Personal Bankruptcy Law: Abuse Prevention versus Debtor Protection 1

Transcription

1 Personal Bankrutcy Law: Ause Prevention versus Detor Protection 1 Michelle J. White University of California, San Diego, and NBER Octoer 2006 Michelle J. White 1 I am grateful to Steven Scroggin, Natalie Martin, Ralh Bruaker, Alain Trannoy, Roger Gordon and Eli Berman for helful comments. 1

2 The need for ankrutcy reform is long over-due and crucial to our Nation s economy and the well-eing of our citizens. Every day that goes y without these reforms, more ause and fraud goes undetected. America s economy should not suffer any longer from the illions of dollars in losses associated with rofligate and ausive ankrutcy filings. House Judiciary Committee Chairman F. James Sensenrenner, Jr. Introduction Personal ankrutcy law has two major--and conflicting economic ojectives. One is to facilitate the oeration of credit markets y roviding a state-sonsored rocedure for unishing default, determining defaulters aility-to-ay, and resolving dets. The other is to rovide detors with consumtion insurance y forgiving some or all of their det when their aility-to-ay is low. In the ast, the unishment ojective was dominant and ankruts were treated as criminals; in various countries they were ut to death, enslaved, dismemered, randed with a T for thief, or confined in detors risons (Efrat, 2006). Modern ankrutcy laws no longer treat ankruts as criminals, ut they still imose reayment requirements and other forms of unishment. Reayment requirements range from France s requirement that ankruts reduce themselves to a overty-level standard of living for 10 years in order to reay, to the U.S. ractice that until 2005 did not require ankruts to reay from earnings at all. Punishments also vary widely. In Britain, ankruts are not allowed to orrow, manage a comany, ractice as a lawyer/solicitor, or hold ulic office for three years after filing (Efrat, 2006). In the U.S., unishment is milder, ut ankruts names are made ulic and the filing can e listed on their credit records for ten years, which can make it more difficult for them to otain credit, rent an aartment, or get certain tyes of jos. The other ojective of ankrutcy is to rovide risk-averse detors with artial consumtion insurance. Detors enefit from orrowing, since it allows them to smooth consumtion and, sometimes, to oen usinesses. However orrowing makes detors vulnerale to adverse shocks, since they may e required to reay at a time when their aility-to-ay is low. Bankrutcy reduces the cost of adverse shocks y roviding a rocedure for discharging some det, which raises detors consumtion when it is low. This tye of insurance is valuale ecause shar reductions in consumtion can e very 2

3 costly to detors and their families--illnesses can turn into disailities if detors cannot ay for medical care, families may ecome homeless if detors cannot ay rent, detors may quit their jos and move in order to avoid their creditors, and detors children may dro out of school in order to work, leading to lower earnings as adults. But when ankrutcy rocedures rovide a high level of det forgiveness and require little reayment, some detors file even when they have not exerienced adverse shocks ehavior that I refer to as oortunistic. The Bankrutcy Ause Prevention and Consumer Protection Act of 2005 (BAPCPA), which went into effect in Octoer 2005, changed U.S. ersonal ankrutcy law from very ro-detor to much more ro-creditor. BAPCPA was assed in resonse to a large increase in ersonal ankrutcy filings in the U.S., from 287,000 er year in 1980, to 718,000 in 1990, to around 1,500,000 er year in each of the years (See tale 1.) An imortant question is whether oortunistic ehavior exlains the raid increase in the numer of filings: did consumers learn that ankrutcy law was very ro-detor and resond y orrowing more and filing for ankrutcy even when they could afford to reay? Creditors grous used the oortunism argument to convince Congressional leaders that reform was needed ecause ankrutcy was eing aused and, since BAPCPA went into effect, ersonal ankrutcy filings have in fact droed sharly. But I argue here that BAPCPA mainly discourages ankrutcy filings y non-oortunistic detors who file ecause their aility-to-ay has fallen. As a result, the adotion of BAPCPA is likely to increase the social cost of det. The aer starts y examining a model of the otimal ersonal ankrutcy rocedure and exloring the asic tradeoffs that determine how much det should e forgiven in ankrutcy and how much ankruts should e required to reay. I then focus on U.S. ankrutcy law oth efore and after the adotion of BAPCPA, asking how it comares to the otimal ersonal ankrutcy rocedure and whether it is effective in discouraging oortunism. I also discuss the market for credit card loans and how credit card loan ricing affects the otimal ankrutcy rocedure. The final section considers 3

4 ankrutcy law in several other countries and what they suggest concerning a more economically efficient aroach to U.S. ankrutcy reform. The ersonal ankrutcy decision and otimal ersonal ankrutcy law How do detors make their ankrutcy decisions and how do the decisions of oortunistic versus non-oortunistic detors differ? Assume that oth tyes of detors file for ankrutcy if doing so increases their utility, ut each tye is assumed to have a non-financial cost of filing ankrutcy stigma--that reresents their cost of earing the unishment that society imoses on ankruts. Oortunists have low stigma costs, so that they file for ankrutcy more often and they engage in strategic ehavior that increases their gain from filing. Non-oortunists have high stigma costs, so that they file for ankrutcy less often and are less likely to anticiate ankrutcy or adjust their ehavior to increase their gain from filing. In addition, some non-oortunists consider filing for ankrutcy only when they have exerienced an adverse shock to their ailityto-ay, so that they do not always file when doing so is financially advantageous. (See Fay, Hurst and White, 2002, for evidence that detors file more often when their financial gain is higher and when their earnings have fallen, which suorts oth tyes of ehavior. See White, 1998, for evidence that not all detors file when their financial gain is ositive. Sullivan et al, 1989, discuss a non-economic view of the ankrutcy decision.) Suose all consumers orrow an amount B at interest rate r in eriod 1 and agree to reay B ( 1 + r) in eriod 2. Consumers aility-to-ay each eriod equals the comined value of their assets lus their earnings in that eriod. In eriod 1, their earnings and assets are assumed to e fixed, ut their earnings or assets in eriod 2 are uncertain. Period 2 is assumed to cover the entire loan reayment eriod. At the eginning of eriod 2, the uncertainty is resolved and consumers--who are now detors-- decide whether to file for ankrutcy. To kee the discussion simle, assume that detors work effort remains the same regardless of whether they file for ankrutcy and that each detor is either an oortunist or a non-oortunist. (In a more general version of the model, detors vary their work effort deending on whether they file and may also 4

5 shift etween non-oortunistic and oortunistic ehavior. See White, 2006.) Detors who file for ankrutcy ay an out-of-ocket cost C, which includes lawyers fees and filing fees, lus a stigma cost S o or S n, which includes the cost of reduced access to credit and all other non-financial costs of ankrutcy. The o and n suscrits denote oortunists and non-oortunists, resectively, where S o < S n. Non-oortunistic detors are assumed to e risk averse, while oortunistic detors may e either risk averse or risk neutral. Now consider the ankrutcy system. If detors file for ankrutcy, the automatic stay stos creditors collection efforts and a fraction 0 < d < 1of det is discharged during the ankrutcy rocess. The discharge roortion reresents oth the fact that certain tyes of dets are not discharged in ankrutcy and that the ankrutcy judge may have the ower to discharge additional det y shortening the reayment eriod (see the discussion elow). Assets u to a value of X are exemt, ut detors must use all of their assets aove X to reay. Earnings u to a value of Y are also exemt, ut only a fraction m of earnings aove Y is exemt, where 0 m 1. m = 1 reresents the fresh start, where detors have no oligation to reay from ost-ankrutcy earnings. Now consider oortunistic detors ankrutcy decision. If an oortunistic detor avoids ankrutcy and reays the det in full, her eriod 2 consumtion will e A + E B( 1+ r), where A denotes the detor s eriod 2 assets and E denotes her eriod 2 earnings. (Individual-level suscrits are omitted.) If the detor files for ankrutcy, her eriod 2 consumtion will e min[ A, X ] + min[ E, Y + m( E Y )] ( 1 d ) B(1 + r) C S o. Because detors work effort is assumed to remain constant, they file for ankrutcy whenever doing so increases their consumtion. Figure 1 shows the oortunistic detors eriod 2 assets on the horizontal axis and their eriod 2 earnings on the vertical axis. The solid line encloses the asset/earnings region in which oortunistic detors gain from filing. The gain region is divided into four su-regions. In su-region I, detors assets and earnings are oth comletely exemt, so that they gain from filing if the det discharged in ankrutcy db ( 1+ r) 5

6 exceeds their comined ankrutcy costs S o + C. In su-region II, detors have nonexemt assets, so that they gain from filing if the value of det discharged exceeds their comined ankrutcy costs lus the value of non-exemt assets A X that they must give u. There is a threshold level of assets, denoted  o, such that oortunistic detors in region II file if their assets are elow  o and not otherwise. In su-region III, detors have non-exemt earnings, so that there is a threshold level of earnings, denoted Ê o, such that they file if their earnings are elow Ê o ut not otherwise. Finally in suregion IV, detors must reay from oth assets and earnings, so that the two thresholds trade off. Overall, oortunistic detors gain from filing for ankrutcy is inversely related to their aility-to-ay, which is oth efficient and equitale. Oortunistic detors also use lanning strategies to increase their gain from filing for ankrutcy. These include changing the form of their assets and/or earnings so that more are exemt in ankrutcy or changing the form of their dets so that more det is discharged in ankrutcy. These strategies increase the values of X, Y and/or d, thus exanding the ankrutcy gain region for oortunistic detors. (Examles of ankrutcy lanning strategies are discussed elow in connection with U.S. ankrutcy law.) If the detor is a non-oortunist rather than an oortunist, then the model changes in two ways. First, the stigma cost term increases from S o to S n, which reduces the size of the ankrutcy gain region in figure 1. The dashed line in figure 1 shows a smaller ankrutcy gain region for non-oortunists. Second, non-oortunistic detors may not file for ankrutcy even when their assets and earnings lace them in the financial gain region some only file if they have also exerienced an adverse shock. These differences imly that non-oortunistic detors are less likely to file than oortunistic detors and also that non-oortunists ankrutcy decisions are less resonsive to changes in the ankrutcy olicy variales. Suose all consumers have the same assets and earnings in eriod 1. Some are oortunists and some are non-oortunists, ut lenders are assumed unale to 6

7 distinguish etween tyes. Also assume (temorarily) that the loan industry is cometitive and lenders make zero rofits. This means that lenders are willing to lend to all consumers as long as an interest rate exists that allows them to make zero exected rofit as of eriod 1. If there are more oortunists in the detor oulation or higher values of the ankrutcy olicy variales d, X, Y, or m, then the interest rate is higher and there may e credit-rationing. How does the ankrutcy system rovide detors with consumtion insurance? Suose the discharge roortion increases from d to d '. This raises all detors consumtion in ankrutcy y ( d ' d) B(1 + r), so that the gain regions in figure 1 ecome larger and more detors of oth tyes file. Creditors resond y raising the interest rate. These changes raise detors consumtion in ankrutcy and lower their consumtion outside of ankrutcy, thus roviding them with additional consumtion insurance. But the increase in d enefits oortunists more than non-oortunists, oth ecause oortunists file more often and ecause their filing rate increases y more than that of non-oortunists. Increases in the exemtion variales X, Y, and m also rovide detors with additional consumtion insurance, ut detors only enefit from an increase in X if they have assets exceeding X (so that they are in su-regions II or IV of figure 1) and they only enefit from an increase in Y or m if they have earnings aove Y (so that they are in su-regions III or IV of figure 1). Thus an increase in d has the advantage that it rovides additional consumtion insurance to the most needy detors, while increases in the other olicy variales mainly rovide additional consumtion insurance to oortunistic detors. (Reductions in detors out-of-ocket cost of filing, C, have the same effect as increases in the value of d.) Now suose the model is generalized to allow detors oth to vary their eriod 2 work effort and to shift from oortunistic to non-oortunistic ehavior, or vice versa. Then an increase in the value of m also causes detors to work more in ankrutcy ecause they kee more of their marginal earnings (assuming that the sustitution effect exceeds the income effect), causes detors file for ankrutcy more often ecause their reayment oligation is smaller, and causes additional detors to ehave oortunistically. Increases in the other ankrutcy olicy variales have less effect on work effort and on the amount of oortunistic ehavior. 7

8 Now consider the otimal values of the ankrutcy olicy variales (see White, 2006, for discussion). Suose the median voter is a non-oortunist, so that the social welfare function considers only non-oortunistic detors references. Increasing the values of any of the ankrutcy olicy variales d, X, Y, and m enefits detors y roviding them with additional consumtion insurance, although as just discussed increases in d rovide the most valuale insurance since detors enefit even when their consumtion is at its lowest level. But detors ay more than the fair rice for ankrutcy-rovided consumtion insurance, ecause lenders charge the fair rice via the interest rate and detors must also ay stigma and out-of-ocket ankrutcy costs when they file. As a result, the marginal value of additional consumtion insurance declines as the insurance level rises and the otimal values of the olicy variales are likely to e lower than their maximum values. The otimal values of all four olicy variales are higher if non-oortunistic detors are more risk-averse, if fewer detors are oortunists, and if the level of oortunism is not highly resonsive to increases in the values of the olicy variales. The otimal values of the earnings exemtions m and Y are also higher if detors ost-ankrutcy work effort is highly elastic to changes in the exemtion levels and lower if the oligation to reay from ost-ankrutcy earnings strongly discourages oortunism. Wang and White (2000) simulated a ankrutcy system similar to the one discussed here. In their model, Y was assumed to e zero and d was assumed to e one, so that only m and X could vary. Detors work effort was allowed to vary deending on whether they file for ankrutcy and the numer of oortunists was also allowed to vary deending on the gain from oortunistic ehavior. Wang and White were articularly interested in determining under what conditions the fresh start for detors (m = 1) is the otimal ankrutcy olicy. Their results showed that, if there were few oortunists, then m = 1 and the fresh start is economically efficient, ut when the level of oortunism increased, the otimal value of m fell to This is ecause imosing a ankrutcy tax on detors ost-ankrutcy earnings is effective in discouraging oortunism, since oortunists ay the tax more often than non-oortunists. However Wang and White never found that a value of m elow.93 was otimal, which suggests that setting m equal to 0 as many countries do--is extremely inefficient. Also the 8

9 otimal value of the asset exemtion X was always elow the maximum level at which credit markets function. Finally, Wang and White found that the various ankrutcy olicy arameters are sustitutes, so that the otimal olicy involves setting one ankrutcy olicy variale at a high level in order to rovide consumtion insurance, while setting the other olicy variale at a lower level to discourage oortunism. U.S. ankrutcy law re-bapcpa Now turn to U.S. ersonal ankrutcy law. I discuss ankrutcy law re-bapcpa in this section and the changes under BAPCPA in the next section. Pre-BAPCPA ankrutcy law. U.S. ankrutcy law rior to BAPCPA differed from the otimal ankrutcy rocedure in that there were (and still are) two searate ankrutcy rocedures, called Chaters 7 and 13, and detors had the right to choose etween them. Under Chater 7, detors were oliged to reay only from non-exemt assets, so that m = 1 and Y = ; while under Chater 13, detors were oliged to reay only from non-exemt earnings, so that X =. Most unsecured det was discharged under oth rocedures, ut student loans, dets incurred shortly efore filing and dets incurred y fraud were not. Detors out-of-ocket ankrutcy costs C were low less than $1,000 for lawyers fees and filing fees. These features made U.S. ankrutcy law very ro-detor, since detors could choose to reay from whichever source they didn t have! They also meant that detors oligation to reay in ankrutcy had little relationshi to their aility-to-ay, since detors aility-to-ay comes rimarily from their earnings, ut most ankrutcy filers filed under Chater 7. Asset exemtions in Chater 7 were (and still are) set y the state in which the detor lives. The most imortant exemtion in most states is the homestead exemtion for equity in owner-occuied homes. It varies widely, from zero in two states to unlimited in Texas, Florida, the District of Columia, and several other states (Elias, 2005). An additional 20 states allow detors an unlimited homestead exemtion if they are married, own their homes in tenancy y the entirety, and only one souse files for ankrutcy. The variation in asset exemtions imlies that detors gain from filing for ankrutcy varies widely deending on where they live, their marital status, whether they own homes, and whether they lan strategically for ankrutcy. 9

10 Detors who filed under re-bapcpa Chater 13 were oliged to roose lans to reay art or all of their det from ost-ankrutcy earnings over three to five years. While some detors roosed to reay sustantial amounts, many roosed to reay an amount equal to the value of their non-exemt assets in Chater 7 if they had non-exemt assets or a token amount otherwise. Detors were not allowed to reay less and they had no incentive to offer more. Only the aroval of the ankrutcy judge--not creditors-- was required. Bankrutcy judges often acceted Chater 13 lans that roosed to reay very little, since most detors otherwise could shift to Chater 7 and reay nothing. This meant that most detors financial gain from filing for ankrutcy under Chater 13 was the same as their gain from filing under Chater 7. Chater 13 also included some secial features that allowed additional det to e discharged known as the suer-discharge. Detors could delay foreclosure of their homes y filing under Chater 13 and their car loans could e discharged to the extent that the loan rincile exceeded the car s market value. Also, dets incurred y fraud and cash advances otained shortly efore filing could e discharged only in Chater 13. These features meant that, for some detors, the financial gain from filing under Chater 13 was higher. The ankrutcy decision re-bapcpa. Suose d, X and C denote the discharge roortion, the asset exemtion, and ankrutcy costs under re-bapcpa Chater 7, where the suerscrits indicate re-bapcpa values. Consider oortunistic detors ankrutcy decision first. Their eriod 2 consumtion level if they file under Chater 7 is X A,0] + E (1 d ) B(1 + r) So C min[ and their consumtion level if they avoid ankrutcy is A + E B( 1+ r). If they have non-exemt assets, then the threshold level of assets at which they are indifferent etween filing or not filing for ankrutcy, denoted  o, must satisfy: d B 1 + r) = Aˆ o X + C + So (. (1) 10

11 Here, the value of det discharged in ankrutcy just equals the value of non-exemt assets that detors must give u lus their comined costs of filing. Oortunistic detors file for ankrutcy if their assets are elow  o. Oortunistic detors also used several well-known re-bapcpa lanning strategies to increase their financial gain from ankrutcy. These included otaining new credit cards and charging as much as ossile on old and new cards (suject to limits on dischargeaility of det incurred shortly efore filing), converting non-exemt assets into exemt home equity y aying down their mortgages or renovating their homes (if the additional home equity would e exemt under the state s homestead exemtion), sheltering non-exemt assets in asset rotection trusts, and/or moving to states that have unlimited homestead exemtions or allow tenancy y the entirety. Oortunistic detors could also file first under Chater 13 and then under Chater 7 a strategy known as filing a Chater 20. The Chater 13 filing gave detors the enefit of the Chater 13 suer-discharge; while shifting to Chater 7 allowed them to avoid using ost-ankrutcy earnings to reay unsecured det. These lanning strategies raised the value of  o and made it more attractive for oortunistic detors to file. Figure 2 shows oortunistic detors ankrutcy gain region as the area to the left of the solid line at  o. Because the earnings exemtion in Chater 7 was unlimited, the gain region has no height limit--detors gained from filing for ankrutcy regardless of how much they earned. In addition, following ankrutcy lanning strategies ushed  o to the right and therefore allowed oortunistic detors to gain from filing even if they had very high asset levels. Oortunistic detors thus could gain from ankrutcy re-bapcpa with oth high earnings and high assets. Now consider non-oortunistic detors. Their ankrutcy gain region is determined in the same way, excet that their stigma costs are higher and they do not make use of ankrutcy lanning strategies. This means that they gained from filing for ankrutcy if their eriod 2 assets turned out to e elow  n, where ˆ <. Figure 2 shows non-oortunistic detors smaller ankrutcy gain region as the area to the left of the dashed line. In addition, some non-oortunistic detors avoided filing for An Aˆ o 11

12 ankrutcy even when their assets and earnings ut them in the gain region, while others filed under Chater 13 and romised to reay most or all of their dets (ut detors rarely comleted these lans). Overall, the consumtion insurance role of ankrutcy was severely undermined y the fact that oortunistic detors could gain from filing even when their aility-to-ay was extremely high. Pre-BAPCPA ankrutcy law encouraged detors to ehave oortunistically and harmed non-oortunistic detors y reducing the availaility of credit, articularly in high-exemtions states such as Texas and Florida (Gro, Scholz and White, 1997, and Berkowitz and White, 2004). However re-bapcpa ankrutcy law had the advantages that it did not distort detors work effort decisions (Han and Li, 2004) and, y reducing downside risk, it encouraged workers to ecome self-emloyed (Fan and White, 2003). Emirical evidence concerning oortunistic ehavior. The major credit card lenders loied heavily for ankrutcy reform on the grounds that most re-bapcpa ankrutcy filers were oortunists who orrowed without intending to reay and harmed other orrowers y filing when they had not exerienced an adverse shock. What evidence is availale to suort these claims? Warren (2003) asked a samle of ankrutcy filers what caused them to file and found that 68% of filers had exerienced an adverse shock in the form of jo loss, illness/injury, forced relocation, or a decline in income. Her results suggest that the remaining detors who did not exerience adverse shocks are oortunists, or aout one-third of all ankrutcy filers. Another form of evidence concerning oortunism is studies that have estimated the roortion of ankrutcy filers that could afford to reay a sustantial roortion of their unsecured det. Barron and Staten (1997) found a figure of aout one-third. But other authors disuted the Barron and Staten results and argued that only a few ercent of ankrutcy filers could afford to reay (see Culhane and Micaela White, 1999, and Flynn and Bermant, 2000). Overall these studies suggest that the roortion of detors who are oortunists could e as low as 3% or as high as one-third. One reason for the amiguity is that many detors roaly qualify as oth tyes. They decide to file for ankrutcy ecause their 12

13 earnings fall, ut efore filing they re-arrange their assets so as to increase their financial gain. U.S. ankrutcy law ost-bapcpa Changes in ankrutcy law under BAPCPA. BAPCPA retained the dualrocedure structure of U.S. ankrutcy law, with detors still oliged to reay only from assets in Chater 7 and only from earnings in Chater 13. But it aolished detors right to choose etween the two rocedures and sustituted a new means test to determine whether detors are allowed to file under Chater 7. Detors whose earnings are too high for Chater 7 must file under Chater 13, where a comlicated new rocedure determines how much they must reay from ost-ankrutcy earnings. The BAPCPA means test is ased on detors average monthly earnings during the six-month eriod rior to filing. Since we reviously defined eriod 2 to cover the entire reayment eriod and the BAPCPA reayment eriod is 5 years, multily detors average monthly earnings rior to ankrutcy y 60 to get re-ankrutcy earnings over 5 years, denoted PE. Detors are allowed to yass the means test and file under Chater 7 if PE < 60MFI, where MFI is the median monthly family income level in the detor s state of residence (adjusted for the detor s family size). Define fixed dollar earnings exemtion over the 5 year reayment eriod, where the suerscrit indicates values under BAPCPA. (How elow.) Detors ass the means test if their non-exemt earnings Y to e the Y is determined is discussed PE Y over the 5 year eriod are less than $6,000 ($100 er month). Comining these conditions, detors qualify to file under Chater 7 if: PE max[ 60MFI, Y + 6,000]. (2) Suose MAX7 denotes the maximum level of earnings at which detors qualify for Chater 7 the right hand side of condition (2). Suose a detor asses the means test and files under Chater 7. Chater 7 under BAPCPA is similar to the re-bapcpa rocedure, since the same state-secific asset exemtions remain in effect and detors are only oliged to use their non-exemt assets 13

14 to reay. But BAPCPA introduced new restrictions on when detors are allowed to use their states asset exemtion. If detors move to a new state less than two years efore filing, then they must use the homestead exemtions in their old states and, if they urchase a house less than 2½ years efore filing, then their homestead exemtion is caed at $125,000. If detors convert non-exemt assets into home equity y aying down their mortgages or renovating their homes, they must do so at least 3 1/3 years or 10 years, resectively, efore filing otherwise the additional home equity will not e exemt (Martin, 2006). But BAPCPA still allows detors to use asset rotection trusts and tenancy y the entirety. BAPCPA also rovided a valuale new asset exemtion for u to one million dollars in tax-sheltered individual retirement accounts (two million dollars for married coules who file for ankrutcy). Suose the Chater 7 asset exemtion under BAPCPA is denoted higher or lower than X. X. For individual detors, X may e either Now suose detors fail the means test and file under Chater 13. BAPCPA aolished detors right to roose their own reayment lans--instead they are required to reay the fixed amount reayment eriod is therefore PE Y. Detors consumtion during the Chater 13 E PE + Y, or the difference etween ost-ankrutcy and re-ankrutcy earnings lus the fixed earnings exemtion. Because detors kee all of their ost-ankrutcy earnings at the margin (m = 1), their ost-ankrutcy work incentives in theory are undistorted. But detors re-ankrutcy work incentives are strongly distorted (see elow), and if their earnings are lower after the ankrutcy filing ( E < PE ), then their consumtion during the reayment eriod might e so low that they refer to sto working comletely. Now turn to the earnings exemtion under BAPCPA, Y. Y is determined y adding three sets of consumtion allowances. The first set covers housing, transort, utilities, food, aarel and ersonal care and secifies a fixed dollar maximum allowance for each. The housing/utilities allowance deends only on the cost of housing in the detor s metroolitan area, the transort allowance deends on the numer of cars the detor owns and the average cost of owning/oerating non-luxury cars, and the food, aarel and ersonal care allowances vary slightly with earnings. These allowances are 14

15 ased on IRS rocedures for collecting from delinquent taxayers. They are intended to force well-off detors either to sharly reduce their consumtion or lose their homes, erhas ecause the IRS elieves that delinquent taxayers often have hidden assets. The second set of allowances covers detors exenditures on income taxes, court-ordered child suort ayments, childcare costs, term life insurance, uninsured health care costs, and several other categories and is also ased on IRS rocedures. While these allowances cover detors actual exenditures rather than maximum sending limits, most of the covered exenses are outside of detors control. Finally, a third set of allowances is not ased on IRS rocedures and covers detors actual exenditures on health and disaility insurance, contriutions to elderly or disaled family memers, additional home energy costs, home security costs, charitale contriutions, telecommunications, contriutions to tax-sheltered retirement lans, and the cost of reaying secured det. (See Assume temorarily that Y is a fixed dollar amount (ut oortunistic detors incentive to send more on these categories is discussed elow). BAPCPA also greatly raised detors costs of filing for ankrutcy. Detors are now required to take a credit counseling course efore filing and a financial management course efore receiving a discharge. They must file detailed financial information with the ankrutcy court, including coies of their tax returns and wage stus--even if they did not file tax returns. Bankrutcy lawyers must certify the accuracy of all the information filed and lawyers can e fined if any of information is found to e false or inaccurate. These new requirements were redicted to raise the comined cost of lawyers fees and filing fees to nearly $3,000 (Elias, 2005). These changes mean that C greatly exceeds C. BAPCPA also eliminated the Chater 13 suer-discharge and lengthened the eriods efore filing during which cash advances and some credit card loans are nondischargeale. These changes mean that the fraction of det discharged in ankrutcy is lower under BAPCPA, or d d <. Finally, BAPCPA also increased the minimum time eriods that must elase etween ankrutcy filings--from 6 to 8 years for Chater 7, from 6 months to 2 years for Chater 13, and from no minimum to 4 years for a 15

16 Chater 13 filing followed y a Chater 7. These changes imly that fewer detors are eligile to file for ankrutcy. The ankrutcy decision under BAPCPA. First consider oortunistic detors ankrutcy decisions under BAPCPA. They are allowed to file under Chater 7 if they ass the means test, which requires that condition (2) holds, or PE MAX7. They gain from filing under Chater 7 if their eriod 2 assets are elow a new asset threshold, denoted  o, which is determined y the condition d B( 1+ r) = Aˆ o X + C + So. The lower rectangle in figure 3a shows the region in which oortunistic detors oth qualify to file under Chater 7 and gain from filing. Note that re-ankrutcy earnings PE, rather than ost-ankrutcy earnings E, are now on the vertical axis in figure 3a. Now turn to oortunistic detors decision to file under Chater 13. Their consumtion in Chater 13 is E PE Y ) + A (1 d ) B(1 + r) C So (, while their consumtion outside of ankrutcy is E + A B( 1+ r). Therefore they gain from filing under Chater 13 if PE Y C S + d B( 1+ r). Suose the right hand side of this o exression is denoted MAX13 the maximum earnings level at which detors gain from filing under Chater 13. Assuming that MAX13 exceeds MAX7, there is a region MAX7 PE MAX13 in which oortunistic detors cannot file under Chater 7, ut gain from filing under Chater 13. Also, detors who file under Chater 13 are not allowed to reay creditors less than they would receive in Chater 7, which means that detors gain from filing under Chater 13 only if their assets are elow the Chater 7 threshold  o. Putting these conditions together, the uer rectangle in figure 3a is the ankrutcy gain region for oortunistic detors filing under ost-bapcpa Chater 13. As shown, the figure assumes that detors always refer Chater 7 over Chater 13. Now consider what strategies oortunistic detors can use under BAPCPA to increase their gain from filing. One strategy involves oortunistic detors working less during the 6 month eriod efore filing, so as to reduce their re-ankrutcy earnings PE. Suose oortunistic detors reduce their earnings y $100 er month during the 6 month eriod. While this lowers their re-ankrutcy consumtion y $600, it lowers 16

17 their oligation to reay in Chater 13 y $6,000 for a 10:1 return! The strong incentive for detors to reduce their re-ankrutcy earnings ersists to the oint that they ass the means test and qualify to file under Chater 7. Oortunistic detors can also avoid Chater 13 y altering their sending atterns, since additional sending on certain categories increases the earnings exemtion dollar-for-dollar. To get a sense of how generous Y is and how much it can e maniulated y oortunistic detors, I first calculated it for hyothetical detors in three states, assuming that all their sending falls within the first and second sets of consumtion allowances. (See White, 2007). I found that the resulting earnings exemtions ranged from aout 5% elow the median family income level in high-income states such as Connecticut to aout 20% aove the median family income level in lowincome states such as West Virginia. Then I recalculated Y Y, assuming that detors increase their sending on categories covered y the third set of consumtion allowances. The additional exenditures included detors uying homes with mortgages or otaining new mortgages on their existing homes, uying a second car or a new car with car loans, uying health insurance, setting u individual retirement accounts, and contriuting to charity. I found that y sending on these categories, oortunistic detors could raise Y to a level equal to at least twice the median family income level. As a result, they can qualify for Chater 7 even if their earnings are as high as the 90 th ercentile of the U.S. income distriution. Strategies that lower PE or raise Y increase the values of oth MAX7 and MAX13, so that they increase the height of oth the Chater 7 and Chater 13 gain regions in figure 3a. As a result, oortunistic detors are more likely to qualify for Chater 7 and gain more from filing under oth chaters. Another strategy that oortunistic detors can use to qualify for Chater 7 is to own or start a usiness efore filing, since BAPCPA exemts detors from the means test if their dets are not rimarily consumer dets. These detors are allowed to file under Chater 7 at any PE level. Now suose oortunistic detors qualify to file under Chater 7 and consider strategies that increase the value of the asset exemtion X. BAPCPA closed off some 17

18 of the most oular re-bapcpa asset lanning strategies detors can no longer ay down their mortgages or renovate their kitchens or move to high-exemtion states efore filing, unless they do so far in advance of their ankrutcy filings. The Chater 20 strategy has also een closed off. However detors can still use Texas and Florida s high homestead exemtions if they already live there and they can still use tenancy y the entirety and asset rotection trusts to shelter assets in Chater 7 (a ill recently introduced in Congress to limit use of asset rotection trusts in ankrutcy is titled the Billionaire s Loohole Elimination Act ). Also, the new exemtion for individual retirement accounts allows detors to shelter high levels of additional assets in Chater 7. While there are limits on the amounts that can e contriuted to these accounts each year, many detors qualify for multile tyes of accounts, so their aggregate contriution limits are high. The limits are articularly high for self-emloyed detors. Using these strategies increases the width of oortunistic detors gain regions in figure 3a, so that they gain from filing for ankrutcy at higher asset levels. Now turn to non-oortunists. They are assumed to make their ankrutcy decisions in the same way, ut their stigma costs are higher and they do not use ankrutcy lanning strategies to increase their financial gain. Their ankrutcy gain region, enclosed y the dashed lines in figure 3, is smaller than that of oortunistic detors. If we comare the re-bapcpa gain regions in figure 2 with the ost-bapcpa gain region in figures 3a and 3, the adotion of BAPCPA imroved oth efficiency and equity y imosing a maximum limit on the earnings at which detors gain from filing for ankrutcy. But the comarison of figures 2 and 3 suggests that BAPCPA harmed non-oortunistic detors y greatly increasing ankrutcy costs, which reduced oth the height and width of their gain region. Many non-oortunistic detors roaly will avoid filing for ankrutcy under BAPCPA simly ecause they cannot afford the high costs of filing. These detors will take the alternative route of defaulting, having their wages garnished, and erhas quitting their jos and moving to avoid their creditors. In contrast, the comarison of oortunistic detors re- versus ost-bapcpa gain regions in figures 2 and 3a suggests that BAPCPA is less likely to discourage filings y oortunists. Oortunists can still gain from filing for ankrutcy even with high asset and earnings levels, although under BAPCPA they need to start lanning for ankrutcy 18

19 earlier. Overall, BAPCPA mainly discourages ankrutcy filings y the worst-off detors. Early evidence on BAPCPA. As shown in tale 1, ankrutcy filings fell from an annual rate of 1.5 million in to an annual rate only 500,000 during the first half of (The filing rate of 2 million in 2005 mainly reflects a rush y detors to file efore BAPCPA went into effect.) During BAPCPA s first year of oeration, aout 5% of ankrutcy filers had earnings aove the median level in their states, comared to 15% efore BAPCPA went into effect (Clifford White, 2006). If we assume that the ankrutcy filers with the highest earnings are the oortunists, then these figures suggest that, during its first year of oeration, BAPCPA discouraged 200,000 oortunists from filing, ut it also discouraged 800,000 non-oortunists. While discouraging oortunism increases economic efficiency, the cost of doing so was high. BAPCPA and the market for credit card loans. Credit card loans are imortant for ankrutcy olicy, oth ecause they have grown raidly and ecause they constitute half of all unsecured det in ankrutcy (Flynn and Bermant, 2003/2004). Between 1990 and 2005, er caita credit card det in the U.S. nearly triled from $960 to $2,700 and, in 2005, the average U.S. household received 45 solicitations for new credit cards. The suly of credit card loans has increased ecause of the rise of credit ureaus that gather financial information on households, comuterized credit scoring models that evaluate orrowers risk, and securitization of credit card det all innovations that lower lenders costs (Moss and Johnson, 1999). Contrary to the model discussed aove, the market for credit card loans in the U.S. is not erfectly cometitive--aout 10 large lenders dominate the credit card industry (Mann, 2007). Another feature of credit card loans that makes them articularly imortant for ankrutcy olicy is that, unlike other tyes of loans, lenders have the right to change the terms of credit card loans at any time. Tyically, lenders comete y offering very favorale terms to new customers including zero annual fees, low or zero introductory interest rates, and rewards such as cash ack or frequent flier miles for charging more. 19

20 Lenders lose money on new customers, ut offset these losses y charging high enalty fees when detors ay late or exceed their credit limits and y drastically raising interest rates when detors charge too much or make only the minimum monthly ayment-- enalty fees of $40 and enalty interest rates of 30% are common. Lenders also set very low minimum monthly ayments--1% of the rincile lus interest and fees for the revious month is common--so that detors who make only the minimum ayment each month reay very slowly and ay large amounts of interest. 2 Lenders make a high fraction of their rofits from these detors. During the 1990 s, increased cometition among credit card lenders caused introductory terms to ecome more attractive, while enalty fees and interest rates rose sharly (Evans and Schmalensee, 1999). These trends have continued during the ast 5 years. A third feature of credit card loans is that lenders have an inefficiently high incentive to offer cards to detors who already are heavily indeted. This is ecause all credit card loans have equal riority, so that the latest lender gets its return mainly at the exense of earlier lenders. Going ack to the model, all consumers were assumed to e identical as of eriod 1 and all received loans on the same terms. But in eriod 2, they have different draws on aility-to-ay. Suose detors with high aility-to-ay ay their credit card ills on time and in full each month, so that they do not ay any interest or fees. But detors with low aility-to-ay make only minimum ayments each month and sometimes ay late, which means that lenders charge enalties and raise their interest rates. Thus the attern of ricing of credit card loans increases the variance of detors eriod 2 consumtion, since loan costs are low for detors who have good draws on eriod 2 aility-to-ay and high for detors who have ad draws. In addition, ehavioral economists argue that detors underestimate how much they will use their credit cards and discount the ossiility of adverse events. The favorale introductory terms thus cause detors to accet additional cards, which means they send more and are more likely to ay enalty 2 Suose a detor orrows $2000 on a credit card at an interest rate of 28%, with a late fee of $39 and a minimum monthly ayment equal to 1% of the rincile lus monthly interest and fees. The detor is assumed to make the minimum ayment each month and ays late twice a year. At the end of 5 years, the detor would still owe 55% of the original loan rincile. Under another commonly-used ractice of setting the minimum monthly ayment at 2.5% of the total amount owed, the rincile increases over time. 20

21 fees and high interest rates (Ausuel, 1991, and Bar-Gill, 2004). But orrowing more on credit cards increases the variance of detors eriod 2 consumtion yet further, since the negative effect of a ad draw on eriod 2 urchasing ower is magnified. Overall, credit card ricing ractices make risk-averse detors worse off y increasing the variance of consumtion. Prior to BAPCPA, detors who exerienced a ad draw on eriod 2 urchasing ower could easily file for ankrutcy, which increased their consumtion levels when adverse events occurred. But under BAPCPA, fewer detors are eligile to file for ankrutcy, while those who file must ay nearly $3,000 in ankrutcy costs and less of their det is discharged. Thus the adotion of BAPCPA reduced the availaility of ankrutcyrovided consumtion insurance just at a time when the changing attern of credit card ricing has made it more valuale to detors. The result is that detors minimum consumtion levels will e lower under BAPCPA and the social costs of det will rise. Bankrutcy Laws in Other Countries Now turn to ersonal ankrutcy laws in France, Germany and Canada. Like the U.S., all three countries have exerienced large increases in consumer det levels and ankrutcy filing rates over the ast 10 to 20 years. See tale 1. Neither France nor Germany had a ersonal ankrutcy law efore the 1990 s, ut all three countries have adoted or changed their ersonal ankrutcy laws. In this section I consider how their ankrutcy laws comare to U.S. ankrutcy law and what they suggest concerning how the efficiency of U.S. ankrutcy law could e imroved. Tale 2 lists characteristics of ankrutcy laws for all four countries. In addition to the four ankrutcy olicy variales already discussed, tale 2 also comares detors out-of-ocket ankrutcy costs, the length of the required earnings reayment eriod, and whether det is discharged automatically at the end of the reayment eriod or the ankrutcy judge has the ower to deny or delay the discharge. All three countries ankrutcy rocedures are roadly similar to the otimal ankrutcy rocedure discussed aove. All have a single ankrutcy rocedure that requires detors to reay from oth assets and ost-ankrutcy earnings. All have simle formulas for determining their fixed asset and earnings exemtions earnings 21

22 exemtions deend mainly on detors family size and location within the country. Detors are oliged to use all of their non-exemt assets and a fixed fraction of their nonexemt earnings to reay for a fixed numer of years. Most countries discharge some det at the time of filing or during the reayment eriod for examle, France discharges loans made to detors when they were already over-indeted. In addition, whatever det remains after detors have met their reayment oligation for the rescried numer of years is discharged. Detors out-of-ocket cost of filing for ankrutcy is lower in all three countries than in the U.S. But although the three countries ankrutcy laws are close in form to the otimal ankrutcy rocedure, their laws could nonetheless e inefficient if the values of the ankrutcy olicy variales are far from the otimal levels. Consider French ankrutcy law first (see Kilorn, 2005). French asset and earnings exemtions are very low. Detors are allowed to kee only asic household goods and they must give u owner-occuied homes and all non-exemt assets. The fixed earnings exemtion covers only a overty-level standard of living of $6,000 for single detors and $13,000 for a family of four with a single earner. Non-exemt earnings are suject to a reayment requirement that quickly increases from 5% to 100% of earnings that exceed $25,000 for a single erson and $28,000 for a family of four with a single earner. The reayment eriod is extremely long--8 to 10 years and, at the end of the eriod, the ankrutcy judge may deny the discharge if the detor has shirked. Because reayment requirements are so high, most lans fail. France also has an informal Chater 7-like ankrutcy rocedure in which all of a detor s dets are discharged immediately, on the grounds that the detor cannot realistically reay anything even over 10 years. (Detors are still required to use their non-exemt assets to reay, if they have any.) This rocedure is used in aout 14% of French ankrutcy filings and ankrutcy trustees recommend it in many more. France also has a high rate of reeat ankrutcy filings around 30%. Harsh reayment lans give ankruts an incentive to sto making ayments and file for ankrutcy again, ecause a new filing 22

23 gives them a fresh chance of receiving an immediate discharge. Since detors out-ofocket cost of filing is zero in France, they can afford to file reeatedly. 3 Thus French ankrutcy law rovides ankruts with a consumtion insurance lottery. Most detors receive a very low level of insurance they must reduce their consumtion to a very low level for a very long eriod of time while continuing to work in order to receive a discharge. But a few detors receive a high level of insurance, since all of their dets are discharged immediately and they have no reayment oligation. Because the French social safety net guarantees a higher minimum standard of living than in the U.S., the need for ankrutcy-rovided consumtion insurance is lower than in the U.S. But the fact that the ankrutcy system is a lottery reduces its value to risk-averse detors. The French ankrutcy system has the advantage that it strongly discourages detor oortunism, since few detors would choose to e suject to its harsh reayment requirements. But those detors who do end u in ankrutcy have strong incentives to quit their jos, move, or at least not seek etter jos during the long reayment eriod. French detors also have a strong incentive to avoid self-emloyment, since the consequences of usiness failure are dire. German ankrutcy law is similar to the French ankrutcy system, ut is somewhat more detor-friendly and it treats detors more uniformly (Kilorn, 2004). The earnings exemtion is higher than a overty-level standard of living (see tale 2) and the reayment eriod is shorter--6 years rather than 8 to 10. Detors are still oliged to use all of their non-exemt assets and earnings to reay and, if they receive inheritances during the reayment eriod, half must e used to reay. But detors who reay as scheduled receive eriodic onuses that discharge u to 25% of their det. Because exemtion levels are higher than in France, most ankruts have no non-exemt assets or earnings and therefore are not oliged to reay anything. Nonetheless, they must make their est efforts to find and hold a jo and must wait for the end of the six year eriod to receive a discharge which the judge may withhold for shirking. There is no second ankrutcy rocedure in Germany, so that detors do not face uncertainty concerning 3 Both the detor and creditors in France must agree to the reayment lans, ut French ankrutcy trustees have the exclusive right to roose reayment lans. Aout 70% of lans are acceted. See Kilorn (2005). 23