VOL 8I NO. I AUSUBEL: CREDIT CARD MARKET ules that determine direct customer rev- off a balance six months after the card enues are set entirely at the bank level; only holder ceases payment on an the interchange fee is set systemwide by the MBNa's charge-off rate is 1-2 percent Master Card and visa organizations lower than that of many large credit card Costs can be divided among interest ex- issuers penses, operating expenses, and loan losses As seen in Table 5, MBNA,s pretax re- Interest expense is determined by market turn on assets(roa)in credit cards equaled interest rates and thus is relatively uniform 7.09 percent in 1985, 6.63 percent in 1986 across banks, as a percentage of outstanding and 4.80 percent in 1987. The 1987 figure balances. Noninterest expenses, which in- for example, interacts with a 42 clude employee salaries, occupancy, equip- rate to yield an after-tax return on assets of ment,and data processing, are also avail- 2. 78 percent. For evidence of the accuracy able from the banks' direct reports and call of this computation, one need look no fur- reports. These noninterest expenses typi- ther than MNC Financials 1987 annual re cally equal 4-6 percent of outstanding bal port nces for large issuers and are mostly(but not entirely) a proper component of total cost. The exception is that the expense of Our credit card operations had an- ther outstanding year. maryland g and marketing costs )should ly be ank, NA(MBna)is by no means considered an investment and thus should typical of the industry, which often is be amortized over a longer period. Never- Over the past five years, MBNa has theless, I have no systematic way to sepa been one of our fastest growing bus rate out these new-account e from lesses. with s2 billion in outstand banks' profits; consequently, I use the entire ings, it continues to be a low cost noninterest expense in my computations. high-volume producer with chargeoffs Observe that this will tend to overstate costs of about 2%0-about half of the indus and understate the returns on assets and try average we think most investors quity. Loan losses are best measured by will find it hard not to be impressed he bank's“ net credit losses”or“net with a business that earns more than charge-offs, which represent the outstand 2.5%(after-tax, 1987)on assets ing balances that the bank newly treats as uncollectable.(Typically, a bank charges By way of comparison, the bank holding company as a whole earned a 1. 36-percent ROa before taxes and a 1.00-percent ROa after taxes in 1987. The holding company, asked for the bank's "net credit losses"(see appendix minus its credit card business, earned less A), The item used from call reports is the"net han a 0. 80-percent roa after taxes in 1987. harge-offs. An alternative measure of es tha could have been used from the call reports is the ank's "provision for loan losses, which is ofter B. The Ordinary Rate of Return in the Banking Industry ank (statistically) expects to charge off in the There are two reasons not to use the figure for provi The pretax return on assets for all u.s sion. First, credit card accounts incur most of their commercial banks during the sample years harge-offs in the initial two years of the life of the and "net charge-offs"(and in fact, some of"net harge-offs""itself) typically represents an expense of rating new accounts and as in the case of market is less ma ctional cost analysis also period. Using losses"in earnings computations r measure of cost, second. s' hat is easily manipulated by the bank: one can use ANC Financial, 1987 Annual Report (dated ion to defer income taxes or a small March 1988) This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
VOL. 81 NO. I AUSUBEL: CREDIT CARD MARKET 59 ules that determine direct customer rev- enues are set entirely at the bank level; only the interchange fee is set systemwide by the MasterCard and Visa organizations. Costs can be divided among interest ex- penses, operating expenses, and loan losses. Interest expense is determined by market interest rates and thus is relatively uniform across banks, as a percentage of outstanding balances. Noninterest expenses, which in- clude employee salaries, occupancy, equip- ment, and data processing, are also avail- able from the banks' direct reports and call reports. These noninterest expenses typi- cally equal 4-6 percent of outstanding bal- ances for large issuers and are mostly (but not entirely) a proper component of total cost. The exception is that the expense of generating a new account (mostly advertis- ing and marketing costs) should properly be considered an investment and thus should be amortized over a longer period. Never- theless, I have no systematic way to sepa- rate out these new-account expenses from banks' profits; consequently, I use the entire "noninterest expense" in my computations. Observe that this will tend to overstate costs and understate the returns on assets and equity. Loan losses are best measured by the bank's "net credit losses" or "net charge-offs," which represent the outstand- ing balances that the bank newly treats as uncollectable.21 (Typically, a bank charges off a balance six months after the card- holder ceases payment on an account.) MBNA's charge-off rate is 1-2 percent lower than that of many large credit card issuers. As seen in Table 5, MBNA's pretax re- turn on assets (ROA) in credit cards equaled 7.09 percent in 1985, 6.63 percent in 1986, and 4.80 percent in 1987. The 1987 figure, for example, interacts with a 42-percent tax rate to yield an after-tax return on assets of 2.78 percent. For evidence of the accuracy of this computation, one need look no fur- ther than MNC Financial's 1987 annual re- port: Our credit card operations had an- other outstanding year. Maryland Bank, N.A. (MBNA) is by no means typical of the industry, which often is the target for criticism and concern. Over the past five years, MBNA has been one of our fastest growing busi- nesses. With $2 billion in outstand- ings, it continues to be a low cost, high-volume producer with chargeoffs of about 2%-about half of the indus- try average. We think most investors will find it hard not to be impressed with a business that earns more than 2.5% (after-tax, 1987) on assets.22 By way of comparison, the bank holding company as a whole earned a 1.36-percent ROA before taxes and a 1.00-percent ROA after taxes in 1987.23 The holding company, minus its credit card business, earned less than a 0.80-percent ROA after taxes in 1987. B. The Ordinary Rate of Return in the Banking Industry The pretax return on assets for all U.S. commercial banks during the sample years 21The follow-up bank credit card survey specifically asked for the bank's "net credit losses" (see Appendix A). The item used from call reports is the "net charge-offs." An alternative measure of losses that could have been used from the call reports is the bank's "provision for loan losses," which is often higher and which may include an allowance for loans that the bank (statistically) expects to charge off in the future. There are two reasons not to use the figure for provi- sion. First, credit card accounts incur most of their charge-offs in the initial two years of the life of the account. Hence, the difference between "provision" and "net charge-offs" (and, in fact, some of "net charge-offs" itself) typically represents an expense of generating new accounts and, as in the case of market- ing expenses, should properly be treated as an invest- ment which is amortized over a longer period. Using "net charge-offs" mitigates this effect and gives a bet- ter measure of cost. Second, "provision" is a quantity that is easily manipulated by the bank: one can use a large loss provision to defer income taxes or a small loss provision to report high current earnings. "Net charge-offs" is less manipulable. The Federal Reserve System's functional cost analysis also uses "net credit losses" in earnings computations. 22MNC Financial, 1987 Annual Report (dated March 1988), p. 4. 23MNC Financial, 1987 Annual Report (dated March 1988), p. 1. This content downloaded from 202.120.224.93 on Sun, 17 Dec 2017 07:42:57 UTC All use subject to http://about.jstor.org/terms
THE AMERICAN ECONOMIC REVIEW MARCH 1991 equaled 0.85 percent in 1983, 0.83 percent cent would imply an ordinary (pretax)re in 1984, 0.90 percent in 1985, 0.80 percent turn on equity of 20 percent per year in 1986, 0.28 percent in 1987, and 1. 14 per- cent in 1988. 4 Taking into account that C. Computations of Ex Post Profitability some areas of banking were effectively taxed for 15 Large Issuers rate than the credit card busi lesses (which were taxed at close to the Several different summary measures of statutory tax rates of 34-46 percent during (ex post) profitability are presented in Ta- this period), it is probably correct to think bles 6 and 7. The first measure of return on of 1.20 percent as the ordinary (pretax)re- assets, RoA ( reported), is precisely the cal turn on assets in the banking industry at culation we illustrated above for MBNA rge One potential flaw in this calculation is that The relationship between the ordinary it relies on the bank,'s own reported cost of rate of return on assets and the ordinary funds. The problem here is that some banks rate of return on equity in the banking may not have been allocating the true op industry depends on the capital require- portunity cost of their low-cost core de ments of banks. For the period 1983-1988, posits (e. g, passbook accounts and non the capital requirement equaled about 6 interest-bearing checking accounts )to their rcent of assets first in 1984-1985, U. S. credit card businesses: in that event. some banking regulators promulgated capital of the profits allocated to the credit card qualing 6 percent of assets for total capital the branch banking business attributable to standards for all commercial bank activities operations would in fact b (and 5.5 percent of assets for primary capi- This difficulty is easily remedied by re tal).Second, actual total equity capital for placing each bank's reported interest ex all insured U.s. commercial banks equaled pense with the standardized index of the 5.96 percent of assets in 1983, 6.01 percent cost of funds defined and defended in Sec n 1984, 6. 17 percent in 1985, 6.21 percent tion Il. My second measure of return on in 1986, 6.06 percent in 1987, and 6.10 per- assets, Roa(adjusted ), is computed by us- ent in 1988(with substantially smaller per- ing an interest expense of COST OF centages for the larger banks). 2 Dividing an FUNdS applied to the nonequity portion of ordinary(pretax) return on assets of 1.20 assets; thus, interest expense as a percent- percent by a capital requirement of 6 per- age of assets equals 94 percent of COST OF FUNDS. If anything, my adjustment tends to reduce systematically the reported re- turns; observe in Table 7 that ROA (re (table 1, ported)exceeds ROA(adjusted)in four out and July 1988, p, 404(table 1). The substantial earnings for 1987 reflect the decision of large ba OI SIX years et aside large sums to cover troubled loans to The return on equity is computed in two ing countries. Excluding international operation different ways in Table 7. The first and most obvious measure, ROE (actual cap), merely slightly exceeded those of 1986 divides(pretax) profits by the actual capital uled to rise, by international agreement, to 8 percent of residing in the credit card bank at the previ rapidly expanding practice of securitizing credit card is a legally distinct entity, its capital is a assets has the effect of removing the credit card ac- eets, thus reducing the (table A 1), and July 1988, p. 405(table 2). For money. 2 ROA(adjusted), as comp from the call report enter banks, equity capital equaled 4.30, 4.56, 4.69, data, also contains a second, minor adjustment Section V) and quity capital equaled 4.76, 5.08, 5.42, 5.50, 5.29, and subtracted a portion of the premium from its profits, 5.29 perce ent of assets in the respective years ROA (adjusted )adds it back in This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
60 THE AMERICAN ECONOMIC REVIEW MARCH 1991 equaled 0.85 percent in 1983, 0.83 percent in 1984, 0.90 percent in 1985, 0.80 percent in 1986, 0.28 percent in 1987, and 1.14 per- cent in 1988.24 Taking into account that some areas of banking were effectively taxed at a lower rate than the credit card busi- nesses (which were taxed at close to the statutory tax rates of 34-46 percent during this period), it is probably correct to think of 1.20 percent as the ordinary (pretax) re- turn on assets in the banking industry at large. The relationship between the ordinary rate of return on assets and the ordinary rate of return on equity in the banking industry depends on the capital require- ments of banks. For the period 1983-1988, the capital requirement equaled about 6 percent of assets. First, in 1984-1985, U.S. banking regulators promulgated capital standards for all commercial bank activities equaling 6 percent of assets for total capital (and 5.5 percent of assets for primary capi- tal).25 Second, actual total equity capital for all insured U.S. commercial banks equaled 5.96 percent of assets in 1983, 6.01 percent in 1984, 6.17 percent in 1985, 6.21 percent in 1986, 6.06 percent in 1987, and 6.10 per- cent in 1988 (with substantially smaller per- centages for the larger banks).26 Dividing an ordinary (pretax) return on assets of 1.20 percent by a capital requirement of 6 per- cent would imply an ordinary (pretax) re- turn on equity of 20 percent per year. C. Computations of Ex Post Profitability for 15 Large Issuers Several different summary measures of (ex post) profitability are presented in Ta- bles 6 and 7. The first measure of return on assets, ROA (reported), is precisely the cal- culation we illustrated above for MBNA. One potential flaw in this calculation is that it relies on the bank's own reported cost of funds. The problem here is that some banks may not have been allocating the true op- portunity cost of their low-cost core de- posits (e.g., passbook accounts and non- interest-bearing checking accounts) to their credit card businesses; in that event, some of the profits allocated to the credit card operations would in fact be attributable to the branch banking business. This difficulty is easily remedied by re- placing each bank's reported interest ex- pense with the standardized index of the cost of funds defined and defended in Sec- tion II. My second measure of return on assets, ROA (adjusted), is computed by us- ing an interest expense of COST OF FUNDS applied to the nonequity portion of assets; thus, interest expense as a percent- age of assets equals 94 percent of COST OF FUNDS.27 If anything, my adjustment tends to reduce systematically the reported re- turns; observe in Table 7 that ROA (re- ported) exceeds ROA (adjusted) in four out of six years. The return on equity is computed in two different ways in Table 7. The first and most obvious measure, ROE (actual cap), merely divides (pretax) profits by the actual capital residing in the credit card bank at the previ- ous year's end. (Since each bank in Table 7 is a legally distinct entity, its capital is a 24Federal Reserve Bulletin, July 1989, p. 462 (table 1), and July 1988, p. 404 (table 1). The substantially lower earnings for 1987 reflect the decision of large banks to set aside large sums to cover troubled loans to develop- ing countries. Excluding international operations, the banks' rates of earnings in 1987 appear to have very slightly exceeded those of 1986. 25At this writing, bank capital standards are sched- uled to rise, by international agreement, to 8 percent of total risk assets in 1992. At the same time, the new and rapidly expanding practice of securitizing credit card assets has the effect of removing the credit card ac- counts from banks' balance sheets, thus reducing the effective capital requirement. 26Federal Reserve Bulletin, July 1989, pp. 474-83 (table A.1), and July 1988, p. 405 (table 2). For money- center banks, equity capital equaled 4.30, 4.56, 4.69, 4.78, 4.33, and 4.42 percent of assets in the respective years. For other banks with $5 billion or more in assets, equity capital equaled 4.76, 5.08, 5.42, 5.50, 5.29, and 5.29 percent of assets in the respective years. 27ROA (adjusted), as computed from the call report data, also contains a second, minor adjustment: to the extent that a bank has purchased credit card portfolios from other banks at a premium (see Section V) and subtracted a portion of the premium from its profits, ROA (adjusted) adds it back in. This content downloaded from 202.120.224.93 on Sun, 17 Dec 2017 07:42:57 UTC All use subject to http://about.jstor.org/terms