hand, they were unable to reject the random-walk hypothesis for the real exchange rate when the United States was paired with each of the following countries— United Kingdom, West Germany, Switzerland and Japan. Frankel (1986) rejected the random-walk hypothesis for the real U.S. dollar/British pound exchange rate using annual data between 1869 and 1984, but was not able to reject the hypothesis using data for 1945-1984. cannot be rejected, then the variable is said to follow a random walk. Using data from various developed countries, recent studies by Darby (1983), A dler and Lehmann (1983), Huizinga (1987), Baillie and Selover (1987) and Taylor (1988) could not reject the unit root hypothesis fo r the real exchange rate in the current float ing rate period and, hence, rejected the notion o f long-run PPP. The issue, nevertheless, remains controversial. One reason is that some researchers have found evidence to reject the random-walk hypothesis in some cases.9 In addition, doubts about the p o w er o f standard tests to discriminate betw een true random walks and near random walks have been raised. For example, Hakkio (1986) demonstrated that, w hen the real exchange rate differs modestly from a random walk, the re sults o f standard tests are biased in favor o f the random-walk hypothesis. In other words, there is a high probability o f failing to reject the ran dom-walk hypothesis even if it is false.10 Another possibility is that the current floatingrate period is too b rie f to assess accurately the validity o f PPP. Lothian (1989), using unit root tests and annual data fo r over 100 years for Japan, the United States, the United Kingdom and France, found that real exchange rates tended to return to their long-run equilibrium values, but that the period o f adjustment was quite long. For example, adjustment periods ranging from three to five years w ere found. Consequently, the current floating-rate period might not be long enough to identify the longrun tendency o f the real exchange rate to re turn to an equilibrium. 10The preceding problem motivated Sims (1988) to develop a new test for discriminating between true and near ran dom walks. Applying this new test, Whitt (1989) was able to reject the hypothesis that the real exchange rate was a random walk. A forthcoming issue of the Journal of Econometrics, however, concludes that the ap propriateness of Bayesian approaches in detecting unit roots remains in doubt because many questions, some re quiring highly technical responses, have not been answered. Consequently, we did not use this technique in our analysis. JANUARY/FEBRUARY 1990 40 APPR O A C H E S T O R EAL EX CHANGE R ATE D ETER M IN ATIO N IN THE LONG RUN Our goal is not to resolve the preceding con troversy about PPP in the long run. Rather, it is to examine, as w ell as extend empirically, the research efforts o f those w ho have provided models that allow fo r the long-run real ex change rate to vary over time. In other words, our goal is to examine the attempts b y resear chers skeptical about PPP in the long run to ex plain movements in the long-run real exchange rate. T w o real approaches and a m onetary ap proach to exchange-rate determination have been used to explain movements in the equilib rium real exchange rate. The first real approach is concerned w ith movements in the real ex change rate that arise from incorporating the difference betw een tradeable and non-tradeable goods prices. Th e other real approach deals w ith the implications o f incorporating a balance o f payments constraint. The m onetary ap proach, in contrast, focuses on the relationship betw een real exchange rates and real interest rates. Tradeables and Non-Tradeables Absolute PPP implies that the equilibrium value o f the nominal exchange rate betw een the currencies o f tw o countries w ould equal the ratio o f the countries’ price levels, w hich is commonly measured by the respective consum er price indexes. Ignoring transportation costs, free international trade eliminates the price dif ference betw een the same good in tw o coun tries; how ever, price differences across coun tries fo r non-traded goods may persist and may change substantially over time. Frequently, this possibility is referred to as PPP holding only fo r internationally traded goods; how ever, one could view this possibility as a substantial m odi fication o f PPP. T o prevent confusion, w e do not call this PPP, but rather characterize it as the law o f one price fo r traded goods. A simple model illustrating this approach is presented below. Let p be the logarithm (log) o f "T h e se indexes suggest that the price level is constructed as follows: P = P‘ ‘ ‘ *| p *T, where the upper-case Ps repre sent levels. Price indexes are not really constructed this way; however, following Hsieh (1982), this construction was chosen to simplify the derivation. Hsieh has argued that his empirical results were not distorted by this assumed construction because he used highly aggregated data. FED ERAL RESERVE BANK OF ST. LOUIS the overall price level, and pT and pNT be the logs o f the price levels o f traded and non-traded goods; an asterisk denotes the foreign country. The overall price level is related to the prices o f tradeable and non-tradeable goods by (2) p = (1 - a)pT + apNT and (3) P * = ( i - p)P; + /?P * T, w h ere a and ft denote the shares o f the nontradeable goods sectors in the economies.11 Assuming the law o f one price fo r tradeable goods, (4) e + pT - p* = 0, w h ere e is the log o f the nominal exchange rate, measured as the foreign currency price o f a unit o f domestic currency.12 By substituting equations 2, 3 and 4 into equa tion 1, the real exchange rate, q, can be w ritten as (5) q = - a ( p T - pNT) + P (p ; - p *T). Thus, the real exchange rate depends on rela tive prices betw een tradeable and non-tradeable goods as w ell as the sizes o f the non-tradeable goods sectors in the tw o countries. Our focus is restricted to the possibility that persistent d if ferences betw een the price changes o f tradeable and non-tradeable goods across the tw o economies can cause real exchange rate movements. T w o main proxies, one using relative prices, the other using output measures, have been us ed to measure the tradeables/non-tradeables distinction. As W o lff (1987) has noted, a stan dard empirical proxy in analyzing relative prices in a w orld with internationally traded and nontraded goods is the ratio o f wholesale prices to consumer prices. The reasoning is straightfor ward. Wholesale price indexes generally pertain to baskets o f goods that contain larger shares o f traded goods than consumer price indexes do. Consumer price indexes tend to contain relative ly larger shares o f non-traded consumer ser vices. T o date, how ever, empirical evidence on 12As a check, using wholesale prices for the prices of traded goods, we found that e + pT- p* was not stationary. Thus, one of the building blocks for this approach does not hold for our data. In addition, even if one were to define the real exchange rate using wholesale rather than consumer prices, PPP would not appear to hold in the long run. 41 the importance o f relative prices in explaining real exchange rate movements is lacking. The Real Approach Using the Balance-of-Paym ents Equilibrium The other proxy fo r the tradeables/nontradeables distinction was highlighted by Balassa (1964). Balassa assumed that the law o f one price held fo r traded goods, that wages in the tradeable goods sector are linked to productivity and that wages across industries are equal. These assumptions cause the price o f non-tradeable goods relative to tradeable goods to in crease m ore over time in a country w ith high productivity grow th in the tradeable goods sec tor than in a country with low productivity growth. Such a productivity differential, in con junction with a general price index that covers both traded and non-traded goods, w ill result in a real exchange rate appreciation fo r fastgrowing countries even with the prices o f traded goods equalized across countries. An alternative real approach to analyze m ove ments in the real exchange rate is to include a balance-of-payments constraint.13 This approach focuses on the theoretical relationship between changes in the equilibrium real exchange rate and changes in the current account. The longrun equilibrium real exchange rate is the rate that equilibrates the current account in the long run. Recall that balance-of-payments accounting ensures that the current account is identical to the negative o f the capital account, which is simply the rate o f change o f net foreign hold ings. Thus, the current account equilibrium in the long run is determined by the rate at which foreign and domestic residents wish to change their net foreign asset positions in the long run. For the empirical application o f the productivi ty approach, Balassa suggested that there should be a positive link betw een the real exchange rate and real per capita gross national product, which assumes that inter-country productivity differences are reflected in per capita income levels. The effect o f shifts in sectoral productivi ty have been investigated by Hsieh (1982) and Edison and Klovland (1987). Hsieh found that real exchange-rate changes fo r W est Germany and Japan could be explained by differences in the relative grow th rates o f labor productivity betw een traded and non-traded sectors fo r these countries and their major trading partners. Similarly, Edison and Klovland, using annual data, found a long-run equilibrium relation be tw een the pound/Norwegian krone real ex change rate and the real output differential and betw een the real exchange rate and the com modity/service productivity ratio differential. The results o f Edison and Klovland raise a number o f interesting questions because the data cover a period that is both long, 1874-1971, and does not encompass the current floatingrate period. Consequently, one is left w ondering w hether 15 years o f data, which require the use o f data m ore frequent than annual observa tions, is sufficient to reach strong conclusions about the current period and w hether Edison and Klovland’s results w ould be altered by data from the current period. Any fundamental econom ic factor that in fluences the current account affects the real ex change rate. Consequently, the long-run equilib rium real exchange rate depends on real fac tors—whose changes can either be anticipated or unanticipated—that cause shifts in the de mand fo r and supply o f domestic and foreign goods. The most notable example is the relative output differential. Relatively faster output grow th domestically w ill induce an appreciation o f the long-run equilibrium real exchange rate. A key aspect o f this approach focuses on the possibility that unanticipated changes in the cur rent account affect the long-run real exchange rate. Unexpected changes in the current ac count are assumed to reflect changes in under lying determinants that, in turn, require offset ting changes in the real exchange rate to ensure current account equilibrium in the long run. A long-run balance-of-payments constraint sug gests that any revisions in expectations about the long-run values o f variables that affect the balance o f payments affect the expected value o f the long-run real exchange rate. As Isard (1983) notes, the substantial changes in the rela tive price o f oil during the 1970s are excellent ex amples o f how unexpected changes in a determi nant o f the current account caused revised expec tations about the long-run real exchange rate. An illustration highlighting the importance o f unanticipated current account changes is pre- 13Examples may be found in Isard (1983) and Frenkel and Mussa (1985). JANUARY/FEBRUARY 1990 42 sented by Dornbusch and Fischer (1980). In their model, a current account surplus causes a rise in wealth through the net in flow o f foreign assets. Assuming the rise in wealth is unantici pated, excess demand in the domestic goods market occurs. In turn, an increase in the real exchange rate is required fo r the new goods market equilibrium. This increase induces the necessary shift from domestic to foreign goods by domestic and foreign consumers to eliminate the excess demand. H ooper and M orton (1982) use this fram ew ork to relate changes in the real exchange rate to economic fundamentals. They use the cumu lated current account as a determinant o f the long-run equilibrium real exchange rate. In their model, unanticipated changes in the current ac count are assumed to provide inform ation about shifts in the underlying determinants that ne cessitate offsetting shifts in the real exchange rate to maintain current account equilibrium in the long run. Consistent with this balance-ofpayments approach, their results indicate that, betw een 1973 and 1978, movements in the cur rent account have been a significant determ i nant o f movements in the real exchange rate fo r the U.S. dollar, predominantly through changes in expectations. The M onetary Approach As m entioned previously, the m onetary ap proach focuses on the relationship betw een real exchange rates and real interest rates. A straightforw ard exposition o f this approach, which can be found in Meese and R ogoff (1988), is based on models developed by Dornbusch (1976), Frankel (1979) and Hooper and M orton (1982). These models are "sticky-price” versions o f the m onetary m odel o f exchange rates; they assume that prices o f all goods adjust slowly in response to disturbances. Thus, tem porary devi ations in the real exchange rate from its longrun equilibrium value (that is, purchasing p ow er parity) are possible. These tem porary deviations necessitate an exchange-rate adjustment mechanism to restore the long-run equilibrium value. A standard as 14For example, a comparison of 0= .6 with 0 = .4 after two periods reveals that, in the former case, the expected dif ference between the actual and long-run equilibrium is .36 of the difference in the current period, while in the latter case the expected difference is .16 of the difference in the current period. FEDERAL RESERVE BANK OF ST. LOUIS sumption is that the deviations are eliminated at a constant rate. The adjustment process can be represented as follows: (6) E,(qt+k - q t+k) = 0k(q, - qt), O < 0 < 1 , w h ere E is an expectations operator, the sub scripts designate the time period, q is the loga rithm o f the real exchange rate, the bars in dicate values that w ould prevail if all prices w ere fully flexible instantaneously and 0 is the speed-of-adjustment parameter. Consequently, there is a monotonic adjustment o f the real ex change rate to the long-run equilibrium, q „ over time with low er values o f 0 indicating a quicker adjustment process.14 Th e long-run equilibrium value changes with random real shocks; how ever, assuming all real shocks fo llow random-walk processes, these shocks do not affect the expected long-run equi librium exchange rate. Consequently, (7) Etqt+k = qt. Substituting equation 7 into equation 6 yields (8) q, = d(E,qt+k - qt) + q„ w h ere 6 = l/(0k - 1) < - 1 . The observed real exchange rate is its tem porary deviation from its long-run equilibrium value plus its long-run equilibrium value. T o complete the model, uncovered interest parity is assumed.15 This assumption is express ed as follows: (9) Etet+k - e, = kr* - kr„ w h ere e is the logarithm o f the nominal ex change rate (foreign currency per domestic cur rency unit), kr t is the k-period nominal interest rate at time t and the asterisk denotes a foreign value. In other words, changes in the nominal exchange rate are directly related to nominal in terest rate differentials. As domestic nominal in terest rates rise relative to foreign rates, the nominal exchange rate o f the domestic country is expected to depreciate. Equation 9 implies that the expected change in the real exchange rate reflects the expected real interest rate differential. In symbols, 15The appropriateness of this assumption can be ques tioned. The uncovered interest parity assumption requires that the forward rate be an unbiased and efficient predic tor of the future spot rate; however, the empirical results summarized by Baillie and McMahon (1989) suggest otherwise. 43 (10) E,(qt+k - q.) = kRt* - kR„ w here the k-period interest rate, kR„ is the dif ference betw een the nominal interest rate less the expected rate o f change in prices. Substi tuting equation 10 into equation 8 yields (11) q, = d(kR* - kRt) + qt. Th erefore, the essence o f the monetary ap proach is that changes in the real exchange rate are directly related to changes in the real in terest differential. As expected real domestic in terest rates rise relative to foreign rates; the real exchange rate o f the domestic country rises as well. Equation 11 provided the foundation fo r vari ous statistical tests by Meese and R ogoff (1988). As noted in the appendix, the measurement o f expected real interest rates is problematic. W hile the sign o f the relationship betw een the long term real interest rate differential and the real exchange rate was consistent with theory, the relationship was not statistically significant. A N EM PIR ICAL AN ALY SIS OF REAL EXCHANGE RATES AN D PO T E N T IA L DETERM INANTS Our empirical analysis proceeds in tw o steps. First, using unit root tests, w e test fo r the stationarity o f the six real exchange rates that result from pairwise combinations o f the fo r eign exchange rates o f the United States, Japan, W est Germany and the United Kingdom. The stationarity o f five potential determinants fo r these exchange rates is examined as well. D e tails on the construction o f these variables are presented in the appendix. Unless noted other wise, w e used monthly data from June 1973 to June 1988 fo r all variables. Thus, in the first step, w e provide additional evidence on the ex istence o f PPP in the long run. Second, w e test fo r cointegration betw een the real exchange rates and each o f the potential determinants. The goal is to identify which variables, if any, from the models that w e have review ed explain variations in the real exchange rate over time. Testing f o r Unit R oots Potential Determinants In summary, the existing literature points to five potential determinants o f long-run real ex change rates that w e use. The real approach identifies three possibilities, tw o based on the tradeables/non-tradeables distinction and one based on the balance-of-payments equilibrium. The proxies to measure the tradeables/nontradeables distinction have used the ratio o f wholesale to consumer prices and real per cap ita gross national product differences, w hile the cumulated current account difference is used fo r the balance o f payments. The other major approach, the m onetary approach, highlights the role o f interest rate differentials. Both short term and long-term interest rate differentials across countries have been used. 16See Trehan (1988) for a basic introduction to the intuition underlying unit roots and cointegration, as well as a practical illustration. W e used the test developed by Dickey and Fuller (1979) fo r testing fo r unit roots.16 In the present case, the test consists o f regressing the first difference o f the variable under considera tion on its ow n lagged level, a constant and, to control fo r autocorrelation, an appropriate num ber o f lagged first differences.17 The coefficient estimate on the lagged level is crucial, because the null hypothesis o f a unit root implies that it is zero. The test-statistic is simply the estimate o f the coefficient divided by its standard error. This test-statistic, which does not have the usual t-distribution, is then compared w ith critical values tabulated in Fuller (1976). The results listed in table 1 show that w e can not reject the null hypothesis o f a unit root fo r any o f the bilateral real exchange rates.18 The percent. Given the nature of the cointegration tests, this caveat applies to these results as well, 17lf lagged first differences are needed, then the test is an “ augmented” Dickey-Fuller test; otherwise, the test is simply a Dickey-Fuller test. The chosen lag length is the smallest lag length for which there is no autocorrelation. 18An important caveat concerning the interpretation of unit root tests is the extremely low power of these tests. Given a sample size of approximately 100 observations, the prob ability of accepting a coefficient of 1.0 on the lagged dependent variable when it is actually 0.95 is roughly 80 JANUARY/FEBRUARY 1990 44 Table 1 Unit Root Tests For Real Exchange Rates and Potential Determinants Countries q PW /PC -PW '/PC* GNP-GNP* TB-TB* RS-RS* RL-RL* UK/US 1.63 0.31 2.20 1.86 2.96’ 2.77 WG/US 1.27 0.13 1.51 2.72 3.151 1.52 JP/US 0.88 1.44 -0 .4 7 0.10 3.66' 1.48 UK/WG 1.56 1.15 2.48 2.08 2.04 2.03 JP/WG 0.84 0.43 0.70 2.56 2.921 2.68 UK/JP 1.27 0.68 0.32 1.91 4.40' 2.73 1 Statistically significant at the 0.05 level. NOTE: The test-statistic reported is minus the regression t-statistic on a in a regression of the follow ing general form: n Ax, * c - ax_, + 1/3, to ., + £,. i« 1 The lag length n is chosen as the smallest value for which no autocorrelation exists. For a sample of 100 observations, the critical value for a significance level of 5 percent is 2.89. real exchange rate measures are nonstationary since there is no tendency fo r the real exchange rates to return to an average value over time. Thus, consistent w ith studies cited previously, w e reject long-run PPP. Table 1 also contains the results o f unit root tests fo r the potential determinants o f real ex change rates. Both proxies used to measure the tradeables/non-tradeables distinction, the dif ference betw een countries o f their ratios o f wholesale to consumer prices (PW/PC-PW*/PC*) and real per capita gross national products (GNP-GNP*), are nonstationary. An identical con clusion is reached fo r the cumulated current ac count difference (TB-TB*), the proxy based on the balance-of-payments approach. In every case, w e cannot reject the null hypothesis o f a unit root. Th e results fo r the tw o proxies based on the m onetary approach are mixed. The difference betw een the short-term interest rates (RS-RS*) appears to be stationary in most cases. In only one case, United Kingdom/West Germany, the null hypothesis o f a unit root is accepted. On the other hand, the difference betw een the long-term interest rates (RL-RL*) appears to be nonstationary because the null hypothesis o f a unit root is accepted in each case. FEDERAL RESERVE BANK OF ST. LOUIS Testing f o r Cointegration Even though the real exchange rate has a unit root, it is possible that there is a long-run rela tionship betw een it and other variables that also contain unit roots. For an equilibrium relation ship to exist betw een these variables, the distur bances that cause nonstationary behavior in one variable must also cause nonstationary behavior in the other variable. To test w hether there is a long-run relation ship betw een variables that contain unit roots, the residuals from an ordinary-least-squares regression betw een the variables can be exam ined fo r stationarity. In other w ords, a DickeyFuller test is perform ed on the residuals resul ting from regressing one variable—the real ex change rate in our case—on a potential determ i nant. Th e first difference o f the residual series is regressed on its lagged level, a constant and an appropriate num ber o f lagged first d iffe r ences. A hypothesis test is perform ed using the coefficient estimate on the lagged level. I f the null hypothesis o f a coefficient o f zero can be rejected, then the residuals are stationary. I f the residuals are stationary, then the variables w ill not drift away from each other. Such variables are said to be cointegrated. Since cointegration 45 Table 2 Cointegration Tests of Real Exchange Rates and Potential Determinants Countries PW /PC-PW '/PC* GNP-GNP* TB-TB* RS-RS* RL-RL* UK/US 1.62 1.64 1.57 1.83 1.71 WG/US 1.49 1.53 1.64 2.45 3.12' JP/US 1.90 1.07 1.07 1.21 1.23 UK/WG 2.76 2.11 2.29 2.11 2.37 JP/WG 2.63 3.501 2.06 1.39 1.33 UK/JP 2.41 1.49 1.32 1.51 1.80 1 Statistically significant at the 0.05 level. tests are oriented tow ard rejecting any long-run relationship, finding cointegration suggests the existence o f a significant statistical link betw een the variables. Table 2 shows the results o f such cointegra tion tests. Overall, there is little evidence to in dicate that the examined variables explain varia tions in the real exchange rate over time. Real exchange rates and the differences betw een ratios o f wholesale to consumer prices across countries do not appear to be cointegrated. Despite augmented Dickey-Fuller statistics that approach the critical value in tw o cases, United Kingdom/West Germany and West Germany/Japan, the residuals are nonstationary in each case. The real exchange rate and the difference be tw een the real per capita gross national pro ducts are cointegrated only fo r W est Germany and Japan. The cointegration tests also reveal that the residuals are nonstationary fo r both the cumulated trade balance and short-term interest rate differences. One interesting result is the significant rela tionship betw een the real exchange rate and the long-term real interest differential fo r the United States and W est Germany. This result contrasts w ith Campbell and Clarida (1987) and Meese and R o go ff (1988) w ho fail to find cointe gration betw een these variables. Since our p ro xy fo r the long-term real interest differential is exactly the same as that used b y Meese and Rogoff, the additional 18 m ore recent months o f data seem to account fo r the different result.19 In figure 2, w e have plotted the long-term real interest differential betw een the United States and Germany and the real mark/dollar exchange rate. As is apparent from the figure, there is a strong link betw een these variables: a higher long-term real interest rate in the United States relative to Germany means a stronger dollar. SUM M ARY Unfortunately, as our review o f a host o f studies reveals, little is known about the deter minants o f real exchange rates in the long run. Our systematic survey o f five potential explana tory variables suggests that no approach to this issue is satisfactory. For example, fo r certain ex change rates, the real approach based on the tradeables/non-tradeables distinction yields evidence o f an equilibrium relationship between 19The sample used by Meese and Rogoff (1988), as stated in their table V, terminated in December 1985, while our sample using long-term interest differentials between West Germany and the United States terminated in June 1987. When we delete these 18 months of observations, the regression t-statistic is 2.77 rather than the 3.12 reported in our table 2. Like Meese and Rogoff, we would fail to find cointegration. JANUARY/FEBRUARY 1990 46 Figure 2 Cointegration of Real Mark/Dollar Exchange Rate and Real Long Interest Rate Differential 1973 the real exchange rate and differences in real per capita gross national product across coun tries. On the other hand, the m onetary ap proach seems to have some value in the W est German/United States case. An equilibrium rela tionship appears to exist betw een the real ex change rate and the difference in long-term in terest rates. In view o f the low p ow er o f unit root tests, this finding is especially noteworthy. Th e fact remains that our knowledge o f the determination o f real exchange rates is meager, at best. The logical question is to ask w h ere research might be directed to expand our knowledge in this area. Numerous explanations, in addition to measurement problems, can be offered fo r the fact that fundamental variables have not yielded good explanations o f real ex change rate movements. One possibility is that the recent period o f floating exchange rates is FEDERAL RESERVE BANK OF ST. LOUIS too brief, especially in view o f our statistical tools, to draw conclusions about the long-run behavior o f real exchange rates. Assuming no change in exchange-rate regime, the passage o f time w ill ultimately rectify this problem. Until sufficient time passes, how ever, it w ould be p re m ature to discard the theoretical approaches that have been proposed. A second possibility is that the existing models are deficient. Our review identified instances in which the assumptions underlying the models did not hold. In addition, since real exchange rates are asset prices, the role o f expectations is an aspect o f the modeling process that deserves additional scrutiny. The failure o f existing models might re sult from the fact that expectations are formed differently than our models suggest. Consequently, the development o f alternative expectations for mation mechanisms might prove productive. 47 A final possibility is that random shocks of various origins, such as oil price shocks, have moved the real exchange rate. While identification o f the real factors that might have affected ex change rates is difficult, Meese and Rogoff (1988) suggest that further attention should be focused on the role o f real shocks. Thus, models utilizing modern real business-cycle research might generate some insights. All o f these “mights” serve as a final rem inder o f how little w e know. REFERENCES Adler, Michael, and Bruce Lehmann. “ Deviations from Pur chasing Power Parity in the Long Run,” Journal of Finance (December 1983), pp. 1471-87. Baillie, Richard T., and Patrick C. McMahon. The Foreign Exchange Market: Theory and Econometric Evidence (Cam bridge University Press, 1989). Baillie, Richard T., and David D. Selover. “ Cointegration and Models of Exchange Rate Determination,” International Journal of Forecasting (1987:1), pp. 43-51. Balassa, Bela. “ The Purchasing-Power Parity Doctrine: A Reappraisal,” Journal of Political Economy (December 1964), pp. 584-96. Campbell, John Y., and Richard H. Clarida. “ The Dollar and Real Interest Rates,” Carnegie-Rochester Conference Series on Public Policy (Autumn 1987), pp. 103-40. _______ . “ On the Mark: A Theory of Floating Exchange Rates Based on Real Interest Differentials,” American Economic Review (September 1979), pp. 610-22. Frankel, Jeffrey A., and Richard Meese. “Are Exchange Rates Excessively Variable?” in National Bureau of Economic Research, Macroeconomics Annual 1987, Stanley Fischer, ed. (MIT Press, 1987), pp. 117-53. Frenkel, Jacob A., and Michael L. Mussa. “Asset Markets, Exchange Rates and the Balance of Payments,” in Ronald W. Jones and Peter B. Kenen, eds., Handbook of Interna tional Economics (Amsterdam: Elsevier Science Publishers B.V., 1985), pp. 679-747. Fuller, Wayne A. Introduction to Statistical Time Series (Wiley, 1976). Hakkio, Craig S. “ Does the Exchange Rate Follow a Ran dom Walk? A Monte Carlo Study of Four Tests for a Ran dom Walk,” Journal of International Money and Finance (June 1986), pp. 221-29. Hooper, Peter, and John Morton. “ Fluctuations in the Dollar: A Model of Nominal and Real Exchange Rate Determina tion,” Journal of International Money and Finance (April 1982), pp. 39-56. Hsieh, David A. “ The Determination of the Real Exchange Rate,” Journal of International Economics (May 1982), pp. 355-62. Huizinga, John. “An Empirical Investigation of the Long-Run Behavior of Real Exchange Rates,” Carnegie-Rochester Conference Series on Public Policy (Autumn 1987), pp. 149-214. Isard, Peter. “An Accounting Framework and Some Issues for Modeling How Exchange Rates Respond to the News,” in Jacob A. Frenkel, ed., Exchange Rates and International Macroeconomics (University of Chicago Press, 1983), pp. 19-56. Cox, W. Michael. “A Comprehensive New Real Dollar Ex change Rate Index,” Federal Reserve Bank of Dallas Economic Review (March 1987), pp. 1-14. Koedijk, Kees, and Peter Schotman. “ Dominant Real Ex change Rate Movements," Journal of International Money and Finance (December 1989), pp. 517-31. Cumby, Robert E., and Maurice Obstfeld. “ International In terest Rate and Price Level Linkages under Flexible Ex change Rates: A Review of Recent Evidence,” in John F.O. Bilson and Richard C. Marston, eds., Exchange Rate Theory and Practice (University of Chicago Press, 1984), pp. 121-51. Lothian, James R. “A Century Plus of Yen Exchange Rate Behavior,” mimeo, New York University. Darby, Michael R. “ Movements in Purchasing Power Parity: The Short and Long Runs,” in Michael R. Darby and James R. Lothian, eds. The International Transmission of In flation (University of Chicago Press, 1983), pp. 462-77. Dickey, David A., and Wayne A. Fuller. “ Distribution of the Estimators for Autoregressive Time Series With a Unit Root,” Journal of the American Statistical Association (June 1979), pp. 427-31. Dornbusch, Rudiger. “ Real Exchange Rates and Macro economics: A Selective Survey,” Scandinavian Journal of Economics (1989:2), pp. 401-32. McNown, Robert, and Myles S. Wallace. “ National Price Levels, Purchasing Power Parity, and Cointegration: A Test of Four High Inflation Economies,” Journal of International Money and Finance (December 1989), pp. 533-45. Meese, Richard, and Kenneth Rogoff. “ Was It Real? The Ex change Rate-lnterest Differential Relation over the Modern Floating-Rate Period,” Journal of Finance (September 1988), pp. 933-48. Sims, Christopher A. “ Bayesian Skepticism on Unit Root Econometrics,” Discussion Paper 3, Institute for Empirical Macroeconomics, Federal Reserve Bank of Minneapolis and University of Minnesota, May 1988. ________“ Expectations and Exchange Rate Dynamics,” Journal of Political Economy (December 1976), pp. 1161-76. Tatom, John A. “ The Link Between the Value of the Dollar, U.S. Trade and Manufacturing Output: Some Recent Evidence,” this Review (November/December 1988), pp. 24-37. Dornbusch, Rudiger, and Stanley Fischer. “ Exchange Rates and the Current Account,” American Economic Review (December 1980), pp. 960-71. Taylor, Mark P. “An Empirical Examination of Long-Run Pur chasing Power Parity Using Cointegration Techniques,” Ap plied Economics (October 1988), pp. 1369-81. Dornbusch, Rudiger, and Paul Krugman. “ Flexible Exchange Rates in the Short Run,” Brookings Papers on Economic Activity (1976:3), pp. 537-75. Edison, Hali J., and Jan Tore Klovland. “A Quantitative Reassessment of the Purchasing Power Parity Hypothesis: Evidence from Norway and the United Kingdom,” Journal of Applied Econometrics (October 1987), pp. 309-33. Frankel, Jeffrey A. “ International Capital Mobility and Crowding-out in the U.S. Economy: Imperfect Integration of Financial Markets or of Goods Markets?” in R.W. Hafer, ed., How Open Is the U.S. Economy? (D.C. Heath and Com pany, 1986), pp. 33-67. Trehan, Bharat. “ The Practice of Monetary Targeting: A Case Study of the West German Experience,” Federal Reserve Bank of San Francisco Economic Review (Spring 1988), pp. 30-44. Whitt, Joseph A. Jr. “ Purchasing-Power Parity and Exchange Rates in the Long Run,” Federal Reserve Bank of Atlanta Economic Review (July/August 1989), pp. 18-32. Wolff, Christian C.P. “ Time-Varying Parameters and the Outof-Sample Forecasting Performance of Structural Exchange Rate Models,” Journal of Business & Economic Statistics (January 1987), pp. 87-97. JANUARY/FEBRUARY 1990 48 Appendix Data Sources and the Construction of Variables For all variables except interest rates, w e use monthly data from International Financial Statis tics fo r June 1973 to June 1988. Recall that q is the log o f the real exchange rate with the con sumer price index o f the respective countries as the deflator. GNP is the log o f real gross na tional product, w h ere nominal gross national product is deflated w ith the gross national p ro duct price deflator. Since gross national product data are not available monthly, w e interpolated from quarterly to monthly observations using industrial production. Note that all the potential determinants are defined as differences, w ith an asterisk indicating the respective value in the other country. PW/PC is the log o f the w h ole sale price index divided by the consumer price index. TB is the cumulated trade balance (that is, the sum from January 1972 to time period t o f the difference betw een exports and imports) divided by gross national product. For the interest rates not involving Japan, the time period is June 1973 to June 1987, while fo r those involving Japan, the time period is July 1977 to June 1987. The nominal short-term interest rates and their sources are as follows: Japan—one-month Gensaki rate from the Bank o f Japan; United Kingdom —one-month interbank deposit rate from the Financial Times', W est Germany—one-month interbank rate from the Frankfurter Allgemeine Zeitung; and United States—the yield on one-month Treasury bills until A pril 1984 and, afterward, the interest FEDERAL RESERVE BANK OF ST. LOUIS rate on three-month Treasury bills from the Federal Reserve Board. The nominal long-term interest rates and their sources are as follows: Japan—average yield to maturity on governm ent bonds w ith constant remaining maturity o f nine years from the Bank o f Japan; United Kingdom—average yield to maturity on governm ent bonds w ith remaining maturity betw een eight and 12 years from the Financial Times; W est Germany—average yield to maturity on governm ent bonds w ith remain ing m aturity over eight years from the Frank fu rte r Allgemeine Zeitung; and United States— yield to m aturity o f governm ent bonds w ith re maining maturity o f 10 years from the Federal Reserve Board. W e calculated the real short term and long-term interest rates, RS and RL, in the same manner as Meese and R o go ff (1988). Thus, actual inflation rates based on the preced ing 12 months as measured by the consumer price index are subtracted from the nominal in terest rates to generate the real rates. As a re sult, the inflation measure does not correspond to the term o f the interest rates. A problem w ith this construction, as w ell as many others, is that negative real interest rates are computed. For example, the long-term (short-term) real in terest rate is negative fo r 25 (42) percent o f the U.S. observations, 38 (30) percent o f the British observations, 0 (9) percent o f the W est German observations and 6 (7) percent o f the Japanese observations. Federal R eserve Bank o f St. Louis Post Office Box 442 St. Louis, Missouri 63166 The R e v ie w is published six times per year b y the Research and Public Information Department o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public f r e e o f charge. Mail requests f o r subscriptions, back issues, or address changes to: Research and Public Information Department, Federal Reserve Bank o f St. Louis, P.O. Box 442, St. Louis, Missouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Bank o f St. Louis o r the Federal R eserve System. Articles herein may b e reprinted provided the source is credited. Please provide the Bank’s Besearch and Public Information Department with a copy o f reprinted material.