CREDIT CARDS AND BANKRUPTCY
TODD J. ZYWICKI
Professor of Law, George Mason University School of Law
Research Fellow, James Buchanan Center for Political Economy
Abstract:
From 1980 to 2005 consumer bankruptcy filings increased five-fold.
Conventional wisdom holds that a primary cause of rising bankruptcy filing rates was
increased household financial distress caused by increased indebtedness caused in turn by
increased credit card borrowing. In 2005, Congress enacted the bipartisan Bankruptcy
Abuse Prevention and Consumer Protection Act (BAPCPA). The legislation was enacted
in response to twenty-five years of rising bankruptcy filings and a perception of
widespread fraud and abuse that threatened the fairness and integrity of the system.
BAPCPA marked the most profound and far-reaching overhaul of America’s bankruptcy
system in over a generation. In the two years since BAPCPA’s enactment, bankruptcy
filings have plunged. From over 2 million filings in 2005, filings plummeted to less than
600,000 in 2006 and 800,000 in 2007.
Critics of the legislation argue that the drop in filings will be temporary, as the
legislation does not address what they believe to be the underlying cause of the rise in
bankruptcy filings in the 1980s and 1990s—excessive consumer debt caused by
profligate lending by credit card issuers especially to risky borrowers. This article
reviews three hypotheses about the relationship between credit cards and bankruptcy: the
substitution model, the “distress” or “overindebtedness model,” and the “strategic model”
and concludes that the conventional wisdom is flawed. A review of empirical evidence
and available data indicates that in fact the growth in credit card lending has not led to an
increase in the consumer debt service ratio or financial distress more generally,
suggesting that the rise in credit card borrowing has been primarily a substitution from
other traditional types of consumer credit, such as retail store credit, personal finance
companies, friends and family, pawnbrokers, and other types of consumer credit.
The article then briefly examines the substitution hypothesis in more depth,
describing how a substitution to credit card debt can bring about a rise in consumer
bankruptcy filings even holding overall consumer debt obligations constant. Finally, the
article examines the rationale and effects of the credit card provisions of BAPCPA for the
substitution and distress models. To date, the response of consumers to BAPCPA has
been consistent with the substitution model, suggesting that with respect to addressing
particular problems regarding the relationship between credit cards and bankruptcy
BAPCPA has been accomplishing its stated purposes.
1
Credit Cards and Bankruptcy
By Todd J. Zywicki*
From 1980 to 2005 consumer bankruptcy filings increased five-fold.
Conventional wisdom holds that a primary cause of rising bankruptcy filing rates was
increased household financial distress caused by increased indebtedness caused in turn by
increased credit card borrowing.
In 2005, Congress enacted the Bankruptcy Abuse
Prevention and Consumer Protection Act (BAPCPA) by overwhelming bipartisan
majorities in both houses of Congress.1 The legislation was enacted in response to
twenty-five years of rising bankruptcy filings and a perception of widespread fraud and
abuse that threatened the fairness and integrity of the system. BAPCPA marked the most
profound and far-reaching overhaul of America’s bankruptcy system in over a generation.
In the two years since BAPCPA’s enactment, bankruptcy filings have plunged. From
over 2 million filings in 2005, filings plummeted to less than 600,000 in 2006 and
800,000 in 2007.
Despite the overwhelming bipartisan support for the legislation in Congress and
its apparent success in accomplishing its intended goals, BAPCPA has been criticized by
many bankruptcy scholars and professionals.
They argue that the rising trend in
bankruptcy filings in the 1980s and 1990s was a manifestation of widespread consumer
distress caused by excessive household debt obligations.2 Although multipronged, a core
*
Professor of Law, George Mason University School of Law; Research Fellow, James Buchanan Center
for the Study of Political Economy. I would like to thank Josh Wright for extremely helpful comments on
earlier drafts of this article as well as workshop participants at Duke, George Mason University,
Georgetown, and Vanderbilt law schools. I would like to thank the Law & Economics Center at George
Mason University School of Law for financial support for this project.
1
The vote in the United States Senate was 74 in favor and 25 opposed, with 1 abstention. The vote in the
House was 302 in favor and 126 opposed with 7 abstentions.
2
See Michelle J. White, Bankruptcy Reform and Credit Cards, 21 J. ECON. PERSPECTIVES 175 (2007) ;
TERESA A. SULLIVAN ET AL., THE FRAGILE MIDDLE CLASS: AMERICANS IN DEBT (2000); Elizabeth Warren,
2
element of the argument is that this economic distress was triggered by more widespread
access to credit cards by higher-risk borrowers over the past three decades. Typically it is
argued that credit cards combine high rates of interest with an “insidious” form of gradual
and subconscious debt accumulation through many routine purchases that draw
consumers into debt unconsciously3 or that credit card issuers prey on consumers’
cognitive biases and errors to trick them into excessive debt.4
Both credit card use and bankruptcies have increased dramatically over the past
several decades, and this rise in credit card use and bankruptcy appear to be correlated—
at least prior to BAPCPA.5
The raw numbers appear staggering at first glance.
Consumers charge over 1 trillion dollars per year on their credit cards and revolve over
$600 billion in credit card debt from month to month.6 Seventy-five percent of American
households have some sort of credit card, compared to seven percent in the 1960s.7
Credit cards have transformed the way we live, shop, and travel. The ubiquity and
importance of credit cards to the modern economy is summed up in the observation of
economist Lewis Mandell that credit cards “have become an essential element of daily
The Bankruptcy Crisis, 73 IND. L.J. 1079, 1082 (1998); RONALD J. MANN, CHARGING AHEAD: THE
GROWTH AND REGULATION OF PAYMENT CARD MARKETS 35-72 (2006).
3
See SULLIVAN ET AL., supra note 2, at 108-40; Juliet B. Schor, Who’s Going Bankrupt and Why?, 79 TEX.
L. REV. 1235, 1238 (2001).
4
See Lawrence M. Ausubel, The Failure of Competition in the Credit Card Market, 81 AM. ECON. REV.
50, 71 (1991); Oren Bar-Gill, Seduction by Plastic, 98 NW. U. L. REV. 1373 (2004).
5
See discussion infra at notes 110-111 and accompanying text. High levels of credit card use have
continued post-BAPCPA even as the number of bankruptcy filings has fallen by half.
6
See BOARD OF GOVERNORS OF THE FEDERAL RESERVE SYSTEM, REPORT TO THE CONGRESS ON PRACTICES
OF THE CONSUMER CREDIT INDUSTRY IN SOLICITING AND EXTENDING CREDIT AND THEIR EFFECTS ON
CONSUMER DEBT AND INSOLVENCY 7 (June 2006) (herinafter FED REPORT).
7
See FED REPORT, supra note, at 6. Seventy-one percent of families today have a bank-type card. “Banktype” card is used in this article to refer to general purpose payment cards such as Visa, Mastercard,
American Express, or Discover.
3
life. With a credit card, you can buy yourself a new car. Without it, you cannot even rent
one.”8
But with credit cards has come a chorus of critics who argue that this growth in
credit card ownership has come with an excessive social cost—too many consumers
overburdened by debt and an annual bankruptcy filing rate that quintupled between 1980
and 2005. Much of the blame for this skyrocketing bankruptcy filing rate is laid at the
door of credit card issuers who, it is said, have extended credit in a profligate manner to
unworthy borrowers and have lured borrowers into financial misery with easy credit, high
interest rates, and an array of hidden fees.
This article reviews the various theories that have been offered to explain the
correlation between credit cards and bankruptcy. Conventional wisdom holds that the
causal link is obvious: that the relationship between credit cards and bankruptcy is
obvious—that the rise in credit card use has produced a generation of overburdened
consumers driven into bankruptcy by profligate and abusive lending by credit card
issuers. This article tests the conventional wisdom and finds it wanting. In particular,
data indicate that credit cards have not in fact led to an increase in household economic
distress for a simple and, upon reflection, obvious reason: The growth in credit card
borrowing has been a substitution from other traditional types of consumer credit, such as
pawnshops, personal finance companies, retail store credit, department store credit cards
and gasoline cards, layaway plans, and other similar products.
Moreover, this
substitution has been driven by rational consumer demand. Credit cards offer lower
interest rates, better terms, ancillary benefits, and move flexibility than any other type of
consumer credit. Although this substitution has also had the unintended consequence of
8
LEWIS MANDELL, THE CREDIT CARD INDUSTRY: A HISTORY at p. xi (1990).
4
increasing bankruptcy filings, it has on the whole been a great boon to consumers and the
economy and it would be unwise to try to reverse it. Although total household has
increased, steady economic growth, record household wealth accumulation, declining
interest rates, and the substitution of high-interest consumer debt with lower-interest debt
has dramatically increased household living standards while leaving their debt service
burden largely unchanged for over twenty-five years.
This article examines competing hypotheses regarding the possible link between
credit cards and bankruptcy. After a brief background on the history of consumer credit
in America and the rise of credit cards within this story, the article turns toward a detailed
exploration of the competing hypotheses.
After establishing the primacy of the
substitution effect with respect to credit card credit, the article explores possible causal
explanations for how a simple change in the composition of household credit holdings
could lead to increased bankruptcies without an increase in the household debt burden.
Finally, the article turns to the effect of BAPCPA on credit card debt to date and the
implications for the future.
I.
“Charge It”: A Brief History of Credit Cards in America
Although credit cards are a relatively new financial innovation, the use of credit is
as old as human society and even predates money itself.9 And as old as credit is the
reality that some borrowers will be unable to repay all that they borrow and that many
others will engage in hand-wringing about others’ excessive borrowing.10 Lendol Calder
refers to this as the “myth of lost financial virtue”—the belief that “earlier generations”
9
SIDNEY HOMER & RICHARD SYLLA, A HISTORY OF INTEREST RATES 3 (3d ed. 1991); Marion Benfield,
Money, Mortgages, and Migraine–The Usury Headache, 19 CASE W. RES. L. REV. 819, 822 (1968)
10
See infra notes 29-36 and accompanying text.
5
were thriftier than the “current generation,” a refrain repeated steadily by almost every
generation since at least mid-nineteenth century America.11 In fact, credit cards have
spawned simply the latest expression of concern about consumer overindebtedness and
reckless lending that has recurred repeatedly in American history.
A.
Consumer Credit in Early America
In pre-Civil War America, most Americans were farmers, living outside major
population centers. Gold and silver coins were scarce. Personal credit, however, was
not, and farmers relied on credit to smooth investment and consumption across the crop
harvesting season. Credit was as important as the Conestoga Wagon in conquering the
west.12
In the decades following the Civil War, a tide of immigrants swept into America,
building the great cities. Most urban dwellers were unskilled blue-collar workers with
unpredictable employment and income, thus the consumer credit industry emerged to
cope with seasonal fluctuations in employment. The emerging American middle class
became homeowners and home furnishers through mortgages and consumer installment
credit. Overall, late nineteenth-century households sought financial assistance from five
major credit sources: pawnbrokers, illegal small-loan lenders, retailers, friends and
family, and mortgage lenders.13 In post-Civil War New York City, for instance, twothirds of the city’s total consumer lending came from small-loan agencies, including loan
11
LENDOL CALDER, FINANCING THE AMERICAN DREAM: A CULTURAL HISTORY OF CONSUMER CREDIT 22
(1999).
12
LEWIS MANDELL, THE CREDIT CARD INDUSTRY: A HISTORY at xii (1990).
13
CALDER, supra note 11, at 38.
6
sharks and forerunners to today’s and “payday” or “wage assignment” lenders.14 Pawn
shops proliferated—in some neighborhoods virtually the entire population had a pawn
ticket at all times and as many as twelve in the winter when factories typically closed
down.15
These various unlicensed lenders charged interest rates that approach 300
percent annually and resorted to embarrassing and aggressive collection practices to
enforce repayment of these illegal debts.16 Counterproductive usury regulations made
operations unprofitable for legitimate lenders thereby driving many urban consumers into
the hands of illegal lenders.17 It was estimated in 1911 that 35 percent of New York
City’s employees owed money to illegal loan sharks. 18 Reviewing the credit market of
this era, former Federal Reserve Chairman Alan Greenspan has described the plight of
lower-income wage earners subject to aggressive and overreaching creditors as one of
“virtual serfdom.”19
Consumer credit expanded following World War I. Credit unions, small local
savings banks, and a national network of licensed consumer finance companies such as
the Beneficial Industrial Loan Corporation and the Household Finance Corporation,
provided consumer loans.20 These were installment loans (often referred to as buying on
14
CALDER, supra note 11, at 49-50.
CALDER, supra note 11, at 44-48.
16
See CALDER, supra note 11, at 48-54. Interest rates on these loans were comparable to modern payday
lenders.
17
Alan Greenspan, Remarks, Economic Development Conference of the Greenlining Institute (Oct. 11,
1997), available in http://www.federalreserve.gov/BoardDocs/Speeches/1997/19971011.htm (hereinafter
Greenspan, Greenlining Remarks).
18
CALDER, supra note 11, at 118.
19
Greenspan, Greenlining Remarks, supra note 17.
20
Greenspan, Greenlining Remarks, supra note 17; Tom Brown & Lacey Plache, Paying with Plastic:
Maybe Not So Crazy, 73 U. CHICAGO L. REV. 63, 72 (2006).
15
7
“time”) meaning that the consumer borrowed a fixed sum (plus interest) with an
obligation to repay the loan over a fixed period in equal installments.21
Beginning with Singer sewing machines, installment credit soon spread to
furniture, pianos, and household appliances, and finally automobiles.22 By the 1930s, the
majority of sales of household furniture, appliances, radios, cameras, and jewelry were
credit sales, as were a substantial percentage of rugs, hardware, sporting goods, and
books (such as encyclopedia and other book sets).23 Financing these purchases through
credit made it possible to acquire and use the goods immediately, rather than having to
save for long periods of time to afford them. Between 1900 and 1939, total consumer
nonmortgage installment debt quadrupled in real dollars, increasing 2-1/2 times during
the decade of the 1920 alone. Martha Olney refers to this unprecedented twin boom in
the purchase of consumer durables and consumer credit the “consumer durables
revolution.”24 David Caplovitz’s study of low-income households, published in 1963,
found that retailers of consumer durables made on average 75 to 90 percent of their sales
on installment credit.25 Two-thirds of families who relocated into public housing projects
in the 1960s purchased their new furniture and major appliances on credit.26 Caplovtiz
also found that more than 70 percent of households used credit to purchase a television
set and that lower and higher-income families were just as likely to use credit to do so.27
21
For example, in a typical installment credit arrangement the debtor may finance a new dining room set
through the furniture store by borrowing $5000 at 8% interest (which may be imputed in the purchase price
rather than explicitly stated) repayable over 60 months in equal payments of $101.38, with the loan secured
by a purchase money security interest in the dining room set itself.
22
MARTHA L. OLNEY, BUY NOW, PAY LATER: ADVERTISING, CREDIT, AND CONSUMER DURABLES IN THE
1920S at p. 85 (1991); MANDELL, supra note 12, at xii.
23
OLNEY, supra note 22, at 100-01 Table 4.4.
24
See OLNEY, supra note 22, at 87-89.
25
DAVID CAPLOVITZ, THE POOR PAY MORE: CONSUMER PRACTICES OF LOW-INCOME FAMILIES 16 (1963).
26
CAPLOVITZ, supra note, at 96.
27
Id. at 88.
8
Overall, Caplovitz found that 81 percent of low-income families in his study made use of
some sort of credit (including installment loans, revolving credit for clothing, buying
from peddlers, or outstanding personal loans), and that 60 percent had consumer debts at
the time of his interview with them.28
The expressions of concern heard today about credit cards were presaged in
similar paternalistic comments about the spread of installment credit.29
Installment
selling was criticized for allegedly inducing overconsumption by American shoppers,
especially supposedly vulnerable groups such as “the poor, the immigrant, and the
allegedly math-impaired female.”30
Rapacious installment sellers were accused of
extending credit to unworthy borrowers, leading them to purchase unnecessary products
and generating debts beyond their means to repay.31 Department stores were criticized
for “actively goad[ing] people into contracting more debt.”32 In 1873 the New York
Times expressed concern that Americans were “Running in Debt” and by 1877 warned
that Americans were “Borrowing Trouble.”33 In 1873 a labor leader bemoaned the
improvidence of America’s consumers, “Has not the middle class its poverty? Very few
among them are saving money. Many of them are in debt; and all they can earn for years,
is, in many cases, mortgaged to pay such debt.”34
An 1899 report concluded that
installment selling “‘lured thousands to ruin’ encouraging people to buy what they could
not pay for and making debt ‘the curse of countless families.’”35
28
CAPLOVITZ, supra note, at 101.
CALDER, supra note 11, at 211.
30
CALDER, supra note 11, at 166.
31
See CALDER, supra note 11, at 182.
32
CALDER, supra note 11, at 217.
33
Quoted in DAVID S. EVANS AND RICHARD SCHMALENSEE, PAYING WITH PLASTIC: THE DIGITAL
REVOLUTION IN BUYING AND BORROWING 101 (2d ed. 2005).
34
CALDER, supra note 11, at 59 (quoting Ira Steward).
35
CALDER, supra note 11, at 213.
29
9
And not merely the poor and improvident were lured into ruin, but upstanding
middle class families as well, as they engaged in a heated rivalry of conspicuous
consumption with their neighbors.36 In 1949 Business Week asked, “Is the Country
Swamped with Debt?” and by 1959 U.S. News and World Report worried that “Never
Have So Many Owed So Much.” In 1940 Harper’s even feared that “Debt Threatens
Democracy.”37
The criticisms of mid-century installment credit mirrored those of credit cards
today: easy access to installment credit allegedly generated overconsumption,
overindebtedness, and finally bankruptcy. Credit customers bought more goods than
cash customers38 and retailers were criticized for enabling shoppers to buy more on credit
than they normally would on cash.39 Installment selling was considered a “menace” that
trapped Americans in “a morass of debt” and was the “first step toward national
bankruptcy,” a further overture to today’s criticisms of credit cards.40 “The youngest
families,” Caplovitz writes, “Are apt to turn to stores for installment credit; as their debts
mount and the payments become increasingly more difficult to meet, they are apt to
obtain loans to pay off their other debts. In this way they obtain relief in the form of
smaller monthly payments at the cost of a longer period of indebtedness.”41 Moreover,
although most Americans believed that installment selling was a “good idea” in general
CALDER, supra note 11, at 215. The term “conspicuous consumption” was coined over a century ago.
See THORSTEIN VEBLEN, THE THEORY OF THE LEISURE CLASS: AN ECONOMIC STUDY OF INSTITUTIONS
(1899). Veblen argues that one effect of conspicuous consumption is a tendency for households to reduce
savings and to rely on debt to live beyond their means.
37
Quoted in EVANS & SCHMALENSEE, supra note 33, at 101.
38
CALDER, supra note 11, at 200.
39
CALDER, supra note 11, at 220; compare MANN, supra note. Caplovitz notes that consumers tended to
pay a higher prices for televisions when purchasing on credit than on cash. CAPLOVITZ, supra note, at 88.
40
CALDER, supra note 11, at 221. The use of debt to purchase consumption goods such as food was
thought to be especially irresponsible. CALDER, supra note 11, at 225.
41
CAPLOVITZ, supra note, at 100.
36
10
and were confident in their own ability to use it responsibly, three out of four also
thought that their neighbors used installment credit excessively42—a judgment mirrored
in modern surveys of consumers about credit card use, in which most consumers assert
confidence in their own ability to use credit cards responsibly but express concern about
the ability of others to do the same.43 And as consumer bankruptcy filings rose during
the 1960s, some commentators and politicians pointed the finger of blame at profligate
installment lending.44 These criticisms of installment credit provide ironic reading today
in light of the modern claim that the ubiquity of credit cards—which have come to
displace installment credit for many consumer transactions—allegedly has produced a
psychology of consumer overconsumption.45
B.
Early Credit Cards
Credit cards were born in the early twentieth century with the introduction of the
first “charge plates” or “charga-plates,” the forerunners of the modern credit card.46
“Open book” credit accounts were long a staple of retailers, especially for the purchase of
nondurables such as groceries, and charge plates formalized those practices. In the postWorld War II era, prototype credit cards emerged, aimed at traveling businessmen, such
as gasoline, hotel, and airline cards.47
42
CALDER, supra note 11, at 235.
Thomas A. Durkin, Credit Cards: Use and Consumer Attitudes, 1970–2000, 86 FED. RES. BULL.
623,628-30 (2000).
44
See Wage Earner Plans Under the Bankruptcy Act: Hearing on H.R. 1057 and H.R. 5771 Before the
House Committee on the Judiciary, 90th Cong. (1967).
45
See, e.g., MANN, supra note 2, at 46; Richard L. Wiener, et al., Consumer Credit Card Use: The Roles of
Creditor Disclosure and Anticipated Emotion, 13 J. EXPERIMENTAL PSYCHOLOGY: APPLIED 32 (2007).
46
CALDER, supra note 11, at 72.
47
MANDELL, supra note 12, at xii. Greenspan reports that hotels began using credit cards as early as 1900
and gasoline cards and large retail department stores were issuing credit cards by 1914. Greenspan,
Greenlining Remarks, supra note 17.
43
11
The formation of Diners Club in 1949 laid the foundation for modern bank cards.
Diners Club was a universal card, in that it was a payment mechanism honored by many
merchants, unlike earlier cards tied to specific retailers of goods and services. It was also
a third-party card, in that Diners Club extended the credit to the customer and paid the
merchant, rather than the merchant issuing the credit directly. Diners Club thus was the
bearer of the risk of nonpayment rather than the restaurant. In return for this assured
payment and the convenience to customers of being able to use Diners Club at a large
number of locations, accepting merchants paid a 7% fee to Diners Club for each use.
But universal third-party cards were slow to take off. The consumer credit market
remained dominated by retail store credit cards through the 1970s.
This continued
domination by retail issuers was primarily because of the presence of usury restrictions
that restricted certain types of consumer lending during this period. Usury regulations
generally produce three types of unintended consequences. First, they encourage lenders
to “reprice” other terms of their credit contracts in order to try to offset the inability to
charge market rates of interest, such as by requiring larger downpayments, higher upfront
fixed fees or annual fees, shorter grace periods, or a myriad of other terms. Second,
usury regulations lead to product substitution, such as by debtors unable to acquire
unsecured credit card credit to switch to less-preferred types of credit, such as pawn
shops or payday lenders. Third, to the extent that it is impossible to fully reprice terms or
substitute to alternative lenders, the end result will be credit rationing, as some borrowers
may be unable to get any legal credit on any terms. All three phenomena appear to have
resulted from the usury regulations imposed in the 1970s.
12
Credit issuers repriced other terms to evade usury regulations, but retailers were
able to do this more effectively than banks. A rapid rise in underlying interest rates in the
1970s combined with usury caps made credit card operations for banks unprofitable
during this period, thus bank-type credit card operations remained modest in scope.48
Banks were able to avoid some of the restrictions by altering other terms of the
cardholder agreement or bundling lending with other services. Banks in states with strict
usury regulations restricted their hours of operation, reduced customer service, tied their
lending operations to other products and services not restricted in price (such as requiring
checking or savings accounts), or imposing higher service charges on demand deposit
accounts or checking account overdrafts.49 Most importantly, to evade usury regulations
credit card issuers imposed annual fees on credit cards, usually ranging from $30-$50.50
Credit card issuers adjusted other terms of the credit contract to compensate for the
inability to charge a market rate of interest, including adjusting grace periods and using
alternate methods for calculating interest charges.51 Credit card issuers also rationed
credit card privileges to only the most credit worthy consumers.52
48
Todd J. Zywicki, The Economics of Credit Cards, 3 CHAPMAN L. REV. 79, 151-55 (2000).
A. CHARLENE SULLIVAN, EVIDENCE OF THE EFFECT OF RESTRICTIVE LOAN RATE CEILINGS ON PRICES OF
FINANCIAL SERVICES (Credit Research Center Working Paper No 36, 1980); see also RICHARD PETERSON
& GREGORY A. FALLS, IMPACT OF A TEN PERCENT USURY CEILING: EMPIRICAL EVIDENCE 33 (Credit
Research Center Working Paper No. 40, 1981).
50
See Zywicki, Economics of Credit Cards, supra note 48, at 152. Because this fee was assessed on
revolvers and transactors alike, it effectively resulted in transactors subsidizing lower interest rates for
revolvers.
51
DAVID EVANS & RICHARD SCHMALENSEE, PAYING WITH PLASTIC: THE DIGITAL REVOLUTION IN BUYING
AND BORROWING 28 (1st ed. 2000); OLNEY, supra note 22, at 132 (“State usury laws were ineffective;
lenders managed to increase effective rates of interest through various fees, penalties, required insurance,
and so on.”).
52
Glenn B. Canner & James T. Fergus, The Economic Effects of Proposed Ceilings on Credit Card Interest
Rates, 73 FED. RES. BULL. 1, 2 (1987).
49
13
Consumers who were unable to get a general-purpose bankcard instead continued
to rely on retail store cards.53 Revolving credit was used primarily for non-durable
consumer purchases, such as clothing, groceries, and milk54 Credit-issuing department
stores had an even more effective way of evading usury restrictions—they could simply
bury the credit losses in the price of the goods they offered and sell the bundled
product.55 The main purpose of department store credit operations was ancillary to retail
operations, such as to build customer loyalty and making shopping more convenient.56
Retailers were willing to and usually did absorb losses in their credit operations, so long
as they furthered these larger goals.57 Cross-subsidization of retail operations by credit
operations by increasing the price of the goods to evade usury restrictions on credit was
an easy step. This provided large retailers with a substantial comparative advantage over
smaller retailers who could not afford to establish and maintain their own credit
operations.58
Although usury regulations thus capped interest rates, they did not prevent
consumers from having to pay high total prices for credit or dramatically reduce access to
credit. Instead, usury regulations transferred wealth from credit issuers who had less
ability to evade usury restrictions (such as small retailers) to issuers (like department
stores) who could evade usury restrictions more easily, such as those who could more
53
PETERSON & FALLS, supra note 49.
CAPLOVITZ, supra note, at 99.
55
See HOMER & SYLLA, supra note 9, at 428; see also CALDER, supra note 11, at 177 (noting practice of
door-to-door peddlers in early-twentieth century America who catered to immigrant families and marked
up the price of the goods sold to cover implied high interest rates). Similarly, pawn shops can simply
adjust the discount price of the goods that are pawned as collateral.
56
MANDELL, supra note 12, at xviii.
57
MANDELL, supra note 12, at xviii; Canner & Fergus, supra note 52, at 2 (“The studies indicated that on
average—not considering profits on associated merchandise sales—such credit card plans consistently
operated at a loss.”)
58
Christopher C. DeMuth, The Case Against Credit Card Interest Rate Regulation, 3 YALE J. ON REG. 201,
238 (1986).
54
14
effectively conceal the real cost of credit in the price of the goods and from higher-risk to
lower-risk borrowers.59 For instance, in those states with the strictest usury restrictions,
consumers also paid significantly higher prices for major appliances, almost all of which
were purchased with installment retail credit.60
Retailers in states with strict usury
regulations also reduced their services to consumers, such as charging for delivery and
gift wrapping or offering fewer choices in their stores.61
C.
Credit Cards Today
In 1978 the Supreme Court decided Marquette National Bank v. First of Omaha
Corp., which effectively deregulated interest rates on credit cards by holding that the
applicable interest rates for nationally-chartered banks would be the home state of the
issuing bank, rather than the consumer.62 The results have been dramatic. In 1970, only
16% of American households had a general-purpose bank-type card; today 71% do.
Ownership of retail store cards initially rose as well through the 1970s, increasing from
45% of households in 1971 to 58% in 1983. This growth reflects the comparative
advantage of retail stores during the high-interest rate 1970s in repricing credit terms to
evade usury caps. Ownership of retail store cards peaked in 1989 at 61% and has
declined steadily since then, reaching 44% by 2004. Ownership of specific use credit
59
See DeMuth, supra note 58, at 238; Daniel J. Villegas, The Impact of Usury Ceilings on Consumer
Credit, 56 S. ECON. J. 126, 140 (1989); William J. Boyes, In Defense of the Downtrodden: Usury Laws?,
39 PUBLIC CHOICE 269, 272 (1982).
60
Canner & Fergus, supra note 52, at 11.
61
Peterson & Falls, supra note 49, at 35 n.5.
62
439 U.S. 299 (1978).
15
cards, such as gasoline and travel and entertainment cards, have shown similar declines in
popularity over this period as they have been supplanted by general use bank cards.63
By effectively eliminating usury regulations, Marquette eliminated the incentives
to engage in term repricing. Beginning in the early 1990s, credit cards eliminated annual
fees on standard cards making pricing more efficient, more consumer-friendly, and
enabling consumers to hold multiple cards simultaneously which spurred heated
competition that has led to lower interest rates, the general elimination of annual fees, and
a proliferation of card benefits. In 1971, households were almost three times as likely to
own a retail store card as a bank-type card; today, that ratio is almost reversed and many
households hold multiple cards.64
The rapid growth in popularity of general purpose credit cards following
deregulation should not be surprising. Consumers historically had to choose among a
limited number of consumer credit alternatives: family and friends, high-interest payday
lenders or check cashers, pawnshops, personal finance companies, rent-to-own retailers,
or illegal loan sharks.65
All of these types of credit have obvious and significant
63
See Brian K. Bucks, Arthur B. Kennickell, and Kevin B. Moore, Recent Changes in U.S. Family
Finances: Evidence from the 2001 and 2004 Survey of Consumer Finances, FED. RES. BULLETIN at p. A31
(2006) (noting that in 2004 95% of credit card holders have bank-type credit cards, while the percentage of
those holding store, gasoline, or travel and entertainment cards have fallen); Kenneth A. Carow & Michael
E. Staten, Plastic Choices: Consumer Usage of Bank Cards Versus Proprietary Credit Cards, 26(2) J.
ECON. & FIN. 216 (Summer 2002); Kenneth A. Carow & Michael E. Staten, Debit, Credit, or Cash: Survey
Evidence on Gasoline Purchases, 51 J. ECON. AND BUS. 409 (1999); see also Ana M. Aizcorbe et al.,
Recent Changes in U.S. Family Finances: Evidence from the 1998 and 2001 Survey of Consumer
Finances, 89 FED. RES. BULL. 1, 25 (2003) (noting 5.2% increase in percentage of families with bank cards,
and 4.8% reduction in percentage with store cards and 3.1% decrease in percentage of households with
gasoline cards over three year period).
64
See Durkin, Credit Cards, supra note 43, at 624. The average American household holds over four credit
cards, which includes store and other cards as well as bank-type cards. See FED. RES. BD., THE
PROFITABILITY OF CREDIT CARD OPERATIONS OF DEPOSITORY INSTITUTIONS 5 (June 1999).
65
Empirical evidence suggests that stricter regulation of consumer credit (and thus reduced access to legal
sources of consumer credit) is correlated with higher levels of illegal borrowing. See POLICIS, THE EFFECT
OF INTEREST RATE CONTROLS IN OTHER COUNTRIES (2004); POLICIS, ECONOMIC AND SOCIAL RISKS OF
CONSUMER CREDIT MARKET REGULATION (2006).
16
drawbacks associated with them when compared to credit cards.
Examining the
alternatives available to consumers indicates why consumers have so readily substituted
toward credit cards.
The primary source of short-term credit for lower-income Americans through
history has been friends and family.66
A recent survey of households in low and
moderate-income areas of Los Angeles, Chicago, and Washington found that 53 percent
of respondents said they would rely on friends or family to borrow $500 for three
months.67 Angela Littwin’s survey of credit use by low-income women found that 93%
had actually borrowed money from friends and family in the past and many had lent
money to friends and family as well.68 Ten percent of her subjects have borrowed only
from friends and family.69 But friends and family may not be able or willing to lend
when needed or in the amounts needed—a reality reinforced by the fact that most social
networks are limited in scope, and so most of the friends and family of low-income
individuals are also low-income and thus have limited funds to lend.
Payday lenders charge annual interest rates as high as 300 percent and generally
require that the borrower have a bank account, which many low-income households do
not.70 Payday loan customers are low to middle-income, relatively young, and less likely
66
CALDER, supra note 11, at 60-64.
ELLEN SIEDMAN, MOEZ HABABOU, AND JENNIFER KRAMER, A FINANCIAL SERVICES SURVEY OF LOWAND MODERATE-INCOME HOUSEHOLDS (2005).
68
Angela Littwin, Beyond Usury: A Study of Credit Card Use and Preference Among Low-Income
Consumers, at 8.
69
Littwin, Comparing Credit Cards, supra note, at 36.
70
LENDOL CALDER, FINANCING THE AMERICAN DREAM; JOHN P. CASKEY, FRINGE BANKING: CHECKCASHING OUTLETS, PAWNSHOPS, AND THE POOR 37-67 (1994); Gregory Elliehausen, Consumers’ Use of
High-Price Credit Products: Do They Know What They are Doing? Networks Financial Institute at Indiana
State University Working Paper 2006-WP-02 (May 2006); see also Ronald J. Mann & Jim Hawkins, Just
Until Payday (working paper, Aug. 2, 2006). It should be noted that even this high rate of interest on a
payday loan may be lower than the implied APR for the fees imposed for a bounced check, which may be
the alternative to a payday advance loan. Se Michael W. Lynch, Legal Loan Sharking or Essential
67
17
to own a home than the average American family.71 Payday loan customers generally are
aware of the prices charged on payday loans, but use payday lenders either to avoid more
expensive problems, such as bounced check fees (which can amount to $50 or more when
the fees of the merchant and bank are combined) or because of an absence of attractive
alternatives, such as credit cards.72 For instance, over half of payday loan customers who
have bank cards had refrained from using them in the past year because they would have
exceeded their credit limit.73 The behavior of payday loan customers suggests that they
are concerned about access to credit and fear being rejected by other lenders.74 Those
who use payday lenders report a relatively high level of satisfaction with the service.75
Despite the seemingly high cost of payday loans there is no indication that payday
lenders are earning economic profits off their activities once risk is taken into account.76
In fact, increased competition among payday lenders in a given market tends to reduce
Service? The Great “Payday Loan” Controversy, REASON (2002); Michael S. Barr, Banking the Poor, 21
YALE J. ON REG. 121, 155 (2004).
71
Barr, supra note 70, at 153.
72
See Donald Morgan and M. Strain, Payday Holiday: How Households Fare after Payday Credit Bans,
FED. RES. BANK OF N.Y., Staff Report no. 309 (Nov. 2007); Gregory Elliehausen & Edward C. Lawrence,
Payday Advance Credit in America: An Analysis of Customer Demand, GEORGETOWN UNIVERSITY,
MCDONOUGH SCHOOL OF BUSINESS, CREDIT RESEARCH CENTER (April 2001).
73
Elliehausen & Lawrence, supra note 72; see also Angela Littwin, Comparing Credit Cards: An
Empirical Examination of Borrowing Preferences Among Low-Income Consumers (working paper);
available in http://ssrn.com/abstract=1014460 (citing Michael S. Barr, et al., Consumer Indebtedness in the
Alternative Financial Services Market U. Mich. Law, Working Paper, April 2007) (noting that payday loan
users use credit cards at same rate as others but were more likely to pay only minimum balance on credit
cards and to have paid late fees).
74
See Paige Marta Skiba & Jeremy Tobacman, Measuring the Individual-Level Effects of Access to Credit:
Evidence from Payday Loans, Working Paper (July 3, 2007). Half of pawnshop borrowers had been
rejected for credit in the past year, 73% of payday advance customers and 46.5 percent of refund
anticipation loan customers were turned down or limited in the past five years. Almost half of payday loan
customers and three-fourths of refund loan customers said that during the past year they thought about
applying for credit but did not because they thought that they would be turned down. Elliehausen, supra
note 70, at 21.
75
Elliehausen & Lawrence, supra note 72.
76
See Paige Skiba & Jeremy Tobacman, The Profitability of Payday Loans, Working Paper (Dec. 7, 2006).
18
the price of payday loans77 and payday loans appear can serve an important role in
improving consumer welfare and quality of life78.
Pawnbrokers require parting with personal items, a hardship in itself, exacerbated
by the fact that those goods often possess little economic value, thus the value received is
less than the replacement cost of the goods.79 Pawn loans are also expensive, with
interest rates of over 200 percent per year.80 Default rates are high.81 In addition,
because the value of the loan is limited by the value of the personal goods pawned, pawn
loans tend to be quite small ($76 on average).82 Jewelry and consumer electronics are the
goods most frequently pawned.83 Those who borrow from pawnbrokers do so because
they generally do not own homes (thus home equity loans are not available) and have
limited alternative sources of borrowing, which primarily include friends or checkcashers.84 In particular, pawn shop borrowers typically have been turned down for a
77
Donald P. Morgan, Defining and Detecting Predatory Lending, FEDERAL BANK OF NEW YORK STAFF
REPORTS, Staff Report No. 273 (January 2007); see also Philip Bond, David K. Musto, & Bilge Yilmaz,
Predatory Lending in a Rational World, Federal Reserve Bank of Philadelphia Working Paper 06-2 (Nov.
2006).
78
Adair Morse, Payday Lenders: Heroes or Villains?, Working Paper (Nov. 20, 2006). Morse finds that
access to payday lending improves welfare across numerous quality-of-life margins, including health
outcomes, reduced crime, and other variables. See also Dean Karlan & Jonathan Zinman, Expanding
Credit Access: Using Randomized Supply Decisions to Estimate the Impacts (working paper, June 25,
2007) (finding in study of South African consumers that access to high-cost credit provided significant
benefits for borrowers across a wide range of economic and “well-being” outcomes, such as higher-quality
diet, improved physical well-being, and more stable employment).
79
CALDER, supra note 11.
80
JOHN P. CASKEY, FRINGE BANKING: CHECK-CASHING OUTLETS, PAWN SHOPS AND THE POOR 36 (1994).
Skiba and Tobacman find that pawn loans have a ninety-day term, with a monthly interest rate of 20
percent on loans from $1-$150 and 15 percent on loans above $150. Skiba & Tobacman, Measuring, supra
note 74.
81
Skiba & Tobacman, Measuring, supra note.
82
Skiba & Tobacman, Measuring, supra note; see also Robert W. Johnson & Dixie P. Johnson,
Pawnbroking in the U.S.: A Profile of Customers 16 (Credit Research Ctr., Monograph No. 34, 1998)
(average loan of $70 with typical range from $35-$260); CASKEY, supra note, at 44 (finding average
pawnshop loan of $50-$70). The average payday loan, by contrast, is $300. Skiba & Tobacman,
Measuring, supra note.
83
CASKEY, supra note, at 37.
84
See PETERSON & FALLS, supra note 49; see also Robert W. Johnson & Dixie P. Johnson, Pawnbroking in
the U.S.: A Profile of Customers 47 (Credit Research Ctr., Monograph No. 34, 1998) (finding that 65.4%
of Americans own their homes, but only 34.6% of those who borrow from pawnshops do so). This absence
19
payday loan, making pawnbrokers a lender of last resort for many. 85 Today, the basic
pawn shop model has been expanded to include automobile title lenders, which operate
similarly to traditional pawn shops.86
Some department stores and retailers of goods such as furniture, appliances,
hardware, lumber, and jewelry, continue to be a source of credit for the purchase of their
goods. These transactions often have high imputed interest rate (often reflected in a
higher price for the goods purchased than would otherwise be the case) and, typically, the
consumer must give a purchase-money security interest in the goods purchased,
permitting their repossession in the event of default. A modern and highly-popular form
of retail credit among low-income consumers is mail-order catalogue shopping, most
notably Fingerhut, which sells a wide variety of household goods on credit.87 Fingerhut’s
prices appear to be much higher than the price for similar goods at traditional retailers.88
Reliance on retail store credit limits consumers’ shopping choices by limiting them to
those sellers who will also offer them credit, restricting choice and price competition for
goods and services.89 Historically, a popular variation for those unable to get store credit
was layaway.90 General bank-type credit cards enable the retail transaction be unhooked
from the credit transaction, thereby facilitating competition in both markets and
of home ownership has a secondary consequence, in that it leads renters to change addresses more often,
which counts as a negative factor in some credit scoring models. Id.
85
Skiba & Tobacman, Measuring, supra note 74. Morgan quotes a CEO of a major pawnshop chain who
states that the rise of payday lending has hurt his company. Morgan, Defining & Detecting, supra note, at
5-6.
86
See Barr, supra note 70, at 164-66.
87
See Littwin, Comparing Credit Cards, supra note, at 41-42. Fingerhut’s customer base has an average
household income below $30,000 and sells almost exclusively on credit, especially to those who have
difficulty obtaining credit elsewhere. Id.
88
Littwin, Comparing Credit Cards, supra note, at 56.
89
Consider, for instance, the ubiquitous used car dealerships that offer credit to those with “bad credit”: the
purchase price of the car usually imputes a very high interest rate into the sales price of the car.
90
Under layaway, the consumer did not obtain possession of the item until it was fully paid for. Layaway
has been largely discontinued today.
20
permitting a better match of consumer preferences, such as with on-line or catalogue
shopping.
Rent-to-own is another alternative.91 Instead of structuring a transaction as a
secured transaction, rent-to-own transactions are leases with an option to buy at the end.
The interest rate is imputed into the lease payment and the overall prices of the goods is
typically 2 to 2.5 times what one would pay in a retail store.92 Styled as a lease rather
than a secured transaction, foreclosure is easier upon default. In fact, rent-to own may be
the true “lender” of last resort (leaving aside illegal loan sharks), as the idea was
originally introduced by a retail appliance store owner for those customers who were
denied credit for appliance purchases.93 This is still the case today, as rent-to-own
arrangements continue to be used primarily by low-income consumers, who are unable to
get other forms of credit at comparably better terms (if at all).94 Rent-to-own customers
are also lower-income, younger, and less likely to be homeowners than others.95
91
See Jim Hawkins, Renting the Good Life, University of Texas School of Law, Law and Economics
Research Paper No. 111. A survey of rent-to-own customers found that that 67% of rent-to-own customers
intended to purchase the merchandise when they began a rent-to-own transaction and 87 percent of those
intending to purchase actually did, suggesting that many consumers use rent-to-own as an alternative to
secured sale. FEDERAL TRADE COMMISSION BUREAU OF ECONOMICS STAFF REPORT, SURVEY OF RENT-TOOWN CUSTOMERS (2000). Seventy-five percent of rent-to-own customers reported being satisfied with
their experience with rent-to-own transactions. Id.
92
Signe-Mary McKernan et al., Empirical Evidence on the Determinants of Rent-to-Own Use and
Purchase Behavior, 17 ECON. DEV. Q. 33, 34 (2003).
93
See Susan Lorde Martin & Nancy White Huckins, Consumer Advocates v. The Rent-to-Own Industry:
Reaching a Reasonable Accommodation, 34 AM. BUS. L.J. 385 (1997). In fact, rent-to-own appears to have
come into being in response to legal bans on cross-collateral clauses for consumer goods, such as in
Williams v. Walker-Thomas.
94
See McKernan et al., supra note 92, at 51; James P. Nehf, Effective Regulation of Rent-to-Own
Contracts, 52 OHIO ST. L.J. 751, 752 (1991); Eligio Pimentel, Renting-To-Own: Exploitation or Market
Efficiency?, 13 LAW & INEQ. J. 369, 394 (1995) (“Consumers who enter in [rent-to-own] transactions have
usually been denied credit by other businesses. They typically resort to the [rent-to-own] arrangement in a
final effort to obtain the merchandise they desire.”).
95
FTC RENT-TO-OWN STUDY, supra note 91. Forty-four percent of rent-to-own customers have credit
cards, but it is not clear how many of them would have exceeded their credit limits had they tried to use
them instead of rent-to-own.
21
The most attractive type of general consumer credit, especially for middle class
borrowers, was an unsecured loan from a personal finance company or local bank.96
Personal loans historically charged interest rates marginally lower than those on credit
cards, but since the early 1990s the interest rates on personal loans have been comparable
to credit cards (both are unsecured loans and therefore the fundamental risk is similar).97
Personal loans also charge higher initiation and underwriting fees than credit cards,
which typically charge no up-front fees.98 The cost of underwriting a personal loan also
placed a practical limitation on the minimum size and duration of a loan to a few
thousand dollars and a multi-year repayment term, even if the consumer only needed a
few hundred dollars for a short period of time, such as to finance a car repair or for a
student to buy books until her student loan check arrives. Eventually, commercial banks
entered the industry as a competing source of consumer credit, especially for installment
loans for the purchase of major consumer durables, such as automobiles.99 In addition to
cost, consumers value the flexible borrowing and repayment terms of credit cards.100
Federal Reserve Board Economist Thomas Durkin has observed that credit cards
have largely replaced the installment-purchase plans that were important to the sales
See CALDER, supra note 11, at 148. Many loans by “industrial lenders” and personal finance companies
were consolidation of outstanding installment loans. Id. at 151.
97
See infra at Figure 9.
98
Credit cards typically have no origination fee and no minimum loan size while personal finance company
loans have substantial up-front original costs. Brito and Hartley reported, “A senior bank officer told us
that the costs to the bank of processing a loan are so high that they cannot afford to make a loan of less than
$3,000 for one year except at interest rates above those charged on credit cards.” They also note, “inquiries
in Houston in February 1992 revealed rates ranging from 17 percent and a $100 fixed fee for a
collateralized 1-year loan at a branch of a major national finance company to over 50 percent for small
loans ($300 maximum) at a local finance company.” 98 Bank loans of similar size and duration “either do
not exist or are available only at terms more onerous than those offered by credit card issuers.” Dagobert
L. Brito & Peter R. Hartley, Consumer Rationality and Credit Cards, 103 J. POL. ECON. 400, 402 (1995).
Applying for a credit card is also easier and more convenient than applying for a personal loan.
99
See OLNEY, supra note 22, at 108.
100
Thomas A. Durkin, Consumers and Credit Disclosures: Credit Cards and Credit Insurance, FED. RES.
BULLETIN 201, 205 (April 2002) (73% of consumer survey respondents state that flexibility of expenditure
and repayment with credit cards makes managing finances easier).
96
22
volume at many retail stores in earlier decades, especially for the purchase of appliances,
furniture, and other durable goods.101 Former Federal Reserve Chairman Alan Greenspan
similarly observed, “[T]he rise in credit card debt in the latter half of the 1990s is
mirrored by a fall in unsecured personal loans.”102 In fact, consumer debt for nonautomobile durable goods, home improvement loans, and “other” nonrevolving credit
(such as unsecured personal loans) declined from a total of about 20 percent of total
household consumer debt in 1977 to 10 percent in 2004.103 Thus, although personal loans
are not as comparatively unattractive as other traditional credit alternatives, they are still
higher cost and less convenient for both consumers and lenders, thus it is not surprising
that credit cards have generally supplanted personal loans.104
Pawn shops, layaway plans, payday lenders, check cashers, personal finance
companies, retail store credit, rent-to-own, loan sharks, friends and family—all have
served as important sources of consumer credit in American history.105 Credit cards are
plainly superior to all of these traditional types of consumer credit. And in fact, research
has demonstrated that those who use these high-priced and inconvenient lending products
do so because they are unable to get a credit card at all or have reached their credit limit
101
Durkin, supra note 43, at 623-24; see also FED REPORT, supra note 6, at 5 (June 2006) (hereinafter FED
REPORTS) (noting that revolving consumer credit has “partly replaced certain types of closed-end
installment credit, principally those types classified as non-automobile durable goods credit, home
improvement loans, and ‘other.’”).
102
Alan Greenspan, Understanding Household Debt Obligations, Remarks Given at the Credit Union
National Association 2004 Governmental Affairs Conference (Feb. 23, 2004), available at
http://www.federalreserve. gov/boarddocs/speeches/2004/20040223/default.htm.
103
FED REPORT, supra note 6, at 5 and Table 4.
104
Arthur B. Kennickell et al., Family Finances in the U.S.: Recent Evidence from the Survey of Consumer
Finances, 83 FED. RES. BULL. 17 (1997) (noting that many lenders have stopped offering general unsecured
lines of credit).
105
Credit cards are also used more frequently in high-crime geographic areas, as credit cards are less
vulnerable to risk of theft than cash. See David B. Humphrey, Lawrence B. Pulley, and Jukka M. Vesala,
Cash, Paper, and Electronic Payments: A Cross-Country Analysis, 28(4) J. MONEY, CREDIT AND BANKING
914, 934 (1996).
23
maximum.106 For instance, although 68% of American households have a bank-type card
and 73% have any credit card, less than half of pawnbroker, rent-to-own, and refund
anticipation loan customers, and only slightly over half of payday advance customers
have credit cards.107 Moreover, 61% of payday loan customers who have a bank card and
a third of tax refund anticipation loan customers with a bank card reported that they had
refrained from using their credit card because they would have exceeded their credit
limit.108 It thus appears that consumers use high-cost lenders only because they are
unable to get access to credit cards and other attractive loan products.109 Plainly, there
has been a massive substitution effect of the introduction of credit cards into the
American economy, leaving only the question of how large this substitution effect has
been as opposed to causing an increase in household indebtedness.
II.
Models of the Relationship Between Credit Cards and Bankruptcy
Several scholars have found a correlation between credit card debt and
bankruptcy.
Domowitz and Sartain, for instance, conclude that “the largest single
contribution to bankruptcy at the margin is credit card debt.”110 Ausubel also finds a
106
Elliehausen, supra note 70, at 19. Angela Littwin similarly finds that low-income consumers would like
to have credit cards, especially to deal with emergencies, but are often unable to get them. See Littwin,
supra note 68, at 5-6.
107
Elliehausen, supra note 70, at 19-20.
108
Id. at 21.
109
For instance, as younger households mature and gain greater credit experience they begin to gain access
to credit cards, at which time they substitute away from these other high-price loans (such as payday loans
and refund anticipation loans) to using credit cards for borrowing small amounts for short periods of time.
Elliehausen finds that low-income consumers who revolve debt on bank cards tend to be older than those
who use payday loans and refund anticipation loans, suggesting that many younger households are unable
to get credit cards and that when they finally do, they prefer them to those sources of credit that they
previously used. Elliehausen, supra note 70, at 21.
110
Ian Domowitz & Robert L. Sartain, Determinants of the Consumer Bankruptcy Decision, 54 J. FIN. 403,
414 (1999).
24
correlation between credit card defaults and bankruptcy.111
Providing a causal
explanation, however, has proven elusive.
Three different models have been postulated to explain the observed correlation
between credit cards and bankruptcy: (1) The “substitution” model, (2) the “distress”
model, and (3) the “strategic” model. The substitution model predicts that bankruptcy
filings could rise as a result of credit cards even if the overall consumer debt burden is
held constant. The distress model postulates that increased use of credit cards will raise
the consumer debt burden and thereby increase consumer bankruptcies. The strategic
model predicts that credit card debt will rise in anticipation of bankruptcy filing, although
it is ambiguous about the overall effect on consumer indebtedness. Consider each model
and its implications.
A.
Substitution Model
The substitution model begins with the observed correlation between the
increased use of credit cards and increased bankruptcy filings, but hypothesizes that this
may be explainable simply by the substitution of increased credit card debt for other
traditional types of consumer credit, such that borrowers are more willing or able to file
bankruptcy and discharge their debt—even holding total consumer debt constant. Even if
overall debt burdens do not rise as the result of credit cards, credit card debt is generally
dischargeable unsecured debt, unlike traditional types of credit which were either
practicably nondischargeable (such as loans from family members or illegal loan sharks)
or secured (such as a loan to finance the purchase of furniture or an appliance).
Consumers also may be more willing to discharge debts owed to national institutional
111
See Lawrence M. Ausubel, Credit Card Defaults, Credit Card Profits, and Bankruptcy, 71 AM. BANKR.
L.J. 249 (1997).
25
lenders, such as Citibank, rather than local businesses. Thus, even if credit cards are
simply a substitute for these other types of consumer credit bankruptcy filings may rise if
the interaction between the bankruptcy laws and developments in consumer credit result
in increased incentives to discharge credit card debt than these other types of credit.
Increased credit card use may be correlated with increased bankruptcy filings, but in
contrast to the distress model, the increased credit card debt does not cause the increase
in bankruptcy filings through the intermediate step of increased debt burdens, but that the
substitution increases the propensity to file, holding debt burdens constant. In contrast to
the strategic model, the increase in consumer credit card debt may be largely exogenously
driven and thus not necessarily a strategic “loading up” decision.
The substitution model implies that the correlation between credit card debt and
bankruptcy filings is a two step decision-making process. First, the debtor may increase
her use of credit card debt for reasons unrelated to a later bankruptcy filing, such as its
comparatively lower interest rate (relative to other available options) or greater
convenience, but not increase her overall debt burden, if the credit card debt is merely a
substitute for other types of debt. Then second, when the debtor is confronted with the
option of filing bankruptcy, the debtor may find bankruptcy to be a more attractive choice
than would be the case in the past. In this model, the mere substitution of credit card debt
for other kinds of consumer debt could result in an increased
propensity to file
bankruptcy, holding overall consumer debt levels constant and thus even if it does not
increase the need to file bankruptcy:
Credit Card Substitution Increased Benefit to Bankruptcy Increased Bankruptcy
26
B.
Distress Model
The “distress model” (or “overindebtedness model”) argues that widespread
access to credit cards has led to an increase in consumer indebtedness and financial
distress and through this to increased consumer bankruptcy filings. Credit cards, it is
argued, lead to a greater rate of overindebtedness compared to earlier ages because of
profligate lending to riskier borrowers or irrational borrowing by consumers in general.
The distress model predicts a causal relationship between increased credit card borrowing
and increased bankruptcy filings through the intermediate step of increased debt. In the
distress model, in contrast to the substitution model, an increase in credit card debt leads
to an increase in bankruptcy filings indirectly, by first increasing overall consumer
indebtedness and thereby subsequently increasing the need for consumers to file
bankruptcy. So the chain of causation may be described as follow:
Credit Card Debt  Increased Debt Increased Distress Increased Bankruptcies
C.
Strategic Model
The “strategic” model essentially reverses the chain of causation from the distress
model. In the strategic model, credit card debt rises in anticipation of bankruptcy.112
Because of information asymmetries, consumers are more aware of their likelihood of
filing bankruptcy than are lenders. Credit cards are uniquely prone to problems of moral
hazard and adverse selection, as once granted they provide cardholders with a line of
credit to draw upon at their discretion with no further approval from the lender. Unlike
many other types of loans, credit cards can be used to finance services and consumption
112
David B. Gross & Nicholas S. Souleles, An Empirical Analysis of Personal Bankruptcy and
Delinquency, 15 REV. FIN. STUD. 319, 324 (2002).
27
and is generally unsecured debt that can be fully discharged in bankruptcy without
surrendering any property. The strategic model draws on these problems of adverse
selection and moral hazard to posit that the correlation between bankruptcy and credit
cards results from the debtor strategically increasing her credit card borrowing in the
prebankruptcy period by “loading up” on credit card debt in order to purchase goods and
services in anticipation of later discharging those debts in bankruptcy. Like the distress
model, a rise in credit card debt temporally precedes a bankruptcy filing and may be
reflected in an increase in total debt; unlike the distress model, however, the rise in credit
card debt cannot be said to have “caused” the bankruptcy filing. Instead, it is because the
debtor has the option to file bankruptcy and discharge the debt effectively causes the rise
in credit card debt, as the debtor strategically loads up his credit cards in anticipation of
discharging those debts:
Expected Bankruptcy Filing Increased Credit Card Debt Bankruptcy
D.
Predictions
Distinguishing among these various hypotheses for purposes of testing can be
difficult, as each predicts that an increase in bankruptcy filings will be preceded by an
increase in credit card debt, with only the direction of causation in question.
Nonetheless, certain data findings would be inconsistent with the predictions of one or
another model. For instance, for the distress model to be correct there must be both an
increase in consumer debt burdens that is caused by a rise in credit card debt, and in turn,
this rise in consumer debt burden must cause a rise in bankruptcy filings.
This
fundamental prediction of a rise in credit card-induced consumer debt burdens is an
28
essential condition of the distress model; thus, a failure to find it would be fatal to the
model.
The substitution model predicts a rise in credit card debt burden over time, but a
relatively constant overall consumer debt burden, as the rise in credit card debt is largely
offset by a decline in other types of consumer debt. The strategic model would be
consistent with either prediction regarding overall consumer debt burdens (rising or
constant) but may be distinguished primarily by the timing of rising consumer debt
burdens. Under the strategic model, one would expect to observe a rapid rise in credit
card debt in the period immediately preceding bankruptcy evidencing a conscious choice
to pay some debts rather than others, or to increase credit card borrowing. The other
models predict no such obvious timing pattern but tend to suggest that credit card debt
accumulation will be more gradual. Both the substitution and strategic models also
predict that defaults on credit cards will not be the same as defaults on secured consumer
loans, such as home mortgages where default will result in foreclosure.
The remainder of this article examines each of these three hypotheses in turn.
Although the distress model of credit card use is widely-accepted in the law review
literature it finds few adherents among economists.113 An examination of the available
data fails to support the predictions of the distress model. Available evidence points
toward the validity of the substitution model as the most plausible model of the
relationship between credit cards and bankruptcy with some marginal effect of the
strategic theory. Contrary to conventional wisdom, credit cards do not appear to have
caused an increase in total household financial distress because although consumers have
increased their use of credit cards as a borrowing medium, this has been primarily a
113
See, e.g., FED REPORT, supra note 6.
29
substitution of credit card debt for other high-interest consumer debt. The increased use
of credit cards thus has changed the composition of household debt, but has not increased
its overall burden.
This substitution of credit cards for traditional types of consumer credit, however,
has had the unintended consequence of making bankruptcy more attractive to consumers
as a response to financial distress and thereby to increase bankruptcy filing rates. Thus,
although credit cards have not spurred an increased need for bankruptcy (indeed, they
may have done the opposite, by making a given level of indebtedness easier to manage
because of their lower cost and flexible repayment terms), they have increased the
incentive for indebted consumers to elect bankruptcy as a response to their financial
difficulties, thereby increasing the propensity to file bankruptcy.
III.
Substitution Model
The first hypothesis regarding the relationship between credit cards and
bankruptcy is the substitution model. The substitution model rests on the premise that a
mere substitution of credit card debt for other forms of consumer debt may lead to
increased bankruptcy filings even holding total consumer debt burden constant, if that
substitution increases the incentives or willingness of consumers to file bankruptcy. In
particular, the substitution model posits that the relationship between credit cards and
bankruptcy as a two step process. First, credit card use may rise for reasons exogenous to
the question of bankruptcy, such as the greater convenience of credit cards or their lower
cost relative to alternatives, and the growth in credit card debt may simply substitute for
other types of debt (such as retail store credit) or informal loans, changing the
30
composition of the household debt portfolio but not increasing financial distress. But
second, this substitution could cause an increased propensity to file bankruptcy in
response to financial distress.
A.
Have Credit Cards Increased Household Debt Burdens?
It is obvious that there has been a substantial substitution of credit cards for other
types of consumer debt. Yet, although credit card debt seems ubiquitous, revolving card
debt is only about seven percent of total household indebtedness. 114
One-half to two-
thirds of American households do not revolve credit card debt or rarely revolve credit
card debt from one month to another.115 As a result, the median credit card debt burden
for American households is zero.116 Of households that do revolve credit card debt, the
median balance is about $2,100, a fairly modest figure when compared to mortgage and
automobile debt. Of course, there are some households with a lot of credit card debt, but
as a first approximation they are the exception.
The best available measure of the household debt burden is the debt-service ratio.
The typical measurement of consumer debt is the debt-to-income ratio, which computes
the ratio of household income to the total amount of consumer debt outstanding. But this
114
See Federal Reserve Release G.19 and Federal Reserve Flow of Funds Accounts of the United States.
Indebtedness on home mortgages constitutes about 83 percent of outstanding household indebtedness.
Installment credit, such as auto loans and student loans, constitute the remainder.
115
See Zywicki, Economics of Credit Cards, supra note 48, at 101; Aizcorbe et al., supra note 63, at 25
(reporting that 55.3% of households pay their credit card bills in full each month); Joanna Stavins, Credit
Card Borrowing, Delinquency, and Personal Bankruptcy, NEW ENG. ECON. REV., July/Aug. 2000, at 15, 20
(58% of households in Survey of Consumer Finances stated that they paid their credit cards in full each
month in the past year); see also Thomas F. Cargill & Jeanne Wendel, Bank Credit Cards: Consumer
Irrationality Versus Market Forces, 30 J. CONSUMER AFF. 373, 379 (1996) (68% of credit card users
“nearly always” pay their full balance every month); FEDERAL RESERVE, REPORT TO CONGRESS, supra
note, at 6 (noting that 56 percent of households nearly always pay their balance in full each month).
116
Anne Kim, Adam Solomon, Bernard L. Schwartz, Jim Kessler, and Stephen Rose, The New Rules
Economy: A Policy Framework for the 21st Century, THE THIRD WAY MIDDLE CLASS PROJECT 9 (Feb.
2007).
31
measurement is misleading because it fails to account for changes in interest rates and
loan maturity rates, which are obviously relevant to one’s ability to pay debt, as well as
the amount that will be borrowed.117 The debt-service ratio measures the amount the
debtor must pay on his debt each month out of his current income and thus does account
for these factors that are ignored by the debt-to-income ratio.118 Households change their
borrowing behavior in response to changes in interest rates—thus a one percentage-point
rise in credit card interest rates is estimated to induce a $70 reduction in total household
credit card debt each month.119
Higher interest rates also lead consumers to more
aggressively shift balances among cards in order to minimize interest payments.120
Historically low interest rates during the past several years have enabled consumers to
borrow more (by taking on great principle obligations), but without increasing their debt
service burden.
As a result of competition, lower interest rates, and the substitution effect of
shopping for lower interest rates, the household debt service ratio has been largely
consistent over time. Disaggregating the debt service ratio into its constituent elements
117
For an extended discussion of the problems with the traditional measures, see Todd J. Zywicki, An
Economic Analysis of the Consumer Bankruptcy Crisis, 99 NORTHWESTERN L. REV. 1463 (2005).
118
The household debt service ratio is the ratio of household debt payments to disposable income and is
analogous to “equity insolvency” in corporate law. For revolving debt, the assumed required minimum
payment is 2-1/2 percent of the balance each month. See Household Debt Service and Financial
Obligations Ratios, About the Release, http://zentralbank.us.releases.housedebt/about.htm. Most consumer
installment loans require payment of 2-3 percent of the balance each month, thus at recent interest rates, the
minimum required payment on a credit card loan would be 13 percent less than the payment on a personal
loan of the same size. Thus, even though the interest rates are similar, the credit card loan has a payment
equivalent to a personal loan with a maturity almost one year longer than that of the typical personal loan.
See Kathleen Johnson, Recent Developments in the Credit Card Market and the Financial Obligations
Ratio, FED. RES. BULLETIN 473, 482 (Autumn 2005). This longer implied maturity is irrelevant to the debtservice ratio, of course, although it would be relevant to a calculation of household net worth. Thus it is
significant that there is no evidence that household net worth (which is equivalent to “bankruptcy” or
“balance sheet” insolvency) has worsened during this period either. See discussion infra at note 129 and
accompanying text.
119
David B. Gross and Nicholas S. Souleles, Do Liquidity Constraints and Interest Rates Matter for
Consumer Behavior? Evidence from Credit Card Data, 117 Q. J. ECON. 148 (Feb. 2002).
120
Id.
32
illustrates the substitution of revolving credit card debt for nonrevolving installment debt
over time, as shown in Figure 1:
Figure 1
Debt Service Ratio by Type of Debt: 1980-2007
0.16
0.14
Proportion of Income
Total DSR
0.12
Mortgage
DSR
0.1
0.08
NonRevolving
DSR
Revolving
DSR
0.06
0.04
0.02
20
07
q1
20
04
q1
20
01
q1
19
98
q1
19
95
q1
19
92
q1
19
89
q1
19
86
q1
19
83
q1
19
80
q1
0
Year
Source: Federal Reserve
As Figure 1 indicates, even though the bankruptcy rate surged from 1980 to 2000, the
debt service ratio was largely constant, before beginning a slight rise since 2000. In fact,
the Total household DSR—even including the mortgage debt service ratio—was actually
lower in 1993 than in 1980—even though the bankruptcy filing rate doubled during that
period. Total DSR was lower in 1998 than in 1987, but the bankruptcy rate almost tripled
during that period. Since 2000, there has been an upward movement in the debt-service
ratio—caused by an increase in mortgage debt—but this upward trend has occurred just
as the bankruptcy filing rate was leveling off after a period of rapid increase.
33
Moreover, it is evident that this rise in the debt service ratio in the past few years
has been in the mortgage debt service ratio, not in consumer credit, and not revolving
debt, for which the debt service ratio has been constant since the mid-1990s.121 This rise
in the mortgage debt service ratio reflects the dramatic expansion of home ownership
during this period as a result of the growth of the subprime lending market, as home
ownership rose from 64% to 69% of families during this period, reaching an all-time
high.122 According to Federal Reserve economists, “[F]ully 100 percent of the increase
in aggregate debt relative to income since 1983 has taken the form of debt on
households’ primary residences…. Meanwhile, aggregate debt associated with credit
cards, consumer installment loans, and other borrowing stayed just below 0.30 of
aggregate household income throughout the last quarter-century.”123 Most of these new
home buyers were marginal purchasers with lower-incomes and higher debt service ratios
that tended to raise the average debt service ratio. During this period, downpayments fell
as well, increasing the amount financed and housing prices generally.124 The end result
was that millions of households moved from renting to owning their houses thereby
121
See Andrew Kish, Perspectives on Recent Trends in Consumer Debt, Payment Cards Center Discussion
Paper, Federal Reserve Bank of Philadelphia 24 (June 2006); see also Alan Greenspan, The Mortgage
Market and Consumer Debt, remarks presented at America’s Community Bankers Annual Convention,
Washington,
D.C.,
Oct.
19,
2004,
available
in
http://www.federalreserve.gov/boardDocs/Speeches/2004/20041019/.
122
United States Census Bureau, Housing Vacancies and Homeownership, Table 14. Homeownership
Rates
for
the
U.S.
and
Regions:
1965
to
Present,
available
in
http://www.census.gov/hhes/www/housing/hvs/historic/histt14.html
123
Karen E. Dynan and Donald L. Kohn, The Rise in U.S. Household Indebtedness: Causes and
Consequences at p. 14, Federal Reserve Board, Finance and Economics Discussion Series, Divisions of
Research & Statistics and Monetary Affairs 2007-37 (emphasis added)).
124
See Francois Ortalo-Magne & Sven Rady, Housing Market Dynamics: On the Contribution of Income
Shocks and Credit Constraints, 73(April) REV. ECON. STUDIES 459 (April 2006). The loan to value ratio
for mortgages has risen during this period as well. See Wenli Li, Moving Up: Trends in Homeownership
and Mortgage Indebtedness, Q1 2005 BUSINESS REV. Q1 at 26, 28 (Philadelphia Federal Reserve Bank,
2005). This too has been largely driven by low-income households, which increased their LTV much faster
than the population at large. Id. At 29-30.
34
acquiring a new wealth accumulation vehicle125 but which also led to a measured increase
in the debt service ratio.126 In fact, even though housing prices rose rapidly during this
period, interest rates plummeted even more rapidly—with the end result being that
consumers could afford more expensive houses at the same or lower monthly payment.127
In addition, the Financial Obligations Ratio for non-homeowners for Fourth Quarter 2007
was lower than in 1986 and only slightly higher than in 1980, further suggesting that the
rise in household indebtedness was for new homeowners rather than extant
homeowners.128
To identify the impact of credit cards on the debt service ratio, we can more
specifically isolate the impact of all consumer debt (revolving and nonrevolving) from
mortgage debt. Removing mortgage DSR from the equation clearly demonstrates the
substitution effect of revolving consumer debt for traditional installment consumer debt,
as shown in Figure 2:
Figure 2
125
See Mark Dorns & Meryl Motika, The Rise in Homeownership, 2006-30 FRBSF ECON. LETTER (Nov. 3,
2006).
126
See Karen Dynan, Kathleen Johnson, and Karen Pence, Recent Changes to a Measure of U.S.
Household Debt Service, FED. RES. BULLETIN 417 (2005).
127
U.S. Housing Market Conditions, Table 11 “Housing Affordability Index,” Department of Housing and
Urban Development. Interest rates on 30-year fixed rate mortgages fell from 8.05 percent in 2000 to 5.8
percent in 2003-05 before rising to 6.4 percent in 2006. Id. at Table 14 “Mortgage Interest Rates, Average
Commitment Rates, and Points, 1973-Present.” In 2000 the average price of existing homes nationwide
was 143,600 and by 2005 the average price was $219,600. In some areas of the country prices almost
doubled. Id. at Table 9, “Existing Home Prices.” Points fell as well during this period. Id. at Table 15.
Nor does it appear that consumers have been shifting substantial consumer debt into home equity loans.
Home equity loans are a relatively small portion of total home mortgage debt—ten percent or less—and
this ratio has been largely constant over the past decade. Federal Reserve Board, Flow of Funds Report,
Table L.218 “Home Mortgages.” In 2005 total mortgage liabilities were over $9 trillion whereas total
home equity loans were about $900 billion. From 2000 to 2005, the percentage of household asset value
has remained largely constant, varying between 56-58% during that period. Id.; see also Christopher L.
Cagan, Mortgage Payment Reset: The Rumor and the Reality 3 (Feb. 8, 2006). In addition, the definition
of “consumer debt” includes things such as medical expenses and student loans.
128
See Federal Reserve, Household Debt Service and Financial Obligations Ratios.
35
Non-Mortgage Debt Service Ratio
0.075
Proportion of Income
0.07
0.065
0.06
0.055
0.05
20
07
q1
20
04
q1
20
01
q1
19
98
q1
19
95
q1
19
92
q1
19
89
q1
19
86
q1
19
83
q1
19
80
q1
0.045
Year
As Figure 2 illustrates, the overall debt service ratio for all consumer, nonmortgage debt (consumer revolving plus nonrevolving debt) has fluctuated in a fairly
narrow band during the period 1980 to 2006—between about 5%-7% of income during
this entire period. In fact, the non-mortgage debt service ratio was actually slightly
higher at the beginning of the data series in 1980 (0.0633) than at the end in the first
quarter of 2006 (0.0616) with local peaks and troughs throughout. There is no upward
trend in the debt service ratio that can explain the quintupling of the bankruptcy filing
during that period.
Further isolating non-mortgage consumer debt into revolving and nonrevolving
components visually illustrates the substitution effect, as shown in Figure 3:
Figure 3
36
Consumer Non-Mortgage Debt Service Ratio: 1980-2007
0.06
Proportion of Income
0.05
0.04
Non-Revolving
DSR
0.03
0.02
Revolving
DSR
0.01
1
07
q
1
20
04
q
1
20
01
q
1
20
98
q
1
19
95
q
1
19
92
q
1
19
89
q
1
19
86
q
1
19
83
q
19
19
80
q
1
0
Year
Examining Figure 3, several points can be noted. First, from the 1980s into the
1990s, there is a clear inverse trend between the revolving debt service ratio and the
nonrevolving debt service ratio, illustrating the substitution effect. Second, from the mid1990s to the present, both the nonrevolving and the revolving debt service ratios have
leveled off and been largely constant, thus the substitution effect appears to have reached
a sort of equilibrium by the mid-1990s. Thus, the overall debt service ratio has remained
largely unchanged during this period.
Moreover, even though this data alone shows a
clear substitution effect between revolving and nonrevolving debt, because of
peculiarities in the way that the debt service ratio is measured, even this measurement
overestimates the contribution of revolving debt to the debt service ratio.
First, wealth has risen more rapidly than income over the past few decades, first
because of the rise in financial wealth from the roaring stock market of the late-1990s,
37
and then the rapid appreciation in housing values into the 2000s.129
When wealth
increases, consumers rationally borrow against and consume some of it. This is known as
the “permanent income hypothesis”130 or “life cycle model”131 of consumption.
According to the permanent income hypothesis, consumers choose their consumption
patterns based on long-term income expectations and not just their current income. An
increase in wealth that is perceived to be permanent is functionally equivalent to an
increase in future income. As a result, consumers will base their consumption on their
wealth and expected permanent income, whether wealth takes the form of intangible
(financial) assets, tangible assets (real property), or human capital (such as investments in
education that raise your expected income stream in the future). An increase in wealth
that is expected to be permanent increases long-run expected income and will therefore
boost consumption today. To the extent that this wealth is otherwise illiquid (such as
human capital or retirement assets) consumer credit is one mechanism for borrowing
against this wealth to increase consumption today, especially when long-term wealth
grows more rapidly than short-term income.
In fact, non-mortgage consumer debt has been about 4% of household wealth for
several decades.132 During periods of rapidly increasing household wealth (such as
during the 1990s) consumers increase their consumption and consumer debt in order to
liquefy some of this accumulated wealth to finance consumption. But because these are
See Greenspan, The Mortgage Market and Consumer Debt, supra note 121, at 3 (noting that “the ratio
of households’ net worth to income has risen to a multiple of more than five after hovering around four and
one-half for most of the postwar period”); see also Zywicki, supra note 117, at 1483-92.
130
See MILTON FRIEDMAN, A THEORY OF THE CONSUMPTION FUNCTION (1957).
131
See Albert Ando & Franco Modigliani, The “Life Cycle” Hypothesis of Saving: Aggregate Implications
and Tests, 53 AM. ECON. REV. 55 (1963); IRVING FISHER, THE THEORY OF INTEREST (1930). An accessible
summary of the empirical evidence is available in Tullio Japelli, The Life-Cycle Hypothesis, Fiscal Policy,
and Social Security, working paper (Feb. 2005), available in http://www.csef.it/papers/modigliani.pdf.
132
Thomas A. Durkin, Discussion, THE IMPACT OF PUBLIC POLICY ON CONSUMER CREDIT 40 (Thomas A.
Durkin & Michael E. Staten eds., 2002).
129
38
capital gains and not earned income, the debt service ratio excludes them from its
calculations, even though it adjusts for the liquefying of the wealth holdings through
increased consumption and taxes. The marginal propensity to consume out of wealth is
larger for increases in house values than financial wealth—every dollar of increased
housing wealth generates about a 6-8 cent increase whereas the marginal propensity to
consume out of financial wealth is only about 2 cents of every additional dollar.133 Thus,
during periods of wealth growth, and especially rapid increases in real estate equity such
as in recent years, consumers tend to borrow against that wealth in order to increase
consumption today, in order to preserve a more stable pattern of consumption through
time. By borrowing against their accumulated wealth holdings to finance consumption
(with no corresponding increase in measured income) this will tend to lead to an
overestimation of the debt service ratio.
Second, the Federal Reserve overestimates the revolving credit debt service ratio
because its methodology inadvertently counts some transactional use as revolving.134
The Federal Reserve measures the debt service ratio on revolving credit by the amount of
revolving credit outstanding at the end of a given month. As a result, the data measures
as part of the outstanding balance on revolving credit, and thus as part of the DSR, those
balances on transactional accounts that will be paid in full at the close of the billing cycle
133
Raphael Bostic, Stuart Gabriel, and Gary Painter, House Wealth, Financial Wealth, and Consumption:
New Evidence from Micro Data, working paper, University of Southern California (Dec. 2005); see also
John D. Benjamin, Peter Chinlov, and G. Donald Jud, Real Estate Versus Financial Wealth in
Consumption, 29(3) J. REAL ESTATE FINANCE AND ECON. 341-354 (2004) (estimating marginal propensity
to consume out of housing wealth as 8 cents per dollar and 2 cents per dollar for financial wealth). See
Greenspan & Kennedy, supra note 126, at 3-6 (summarizing studies).
134
Kathleen W. Johnson, Convenience or Necessity? Understanding the Recent Rise in Credit Card Debt,
finance and economics Discussion Series, Divisions of Research & Statistics and Monetary Affairs, Federal
Reserve Board, Working Paper 2004-47; see also Kathleen W. Johnson, The Transactions Demand for
Credit Cards, 7 B.E. J. OF ECON. ANALYSIS & POL’Y, Issue 1, Article 16 (2007), available in
http://www.bepress.com/bejeap/vol7/iss1/art16; see also William Hempel, Discussion, in THE IMPACT OF
PUBLIC POLICY ON CONSUMER CREDIT, supra note 132, at 66, 67.
39
but happen to be outstanding at the time of reporting. Moreover, this measurement error
has been exacerbated over time because transactional or “convenience” use of credit
cards has been rising more rapidly than true revolving credit. During the past 15 years,
transactional use grew by approximately 15% per year (primarily because of various
incentives such as frequent flier or rebate programs), whereas the amount borrowed on
credit cards grew only about 6 ½% per year. The percentage of credit card transactions
that are paid off at the end of each month also has risen as has the percentage of credit
card holders who are transactors.135 In addition, the median monthly charge amount for
transactional users has risen over four times more rapidly for transactional users than for
revolvers.136 The overestimation that results from measuring transactional credit card use
as credit card borrowing for purposes of the debt service ratio tends to overstate the
revolving credit debt service ratio by approximately ten percent, an effect that has
doubled in the past decade as a result of the rapid rise of transactional use relative to
revolving.137
In addition, because transactional users do not have to pay for their purchases
until the end of the billing period plus the grace period after receiving their bill, they have
the opportunity to take advantage of interest rate “float” during the time between their
purchase and actual payment of the obligation, which may be as long as 45-60 days.138
During that period, a transactional user essentially receives a free loan from the credit
135
The percentage of convenience users relative to revolvers has risen steadily over time as credit cards
have replaced checks and cash as a transaction medium. See Johnson, Transactions Demand, supra note
134; see also Delinquency on Consumer Loans: Testimony Before the House Comm. on Banking and Fin.
Servs., 104th Cong. 1 (1996) (statement of Lawrence B. Lindsey, Member, Bd. of Governors of the Fed.
Reserve Sys.).
136
Johnson, Convenience or Necessity?, supra note 134.
137
See Johnson, Convenience or Necessity?, supra note 134. It may be as high as fifteen percent. See
Kish, supra note, at 31.
138
If, for instance, a purchase is made at the beginning of a new billing period but the bill is not paid until
the end of the grace period.
40
card issuer at zero percent interest during which time those same funds can be invested in
assets that generate a positive return, even if only a money market account or similar
safe, short-term investment.139 Revolvers are more likely to make use of debit cards than
are nonrevolvers, which can be explained by the fact that revolvers do not receive the
benefit of interest-rate float because they are required to pay the full interest on the
account, so they rationally use debit cards instead.140 The greater ability to plan the
timing of expenditures also enables those with credit cards to carry smaller precautionary
balances in their checking accounts than those without credit cards, enabling them to
invest their money in interest-bearing accounts during the period of the interest-free
float.141
In fact, despite these factors that that bias toward an overestimation of credit card
credit, there still is no discernible increase in the debt service ratio for nonmortgage
consumer debt over the past 25 years.142 Indeed, the DSR for installment credit would
have fallen still further over time, but for an unexpected jump in installment debt around
139
Technically the interest rate is slightly negative because of the time value of money.
Jonathan Zinman, Debit or Credit? (Working Paper, Dartmouth College Aug. 10, 2006). Researchers
have found similar results in Australia: transactors use credit cards for payments almost twice as frequently
than revolvers (22 percent of transcations compared to 12 percent) and revolvers are substantially more
likely to use debit cards than transactors. See David Emery, Tim West, and Darren Massey, Household
Payment Patterns in Australia, in RESERVE BANK OF AUSTRALIA, PROCEEDINGS OF PAYMENT SYSTEMS
REVIEW
CONFERENCE
139,
158-59
(Nov.
29,
2007),
available
in
http://203.34.60.68/PaymentsSystem/Publications/PaymentsInAustralia/PaymentsSystemReviewConferenc
e/2007/index.html. Brown and Plache find that 62 percent of revolvers who have a general purpose debit
card actually used that card whereas only 37 percent of nonrevolvers used their debit card. See Tom Brown
& Lacy Plache, Paying with Plastic: Maybe Not So Crazy, 73 U. CHICAGO L. REV. 63, 84 (2006).
141
EVANS & SCHMALENSEE, supra note 33, at 91-93; Johnson, Transactions Demand, supra note 134, at 3.
A full accounting of the revolving credit DSR would include the ability to earn interest by maintaining
smaller precautionary balances.
142
See Hempel, supra note 134, at 67 (“[C]onsumer credit has been fairly constant relative to income over
the past 30 years, but the composition has changed.”). Some of the growth in credit card credit also may
have been a substitution away from informal types of credit that traditionally have been poorly measured,
thus if the substitution is to better-measured types of credit, this would tend to artificially inflate the debtservice ratio as well. A recent study of low and moderate-income households found that 15 percent of
those in the study reported that they had previously obtained financing from institutions generally excluded
by aggregate statistics, such as pawn shops, payday lender, and rent-to-own establishments. See SIEDMAN,
ET AL., supra note 67.
140
41
2000, caused by two developments on margins where credit cards are poor substitutes for
installment loans.
First, the low-interest rate environment in the economy, combined with
promotional zero and low interest rate deals on automobile loans beginning around 2000
led to many consumers who had been leasing automobiles to switch to purchasing
automobiles, thereby leading to an increase in automobile installment debt.143
The
majority of installment borrowing today is for vehicle purchases.144 Thus, there was a
slight increase in the total amount of household installment credit for automobiles
beginning around 2000, but this was offset by a corresponding decline in automobile
lease commitments, thus there was no worsening of household financial condition, even
though this too tended to inflate the debt service ratio for installment debt.145 Second,
there has been a dramatic rise in student loan obligations (which are installment loans)
over time.146 From 1977 to 2004, total outstanding balances on student loans rose from
one percent of outstanding balances on consumer nonmortgage credit accounts to 21
percent, a level almost equal to total outstanding balances on credit cards with both
trailing only automobile loans as a percentage of household nonmortgage consumer
debt.147 Leaving aside automobiles and mortgages, student loan debt rose from 28% of
143
See Aizcorbe et al., supra note 63, at 24. The growing popularity of sport utility vehicles, which were
both more expensive and depreciate more slowly than cars, simultaneously increased indebtedness and
increased household assets. Id. at 17.
144
See Bucks, et al., supra note 63, at p. A30. From 2001 to 2004 the percentage of nonmortgage
consumer debt attributable to vehicle purchases rose form 54.8 percent to 55.5 percent.
145
See Dynan, et al,, supra note 126, at 421; see also Federal Reserve Release G.20 “Finance Companies”
(noting peak in outstanding receivables on motor vehicle leases in August 2000, followed by decline over
next several years).
146
See Aizcorbe et al., supra note 63, at 24. From 2001 to 2004, balances on student loans rose from 22.2
percent of household installment debt to 26.0 percent.
147
FED REPORT, supra note 6, at 8, Table 4. Credit card balances rose from 1970 to 1995 reaching 26% of
total outstanding household debt before falling from 1995 to 2004 from 26% to 22%. Id.
42
outstanding consumer credit to 58%.148 By contrast, personal loans fell from 28% to 8%
of nonauto, nonrevolving consumer loans, as they were “replaced” by credit card debt.149
Average interest rates on consumer loans generally fell steadily during this period as
well, thereby reducing the debt service ratio, even as total indebtedness rose.150 Thus,
there is no evidence that the increase in credit card debt has made households worse off,
as low interest rates and the substitution effect have kept the debt service ratio largely
constant for 25 years.
Evidence of the substitution effect is also found among bankruptcy filers.
Consider the following data drawn from Sullivan, et al.:
Figure 4
Debt of Bankruptcy Filers
80000
70000
60000
Dollars
50000
Total Debt
Secured Debt
40000
Unsecured Debt
30000
Credit Card Debt
20000
10000
0
1981
1991
Year
1997
148
Dynan, et al., supra note 126, at 418.
Id.
150
Id. at 419.
149
43
As illustrated in Figure 4, from 1981 to 1997 the average amount of total debt
held by bankruptcy filers remained constant, but the ratio of credit card debt to total
unsecured debt increased, suggesting a substitution between credit card debt and other
unsecured debt. Sullivan, et al., find that in 1981, total debt for bankruptcy filers was
$68,154, of which unsecured debt was $27,365.151 By 1997, mean total debt among
bankruptcy filers had actually fallen slightly to $61,320 and unsecured debt rose slightly
to $29,529. Although total debt and total unsecured debt remained relatively constant,
mean credit card debt among bankruptcy filers rose from $3,635 to $14,260 during this
period and median credit card debt rose from $2,649 to $9,345.152 Thus, the substitution
effect is evident among bankruptcy filers specifically, as credit card debt has risen even
as total debt and total unsecured debt have remained largely constant. Credit card debt
nonetheless remains a small fraction of overall household debt for bankruptcy filers.
B. Credit Cards and Substitution by Low Income Families
Some have suggested that even if credit cards haven’t increased the consumer
debt burden for all or the average family, greater access of low-income and riskier
borrowers to credit cards has increased the prevalence of economic distress and
bankruptcy among this subclass of consumers.153 Economic theory and available data are
inconsistent with this hypothesis. Again, this appears to be because the use of credit
151
See SULLIVAN, ET AL., FRAGILE, supra note 2, at 66, Table 2.4. All values are in 1997 dollars.
Id. at 122, Table 4.1. The 1981 figures include only bank-type cards whereas the 1997 figures include
all credit card debt. As noted, during this period there was a general substitution from other types of credit
cards to bank-type cards, thus the 1981 figures may underestimate total credit card debt. In 1991, the mean
debt for bank-type cards only among bankruptcy filers was $11,529, thus using the same category as 1981
there was plainly a large increase in bank-type card debt during the 1980s.
153
See Warren, supra note 2, at 1083; Susan L. DeJarnatt, Once is Not Enough: Preserving Consumers’
Rights to Bankruptcy Protection, 74 IND. L.J. 455, 499 (1999); Bernard R. Trujillo, The Wisconsin
Exemption Clause Debate of 1846: An Historical Perspective on the Regulation of Debt, 1988 WISC. L.
REV. 747, 749 (1998).
152
44
cards among low-income consumers has resulted from a substitution from other types of
credit, rather than an increase in debt burdens. As a report of the Chicago Federal
Reserve Bank concluded, “The increase in the credit card debt burden for the lowest
income group appears to be offset by a drop in the installment debt burden. This suggests
that there has not been a substantial increase in high-interest debt for low-income
households, but these households have merely substituted one type of high-interest debt
for another.”154 Bird, Hagstrom, and Wild similarly conclude that although credit card
ownership among lowest-income households doubled between 1983 and 1995 and the
average balance for those who revolved increased substantially as well, “the data reveal
little direct evidence that the increase in [credit card] debt has caused an increase in
economic distress, either for poor households or households at large.”155 Indeed, they
found that credit card delinquencies among poor households fell during this period.156
Two studies have empirically examined the hypothesis that rising bankruptcies
reflected increased lending to riskier borrowers and found little support. The first, by
economists Donald P. Morgan and Ian Toll, concludes, “If lenders have become more
willing to gamble on credit card loans than on other consumer loans credit card chargeoffs should be rising at a faster rate [than non-credit card consumer loans] . . . . Contrary
to the supply-side story, charge-offs on other consumer loans have risen at virtually the
same rate as credit card charge-offs.”157 This “suggest[s] that some other force [other
154
Wendy M. Edelberg & Jonas D. M. Fisher, Household Debt, CHI. FED. LETTER, Nov. 1997, at 1, 3
(1997); see also id. at 4 (“[I]ncreases in credit card debt service of lower-income households have been
offset to a large extent by reductions in the servicing of installment debt.”).
155
Edward J. Bird, Paul A. Hagstrom, & Robert Wild, Credit Card Debts of the Poor: High and Rising, 18
J. POL’Y ANALYSIS & MANAGEMENT 125, 132 (1999).
156
Id.
157
Donald P. Morgan & Ian Toll, Bad Debt Rising, CURRENT ISSUES ECON. & FIN. (Research & Mkt.
Analysis Group, Fed. Reserve Bank, New York, N.Y.), Mar. 1997, at 1, 3. Consistent with the argument
presented in the text here, Morgan and Toll conclude that increased consumer demand for credit cards
45
than extension of credit cards to high-risk borrowers] is driving up bad debt.”158 A
second study, by David B. Gross and Nicholas S. Souleles, concludes that changes in the
risk-composition of credit card loan portfolios “explain only a small part of the change in
default rates [on credit card loans] between 1995 and 1997,” finding instead that rising
bankruptcies occurred entire population of borrowers became more risky and prone to
filing bankruptcy over that time period.159
Compared to the population at large, low-income and younger households face a
limited number and variety of credit options, thus credit cards are an especially attractive
compared to available alternatives. Research by economist Gregory Elliehausen finds
that users of high-price credit products such as pawnbrokers, rent-to-own, automobile
title loans, and tax refund anticipation loans “are more likely than higher income groups
to have limited discretionary income after necessities and be more vulnerable to
unexpected expenses.”160 These are precisely the types of high-cost loans that consumers
substitute away from when given the opportunity, as well as other non-“fringe” types of
credit, such as layaway, retail store credit, and personal finance companies.
Available data also supports the hypothesis that the growth in credit card use by
low-income households has been a substitution away from these comparatively lessattractive types of consumer credit. The Federal Reserve’s debt service ratio data does
not provide disaggregated data by income or other subgroups. The Survey of Consumer
relative to other forms of consumer credit is driving the increase in credit card debt, not a supply-side shift.
Id.
158
Id. at 4.
159
Gross & Souleles, supra note 112.
160
Gregory Elliehausen, Consumers’ Use of High-Price Credit Products: Do They Know What They are
Doing? Networks Financial Institute at Indiana State University Working Paper 2006-WP-02 at p.18 (May
2006).
46
Finances, however, does provide a picture of the debt service ratio of households by
income group, as shown in Figure 5:
Figure 5
Debt Service Ratio (Lowest Quintile) and Bankruptcy
20.0
Debt Service Ratio
18.0
12
17.0
10
16.0
15.0
8
14.0
6
13.0
12.0
Bankruptcies/1000 Households
14
19.0
DSR Lowest
Quintile
Bankruptcies
per 1000
Households
4
1992
1995
1998
2001
2004
2006
Year
Source: Survey of Consumer Finances
As can be seen, over the past 15 years the debt service ratio for lowest income
quintile households has fluctuated between 16-19 percent, with no discernible trend. Nor
is there any obvious relationship with bankruptcy filings—the DSR for lowest-quintile
households rose from 1992 to 1995, while the bankruptcy filing rate actually fell slightly
during that period. Then the DSR for this group fell from 1995 to 2001, even though the
bankruptcy filing rate jumped. Over the decade from 1992 to 2001 the DSR for lowest
income households actually fell, even though the overall consumer bankruptcy filing rate
rose substantially. Moreover, the growth in the DSR of low income households from
2001-2004 years is attributable to the rapid growth in home ownership among low-
47
income families during that time, thus the growth primarily has been in the mortgage
component of the DSR rather than the consumer credit DSR.161 In fact, the percentage of
households carrying a balance on their credit cards actually declined between 2001-2004
for the lowest two income groups, the lowest wealth group, the youngest age group, and
nonwhite or Hispanic families.162 The ratio of household debt payments to income for
lowest quintile households is also similar to that for all other income ranges except the
highest decile.163
The Survey of Consumer Finances also calculates the percentage of households
that are in “financial distress” as measured by the number of households for which 40
percent or more of their income is dedicated to servicing debt. By this measurement,
there is no discernible evidence that the percentage of low-income households in distress
has risen over the past 15 years as a result of greater access to credit cards; indeed, in
2004 the percentage of low income households in financial distress was at its lowest level
of any time during this period, even though the overall bankruptcy rate had risen
dramatically during that time, as shown in Figure 6:
Figure 6
161
See discussion in sources cited in note 121 and accompanying text.
Bucks, Kennickell, & Moore, supra note 144, at p. A30.
163
See Brian W. Cashell, Rising Household Debt: Background and Analysis, CRS REPORT FOR CONGRESS
11 (Apr. 20, 2007).
162
48
Lowest Income Quintile Households with Financial Distress
30.5
16
30.0
14
29.5
12
29.0
10
28.5
28.0
8
27.5
6
Lowest Quintile
Households with DSR
> 40%
Bankruptcy Filing
Rate
27.0
4
26.5
2
26.0
25.5
0
1992
1995
1998
2001
2004
2006
Again, there is no obvious relationship between the percentage of low-income
households in financial distress and the bankruptcy filing rate, suggesting that the
increased access to credit cards has not led to an increase in financial distress. Consider
one indicia of whether risk composition has changed, the delinquency rate on credit card
loans, as shown in Figure 7:
Figure 7
49
Delinquency Rate, Credit Card Loans
Delinquency Rate
6
5.5
5
4.5
4
3.5
20
07
Q1
20
05
Q1
20
03
Q1
20
01
Q1
19
99
Q1
19
97
Q1
19
95
Q1
19
93
Q1
19
91
Q1
3
Year
Source: Federal Reserve
As can be seen in this chart, delinquency rates have fluctuated over time, but there is no
upward trend in delinquency rates, and, in fact, there is a slight downward trend over the
course of the reported data, depending on the starting point.164 Credit card delinquencies
closely follow the business cycle. Similarly, the pattern in credit card delinquency rates
is comparable to delinquency rates on other consumer loans over the same time period,
suggesting that the riskiness of credit card borrowers is comparable to that for other types
of credit, as shown in Figure 8165:
Figure 8
164
The Federal Reserve did not compile data on credit card delinquency rates until 1991. Other measures
of delinquency on credit card loans also fail to find a discernible upward trend in delinquencies on credit
card loans over time, as one would expect if credit was being extended to riskier borrowers. See Bucks,
Kennickell, & Moore, supra note 144, at p. A34 & n. 49. 1991 was a recession year, so delinquency raters
were higher during that period, as with during the 2001-02 period, so whether the trend is actually
downward depends on the starting point. On the other hand, as noted, there is no discernible upward trend
over the period that could account for the dramatic rise in bankruptcy filings during this era.
165
As can be readily observed, delinquency on credit card loans is systematically higher than other
consumer loans for reasons unrelated to the risk composition of the pool of borrowers, such as the greater
difficulty in collection, the unsecured nature of credit card credit, and greater problems of moral hazard and
adverse selection.
50
Delinquency Rate Consumer Loans
Delinquency Rate
6
5
Credit Card
4
All Consumer
3
Non-Credit
Card
2
1
20
07
Q1
20
05
Q1
20
03
Q1
20
01
Q1
19
99
Q1
19
97
Q1
19
95
Q1
19
93
Q1
19
91
Q1
0
Year
Source: Federal Reserve
A report of the Federal Reserve concludes that improvements over time in riskscreening technology by card issuers and account management techniques, such as
controls on credit limits, have enabled credit card issuers to continually extend credit to
wider swaths of society while at the same time effectively controlling risk.166
Professor Angela Littwin argues, however, that there has been a “low degree of
substitution” between credit cards and fringe lending for lower-income households.167
She reaches her conclusion through interviews of low-income women who express a
preference for the use of fringe lending products such as pawnshops and payday lenders
instead of credit cards.
From this she infers that low-income consumers have not
substituted much borrowing activity from fringe lending products to credit cards, and
thereby concludes that the substitution hypothesis is invalid.
166
167
FED REPORT, supra note 6, at 10.
Littwin, Comparing Credit Cards, supra note, at 18-19.
51
But while providing interesting insights into the preferences of low-income
consumers, Littwin’s analysis does not actually address the substitution hypothesis. The
substitution hypothesis is not a theory of consumer preferences, but rather an explanation
for fact-based observations that appear anomalous—credit card debt has increased
substantially over time but that overall consumer debt burdens have not increased as a
result. Littwin does not appear to question the fact that credit card borrowing by lowincome consumers has risen (notwithstanding their expressed dislike for credit cards),
and it is unclear whether she believes the overall consumer debt burden has risen. Thus,
her analysis doesn’t actually speak to the validity of the substitution hypothesis.
Professor Littwin’s argument seems to be that low-income consumers have not
substituted much debt from fringe lending products to credit cards, which she takes as
rebutting the substitution hypothesis. Her exact argument is a bit opaque, but there seem
to be two possible constructions. Neither is consistent with available data.
First, Littwin may be arguing that the substitution hypothesis is invalid because
low-income consumers use both fringe lending and credit cards, so increased use of credit
cards has not led to a decreased use of fringe lending products. Thus, she may be arguing
that credit card borrowing has been additive to fringe borrowing, not a substitute; or
alternatively, that credit cards and fringe lending are substitutes, just not on a one-to-one
basis such that a one dollar increase in credit card borrowing does not result in an equal
decrease in fringe borrowing.168 But if this is so, then the increase in credit card
168
This may be what Littwin has in mind when she refers to fringe lending and credit cards as being
“complements” rather than “substitutes,” which is that credit card borrowing is additive to fringe lending
rather than a one-to-one substitute. “Complements” are those goods such that increasing the price of one
good reduces demand for the other, e.g. a negative cross-price elasticity on the margin. Classic examples
of complements include peanut butter and jelly or coffee and Cremora—i.e., if the price of peanut butter
rises, demand will fall not just for peanut butter but also for jelly because they go together. “Substitutes”
are those goods such that increasing the price of one good increases demand for the other. For instance,
52
borrowing should be reflected in an increase in the overall consumer debt burden. But
Littwin provides no evidence contrary to the data reported above that indicates that the
debt-service burden for consumer credit has remained constant over time.
Second, she may be arguing that the cross-elasticity of demand between credit
cards and fringe lending is low (i.e., inelastic) and there has been minimal substitution
from fringe borrowing to credit card borrowing among low-income households, which
would explain the minimal increase in measured consumer debt burdens. If this is her
argument, then it would be consistent with the data indicating that there has been no
increase in the overall consumer debt-service burden. But this seems to be inconsistent
with the apparent increase in the use of credit card use by low-income households over
the past 25 years. Again, this is a statistical question, not one of expressed consumer
preferences. Nor does she provide an explanation as to why the this subset of society
would hold preferences that deviate so dramatically from the preferences of other
socioeconomic groups, such as the middle class, for which credit cards are immensely
popular.
Moreover, the finding that consumers use both credit cards and other types of
consumer credit not only is consistent with the substitution hypothesis, but is exactly
what the hypothesis would predict—consumers will hold portfolios of all types of credit
products and will switch among them depending on their relative prices. So if the cost of
credit card borrowing goes up, consumers at the margin will switch from credit cards to
fringe borrowing products and vice-versa. As noted, there is some evidence that suggests
increasing the price of driving (such as by increasing the cost of parking) will increase demand for taking
the bus, which is a substitute for driving. A “complements” model of consumer credit, therefore, would
predict that an increase in the price of fringe borrowing would lead to a decrease in the use of credit card
borrowing, and that a decrease in the price of fringe borrowing would also lead to a decrease in demand for
credit card borrowing. This doesn’t see to be a plausible model of consumer credit usage.
53
that consumers who use both credit cards and fringe borrowing generally use their credit
cards first, but that when they run up against their credit limits and would face over-limit
charges, they switch over to fringe borrowing. Without knowing the pattern of consumer
use—e.g., do they use their credit cards more or less when the price of fringe lending
rises—Professor Littwin cannot determine the degree of substitutability among different
products.169
C.
Understanding the Substitution Effect
The growth of credit card credit thus has resulted from a substitution by
consumers of credit cares for other, less-attractive forms of consumer credit.
Nonetheless, many commentators nonetheless insist that the growth in credit card use as a
source of revolving credit is irrational in light of the “high” interest rates charged on
credit cards. But this is based on a misunderstanding of the reasons for the growth in
credit card use. Figure 9 displays interest rates for credit cards (both generally and for
those accounts actually assessed interest) and their most competitive comparable credit
option, a 24-month personal loan:
Figure 9
169
Her regression data provides no information on cross-elasticities of demand nor do her data include any
sort of price change or analytically similar change in supply or demand conditions that would allow
identification of a cross-price effect.
54
Credit Cards v. Personal Loans
20
19
Interest Rate
18
24 Month Personal
Loan
17
16
All Credit Card
Accounts
15
Credit Card
Accounts
Assessed Interest
14
13
12
19
72
19 -02
74
19 -02
76
19 -02
78
19 -02
80
19 -02
82
19 -02
84
19 -02
86
19 -02
88
19 -02
90
19 -02
92
19 -02
94
19 -02
96
19 -02
98
20 -02
00
20 -02
02
20 -02
04
20 -02
06
-0
2
11
Year
As can be seen, over time the difference between interest rates on short-tem personal
installment loans and credit card accounts is small and has narrowed over time. Indeed,
in recent years the interest rate on credit card accounts has frequently fallen below that of
short-term personal loans, and the interest rates on those accounts actually assessed
interest frequently is even lower than for all accounts.170 Moreover, once initiation and
application fees on personal loans are taken into consideration the total cost of personal
loans is almost certainly higher overall.171
Some commentators nonetheless have argued that this substitution by consumers
to greater reliance on credit cards is evidence of widespread consumer irrationality rather
170
See also Brown & Plache, supra note 20, at 80 (noting that higher percentage of nonrevolvers have
credit cards with APRs greater than 10% than do revolvers). Further anecdotal evidence is a survey
published each Friday in the Washington Times of area consumer banking rates each Friday, which show
similar results. On September 14, 2007, for instance, the prevailing interest rate in the Washington, D.C.,
area on personal loans was one percentage point higher than on credit cards (10.75% for personal loans
versus 9.75% for credit cards); at various other times the spread has been larger but reported credit card
rates consistently have been lower. See WASHINGTON TIMES, p. C10 (Sept. 14, 2007).
171
See supra note 98.
55
than a beneficent process of market competition.172 But these arguments ignore the very
possibility of a substitution effect, implicitly assuming that all debt has been piled upon
preexisting consumer debt burdens. It also is implicitly assumed that there must have
been an increase in debt burdens, both because of an increase in indebtedness as well as a
belief that credit cards impose higher interest rates than the types of credit that they
replaced. Moreover, as illustrated in Figure 9, the substitution of relatively lower-cost
credit cards when compared to available alternatives is perfectly consistent with
consumer rationality. Professor Mann does consider the substitution effect and generally
finds that increases in credit card debt do not generate increases in overall consumer debt,
a finding that is consistent with the substitution hypothesis, but not the distress
hypothesis.173
It is also asserted that credit cards are uniquely prone to consumer irrationality
and overspending.
But this argument usually is not based on a comparison to the
alternative types of consumer credit that they replaced, such as installment or “open
book” store credit or even retail store credit cards, which were widely-owned in the
1970s and which were subject to identical criticisms in earlier generations.174 Professor
White argues that credit cards are more prone to biases of “hyperbolic discounting” than
installment credit.175 This hypothesis is possible—White provides no evidence to support
it—but it is not particularly plausible.
Why would a consumer be more prone to
hyperbolic discounting bias when a purchase is made on a credit card with the full
172
See Bar-Gill, supra note 4. Bar-Gill offers no empirical support for his contentions and his theoretical
model rests on some problematic assumptions. See Joshua D. Wright, Behavioral Law and Economics,
Paternalism, and Consumer Contracts: An Empirical Perspective, 2 N.Y.U. J. LAW & LIBERTY 470, 48588 (2007).
173
See MANN, supra note 2, at 57, Table 4.3.
174
See discussion supra at note 47 and accompanying text.
175
White, supra note 2, at 181-82.
56
balance to appear on the statement and become due in full the next month as opposed to
an installment loan where the full price of the loan is concealed in monthly payments that
may stretch out over many months or even years, especially when the price of credit is
obscured in the price of the goods?176 White argues that the difference is that the
payments under installment loans are regularized, but if there is a question of hyperbolic
discounting presumably it would be salient at the time the loan is made, not when it is
repaid, especially when the payments are to be made two or three years into the future.
Student loans, for instance, are installment loans but it would be difficult to argue that
students anticipate the full cost of those loans more rationally than for credit cards.
Moreover, unlike many installment loans, credit card loans can be easily refinanced for a
better interest rate by switch balances to lower-rate cards. As a pure matter of theory,
installment loans seem, if anything, more prone to hyperbolic discounting than credit
cards.
The realities of installment loans reinforce this conclusion. As previously noted,
installment credit was criticized by contemporaries on precisely the same grounds that
credit card credit is criticized today—that it led consumers into a treadmill of debt. The
realities of consumer installment lending practices tended to reinforce this conclusion.
By requiring monthly or even weekly repayment schedules, merchant installment lenders
created multiple opportunities to interact with the customer. For example, Caplovitz
observed that merchant lending contracts often required weekly payments which the
176
For instance, consumers are often unable to understand the full cost of traditional installment loans such
as the APR and related terms. See JAMES M. LACKO & JANIS K. PAPPALARDO, FED. TRADE COMM’N.,
IMPROVING CONSUMER MORTGAGE DISCLOSURES: AN EMPIRICAL ASSESSMENT OF CURRENT AND
PROTOTYPE DISCLOSURE FORMS 35 (2007); Jinkook Lee and Jeanne M. Hogarth, The Price of Money:
Consumers’ Understanding of APRs and Contract Interest Rates, 18 J. PUB. POL’Y AND MARKETING 66
(1999); Diane Hellwig, Comment, Exposing the Loansharks in Sheep’s Clothing: Why Re-Regulating the
Consumer Credit Market Makes Economic Sense, 80 NOTRE DAME L. REV. 1567, 1591-92 (2005)
(summarizing studies).
57
customer usually would bring to the store and pay in person. Caplovitz observes that this
served several functions for the merchant.177
First, by dividing the repayment into
weekly payments it rendered the actual payments relatively small, which helped “to
create the illusion of ‘easy credit.” Although the concept of “hyperbolic discounting” had
not been labeled at that time, Caplovitz argues that this structure of payments reflected
that concept, because “Consumers are apt to think more of the size of the payments than
of the cost of the item or the length of the contract.”178 Caplovitz also argues that those
who shopped with installment credit had diminished levels of self-restraint and were less
able to defer gratification than others.179
Caplovitz also notes that those repeated
interactions also enables merchants to get to know their customers better, thereby making
it easier to monitor changes in their risk profiles.180
But Caplovitz notes another implication of the installment system—each time the
consumer would enter the store to make a payment that provided another opportunity for
the merchant to sell additional items to the consumer, a transaction that is made that
much easier because the customer is already in the store “where he can look at the
merchandise.”181 As a result, critics of the installment system claimed that it tended to
promote an endless cycle of debt by consumers, made all the easier by the way in which
the merchant-customer interaction was structured.
This is a completely different
relationship from the merchant-customer relationship when credit cards are involved.
As noted, complaints about the perceived irrationality or short-sightedness of
“other consumers” is as ubiquitous as credit itself, whether the product was installment
177
DAVID CAPLOVITZ, THE POOR PAY MORE: CONSUMER PRACTICES OF LOW-INCOME FAMILIES 24 (1963).
Id.
179
Id. at 118.
180
Id.at 24.
181
Id. at 25.
178
58
loans in mid-Twentieth Century America or credit cards today. Consumers today surely
are at least as sophisticated at using and shopping for credit as in the past and the ubiquity
of credit advertising has made informed shopping easier than ever.182 In fact, consumer
behavior involving credit cards appears to be generally consistent with rational economic
behavior. Revolvers are more aware of their interest rates and more likely to comparison
shop among cards on that basis than others, and those who carry larger balances are even
more likely to be aware of their interest rate and comparison shop on this term than those
who revolve smaller balances.183 Revolvers are more likely than convenience users to
read credit card solicitation material, and a larger proportion of revolvers said that they
would apply for a card with a lower rate if it were offered, and the larger the outstanding
balance the more likely the cardholder would be apply for a lower-rate card.184
Revolvers are more likely to hold a credit card with an annual fee but a lower interest rate
than are transactional users.185 In fact, as illustrated above in Figure 9 and as others have
found, this competition is so intense that credit card pricing today actually illustrates an
inversion of interest rates—revolvers actually tend to have interest rates that are lower on
182
Johnson notes, for instance, that over the past decade credit card users have become less myopic in
terms of their household financial planning. See Johnson, Transactions Demand, supra note 134, at 13-15.
Moreover, as noted, increased competition among payday lenders tends to decrease the price of these loans,
suggesting some degree of shopping behavior even among those borrowers. See supra note 77, and
accompanying text. In a hyper-competitive environment such as the credit card environment, shopping is
even easier for consumers.
183
See Thomas A. Durkin, Credit Card Disclosures, Solicitations, and Privacy Notices: Survey Results of
Consumer Knowledge and Behavior, FEDERAL RESERVE BULLETIN p. A 109, 112-A115 (2006) (80% of
revolvers examine APR frequently compared to 40% of transactors); Zywicki, Economics of Credit Cards,
supra note 48, at 104-09; Randall J. Pozdena, Solving the Mystery of High Credit Card Rates, 42 FRBSF
WEEKLY LETTER 2 (1991).
184
Durkin, supra note 183, at A117; Glenn B. Canner & Charles A. Luckett, Developments in the Pricing
of Credit Card Services, 78 FED. RES. BULL. 652, 663 (1992); Paul Calem & Loretta Mester, Consumer
Behavior and the Stickiness of Credit card Interest Rates, 85 FED. RES. BULLETIN 333 (1988); see also
Darryl E. Getter, Consumer Credit Risk and Pricing, 40 J. CONSUMER AFFAIRS 41, 57-60 (2006); Sha
Yang, Livia Markoczy, & Min Qi, Unrealistic Optimism in Consumer Credit Card Adoption, 28 J. ECON.
PSYCH. 170, 177 (2007).
185
See Brown & Plache, supra note 20, at 79-80.
59
average than nonrevolvers.186
According to the Survey of Consumer Finances the
median interest rate on the household credit card with the largest balance was 11.5
percent in 2004, a drop of 3.5% from 2001.187 Consumers also have become increasingly
savvy about exploiting “teaser rate” offers by “card surfing” from one teaser rate card to
the other.188 Those who do not revolve balances, by contrast, tend to focus on other
terms of credit card contracts, such as the grace period for payment, benefits such as
frequent flier miles, and whether there is an annual fee, just as standard economic theory
would predict.189
There also is no evidence that borrowers systematically underestimate their
likelihood of credit card borrowing or the cost of it.190 Most consumers choose the credit
card plan that is most suitable for their needs and those who do not tend to learn fairly
rapidly from their mistakes and switch to a more appropriate card, with those who made
the biggest mistakes being the ones most likely to switch.191 In fact, with respect to credit
cards empirical research indicates that where consumers err, they do so by overestimating
their likelihood of revolving rather than underestimating.192 Consumers are less likely to
186
Brown & Plache, supra note 20; Howard Beales & Lacey L. Plache, Rationality, Revolving, and
Rewards: An Analysis of Revolving Behavior on New Credit Cards, Working Paper (April 2007).
187
Daryl E. Getter, The Credit Card Market: Recent Trends, Funding Cost Issues, and Repricing Practices,
CRS REPORT FOR CONGRESS 2-3 (Feb. 27, 2008).
188
See Zywicki, Economics of Credit Cards, supra note 48, at 107-08.
189
Durkin, supra note 183, at A112; Canner & Luckett, supra note 184.
190
Canner & Luckett, supra note 184 , at 665; Cargill & Wendel, supra note 115, at 386.
191
See Brown & Plache, supra note 20; Sumit Agarwal, Souphala Chomsisengphet, Chunlin Liu, &
Nicholas S. Souleles, Do Consumers Choose the Right Credit Contracts?, Fed. Res. Bank of Chicago
Working Paper WP 2006-22 (Oct. 23, 2006); Sumit Agarwal, John C. Driscoll, Xavier Gabaix, & David
Laibson, Stimulus and Response: The Path from Naivete to Sophistication in the Credit Card Market,
Working Paper (Aug. 20, 2006). This “learning” phenomenon has been observed with other consumer
contracts as well. See Eugenio Miravete & Ignacio Palacios-Huerta, Rational Attention in a Repeated
Decision Problem, Working Paper (Sept. 2004).
192
Agarwal, et al., find that consumers rarely erred in choosing the no-annual fee, higher interest rate cards,
correctly predicting that they would not revolve. Instead, they disproportionately erred in choosing the
high-fee, low-interest rate card, and then failing to revolve enough debt to justify the payment of the annual
fee. Agarwal, et al., Do Consumers Choose, supra note 191, at 8-12. One study that purports to find an
“unrealistic optimism” bias in credit card borrowing did not try to determine whether there is also
60
revolve on higher-APR credit cards and are more likely to revolve where they pay an
annual fee in exchange for a lower interest rate.193 Those who pay no annual fee are the
least likely to revolve balances, indicating that consumers are not stockpiling credit cards
on which they are later induced to resolve balances.194 When a consumer obtains a new
credit card, the primary predictor of whether she will revolve on that card is whether she
revolved on the old card, suggesting that getting a new credit card does fundamentally
change consumer behavior or “seduce” consumers into revolving.
Moreover, a
cardholder becomes less likely to revolve balances the longer they hold their card. In
addition, consumers who hold rewards cards are less likely to revolve than those who do
not, thereby suggesting that the promise of these rewards does not induce a borrower to
short-sightedly “overconsume” and thereby unconsciously pile up debt.
It is also asserted that credit cards are an “expensive” form of consumer credit.
The basis for this claim is unclear, as those who assert it rarely state what alternatives
they are comparing. This belief may be the residue of an older concern that credit card
interest rates historically were “stickier” than other forms of consumer credit, in the sense
that underlying changes in the cost of funds were not reflected in the interest rates on
credit cards. This led some to conclude that there was a failure of competition in the
credit card market that led to large economic profits for credit card issuers. 195 In fact,
credit card interest rates were “stickier” in prior eras.
But this was best explained by
unrealistic pessimism or the relative frequency of optimism and pessimism bias. See Yang, et al., supra
note 184.
193
Beales & Plache, supra note 186.
194
Beales & Plache, supra note 186.
195
Ausubel, Failure of Competition, supra note 4.
61
standard economic factors, not because of a failure of competition in the credit card
industry or consumer irrationality.196
First, it is assumed that the interest rate is the only margin on which credit card
issuers compete, so that changes in the cost of funds will be directly reflected in changes
in credit card interest rates. But a majority of American households are transactional
users who rarely or never revolve balances on their cards, and thus would prefer
decreased fees, better customer service, or increased benefits such as frequent flyer miles
rather than a cut in the interest rate.197
Second, credit cards have higher operating costs than other types of credit,
including costs for activities such as ongoing customer support, account acquisition,
dispute resolution services, fraud protection, account administration, and the
administrative burden of processing a large number of relatively small transactions. For
credit cards, the cost of funds are approximately 30-40% of total costs, with charge-offs
amounting to as much as 30-40% and operating costs 20-30%.198 By contrast, the cost of
funds account for 60 percent of total expenses for installment consumer lending, about 70
196
See Zywicki, Economics of Credit Cards, supra note 48, at 110-28 from which this discussion is drawn.
See supra note 115; FEDERAL RESERVE BOARD, THE PROFITABILITY OF CREDIT CARD OPERATIONS OF
DEPOSITORY INSTITUTIONS 3 (Aug. 1998); see discussion in Zywicki, Economics of Credit Cards, supra
note 48, at 101-04.
198
Canner & Luckett, supra note 184, at 658; see also William F. Baxter, Section 85 of the National Bank
Act and Consumer Welfare, 1995 UTAH L. REV. 1009, 1016 (1995) (“[T]he cost of funds for credit card
lending comprises less than half, and possibly as little as one-quarter, of total costs (compared with sixty to
eighty percent of total costs for other types of bank lending) . . . .”); Johnson, Recent Developments, supra
note 161, at 477 n.7 (Autumn 2005) (citing reports in Credit Card News); GENERAL ACCOUNTING OFFICE,
CREDIT CARDS: INCREASED COMPLEXITY IN RATES AND FEES HEIGHTENS NEED FOR MORE EFFECTIVE
DISCLOSURES TO CONSUMERS 100 (Sept. 2006) [“GAO REPORT”]. Average operating expenses for credit
card lenders, as a percentage of total assets for banks, are over 9 percent versus 3.44 percent average for
other consumer lenders. Id. at 101. The actual ratios depend primarily on the charge-off rate, which is
higher during recessions. See EVANS & SCHMALENSEE, supra note 33, at 224.
197
62
percent for commercial lending, and nearly 80 percent for mortgage lending. 199 Thus, as
a purely mathematical matter, credit card interest rates would be expected to be less
responsive to a change in the cost of funds than other types of consumer credit because
those costs comprise less of the final price. In fact, credit card interest rates traditionally
were much less responsive to changes in cost-of-funds rates than other forms of
consumer lending, such as personal loans, car loans, and mortgages, regardless of
whether the cost of funds was rising or falling.200
Third, credit card loans are inherently riskier than other types of consumer credit.
Charge-off rates are higher and thus comprise a larger percentage of credit card
operations. Thus, when credit card charge-offs increase, the spread charged between the
underlying cost of funds and the interest rate increases.201 The percentage of credit card
charge-offs resulting from bankruptcy increased from 10 percent of all charge-offs in
199
Canner & Luckett, supra note 184, at 658; see also id. at 657; ALEXANDER RASKOVICH & LUKE FROEB,
HAS COMPETITION FAILED IN THE CREDIT CARD MARKET? 5 (U.S. Dep’t of Justice, Economic Analysis
Group Discussion Paper EAG 92-7, 1992).
200
See Zywicki, Economics of Credit Cards, supra note 48, at 111-19. This regularity is not unique to
credit cards. Interest rates on small loans of shorter duration have always been higher than longer-term,
higher-value loans, primarily because of the need to recoup the costs associated with granting and servicing
the loan. Thus, credit card interest rates are no stickier than the rates on other consumer loans, once
adjustments are made for differences in loan default rates. RASKOVICH & FROEB, supra note 199, at 6-11.
Even in the Middle Ages the interest rates charged by pawnshops (including publicly endowed pawnshops)
were much higher than interest rates charged on long-term loans secured by farms or estates. HOMER &
SYLLA, supra note 9, at 424-25. See also HOMER & SYLLA, supra note 9, at 427; see also DeMuth, supra
note 58, at 228; James M. Ackerman, Interest Rates and the Law: A History of Usury, 1981 ARIZ. ST. L.J.
61, 89 (1981) (”Administrative costs, however, are not directly related to loan size; it costs nearly as much
to do the paper work on a $100 loan as on a $100,000 loan. . . . Thus, small loans are more costly to make
and require a higher rate of interest.”).
201
See Adam B. Ashcraft, Astrid A. Dick, and Donald P. Morgan, The Bankruptcy Abuse Prevention and
Consumer Protection Act: Means-Testing or Mean Spirited? Working Paper, Federal Reserve Bank New
York (Dec. 19, 2006). This cost-based stickiness was reinforced by consumer demand decisions that
further reinforced this stickiness. When the cost of funds rate falls, interest rates on other types of
consumer loans fall faster than interest rates on credit cards. This causes low-risk borrowers to substitute
away from credit cards to these other types of loans, such as home equity loans. This leaves the remaining
pool of credit card borrowers riskier than before, which leads to increased charge-offs, which dampens the
downward trend in interest rates. Thus, for credit card lending, the charge-off rate will tend to be inversely
related to the cost of funds rate. See RASKOVICH & FROEB, supra note 199, at 7-9.
63
1982 to 50 percent in 2002.202
Credit card debts also are more difficult to collect than
other types of loans in the event of default.203 Almost all credit card debt is unsecured
and given the small value of most credit card purchases and the consumptive nature of
many of them (food, travel, and entertainment) it is simply not feasible for the credit to
take a security interest in most of the things purchased with credit cards. Moreover,
credit card issuers will have little ability to collect any judgment against the debtor,204
and at least prior to BAPCPA, credit card issuers collected little of their debt in
bankruptcy.205
Moreover, because credit cards provide a standing line of credit that can be drawn
upon at the debtor’s discretion, they present particularly difficult problems of moral
hazard and adverse selection. Credit card issuers issue credit to a consumer based on the
consumer’s financial condition at that time, and often a given consumer will be a very
good risk at the time the card and the line of credit are issued. But consumers are most
likely to draw on credit cards as a source of credit at precisely the moment where they
receive a financial setback that causes them to switch from being low-risk to high-risk
borrowers, such as job loss or an unexpected large liability. 206 This option for a debtor to
be able to switch his risk profile at his option raises problems of moral hazard and
adverse selection that appear to be more severe than for other types of credit.207
202
EVANS & SCHMALENSEE, supra note 33, at 225.
Canner & Luckett, supra note 184, at 657; EVANS & SCHMALENSEE, supra note 33, at p. xi.
204
Although the credit card lender can use the threat of impaired credit ratings as an ex ante device to
reduce the incentives for debtor misbehavior, threats to impair the credit rating of a borrower who has
already defaulted does not assist in the actual recovery of outstanding balances. Pozdena, supra note 183,
at 2.
205
Edith H. Jones & Todd J. Zywicki, It’s Time for Means-Testing, 1999 B.Y.U. L. REV. 177, 185-192
(summarizing studies).
206
Pozdena, supra note 183, at 2.
207
Mortgage or automobile lenders, for instance, use collateral to limit the threat of similar opportunism by
debtors.
203
64
Nevertheless, despite the historic stickiness of credit card interest rates, beginning
in the early-1990s credit cards in fact did become much more responsive to changes in
the cost of funds. In the first quarter of 1991 credit card interest rates hit their all-time
high of an average rate of 18.28%, but by the trough in the cost of funds rate in 2003
credit card interest rates fell to about 12%. The Federal Reserve further notes that in
1991 only 11 percent of responding banks reported interest rates below 16 percent on
their largest credit card plans, but by 2005 80 percent did so.208 Moreover, changes in
credit card interest rates have become much more closely correlated with changes in the
underlying cost of funds.209 Evans and Schmalensee observe that credit card interest
rates were about ten times more responsive to the cost of funds in the past decade than
during the 1980s.210
What happened during this period to make credit cards more responsive to
changes in general market conditions?
First, the initial wave of post-Marquette competition led to the elimination of
annual fees on credit cards and increased consumer benefits, rather than competition on
interest rates. Annual fees disappeared quickly and today few standard cards charge
them—ubiquitous before 1990 by 2002 only two percent of customers not enrolled in a
208
FED REPORT, supra note 6, at 7.
For the period 1972-2006, the correlation coefficient between interest rates on 30-year mortgages and
the cost of funds rate has been 0.865; the correlation coefficient between 48-month automobile loans and
the federal funds rate has been 0.866. The correlation coefficient during that same period for interest rates
on 24-month personal loans has been 0.727. By contrast, the correlation coefficient on credit cards loans
during that period has been only 0.633. Beginning with the final quarter of 1994 to the present, the
correlation coefficient on credit card interest rates and the cost of funds rate rose to 0.88, and for credit card
accounts actually assessed interest, the correlation coefficient was even slightly higher. See also Johnson,
Recent Developments, supra note 161, at p. 477 (noting that correlation between credit card interest rates
and the prime rate was only 0.09 during the 1980s and early 1990s but has risen to 0.90 from mid-1990s to
present).
210
EVANS & SCHMALENSEE, supra note 33, at 239-40.
209
65
rewards program paid an annual fee.211 Annual fees are an especially unpopular form of
credit card term pricing among consumers. A recent survey found that “no annual fee”
was the most important term on which consumers selected credit cards—higher than any
other single term, including the interest rate.212
The unpopularity of annual fees is
evidenced by the fact that when annual fees were first imposed in 1980, consumers
canceled over nine million bank cards that year, amounting to some 8% of the
outstanding total.213
Because issuers were competing on annual fees because of their
unpopularity with consumers, interest rate competition was minimal during this time, and
interest rates remained relatively sticky.
This led to a second wave of price innovation in credit cards—more aggressive
competition on interest rates. At the same time, credit card issuers began to develop
more sophisticated risk-assessment models that enabled them to more accurately offer
interest rates more closely-tailored to different risk tranches.
This heightened
competition on interest rates eventually led to a gradual introduction of adjustable rate
credit cards into the market, with the interest rate tied to a change in some underlying
interest rate. In 1989, variable rate credit cards accounted for only about three percent of
credit card accounts. By 1994 that share had grown to about 60 percent and by 2005
approximately 93 percent of cards had variable interest rates.214 Revolutionary changes
in communications technology also have reduced the processing costs associated with
credit card transactions.
Electronic processing of credit card receipts supplanted
211
JONATHAN M. ORSZAG & SUSAN H. MANNING, AN ECONOMICS ASSESSMENT OF REGULATING CREDIT
CARD FEES AND INTEREST RATES (Oct. 2007); Mark Furletti, Credit Card Pricing Developments and their
Disclosure, FED. RES. BANK OF PHILADELPHIA DISCUSSION PAPER (Jan. 2003).
212
EVANS & SCHMALENSEE, supra note 33, at 218; see also Durkin, supra note 100, at 203 (95% of survey
respondents report that amount of annual fee “very” or “somewhat important” compared to 91% for interest
rate).
213
MANDELL, supra note 8, at 78.
214
GAO REPORT, supra note 198, at 15; Johnson, Recent Developments, supra note 161, 477-78
66
traditional paper receipts during this period, dramatically reducing the processing costs of
transactions and loan servicing. Improvements in credit scoring and other individual risk
assessment models may also have tended to permit greater tailoring of interest rates to
actual risk, as rather than merely charging one interest rate to all borrowers it has become
increasingly easy to discriminate among different borrowers based on risk characteristics.
A third wave of credit card pricing innovation followed the adoption of variable
interest rates, namely an increase in behavior-based pricing terms such as increased use
of late fees, overdraft charges, and similar fees tailored to the borrower’s actual risk
profile rather than relying solely on interest rates. Interest rates are one term that can be
adjusted in light of borrower risk, but they are only an ex ante prediction of riskiness,
thus they are an imperfect price term. Risk-based fees, by contrast, are based on the ex
post level of observed behavior, thus ameliorating the extreme problems of adverse
selection and moral hazard previously discussed.215
The only systematic empirical study of these fees to date concludes that these fees
are risk-based and complement interest rates for efficient risk pricing.216 They find that a
one standard deviation increase in bankruptcy per capita leads to an increase in penalty
fees of $0.62 to $1.31. Similarly, a one standard deviation change in the chargeoff ratio
was found to change late fees in a range of $4.35 to $7.57. In addition, they find that a 1
basis point reduction in card interest rates will result in an increase in penalty fees of
between 0.88 and 4.11 cents. Thus, in their study, a one standard deviation in credit card
interest rates (273 basis points) was estimated to change late fees by $2.40. Moreover,
215
See Testimony of Professor Todd J. Zywicki before United States House of Representatives, Financial
Services Committee, Subcommittee on Financial Institutions and Consumer Credit, Hearing on “Credit
Card Practices: Current Consumer and Regulatory Issues,” April 26, 2007.
216
See Nadia Massoud, Anthony Saunders, and Barry Scholnick, The Cost of Being Late: The Case of
Credit Card Penalty Fees, working paper (January 2006); see also GAO REPORT, supra note 198, at 103.
67
they found no evidence that assessed penalties were larger for low-income borrowers,
after controlling for risk.
This use of more sophisticated and risk-based pricing techniques has made
possible more flexible interest rates as well. In addition, consumers who pay these fees
do not appear to be surprised by their existence, but generally are aware of them before
they enter into the transaction that triggers the fee.217 The spread between credit card
interest rates and the underlying cost of funds has narrowed over time as well, a trend that
has accelerated in recent years as the combination of more flexible interest rates and
greater use of risk-based behavioral fees has allowed credit card issuers to more
accurately price risk.218
These behavior-based fees appear to be motivated by efficient risk pricing and are
not simply a new form of consumer abuse. Return on assets for credit card banks have
been largely constant over the past two decades, suggesting that the tradeoff between
greater use of behavior-based fees and lower interest rates has not been driven by a desire
to “fool” consumers with hidden fees that result in economic profits in the banking
industry.219 Return on assets in the credit card industry have been stable even though
217
See Durkin, Credit Card Disclosures at p. A114.
See Zywicki, Testimony, supra note 215; see also Getter, supra note 184. Block-Lieb and Janger argue
that this narrowing of spreads during a period of rising bankruptcy filings evidences a failure of
competition in the credit card market, as spreads should increase as bankruptcies increase. See Susan
Block-Lieb & Edward J. Janger, The Myth of the Rational Borrower: Rationality, Behavioralism, and the
Misguided “Reform” of Bankruptcy Law, 84 TEX. L. REV. 1481 (2006). But this analysis misses the point
for several reasons. First, more sophisticated economic analysis has found a relationship between default
rates on credit cards and the interest rate spread. See supra note 201, and accompanying text. Second,
costs on other elements of credit card operations were declining, notably a drop in the cost of processing
transactions because of advances in information technology, leading to a reduction in costs at the same time
that bankruptcy losses were rising. Thus, Block-Lieb and Janger’s prediction would be correct only if all
other costs remained constant. Third, as noted, there has been a change in price terms as credit card lenders
have engaged in greater price discrimination such as imposing behavior-based fees on demonstrated riskier
borrowers rather than using the crude instrument of interest rates. Thus, Block-Lieb and Janger’s
observation is questionable both theoretically and empirically. See also Wright, supra note 172, at 485-88.
219
See Wright, supra note 172, at 476-77 (noting logical condition of a zero-profit equilibrium).
218
68
there has been a steady rise in the returns of other bank commercial and consumer
lending activities during this period.220 Issuers have fundamentally changed their pricing
models many times in the past 25 years: beginning with generally-assessed annual fees
and high fixed interest rates, then to more risk-based but fixed interest rates, and finally
to more flexible interest rates in combination with the use of behavior-based fees. Each
of these pricing innovations appears to be more efficient risk-based pricing than the mix
of terms they replaced.
D.
The Substitution Hypothesis of Credit Cards and Bankruptcy
The best available evidence thus suggests that the growth in credit card debt over
the past decades has been a substitution away from other types of consumer credit and
that this dynamic holds equally for lower-income households as well as others. This
result is tentative, of course, and the discussion in this article has relied primarily on
broad macroeconomic data. Nonetheless, it appears that the substitution theory has the
strongest explanatory power, both as an explanatory and empirical matter.
But to observe that the substitution effect best explains the growth of credit card
debt over time raises the subsequent question of how this translates into increased
bankruptcy filings. In general, the rise in consumer bankruptcy filings resulted from
three factors: changes in the economic costs and benefits of bankruptcy that provided
increased incentives to file bankruptcy, changes in social norms that reduced the social
220
See GAO REPORT, supra note 198, at 76; ORSZAG & MANNING, supra note 211, at 29. For a discussion
of the special difficulties in inferring credit card “profits” from the standard analysis of “return on assets”
used in the banking industry, which is an accounting measure of profit not an economic measure, see
Zywicki, The Economics of Credit Cards, supra note 48, at 128-46. Although the return on assets is not the
same as economic profits, it is being used to examine trends over time not the absolute level of “profits”
thus the constant return on assets in the consumer banking industry for credit card operations suggests an
absence of economic profits as a result of shifting price terms over time.
69
opprobrium or “stigma” associated with filing bankruptcy, and changes in the nature of
consumer credit toward more national and impersonal credit relations that increased the
willingness of borrowers to file bankruptcy.221 Credit card substitution independently
contributes to several of these three factors by explaining how increased use of credit
cards can generate increased bankruptcy filings even holding total household debt
burdens constant, so I will briefly focus on those particular factors here even though a
more general discussion of the causes of the rise in bankruptcy filings goes beyond the
scope of this article.
First, in contrast to credit cards, many of the traditional types of credit that were
replaced were secured debt or otherwise nondischargeable, such as pawn shops. Credit
card debt, by contrast, is unsecured thus consumers can file bankruptcy and keep the
goods and services purchased. Bankruptcy also will be of little use in dealing with
informal loans, such as from friends and family, as the obligation to repay those debts is
usually moral and extralegal rather than legal.
This substitution of dischargeable
unsecured debt for various forms of secured or de facto nondischargeable debt increases
the benefit of filing bankruptcy. Moreover, the decline in credit card interest rates
relative to other types of credit has permitted consumer to borrow (and thereby consume)
more with a given amount of income and wealth, thus consumers have larger amounts of
debt that they can discharge than in the past.
Second, developments in consumer credit markets have made the negative
consequences of filing bankruptcy may be less severe today than in the past.
Traditionally it was perceived that filing bankruptcy would cripple the ability to acquire
221
See Todd J. Zywicki, Institutions, Incentives, and Consumer Bankruptcy Reform, 62 WASHINGTON &
LEE L. REV. 1071 (2005).
70
new credit following bankruptcy. Today, however, there have been changes in credit
markets that have made credit more available to former bankruptcy filers. One survey
from the early 1990s found that over 16% of bankruptcy filers were able to gain
unsecured credit within one year after filing bankruptcy and over 55% within five
years.222 A survey from the late 1990s found that three-quarters of bankruptcy filers have
at least one credit card within a year after filing223 and a more recent study also finds that
most bankrupts can gain access to credit cards soon after filing bankruptcy.224
Bankruptcy filers are able to gain access to a broad cross-section of revolving credit as
well as installment lenders.225 Filing under chapter 7 for bankruptcy creates a multi-year
bar to refilling, making many bankruptcy filers lower-risk than those otherwise similarlysituated.226
This increase in competition not only reduces the negative effects of filing
bankruptcy but also makes it more difficult for creditors to “ostracize” those who file
bankruptcy.
Creditors face a collective action problem in withholding credit from
borrowers who file bankruptcy. Lenders may prefer as a group to ostracize borrowers
who file bankruptcy, but in practice each lender has an individual incentive to lend to
a debtor who files bankruptcy at the right price.227 Consistent with this collective
action problem, bankruptcy filers who reacquire credit are more likely to obtain credit
222
See MICHAEL STATEN, THE IMPACT OF POST-BANKRUPTCY CREDIT ON THE NUMBER OF PERSONAL
BANKRUPTCIES 10–11, (Credit Research Ctr., Krannert Graduate Sch. of Mgmt., Purdue Univ., Working
Paper No. 58, 1993) (arguing that for various reasons this estimate probably underestimated access to credit
at that time), available at http://www.msb.edu/prog/ crc/pdf/wp58.pdf.
223
VISA, CONSUMER BANKRUPTCY: ANNUAL BANKRUPTCY DEBTOR SURVEY (1997).
224
Katherine Porter, Bankrupt Profits: The Credit Industry’s Business Model for Postbankruptcy Lending,
Working Paper (2007).
225
STATEN, supra note 222, at 11–12.
226
BAPCPA extends the bar period for refilling from six years to eight years.
227
This problem is exacerbated when the existing group cannot restrain entry by new lenders who can enter
the market to serve those ostracized by incumbents. See RICHARD A. POSNER, ECONOMIC ANALYSIS OF
LAW § 8.5 (6th ed. 2003) (discussing the free-rider problem that can arise in the enforcement of social
norms).
71
from a new lender rather than a prebankruptcy lender. 228 Moreover, barriers to entry
are low in consumer credit markets, especially with the invention of nonbank finance
companies, and, in recent years, the greatest amount of entry appears to have occurred in
the subprime market, which specializes in lending to consumers with previous
bankruptcies and tarnished credit.229 Moreover, credit and retail sales traditionally were
tied together, such that a borrower who failed to pay his credit bills also would be unable
to purchase goods on credit in the future. The unhooking of credit from goods sales
means that consumers can be less concerned that filing bankruptcy will prevent them
from obtaining necessary goods and services.
Third, credit cards have made consumer credit relations less “personal” in nature,
decreasing the sense of trust and moral obligations associated with paying debts.
Consumer credit historically was a highly localized and personalized transaction, for
example, a local department store or Main Street tailor selling goods to their customers
on credit.230 Borrowers were often familiar with the owner of the lending store or bank in
a personal and social setting. Where the credit relationship is embedded in a context of a
social and economic relationship, it is more likely that a trust relationship will arise
between the parties that will bring with it a moral obligation.231
Credit card transactions, by contrast, are conducted with large national banks and
South Dakota and Delaware based credit card issuers such as Citibank and MBNA.232
228
Staten, supra note 222, at 12. Nor did it make a difference whether a debtor discharged his debts in
Chapter 7 or filed Chapter 13 and presumably attempted to repay some of his pre-petition debts. Id. at 16.
See also Porter, supra note 224.
229
Zywicki, Economics of Credit Cards, supra note 48, at 128–29.
230
See CALDER, supra note 11, at 37–42 (discussing the rise of retail credit).
231
Rafael Efrat, The Moral Appeal of Personal Bankruptcy, 20 WHITTIER L. REV. 141, 159 (1998).
232
See Jeremy Shearmur & Daniel B. Klein, Good Conduct in the Great Society: Adam Smith and the Role
of Reputation, in REPUTATION: STUDIES IN THE VOLUNTARY ELICITATION OF GOOD CONDUCT 29 (Daniel
B. Klein ed., 1997).
72
Impersonal credit relations, such as dealing with these institutional lenders, are less likely
to evolve into high-trust relationships, and these weaker extralegal constraints make
individuals more willing to breach those contractual promises.233 An individual is less
likely to feel himself bound in a trust relationship with his credit card issuer than he
would be if he purchased a suit on store credit from his local tailor, where he knows the
owner and employees personally and thus will be more aware of the hardship imposed by
a breach of trust by failure to pay his debt.234 Low-trust relationships such as between
modern credit card issuers and consumers therefore are more prone to opportunism than
high-trust relationships. Bankruptcy filers acknowledge that although they generally feel
guilty about not being able to pay their debt to their creditors, “many had less guilt
feelings about not paying large creditors or creditors they did not know personally.”235
Indeed, as David Skeel observes, part of the impetus for the 1898 Bankruptcy Act was
the concern of merchants who engaged in interstate commerce that when debtors ran into
financial trouble they “played favorites” with their creditors, preferring “family members
and local creditors, not the out-of-state merchants.”236
There is also a widely-shared impression that there has been a change in social
norms regarding bankruptcy such that bankruptcy has become a less shameful act,
which also would reduce the social sanction and disincentive for filing. 237
233
For
Efrat, supra note 231, at 159.
Efrat, supra note 231, at 159. This same analysis could apply to the development of trust relationships
by creditors, but the primary constraint on lender opportunism are contracts and other more formal
institutions, including legislation and regulation, as well as repeat-dealing and reputation effects, thus trust
seems much less relevant on the lender’s side of the transaction.
235
Rafael Efrat, The Evolution of Bankruptcy Stigma, 7 THEORETICAL INQUIRIES IN LAW 365 (2006) (citing
VISA, QUALITATIVE RESEARCH: BANKRUPTCY PROCESS 9 (1997)).
236
DAVID A. SKEEL, JR., DEBT’S DOMINION: A HISTORY OF BANKRUPTCY LAW IN AMERICA 36 (2001).
237
See Rafael Efrat, Bankruptcy Stigma: Plausible Causes for Shifting Norms, 22 EMORY BANKRUPTCY
DEVELOPMENTS J. 481, 485-89 (2006). But see Teresa A. Sullivan, Elizabeth Warren, & Jay Lawrence
Westbrook, Less Stigma or More Financial Distress: An Empirical Analysis of the Extraordinary Increase
234
73
instance, a 1960s study found almost 25% of petitioners felt ashamed about filing
bankruptcy,238 a figure that fell to 10% in a 1995 survey 239. This decline in stigma
regarding bankruptcy is consistent with broader social trends since the 1960s that
have tended to shift blame for individual failure away from personal responsibility to
environmental factors 240 as well as a variety of other bankruptcy-specific factors, such
as the decision in the 1978 Code to replace the traditional pejorative term “bankrupt”
with the more neutral term “debtor... as a means of reducing the stigma connected with
the term bankrupt.”241
Bankruptcy attorneys may also have played a role in changing norms about
bankruptcy, especially with respect to credit cards. In general, debtors’ attorneys seem to
be somewhat more hostile toward creditors than are their clients and are especially
dismissive of the belief that there is a moral, as opposed to purely legal, obligation to
repay creditors, especially institutional lenders such as credit card issuers. One lawyer
observed, “Some people feel there is a moral issue; frankly I don’t.”242 Another lawyer
stated, “My attitude is—the law is there. The credit card companies charge 20% interest.
Discharge is a risk of doing business. I don’t feel bad about it. Some debtors feel so
harassed. Some debtors say they feel bad about discharging debt, and I wonder if they
do. Some are overly emotional, and I’m thinking, ‘What’s the big deal?’ Especially with
in Bankruptcy Filings, 59 STANFORD L. REV. 213 (2006) (concluding that there has been no measurable
decline in social stigma associated with bankruptcy), compare TERESA A. SULLIVAN, ELIZABETH WARREN
& JAY WESTBROOK, AS WE FORGIVE OUR DEBTORS: BANKRUPTCY & CONSUMER CREDIT IN AMERICA 337
(1989) (suggesting that a decline in social stigma might explain rising bankruptcy filing rates).
238
See DAVID T. STANLEY & MARJORIE GIRTH, BANKRUPTCY: PROBLEM, PROCESS, REFORM 67-68 (1971).
239
Tahira K. Hire & Kyle L. Kostelecky, Pilot Study of Consumer Debtors Provides New Insight—What
Influences Debtors’ Attitudes? 14 AM. BANKR. INST. J., Apr. 1995, at 28, 36, available at 14-APR
AMBKRIJ1 (Westlaw).
240
See Efrat, supra note 235, at 490.
241
S. REP. NO. 95-589, at 23 (1978), as reprinted in 1978 U.S.C.C.A.N. 5787, 5809.
242
Jean Braucher, Lawyers and Consumer Bankruptcy: One Code, Many Cultures, 67 AM. BANKR. L.J.
501, 523 (1993).
74
credit cards—it’s not like a friend or a relative.”243 Many attorneys persuade clients that
the sacrifices required to repay debts will impose an undue hardship on their families, and
thus discharge of debts is morally appropriate.244
IV.
The Distress Model
There is little doubt that some consumers overborrow on credit cards and as a
result are forced into filing bankruptcy, just as some members of earlier generations
overused installment credit or pawnbrokers. In this sense, there is nothing inherently new
about credit cards—they are simply another type of consumer credit that the
overwhelming number of consumers use responsibly and to improve their lives, but some
percentage of consumers overuse. For the distress model to be able to explain the rise in
bankruptcy filings over the past 25 years, therefore, it must be demonstrated that greater
use of and more widespread access to credit cards has created some new systematic
tendency for consumer overindebtedness relative to prior eras. But, as noted, increased
access to and use of credit cards is not reflected in an increase in household debt service
burdens and thus the factual predicate seems to be absent.
In general, expositors of the distress theory of credit cards have failed to address
two important threshold issues. First, they generally have been silent or unaware of any
substitution effect, thus they have provided no explanation for why credit cards are a
unique contributor to consumer financial distress. White cites the dramatic rise in credit
card use over the past two decades245 as does Sullivan, et al.246 but neither mentions the
243
Id.
Id. at 509.
245
White, supra note 2.
244
75
offsetting decline in installment debt. Moss and Johnson simply assert that the rise in
consumer bankruptcies has been driven by increased credit card lending to lower-income
and riskier borrowers, but do not consider the substitution effect.247
A second problem with prior studies is that they generally assert—but do not
demonstrate—that any correlation between credit card debt (or credit card defaults) and
bankruptcy is caused by increased financial distress. Ausubel, for instance, finds a
correlation between credit card defaults and bankruptcy and assumes that the credit card
defaults cause the subsequent bankruptcy, or alternatively that the credit card defaults
evidence underlying financial distress caused by excessive credit card debt that then
causes the subsequent bankruptcy.248 A study of trends in credit cards and bankruptcy in
Canada similarly asserts the direction of causation and thus suffers from similar flaws.249
Both the substitution model and the distress model are consistent with the
observed correlation between credit card debt and bankruptcy. Given that the direction of
causation is indeterminate as an a priori matter, the causal relationship between credit
cards and bankruptcies must be demonstrated, not merely assumed. Establishing a causal
Sullivan, et al., observe that “credit card usage has grown fastest in recent years among debtors with the
lowest incomes,” but do not make any mention of whether this increased credit card usage has resulted
from a decrease in the usage of other types of consumer credit. See SULLIVAN, ET AL., FRAGILE, supra note
2, at 136. They do not ask, for instance, whether retailers or personal finance companies were previously
servicing these consumers. As noted above their own data strongly indicates the presence of a substitution
effect from other types of consumer credit. See supra at Figure 4.
247
David A. Moss & Gibbs A. Johnson, The Rise of Consumer Bankruptcy: Evolution, Revolution, or
Both?, 73 AM. BANKR. L.J. 311, 314 (1999). Moss and Johnson acknowledge that bankruptcies have
increased more rapidly than consumer indebtedness in recent years and simply assume—but do not
demonstrate—that this acceleration in the so-called multiplier effect between debt and bankruptcy is caused
by an increase in credit card lending to higher-risk borrowers. They offer no support for this assertion. The
finding of an increased multiplier effect is consistent with the substitution hypothesis.
248
Ausubel, supra note 111. Ausubel also notes that credit card defaults are closely correlated with the
business cycle. Ausubel, supra note 111, at 253. But there is actually an inverse relationship between
credit card borrowing and unemployment rates—consumers are actually less likely to revolve credit card
balances when they suffer unemployment, thus the likelihood of default as a result of unemployment is not
preceded by an increase in indebtedness. See Beales & Plache, supra note 193.
249
Dianne Ellis, The Effect of Consumer Interest Rate Deregulation on Credit Card Volumes, Charge-Offs,
and the Personal Bankruptcy Rate, BANK TRENDS (Pub. Info. Ctr,, Fed. Deposit Ins. Corp., Washington,
D.C.), Mar. 1998, at 1, available at http://www.fdic.gov/bank/analytical/bank/bt_9805.html.
246
76
link is difficult because of the endogenous nature of the relationship between credit cards
and bankruptcy. Although higher levels of credit card debt may cause higher levels of
bankruptcy filings, more liberal bankruptcy laws increase the willingness of debtors to
borrow, and to borrow on unsecured debt such as credit cards in particular.250 The
relative ease with which a debtor anticipates that she can discharge debt ex post in part
causes the level of borrowing ex ante. To establish the distress model it is necessary to
control for this problem of endogeneity in determining the nature of the causal
relationship between credit card debt and bankruptcy.
The only scholar to acknowledge and try to address the problems of endogeneity
and the substitution effect is Professor Ronald Mann. Mann claims to find a strong
causal relationship between credit card debt and personal bankruptcy at the
macroeconomic level, thereby supposedly demonstrating the validity of the distress
model.251 Mann collected annual cross-sectional data on credit card debt, credit card
spending, consumer debt, and consumer bankruptcy for 12 or 13 years for five
countries—Australia, Canada, Japan, United Kingdom, and the United States. He then
ran ordinary least squares regressions to test various relationships among the data. Mann
concludes that his statistical analysis demonstrates the validity of the distress model of
credit cards. In fact, however, there are several major problems with Mann’s statistical
analysis as well as his interpretation of his own findings. Although he claims to have
established a causal relationship, and not merely correlation, he has failed to do so.
And, of course, a “softer” bankruptcy regime makes lenders less likely to extend unsecured credit, so the
overall equilibrium level of borrowing is ambiguous as an a priori matter. In this sense, the tradeoff
involving the rigor of the bankruptcy regime is identical to that regarding contract enforcement generally,
with the only difference being that contract rules govern the enforceability of contracts ex ante whereas
bankruptcy laws govern the option to nullify a contract ex post. See Charles Goetz & Robert E. Scott,
Enforcing Promises: An Examination of the Basis of Contract, 89 YALE L.J. 1261 (1980).
251
MANN, supra note 2.
250
77
Indeed, to the extent that any supportable conclusions can be drawn from his data set and
statistical analysis, his findings actually tend to support the substitution model rather than
the distress model.
Mann acknowledges the endogeneity problem and his study is specifically
designed to try to address it. Mann’s proffered solution is to test the relationship between
credit card debt (CCD) lagged by one, two, or three years, and bankruptcy filings (BKR).
When Mann performs these regressions, he finds a statistically significant relationship
between credit card debt and bankruptcies, both in the same period, but also when lagged
for each of the three periods.252 Mann asserts that lagging the CCD variable solves the
endogeneity problem, thereby enabling him to draw inferences about the causal direction
of the relationship between CCD and bankruptcies.
But lagging the observations on CCD in this data set does not solve the problem
of endogeneity in his data.253 To understand why not, consider the following exercise.
To test whether lagging CCD establishes a determinate relationship on the direction of
causation to BKR, I used Mann’s same data set and simply reversed the hypothesized
statistical relationship, specifying BKR as the independent variable and CCD as the
dependent variable lagged by one, two, and three years. If it is true that regressing lagged
CCD against bankruptcies solves the endogeneity problem as a statistical matter, then it
should not be possible to reverse the equation and get meaningful results. The results are
presented in Table 1:
Table 1: Bankruptcy and Credit Card Debt
252
MANN, supra note 2, at 70, Table 5.1.
Several of the theoretical and empirical problems with Mann’s analysis are discussed in Wright, supra
note 172, at 489-91. For instance, Wright notes that the data itself is merely a scatterplot of 65
observations—as if each point is a separate country—and is neither a panel regression of country-level data
nor a country-specific time series. Id. So it is not clear what it means to impose a “lag” with this data.
253
78
Coeff
CONSTANT
N
R2
*** p<0.001
No LAG
0.492***
(.024)
-(191.210)
62
.87
LAG1
0.513***
(.025)
-(182.750)
59
.87
LAG2
0.533***
(.027)
-(172.42)
54
.87
LAG3
0.557***
(.029)
-(162.96)
49
.87
As can be seen in Table 1 regressing bankruptcies (lagged 1, 2, and 3 years) on credit
card debt consistently produces strongly statistically significant results. Moreover, the rsquared values for these regressions are virtually identical to the r-squared values in
Mann’s reported regressions. In other words, if it is true that simply “lagging” the
independent variable (in this case bankruptcy filings) solves the endogeneity problem as a
statistical matter, then Table 1 demonstrates that bankruptcy filings “cause” subsequent
credit card indebtedness one, two, and three years later with an increasing causal effect as
the lag increases.254
The explanation for this incongruous finding is obvious—
contemporaneous and lagged credit card debt are collinear (as are contemporaneous and
lagged bankruptcies), thus adding a lag does not resolve the problem.255 I should stress
that Table 1 should not be read to suggest that an increase in bankruptcies “causes”
increases in credit card debt one, two, or three years later, but rather to demonstrate the
statistical point that the endogeneity problem in the credit card-bankruptcy relationship
by asserting CCD to be independent and attaching lags to assumed independent variables.
As a statistical matter, both Professor Mann’s results and those in Table 1 are equally
valid—or invalid—in establishing the direction of causation.256
254
This is how Professor Mann interprets his analogous findings. See MANN, supra note 2, at 64-65.
The correlation coefficient among CCD, d1CCD, d2CCD, and d3CCD with one another is 0.99
256
Professor Mann suggests that the endogeneity issue can be resolved as a matter of a priori logic. See
MANN, supra note 2, at 68. But it is precisely because a priori reasoning cannot resolve the endogeniety
problem that it must be addressed statistically, not logically.
255
79
Professor Mann’s approach illustrates the well-known post hoc ergo propter hoc
fallacy in economics.257 Simply because one observation in a data series comes prior to a
later observation does not mean that the earlier observation causes the later. Where
expectations about future events are incorporated into today’s actions—a concept known
as “rational expectations” in economics—merely lagging the variable that is asserted to
be independent will not solve the endogeneity problem because it actually is not
independent.258 So if borrowers incorporate their expected willingness to file bankruptcy
in the future into their borrowing decisions today, the level of debt will still be
endogenous to the subsequent bankruptcy decision.
Consider an example to illustrate the fallacy. BAPCPA was signed into law in
April 2005 but most of its provisions did not become effective until October 17, 2005. In
the two weeks preceding BAPCPA’s effective date, roughly 500,000 Americans filed
bankruptcy, including 350,000 in the week before the effective date alone, over 10 times
the typical weekly filing rate. If one were simply to lag bankruptcy filings by one or two
weeks preceding BAPCPA then one would infer that this spike in bankruptcy filings
“caused” BAPCPA, when of course, causation ran in the exact opposite direction—
bankruptcy filings rose because it was anticipated that BAPCPA was about to become
257
James Tobin, Money and Income: Post Hoc Ergo Propter Hoc?, 84 Q. J. ECON. 301, 302–03 (1970).
The phrase literally translates as “after this, therefore because of this.” It “assumes or asserts that if one
event happens after another, then the first must be the cause of the second.” As one commenter observes, it
“is a particularly tempting error because temporal sequence appears to be integral to causality. The fallacy
lies in coming to a conclusion based solely on the order of events, rather than taking into account other
factors that might rule out the connection.”
Post hoc ergo propter hoc, WIKIPEDIA,
http://en.wikipedia.org/wiki/Post_hoc_ergo_propter_hoc (visited August 21, 2007).
258
See Thomas J. Sargent, Rational Expectations, in THE CONCISE ENCYCLOPEDIA OF ECONOMICS,
available in http://www.econlib.org/LIBRARY/Enc/RationalExpectations.html.
80
effective. Thus, even though BAPCPA temporally came subsequent to the filing wave, it
seems quite clear that BAPCPA nonetheless caused the prior filing spike.259
In addition to these methodological issues there are problems with Mann’s
interpretation of his results in that there are almost certainly relevant variables omitted
from his regression and the failure to account for them causes spurious statistical results.
In particular, both credit card debt and bankruptcy may be correlated with one another or
are both causally endogenous to some other variable or variables. The presence of
omitted variables might explain, for instance, why the regressions can be specified with
either bankruptcies or credit card debt as the dependent variable and still generate
statistically significant results. Further evidence of the probable presence of omitted
variable bias is the very large coefficient values on the constants in his regressions and
the unusually high r-squared value generated in these extremely simple linear regressions
with only 13 observations from five different countries with such variable borrowing and
bankruptcy patterns.260
An examination of the data reveals that one source of omitted variable bias is a
simple upward trend line that may be a proxy for some omitted variable that is correlated
with both bankruptcies and the various consumer credit variables, such as income.261 To
consider the possibility of an omitted trend variable, using Mann’s data I ran regressions
using YEAR as the independent variable, first with levels of credit card debt (CCD) then
with bankruptcy levels (BKR) as the dependent variable. In both regressions YEAR
259
Nor is this just randomness or irrational panic by consumers One study of the effects of BAPCPA
observed that no similar filing wave preceded the enactment of the 1978 bankruptcy code, suggesting that
consumers rationally (and correctly) perceived that the 1978 law would make it easier to file bankruptcy.
See Ashcraft, et al., supra note 201.
260
See Wright, supra note 172, at 490.
261
Japan appears somewhat anomalous compared to the other countries in Mann’s data set as both
consumer credit and credit card debt rise over the first half of the period under examination and then fall
steadily afterwards, yet bankruptcies rise throughout the entire period.
81
turned out to be statistically significant at the 95% level of confidence, suggesting the
possible presence of some exogenous trend variable that has caused both CCD and BKR
to increase over time.
A second possible source of omitted variable bias is the presence of countryspecific factors that may be causing particular relationships between CCD and BKR, such
as differences in the countries’ respective bankruptcy regimes, consumer credit policies,
or cultural attitudes regarding debt and bankruptcy. To test for this using Mann’s data, I
ran a regression with only the country dummy variables as the independent variable and
BKR as the dependent variable. Each country dummy variable was extremely significant
at the 0.000 level with an r-squared value of 0.8786. In short, a regression that uses only
country dummy variables and excludes all of Mann’s consumer credit variables produces
a high degree of statistical significance and an r-squared value virtually identical to the rsquared value he found in his regression of CCD on BKR.
I then combined these two possible sources of omitted variable bias into a
regression that specified just these two variables—YEAR and the country dummy
variables—as the independent variables and BKR as the dependent variable (and
excluded all consumer credit variables). Under this specification, all of the country
dummies remain significant at the 0.000 level, YEAR is borderline significant at the
0.059 level, and the r-squared value is 0.9382. In short, accounting for only the upward
trend lines in the data and country dummy variables, and omitting every consumer credit
variable from the regression, the r-squared value is substantially higher than Mann finds
for his linear regressions on CCD and approximates that of his most extensive regression.
I then added the country GDP variable (included in Mann’s data) to the regression and
82
ran it again omitting all consumer credit variables. The r-squared value was 0.9425. The
country dummy variables all remained statistically significant at a 0.000 level of
significance, YEAR was equally significant, and GDP was statistically significant at the
.01 level of significance.
I took the simple model with a time trend (YEAR) and country dummy variables
and added in CCD. Under this specification of the model, the r-squared value is 0.9503,
but all of the independent variables are statistically insignificant, including CCD. I then
added to this same regression GDP—regressing YEAR, country dummies, CCD, and
GDP as independent variables and BKR as the dependent variable—and somewhat
surprisingly, GDP returned a statistically significant value (at the .01 level) although the
other variables—including CCD—remained statistically insignificant.262 In other words,
after controlling for a trend line and country differences, CCD becomes statistically
insignificant when added to the regression.
Again, I wish to stress that this discussion should not be interpreted to suggest
that it disproves the hypothesis that increases in credit card debt leads to increases in
bankruptcy filings or proves that country-specific or GDP variables predict changes in
262
Mann does not report this particular regression in his book. In Table 5.5, he reports a similar regression
with all three consumer credit variables—CCD, level of credit card spending (CCS), and level of consumer
debt (COND). He finds borderline significance for CCD in that regression. This result seems to be an
artifact of a curious and unexplained relationship between CCS and CCD in his data. When CCS and CCD
are run separately, they consistently generate positive and significant results on bankruptcies. When they
are run together in the same regression, the sign on CCS switches to negative and borderline significance
and CCD remains positive and significant. As noted, running CCD without CCS in the regressions with
country dummies and macroeconomic variables results in CCD being statistically insignificant. It is only
when CCS is included in the regression, with its peculiar negative coefficient, that CCD becomes
statistically significant. There may be multicollinearity in the data between those two variables that is
creating a spurious significance for CCD when they are run together, e..g, the negative coefficient on CCS
increases the positive coefficient on CCD. Mann reports variance inflation factor (VIF) values with mean
values of approximately 6 and an examination of the data indicates maximum VIF values in a similar
range. Statistics books indicate that a “rule of thumb” is that multicollinearity is present when the mean
VIF value exceeds unity or the maximum VIF exceeds 10. See CHRISTOPHER F. BAUM, AN INTRODUCTION
TO MODERN ECONOMETRICS USING STATA 85 (2006).
83
bankruptcy filing rates, but to recognize that the data set and simple linear regression
techniques that Professor Mann uses simply cannot establish the statistical relationships
that he is seeking.263 Mann’s conclusion that his empirical analysis demonstrates a
determinate causal relationship between credit card debt and bankruptcy appears to be a a
spurious result caused by one or more omitted variables.
But in addition to these statistical problems there is also a logical difficulty in
Mann’s interpretation of his findings. Mann finds that “[even if] credit card spending and
consumer debt are held constant, an increase in credit card debt—a shift of consumer
borrowing from noncard borrowing to card borrowing—is associated with an increase in
bankruptcy filings.”264 From which he concludes, “[T]he data… bear out the idea that
credit cards are unique contributors to the overindebtedness problem, an idea that is
inconsistent with the view that credit cards merely substitute for other less efficient forms
of consumer lending. The data indicate that credit card debt correlates with subsequent
increases in consumer bankruptcy, even when overall borrowing is held constant.”265 But
this conclusion simply misinterprets the statistical findings. If consumer debt is being
held constant, then by definition increased indebtedness cannot explain the relationship
between credit cards and bankruptcy. In fact, in some of the more extensive forms of his
regressions Mann finds no statistically significant correlation between credit card debt
and consumer indebtedness, but only between credit cards and bankruptcy directly—a
finding that, if true, actually supports the substitution theory not the distress theory.266
263
See Wright, supra note 172, at 491. One conventional solution to the simultaneity problems here would
be to use an instrumental variable estimator instead of simple linear techniques. See id. at 490 n.64.
264
MANN, supra note 2, at 66 & 71 Table 5.3.
265
MANN, supra note 2, at 182 (emphasis added).
266
MANN, supra note 2, at 57 Table 4.3. Again, it should be cautioned that both the limitations of the data
set itself as well as the crude linear regression techniques that are used suggest extreme caution in
interpreting the results. Indeed, this caveat is especially appropriate here, as at one point Mann finds the
84
Despite the widespread acceptance by legal scholars of the distress model of the
relationship between credit cards and bankruptcy, there is no solid evidence to support
the purported causal relationship.
Most adherents have failed to address or even
acknowledge the substitution effect or the problem of endogeneity and the one effort
addressed to those problems fails to deal with them effectively, and the only effort that
does actually ends up supporting the substitution theory to the extent that it establishes
statistically valid results at all.
V.
Strategic Model
A third hypothesis to explain the observed correlation between credit cards and
bankruptcy is the strategic model or the “loading up” hypothesis.267 The strategic model
is essentially the inverse of the distress model—rather than excessive credit card debt
causing a subsequent rise in bankruptcy filings the strategic model argues that credit card
debt rises in anticipation of an anticipated bankruptcy filing.
Strategic behavior involving credit cards may manifest itself in three ways. First,
debtors will have an incentive to “load up” their credit card on the eve of bankruptcy,
especially by purchasing goods and services that can be consumed or retained in
coefficient on the relationship between lagged credit card spending and bankruptcies to be negative (and
significant at the 0.10 level on a two-year lag) which would suggest that an increase in credit card spending
causes a reduction in bankruptcy filings two years later. Professor Mann avoids this absurd conclusion by
quite sensibly noting that this finding probably reflects consumer expectations about economic conditions
that lead them to spend more when they are optimistic about their economic situation. MANN, supra note 2,
at 67. But applying a rational expectations interpretation to explain this particular anomaly undercuts his
entire methodological approach of assuming that a lag cures the endogeneity problem without reference to
expectations.
267
See Carol C. Bertaut & Michael Haliassos, Credit Cards: Fact and Theories, in THE ECONOMICS OF
CONSUMER CREDIT (Giuseppe Bertola, Richard Disney, and Charles Grant eds., Forthcoming 2007)
(surveying theoretical and empirical literature).
85
bankruptcy.268 Credit cards are an open line of credit that can be drawn against at the
borrower’s discretion and that generally can be discharged in bankruptcy. By contrast,
some unsecured debts are not dischargeable in bankruptcy,269 and secured debts, such as
home and auto loans are minimally affected. Second, a debtor who is unable to pay all
of his bills can consciously choose which of his many debts to pay in the pre-bankruptcy
period, and specifically can choose to pay secured or nondischargeable debts while
electing to allow dischargeable debts, such as credit card debt, to go unpaid. Third, a
debtor can increase the value of her exempt assets (such as increasing exempt equity in
her home) rather than pay outstanding credit card bills.270 Regardless of the scenario, a
core prediction of the strategic model is that an increase in credit card debt will occur in
the pre-bankruptcy period.
Some behavior involving credit cards and bankruptcy can be explained by
strategic behavior.
Gross and Souleles, for example, find that in the year before
bankruptcy borrowers significantly increase the use of their credit cards versus earlier
periods, running up their balances rapidly in the period leading up to bankruptcy. 271 This
finding is inconsistent with those who identify credit cards as presenting a special
problem for consumers because of the gradual, subconscious, and “insidious” manner in
11 U.S.C. §523(a)(3) creates a presumption of nondischargeability incurred for the purchase of “luxury
goods and services” above a certain threshold within a certain number of days preceding a bankruptcy
filing. Debts incurred for non-luxury goods and services are unaffected, of course, as are even those
charges incurred for luxury goods outside of the 90 day window (70 days for cash advances). In one
notable example a debtor charged several discretionary car repairs and purchase new tires a few weeks
before filing bankruptcy then discharged the credit card debts in bankruptcy. See Manuel Perez-Rivas &
Martin Weil, Massie Filed for Bankruptcy; Montgomery Schools Halt Consideration of Finalist, WASH.
POST, May 4, 1999, at A1; Bernard Dagenais, Bankruptcy: Not Quite a Free Ride, WASH. TIMES, May 10,
1999, at D3. Although the timing of the repairs plainly was strategic it was nonetheless dischargeable.
269
See 11 U.S.C. §523(a). This includes debts for things such as taxes, alimony and child support,
fraudulently-induced debts, student loans, and others.
270
See ANDREAS LEHNERT & DEAN M. MAKI, CONSUMPTION, DEBT, AND PORTFOLIO CHOICE: TESTING THE
EFFECT OF BANKRUPTCY LAW, at 33 (Fed. Res. Bd., Working Paper No. 2002-14, 2002). Conversion of
nonexempt assets to exempt assets is generally permissible under the Bankruptcy Code.
271
Gross & Souleles, supra note 112, at 338.
268
86
which credit card debt purportedly accumulates over time.272 That the rise in credit card
debt is rapid and concentrated in the period immediately preceding bankruptcy suggests
that credit card indebtedness does not cause bankruptcy in some cases, but that the debtor
is already on the way toward bankruptcy when the credit card borrowing begins, and is
either acting strategically or is tapping her credit line of last resort.273
Andreas Lehnert and Dean Maki also find some evidence to support the strategic
model, concluding that generous bankruptcy exemptions tend to lead to increased statelevel bankruptcy filing rates but also to a greater tendency of households to
simultaneously hold more low-return liquid assets and to owe high-cost unsecured debt,
such as credit cards.274 This suggests that these households could pay down some of their
credit card debt but choose not to and, absent the ability to file bankruptcy it would be
rational to use some of their low-return savings to reduce their high-cost debt obligations.
With bankruptcy as an option, however, consumers can retain their assets through
exemptions while discharging their unsecured debt, thereby providing an incentive to
make the strategic decision to simultaneously carry both more assets and more unsecured
debt, as the strategic model would predict. In fact, households in states with higher
exemptions are more likely to simultaneously hold liquid assets and higher unsecured
debt. Paula Lopes similarly finds that some consumers borrow strategically with the
272
See supra note 3 and accompanying text.
The conclusion that the debtor is acting strategically is, of course, subject to an endogeneity critique as
well, as the debtor may be taking on credit card debt to try to avoid filing bankruptcy. If this is the case,
then the relationship between credit card debt and bankruptcy is spurious, in that it suggests that the debtor
was heading toward bankruptcy as a result of other factors and that the ability to borrow on credit cards
simply temporarily postponed the inevitable bankruptcy filing but did not cause it.
274
Andreas Lehnert and Dean M. Maki, Consumption, Debt and Portfolio Choice: Testing the Effect of
Bankruptcy Law, working paper (Feb. 20, 2002).
273
87
intention of defaulting in the near future and that their likelihood of strategic borrowing
depends on the relative generosity of exemptions.275
Ning Zhu also finds support for the strategic model of credit card use.276 He finds
that bankruptcy filers exhibit consumption patterns similar to those who do not file
bankruptcy, but that the pattern of credit use for bankruptcy filers differs from that of
non-bankruptcy filers. Twenty-six percent of the debt reported by bankrupt households is
credit card debt as compared to 15% for non-bankrupt households. By contrast, only 5%
of the credit for bankrupt households is bank credit versus 21% for non-bankrupt
households.277 This suggests that even though there has been a general substitution of
credit card credit for bank credit among American households, this substitution has been
greater for those households that file bankruptcy than for others.278 This may reflect a
strategic decision on the part of bankrupt households to make greater use of credit card
credit in anticipation of bankruptcy. Moreover, Zhu finds that the median mortgage debt
for bankrupt and non-bankrupt households is similar as are their median house values.
Some general data also support the strategic model. For instance, although credit
card defaults have risen in tandem with bankruptcy filings,279 defaults on secured home
and auto loans remained steady during most of this period.280 In fact, over the past 25
years, the general trend on chargeoffs on credit card loans has been upward over time; by
contrast, the charge-off rate on residential real estate (mortgages and home equity lines of
275
Paula Lopes, Credit Card Debt and Default over the Life Cycle, Working Paper, Financial Markets
Group, London School of Economics (2003).
276
Ning Zhu, Household Consumption and Personal Bankruptcy, working paper (Feb. 2007).
277
Id. at 16.
278
See also SULLIVAN, ET AL., FRAGILE, supra note 2, at 120. Zhu interprets these findings as supporting
the strategic model, but they also could be consistent with the substitution model if it is simply the result of
decisions to substitute credit card debt for other debt in light of an absence of attractive alternatives.
279
See Ausubel, supra note 111, at 249.
280
See Durkin, Discussion, supra note 132, at 38-39.
88
credit) have generally trended downward, even with the recent rise in chargeoffs in the
residential real estate market, as shown in Figure 10:
Figure 10
9
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
8
Credit Cards
7
6
5
4
3
2
1
Credit Card
Mortgage
20
07
Q1
20
05
Q1
20
03
Q1
20
01
Q1
19
99
Q1
19
97
Q1
19
95
Q1
19
93
Q1
19
91
Q1
0
Mortgages
Chargeoffs Trends: Credit Cards v. Mortgages
Year
Source: Federal Reserve.
Figure 10 identifies trends in charge-off rates for residential real estate on one
hand and credit card loans on the other.281 As is seen, until the mid-1990s, the charge-off
rates on the two types of loans tended to move in tandem. During the period of rapidly
rising bankruptcies in the mid to late-1990s, however, the charge-off rate on credit card
loans rose dramatically, even as the charge-off rate on residential real estate loans
declined.
If households were filing bankruptcy because of economic distress they
presumably would be unable to pay both credit card loans and mortgages, and thus
charge-off rates should move in tandem with one another.
281
Linear trendlines have been added for clarity.
89
Instead, the trends are
observed to be moving in opposite directions, suggesting that consumers have become
more likely to default on credit card loans than on residential real estate loans.
One plausible explanation for these divergent trends is that, when forced to
choose, stressed consumers have consciously chosen to pay their secured mortgages
rather than their dischargeable unsecured credit card debts in the pre-bankruptcy period.
Moreover, if rising credit card charge-offs reflected an increase in financial distress, then
charge-offs would be expected to be correlated with a rise in delinquency rates on credit
card loans. But as indicated in Figure 6 above, the delinquency rate on credit cards has
shown no similar upward trend over time, and has actually trended slightly downward
since 1991, suggesting that the rise in credit card charge-offs have not been the result of
an increasing number of consumers behind in their payments or otherwise struggling to
pay their credit card bills.
Like
mortgages,
default
rates
on
student
loans,
which
are
largely
nondischargeable under the Bankruptcy Code, have also declined dramatically over time
from a peak of over 22% in 1990 to 5.1% in 2004282 even though there has been a
dramatic rise in outstanding student loan debt during this period.283 The declining default
rate on this nondichargeable debt obligation, compared to the rise in credit card
chargeoffs, is also consistent with the strategic model.284
282
See United States Department of Education, Official Cohort Default Rate for Schools,
http://www.ed.gov/offices/OSFAP/defaultmanagement/defaultrates.html.
283
See supra notes 146-147 and accompanying text.
284
Mann purports to disprove the strategic model in his statistical analysis, finding that the coefficient
value on credit card debt (CCD) as a predictor of bankruptcy filings increases as he extends the lag period
on his regression. MANN, supra note 2. But this conclusion is not supportable from his finding. He does
not establish that the coefficient values on his lags differ in a statistically significant manner from one
another, meaning that the data do not disprove the hypothesis that they are statistically equivalent to one
another. In fact, a series of pairwise tests of the various coefficient values finds the coefficient value on
each LAG term to be within the confidence interval for surrounding coefficients, raising doubt about
whether they could be shown to be statistically different from each other. Testing Mann’s conclusion is
90
Overall, this all provides some support for the strategic hypothesis.
More
generally, it seems clear that consumers have become more sophisticated over time at
pre-bankruptcy planning and deciding which debts to pay and which ones on which to
default, as evidenced by the rising charge-off rates on credit card loans while mortgages
and student loans default rates have fallen. Moreover, credit card charge-offs have risen
over time, even though credit card delinquencies have not. Although this suggests that
strategic credit card use explains some of the relationship between credit card debt and
bankruptcy it seems obvious that some consumers are forced into bankruptcy by
excessive credit card debt thus it still seems unlikely that the strategic model can explain
the full relationship.
VI.
BAPCPA and Credit Cards
If the substitution theory is correct, then it suggests that an important byproduct of
the growth of credit card use and borrowing will be an increase in the bankruptcy filing
rate, holding everything else constant. This substitution itself may be the result of factors
fundamentally unrelated to bankruptcy, such as the lower operating costs of credit cards
for issuers and consumer demand for credit cards relative to other traditional types of
consumer credit.
The substitution effect, therefore, will increase the demand for
bankruptcy and lead to an increase in bankruptcy filing rates.
difficult because of his unorthodox specification of the regression, which rests on his decision to specify
each sequential LAG term as a separate regression, notwithstanding the extremely high degree of
multicollinearity among them. Using the more conventional statistical technique of testing all lag terms
simultaneously to control for multicollinearity results in the two longer lag terms being statistically
insignificant and the shorter terms borderline statistically significant. The relevant p values for the
variables are CCD=0.064, d1CCD=0.084, d2CCD=0.233, d3CCD=0.810.
91
If this is so, then there are several logical consequences that follow.
First,
although higher bankruptcies will likely result as a byproduct of increased credit card
lending, it would be economically inefficient to adopt policies that will encourage credit
card lenders to restrict the issuance of credit, especially to lower-income borrowers.
Doing so would have the unintended consequence of encouraging borrowers to shift back
from credit card lending to various higher-cost types of installment and other lending,
with negative welfare effects. Although this would likely have the effect of reducing
bankruptcy filing rates it would be an inefficient way of bringing this about, as it uses a
restriction in the supply of credit to try to solve what is fundamentally an issue of
bankruptcy policy.
A more efficient response would be to amend the bankruptcy code directly to
address the unintended consequences that have led to an increase in bankruptcy filings.
Under this approach amending the Bankruptcy Code can be understood as an institutional
response to address the various factors that have raised the incentives for filing
bankruptcy in recent decades.285
This would preserve the general social benefits
associated with widespread access to credit cards but restrain the problems of
opportunism, moral hazard, and adverse selection that have accompanied this growth of
credit card lending.
The latter approach is the one chosen by BAPCPA, which addresses the problems
created for the bankruptcy system by increased use of credit cards. This Part of the
article addresses the impact of BAPCPA on the treatment of credit cards in bankruptcy.
A.
285
The Effects of BAPCPA on Credit Cards in Bankruptcy
See Zywicki, Institutions, supra note 221.
92
BAPCPA makes several changes to the treatment of credit cards in bankruptcy,
all of which are designed to address the problems of moral hazard and adverse selection
associated with credit card use and bankruptcy. First, BAPCPA provides a variety of
new safeguards to reduce fraud in bankruptcy, such as requiring debtors to file tax
returns, pay advices, and other financial information that will make it easier to detect
fraud and concealment of assets by debtors.286
Second, before a debtor can file a
bankruptcy petition he must first attend consumer credit counseling to determine whether
he should enroll in a debt management plan instead of filing bankruptcy.287 Third, the
United States Trustee is given greater powers to investigate and pursue fraud through
enhanced powers to conduct random audits of debtors’ cases.288
Fourth, BAPCPA
imposes a new system of “means-testing” eligibility for chapter 7 relief.289
Under
BAPCPA, a debtor who makes above the state median income (adjusted for family size)
and has the ability to repay a substantial portion of his or her non-priority unsecured debt
in a chapter 13 repayment plan would be presumed to have to file under chapter 13 rather
than chapter 7 unless he or she can show extraordinary circumstances that make chapter
13 inappropriate. Although this provision was estimated to affect only about 10% of
bankruptcy filers, by targeting those with the highest repayment capacity, it was
estimated that this provision would recover $4 billion in bankruptcy that was previously
discharged.290 Fifth, BAPCPA makes nondischargeable more credit card debt for cash
advances and for the purchase of “luxury goods and services” in the pre-bankruptcy
period, lengthening the relevant period and reducing the threshold amount to trigger the
286
Bankruptcy Abuse Prevention and Consumer Protection Act of 2005 §315.
BAPCPA §106.
288
BAPCPA §603.
289
BAPCPA §102.
290
See Jones & Zywicki, supra note 205, at 186-92 (summarizing studies).
287
93
nondischargeability presumption.291
Sixth, BAPCPA places new limits on abusive
homestead exemptions, reducing the ability of debtors to protect assets in bankruptcy.292
Finally, BAPCPA expands the protection for credit card issuers in situations where
debtors use credit cards to pay debts that would be nondischargeable, such as for taxes.293
Each of these provisions will tend to increase the collection of credit card debts in
bankruptcy.294
B.
Implications of BAPCPA for Credit Cards and Bankruptcy
BAPCPA, therefore, makes some credit card debt more difficult to completely
discharge, both directly and indirectly.
BAPCPA thus should have two expected
economic results. First, it should increase the ex post financial recovery by credit card
issuers from those consumers who file bankruptcy and are impacted by the various
provisions related to means-testing, reduced fraud, and more stringent restrictions on
loading-up behavior of “luxury goods and services” and cash advances.
Second, these protections should reduce the incentives to file bankruptcy, and by
reducing the risk of lending to certain categories of debtors it should lead to reduced
credit card interest rates and increased credit supply to certain categories of borrowers.
On the other hand, this reduction in interest rates and increased supply of unsecured
BAPCPA §310. Credit card debt is generally dischargeable unless incurred for the purchase of “luxury
goods or services” within a short period preceding bankruptcy. 11 U.S.C. §523(a)(2)(C). Prior to
BAPCPA, the luxury goods or services had to be owed to a single creditor and aggregate more that $1,225
on or within 60 days of bankruptcy, or cash advances aggregating more than $1,225 within 60 days of
bankruptcy. BAPCPA amends this section to reduce the threshold for “luxury goods and services” to $500
days and extends the period length to 90 days, and reduces the cash advances threshold to $750 and extends
the period to 70 days.
292
BAPCPA §§308, 322.
293
BAPCPA §314.
294
Other provisions will have the likely effect of reducing collections for credit card issuers. In particular,
BAPCPA increases the rights in bankruptcy of certain secured creditors (notably automobile lenders) and
for certain priority debts, such as collection of domestic support obligations and taxes. The improved
position of these senior creditors will reduce the amounts available for distribution to unsecured creditors.
291
94
credit theoretically could lead to an increase in indebtedness and a subsequent increase in
bankruptcies.
As a result, the overall impact of BAPCPA on bankruptcy filings is
ambiguous as an a priori matter, and must be resolved empirically.
1.
The Effect of BAPCPA on Bankruptcy Filings
Each of the different theories discussed generate different predictions about the
likely impact of BAPCPA on bankruptcy filing rates.
Most straightforwardly, the
substitution model predicts that by reducing the incentives to file bankruptcy ex post,
BAPCPA will be expected to reduce the number of bankruptcy filings by encouraging
debtors confronted with financial distress to consider alternatives.
The distress theory, by contrast, predicts that BAPCPA will have little impact on
bankruptcy filing rates or may even have the unintended consequence of increasing
bankruptcy filing rates. At its simplest, the distress theory assumes that bankruptcy
filings are fundamentally involuntary in nature in that the “overwhelming majority of
filers are in such distress that they are all but compelled to file” and thus will have no
impact on bankruptcy filing rates.295 Others have argued that by reducing the risk of
lending BAPCPA may actually have the unintended consequence of actually increasing
the number of bankruptcy filings.296 Those who take this position argue that by reducing
lender risk in bankruptcy, BAPCPA will encourage lenders to extend credit to
Ronald J. Mann, Bankruptcy Reform and the “Sweatbox” of Credit Card Debt, 2007 U. ILL. L. REV. 375
(2007).
296
See Robert L. Lawless, The Paradox of Consumer Credit, 2007 U. ILL. L. REV. 347 (2007).
295
95
increasingly risky borrowers, which in turn will increase in the number of consumers in
financial distress and therefore to an increase in the number of bankruptcy filings.297
Although BAPCPA has been in operation only about two years, bankruptcy
filings have fallen dramatically since it became effective and are still only about half of
pre-BAPCPA levels despite problems in the economy and consumer credit markets. At
the same time, there is no evidence that BAPCPA is spawning hardship by making it
unduly difficult for needy filers to be able to file. This sharp decline in filing rates is
difficult to reconcile with the distress model but is consistent with the substitution model
which predicts that reducing the incentives to file bankruptcy will result in fewer filings.
The most likely reason for the inability of the distress model to accurately predict this
development is the failure of the adherents to the distress model to fully and accurately
understand the actual provisions of BAPCPA and the impact they are likely to have on
consumers. Those borrowers whose behavior will be affected by BAPCPA are not those
typically considered to be riskier marginal consumers. Instead, the relevant provisions of
BAPCPA dealing with credit cards are aimed at higher-income borrowers impacted by
means-testing and those borrowers likely to be affected by the more stringent
requirements related to the purchase of “luxury goods or services” or those with large
cash advances in the pre-bankruptcy period. The likely reason for the decline in filings is
that BAPCPA’s provisions address problems of opportunism by a particular segment of
Ausubel, supra note 111, at 269 (“Paradoxically, we conclude that the likely effect of further limiting
the dischargeability of credit card debt in bankruptcy is an increase in the frequency with which consumers
are buried under mountains of credit card debt”); id. at 270 (“[N]ew limitations on the dischargeability of
credit card debt are likely to have unintended consequences and will probably only worsen the problem of
consumer overextension.”); Gene Koretz, Reform That Could Backfire: Why Bankruptcies Will Increase,
BUSINESS
WEEK
(May
21,
2001),
available
in
http://www.businessweek.com/magazine/content/01_21/c3733035.htm; Mark M. Zandi, Easy Credit,
Profligate Borrowing, Tough Lessons, REGIONAL FIN. REV. 16 (Jan. 1997) (cited in NATIONAL
BANKRUPTCY REVIEW COMMISSION, REPORT Chapter 1 (Oct. 20, 1997).
297
96
borrowers—those higher-income debtors most prone to opportunism—rather than simply
reducing risk across the board for all borrowers.
This latter group is likely to be composed disproportionately of higher-income
borrowers as well, who are likely to have sufficiently high credit lines and opportunities
to consume “luxury goods and services.” As a result, these provisions are not likely to
have a substantial effect of leading to increased bankruptcies as BAPCPA alone is not
likely to spur a substantial increase in lending to higher-risk consumers. To the extent
that it might have such a result, as with previous consumer lending innovations BAPCPA
will likely result in some substation away from alternative higher-cost loans to credit
cards, thereby resulting in no increase in financial distress.
2.
The Effect of BAPCPA on Credit Cards in Bankruptcy
Ronald Mann argues that the primary effect of BAPCPA will not be to reduce
filings by reducing fraud and abuse, but rather to transfer wealth from debtors to credit
card issuers by substantially prolonging the length of time necessary before individuals
can file bankruptcy, a period during which a financially-stressed debtor will be accruing
interest and presumably other fees on her account. Mann refers to this period as the
bankruptcy “sweatbox.”298 He writes, “[T]he most important effect will be to slow the
time of inevitable filings by the deeply distressed, allowing issuers to earn more revenues
from these individuals before they file.”299 Mann does not explain how exactly credit
card issuers will be earning “more revenues” from deeply distressed consumers during
the “sweatbox” period.
298
299
It seems to follow from his premise that the borrowers in
Mann, supra note 295.
Id. at 379.
97
question are so distressed as to be unable to pay their debts at all, suggesting that the
overwhelming impact of the accrual of new interest and fees will be to simply increase
the eventual bankruptcy claims of credit card issuers—and other creditors who remain
unpaid during the same time period—not to actually increase the receipts of credit card
issuers. But Mann’s reasoning here is a bit unclear. He states that borrowers “often pay
until close to the bankruptcy date” but also that most debtors are in hopeless financial
distress at the time they file.300 These two claims seem to be inconsistent.
More importantly, although Mann asserts that BAPCPA will substantially delay
the ability of distressed debtors to file bankruptcy, his claim is based largely on
generalized assertions about the effects of BAPCPA rather than a concrete analysis of its
provisions. He enumerates several provisions of BAPCPA that increase the costs or
delay of filing bankruptcy, such as increasing the period between repeat chapter 7 filings
from six to eight years, raising the costs of filing bankruptcy, increasing some of the
administrative hurdles (such as requiring pre-filing credit counseling), and by reducing
the benefits of filing bankruptcy. He makes little effort to distinguish those reforms that
may be justified by legitimate efforts to reduce fraud and abuse, such as increasing the
period between repeat filings or administrative burdens to improve the integrity of the
system, from those reforms that arguably are designed solely to increase the heat of the
“sweatbox.”301 He asserts that the combined impact of all of these provisions would be to
create a delay of “several months” before distressed households could actually file
300
.Id. at 397 & n. 106.
It appears that Mann simply dismisses the legitimacy of any of these reforms on the basis that they are
unfounded because of an absence of fraud or abuse in the system. See id. at 379-84.
301
98
bankruptcy.302 He predicts that BAPCPA will have “only a minor effect on the number
of people who choose to file, although it will have a notable effect on when they file.”303
Mann offers little detailed theory or empirical evidence to support the sweatbox
theory—indeed, he acknowledges that his discussion is “speculative” and “depends on a
variety of empirical assumptions about the behavior of distressed borrowers that are
difficult to verify empirically.”304 To date, there is no evidence that BAPCPA has either
prevented needy filers from filing or substantially delayed them from filing, which would
be necessary for the sweatbox effect.305 First, Mann’s theory rests on the premise that
BAPCPA makes bankruptcy more difficult and a more extended process for many filers.
Many of the reforms affect post-filing behavior, thus they seem to have no relevance to
the pre-filing “sweatbox.”
For those that do affect pre-filing behavior, Mann
overestimates the costs that they impose. Although BAPCPA adds some new complexity
or delay for some filers, for the vast majority of filers the increases are quite small.
For example, Mann notes the requirement of pre-bankruptcy credit counseling as
a precondition for filing bankruptcy.306 In practice, however, this requirement is not
onerous and it is doubtful that it has deterred or slowed many consumers from seeking
bankruptcy relief. According to a study by the General Accounting Office, the cost and
delay associated with seeking consumer credit counseling has been minimal.307 Ninety
percent of bankruptcy clients fulfill the requirement on the spot by telephone or Internet
302
Id. at 393.
Id. at 397.
304
Id. at 396.
305
See Testimony of Professor Todd J. Zywicki, Testimony before United States Senate, Committee on the
Judiciary, Subcommittee on Administrative Oversight and the Courts, “Oversight of the Implementation of
the Bankruptcy Abuse Prevention and Consumer Protection Act” (Dec. 5, 2006).
306
Id. at 395.
307
See GENERAL ACCOUNTING OFFICE, BANKRUPTCY REFORM: VALUE OF CREDIT COUNSELING
REQUIREMENT IS NOT CLEAR (April 2007).
303
99
with little inconvenience or delay.308
Many lawyers have even set aside rooms,
telephones, and computers in their offices to permit debtors to fulfill the credit counseling
requirement with minimal delay and disruption.309 Although the typical fee for credit
counseling and a bankruptcy certificate is $50, many indigent filers have these fees
waived (as high as 26% for some counseling agencies).310 Counseling agencies have
been almost uniformly prompt and competent in fulfilling their responsibilities.
Other
obligations under BAPCPA, such as the requirement of filing tax returns and pay advices
with one’s petition, increase paperwork burdens but it’s not clear why that would
necessarily increase delay. Similarly, the extension of the refilling bar from six to eight
years will likely affect only a small number of filers.311
Available data also is inconsistent with Mann’s theory. One logical prediction of
Mann’s theory would be that post-BAPCPA credit card delinquencies would rise rapidly
as the predicted effect of BAPCPA should be to create a growing class of debtors who
need to file bankruptcy (or at least would have been able to file bankruptcy pre-
308
Id. at 21. Recent research indicates, somewhat surprisingly, that the efficacy of credit counseling is not
affected by whether it is conducted by telephone versus in person. See Michael Staten and John Barron,
Evaluating the Effectiveness of Credit Counseling, Phase One: The Impact of Delivery Channels for Credit
Counseling Services (May 31, 2006); but see Noreen Clancy & Stephen J. Carroll, Prebankruptcy Credit
Counseling 17 (2007) (concluding that the “literature generally, but not always, suggests that in-person
counseling is more effective than other modes of delivery).
309
GAO, CREDIT COUNSELING, supra note 307, at 37. The GAO expressed concern about the marginal
value associated with consumer credit counseling, especially in these situations, but this is distinct from the
concerns of the “sweatbox” theory, which contends that the requirement substantially increases the
marginal cost of filing bankruptcy.
310
GAO, CREDIT COUNSELING, supra note 307.
311
There are a relatively modest number of repeat chapter 7 bankruptcy filers and although it is not known
exactly how many people would need to refile more than six years but less than eight years after an initial
filing, it is difficult to imagine that this will impact a substantial number of bankruptcy filers. See Jean M.
Lown, Serial Bankruptcy Filers No Problem, 26 AM. BANKR. INST. J. 36 (June 2007) (estimating that about
4 percent of bankruptcy filers are repeat filers). On the other hand, many of those who are repeat filers are
actually serial filers who file many times. See Jean M. Lown, Serial Bankruptcy: A 20-Year Study of Utah
Filers, 25 AM. BANKR. INST. J. 24 (Feb. 2006).
100
BAPCPA), but are unable to do so post-BAPCPA.312 This dynamic should be captured in
a growing number of delinquent debtors who are struggling to make their payments. In
fact, delinquency rates on credit card loans post-BAPCPA appear to be comparable or
lower than delinquency rates for many pre-BAPCPA.313
VII.
CONCLUSION
This article has examined the conventional belief that the growth in the
availability of credit cards over time has caused a direct increase in consumer bankruptcy
filing rates by increasing the frequency of household financial distress. Analysis of
available data and empirical studies fail to support the hypothesis. Instead, the growth in
credit card debt over the past three decades reflects a substitution of credit cards for more
expensive and less attractive alternative forms of credit such as personal finance
companies, pawn shops, and retail store credit, have now found themselves with a general
purpose credit vehicle that can be tapped whenever necessary. As a result, there has been
a general increase in credit card credit and availability, but no evident increase in overall
consumer indebtedness (as measured by the debt service ratio) or risk. On the other
hand, by substituting unsecured debt for secured debt and nationalized and
depersonalized types of consumer credit for traditional forms of credit embedded in local
relationships, the substitution to credit cards has increased the propensity of consumers to
file bankruptcy in response to financial distress.
312
Efforts to isolate the possible impact of BAPCPA on credit card delinquency are confounded because on
January 1, 2006—the first quarter after BAPCPA became effective—new federal regulatory guidelines
requiring higher mandatory minimum payments on credit cards went into effect, dramatically increasing
the mandatory minimum. See Credit Card Minimum Payments Set to Double Beginning January 1, 2006,
http://www.send2press.com/newswire/2005-12-1205-005.shtml. Credit card delinquencies jumped a small
amount after this regulatory guidance became effective, but have remained largely constant since 2006Q1.
313
See Figure 6 supra (current delinquency rates on credit card loans are lower than many earlier periods).
101
Strategic behavior by borrowers also may explain some of the rise in credit card
debt and the relationship between credit card debt and bankruptcy. Because of problems
of asymmetric information borrowers know their likelihood and timing of bankruptcy
better than creditors. As such, the strategic theory predicts that debtors would increase
their use of dischargeable credit card debt in the period immediately preceding
bankruptcy as well as disproportionately choosing to default on credit card debt relative
to secured and nondischargeable debt. In fact, credit card chargeoffs have risen over time
even as chargeoffs on household mortgages and nondischargeable student loans have
fallen. Bankruptcy filers also make greater use of credit card debt than other households.
All of this suggests some support for the strategic model.
Finally, this article has examined the impact of BAPCPA’s reforms on credit
cards. Critics of BAPCPA argued that because it failed to address the underlying causes
of the rise in bankruptcies, namely the extension of excessive credit card debt to high-risk
borrowers, BAPCPA would have minimal effect on bankruptcy filings or even result in
higher filings. Two and a half years after BAPCPA became effective, however, filings
are still approximately one-half of their steady-state level pre-BAPCPA.
Critics of
BAPCPA also argued that even if BAPCPA did not reduce filings, by making it more
difficult for borrowers to file, thereby drawing out their period of default and their time in
the “bankruptcy sweatbox.” These predictions also appear unfounded to date. These
predictions, however, seem to have been based on a misunderstanding of the actual terms
of BAPCPA.
Credit is as old as human society. And as old as credit is the concern expressed
by some that “other people” cannot be trusted with credit and that they are being tempted
102
into overconsumption and overindulgence and pining for a “golden age” of purported
frugality and aversion to consumer debt. This golden age never existed, at least in
America. The criticisms heard today of credit card issuers are identical to those of
installment lenders of the 1940s and before that still other perceived irresponsible use of
consumer credit. It is time to move beyond the conventional wisdom and to understand
how Americans actually use consumer credit and why they have rationally chosen to
increase their use of credit cards. Only then will we fully understand the relationship
between credit cards and bankruptcy and ensure that any reforms will increase consumer
welfare.
103