American Economic association The Failure of Competition in the Credit Card Market Author(s): Lawrence M. Ausubel Source: The American Economic Review, Vol. 81, No. 1(Mar, 1991), pp. 50-81 Published by: American Economic Association StableUrl:http://www.jstor.org/stable/2006788 Accessed: 17-12-201707: 42 UTC JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact supportejstor org Your use of the JSTOR archive indicates your acceptance of the Terms Conditions of Use, available at http://about.jstor.org/terms American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic review STOR This content downloaded from 202. 120.224.93 on Sun, 17 Dec 201707: 42: 57UTC Allusesubjecttohttp:/aboutjstor.org/terms
American Economic Association The Failure of Competition in the Credit Card Market Author(s): Lawrence M. Ausubel Source: The American Economic Review, Vol. 81, No. 1 (Mar., 1991), pp. 50-81 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/2006788 Accessed: 17-12-2017 07:42 UTC JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic Review 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 Failure of Competition in the credit Card Market By LAWRENCE M. AUSUBEL* The bank credit card market, containing 4,000 firms and lacking regulatory barriers, casually appears to be a hospitable environment for the model of perfect competition. Nevertheless, this article reports that credit card interest rates have been exceptionally sticky relative to the cost of funds. Moreover, major credit card issuers have persistently earned from three to fiue times the ordinary rate of return in banking during the period 1983-1988. The failure of be partly attributable to credit card choices without taking account of the very high probability that they will pay interest on their outstanding balances (JEL 315, 612) This article presents and discusses a col- strained inputs, significant sunk costs, or lection of data which is paradoxical within significant barriers to entry. Finally, there is the paradigm of perfect ition The no evidence of any explicit collusion on market studied the bank credit card indus- price or quantity in the United States, contains literally ch a favorable market do 4,000 firms who sell a relatively homoge- tion, or one not even half so optimistic, neous good to 75 million consumers. The many economists would prefer to presume ten largest firms account for only about that the market must behave as a competi wo-fifths of market share. Firms have his- tive spot market in continuous equilibrium orically operated without regulatory bar- It is the purpose of this article to argue that riers to conducting business across state this presumption is empirically unjustified lines-and at least 20 firms aggressively so- in the market for bank credit cards in the licit business on a national scale. firms have 1980 s Section I outlines the market struc also operated in the virtual absence of price ture of the bank credit card industry. Sec gulations for most of a decade. There do tion Il offers empirical evidence of extreme not appear to be any particularly con- price stickiness in credit card interest rates Section Ill provides direct profit data on dustry, arguing that the 50 large credit card issuers have earned from three Department of Managerial Economics and deci- to five times the ordinary rate of return for gement, Northwestern University, Evanston, IL 60208 the banking industry during the period The author acknowledges the support of the Kellogg 1983-1988 Section IV examines profits over School,s Banking Research Cen he Lynde and a larger sample of banks and a longer time Harry Bradley Foundation, and the C. V. Starr Center period. Section V presents additional data at New York university and appreciates the diligent on resales of credit card portfolios between Alan Blinder, Charles calomiris, Raymond Deneckere, banks, suggesting that the extraordinary Peter Diamond, Stuart Greenbaum, Robert Johnson, profits exist ex ante as well as ex post (and Meetings, the Econometric society meetings, the explanations for price stickiness and supra western University Summer Industrial Organizat Conference, the Federal Reserve Bank of Chicago, tion VIll briefly discusses the extent of wel New York University, Princeton University, and the fare loss in the market and the merits of the officers of 21 major banks who cooperatively re- regulation to correct market failure. Con clusions are presented in Section IX. This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
The Failure of Competition in the Credit Card Market By LAWRENCE M. AUSUBEL* The bank credit card market, containing 4,000 firms and lacking regulatory barriers, casually appears to be a hospitable environment for the model of perfect competition. Nevertheless, this article reports that credit card interest rates have been exceptionally sticky relative to the cost of funds. Moreover, major credit card issuers have persistently earned from three to five times the ordinary rate of return in banking during the period 1983-1988. The failure of the competitive model appears to be partly attributable to consumers, making credit card choices without taking account of the very high probability that they will pay interest on their outstanding balances. (JEL 315, 612) This article presents and discusses a col- lection of data which is paradoxical within the paradigm of perfect competition. The market studied, the bank credit card indus- try in the United States, contains literally 4,000 firms who sell a relatively homoge- neous good to 75 million consumers. The ten largest firms account for only about two-fifths of market share. Firms have his- torically operated without regulatory bar- riers to conducting business across state lines-and at least 20 firms aggressively so- licit business on a national scale. Firms have also operated in the virtual absence of price regulations for most of a decade. There do not appear to be any particularly con- strained inputs, significant sunk costs, or significant barriers to entry. Finally, there is no evidence of any explicit collusion on price or quantity. Given such a favorable market descrip- tion, or one not even half so optimistic, many economists would prefer to presume that the market must behave as a competi- tive spot market in continuous equilibrium. It is the purpose of this article to argue that this presumption is empirically unjustified in the market for bank credit cards in the 1980's. Section I outlines the market struc- ture of the bank credit card industry. Sec- tion II offers empirical evidence of extreme price stickiness in credit card interest rates. Section III provides direct profit data on the industry, arguing that the 50 largest credit card issuers have earned from three to five times the ordinary rate of return for the banking industry during the period 1983-1988. Section IV examines profits over a larger sample of banks and a longer time period. Section V presents additional data on resales of credit card portfolios between banks, suggesting that the extraordinary profits exist ex ante as well as ex post (and that bankers expect the profitability to per- sist). Section VI explores some theoretical explanations for price stickiness and supra- normal profits. Section VII calculates what would be "competitive" interest rates. Sec- tion VIII briefly discusses the extent of wel- fare loss in the market and the merits of regulation to correct market failure. Con- clusions are presented in Section IX. *Department of Managerial Economics and Deci- sion Sciences, J. L. Kellogg Graduate School of Man- agement, Northwestern University, Evanston, IL 60208. The author acknowledges the support of the Kellogg School's Banking Research Center, The Lynde and Harry Bradley Foundation, and the C. V. Starr Center at New York University and appreciates the diligent research work of Gail Eynon and Paul Palmer. I thank Alan Blinder, Charles Calomiris, Raymond Deneckere, Peter Diamond, Stuart Greenbaum, Robert Johnson, Charles Kahn, Robert Porter, and three anonymous referees for helpful comments. I also thank seminar participants at the American Economic Association Meetings, the Econometric Society Meetings, the NBER Economic Fluctuations Conference, the North- western University Summer Industrial Organization Conference, the Federal Reserve Bank of Chicago, New York University, Princeton University, and the University of Delaware. Special thanks are also due to the officers of 21 major banks who cooperatively re- sponded to my requests for data. 50 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
VOL. 8I NO. I AUSUBEL: CREDIT CARD MARKET I. The Bank Credit Card Market: Is 4.000 nection with the credit card industry apply also to other forms of consumer borrowing (especially other unsecured credit) Credit cards are the currency of late If visa and Master Card were the relevant 20th-century America. The aggregate charge levels of business to examine, then two firms volume on plastic in the United States was would control a substantial part of the credit half of this total-$165 billion in volume- ness decisions are made at the level of the was charged on Master Card and Visa credit issuing bank. Individual banks own their cards(the primary focus of this article), and cardholders' accounts and determine the in volume was growing at well over 10 percent terest rate, annual fee, grace period, credit per year. The remaining volume arose limit, and other terms of the accounts. (only largely from similar credit cards (e. g, the charges such as the"interchange fee"from Discover and Optima cards), "travel and the merchant's bank to the cardholder's entertainment"cards (e.g, the American bank are standardized, and the cardholder's Express card), and retail cards (e. g, depart- bank appears only to break even on such ment store and oil company cards) charges. Moreover, there is absolutely no borrowing via credit cards (and all con- indication that the Master Card and Visa sumer borrowing) is also significant and has organizations serve to facilitate collusion on een even more of a growth industry Out- other prices. )In essence, Master Card In standing U.S. balances on revolving credit ternational and Visa U.S.a. are organiza accounts equaled $203 billion at year-end tions largely irrelevant to this discussion; 1989, up from only $70 billion in 1982.3"firms"will henceforth refer to the issuing More than S130 billion of this total con- banks sisted of Master Card and visa balances The market for MasterCard and Visa nore than a threefold increase from 1982, cards, thus, is relatively unconcentrated the and bank card balances were still increasing top ten firms control only about two-fifths of at more than a 15-percent annual rate. 4 the market, and the next ten firms control Overall outstanding consumer installment only one-tenth of the market(see Table 1) credit balances in the United States reached moreover the market is exceptionally broad S717 billion, up from $356 billion in 1982; a bank that ranked number 100 in 1987 still it is worth observing that many of the con- had approximately 160,000 active accounts, siderations explicitly discussed here in con- $125 million in outstanding balances, and $250 million in annual charge volume Unlike most aspects of American bank- ing, the credit card business has historically Moreover, Americans were estimated to have made 9.1 billion credit card transactions in 1987(The Nilson operated free of interstate banking and Report, Number 428, May 1988, p 5) me in 1987 consisted of $138 billion in counted for 59 percent mOreover, the observed interest rate behavior does of this value and Master Card accounted for the not seem to fit the conventional view of collusive haining 41 percent (The Nilson R prIc nd 1985, three major issuers(CI February 1988, p. 6, and Number 423, March 1988, pp ufacturers Hanover, and Maryland Bank interest rates on standard cards to the fEderal Reserve Boards series of Consumer In- 17.5-17,9-percent range ar from this stallment Credit, as published in Federal Reserve Bul- industry price war, other major issuers (e.g, Citibank tin,April 1990, table 1.55, line 15(and previous and First go)steadfastly maintained 19.8-percent lines 16, 19, cial bank m2计 three price-cutters announced rate increases, appa ently find on that the earlie 27, 1989, p. 32; Wall Street Journal, March 22, 1989, p FEderal Reserve Bulletin, April 1990, table 1.55, line B1) (June1987),p.7. This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
VOL. 81 NO. 1 AUSUBEL: CREDIT CARD MARKET 51 I. The Bank Credit Card Market: Is 4,000 Enough for Competition? Credit cards are the currency of late 20th-century America. The aggregate charge volume on plastic in the United States was estimated at $375 billion in 1987.1 Almost half of this total-$165 billion in volume- was charged on MasterCard and Visa credit cards (the primary focus of this article), and volume was growing at well over 10 percent 2 per year. The remaining volume arose largely from similar credit cards (e.g., the Discover and Optima cards), "travel and entertainment" cards (e.g., the American Express card), and retail cards (e.g., depart- ment store and oil company cards). Borrowing via credit cards (and all con- sumer borrowing) is also significant and has been even more of a growth industry. Out- standing U.S. balances on revolving credit accounts equaled $203 billion at year-end 1989, up from only $70 billion in 1982.3 More than $130 billion of this total con- sisted of MasterCard and Visa balances, more than a threefold increase from 1982, and bank card balances were still increasing at more than a 15-percent annual rate.4 Overall outstanding consumer installment credit balances in the United States reached $717 billion, up from $356 billion in 1982;5 it is worth observing that many of the con- siderations explicitly discussed here in con- nection with the credit card industry apply also to other forms of consumer borrowing (especially other unsecured credit). If Visa and MasterCard were the relevant levels of business to examine, then two firms would control a substantial part of the credit card market. However, most relevant busi- ness decisions are made at the level of the issuing bank. Individual banks own their cardholders' accounts and determine the in- terest rate, annual fee, grace period, credit limit, and other terms of the accounts. (Only charges such as the "interchange fee" from the merchant's bank to the cardholder's bank are standardized, and the cardholder's bank appears only to break even on such charges. Moreover, there is absolutely no indication that the MasterCard and Visa organizations serve to facilitate collusion on other prices.6) In essence, MasterCard In- ternational and Visa U.S.A. are organiza- tions largely irrelevant to this discussion; "firms" will henceforth refer to the issuing banks. The market for MasterCard and Visa cards, thus, is relatively unconcentrated. The top ten firms control only about two-fifths of the market, and the next ten firms control only one-tenth of the market (see Table 1). Moreover, the market is exceptionally broad. A bank that ranked number 100 in 1987 still had approximately 160,000 active accounts, $125 million in outstanding balances, and $250 million in annual charge volume.7 Unlike most aspects of American bank- ing, the credit card business has historically operated free of interstate banking and 1Moreover, Americans were estimated to have made 9.1 billion credit card transactions in 1987 (The Nilson Report, Number 428, May 1988, p. 5). 2U.S. volume in 1987 consisted of $138 billion in sales slips (i.e., charged goods and services) and $27 billion in cash advances. Visa accounted for 59 percent of this value and MasterCard accounted for the re- maining 41 percent. (The Nilson Report, Number 422, February 1988, p. 6, and Number 423, March 1988, pp. 4-5). 3Federal Reserve Board's series of Consumer In- stallment Credit, as published in Federal Reserve Bul- letin, April 1990, table 1.55, line 15 (and previous issues). 4Federal Reserve Bulletin, April 1990, table 1.55, lines 16, 19, and 21. Revolving credit held by commer- cial banks, savings institutions, and pools of securitized assets consists almost entirely of MasterCard and Visa balances. 5Federal Reserve Bulletin, April 1990, table 1.55, line 1 (and previous issues). 6Moreover, the observed interest rate behavior does not seem to fit the conventional view of collusive pricing. Around 1985, three major issuers (Chase Man- hattan, Manufacturers Hanover, and Maryland Bank) reduced their interest rates on standard cards to the 17.5-17.9-percent range. Far from this triggering an industry price war, other major issuers (e.g., Citibank and First Chicago) steadfastly maintained 19.8-percent rates on most accounts, without apparent detriment to their customer bases. Finally, in the spring of 1989, the three price-cutters announced rate increases, appar- ently finding without facing retaliation that the earlier cuts had been unprofitable (The New York Times, April 27, 1989, p. 32; Wall Street Journal, March 22, 1989, p. B1). 7The Nilson Report, Number 406 (June 1987), p. 7. 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 TABLE 1-TOP TEN ISSUERS OF MASTERCARD AND VISA CARDS. 1987 Percentage market Percentage market share Number of share(by number (by outstandi Bank of accounts) (S billion) Citibank 153B Chase manhattan Bank S5. 4B Wells Fargo Bank 1,800,00 28 Marylan 1,800,000 Marine Midland Bank 1,700,00 Chemical bank 1,500,000 Associates National Bank Top ten 414 Second ten Total 118,900,00 100 Sources: Individual banks' numbers of accounts surveyed by American Banker, March 1, 1988, pp. 1-2)and Credi ard News(August 15, 1988, pp. 4-16); total number of accounts from Nilson Report(Number June 1987), p. 4). Individual banks' outstanding balances based on American Banker(September 21, 1987, p. 43 [call report datap utstanding at Greenwood Trust Co [ Discover card] and American Express Centurion Bank). Data reported for December 31, 1986, adjusted for acquisitions effective in 1987. Conflicts between sources were resolved using best branch banking restrictions. Indeed, the Court held that only the usury ceiling of the largest issuers today conduct truly national state in which the bank is located, and not businesses. For example, Maryland Bank that of the state in which the consumer is (ranked number seven in Table 1)conducts located, restricts the interest rate the bank business in all 50 states and has only five may charge. This gave banks the option of percent of its accounts in its home state. shifting their credit card operations to cent of its business is concentrated are Cali- without usury laws. By 1982, amid fornia(10.7 percent), Texas (6.7 percent), quette-created bank pressure and histori Pennsylvania(6.0 percent), and New Jersey cally high market interest rates, most lead ing banking states had relaxed or repealed In the past, credit card issuers were con- their interest rate ceilings. Meanwhile rained by state usury laws. However, the South Dakota and Delaware had estab U.S. Supreme Court's December 1978 Mar- lished themselves as attractive homes- quette decision paved the way for the practi- away-from-home for credit card issuers cal elimination of price regulations The While a number of states maintain bindin usury laws at this writing (most notably Arkansas, with a ceiling of five percentage Prospectus for Maryland Bank, N.A. Credit C points above the Federal Reserve discount Trust 1987-A, December 9, 1987, pp 17- rate), essentially all major issuers can pur Marquette National Bank v. First of Omaha Service sue business in those states free of restric- Corporation, 439 U.S. 299(1978). The Marquette de on applies to credit cards issued by nationally char. tered banks, but not to retail cards (e. g, oil company credit cards). The decision explicitly permits banks to ther customer fees. at this writing, at the behest of export"their interest rates; banks have interpreted the lowa Attorney General, courts are considering this also to permit the export"of annual fees and whether this rule does indeed apply to fees This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
52 THE AMERICAN ECONOMIC REVIEW MARCH 1991 TABLE 1 -Top TEN ISSUERS OF MASTERCARD AND VISA CARDS, 1987 Percentage market Outstanding Percentage market share Number of share (by number balances (by outstanding Bank accounts of accounts) ($ billion) balances) Citibank 10,000,000 8.4 $15.3B 16.3 Chase Manhattan Bank 5,000,000 4.2 $5.4B 5.8 Bank of America 4,800,000 4.0 $5.2B 5.5 First Chicago 4,500,000 3.8 $4.6B 4.9 Manufacturers Hanover 3,300,000 2.8 $2.0B 2.1 Wells Fargo Bank 1,800,000 1.5 $2.8B 3.0 Maryland Bank 1,800,000 1.5 $1.7B 1.8 Marine Midland Bank 1,700,000 1.4 $1.4B 1.5 Chemical Bank 1,500,000 1.3 $1.3B 1.4 Associates National Bank 1,200,000 1.0 $1.0B 1.1 Top ten 35,600,000 30.0 $40.7B 43.4 Second ten 11,500,000 9.7 $9.0B 9.6 Total 118,900,000 100.0 $93.9B 100.0 Sources: Individual banks' numbers of accounts surveyed by American Banker, (March 1, 1988, pp. 1-2) and Credit Card News (August 15, 1988, pp. 4-16); total number of accounts from Nilson Report (Number 406 [June 1987], p. 4). Individual banks' outstanding balances based on American Banker (September 21, 1987, p. 43 [call report data]) and Nilson Report (Number 406 [June 1987], pp. 4-5); total outstanding balances from Federal Reserve Bulletin (December 1988, table 1.55; revolving credit outstanding at commercial banks and savings institutions, minus loans outstanding at Greenwood Trust Co. [Discover card] and American Express Centurion Bank). Data reported for December 31, 1986, adjusted for acquisitions effective in 1987. Conflicts between sources were resolved using best available information. branch banking restrictions. Indeed, the largest issuers today conduct truly national businesses. For example, Maryland Bank (ranked number seven in Table 1) conducts business in all 50 states and has only five percent of its accounts in its home state.8 The only states where more than five per- cent of its business is concentrated are Cali- fornia (10.7 percent), Texas (6.7 percent), Pennsylvania (6.0 percent), and New Jersey (5.8 percent). In the past, credit card issuers were con- strained by state usury laws. However, the U.S. Supreme Court's December 1978 Mar- quette decision paved the way for the practi- cal elimination of price regulations.9 The Court held that only the usury ceiling of the state in which the bank is located, and not that of the state in which the consumer is located, restricts the interest rate the bank may charge. This gave banks the option of shifting their credit card operations to wholly owned subsidiaries situated in states without usury laws. By 1982, amid Mar- quette-created bank pressure and histori- cally high market interest rates, most lead- ing banking states had relaxed or repealed their interest rate ceilings. Meanwhile, South Dakota and Delaware had estab- lished themselves as attractive homes- away-from-home for credit card issuers. While a number of states maintain binding usury laws at this writing (most notably, Arkansas, with a ceiling of five percentage points above the Federal Reserve discount rate), essentially all major issuers can pur- sue business in those states free of restric- 8"Prospectus for Maryland Bank, N.A., Credit Card Trust 1987-A," December 9, 1987, pp. 17-18. 9Marquette National Bank v. First of Omaha Service Corporation, 439 U.S. 299 (1978). The Marquette deci- sion applies to credit cards issued by nationally char- tered banks, but not to retail cards (e.g., oil company credit cards). The decision explicitly permits banks to "export" their interest rates; banks have interpreted this also to permit the "export" of annual fees and other customer fees. At this writing, at the behest of the Iowa Attorney General, courts are considering whether this rule does indeed apply to fees. 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
VOL 81 NO. I AUSUBEL: CREDIT CARD MARKET tion. It is fair to say that the bank credit mailing(11 responses) included a request card market in the United States was func- for direct profit calculations, which were tionally deregulated in 198 provided by seven banks. Appendix a pro- vides details of the construction of the I. Credit Card Interest Rate Behavior BCCS. Table 2 includes the size distribution of banks that reported data. Respondents A. Sticky Interest Rates were promised anonymity. The most aesthetically pleasing way for The cost of funds is obviously the primary an economist to determine the cost of funds determinant of the marginal cost of lending is to"let the market decide it. "In the case via credit cards, and it is usually the only of credit cards, this is feasible because of omponent of marginal cost that varies the phenomenon of credit card securitiza- idely from quarter o quarter. Thus, a tion. Consistently, during 1987-1989, model of continuous spot market equilil credit-card-backed securities offered yields rium would predict a substantial degree of in the vicinity of 0. 75 percent above those of onnection between the interest rate Treasury securities with comparable maturi charged on credit cards and the banks ' cost ties 2 Meanwhile the visa systemwide av- of funds. However, Figure 1, which com- erage cardholder payment rate (i.e, card- pares credit card interest rates with the cost holder payments as a percentage of out of funds, displays stark empirical rejection standing balances) ranged from 13 to 17 of this prediction. Credit card interest rates percent per month during the years were highly sticky during the period 1983-1987, implying an average maturity for 1982-1989 and, in fact, were virtually con- credit card receivables of 6-8 months. 3To ervative. i will define the cost of In this section, credit card interest rates funds to equal the one-year Treasury bill are captured by two distinct sets of data: yield 4 plus 0.75 percent, averaged over ne aggregated and one disaggregated. The each quarter. This series is also plotted in first set of data is the Federal Reserve Bul- Figure 1 letin series for credit card interest rates The proposition that interest rates are based on the Federal Reserve Board's quar- sticky can be formally supported by regress terly survey of banks. Reported are arith- ing credit card interest rates on the cost of metic averages of each bank's"most com mon"rate charged during the first week of each mid-quarter month This series is plotted in Figure 1. The second set of data IsEe, for example, "Credit Card Bonds are Hot (and much of the empirical discussion of 16. 1990.p. cIn: Credit icard the authors own bank credit card survey (November 15, 1988), p. 7; Credit Card management BCCS)of 58 of the largest bank issuers of May/June 1988, p. 34 credit cards. The first mailing(21 responses) asked primarily for pricing and cost data: it rate is Standard Poor's Asset-Backed Securitization generated a quarterly interest rate series for CreditReview, March 16, 1987, p. 19. Individual banks rospectuses have reported cardholder payment rates 17 credit card issuers and an annual loan- of 9-23 percent per month, never implying an average loss series for 10 issuers. The follow-up tuses in Appendix B). This impression was substanti. maturity of more port quote INDeed, the al Bank( First Chicago' s Delaware ary, listed fourth in Table 1)as Reserve bullet the period 1982-1989 is 18.85 saying that his bank finances its credit card portfolio ted is 17.77 percent(fourth quarter, 1988/lue re- with a variety of financial instruments with combined 985) and the lowest va maturit nt to a 145-day duration (Credir Federal Reserve Boards G. 19 stat April 5, 1990; Federal Reserve Bulletin, April 1990 table 1.56, line 4(and previous issue April 1990, table 1.35 line 21(and previous This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
VOL. 81 NO. 1 AUSUBEL: CREDIT CARD MARKET 53 tion. It is fair to say that the bank credit card market in the United States was func- tionally deregulated in 1982. II. Credit Card Interest Rate Behavior A. Sticky Interest Rates The cost of funds is obviously the primary determinant of the marginal cost of lending via credit cards, and it is usually the only component of marginal cost that varies widely from quarter to quarter. Thus, a model of continuous spot market equilib- rium would predict a substantial degree of connection between the interest rate charged on credit cards and the banks' cost of funds. However, Figure 1, which com- pares credit card interest rates with the cost of funds, displays stark empirical rejection of this prediction. Credit card interest rates were highly sticky during the period 1982-1989 and, in fact, were virtually con- stant.'0 In this section, credit card interest rates are captured by two distinct sets of data: one aggregated and one disaggregated. The first set of data is the Federal Reserve Bul- letin series for credit card interest rates, based on the Federal Reserve Board's quar- terly survey of banks. Reported are arith- metic averages of each bank's "most com- mon" rate charged during the first week of each mid-quarter month." This series is plotted in Figure 1. The second set of data (and much of the empirical discussion of this and the next section) is derived from the author's own bank credit card survey (BCCS) of 58 of the largest bank issuers of credit cards. The first mailing (21 responses) asked primarily for pricing and cost data; it generated a quarterly interest rate series for 17 credit card issuers and an annual loan- loss series for 10 issuers. The follow-up mailing (11 responses) included a request for direct profit calculations, which were provided by seven banks. Appendix A pro- vides details of the construction of the BCCS. Table 2 includes the size distribution of banks that reported data. Respondents were promised anonymity. The most aesthetically pleasing way for an economist to determine the cost of funds is to "let the market decide it." In the case of credit cards, this is feasible because of the phenomenon of credit card securitiza- tion. Consistently, during 1987-1989, credit-card-backed securities offered yields in the vicinity of 0.75 percent above those of Treasury securities with comparable maturi- ties.12 Meanwhile, the Visa systemwide av- erage cardholder payment rate (i.e., card- holder payments as a percentage of out- standing balances) ranged from 13 to 17 percent per month during the years 1983-1987, implying an average maturity for credit card receivables of 6-8 months.13 To be conservative, I will define the cost of funds to equal the one-year Treasury bill yield14 plus 0.75 percent, averaged over each quarter. This series is also plotted in Figure 1. The proposition that interest rates are sticky can be formally supported by regress- ing credit card interest rates on the cost of 10Indeed, the highest value reported in the Federal Reserve Bulletin series in the period 1982-1989 is 18.85 percent (first quarter, 1985) and the lowest value re- ported is 17.77 percent (fourth quarter, 1988). "1Federal Reserve Board's G.19 statistical release, April 5, 1990; Federal Reserve Bulletin, April 1990, table 1.56, line 4 (and previous issues). 12See, for example, "Credit Card Bonds are Hot, but Maybe Stingy on Yield," Wall Street Journal, April 16, 1990, p. C1; Credit Card News, Volume 1, Number 3 (June 15, 1988), p. 2, and Volume 1, Number 14 (November 15, 1988), p. 7; Credit Card Management, May/June 1988, p. 34. O3The source of the systemwide cardholder payment rate is Standard & Poor's Asset-Backed Securitization CreditReview, March 16, 1987, p. 19. Individual banks' prospectuses have reported cardholder payment rates of 9-23 percent per month, never implying an average maturity of more than one year (see list of prospec- tuses in Appendix B). This impression was substanti- ated by a trade-publication report quoting the chair- man of FCC National Bank (First Chicago's Delaware credit card subsidiary, listed fourth in Table 1) as saying that his bank finances its credit card portfolio with a variety of financial instruments with combined maturities equivalent to a 145-day duration (Credit Card News, March 15, 1989, p. 2). "Federal Reserve Bulletin, April 1990, table 1.35, line 21 (and previous issues). 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 CredIt Card Interest Rate 10% Cost of Funds 0L⊥⊥1111⊥⊥⊥⊥⊥⊥⊥⊥ 19821983198419851988198719881989 FIGURE 1. STICKY CREDIT CARD INTEREST RATES, 1982-1989 Credit card interest rate is the quarterly Federal Reserve System series; cost of s the quarterly car Treasury bill yield plus 0. 75 percen funds. First, using the Federal Reserve se- rate is regressed on its own lagged value ries, an aggregate credit card interest rate is the lagged cost of funds egressed on its own lagged value, the lagged variable for that bank.The of the ost of funds, and a constant. Second, a linear regressions are reported in Table 3 more thorough regression can be run using Note that, in the second regression, every the author's bCCs series for the 17 individ- coefficient has a t statistic of at least 6 al banks: each banks credit card interest while inclusion of additional variables with TABLE 2-SIZE OF BANK CREDIT CARD ISSUERS FOR WHICH DATA ARE REPORTED Number of banks Number of banks Number of banks 1987 ranking with BCcs with bccs of active accounts mputation estimation of profits rate series 3 2 41 51+ 0 2 This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
54 THE AMERICAN ECONOMIC REVIEW MARCH 1991 20% ....... Credit Card Interest Rate 5% _ 1982 1983 1984 1985 1988 1987 1988 10% Year FIGURE 1. STICKY CREDIT CARD INTEREST RATES, 1982-1989 Notes: Credit card interest rate is the quarterly Federal Reserve System series; cost of funds is the quarterly one-year Treasury bill yield plus 0.75 percent. funds. First, using the Federal Reserve se- ries, an aggregate credit card interest rate is regressed on its own lagged value, the lagged cost of funds, and a constant. Second, a more thorough regression can be run using the author's BCCS series for the 17 individ- ual banks: each bank's credit card interest rate is regressed on its own lagged value, the lagged cost of funds, and a dummy variable for that bank. The results of these linear regressions are reported in Table 3. Note that, in the second regression, every coefficient has a t statistic of at least 6, while inclusion of additional variables with TABLE 2-SIZE OF BANK CREDIT CARD ISSUERS FOR WHICH DATA ARE REPORTED Number of banks Number of banks Number of banks Number of banks 1987 ranking with BCCS with call with prospectuses with BCCS by number reports enabling reports enabling enabling reports of of active accounts computation computation estimation interest (1 = largest) of profits of profits of profits rate series 1-10 1 3 4 2 11-20 2 3 0 5 21-30 2 1 1 3 31-40 1 1 2 1 41-50 1 1 1 4 51+ 0 0 0 2 Total: 7 9 8 17 Sources: Author's bank credit card survey (BCCS), consolidated reports of condition and income (call reports), and prospectuses and registration statements. Ranks according to The Nilson Report, Number 406 (June 1987), pp. 6-7. 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
VOL 8I NO. I AUSUBEL: CREDIT CARD MARKET TABLE 3--ORDINARY LEAST-SOUARES REGRESSION OF CREDIT CARD INTEREST RATE ON COST OF other lags tended to cause some coefficients FUNDS AND LAGGED CREDIT CARD INTEREST RATE to become insignificant. To aid in compar- QUARTERLY, 1982-1987) ing the results of the two regressions, the Fed series is used only for the period Federal Reserve 1982-1987; using 1982-1989 data yields the Variable same conclusior survey data survey data The coefficient on the cost of funds while statistically significant in each of the two (000584) .00896) regressions, is economically insignificant CREDIT CARD INTEREST Whereas a competitive-spot-market model RATE (0.0444) (0.0326) would predict a coefficient near 1, the re- gressions using aggregated and disaggre gated data yielded coefficients of only 0.042 Bank-1 dummy and 0.054, respectively. It takes many years Bank-2 dummy (0.659) for the price to adjust to changes in marginal (0.594) cost when the rate of adjustment is only on Bank-3 dummy he order of 5 percent per quarter. (0.719) Bank-4 dummy B. Nonprice Competition Bank-5 dummy (0.651) The credit card industry has defended Bank-6 dumm high interest rates in the mid-to-late 1980,s, in part, by asserting that the increased Bank-7 dummy spread between the credit card interest rate Bank-8 dumn and the cost of funds had been caused by an (0.595) ncrease in the industry's rate of bad loans Bank-9 dummy The loan-loss data from the authors bccs (0.644) Bank-10 dummy indicate that, in the period 1982-1987, the (0.634) charge-off rate actually did increase roughly Bank-1l dummy coincident with the increase in the interest (0.589) rate spread (see Table 4). However, higher Bank-12 dummy (0.595) loan losses are an explanation for the higher interest rate spreads only if one believes (0.586) that the latter are solely determined by costs. If credit card interest rates are determined (0.577) otherwise then the causation may run in the (0.636) reverse direction: an increased interest rate Bank-16 dummy spread may cause an increase in charge-offs Suppose, for example, that a bank can Bank-17 dummy select both its interest rate and the default (0.707) risk of its marginal customer. By choosing a Number of higher marginal default rate, the bank in creases its total number of loans but also its 0.937 charge-off rate(the average default rate Durbin h 0.10 Suppose that the bank first selects its inte Notes: CREDIT CARD INTEREST RATE is the de- est rate and then its marginal default rate pendent variable in each regression. COST OF FUNDS Profit maximization requires the bank to set defined as the yield on one-year Treasury bills plus 75 percent. Observe that there is no cross-firm varia ion in this variable, so that year dummy variables ded in the Federal reserve bulletin Banks included in the au t of 0.0439 on COSt UNDS,a 0.864 on CREDIT CARD standard errors NTEREST R This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
VOL. 81 NO. 1 AUSUBEL: CREDIT CARD MARKET 55 TABLE 3-ORDINARY LEAST-SQUARES REGRESSION OF CREDIT CARD INTEREST RATE ON COST OF FUNDS AND LAGGED CREDIT CARD INTEREST RATE (QUARTERLY, 1982-1987) Federal Reserve Bank Board credit card Variable survey data survey data COST OF 0.0422 0.0540 FUNDS-1 (0.00584) (0.00896) CREDIT CARD INTEREST 0.895 0.685 RATE-1 (0.0444) (0.0326) Constant 1.51 (0.807) Bank-1 dummy 5.75 (0.659) Bank-2 dummy 5.11 (0.594) Bank-3 dummy 6.38 (0.719) Bank-4 dummy 5.79 (0.657) Bank-5 dummy 5.70 (0.651) Bank-6 dummy 4.18 (0.500) Bank-7 dummy 5.69 (0.645) Bank-8 dummy 5.12 (0.595) Bank-9 dummy 5.61 (0.644) Bank-10 dummy 5.44 (0.634) Bank-11 dummy 5.00 (0.589) Bank-12 dummy 5.12 (0.595) Bank-13 dummy 4.99 (0.586) Bank-14 dummy 4.88 (0.577) Bank-15 dummy 5.50 (0.636) Bank-16 dummy 5.22 (0.593) Bank-17 dummy 6.17 (0.707) Number of observations: 24 408 R 2: 0.96 0.937 Durbin h: - 0.69 0.10 Notes: CREDIT CARD INTEREST RATE is the de- pendent variable in each regression. COST OF FUNDS is defined as the yield on one-year Treasury bills plus 0.75 percent. Observe that there is no cross-firm varia- tion in this variable, so that year dummy variables cannot be included in the second regression equation. Banks included in the author's bank credit card survey were assured anonymity. Numbers in parentheses are standard errors. other lags tended to cause some coefficients to become insignificant. To aid in compar- ing the results of the two regressions, the Fed series is used only for the period 1982-1987; using 1982-1989 data yields the same conclusions.15 The coefficient on the cost of funds, while statistically significant in each of the two regressions, is economically insignificant. Whereas a competitive-spot-market model would predict a coefficient near 1, the re- gressions using aggregated and disaggre- gated data yielded coefficients of only 0.042 and 0.054, respectively. It takes many years for the price to adjust to changes in marginal cost when the rate of adjustment is only on the order of 5 percent per quarter. B. Nonprice Competition The credit card industry has defended its high interest rates in the mid-to-late 1980's, in part, by asserting that the increased spread between the credit card interest rate and the cost of funds had been caused by an increase in the industry's rate of bad loans. The loan-loss data from the author's BCCS indicate that, in the period 1982-1987, the charge-off rate actually did increase roughly coincident with the increase in the interest rate spread (see Table 4). However, higher loan losses are an explanation for the higher interest rate spreads only if one believes that the latter are solely determined by costs. If credit card interest rates are determined otherwise, then the causation may run in the reverse direction: an increased interest rate spread may cause an increase in charge-offs. Suppose, for example, that a bank can select both its interest rate and the default risk of its marginal customer. By choosing a higher marginal default rate, the bank in- creases its total number of loans but also its charge-off rate (the average default rate). Suppose that the bank first selects its inter- est rate and then its marginal default rate. Profit maximization requires the bank to set 15With the 1982-1989 Federal Reserve Bulletin se- ries, one obtains a coefficient of 0.0439 on COST OF FUNDS_ 1, a coefficient of 0.864 on CREDIT CARD INTEREST RATE_ 1, and a constant of 2.06. 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 TABLE 4-LOAN LOSSES ON CREDIT CARDS OF TEN SURVEY BANKS DURING 1976-1987 Ill. The Ex Post Profitability of the Credit D THE SPREAD BETWEEN CREDIT CARD INTEREST RATES AND THE COST OF FUN As seen in Section I, the credit card m ket of the 1980,'s possessed most of the Inte orea te usual prerequisites for invoking the model Year percentage (percentage) of perfect competition. A perfectly competi- 10 run economic profits for“ marginal”frms. 10.36 Moreover, since free entry into the industry 1979 is possible and no input appears to be in scarce supply, there is no credible source of rents to distinguish“ inframarginal” firms from"marginal"firms. Thus, the competi tive model would predict that all credit card issuers earn zero long-run economic profits 11.55 Many models of imperfect competition which preserve the free-entry assumption Source: Author 's bank credit card survey (Appendix a, would also yield the zero-profit prediction Table Al, questions 7 and 1). By way of contrast, the interest rate stick iness documented in the previous section suggests that credit cards must become ex- ts marginal default rate equal to the dif- traordinarily profitable whenever the cost of ference between the interest rate it charges funds drops. Indeed, in this section, I will and the marginal cost (net of defaults) of present a rather paradoxical set of data lending funds. (The net marginal cost should which indicates that returns from the credit equal the cost of funds plus a constant that card business were several times greater is fairly stable in the short run. )Thus, the than the ordinary rate of return in banking prediction is that an optimizing bank should during the years 1983-1988 set its marginal default rate equal to the At the same time this profitability data dent reason why credit card interest rates interpreted. On do u that the above evidence interest rate spread plus a constant. will help to assure Suppose now that there is an indeper of interest rate stickiness has been correctly might have thought to ar- fail to fall when general market interest gue that price rigidity is consistent with rates decline(for example, see Section VI below ). The logic of the previous paragraph dictates that loan losses will subsequently cards, there is stickiness despite a deregulated increase. If firms do not compete and drive environment. Second, under regulation, the airlines price down toward marginal cost, they are appi Free entry is, a reasonable depiction of a credit ently competed away their profits. likely instead to compete and drive marginal card market in which 4,087 banks (and other deposit cost up toward price, o in the form of is ing cards to less credit-worthy customers institutions) already issued their own Visa card similar (largely overlapping) number issued thei Master Cards in September 1987. All of these instit ons could legally offer accounts to customers any where in the United States. Nonmember institutions LA related argu could join the Visa system by paying a fairly trivial eorge w. hat the ci rd's one thousand dollars, according to a visa official (Only price regulations, at a time when the introduction of jet the assets of the subsidiary that issues the cards, and engines reduced the fundamental cost of air trans- not the assets of the holding company are figured into portation es to compete and drive this formula. Furthermore, it would seem strained to p to price by er passengers o n requires airplane. The ar many years or tha put is in supply, given First. in the airli the deluge of cre solicitations made by banks in been caused by price regulation, whereas with credit recent years This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
56 THE AMERICAN ECONOMIC REVIEW MARCH 1991 TABLE 4-LoAN LOSSES ON CREDIT CARDS OF TEN SURVEY BANKS DURING 1976-1987 (FEWER THAN NINE DURING 1976-1978 AND 1986-1987) AND THE SPREAD BETWEEN CREDIT CARD INTEREST RATES AND THE COST OF FUNDS Average charge-off Interest rate rate spread Year (percentage) (percentage) 1976 1.15 10.57 1977 0.99 10.36 1978 1.27 8.11 1979 1.44 5.78 1980 2.04 5.08 1981 1.48 2.94 1982 1.67 6.42 1983 1.32 8.79 1984 1.36 7.69 1985 1.94 9.82 1986 3.01 11.55 1987 2.60 10.43 Source: Author's bank credit card survey (Appendix A, Table Al, questions 7 and 1). its marginal default rate equal to the dif- ference between the interest rate it charges and the marginal cost (net of defaults) of lending funds. (The net marginal cost should equal the cost of funds plus a constant that is fairly stable in the short run.) Thus, the prediction is that an optimizing bank should set its marginal default rate equal to the interest rate spread plus a constant. Suppose now that there is an indepen- dent reason why credit card interest rates fail to fall when general market interest rates decline (for example, see Section VI, below). The logic of the previous paragraph dictates that loan losses will subsequently increase. If firms do not compete and drive price down toward marginal cost, they are likely instead to compete and drive marginal cost up toward price,16 in the form of issu- ing cards to less credit-worthy customers. III. The Ex Post Profitability of the Credit Card Market As seen in Section I, the credit card mar- ket of the 1980's possessed most of the usual prerequisites for invoking the model of perfect competition. A perfectly competi- tive model would at least predict zero long- run economic profits for "marginal" firms. Moreover, since free entry into the industry is possible and no input appears to be in scarce supply,17 there is no credible source of rents to distinguish "inframarginal" firms from "marginal" firms. Thus, the competi- tive model would predict that all credit card issuers earn zero long-run economic profits. Many models of imperfect competition which preserve the free-entry assumption would also yield the zero-profit prediction. By way of contrast, the interest rate stick- iness documented in the previous section suggests that credit cards must become ex- traordinarily profitable whenever the cost of funds drops. Indeed, in this section, I will present a rather paradoxical set of data which indicates that returns from the credit card business were several times greater than the ordinary rate of return in banking during the years 1983-1988. At the same time, this profitability data will help to assure that the above evidence of interest rate stickiness has been correctly interpreted. One might have thought to ar- gue that price rigidity is consistent with "A related argument was made in the context of airline regulation. George W. Douglas and James C. Miller (1974) argued that the Civil Aeronautics Board's price regulations, at a time when the introduction of jet engines reduced the fundamental cost of air trans- portation, led airlines to compete and drive their costs up to price by placing fewer passengers on a given airplane. The arguments differ in two important ways. First, in the airline industry, price rigidity may have been caused by price regulation, whereas with credit cards, there is price stickiness despite a deregulated environment. Second, under regulation, the airlines apparently competed away their profits. 17Free entry is a reasonable depiction of a credit card market in which 4,087 banks (and other deposit institutions) already issued their own Visa cards and a similar (largely overlapping) number issued their own MasterCards in September 1987. All of these institu- tions could legally offer accounts to customers any- where in the United States. Nonmember institutions could join the Visa system by paying a fairly trivial entry fee: six dollars per million dollars in assets, plus one thousand dollars, according to a Visa official. (Only the assets of the subsidiary that issues the cards, and not the assets of the holding company, are figured into this formula.) Furthermore, it would seem strained to argue either that adjustment to the "long run" requires many years or that some input is in scarce supply, given the deluge of credit card solicitations made by banks in recent years. 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
VOL. 81 NO. I AUSUBEL: CREDIT CARD MARKET competitive spot markets, if unobservable Bank credit card increases in quality exactly offset reductions low-up survey n factor costs. The profitability data enable performed direct one to dismiss this possibility: profits, in of the 50 larges fact, dramatically rose at the time that the cards cost of funds dropped Call reports: Profitability data for another It is possible to object to the following nine of the extracted reported, by their very nature, represent FDIC call reports filed by the banks with the ex post profits. Perhaps(especially since the Prospectuses: Partial data on profitability for sample period is during a cyclical boom)the an additional eight large banks were ob- observed profits are merely a very favorable tained from filings with the SeC in con- realization of a random variable whose nection with the sale of credit-card ex ante returns were quite ordinary Second backed securities it might be thought that, while the credit card market was extremely profitable in Respondents to the author's survey were the years 1983-1988, the market has now promised anonymity (but details of the con quilibrated and henceforth normal returns struction are provided in Appendix A). The will be observed. Third, the profitability call reports and prospectuses are part of the figures might be derived from accounting public record. Table 2 reports the size dis- data that either are being misinterpreted or tribution of banks included in each of the are systematically misstating true economic survey, call report, and prospectus samples I consider each of these concerns else A. An Illustrative Profit Calculation where in the paper. In Section IV, I briefly discuss an additional source of evidence(the As will be detailed in the next two sub Federal Reserve Systems functional cost sections, earnings in the banking industry analysis), which, while significantly less reli- are usefully expressed as a percentage of able than the other data (in this author's assets: returns on assets are linked with opinion), gives profits over a longer period returns on equity by the banking system's that includes the previous cyclical down- capital requirements. Before reporting turn. In Section V, I introduce another in- mary profit figures for 15 and estimates for dependent set of data which examines eight of the 50 largest issuers, I will examine sale prices of credit card portfolios between in detail the components of revenues and banks and finds that they trade at large costs for one individual credit card issuer, I premia. The latter data indicate that ex ante consider here Maryland Bank, Na returns from credit cards are quite large (MBNA), the Delaware-based credit card and, since they are based on market valua- arm of MNC Financial. which is ranked tions, should help allay any fears that the seventh in Table 1. 18 This institution was accounting data are being misinterpreted. selected because more public information Finally, it should be recalled from Table 4 exists on its credit card operations than on that the interest rate spread was quite any other banks: MBNA, which is required healthy except for a brief period around to file its own call report, has credit card 1981 and that this brief spell of unprof- loans exceeding 92 percent of its assets, and ability can be attributed to banks not hav- it has also made several credit-card-backed ng yet established credit card subsidiaries securities offerings exploiting the Supreme Court's marquette ecision. This episode does not seem likely I MNC Financial is the 39th largest U.S ussed in this section originate from three National Bank. nthe the csr orate parent of Maryland The ex post profit data re and. MBna was founded in Newark. Di independent sources and were assembled by 1982, apparently to avoid Marylands usury the author also the text near footnote 9 This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
VOL. 81 NO. 1 AUSUBEL: CREDIT CARD MARKET 57 competitive spot markets, if unobservable increases in quality exactly offset reductions in factor costs. The profitability data enable one to dismiss this possibility: profits, in fact, dramatically rose at the time that the cost of funds dropped. It is possible to object to the following analysis on several grounds. First, the data reported, by their very nature, represent ex post profits. Perhaps (especially since the sample period is during a cyclical boom) the observed profits are merely a very favorable realization of a random variable whose ex ante returns were quite ordinary. Second, it might be thought that, while the credit card market was extremely profitable in the years 1983-1988, the market has now equilibrated and henceforth normal returns will be observed. Third, the profitability figures might be derived from accounting data that either are being misinterpreted or are systematically misstating true economic profits. I consider each of these concerns else- where in the paper. In Section IV, I briefly discuss an additional source of evidence (the Federal Reserve System's functional cost analysis), which, while significantly less reli- able than the other data (in this author's opinion), gives profits over a longer period that includes the previous cyclical down- turn. In Section V, I introduce another in- dependent set of data which examines re- sale prices of credit card portfolios between banks and finds that they trade at large premia. The latter data indicate that ex ante returns from credit cards are quite large and, since they are based on market valua- tions, should help allay any fears that the accounting data are being misinterpreted. Finally, it should be recalled from Table 4 that the interest rate spread was quite healthy except for a brief period around 1981 and that this brief spell of unprof- itability can be attributed to banks not hav- ing yet established credit card subsidiaries exploiting the Supreme Court's Marquette decision. This episode does not seem likely to be repeated. The ex post profit data reported and dis- cussed in this section originate from three independent sources and were assembled by the author. Bank credit card survey: The author's fol- low-up survey yielded profit calculations performed directly by executives of seven of the 50 largest bank issuers of credit cards. Call reports: Profitability data for another nine of these issuers were extracted from call reports filed by the banks with the FDIC. Prospectuses: Partial data on profitability for an additional eight large banks were ob- tained from filings with the SEC in con- nection with the sale of credit-card- backed securities. Respondents to the author's survey were promised anonymity (but details of the con- struction are provided in Appendix A). The call reports and prospectuses are part of the public record. Table 2 reports the size dis- tribution of banks included in each of the survey, call report, and prospectus samples. A. An Illustrative Profit Calculation As will be detailed in the next two sub- sections, earnings in the banking industry are usefully expressed as a percentage of assets: returns on assets are linked with returns on equity by the banking system's capital requirements. Before reporting sum- mary profit figures for 15 and estimates for eight of the 50 largest issuers, I will examine in detail the components of revenues and costs for one individual credit card issuer. I consider here Maryland Bank, N.A. (MBNA), the Delaware-based credit card arm of MNC Financial, which is ranked seventh in Table 1.18 This institution was selected because more public information exists on its credit card operations than on any other bank's: MBNA, which is required to file its own call report, has credit card loans exceeding 92 percent of its assets, and it has also made several credit-card-backed securities offerings. 18MNC Financial is the 39th largest U.S. bank hold- ing company and the corporate parent of Maryland National Bank, the largest commercial bank in Mary- land. MBNA was founded in Newark, Delaware, in 1982, apparently to avoid Maryland's usury law. See also the text near footnote 9. 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 REVEn MARCH 1991 TABLE 5-COMPONENTS OF PROFITS FOR MARYLAND BANK, NA Finance charges 1492 13.21% 158% Other customer charges 1.42% 1.17% Total revenue. Net charge-offs 109% 1.77% 1.80% rofi urces: Consolidated reports of condition and income (call reports), prospectuses, and registration statements for Maryland Bank, N.a. MBNA's credit card operations(and their An item-by-item profit calculation for profitability) are fairly typical of major is- MBNA is displayed in Table 5. As is typical suers, with the exception that the bank has for credit card issuers, the single largest stressed the concept of " affinity credit component of revenue is the finance charge cards, whereby cards are marketed to (which, for MBNA, derives from annual members of professional organizations, fra- percentage rates of 14.5-18.9 percent, de- ternal orders, and cause- related groups (with pending on the account). Despite the fact the organizations'endorsements). As a con- that the bank provides a 25-day grace pe- sequence, its interest rates are somewhat riod during which no finance charge is as- lower and its customers are somewhat more sessed if the account balance is paid in full redit-worthy than average. Indeed it may more than 80 percent of the banks credit interest readers that, during the period when card outstanding balances do accrue inter his article was undergoing the journals re- est. The drop in finance-charge revenues iew process, MBNA's marketing agent pro- displayed in Table 5 is largely attributable posed to establish an official American Eco- to the banks decision to reduce the interest nomic Association Visa card. This card rates on some of its accounts during would have carried a $20 annual fee($40 1985-1987 for a gold card)and an 189-percent annual MBNA also derives direct customer rey. ee: the AEa would have received s1 for enues from the annual fee and other each account opened, $3 for each account tomer charges(e.g, $15 late payment, over- renewed, and $0. 25 per retail transaction. limit, and returned-check charges). Indirect MBNA S agent estimated that 1, 000 cards revenues are derived from the interchange would be issued, generating $13 in revenue fee, the portion of the merchant discount per card per year for the AEA. However, that is paid to the customer's bank. It is the AEAs executive committee, concerned worth reemphasizing that the price sched that the aea" would be viewed as endors- ing a specific credit card by entering into such a contract, voted against establishing tary to the Executive Committee. I thank Orley Ashen the affinity card program. 9 felter and C. Elton Hinshaw for providing this informa A good rule of thumb mentioned in credit card ade publications is that 90 percent of i Draft minutes of the March 23, 1990, meeting of overall outstanding balances accrue interest. See the the AEA Executive Committee: Report of the Secre- discussion in Section VI-C. This content downloaded from 202. 120. 224.93 on Sun. 17 Dec 201707: 42: 57 UTC Allusesubjecttohttp:/about.jstor.org/terms
58 THE AMERICAN ECONOMIC REVIEW MARCH 1991 TABLE 5-COMPONENTS OF PROFITS FOR MARYLAND BANK, N.A. Components 1985 1986 1987 Finance charges 16.66% 14.92% 13.21% Annual fees 1.40% 1.58% 1.29% Other customer charges 1.10% 1.42% 1.17% Interchange fees 3.06% 3.00% 2.92% Total revenue: 22.22% 20.92% 18.60% Interest expenses 9.57% 7.80% 7.13% Noninterest expenses 4.47% 4.71% 4.87% Net charge-offs 1.09% 1.77% 1.80% Total cost: 15.13% 14.28% 13.80% Return on assets (pretax profits expressed as a percentage of outstanding balances) 7.09% 6.63% 4.80% Sources: Consolidated reports of condition and income (call reports), prospectuses, and registration statements for Maryland Bank, N.A. MBNA's credit card operations (and their profitability) are fairly typical of major is- suers, with the exception that the bank has stressed the concept of "affinity credit cards," whereby cards are marketed to members of professional organizations, fra- ternal orders, and cause-related groups (with the organizations' endorsements). As a con- sequence, its interest rates are somewhat lower and its customers are somewhat more credit-worthy than average. Indeed, it may interest readers that, during the period when this article was undergoing the journal's re- view process, MBNA's marketing agent pro- posed to establish an official American Eco- nomic Association Visa card. This card would have carried a $20 annual fee ($40 for a gold card) and an 18.9-percent annual fee; the AEA would have received $1 for each account opened, $3 for each account renewed, and $0.25 per retail transaction. MBNA's agent estimated that 1,000 cards would be issued, generating $13 in revenue per card per year for the AEA. However, the AEA's executive committee, concerned that the AEA "would be viewed as endors- ing a specific credit card by entering into such a contract," voted against establishing the affinity card program.'9 An item-by-item profit calculation for MBNA is displayed in Table 5. As is typical for credit card issuers, the single largest component of revenue is the finance charge (which, for MBNA, derives from annual percentage rates of 14.5-18.9 percent, de- pending on the account). Despite the fact that the bank provides a 25-day grace pe- riod during which no finance charge is as- sessed if the account balance is paid in full, more than 80 percent of the bank's credit card outstanding balances do accrue inter- est.20 The drop in finance-charge revenues displayed in Table 5 is largely attributable to the bank's decision to reduce the interest rates on some of its accounts during 1985-1987. MBNA also derives direct customer rev- enues from the annual fee and other cus- tomer charges (e.g., $15 late payment, over- limit, and returned-check charges). Indirect revenues are derived from the interchange fee, the portion of the merchant discount that is paid to the customer's bank. It is worth reemphasizing that the price sched- 19Draft minutes of the March 23, 1990, meeting of the AEA Executive Committee; Report of the Secre- tary to the Executive Committee. I thank Orley Ashen- felter and C. Elton Hinshaw for providing this informa- tion. 20A good rule of thumb mentioned in credit card trade publications is that 90 percent of an issuer's overall outstanding balances accrue interest. See the discussion in Section VI-C. 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