Elizabeth Warren⁽¹⁾

Americans love records, statistics, and amazing numbers. Home runs, rushing yardage, presidential-approval ratings, smoking-related deaths, and murder rates--all offer great material for attention-grabbing headlines. In 1996, the bankruptcy system stepped into the spotlight by supplying an attention-grabbing number of its own: more than a million bankruptcies were filed in a single year.⁽²⁾ The filing rate had increased sharply over the previous year, but this was nothing new. Rather, the magic of "one million" made bankruptcy a newsworthy subject to be covered in virtually every newspaper, news magazine, and national television news program during the year. Bankruptcy, like baseball and public-opinion polls, had become a sport of numbers, perhaps not widely understood, but the subject of breathless reports, as the benchmark of a million-plus consumer filings was passed.

The million-filing mark might have been a short-lived subject for the popular press but for one factor: the filing rate gave the credit industry an opportunity to plead its case publicly that Congress should significantly restructure the bankruptcy laws. The fact that one million families filed for bankruptcy signified a crisis, namely that it is too easy to file for bankruptcy. The credit industry also offered the blueprint for a solution: make it more difficult for families to file for bankruptcy.

Business bankruptcy filings are not part of the alleged crisis. Perhaps a robust economy or the increasing sophistication of parties to craft out-of-court workouts has kept the business bankruptcy system out of the news. Business bankruptcy cases account for only about four percent of all bankruptcy filings and the number of business filings has been dropping gradually for nearly a decade.⁽³⁾ A single business filing can have a powerful impact on workers, suppliers, and a local economy, but reporting tends to focus on the details of a particular case, not on an alarming trend. Outside the rarified corridors of academia, few people are calling publicly for the overhaul of the business bankruptcy system.

The consumer bankruptcy system, however, has been subjected to vigorous and repeated attack. Pundits implicitly assume that it is fine for the current system to permit businesses in financial trouble to stiff each other and perhaps even to dump their employees and wiggle out of their obligations to consumers, but the fact that Mary Beth and Kevin Rourke discharge their debt to Citibank signifies that something is terribly wrong in a well-ordered legal system.

The line of argument that casts families as villains--or at least as suspects--starts and ends with the sharp rise in consumer bankruptcy filings. Why would a million families file for bankruptcy in a time of unparalleled prosperity? Why would bankruptcy filings rise by more than twenty percent each year from 1994 through 1997 when inflation and unemployment were both falling? The obvious answer, according to the credit industry, is that consumers must be taking advantage of their creditors and using the bankruptcy system to shuck legal responsibilities they otherwise could have met easily. The credit industry has concluded that the system is full of these people engaging in this abusive activity. We are in a crisis, say the creditors, and a crisis means that the time has come for reform that will limit the last resort of the bankruptcy discharge for American families.

The argument for sweeping change is powerful in its simplicity. It rests on a single premise: an assumed connection between one-million-plus filings and consumer abuse. If the assumption is erroneous, then the clamor for radical and costly bankruptcy reform is misplaced. Indeed, a different explanation for the rise in filing might support a very different view of the financial health of American families and the function of the bankruptcy system.

I. Why Are Consumer Bankruptcies Increasing?

The actual explanation for the rise in consumer bankruptcies is at once both elegantly simple and devilishly complex. The simple explanation is that consumer bankruptcies are rising because consumers' debts are rising faster than their incomes. More debt means more families are in trouble or are vulnerable to trouble. The complex answer explores why debts are outstripping income for an increasing number of Americans.

The easiest way to think of the role of rising consumer debt in consumer failures is to envision American families as hikers. Through life, the hike is sometimes uphill and sometimes downhill, and the path is sometimes strewn with obstacles. As they hike, the adults may acquire debt, which is much like putting rocks in their backpacks. If the packs have only a few rocks, the family can withstand most of the events and calamities they are likely to encounter. But as the pack grows heavier, smaller and smaller misfortunes will cause the hikers to fall, unable to continue the hike. The only way they can get back on their feet is by emptying some of the rocks from their packs. More Americans' packs are getting heavier with debts, and more of them find that they can continue only if they unload some of their rocks by declaring bankruptcy.

The macrodata⁽⁴⁾ are unambiguous about the best predictor for consumer bankruptcy. Consumer bankruptcy filings rise and fall with the levels of consumer debt. After a comprehensive analysis of household indebtedness, the Congressional Budget Office ("CBO") concluded that "[t]he incidence of bankruptcy in the adult population closely follows the indebtedness of the household sector."⁽⁵⁾ As more families carry more debt, a larger percentage of them end up in the bankruptcy courts.⁽⁶⁾ The CBO graph reproduced on the following page illustrates the correlation.

Nonbusiness Bankruptcy Cases and Debt

Dotted areas are recessions as defined by the National Bureau of Economic Research.

The fact that bankruptcies are rising during a time of prosperity is readily explainable as well. The CBO noted that a rise in consumer bankruptcies "during periods of economic expansion is not really surprising"⁽⁷⁾ since household debt also increases during economic expansion.

Other detailed studies of time-series data explore a range of economic factors in greater detail, but they reach a similar conclusion. Economics professor Jagdeep Bhandari and business professor Lawrence Weiss reviewed data about economic conditions and bankruptcy filing rates from 1946 through 1991. They note the high positive correlation between consumer debt and bankruptcy filings (both rise together) and the high negative correlation between income and bankruptcy filings (as one rises the other falls). They concluded, "[o]ur evidence indicates that the increase in the number of bankruptcy filings is primarily due to the increased level of debt as a percentage of income."⁽⁸⁾

Two economists, Professor Ian Domowitz and Professor Elie Tamer, developed a statistical model to estimate bankruptcy filing rates.⁽⁹⁾ They determine that from 1939 to 1994 bankruptcy legislation favorable to debtors has had little effect on bankruptcy filings.⁽¹⁰⁾ These researchers suggest that consumers do not take advantage of more favorable bankruptcy laws;⁽¹¹⁾ instead, bankruptcy filings are driven by economic factors.

Economist Lawrence Ausubel developed a related explanation of rising bankruptcy filing rates. While he documents the relationship between high consumer debt generally and rising consumer bankruptcies, he focuses more specifically on the tie between credit card default rates and bankruptcy filing rates. He notes that the rate of consumer bankruptcies is "astonishingly highly correlated with the rise in credit card defaults."⁽¹²⁾ Professor Ausubel explains:

The profit margins of credit card issuers substantially increased beginning in 1982, as a result of the functional deregulation of credit card interest rates coupled with prevalent consumer behavior. This has created incentives for card issuers to relax their credit standards, in turn leading to a secular increase in the rate of credit card defaults.⁽¹³⁾

When the cards are highly profitable, credit card issuers have a strong incentive to distribute them to marginal borrowers and to borrowers already loaded with debts, which increases both the issuer's profits and its loan defaults. Card profitability may increase for a number of reasons. About three-quarters of credit card revenues come from interest payments,⁽¹⁴⁾ so profitability increases if the same customers increase their debt levels, if new customers begin to carry debt, or if the rates the company charges increase more than their cost of funds. Any one of these routes to profitability could also be a route to bankruptcy, and Professor Ausubel demonstrates that high credit card profitability correlates closely with increases in defaults and bankruptcies.

Research by the Federal Deposit Insurance Corporation ("FDIC") demonstrates the same correlation, observing that interest rate deregulation in the mid-1980s led to a substantial expansion in credit availability and an increase in personal bankruptcies.⁽¹⁵⁾ The FDIC concludes: "A tightly regulated world, marked by a low level of personal bankruptcies and restricted access to credit, was exchanged for a deregulated world marked by expanded access to consumer credit and a higher level of personal bankruptcies."⁽¹⁶⁾

The CBO data, the Bhandari and Weiss data, the Ausubel data, and the FDIC research all demonstrate the correlation between rising levels of consumer debt and rising bankruptcy rates. The simple explanation of the rise in filings-- bankruptcies rise as household debt rises--is undeniable. As their packs get heavier, more Americans fall and cannot get back on their feet without lightening their loads.

B. Why So Much Debt?

Why have Americans taken on so much consumer debt relative to their incomes? Some incur the debt with little thought about how it adds up, perhaps like the grasshopper who never thought about the coming winter. Others take on substantial debt, but they keep it within manageable ranges until an unavoidable financial obligation--a significant medical bill or a car accident--pushes the aggregate debts to staggering levels. Still others have large debts they could maintain until they were downsized out of a well-paying job or faced a cutback on overtime. The mismatch between debt and income, either because the debts rose or the income fell, is fundamental to debtors' economic collapse.

There are many rocks along the hiking trail that can cause economic collapse. My coauthors, Dr. Teresa Sullivan and Professor Jay Westbrook, and I have tried to identify specifically the fractures in middle-class economic stability that have led to the debt-income imbalances that force so many debtors into bankruptcy. In our forthcoming book, The Fragile Middle Class: Americans in Financial Crisis, we explore debtors' own explanations for their economic failures along with data on economic pressures on most Americans. A detailed analysis of the kinds of problems that trigger the economic collapse of more than a million families each year is beyond the scope of this paper. For this paper, the focus is on the allegation that the rise in consumer bankruptcy can be attributed to rising consumer abuse of the bankruptcy system.

II. Evidence of Abuse

Notwithstanding the strong statistical evidence that the rise in consumer bankruptcy filings is linked to the increase in debt loads per household, the credit industry--and many other observers--place the blame for the rise in filings on increased consumer abuse. According to this view, increasing willingness of American families to take advantage of their creditors by filing bankruptcy has prompted the increase in consumer filings. To sustain this assertion, however, it is essential to develop evidence of systematic abuse. Outrageous anecdotes may highlight legal loopholes that should be closed, but they cannot constitute the sole basis for radical changes to the policy and structure of an entire system.⁽¹⁷⁾ For example, if the law permits an individual to convey all her property to a self-settled trust while she shields millions of dollars from her creditors, the practice must be curtailed even if it is a rare event because it is unfair to creditors and a violation of underlying bankruptcy principles. Indeed, the National Bankruptcy Review Commission recommended a number of changes to the consumer bankruptcy system that fit this category: there was no evidence that the identified abusive practices were widespread, but the bankruptcy code should not condone certain kinds of debtor and creditor behavior.

The current national debate is premised on the existence of a crisis. By the terms of the current debate, significant changes in the structure and policy underpinnings of the present system are justified by a calamitous rise in filings that signals widespread abuse. To presume a crisis from only the cold filing numbers is too broad a logical leap. Some form of systematic abuse must be demonstrated. If the system is experiencing occasional abuses at the margin, then the debate--and the proposed solutions--should be shaped very differently.

A. Macrodata

At the outset, an allegation of widespread abuse will be difficult to document because the available macrodata offer an adequate explanation for the rise in consumer bankruptcies. One can fully account for the current rise in filings and leave no room for the inference that the rise is prompted by abusive filers. If debtors behave precisely as they have for decades--declaring bankruptcy more often when they carry more debt and declaring bankruptcy less often when they carry less debt--the rise in consumer filings is entirely predictable. Since the rise in consumer filings is a predictable function of debt, then the rise in filings should have been considerably sharper if the system were now rife with abuse. To rebut the inference that the rise is linked to increased debt loads, as it has been for decades, evidence of abuse must be strong enough to demonstrate enough systematic abuse to account for a significant rise in filings.

The major effort to develop macroeconomic evidence of some less-mentioned causes of bankruptcy was a study initiated by Visa, Inc.⁽¹⁸⁾ The study examined "[o]ver 65 macroeconomic and social variables with potential explanatory value . . . to explain trends in personal bankruptcy filings both at the state and national levels."⁽¹⁹⁾ The study covered a sixteen-year period, and purports to explain "99 percent of the variation in bankruptcy filing trends."⁽²⁰⁾ The study hints only indirectly at widespread abuse in its conclusion that a significant portion of the rise in consumer filings is based on unmeasured "social factors" such as "changes in the bankruptcy laws, the reduced stigma associated with filing for personal bankruptcy and broader advertising of legal assistance with bankruptcy filings."⁽²¹⁾ Visa concedes that "[m]any social factors are not easily quantifiable, as consistent, historical data do not exist."⁽²²⁾ To deal with this lack of data, the Visa researchers simply assumed that anything that could not be explained by other economic variables must be attributable to those of social variables.

This study was submitted to the National Bankruptcy Review Commission to support the adoption of credit-supported means testing, evidently on the assumption that evidence of the effects of social variables, which combine such factors as lawyer advertising and declining stigma, suggest a constellation of abuse.⁽²³⁾ The Commission in turn asked the CBO to review the study. A review by Mr. Kowalewski, Chief, Financial and General Macroeconomic Analysis Unit, Macroeconomic Analysis Division, CBO, concluded that the method of analysis employed by the Visa researchers was "unsound"⁽²⁴⁾ and reported that "[t]he results of my analysis suggest the opposite--the model is misspecified and hence is an unreliable explanation of personal bankruptcy filings. . . . I also found the report's claim about the importance of 'social factors' that Visa believes affect personal bankruptcy filings to be unfounded."⁽²⁵⁾ Even a modest, indirect claim of rising abuse could not be sustained by the macrodata.⁽²⁶⁾

As the Visa efforts demonstrate, identifying the variables that will demonstrate abuse can be exceedingly difficult. One reason is that the very concept of abuse is slippery. One person's example of abuse may be another person's legitimate use of the system. Filing bankruptcy on the eve of a home mortgage foreclosure is, in the eyes of some, an abusive use of bankruptcy and, in the eyes of others, a legitimate use of a statutory provision designed to halt debt collection. Without venturing deeply into the heavily moralistic debate about the elements of abuse, there appears to be some joinder in the debates that has created a working definition of abuse: if debtors use bankruptcy when they could repay their debts, then they have abused the system and have imposed costs on their bill-paying counterparts who do not declare bankruptcy. The central idea is that people who use bankruptcy ought to be those in serious financial trouble.

B. Microdata

The microdata⁽²⁷⁾ suggest that the people who file for bankruptcy generally are those who need it. The work of my coauthors, Dr. Teresa Sullivan and Professor Jay Westbrook, and I show this to be the case. In the early 1990s, for example, the typical family filed for bankruptcy with an annual combined income of about $21,200 (in 1997 dollars),⁽²⁸⁾ compared with a median family income of about $42,400 (in 1997 dollars) for those not in bankruptcy.⁽²⁹⁾ For many, their incomes represented a sharp drop over the preceding year as they coped with unemployment, downsizing, and self-employment as independent contractors.⁽³⁰⁾ At the same time, their debts were staggering, far outstripping debt loads for most Americans.⁽³¹⁾ The ratio of consumer debt-to-income (which omits home mortgage debt), shows that at the median these debtors owed more than their annual income to doctors, hospitals, credit card issuers, finance companies, and other consumer lenders.⁽³²⁾ On average, they owed so much in short-term, high-interest debt that every fifth paycheck would have to be used exclusively for interest payments on their debts just to stay even.⁽³³⁾ If they could not afford to give up one paycheck every five weeks, then they would simply fall deeper into debt even if they never incurred another doctor's bill or another credit card charge or picked up any new rocks.

These data sketch a bleak picture for the typical debtor in bankruptcy, which is reported in far more detail elsewhere,⁽³⁴⁾ but the data offer only indirect information on abuse. These data suggest that any abuse is confined to few enough debtors that they affect neither the mean nor the median financial reports emerging from the bankruptcy courts.

To say, however, that most of the debtors are in financial trouble does not definitively answer the question of whether consumers abuse the bankruptcy system. The credit industry has funded its own microdata analysis of consumer bankruptcy that has been the primary basis for its bankruptcy reform efforts. Continuing its longtime affiliation with the Credit Research Center ("CRC"),⁽³⁵⁾ originally located at Purdue University and recently relocated to Georgetown University with its director, Michael Staten, the credit industry commissioned a study of debtors who filed for bankruptcy in 1996.⁽³⁶⁾

The credit industry study purported to test directly whether debtors in Chapter 7 "could have funded meaningful Chapter 13 repayment plans" by examining the incomes and debts of debtors currently in Chapter 7.⁽³⁷⁾ The study collected detailed information from 4000 files in major American cities and assembled the data to construct direct evidence of abuse.⁽³⁸⁾ As such, the study focused more tightly than other studies on the working definition of abuse--ability to repay.

The credit industry study used information from the debtors' reports of their incomes and expenses, but the researchers calculated "ability to repay" by including some expenses and excluding others from the reported family budget. So, for example, a debtor's continuing payments on certain secured debts, such as car loans or appliance loans, were omitted from a debtor's budget expenses unless a formal reaffirmation could be found in the files; continuing payments of nondischargeable debt, such as educational loans, evidently were also omitted.⁽³⁹⁾ All household expenses were assumed to hold constant for five years, with no adjustment for inflation or addition of a budget entry to pay car repairs or future emergencies.⁽⁴⁰⁾ Family income was computed from all sources, including alimony and income from a small business. Each debtor's income was assumed to continue without interruption even though the report indicates that more than 68% of the debtors experienced an "income decline" suggesting a period of unemployment prior to filing.⁽⁴¹⁾ Debts were calculated without interest charges, including the debts that were reaffirmed. Matching income-after-adjusted-expenses to the listed debts, the credit industry study concluded that 40% of the debtors could pay at least a dollar of debt, and that over a five-year period, 13.7% of the unsecured debt listed by this 44% of the Chapter 7 debtors in bankruptcy could be repaid.⁽⁴²⁾ The report concluded that even more families could pay if they sold their cars and their furniture.⁽⁴³⁾

The credit industry's press releases widely touted the finding that 40% of debtors could repay some of their debts. This assertion made it into the popular press as a "fact" without attribution to its origin. Newsweek magazine, for example, ran an article entitled "Deadbeat Nation" in which it noted that creditors blame debtor abuse for the rise in bankruptcies: "one study says 45 percent of bankruptcy filers could pay back much of their debt."⁽⁴⁴⁾ Nowhere does the article point out that the quoted statistic comes from the industry's own study. Three months later, The Los Angeles Times, lamenting creditor abuse said, "[m]uch of [consumers'] debt relief is simply unwarranted. Experts say that about 45% of Americans who seek complete relief from their debts through Chapter 7 personal bankruptcy filings could afford to pay off an average of one-third of their debts within three years."⁽⁴⁵⁾ Again, there is no indication in the piece that the only "experts" who claim such numbers are those funded by the credit industry. USA Today picked up the same information for its editorial on consumer bankruptcy: "Purdue University researchers found a third of debtors could pay some of their bills but don't."⁽⁴⁶⁾ Former Senator Lloyd Bentsen wrote an op-ed for The Washington Times proclaiming, "[a]ccording to a Purdue University study, nearly half of the people who file for bankruptcy could repay a significant amount of their outstanding obligations, but instead choose to renege."⁽⁴⁷⁾ Mr. Bentsen also failed to disclose in his op-ed that he had been hired earlier that month by the credit industry to lobby his former colleagues in the Senate to adopt harsher bankruptcy laws. His conclusions about bankruptcy abuse were cited in turn by Congressman Gekas in his press release accompanying introduction of a means-tested bankruptcy bill, again without notation of Mr. Bentsen's lobbyist status or the credit industry as the source of the underlying research.⁽⁴⁸⁾

While it does not highlight this finding, the credit industry report indicates what proportion of the debt listed in Chapter 7 would be repaid. If each debtor were examined in a complex infrastructure at no cost to the creditors, and if all the optimistic assumptions of the credit industry researchers held true, the industry could collect about 1.21% of its outstanding "nonhousing debt" each year for five years.⁽⁴⁹⁾ It is interesting to speculate whether this return would outweigh the administrative costs to the creditors of maintaining an open, interest-free account for five years. In order to produce that return for the creditors, the taxpayers would be charged with building an elaborate bureaucratic structure to review the debtors' files and with providing hearings for debtors who dispute their creditors' claims that they could repay. Whether the increase in administrative costs, including trustee fees, debtors' and creditors' lawyers' fees, court costs, filing and tracking costs, and mailing costs would outweigh any increase in creditor collection is beyond the scope of the credit industry study.

The credit industry pressed its study, then only in interim form, on both the National Bankruptcy Review Commission and on Congress. Some concerns about the repayment data led members of Congress and the Chair of the Commission to request technical reviews of the study. Both the CBO and the General Accounting Office ("GAO"), in their roles evaluating statistical information bearing on policy questions for Congress, reviewed the credit industry study. The CBO review, conducted by Kim Kowalewski, Chief, Financial and General Macroeconomic Analysis Unit, Macroeconomic Analysis Division, raised concerns about the abbreviated reporting of the data that "make it impossible to determine the reliability of the study's findings."⁽⁵⁰⁾ The analysis raised questions regarding unfounded assumptions about current income, future income, current expenses, future expenses, amount of indebtedness, the representativeness of the sample, data-collection techniques, sample weighting methods, and inaccurate data comparisons.⁽⁵¹⁾ For multiple reasons therefore, the CBO report concluded that the credit industry study was "misleading" and it "overstat[ed]" capacity to repay.⁽⁵²⁾ The coauthor of the study, Michael Staten, made an angry rebuttal to the CBO analysis before the report was made public.⁽⁵³⁾ Later, Mr. Kowalewski reviewed Dr. Staten's comments, and further analyzed some new data presented, discovering added evidence that the study "overstate[s the debtors'] repayment capacity."⁽⁵⁴⁾

With a broader time frame in which it could review the study, the GAO could make an even more detailed assessment that raised additional questions regarding unfounded assumptions about both current and future income, calculation of expenses, and accuracy of the debt figures.⁽⁵⁵⁾ In addition, the GAO report raised questions about sampling techniques and data collection.⁽⁵⁶⁾ Like the CBO, the GAO noted that because the report contains too little basic statistical information, "standard statistical methods, such as the calculation of statistical error rates, cannot be used to evaluate the likely accuracy of the results."⁽⁵⁷⁾ The GAO report expressed concern that the report overstated debtors' ability to repay.⁽⁵⁸⁾ The GAO report concluded that the "[s]tudy's [f]undamental [a]ssumptions [w]ere [n]ot [v]alidated" and that the credit industry report does "not provide a sound basis" for its conclusion about how many debtors in bankruptcy could repay their debts.⁽⁵⁹⁾ A number of academics who reviewed preliminary reports of the study reached a similar conclusion.⁽⁶⁰⁾

Once again, Dr. Staten published a detailed rebuttal to the report even before the report was made public. Dr. Staten's rebuttal, published on the American Bankruptcy Institute's web page as early as January 26, 1998, preceded the public release of the GAO report by more than a week.⁽⁶¹⁾ The response, however, does not address many of the objections raised by the GAO report, a point noted by the GAO.

When the credibility of the CRC study was challenged, the credit industry responded by hiring more studies. The industry paid accounting firm Ernst & Young ("E&Y") to corroborate the CRC study. The GAO included the E&Y work in its analysis, concluding that the E&Y study suffered from the same deficiencies as the CRC study.⁽⁶²⁾ Undeterred, the credit industry paid for yet another study to be released as the GAO was making its unfavorable report on the earlier studies.⁽⁶³⁾ This study, by WEFA Group, purported to make a "conservative" estimate of the losses imposed by the bankruptcy system. At the request of Congress, the GAO evaluated this study in detail, concluding once again that the validity of the study was not established.⁽⁶⁴⁾ The GAO was critical of the assumptions used in the study, the methodology employed, the lack of documentation, and the breadth of the conclusions.⁽⁶⁵⁾

The CBO and GAO reviews of the CRC study sponsored by the credit industry are reminiscent of the credit industry's prior efforts to collect data to persuade Congress and the public of the existence of a bankruptcy crisis. Claiming that the rise in consumer bankruptcies in the early 1980s demonstrated abuse of a liberal new bankruptcy law, the credit industry lobbied Congress to restrict Chapter 7 eligibility by adopting a means test. The resulting bill, S. 2000, was highly controversial, and the credit industry turned to the same CRC to develop a study to demonstrate debtor abuse. The study, quickly distributed in Congress and widely quoted in the press,⁽⁶⁶⁾ purported to show that more than thirty percent of the bankrupt debtors could have repaid their debts in full and that $1.1 billion was discharged each year by debtors who could have repaid.⁽⁶⁷⁾ There was no CBO or GAO review of the 1981 credit industry study, but my coauthors, Teresa Sullivan and Jay Westbrook, and I reviewed that study.⁽⁶⁸⁾ We concluded:

The Purdue Study purports to be a scientific study, offering only "sterile" data. As a scientific study it is deeply flawed. The Study lacks crucial expertise, is designed incorrectly, asks a series of inartful questions, gathers its data improperly, misanalyzes the statistical data and draws erroneous and biased inferences from the data analysis. Moreover, error after error increases the count of the debtors who "could pay" and the amount of debt that could be recovered under S. 2000.⁽⁶⁹⁾

This critique came too late to have any meaningful influence on the debates, but the credit industry was not successful in persuading Congress to adopt a means test. Nonetheless, the credit industry got its money's worth. The bankruptcy laws were amended to include a package of credit industry proposals that made bankruptcy more difficult for individuals and permitted creditors to seek debt reaffirmation more aggressively.⁽⁷⁰⁾ With bankruptcy filings approaching 300,000, Congress and the public were persuaded that there was a bankruptcy crisis requiring a credit-industry-proposed solution.

The credit industry study has been discredited, but there is another approach using microdata to detect rising consumer abuse. Data collected over time can help detect shifts in the population of people who declare bankruptcy. If the sharp rise in consumer bankruptcies is attributable to a sharp increase in debtor abuse, then the basic financial profiles of the filers ought to change over time: the addition of more affluent debtors should increase both the mean and the median incomes or lower the relative size of the debts, demonstrating their relatively better ability to repay debts.

To test the hypothesis that an increase in debtor abuse accounts for the rise in bankruptcy filings, it is possible to compare the financial profiles of the debtors who filed over a seventeen-year period: 1981, 1991, 1995, and 1997. The 1981 data are from the Consumer Bankruptcy Project, a study my coauthors and I conducted of 1557 debtors in ten judicial districts. This study is reported in detail in our book, As We Forgive Our Debtors: Bankruptcy and Consumer Credit in America.⁽⁷¹⁾ The 1991 data are from the Consumer Bankruptcy Project II, a study by the same authors of 2400 debtors in sixteen districts for which we collected additional financial information for about 750 debtors in five of the districts. Details of this study are reported in Consumer Debtors Ten Years Later: A Financial Comparison of Consumer Bankrupts 1981-1991.⁽⁷²⁾ The 1995 data were developed by Professors Marianne Culhane and Michaela White as part of the Bankruptcy Reaffirmation Project. They collected data on more than a thousand Chapter 7 cases drawn from seven districts. A larger report of their findings will be forthcoming in later scholarly publications, but they were willing to make debt and income data available from their study to make the multi-year study possible.⁽⁷³⁾ The 1997 data come from the efforts of the Honorable Barbara Sellers, Bankruptcy Judge for the Southern District of Ohio. She provided detailed raw data from a sample of 150 cases filed in 1997 in the Southern District of Ohio, which I analyzed for this comparison.⁽⁷⁴⁾ The comparisons are illustrated in Figure 2 on the following page and reported in more detail in Table 1 of the Appendix.⁽⁷⁵⁾

Figure 2

The comparisons are not ideal. If we had had more funding, my coauthors and I would have collected data from more districts initially and we would have continued to collect data from these districts over time. The 1981 and 1991 data reported here are from the same three districts in Illinois, Pennsylvania, and Texas. The 1995 data are from California, Colorado, Georgia, Massachusetts, Nebraska, North Carolina, and Wisconsin. The 1997 data are from Ohio. Because we have both different districts and different time periods in some of the studies, it is difficult to determine whether the differences were caused because Ohio differed from Texas or whether a 1981 sample differed from a 1997 sample or both. Moreover, it is always difficult to rely on the analysis of a limited number of districts; the districts sampled may differ from all the other districts that were not sampled.⁽⁷⁶⁾ Nonetheless, the comparison among the eleven districts is a little like lighting a candle; it provides limited light and it is more productive than cursing the darkness.⁽⁷⁷⁾

Notwithstanding the limitations of the data, the suggestion of a trend is strong: since 1991, the incomes of the debtors in bankruptcy have declined. If there were growing widespread abuse, the debtors who filed in later years should be on average somewhat more prosperous than their counterparts who filed earlier. Instead, income among the pool of bankruptcy debtors declines significantly. These data strongly suggest that the debtors as a group are in worse financial condition when they file for bankruptcy, as least as measured by income.

Of course, poorer debtors might still abuse the system. If debtors' incomes were declining, but their debts were declining even faster, these debtors might be accused of taking "bankruptcies of convenience." Such a decline in debt relative to income might suggest that an increase in abusive filings accounts for the rise in bankruptcies. Instead, however, these data suggest that not only are the debtors' incomes declining, their debt-to-income ratios are not improving.⁽⁷⁸⁾

Figure 3

These data, although limited, are inconsistent with the hypothesis of increasing debtor abuse as an explanation of the rise in consumer bankruptcies. Debtors in 1995 and 1997 had lower incomes than their counterparts in 1991, who had lower incomes than their counterparts in 1981. The debtors' debt-to-income ratios have not improved. To the extent the pool of debtors in bankruptcy is changing, the data suggest that the pool of bankrupt debtors includes a growing number of low-income debtors. The debt picture is more difficult to read, but total debt remains statistically constant, although the composition of secured and unsecured debt may be changing as more debtors have a larger fraction of unsecured debt when they file for bankruptcy.

It would be too speculative to draw a firm conclusion from the limited data points in these microdata studies, but the available income data are consistent with reported changes in credit-granting policies. The growth in consumer credit in recent years has been largest among the poorest Americans. The Federal Reserve's Survey of Consumer Finance notes that while debt burdens are generally falling for families with incomes above $50,000, families with incomes below $10,000 are increasing their debt loads.⁽⁷⁹⁾ Banking industry analysts attribute rising credit card defaults in part to the increase in credit solicitations to lower-income Americans. "[T]he issuers, in our opinion, have chosen to extend credit to individuals in a lower stratum of the creditworthiness spectrum compared with prior cycles."⁽⁸⁰⁾ The fastest-growing credit issuers in the past five years have been companies that specialize in loans to borrowers with poor credit histories, the so-called subprime market.⁽⁸¹⁾ Although losses are substantial, effective interest rates of 18 to 40%⁽⁸²⁾ make this kind of lending profitable. Well-known credit issuers have entered the subprime credit market. In 1996, Ford Credit, for example, announced an "ambitious plan" to enter the subprime lending market.⁽⁸³⁾ Other mainstream lenders, including FDIC-insured banks, have entered this market as well.⁽⁸⁴⁾ As more creditors extend loans to lower-income people and to people out of work and otherwise in credit trouble, it would not be surprising if the bankruptcy pool expanded to include people with lower and lower incomes. The microdata do not prove this point, but they are certainly consistent with it.

The credit industry study collected information on incomes and debts in 1996. Notwithstanding the criticism leveled at the study by the GAO⁽⁸⁵⁾ and the CBO,⁽⁸⁶⁾ I would have included their data as another point in time in the comparison of debtors over time. The problem is that the CRC study does not report enough information to enable valid statistical comparisons with other data. The CRC study includes reports on both income and debts, which are essential to all of its calculations, but insufficient data, such as the number of valid responses in each cell and the standard deviation of the response, to make any valid comparisons with other data. Moreover, income and debt are disaggregated in ways that make direct comparisons problematic.

Another study employing analysis of data directly from the bankruptcy files comes from economics professor Ian Domowitz of Northwestern, along with Robert Sartain of Abt Associates, who analyzed information about bankrupt debtors from a sample drawn by the GAO in 1980 from cases filed in New York, Ohio, Kentucky, and California.⁽⁸⁷⁾ Domowitz and Sartain analyzed 827 cases, supplementing the information with data from the Survey of Consumer Finance, to see how bankrupt debtors differed from those in the population generally.⁽⁸⁸⁾ Among their conclusions:

In terms of the decision to declare bankruptcy, health problems leading to substantial medical debt are found to constitute the most important factor in assessing the impact of household conditions. The impact of medical debt on the population incidence of filings may not be large, however, in the sense that a small percentage of the population has such debt in large amounts relative to income.

The largest single contribution to bankruptcy at the margin is credit card debt relative to income. An increase in credit card debt from the population average to the level of a typical bankrupt household is predicted to cause a six-fold increase in the conditional probability of filing for protection.⁽⁸⁹⁾

Professor Domowitz also conducted a time-series analysis for cases filed through 1995, and he has spent some time looking at ability to repay. He describes his work as not a major study in itself, but instead as "confirmatory of evidence from other researchers."⁽⁹⁰⁾ He concludes that "Chapter 7 debtors able to repay substantially constituted a very small fraction of the debtor sample."⁽⁹¹⁾

The macrodata reported earlier provide a general explanation for the increase in consumer filings--the rise in filings accompanies the rise in consumer debt. This explanation is reinforced by the microdata examining the files of individual debtors. The available data does not support the leap from the rise in bankruptcy filings to the conclusion that we are in the midst of a bankruptcy crisis.

III. The Crisis

Table 1
Distribution of Income and Debts for Chapter 7 Debtors
in 1981, 1991, 1995 and 1997, Adjusted to 1997 Dollars⁽⁹²⁾

1981⁽⁹³⁾

Total Secured Unsecured

Distribution Income Debt Debt Debt

Mean 25,961 69,017 36,209 32,805

s.d. 16,661 101,258 66,983 62,157

25th percentile 15,107 16,601 2,182 8,168

Median 23,254 35,434 13,596 14,666

75th percentile 35,314 75,984 45,904 29,134

Valid N 715 916 917 915

Missing 202 1 0 2

1991⁽⁹⁴⁾ (Three Districts)

Total Secured Unsecured

Distribution Income Debt Debt Debt

Mean 21,444 57,384 26,992 30,094

s.d. 13,610 67,163 46,225 43,739

25th percentile 13,660 18,197 0 11,477

Median 18,807 30,336 6,906 19,550

75th percentile 27,089 73,442 32,977 31,991

Valid N 249 254 257 254

Missing 8 3 0 3

1995⁽⁹⁵⁾

Total Secured Unsecured

Distribution Income Debt Debt Debt

Mean 19,943 73,981 36,757 37,224

s.d. 12,550 97,809 73,548 52,185

25th percentile 11,981 19,836 0 13,213

Median 17,958 40,369 7,364 23,021

75th percentile 25,528 90,255 44,253 39,791

Valid N 1,033 1,035 1,035 1,035

Missing 2 0 0 0

1997⁽⁹⁶⁾

Total Secured Unsecured

Distribution Income Debt Debt Debt

Mean 17,765 51,994 19,895 30,230

s.d. 9,614 47,912 26,478 38,181

25th percentile 11,664 21,933 2,268 15,816

Median 17,652 37,455 9,217 19,915

75th percentile 23,496 62,091 28,050 32,025

Valid N 75 75 75 75

Missing 0 0 0 0

Statistical Tests (t-value)

Total Secured Unsecured

Income Debt Debt Debt

1997 vs. 1995

t-value 1.472 1.929⁽⁹⁷⁾ 1.976⁽⁹⁸⁾ 1.139

1997 vs. 1991 (3)

t-value **2.182² 0.648¹ 1.269 0.024

1997 vs. 1981

t-value ***4.186⁽⁹⁹⁾ 1.442 **2.095² 0.353

1995 vs. 1991 (3)

t-value 1.664¹ ***2.560³ **2.031² **2.011²

1995 vs. 1981

t-value ***8.605³ 1.100 0.171 1.706¹

1991 (3) vs. 1981

t-value ***3.853³ 1.728¹ **2.072² 0.652

Source: Consumer Bankruptcy Project, Phases I and II, The Bankruptcy Reaffirmation Project, The Honorable Barbara Sellers.

Table 2

Distribution of Total Debt/Income Ratio for Chapter 7 Debtors⁽¹⁰⁰⁾

1981⁽¹⁰¹⁾

Total

Debt/Income

Distribution Ratio

Mean 3.56

s.d. 11.89

25th percentile 0.71

Median 1.42

75th percentile 2.58

Valid N 686

Zero Income 29

Missing 202

1991⁽¹⁰²⁾ (Three Districts)

Total

Debt/Income

Distribution Ratio

Mean 2.82

s.d. 3.70

25th percentile 1.07

Median 1.70

75th percentile 3.22

Valid N 239

Zero Income 7

Missing 11

1995⁽¹⁰³⁾

Total

Debt/Income

Distribution Ratio

Mean 4.87

s.d. 23.61

25th percentile 1.23

Median 2.37

75th percentile 4.55

Valid N 988

Zero Income 45

Missing 2

1997⁽¹⁰⁴⁾

Total

Debt/Income

Distribution Ratio

Mean 2.61

s.d. 2.03

25th percentile 1.13

Median 2.32

75th percentile 3.08

Valid N 69

Zero Income 6

Missing 0

Total

Debt/Income

Statistical Tests (t-tests) Ratio

1997 vs. 1995 0.794

t-value

1997 vs. 1991 (3) 0.452

t-value

1997 vs. 1981 0.662

t-value

1995 vs. 1991 (3) 1.337

t-value

1995 vs. 1981 1.340

t-value

1991 (3) vs. 1981 0.945

t-value

Source: Consumer Bankruptcy Project, Phases I and II, The Bankruptcy Reaffirmation Project, The Honorable Barbara Sellers.

Table 3

Distribution of Total Nonmortgage Debt/Income Ratios

for Chapter 7 Debtors⁽¹⁰⁵⁾

1981⁽¹⁰⁶⁾

Total Nonmortgage

Debt/Income

Distribution Ratio

Mean 2.30

s.d. 7.52

25th percentile 0.57

Median 0.87

75th percentile 1.69

Valid N 685

Zero Income 29

Missing 203

1991⁽¹⁰⁷⁾ (Three Districts)

Total Nonmortgage Debt/Income

Distribution Ratio

Mean 1.93

s.d. 2.50

25th percentile 0.82

Median 1.25

75th percentile 1.93

Valid N 174

Zero Income 7

Missing 76

1995⁽¹⁰⁸⁾

Total Nonmortgage

Debt/Income

Distribution Ratio

Mean 3.61

s.d. 23.31

25th percentile 0.98

Median 1.55

75th percentile 2.76

Valid N 988

Zero Income 45

Missing 2

1997⁽¹⁰⁹⁾

Total Nonmortgage Debt/Income

Distribution Ratio

Mean 1.92

s.d. 1.29

25th percentile 1.05

Median 1.64

75th percentile 2.61

Valid N 69

Zero Income 6

Missing 0

Total Nonmortgage Debt/Income

Statistical Tests (t-tests) Ratio

1997 vs. 1995 0.602

t-value

1997 vs. 1991 (3) 0.033

t-value

1997 vs. 1981 0.419

t-value

1995 vs. 1991 (3) 0.949

t-value

1995 vs. 1981 1.420

t-value

1991 (3) vs. 1981 0.639

t-value

Source: Consumer Bankruptcy Project, Phases I and II, The Bankruptcy Reaffirmation Project, The Honorable Barbara Sellers.

1. Leo Gottlieb Professor of Law, Harvard Law School. Melissa Jacoby read and commented on an earlier draft of this paper, providing a number of insights that are reflected in the final version of this paper. Dr. Teresa Sullivan and Professor Jay Lawrence Westbrook, my longtime coauthors, not only helped produce much of the data that are central to the discussion in this paper, they also provided useful comments in the construction of this paper. I offer heartfelt thanks to all three and add the usual caveat that they are not to be blamed for my errors.

In addition, Professors Marianne Culhane and Michaela White were extraordinarily generous in permitting me access to their 1995 data from the Reaffirmation Project. The Honorable Barbara Sellers collected the raw data from the 1997 Ohio sample, which she offered to the National Bankruptcy Review Commission. She was kind enough to permit me to analyze the data and report on them here. The multiyear, multidistrict comparison reported in this paper would have been impossible without the extraordinarily valuable contributions of people who have labored hard to bring reliable data into the bankruptcy debates. For Professors Sullivan, Westbrook, Culhane, and White and Judge Sellers, I am in your debt.

2. See News Release, Calendar Year Shows Bankruptcy Filings up 27 Percent over 1995, Administrative Office of the U.S. Courts, Mar. 18, 1997 (on file with author).

3. See id. Although the Administrative Office gives both business and nonbusiness filings, the question of what constitutes a "business" is more complex than the clearly divided numbers would suggest. In our earlier work, for example, we found that in a sample of "nonbusiness" files in Chapter 7 and Chapter 13, that as much as 20% of the sample involved the debts of failed businesses. See Teresa A. Sullivan, Elizabeth Warren & Jay L. Westbrook, As We Forgive Our Debtors: Bankruptcy and Consumer Credit in America 111 (1989) [hereinafter As We Forgive].

4. Macrodata are aggregated data about large economic trends such as bankruptcy filing rates, consumer debt, inflation, unemployment, changes in the gross national product, and so on that correlate certain trends, such as the rise and fall in consumer bankruptcy filings, with other variables.

5. The Increase in Personal Bankruptcy and the Crisis in Consumer Credit: Hearing Before the Subcomm. on Admin. Oversight and the Courts of the Senate Comm. on the Judiciary, 105th Cong. 39 (1997) (statement of Kim J. Kowalewski, Chief, Financial and General Macroeconomic Analysis Unit, Macroeconomic Analysis Division, CBO).

6. See id.

7. Id.

8. Jagdeep S. Bhandari & Lawrence A. Weiss, The Increasing Bankruptcy Filing Rate: An Historical Analysis, 67 Am. Bankr. L.J. 1, 1 (1993).

9. See Ian Domowitz & Elie Tamer, Two Hundred Years of Bankruptcy: A Tale of Legislation and Economic Fluctuations (May 1997) (unpublished manuscript, on file with the Department of Economics at Northwestern University).

10. See id. at 1-5.

11. See id. at 37.

12. Lawrence M. Ausubel, Credit Card Defaults, Credit Card Profits, and Bankruptcy, 71 Am. Bankr. L.J. 249, 250 (1997).

13. Id. at 251.

14. Finance charges contributed 75.1% of credit card revenues of Visa and MasterCard issuers in the second quarter of 1992. See David S. Evans & Richard L. Schmalensee, The Economics of the Payment Card Industry 55 fig.3 (1993) (citing data from the Visa U.S.A., Inc. Profit Analysis Report (quarter ending June 30, 1992)); see also Lawrence M. Ausubel, The Credit Card Market, Revisited 24 n.31 (July 20, 1995) (unpublished manuscript, on file with author).

15. See Diane Ellis, The Effect of Consumer Interest Rate Deregulation on Credit Card Volumes, Charge-Offs, and the Personal Bankruptcy Rate, Bank Trends (Federal Deposit Ins. Corp., Washington, D.C.), Feb. 1998, at 1.

16. Id.

17. For a discussion of legislation-by-anecdote in the bankruptcy field, see Teresa A. Sullivan, Elizabeth Warren & Jay L. Westbrook, The Use of Empirical Data in Formulating Bankruptcy Policy, Law & Contemp. Probs., Spring 1987, at 195.

18. See generally Visa U.S.A., Inc., Consumer Bankruptcy: Causes and Implications (1996).

19. Id. at 1.

20. Id.

21. Id. at 13.

22. Id.

23. See id. Of course, some might see increased access to lawyers and awareness of legal rights as an indication that the system is functioning more nearly as it was initially intended.

24. Memorandum, "Evaluations of Three Studies Submitted to the National Bankruptcy Review Commission," from Kim J. Kowalewski to Brady Williamson, Chairman, National Bankruptcy Review Commission 1 (Oct. 6, 1997).

25. Id. Kowalewski notes that although Visa admits it cannot measure "social factors," Visa nonetheless concludes that a decline in social stigma, increased legal advertising, and increased availability to debtors who have declared bankruptcy in the past play an important role in bankruptcy filings. Kowalewski explains:

Visa bases its conclusion on the apparent statistical significance and positive sign of the coefficient on their variable USTREND [Visa's "social" variable]. However, this variable does not measure any of these social factors; it is a simple trend that bears no necessary relationship to these factors.

The figure on the next page displays the variable USTREND and the number of personal bankruptcy filings per 1000 adults, both from the Visa data set. Even though they had no observations on the various social factors they cited, Visa assumed that the social factors behaved in a surprising way: the factors were assumed to have been unchanged between the fourth quarter of 1980 and the first quarter of 1985 and between the second quarter of 1992 and the first quarter of 1996, but to have risen between these two periods. Visa's report does not explain how they arrived at this unusual assumption, but the close correspondence between the two series, particularly the dates when the trends in the two series change, strongly suggests that USTREND was constructed specifically to follow the trends in the personal bankruptcy rate. Consequently, Visa's conclusion about the importance of social factors is unfounded. By defining their social trends proxy to mimic the broad trends of the variable they were attempting to explain, they arrived at their conclusion by assumption.

Id. at 6-7 (footnote omitted). In the footnote, Kowalewski commented, "[i]t is interesting to observe that because the social trends variable is flat during 1995 and early 1996, Visa believes that their social factors played no role behind the increase in personal bankruptcy filings in that period." Id. at 6 n.4.

26. Visa officials responded angrily to the CBO analysis of their work. The Senior Vice President of Visa wrote the Chairman of the National Bankruptcy Review Commission demanding that the report from CBO be suppressed and charging that the analysis of the study had been made in secret. See Letter from Thomas A. Layman, Senior Vice President, Visa U.S.A., Inc., to Brady Williamson, Chairman, National Bankruptcy Review Commission (Oct. 14, 1997) (on file with author). Visa subcontractors who conducted the research also wrote to protest the criticisms. See Mark Lauritano, WEFA Group, A Rebuttal to Kim J. Kowalewski's "Evaluation on Three Studies Submitted to the National Bankruptcy Review Commission" Dated October 6, 1997 (no date) (unpublished manuscript, on file with author).

27. Microdata are data developed from debtors' files on a case-by-case basis and added together to create a statistical profile of the average debtor.

28. See Teresa A. Sullivan, Elizabeth Warren & Jay L. Westbrook, Consumer Debtors Ten Years Later: A Financial Comparison of Consumer Bankrupts 1981-1991, 68 Am. Bankr. L.J. 121, 128 tbl.1 (1994) (reporting on a 1991 empirical study of consumer debtors) [hereinafter Ten Years Later]. The text reports median income in order to be comparable to the census data, which report only median incomes. The mean income in 1981 was somewhat higher at $23,927. All the numbers have been recalculated to 1997 to adjust for the effects of inflation. To translate a number gathered in 1991 to 1997 dollars, it was multiplied by 1.1784141.

29. To calculate this, we relied on Bureau of the Census, U.S. Dep't of Commerce, Statistical Abstract of the United States 1994 at 469 tbl.715 (1994) (displaying table entitled "Money Income of Families--Median Income, by Race and Hispanic Origin, In Current and Constant (1991) Dollars: 1970-1992"), and we reported the data in 1997 constant dollars.

30. See Ten Years Later, supra note 27, at 129-33.

31. See id. at 135.

32. See id. at 128-39. The article includes detailed reports about the incomes, assets, debt, debt-to-income ratios, and other financial information about the debtors who filed in bankruptcy both in 1991 and 1981.

33. Actually, it may be too optimistic to estimate that debtors in bankruptcy would have to devote only every fifth paycheck to pay the interest on their outstanding consumer debt. Despite low inflation, average interest rates climbed at least one percentage point during 1997, putting the rate at 18.8%, according to one industry research firm, RAM Research. According to the same research company, penalty charges have gone up more than 50% during the year. For customers who have been late on the payments, as many bankruptcy debtors presumably have been, interest rates go as high as 30% in some banks. Associates National Bank in Dallas, for example, sets a 29.99% interest rate for its highest-risk customers, according to bank officials. See Karen Hube, Now, Time to Dig Out from 1997 Debt, Wall St. J., Jan. 2, 1998, at 17. If the debtors who file for bankruptcy had been using credit from the subprime market, they could have been facing interest rates as high as 40%. See Robyn Meredith, Will Ford Become the New Repo Man?, N.Y. Times, Dec. 15, 1996, at A1. If the debtors in bankruptcy were obligated to pay interest rates that exceeded the national averages, they would have to devote an even larger share of their income to interest charges.

34. See Ten Years Later, supra note 27, at 138 tbl.3. Again, the median debt-to-income ratio is reported in the text, this time for comparability with the income figures. The mean debt-to-income ratio is higher at 1.62, meaning that, at the mean, a debtor in bankruptcy owed about a year and seven months' income in short-term, high-interest debt.

35. The credit industry has paid the CRC to conduct a number of studies, such as a study on the effects of postbankruptcy credit on bankruptcy filings and the effect of waiving finance charges in credit counseling. Perhaps the best-known study commissioned by the credit industry from CRC is the 1981 study often referred to as the Purdue Report. See discussion infra note 66.

36. See John M. Barron & Michael E. Staten, Personal Bankruptcy: A Report on Petitioners' Ability-to-Pay (Oct. 6, 1997) (unpublished manuscript, on file with author).

37. Id. at 2.

38. See id. at 6.

39. I say "evidently" because the explanation of what was omitted from the expense category was not always clear. It would appear, for example, by reading the table of expenses that car loans were always included because of the entry made in Table 5. See id. at 17 tbl.5. It is only in the text several pages away that it is possible to infer that only reaffirmed debts were included in the expense calculation. See id. at 11-12.

40. Because the documentation is so slim, it is difficult to know exactly how the credit industry's researchers calculated expenses; I have made guesses based on the reported information.

41. See Barron & Staten, supra note 35, at 9-10.

42. See id. at 25 & tbl.11, 26 tbl.12.

43. See id. at 28 tbl.14.

44. Daniel McGinn, Deadbeat Nation, Newsweek, Apr. 14, 1997, at 50, 50.

45. Editorial, 'Last Resort' Is Coming First, L.A. Times, July 28, 1997, at B4.

46. Editorial, Lax Bankruptcy Laws Make Everyone Pay, USA Today, June 12, 1997, at 14A.

47. Lloyd Bentsen, Editorial, Get Tough on Bankruptcy Laws, Wash. Times, Sept. 19, 1997, at A19.

48. See Press Release, Statement of Chairman George W. Gekas, Judicial Subcommittee on Commercial and Administrative Law, Bankruptcy Reform Act of 1998, Office of U.S. Representative George W. Gekas, Feb. 3, 1998, at 3-4 (on file with author).

49. The study's report of the data are not entirely clear, but it appears that the yearly repayment is summed to show a total repayment of 13.7%, and the number reported in the text is calculated from this number. See Barron & Staten, supra note 35, at 26 tbl.12. The reported five-year repayment of 13.7% of the debt for 44% of the Chapter 7 debtors, works out to 2.7% per creditor for each of the nonhousing creditors (as classified by the report). This figure was reached as follows. There are 2441 Chapter 7 filers in the study, see id. at 7 tbl.2, of whom 44% can repay something, see id. at 25. Thus, 2441 * 0.44 = 1074 can repay. The mean repayment available is $6848, see id. at 26 tbl.12; thus the total that can be repaid is $6848 * 1074, or $7,354,752. The authors report that the $6848 payment represents 13.7% repayment over 5 years. See id. The $7,354,752 is 13.7% of $53,684,321, which we may take to be the unpaid nonhousing, nonpriority debt of the "can pay" debtors.

The figure of $53 million approximates a number that we can reach independently by adding up the mean nonhousing debt figures from Table 8 and multiplying by 1074: $5992 vehicle debt + $2885 other secured debt + $41,228 unsecured debt = $50,105 mean debt of the can-pays. See id. at 21 tbl.8. Multiplied by the 1074 debtors, the figure is $53,812,770. Duplicating the $53 million figure in this way suggests that the mean figures in Table 8 are intended to be applied more generally, so we will continue with a few more applications.

Continuing with the assumption that the mean debt figures presented in Table 8 apply to the entire sample and not just the can-pay debtors, then $50,105 * 2441 (total sample) yields a total nonpriority, nonhousing debt for the sample of $122,306,305 total debt owed. The repayment figure of $7 million is thus 6% of the $122 million owed. Spread over 5 years, see id. at 26 tbl.12, this amounts to 1.2% of the debt paid each year. In terms of the total debt owed, this sum is trivial.

In terms of the can-pay debtors, on the other hand, the repayment is considerable. Again working from the mean figures in the tables, this sum amounts to about $114 a month ($6848 over 60 months, or $1368 for 12 months) or about 8% of their mean net annual take-home pay of $16,470. See id. at 16 tbl.4. Looking at their annual budgets, see id. at 17 tbl.5, which already exceed the net take-home by $7739, it is not clear how the debtors could further economize, but here are some possibilities:

• they could eliminate most transportation ($1425);

• they could eliminate one-third of their food expenditures of $4052;

• they could eliminate laundry ($404) AND clothing ($846); or

• they could eliminate a telephone ($713) AND medical and dental expenses ($623).

50. Kowalewski, supra note 23, at 8.

51. See id. at 8-10.

52. Id. at 9.

53. See Letter from Michael Staten, Director, CRC, to Brady Williamson, Chairman, National Bankruptcy Review Commission 1 (Oct. 14, 1997) (on file with author) (explaining that he had been "informed" of the CBO review on October 6). The CBO review was made public on October 20 when the report of the National Bankruptcy Review Commission was published.

54. Memorandum, "Follow-Up Comments on Dr. Staten's Study," from Kim J. Kowalewski, to Brady Williamson, Chairman, National Bankruptcy Review Commission 3 (Oct. 10, 1997) (on file with author).

55. See U.S. Gen. Accounting Office, GAO/GGD-98-47, Personal Bankruptcy: The Credit Research Center Report on Debtors' Ability to Pay 1 (1998) [hereinafter GAO Report].

56. See id. at 2-6.

57. Id. at 6.

58. See, e.g., id. at 4-5.

59. Id. at 6.

60. See Letter from Ian Domowitz, Professor, Department of Economics and Institute for Policy Research, Northwestern University, to Brady Williamson, Chairman, National Bankruptcy Review Commission 3 (June 9, 1997) (on file with author) (questioning the study's calculations and concluding "Dr. Staten should make the full data set available to the Commission itself, which can then decide how best to go about verifying the results"); Letter from Professors Marianne Culhane & Michaela White, Creighton University School of Law, to Brady Williamson, Chairman, National Bankruptcy Review Commission 1 (June 11, 1997) (on file with author) (providing a detailed analysis of data collection and concluding that the "VISA study seriously overstates the repayment capacity of most chapter 7 debtors"); Statement of Professor William C. Whitford, University of Wisconsin School of Law, to the National Bankruptcy Review Commission (Jan. 23, 1997) (on file with author) (explaining deficiencies in the data in sampling and calculations).

61. See John M. Barron & Michael E. Staten, Response to GAO by Credit Research Center Authors, available in American Bankr. Inst., ABI World: Reports & Testimony (visited May 25, 1998) <http://www.abiworld.org/research/researchfront.html>.

62. See U.S. Gen. Accounting Office, GAO/T-GGD-98-79, Personal Bankruptcy: The Credit Research Center and Ernst & Young Reports on Debtors' Ability to Pay 6-9, 11-12 (1998) (testimony of Richard M. Stana, Associate Director, Administration of Justice Issues, General Government Division, U.S. General Accounting Office).

63. See WEFA Group, The Financial Costs of Personal Bankruptcy (1998).

64. See Letter from Richard M. Stana, Associate Director, Administration of Justice Issues, U.S. General Accounting Office, to Hon. Martin T. Meehan, U.S. House of Representatives 5 (Apr. 23, 1998) (on file with author).

65. For this review, there were no angry denials by the authors. The GAO reports, "We attempted to discuss our observations about the report with the WEFA report's authors, but were unable to reach anyone to do so." Id. at 1.

66. For a brief discussion of the credit industry's earlier efforts to persuade Congress to adopt a means test for all debtors in bankruptcy, see Elizabeth Warren, Reducing Bankruptcy Protection for Consumers: A Response, 72 Geo. L.J. 1333 (1984). The credit industry "simultaneously commissioned a study of bankrupt debtors, hired a prestigious Madison Avenue advertising firm to publicize the results of the study, and lobbied for federal legislation to remedy the problems that it was certain the study would reveal." Id. at 1333 n.4.

67. See 1 Credit Research Ctr., Purdue Univ., Consumer Bankruptcy Study 88-91 (1982). The figure was initially reported as $1.6 billion and then corrected to $1.1 billion after the report was published, but both numbers continued to circulate through the halls of Congress.

68. See Teresa A. Sullivan, Elizabeth Warren & Jay L. Westbrook, Limiting Access to Bankruptcy Discharge: An Analysis of the Creditors' Data, 1983 Wis. L. Rev. 1091.

69. Id. at 1145 (footnote omitted). One of the CRC researchers responded to the critique, claiming that the study's estimate that $1.1 billion in debt is discharged that could be repaid was defensible. See A. Charlene Sullivan, Reply: Limiting Access to Bankruptcy Discharge, 1984 Wis. L. Rev. 1069. We argued in a rejoinder that the CRC's own data seriously undermined the study's widely publicized claims and that the defense of the study did not restore its credibility. See Teresa A. Sullivan, Elizabeth Warren & Jay L. Westbrook, Rejoinder: Limiting Access to Bankruptcy Discharge, 1984 Wis. L. Rev. 1087 [hereinafter Rejoinder].

70. For a discussion of the connection between the credit industry's 1981 study and the subsequent adoption of creditor-sponsored amendments to the bankruptcy code, see Rejoinder, supra note 68, at 1087-103.

71. As We Forgive, supra note 2. The ten districts studied were the Northern District of Illinois, the Central District of Illinois, the Southern District of Illinois, the Eastern District of Pennsylvania, the Central District of Pennsylvania, the Western District of Pennsylvania, the Eastern District of Texas, the Western District of Texas, the Southern District of Texas, and the Northern District of Texas.

72. Ten Years Later, supra note 27. The 16 districts studied were the 10 of the 1981 study plus the Northern District of California, the Central District of California, the Eastern District of California, the Southern District of California, the Middle District of Tennessee, and the Western District of Tennessee.

73. Professors Culhane and White randomly sampled cases from seven districts: the Northern District of California, the District of Colorado, the Northern District of Georgia, the District of Massachusetts, the District of Nebraska, the Middle District of North Carolina, the Western District of Wisconsin. The researchers analyzed 1050 cases either filed or later converted to Chapter 7. The sample was limited to cases filed by an individual or married couple in calendar year 1995, in which schedules had been filed. For the analysis presented here, outliers were eliminated from the sample following the same protocols as Sullivan, Warren, and Westbrook in As We Forgive, supra note 2, with dollar figures adjusted for inflation, to enhance the comparability of the data. See Marianne B. Culhane & Michaela M. White, Preliminary Results of the Bankruptcy Reaffirmation Project 3 (Sept. 25, 1997) (unpublished manuscript, on file with author); data reported infra on tables in Appendix.

74. The 1997 database was constructed from data supplied by the Honorable Barbara Sellers. She selected the first 100 Chapter 13 cases and the first 50 Chapter 7 cases assigned to her during 1997. Every case in that group was studied. Assignment of cases in the Southern District of Ohio is random among the judges. Judge Sellers used only her own cases in order to eliminate any differences in outcomes that might be influenced by the judge.

Judge Sellers recorded about 75 pieces of information for each case. We received her handwritten sheets of information in June 1997. Following the same protocols that were used to code and report data from the Consumer Bankruptcy Project Phase I and Phase II, we coded the data into machine-readable form. The codes for the reasons for bankruptcy are on file with the authors.

Becase Judge Sellers's selection of cases produced a disproportionate share of Chapter 13 cases, analysis of the data from this district was weighted in the 75-25 proportion of Chapter 7 to Chapter 13 cases that existed in 1997 in her district. In order to enhance the accessibility of our data for other researchers, we did not sample the Chapter 13 cases, employing instead a straightforward weighting process. For this report, however, the comparison is among Chapter 7 cases exclusively, so there is no need to weight the data.

75. Figure 2 is a graphic representation of the data presented in more detail in Table 1 in the Appendix.

76. Although it is more expensive, we are able to correct for another statistical problem: the possibility of inaccurate sampling of the district chosen. In each district we sampled, we took proportionate cases from all places in which cases were filed. This means, for example, that our sample of cases from the Northern District of Texas includes cases from Dallas, Ft. Worth, and Lubbock. Any researcher who goes only to the major city risks biasing the sample to represent only filers in large metropolitan areas and not their counterparts in more remote locations who may, in fact, differ sharply. This appears to be one of many problems with sample selection in the most recent credit industry study which appears to be based on samples from only one location for each district. Barron & Staten, supra note 35, at 7, seems to say that all data collections took place in the identified major cities, thereby eliminating any cases from smaller cities and towns. We have also studied seasonality biases, nonresponse biases, and statistical outliers. These technical problems are discussed in Ten Years Later, supra note 27, at 147-54.

77. One direct comparison can be made. Three of the ten districts reported on in 1981 were also reported on in the 1991 study. The three districts are the Northern District of Illinois (Chicago), the Eastern District of Pennsylvania (Philadelphia), and the Western District of Texas (San Antonio). The results of these comparisons are reported in Ten Years Later, supra note 27. The findings reported there are consistent with the findings reported below, suggesting--but certainly not establishing--that the differences observed are differences over time rather than differences among districts. For more discussion of interdistrict variation, see id. at 151-54.

78. Figure 3 is a graphic representation of the data presented in more detail in Table 3 in the Appendix. Table 2 in the Appendix is reproduced to demonstrate the patterns when total debt is compared with income.

79. See Arthur B. Kennickell et al., Family Finances in the U.S.: Recent Evidence from the Survey of Consumer Finance, 83 Fed. Reserve Bull. 1, 21 & tbl.14 (1997).

80. George M. Salem & Aaron C. Clark, Bank Credit Cards: Loan Loss Risks Are Growing, Banking Industry Rep. (Gerard Klauer Mattison & Co., New York, N.Y.), June 11, 1996, at 7.

81. See Dow Jones & Co., Shrinking Profits Drive Mortgage Banker into B&C Loans, Cap. Markets Rep., Oct. 7, 1997, available in WESTLAW, Allnewsplus Database (reporting an increase from $80 billion in subprime loans in 1992 to $150 billion in 1996).

82. See Robyn Meredith, Will Ford Become the New Repo Man?, N.Y. Times, Dec. 15, 1996, § 3, at 1.

83. Id.

84. See Barbara A. Rehm, In a First, FDIC Warns Banks About Dangers of Subprime Lending Series: 5, Am. Banker, May 13, 1997, at 2.

85. See GAO Report, supra note 54, app. I, at 37-46.

86. See Kowalewski, supra note 23, at 1.

87. See Ian Domowitz & Robert L. Sartain, Determinants of the Consumer Bankruptcy Decision 1 (Jan. 1997) (unpublished manuscript, on file with the Department of Economics at Northwestern University).

88. See id. at 7. Professor Domowitz responded to criticisms from Dr. Staten about the "age" of his data in a detailed letter to the National Bankruptcy Review Commission. See Domowitz, supra note 59. Professor Domowitz pointed out that his work, unlike Dr. Staten's, covers a long time period that includes the "ebbs and flows of filings with the business cycle." Id. at 2.

89. Domowitz & Sartain, supra note 86, at 2.

90. Domowitz, supra note 59, at 2.

91. Id.

92. Cases with extreme values on assets, total debt, or income are removed. Outliers removed are beyond the cut off points used in As We Forgive, supra note 2, adjusted to 1997 dollars using the Consumer Price Index. To adjust for inflation, the original 1981 figures were multiplied by 1.7656773. The original 1991 figures were multiplied by 1.1784141. The original 1995 figures were multiplied by 1.0531496. Adjusted cut-off points to determine outliers were Income > 114,769; Assets > 882,838; Total Debt > 882,838.

93. See As We Forgive, supra note 2; 1981 data reported in 1997 dollars.

94. Consumer Bankruptcy Project, Phase II; 1991 data reported in 1997 dollars.

95. The Bankruptcy Reaffirmation Project; 1995 data reported in 1997 dollars.

96. Ohio Data (Judge Sellers); 1997 data reported in 1997 dollars.

97. Significant at p < .10 (t > 1.645).

98. Significant at p < .05 (t > 1.960).

99. Significant at p < .01 (t > 2.326).

100. Cases with extreme values on assets, total debt, or income are removed. Outliers removed are beyond the cut off points used in As We Forgive, supra note 2, adjusted to 1997 dollars using the Consumer Price Index. Adjusted cut-off points to determine outliers were Income > 114,769; Assets > 882,838; Total Debt > 882,838.

101. See As We Forgive, supra note 2.

102. Consumer Bankruptcy Project, Phase II.

103. The Bankruptcy Reaffirmation Project.

104. Ohio Data (Judge Sellers).

105. Cases with extreme values on assets, total debt, or income are removed. Outliers removed are beyond the cut off points used in As We Forgive, supra note 2. Adjusted cut-off points to determine outliers were Income > 114,769; Assets > 882,838; Total Debt > 882,838.

106. See As We Forgive, supra note 2.

107. Consumer Bankruptcy Project, Phase II.

108. The Bankruptcy Reaffirmation Project.

109. Ohio Data (Judge Sellers).