Home News RoomBankruptcy Research CenterBankruptcy Reports, Research and Testimony General Testimony

• Abstract
• Appendix

1. Introduction and Summary

This study calculates the effects of the needs-based bankruptcy provision of H.R. 3150, the "Bankruptcy Reform Act of 1998," on Chapter 7 bankruptcy filers with a new database of bankruptcy petitions. The analysis also replicates, with the new data, the Georgetown methodology of calculating petitioners’ ability-to-repay their non-housing debt.³ The additional data used is a sample of 5,722 Chapter 7 bankruptcy petitions filed in four bankruptcy courts, mainly during 1992 and 1993. The Courts studied were Boston, Chicago, Los Angeles, and Nashville.  ⁴

The needs-based provision considered requires petitioners with incomes above 75% of the national median to file under Chapter 13, if they can repay at least 20% of their unsecured non-priority debts within 5 years, after making secured and priority debt payments. In the new database, 8-14 percent of Chapter 7 filers from four cities would have been impacted by the needs-based provision and required to file Chapter 13 --had it been in effect in 1992-93.

The Ernst and Young analyses provided here corroborate the Georgetown study finding that "a sizable minority of Chapter 7 debtors could make a significant contribution toward repayment of their non-housing debt over a 5-year period."⁵ This outcome is important because the new database is from a different time period, and includes two different cities than those chosen by professors Barron and Staten.⁶

This study is organized into seven sections, beginning with this introduction and ending with an appendix. Section 2 gives a general description of the bankruptcy petition database used in the new analysis. Section 3 presents ability-to-repay results under the needs-based bankruptcy provision of H.R. 3150 and a city-by-city impact on Chapter 7 petitioners. Section 4 summarizes the Georgetown methodology for calculating ability to repay non-housing non-priority debt -- including modifications necessary to apply their concept to the new database. Section 5 presents the actual Georgetown concept applied to the new database and compares the results with those obtained in the Barron and Staten paper. Additional observations about data collection methods are made in Section 6; this is followed by a brief concluding section (Section 7). Finally, further details on the methods used to calculate repayment ability under the needs-based provision of H.R. 3150 are presented in the appendix; the appendix also contains in-depth tabulations on the impact of the needs-based provision by city and for all four cities combined.⁷

2. Description of the New Bankruptcy Petition Sample

The courts chosen in the study were selected to contrast practices in different regions of the country, with different caseloads and legal settings: Los Angeles, in the Central California District, is the court with the largest caseload nationally and therefore a natural to include.⁸ Chicago, from the Northern Illinois District, was taken to "represent" the Midwest. Nashville "represents" a smaller, less urban court and is a court with a disproportionately high fraction of Chapter 13 filings. Boston, a court of moderate size, was chosen to "represent" the east coast.⁹ Courts in Florida and Texas were also considered, but for various logistic and cost reasons the collection of petitions from these districts was not completed.

The U.S. Federal Trustee (regional administrative) Offices were the source of the petitions obtained for the study. To keep tight control of the process, all cases were taken sequentially during the data collection period until the desired number was attained (Table 1).¹⁰

Table 1. Data Collection Period and Sample Sizes by Court

To reduce the cost of data collection, petition copies were obtained after the trustee office was finished with them. They were routinely retrieved by an outside vendor and sent to Visa, where the data entry took place on site.

Controls were employed in the key entry step and numerous data editing steps were taken to assure a high quality result. Not all the petitions selected were found to be usable. Some petitions were "face sheet" only filings and hence had no useful petition information; others were incomplete to the extent that the financial condition of the petitioner could not be determined and also had to be dropped.¹¹

One final point (also covered in Section 6) -- all the reported data in the study were taken from the actual petitions filed by debtors. As a rule, these data are unaudited unless the petition is challenged by a creditor or trustee. Of course, the Chapter 7 filers whose petitions are examined here may have an incentive to make themselves appear insolvent by overstating expenses and understating income, especially when they are filing a "no asset" case.¹² This may occur where the Courts have elected not to dismiss the petition under the substantial abuse provision of Section 707(b), even if the debtor has an obvious ability to repay a substantial portion of his or her debt.

3. Repayment Ability in the New Database under H.R. 3150’s Needs-based Provision

The analysis in this section focuses only on the needs-based proposal of H.R. 3150, which requires petitioners with monthly income of at least 75% of the national median for families of comparable size to file under Chapter 13, if they have "monthly net income"¹³ of at least $50 and the capacity to repay 20% or more of their unsecured non-priority debts within 5 years, after making secured and priority debt payments.¹⁴ Table 2 on the next page summarizes what was found. Charts 1 and 2 show the same information graphically.

To construct Table 2, the Visa petition database was used to calculate the ability of bankruptcy petitioners to repay their debt under the needs-based¹⁵ bankruptcy provision. H.R. 3150’s needs-based provision uses living expenses calculated from the U. S. Internal Revenue Service’s Collection Financial Standards¹⁶ to calculate repayment ability as a percentage of unsecured non-priority debts. As already noted, in addition, the needs-based proposal has an exemption for families earning less than 75% of the national median income, so no repayment ability is calculated for all such families. The methodology used to calculate repayment ability under H.R. 3150’s needs-based provision is described in the appendix.

1/ Repayment rates from the 1991-1993 Visa database, calculated on the basis of: petitioners' reported income; mortgage and alimony payments; geographically determined IRS allowances for rent, transportation & living expenses; and 60 month interest-free amortization of outstanding priority debt and secured non-mortgage debt. See assumptions in text.

2/ Chapter 7 filers are impacted by H.R. 3150’s needs-based bankruptcy proposal if they have income > 75% of national median income for families of comparable size, net monthly income¹⁷ > $50, and can repay at least 20% of their unsecured non-priority debt over 60 months after making secured and priority debt payments. The 1992 and 1993 national median income data by family size were obtained from the U.S. Bureau of the Census. Income was assigned using average family size for families comprised of married couples; and individuals and single parent families.

3/ Repayment of secured debt is shown over 60 months. For housing debt, repayment is taken as stated by the petitioner unless the outstanding balance would have been paid off in less than 60 months. All other non-housing secured and priority debt is amortized over 60 months. In practice, petitioners may continue to repay outstanding secured debts beyond the 60 month repayment plan, which would increase the share of secured and total debt repaid.

4/ Average was weighted by 1993 annual Chapter 7 filings in the four courts: 11,921 (Boston), 21,738 (Chicago), 74,967 (Los Angeles), and 4,774 (Nashville). Judicial Business of the U. S. Courts, Report of the Director Admin. Office of the U. S. Courts, op. cit.

As shown in Table 2 (and Chart 1), from 8-14 percent of Chapter 7 filers in the four cities studied would have been required to file under Chapter 13, if H.R 3150’s needs-based bankruptcy provision had been in effect in 1992 and 1993. These filers would have had the ability to repay 55 to 72 percent of their total debts, if income remained unchanged relative to expenses and liabilities during the 60 month repayment period.¹⁸ When broken down by type, the corresponding repayment figures are 52-71 percent for secured and priority debt owed and 63-85 percent for unsecured non-priority debt owed. Table 2 also presents the average share of Chapter 7 filers (11.8%) and debt repayable (62%) across the four cities. (More details on the distribution of repayment ability across debtors are provided in Tables A1 through A5 in the appendix.)

The estimation of the impact of the needs-based bankruptcy proposal in this paper is subject to a number of limitations and assumptions. The major ones, in addition to those usual to sample studies,¹⁹ are:

2. The calculations use petition data submitted by petitioners -- that is, all data were taken from the actual petitions, as filed by debtors. While submitted under oath, these data are unaudited, unless the petition is challenged by a creditor or trustee. As noted elsewhere, the Chapter 7 filers whose petitions are examined here may have an incentive to exaggerate their financial distress by overstating expenses and understating income. In addition, the data do not account for amendments made by debtors to their petitions.

3. The repayment calculations assume that the petitioners’ future income relative to expenses and liabilities during the five year period is the same as in the 12 months prior to declaring bankruptcy. For petitioners whose income relative to expenses and liabilities would be higher over the five year period, this assumption would underestimate repayment ability. Conversely, the assumption would overestimate the repayment ability of petitioners whose income relative to expenses and liabilities would be lower during the five year period.

4. The repayment calculations assume all secured and priority debt is paid first, whether reaffirmed or not. That is, repayment ability is measured as a share of total unsecured non-priority debt.

5. To make the calculations using the new database, some imputations were required due to lack of information. These include imputations for family size and housing debt. ²⁰
6. The repayment calculations do not include potential administrative costs incurred in repaying debt. While these costs could somewhat reduce repayment ability, many administrative costs are incurred before the bankruptcy filing. Consequently, it is thought that administrative costs would impact ability to repay debt for a minimal number of filers. This issue is discussed more fully in the appendix.

4. Georgetown Repayment Ability Measure

This section briefly describes the Georgetown repayment ability methodology -- first, as originally employed in their research and then, as modified by us.

The original Georgetown repayment ability measure is based on the following formulas:

A detailed explanation, including the key assumptions underlying the repayment capacity, is included in the Georgetown paper. The ability to repay non-housing, non-priority debt was calculated using a methodology that replicated, to the extent possible using the new database, the Georgetown calculation.²³

Three modifications were made to the Georgetown calculation when applying their concept to the new database: (1) Annual net income was calculated based on the net take-home wage plus non-wage income, less other Federal taxes paid. The new database only included take-home wage income net of all payroll deductions, including some saving deductions such as 401(k) contributions, while the Georgetown calculation did not subtract out saving deductions. Georgetown estimated the federal tax liability on other income to calculate the current tax, while the new database calculation takes the reported other Federal income taxes paid, which may include taxes from prior years. It should be noted that, to the extent either of these approximations would consistently influence the estimates, both would tend to lead to an understatement of the ability to repay. (2) Housing debt was estimated for the new database calculation, since details on liabilities were limited to total debt, secured debt and unsecured debt. Housing debt was estimated by multiplying secured debt by the ratio of real assets to total assets. As shown in Charts 4 and 5 (in Section 5), this approximation worked well²⁴ for these data. (3) Priority debt for the new database calculation included back taxes, alimony and student loans. Some priority debt items, included by Georgetown -- such as past-due child support -- were not available from the new database. Everything else being equal, this last approximation would lead to an overstatement of the repayment capacity.

5. Georgetown Repayment Ability in the New Database

Findings about the ability to repay non-priority, non-housing debt by Chapter 7 petitioners from both the new database and the Georgetown sample are presented in Table 3 and Chart 3. The new database estimates of repayment ability support the findings of Barron and Staten that a "sizable minority" of Chapter 7 petitioners could repay some or all of their non-housing liabilities.

The calculations of Chapter 7 petitioners’ repayment ability from the new database are somewhat higher, in most cases, than those reported in the Georgetown study. For example, Barron and Staten report that 25% of Chapter 7 filers in their 1996 database had the ability to repay at least 30% of their debt. In the new database, the corresponding figures range upwards from 33% for Boston to 55% for Chicago. (The bolded line in Table 3 has the new database repayability ratios for each court at the 25% point.)

6. Data Limitations

The Visa Bankruptcy Petition Study was conducted with considerable care. Nonetheless, as in virtually all empirical investigations, the effort suffers to some extent from sampling and nonsampling limitations. In this section, an attempt to quantify them is given. Specifically discussed are (1) missing information and reporting errors on the petitions used, (2) undetected data keying errors, (3) sampling errors and biases, and (4) the extent to which the results can be generalized given the courts chosen and the periods of sampling used.

(1) Reporting errors and missing information. -- The repayment ability measures used with the new database rest on a number of assumptions. These assumptions, including taking reported information from the petitioners’ statements at face value, are described in the Georgetown report or for the needs-based proposal in the appendix. In the Visa Bankruptcy Petition Study, as was the case with the Georgetown work, all the data were taken from the actual petitions as filed by debtors. These data are unaudited, unless the petition is challenged by a creditor or trustee. Again, the Chapter 7 filers, whose petitions are examined here, may have an incentive to make themselves appear insolvent by overstating expenses and understating income.

 CHART 4

 CHART 5

For the Georgetown repayment calculation, adjustments had to be made because of data availability differences between the two databases. The item content of the new database was less extensive than in the Georgetown data. To calculate repayability ratios for the new data, some items obtained by Georgetown had to be imputed. A notable example was housing debt. Since Ernst & Young had access to the Georgetown data for Chicago and Los Angeles, the effect of the housing debt imputation was tested. Charts 4 and 5 show what a small difference occurred.²⁵

Another way to examine the computational compromises forced on the Visa Bankruptcy Petition Study is to directly compare what the repayability ratio would be in the Georgetown data, if they had used the new database imputations. This has been done in Table 4 below.

The last column in Table 4, as in Table 3, is from the Georgetown report but the Chicago and Los Angeles figures were calculated by Ernst & Young using data from the Georgetown database. Ernst & Young’s repayment ability calculations for Chicago and Los Angeles used the same imputations that had to be employed in the Visa Bankruptcy Petition database. Since the calculations for Chicago and Los Angeles in Tables 3 and 4 use the same methodology, comparisons between these Tables for Chicago and Los Angeles are due mainly to sampling errors and timing differences.

Again, notice from Table 4 that 25% of Chapter 7 filers in the Georgetown 1996 database had the ability to repay at least 30% of their debt (48% for Chicago and 33% for Los Angeles). In the 1991-93 Visa data base, the corresponding figures were 55% for Chicago and 43% for Los Angeles. (The bolded line in Table 4 shows, as in Table 3, the repayability ratio for each court at the 25% point.)

(2)Data Capture Errors -- Tight controls were employed in the key entry step for the new database and numerous data editing steps were taken to assure a high quality result. Employees performed data entry after they were trained on a Paradox Form data entry system. The Paradox data entry form included built-in field checks for each data item using the following methods: reasonable range checks for each entry, table-driven valid responses for each entry, and summation fields checked (against the individual items entered for expenses, income, assets, secured debt, and unsecured debt).

This approach minimized data entry errors throughout the process. In addition, extra "summation" fields were placed on the entry form for use in validating the information prior to analysis. All data fields were further examined for validity during the analysis phase itself. Whenever data entry errors were detected, an attempt was made to correct the fields. If there was no way to determine the correct entry, the petition was eliminated from the analysis.

Despite the evident care taken in mounting the new database, keying and other data capture errors occurred. Since original source documents were retained, data capture problems were measured by comparing keyed information to the source documents. A small sample was drawn to do the required testing, so this source of uncertainty could be quantified. We found an error rate of about 1% in this test of the final file. The errors found exhibited no systematic pattern; in fact, because the errors tended to cancel each other out, the mistakes found had little discernible effect on the repayability ratio.

(3) Sampling Errors –Block samples were taken of petition filings over time in each of the four courts in the study, as displayed in Table 1. Because the samples are not strictly random, we cannot calculate error bounds on the estimates produced without making assumptions. Still, the samples chosen are large and hence one would expect statistically stable results.

To quantify this intuition, we divided the samples in each court into ten random subsamples or "replicates." This gives a ready way of seeing the variability in the repayability ratio, for example (under the assumption that the observations are independent and identically distributed).

Because the filings in any single block of days might be more alike than a random set of days, the sampling error measures from the replicates were examined for this potential lack of independence, using time series methods (see Chart 6). The autocorrelations found were so weak ²⁶ that in the end we made no adjustment for this possibility.

In any case, the margin of error calculations developed showed the following 95% confidence bounds for the repayability ratio of the top 25% of filers - - the group we continue to use for illustrative purposes:

Notice that, except for Nashville, the large samples taken provided tight error bounds. Even for Nashville, moreover, all the repayability ratios are highly significantly different from zero. Tests for differences by court are also significant between Boston and Chicago, Chicago and Los Angeles, and Los Angeles and Boston. There is no statistically significant difference between Nashville and any of the other courts.

Again, using the same assumptions as above, we can test in Chicago and Los Angeles to see if there are statistically significant differences between the Georgetown data in Table 4 and the new data base from Table 3 ²⁷ None were found to exist -- providing further evidential weight for what has been said earlier about the strength of the corroboration between the two studies.

(4) Generalization – Procedural decisions made in choosing the four courts do not allow us to make statistically-based statements about bankruptcy petitioners in general. The results, by themselves, cannot technically be used to project to other courts or even other time periods in the same court (without assumptions being made).

The approach taken by the Visa study (and by the Georgetown work, as well) can be characterized as an experimental design, rather than an enumerative sampling one. This distinction is important in how the data can be used and what role they could play in modeling the U.S. bankruptcy process.

To have chosen courts for a nationally representative sample, a multistage stratified probability approach would be usual.²⁸ The courts were chosen for the Visa study purposively -- in keeping with a fixed analysis of variance or experimental design approach.²⁹ Incidentally, four courts, even if chosen as part of a probability sample design, would generally be too few to achieve stable national probability sample estimates. Still, combined with the Georgetown results, 15 courts have been studied -- all having sizable percentages of Chapter 7 petitioners who appear to be able to repay a large portion of their debts.

The Visa data are different from those collected by professors Barron and Staten in that they were obtained over a long period of time in each of the four courts studied. This affords some protection against potential periodicity in repayability. In light of observations made elsewhere³⁰ we tested for seasonality in the new database. Chart 6 plots repayment ability against time in each of the courts. There appears to be no evidence of seasonality, day of the month or day of the week effects.

7. Concluding Comments

We believe that the new data have considerable value. For the four cities examined, H.R. 3150’s needs-based provision ³¹ would have impacted from 8 to 14 percent of Chapter 7 filers in the study. Moreover, these filers could have repaid 55 to 72 percent (average 62 percent) of their total debts over 5 years. When broken down by type of debt for filers impacted, the corresponding repayment figures are 52-71 percent for secured and priority debt owed, and 63-85 percent for unsecured non-priority debt.

Second, the new data definitely corroborate the Georgetown results that "a sizable minority of Chapter 7 debtors could make a significant contribution toward repayment of their non-housing debt over a 5-year period." The corroboration occurs because we have been able to add to the courts and time periods where evidence for sizable repayability exists. In our view, the combined results are suggestive of a potentially widespread phenomenon.

Footnotes

¹ This study was funded by Visa U.S.A. and MasterCard International.Return to Text

² John. M. Barron and Michael E. Staten, "Personal Bankruptcy: A Report on Petitioners’ Ability to Pay" (October 6, 1997).Return to Text

³ The full details of the Georgetown methodology are found in Barron & Staten’s paper, entitled "Personal Bankruptcy: A Report on Petitioners’ Ability-to-Pay," dated October 6, 1997, and will not be repeated here. To provide a context of our new results, however, we have provided in the body of the paper (see Section 4), a definition of the Georgetown concept and how it was operationalized. Direct numerical comparisons of our results with Georgetown are given too (in Section 5).Return to Text

⁴ The data base is described fully in the report "Bankruptcy Petition Study" by Visa USA, Inc., June 1997. A brief description is also provided in Section 2 of this paper.Return to Text

⁵ Barron and Staten, op. cit.Return to Text

⁶ The Georgetown work was based on data from 13 courts, two of which – Chicago and Los Angeles – were the same as in the Visa data, albeit as noted for a different period.Return to Text

⁷ The cities were combined in keeping with the 1993 petition volumes in each of the four courts studied. The next footnote provides these counts. Return to Text

⁸ One of the main sources for this statement is the publication series, Judicial Business of the U. S. Courts, Report of the Director Admin. Office of the U. S. Courts. Table F-2 in the report shows that Chapter 7 volumes for 1995 in the four selected courts are 13,473 (Boston), 27,682 (Chicago), 79,075 (Los Angeles), and 5,378 (Nashville). The overall 1995 Chapter 7 volume was 761,652. Comparable figures for 1993 are 11,921 (Boston), 21,738 (Chicago), 74,967 (Los Angeles), and 4,774 (Nashville). The overall 1993 Chapter 7 volume was 602,907.Return to Text

⁹ The word "represent" has been put in quotes because the actual choice of bankruptcy districts was done purposively, using what might be characterized as an experimental design rather than a randomization-based sampling approach. More is said about this in Section 6. Return to Text

¹⁰ For Los Angeles, the sample sought was 2,500. More initially were pulled and subsampling was needed to obtain the values in Table 1. There was also some subsampling for the Boston court. While the collection of petitions at the Trustee Offices seems basically to have worked well, in Boston and Los Angeles, filings from satellite courts were generally not made available.Return to Text

¹¹ About 5% of the Chapter 7 petitions selected had financial information that was too incomplete to use. Return to Text

¹² About 95% of Chapter 7 filings nationally are "no asset" cases. A "no asset" case refers to the situation where there are no funds remaining to be dispersed to creditors, after the debtor is allowed to keep all exempt assets.Return to Text

¹³ Monthly income less certain IRS standard expenses and monthly payments for secured and priority debts.Return to Text

¹⁴ A detailed description of the methodology used for the repayment calculations is provided in the appendix, along with an extensive set of detailed tables measuring impacts. Return to Text

¹⁵ The needs-based bankruptcy repayment calculation is similar to, but with different rules than, the non-housing, non-priority debt repayment calculation presented by Professors Barron and Staten (see Section 4). Return to Text

¹⁶ See appendix for the details and a full citation.Return to Text

¹⁷ Monthly income less expense certain IRS standard expenses and monthly payments for secured and priority debts.Return to Text

¹⁸ However, it should be noted that these figures understate the repayability of the total debt over the life of the loan. For example, a petitioner with a mortgage would likely continue to make payments after the five year period called for in the needs-based proposal. Return to Text

¹⁹ In Section 6, there is a full discussion of the procedures used in processing the petitions and the quality impact that may have had on the results, plus an assessment of the importance of sampling error.Return to Text

²⁰ For more detail on the imputations and the repayment ability methodology, see the appendix.Return to Text

²¹ All expenses, including housing costs, excluding taxes and installment debt payments.Return to Text

²² All unsecured priority claims.Return to Text

²³ A detailed explanation, including the key assumptions underlying the repayment capacity, is included in the Georgetown paper. Barron and Staten, op. cit.Return to Text

²⁴ As part of the research, a portion of the Georgetown data – specifically, for the two courts common to both studies (Chicago and Los Angeles) -- was made available to Ernst and Young, with all identifiers removed. This file made possible some direct comparisons of differences between the estimates of housing debt and what difference arose as a result (See Section 6).Return to Text

²⁵ Using the Georgetown data, in Chart 4 a scatterplot is provided of the imputed mortgage debt versus the actual mortgage debt (where imputed mortgage debt was obtained by multiplying total secured debt by the ratio of real estate assets to total assets). Notice that the great bulk of the points lie close to (or on) the 45⁰ line of agreement. Chart 5 carries the analysis one step further and compares in another scatterplot what agreement exists between the repayability ratios themselves, when imputed or actual mortgage debt is used in the calculation. As earlier in Chart 4, most of the points lie close to (or indeed on) the 45⁰ line of complete agreement.Return to Text

²⁶ A look at Chart 6 provides a sense of this. The time series plots show great stability over time and, after adjusting for a court effect and for differences in sample sizes, appear to be almost identical and with little evidence of a trend or pattern of any sort. This is unlike the number of petitions, which have been rising steeply over time.Return to Text

²⁷ To do this testing we had to take into account the fact that the Georgetown samples for Chicago and Los Angeles were only a little over 200 observations each. This means that, ceteris paribus, the margin of error would be roughly 2 times greater that that for the Visa data for Chicago and over 3 times greater than for the Visa data from Los Angeles. Return to Text

²⁸ As, say, in Cochran (1977), Sampling Techniques, Wiley.Return to Text

²⁹ For example, Scheffe (1958), Analysis of Variance, Wiley. Return to Text

³⁰ United States Government Accounting Office (February 1998), "Personal Bankruptcy: The Credit Research Center Report on Debtors’ Ability to Pay."Return to Text

³¹ Impacts of other provisions of H.R. 3150 were beyond the scope of this study.Return to Text