Who Files for Bankruptcy? State Laws and the Characteristics of Bankrupt Households

Transcription

1 Who Files for Bankruptcy? State Laws and the Characteristics of Bankrupt Households By MICHELLE M. MILLER The characteristics of bankrupt households such as incoe and asset levels vary widely across states. This paper asks whether these variations can be attributed to state exeption laws or state garnishent laws. Using a new household-level dataset, I find that high exeption levels encourage high asset households to file for bankruptcy while high garnishent rates encourage low incoe households to file for bankruptcy. These results are supported by a theoretical odel in which households choose between repayent, bankruptcy, and non-response which occurs when households siply walk away fro their bills, allowing creditors to garnish their wages and seize their assets. In, over.5 illion households filed for bankruptcy, discharging ore than $459 billion in debt and rivaling Medicare as one of the country s largest transfer of wealth progras Report of Statistics Required by the Bankruptcy Abuse Prevention and Consuer Protection Act of 5, Table X. Notably, the percentage of households filing for bankruptcy varies treendously across states. In Texas, for exaple, 5.6 of every, households filed for bankruptcy in copared to 7.7 per, households in Tennessee. Nuerous papers have studied the cross-state differences in filing rates, but to date no study has adessed an equally iportant question why do the characteristics of bankrupt households vary across states? For exaple, no study has asked why the edian incoe of bankrupt households in Texas was $37,78, whereas the coparable figure in Tennessee was a ere $8,895 Report of Statistics Required by the Bankruptcy Abuse Prevention and Consuer Protection Act of 5, Table X. Or, as another exaple, no study has adessed why the average bankrupt household in Texas had nearly $, in assets while the coparable figure in Tennessee was Rutgers Business School, Departent of Finance and Econoics, Washington Park, Roo 54, Newark, New Jersey 7 eail: iller@business.rutgers.edu. I would like to thank Robert A. Margo, Lars J. Lefgren, Randall P. Ellis,and Erik Hurst for their helpful feedback. In addition, I would like to thank attorneys in the Business Reorganization and Restructuring group at Jones Day for their coents and legal expertise. Finally, I a grateful to the seinar participants at Boston University, Rutgers Business School and the St. Louis Federal Reserve Bank. All reaining errors are y own.

2 only $84, Report of Statistics Required by the Bankruptcy Abuse Prevention and Consuer Protection Act of 5, Table X. These cross-state differences suggest that soe households who seek bankruptcy protection in one state, are resorting to alternative easures in other states. This paper exaines whether state exeption and garnishent laws can explain the cross-state differences in the characteristics of bankrupt households. Using a new dataset of households who filed for bankruptcy on February, 7, I find that high exeption levels encourage high asset households to file for bankruptcy, whereas high garnishent rates encourage low incoe households to file for bankruptcy. This eans that state exeption and garnishent laws ipact who files for bankruptcy. In other words, these state laws fundaentally affect the extent of insurance offered by the bankruptcy syste. This finding has iportant policy iplications--- by changing their exeption and garnishent laws, states can change the characteristics of bankrupt households. Indeed, as bankruptcy is a federal procedure, the federal governent ay wish to enact legislation, so as to haronize bankruptcy use across states. My findings are supported by a theoretical odel in which households choose between three options: repayent, bankruptcy, and non-response which occurs when households siply walk away fro their debts, allowing creditors to garnish their wages and seize their assets. This odel is a new addition to the literature whereas prior papers provide epirical evidence that households choose between repayent, bankruptcy and non-response Agarwal, Liu, and Mielnicki 3; Dawsey and Ausubel 4, existing theoretical odels largely ignore the option of non-response Gropp, Scholz and White 997; Nelson 999; Adler, Polak and Schwartz ; Wang and White ; Athreya ; White 5; Li and Sarte 6; Pavan 8. Including non-response changes the household s utility axiization proble as well as the lender s profit axiization proble. As a result, y odel akes unique predictions about who files for bankruptcy. For exaple, y odel predicts that bankruptcy is ore popular aong iddle rather than lower incoe households. While recent statistics have shown this to Even White 998, ultiately views bankruptcy as the household s only alternative to repayent. In White 998, households first choose between full repayent and default. If a household defaults, the creditor ay garnish the household s wages. In response to garnishent, a household will either repay its debts in full or file for bankruptcy. Thus, ultiately, households either repay in full or file for bankruptcy. In y odel, households choose between three options: full repayent, bankruptcy or non-response. When a household chooses non-response, a creditor will garnish its wages and seize its non-exept assets. However, the household will not react to these collection efforts; this is different fro White 998 in which households respond to collection efforts. As the household s actions are different, the payoffs of both debtors and creditors are different than those seen in White 998.

3 be true Sullivan, Warren, and Westbrook ; Sullivan and McIntyre, previous theoretical odels did not support this finding. To be sure, nuerous papers have explored the ipact of exeption laws on bankruptcy filings, albeit with conflicting results--- soe papers found that high exeption laws do not ipact bankruptcy while other papers found that high exeption laws encourage bankruptcy. For exaple, using aggregate data, Peterson and Aoki 984, Fisher, Weiss, Bhandari and Robins, and Lefgren and McIntyre 9 regressed an area s filing rate on its exeption level. As the coefficient of interest was insignificant, these studies concluded that high exeption laws do not ipact bankruptcy. As explained below however, aggregate data can only estiate the ipact of state laws on the average household. Indeed, in this paper, I also show that ost people are unaffected by exeption levels. However, in other studies using household level data, it is evident that soe households are affected by exeption levels. For exaple, White 987, Dawsey and Ausubel 4, Agarwal, Liu, and Mielnicki 3, and Lin and White regressed a household s bankruptcy decision on the state s exeption level. As the coefficient of interest was positive and significant, these studies concluded that high exeption laws encourage bankruptcy. These regressions, however, only included level ters as opposed to interaction ters. As a result, they overstate the ipact of exeption laws on low asset households and understate the ipact of exeption laws on high asset households. By including both level and interaction ters, this paper accurately estiates the heterogeneous ipact of exeption laws on bankruptcy; it shows that whereas high exeption laws have little ipact on the average household the average household has few assets, they have a significant ipact on households with high assets. I. Institutional Background Under Article I, Section 8 of the United States Constitution, Congress has the authority to establish unifor laws on the subject of bankruptcy throughout the United States. Although bankruptcy is governed by federal law, states have enacted a variety of statutes which are thought to influence a household s bankruptcy decision. This section begins with a discussion of As an exception, using household level data fro the Panel Study of Incoe Dynaics PSID,Fay, Hurst, and White exained whether the likelihood of filing depends on the financial benefit of bankruptcy. Financial benefit is defined as the debt that can be discharged less nonexept assets a debtor loses by filing. Unfortunately, with this definition, the ipact of exeption levels cannot be disentangled fro additional units of assets or debts. 3

4 state exeption and garnishent laws, detailing the aatic variations across states. Then it continues with a discussion of federal bankruptcy law, noting how each type of state law ay ipact a household s bankruptcy decision. A. State Exeption Laws Exeption laws protect an individual s unsecured assets both inside and outside of bankruptcy. There are two types of exeption laws: personal exeption laws and hoestead exeption laws. Personal exeptions allow individuals to retain personal property such as jewelry and appliances while hoestead exeption laws protect a household s residence. In this paper I concentrate on hoestead exeption laws, as they are substantially larger and ore readily quantified. 3 These exeption levels vary significantly across states--- five states Florida, Iowa, Kansas, Oklahoa and Texas have unliited hoestead exeptions eaning that literally an unliited aount of hoe equity is protected fro seizure by creditors. On the other hand, three states Maryland, New Jersey and Pennsylvania have no hoestead exeptions; in these states, creditors can seize all hoe equity. It should be noted that exeption laws protect households against involuntary liens or unsecured debts. However, they do not protect households against voluntary liens or consensual secured debts. Thus, exeption laws do not protect households against foreclosure. Property exeption laws have a long history in the United States. The first hoestead exeption law was passed in Texas then the Republic of Texas in 839. Coupled with free land, Texas hoped hoestead exeptions would attract Aerican settlers Hynes, Malani and Posner 3. 4 By 86, nearly every state in the nation had adopted a hoestead exeption law. And the regional patterns which existed in the 86s still persist today. Hynes, Malani, and Posner 3 deonstrate that the best predictor of a state s current exeption level is its 3 Personal property exeptions list the type of personal property an individual can protect in bankruptcy. Typically, they refer to categories of basic necessities, like food, clothes, furnishings or tools of trade, but soeties they refer to specific ites, like herds of sheep or ilitary unifors. However, unlike hoestead exeption laws, ost personal exeption laws do not place a dollar value on protected assets; as a result, they are difficult to quantify. In Delaware for exaple, a household s clothing, jewelry, books, faily portraits, piano, leased organs, sewing achines, burial plot, and church pew are all protected by personal exeption laws. Siilarly, in Texas, personal exeption laws protect the faily Bible, household pets, two horses with saddle, blanket and bridle for each, twelve head of cattle, fowl, and any food on hand for these anials. Using several different calculation techniques, Hynes, Malani and Posner 3 deterined that the average personal exeption law protects approxiately $6, of assets. By coparison, the average hoestead exeption law protects approxiately $5, of assets. Given their size, I a generally not concerned about oitted variable bias. 4 Iigration guides specifically advertised a generous hoestead exeption, good soil, and a cliate suitable for cotton. 4

5 historical exeption level. Indeed, any of the current laws still contain extreely archaic provisions. 5 B. State Garnishent Laws When a household fails to repay its debts, creditors ay garnish the household s incoe. Garnishent can be taken for any type of debt--- coon exaples include defaulted child support, taxes, court fines, and student loans. Wages, salaries, coissions, bonuses, and incoe fro retireent progras can all be garnished. Garnishent procedures first appeared in Maryland s State Statue in 683. Soon other states and territories passed siilar rulings. In the 96s, concern arose regarding garnishent abuses. Therefore, in 968, Congress enacted the Federal Consuer Credit Protection Act CCPA. According to this Act, 75 percent of wages or 3 ties the federal iniu wage per week whichever is higher is protected fro garnishent. 6 In addition, the CCPA prohibits eployers fro firing a worker due to garnishent. The CCPA also allows states to enact their own garnishent laws, provided that these laws protect a greater portion of borrowers wages than the federal share. Four states--- Florida, Pennsylvania, South Carolina, and Texas--- currently prohibit wage garnishent except for debts related to taxes, child support, federally guaranteed student loans, court-ordered fines or restitution for a crie the debtor coitted. An additional states have thresholds that are higher than the federal law. C. Personal Bankruptcy in the United States The nuber of Aerican households seeking bankruptcy relief has increased substantially over the past three decades. While less than 5, filed for bankruptcy in 978, over illion households filed for bankruptcy in 5. In response, Congress enacted the Bankruptcy Abuse Prevention and Consuer Protection Act of 5 BAPCPA. 7 5 In Oklahoa, for exaple, a debtor can exept a gun, twenty head of sheep, and all provisions and forage on hand; see 3 Okl. St.. 6 On February, 7, the federal iniu wage was $5.5. Thus, at the tie y saple was collected, at least $ x $5.5 of weekly wages was exept fro garnishent. When an eployee s weekly wages exceeded $54.5 but were less than $6., only the aount over $54.5 could be garnished. For exaple, if an eployee earned $65 in a particular week, only $5.5 could be garnished. When an eployee s earnings were $6. or ore in a given week, up to 5 percent of those earnings could be garnished. For an eployee earning $5. a week, 5 percent of his earnings or $6.5 could be garnished while $87.5 had to be paid to the eployee. 7 The BAPCPA ade several aendents to the Bankruptcy Code in hopes of curbing abusive filings. For exaple, the BAPCPA hoped to decrease the nuber of repeat filers. However, Miller and Miller 8 found that the law had virtually no ipact on the rate of repeat filings. 5

6 For households facing serious debt, there are any advantages to filing for bankruptcy. First, through the bankruptcy process, households are able to discharge legally default on uch of their debt--- the average household discharges approxiately $36, of debt upon filing for bankruptcy Culhane and White 999. In addition, once the household files for bankruptcy, all wage garnishent ust stop. In states where a sizeable portion of wages are garnished, this is clearly beneficial to the household s cash flow. However, in states where only a sall portion of wages can be garnished, bankruptcy ay be less advantageous. Regardless of the financial benefits, bankruptcy is a costly endeavor. In addition to the nearly $3 filing fee, households ust pay attorney fees, which average $,83 Palank 8. Moreover, there are non-pecuniary costs to bankruptcy, including the stiga of bankruptcy and future restrictions fro the credit arket. And finally, households ust either forfeit a portion of their assets or a portion of their future incoe. The United States has two priary procedures for personal bankruptcy--- Chapter 7 and Chapter 3. Chapter 7 is intended for households with little or no incoe while Chapter 3 is designed for households with regular incoe. Chapter 7: Liquidation.---Under Chapter 7 of the Bankruptcy Code, households ust liquidate all of their nonexept assets; these are the assets above the exeption levels discussed above. 8 As entioned, five states have unliited hoestead exeption levels; in these states debtors will not be forced to sell their house to payoff their creditors. However, in the three states with no exeption level, households ust surrender their hoe. Clearly, bankruptcy is ore advantageous in states with higher exeption levels. After nonexept assets are liquidated, the proceeds are distributed aongst the household s creditors. The case is then closed and the household s reaining debts are discharged. Most unsecured debts, including credit card debts, installent loans, edical debts, unpaid rent and utility bills, tort judgents, and business debts, can be discharged under Chapter 7. Under The authors found that the law did not appear to affect who repeatedly filed for bankruptcy; the financial description of repeat filers reained the sae before and after the BAPCPA. Instead, the BAPCPA sees to have increased the tie debtors waited between filings. The BAPCPA also hoped to prevent high incoe households fro filing under Chapter 7 of the Bankruptcy Code. In particular, the eans test required that households earning incoe above the state edian level file under Chapter 3. However, ost papers have found this incoe restriction rarely applies Flynn and Berant, ; Culhane and White, 999; Tabb and McClelland, 7; White, 7. 8 In soe states a debtor is required to file a declaration of hoestead exeption for such exeption to be enforceable. However, the household can file this declaration post-petition, so as a practical atter, this requireent does not often interfere with a debtor's ability to clai a hoestead exeption. Accordingly, this paper does not ake a distinction between states which autoatically provide hoestead exeptions, and those which require the filing of a hoestead declaration In re Michael, 63 F.3d 56 9th Cir

7 Chapter 7 of the Bankruptcy Code, the debtor ay keep all of his future earnings. However, not all households can file for Chapter 7 relief. Households who filed for Chapter 7 within the past seven years are ineligible. 9 Additionally, the BAPCPA precludes debtors with incoe above their state s edian incoe fro filing under Chapter 7. Chapter 3: Debt Reorganization.---Under Chapter 3, households retain all of their assets and instead agree to repay soe of their debts fro future earnings. Debtors pay their projected onthly disposable incoe the difference between their onthly incoe and onthly budgeted living expenses into the Chapter 3 repayent plan. After sixty onths, the case is closed, and any reaining debts are discharged. It is iportant to note that the repayent plan ust copensate creditors at least as uch as they would receive under Chapter 7. All debtors are peritted to file under Chapter 3 of the Bankruptcy Code. Debtors can file for Chapter 3 bankruptcy every two years. In 7, less than 4 percent of cases were filed under Chapter 3 of the Bankruptcy Code. The ajor benefit of filing a Chapter 3 is asset retention. In addition, a Chapter 3 bankruptcy only reains on a debtor s credit report for seven years; by coparison, a Chapter 7 bankruptcy reains on a debtor s credit report for ten years. Thus, there is reason to believe that Chapter 3 filers have better access to future credit. But, copared to Chapter 7, Chapter 3 is an extensive and lengthy process--- under Chapter 3, households ust ake onthly payents for five years. In contrast, Chapter 7 cases typically last less than six onths. 3 As shown in Lefgren, McIntyre and Miller, the best observable predictor of chapter choice is the consuer s attorney; households are likely to file under Chapter 3 if they consult a bankruptcy lawyer who specializes in Chapter 3 cases. To conclude this section, Table provides a suary of state exeption and garnishent laws; with this table, the cross-state variation in these laws becoes apparent. In ters of hoestead exeptions, while five states have unliited hoestead exeptions, three states have 9 Specifically, the Bankruptcy Code states that debtors are only eligible for a Chapter 7 discharge every eight years. The Chapter 3 Trustee distributes these funds between the household s creditors. Specifically, if a debtor has previously filed for under Chapter 3 of the Bankruptcy Code, he is eligible for another discharge after two years. And if the debtor has previously filed under Chapter 7 of the Bankruptcy Code, he is eligible for a Chapter 3 discharge after four years. If a household fails to ake the court andated payents the case is disissed and the household is once again liable for all its debts. Lefgren, McIntyre and Miller 8 report that 6 percent of bankruptcies filed under Chapter 3 of the Bankruptcy Code are disissed largely due to nonpayent on the debtor s part. 3 In fact, if a household does not have any non-exept assets, the case will typically only last one onth. 7

8 none. And although 4 states allow garnishent up to the federal liit, 6 states have set thresholds higher than the federal law, including four states which prohibit wage garnishent altogether. 4 II. Literature Review To be sure, nuerous papers have explored the ipact of state laws on bankruptcy filings; however, none of these papers studied their heterogeneous ipact. For exaple, using aggregate data, any papers regressed state Apilado, Dauten and Sith 978; Shiers and Williason 987; Weiss, Bhandari and Robins ; Fisher, district Buckley and Brinig 998, county White 987, or zip code Lefgren and McIntyre 9 filing rates on legal variables. Most of these papers found no statistical relationship between exeption laws and bankruptcy. 5 However, aggregate data cannot estiate the heterogeneous ipact of state laws across households--- aggregate data only captures the ipact of state laws on the average household. Indeed, in this paper, I also show that the average household is unaffected by exeption levels; this is because the average household have few assets. However, in other studies using household level data, it is evident that soe households are affected by exeption levels. These papers regressed a household s bankruptcy decision on state laws. For exaple, using individual-level data fro a credit card issuer, Dawsey and Ausubel 4 and Agarwal, Liu, and Mielnicki 3 found that high hoestead levels and high garnishent rates encourage bankruptcy. Siilarly, using data fro the Hoe Mortgage HMDA, Lin and White found that exeption laws increase bankruptcy. Because theses regressions included level, but not interaction, ters, they overstate the ipact of exeption laws on low asset households and understate the ipact of exeption laws on high asset households. By including both level and interaction ters, this paper accurately estiates the heterogeneous ipact of exeption laws on bankruptcy. It should be noted that using household-level data fro the Panel Study of Incoe Dynaics PSID, Fay, Hurst and White exained whether the likelihood of filing depends on the financial benefit of 4 As shown in Lefgren and McIntyre 9 these state laws are not correlated. It is not the case that a state protects households fro their creditors by passing both high hoestead exeption laws and low garnishent rates. Indeed the correlation between these two laws is.. 5 One notable exception is White 987 which found that high exeption levels were correlated with high bankruptcy rates. However, the precision of the results has been debated. As entioned in Lefgren and McIntyre 9, because the analysis is perfored at the county level, and the author did not correct the standard errors for clustering within states, it is likely that the precision of the estiates is overstated. 8

9 bankruptcy. Financial benefit is defined as the debt that can be discharged less nonexept assets a debtor loses by filing. Unfortunately, with this definition, the ipact of exeption levels cannot be disentangled fro additional units of assets or debts. This paper also builds on the theoretical literature on personal bankruptcy. Whereas prior papers provide epirical evidence that households choose between repayent, bankruptcy and non-response Agarwal, Liu, and Mielnicki 3; Dawsey and Ausubel 4, 6 existing theoretical odels largely ignore the option of non-response Gropp, Scholz and White 997; Nelson 999; Adler, Polak and Schwartz ; Wang and White ; Athreya ; White 5; Li and Sarte 6; Pavan 8. Including non-response changes the household s utility axiization proble as well as the lender s profit axiization proble. As a result, y odel akes unique predictions about who files for bankruptcy. For exaple, prior odels including Lehnert and Maki predict that households file for bankruptcy if incoe falls below a certain level. However, recent statistics have shown that bankruptcy is ore popular aong iddle rather than lower incoe households Sullivan, Warren, and Westbrook ; Sullivan and McIntyre. By including non-response, y odel predicts that bankruptcy is ore popular aong iddle rather than lower incoe households. III. Model In this section, I develop a two-period odel of bankruptcy. The purpose of this odel is to show how state exeption and garnishent rates affect a household s bankruptcy decision. In the first period, households learn their first period incoe, and choose their consuption, assets, and debts. 7 In the second period, ared with their assets and debts, households learn their second period incoe, and decide whether to file for bankruptcy. Risk neutral lenders are assued to ake zero profits. In order to be consistent with bankruptcy law, assets represent hoe equity while debt represents unsecured loans such as credit card debts. This odel does not distinguishing between the value of the hoe and the aount of the ortgage; in order to ake this distinction, 6 Agarwal, Liu, and Mielnicki 3 and Dawsey and Ausubel 4 use the ter inforal bankruptcy to refer to non-repayent without the benefit of forally filing for bankruptcy. 7 Without bankruptcy, a consuer will either borrow or save; he will not do both. However, when bankruptcy is available, households ay borrow and save siultaneously. Although debt requires paying an interest rate preiu, the prospect of not having to repay the debt in full lowers this preiu. 9

10 the odel would need to include an additional choice variable. Again, the purpose of this odel is to show how state exeption levels ipact a household s bankruptcy decision. As exeption levels only apply to non-consensual secured debts, this siplification should not be worrisoe. 8 Extending the odel to include the value of the hoe and the aount of the ortgage is left to future works. This odel akes two extensions to previous works. First, it expands the traditional oneperiod bankruptcy odel to a two-period odel. Thus, the odel allows assets, debts, and the interest rate to depend on the state s exeption or garnishent laws; any earlier works have assued that these variables are exogenous. 9 Additionally, in this odel, households choose between three options--- repayent, bankruptcy and non-response. Previous odels have viewed bankruptcy as the consuer s only alternative to repayent. However, households can siply ignore collection letters and bills--- such behavior is deeed non-responsive. Nonresponsive households are subject to both wage garnishent and asset seizure. While Agarwal, Liu, and Mielnicki 3 and Dawsey and Ausubel 4 discuss non-response, it has never been explicitly odeled. Non-response is an iportant alternative to consider. In 7, when this data for this paper was collected, 7. percent of U.S. failies reported to be at least 6 days behind on one of their loans Bucks et al. 9. A. The Household s Proble In this odel, risk averse households live for two periods and axiize their expected lifetie utility. In the first period, households receive incoe. However, second period incoe is uncertain; is uniforly distributed between and. In the first period, households choose their first period consuption C, assets A which can be used for future consuption, and debt D. Note that lenders charge interest rate r on a loan; as discussed below, I allow the interest rate to depend on the household s assets and debts as well as the state exeption level 8 This odel does not adess consensual secured loans. Again, the purpose of this odel is to show how state exeption levels ipact a household s bankruptcy decision. Fundaentally, hoestead exeptions do not protect a hoeowner against foreclosure by a ortgagee. Hoestead exeptions apply only against involuntary liens or unsecured debts. Because exeptions are not relevant to consensual secured debt, ortgages are not odeled here. 9 Repetto 998 and Pavan 8 evaluated the ipact of exeption laws on household asset accuulation. However, neither paper considered the ipact of garnishent rates. Furtherore, neither paper included non-response in their theoretical odel.

11 and garnishent rate. Thus, the household s first period budget constraint is given by C A D. In the second period, ared with assets A and debt D+r, a consuer learns his incoe and decides whether to file for bankruptcy. If he files for bankruptcy under Chapter 7, the consuer ust surrender nonexept assets and pay a filling fee. All reaining debts are then discharged forgiven. Thus, if the consuer files for bankruptcy under Chapter 7 his consuption is given by C + ina,e - F where E represents the state exeption level and F denotes the filing fee. If the consuer files for bankruptcy under Chapter 3 of the Bankruptcy Code, he ust surrender his incoe and pay a filing fee. However, the consuer ay keep all its assets. Again, all reaining debts are then discharged forgiven. Thus, the consuer s consuption under Chapter 3 of the Bankruptcy Code is given by C A F. According to the Bankruptcy Code, households ust repay creditors at least as uch in a Chapter 3 repayent plan as in a Chapter 7. In addition, creditors can petition to convert a Chapter 7 into a Chapter 3 if it is in their best interest. Therefore, if the household files for bankruptcy, its consuption is given by C in[ in A, E F, A F]. According to this equation, and in accordance with the Bankruptcy Code, higher incoe households file under Chapter 3 rather than Chapter 7. If a consuer does not file for bankruptcy, his second period consuption is given by C A Q, A, D, r. When the household repays its debt in full Q,A,D,r = D+r. When the household chooses non-response, Q,A,D,r=g +axa-e,, where g denotes the state s garnishent rate. Recall that when a consuer fails to repay his debts, he is subject to This set-up assues that assets pay an interest return noralized to unity while debts carry a gross interest rate of r. This siplification does not alter y odel s predictions. In addition to the filing fee, F incorporates attorney fees, the stiga of bankruptcy, and the cost of future credit arket exclusion. More realistically, when filing under Chapter 3, a consuer surrenders a portion of its incoe and his consuption is given by C β + A - F. According to the eans test, a household can deduct standardized expenses fro its incoe; standard deductions can be taken for healthcare, housing, utilities, vehicle expenses, and public transportation expenses. In addition, to the standard deductions, a household can also deduct actual spending on taxes, payroll deductions, ter life insurance, court-ordered payents, utilities and health care. Therefore, as entioned above, it ay be ore realistic to state that when filing under Chapter 3, a household s consuption is given by C β + A F. However, the siplification in the text does not alter y odel s predictions.

12 wage garnishent at rate g and creditors ay sue the household for his non-exept assets, axa-e,. Thus, Q,A,D,r=in[g + axa-e,, D+r]. 3 Therefore in the second period, households file for bankruptcy when in[ in A, E F, A F] A Q, A, D, r or equivalently, when in[ in A, E, A] F A Q, A, D,. r For descriptive purposes, Figure depicts the household s second period decisions for a household with sufficiently large assets A>E and sufficiently large debts D+r>F. The vertical axis depicts second period consuption while the horizontal axis depicts all the possible realizations of second period incoe. Consuption can be divided into three regions. In the F left-ost region, where, households are non-responsive--- they do not repay their debt g and, as a result, wage garnishent and asset seizure take place. These households avoid bankruptcy and its high filing fees without repaying their debts in full. In the iddle region where F g F g D r F households file for bankruptcy. In particular, households with A E will file under Chapter 7 of the Bankruptcy Code while households with A E D r F will file under Chapter 3. Finally, in the right-ost region where D r F households repay their debts in full. As seen clearly in this figure, the odel predicts that bankruptcy is priarily used by iddle incoe households. With the second period decisions detailed above, the household s axiization probles can 4, 5 be written as: 3 Households who choose non-response ay also suffer fro non-pecuniary repercussions such as stiga or future exclusion fro the credit arket. One could write Q,A,D,r=in[g+ axa-e,-f, D+r] where f incorporates the stiga of non-response and the cost of future credit arket exclusion. Because f < F, this siplification does not alter the odel s predictions. 4 I assue that the household s utility is only defined over their consuption. For an alternative odel see Livshits, MacGee and Tertilt 7. 5 It is iportant to note that households do not choose their exeption level E. According to Section 548b3A of the Bankruptcy Code, debtors cannot anipulate hoestead exeptions by oving to another state. This section states that if a debtor oved within the past 73 days, he is bound by the exeption level of the state in which he previously resided.

13 ax U A, D A D D r F ax AE, F / g U A F U g D r F in A, E U A D r ax AE, F / g U in A, E F The household s first order condition with respect to debt is as follows: 6 U A D r D U A D r. dd D r F The first ter of this equation denotes the first period utility gain associated with an additional unit of debt. An additional unit of debt only affects second period utility if the household chooses repayent; the second ter of this equation captures the second period utility loss associated with additional units of debt, integrated over the region of repayent. The first order condition with respect to assets is: U A D F / g U g A d in A, E * [ D r F ax A E,] U A F D ax AE, F D r F d in A, E U ' A F * / g U A D r The first ter of this equation represents the first period utility loss associated with an additional unit of assets. An additional asset increases second period utility if the household chooses nonresponse and has assets below the exeption level, bankruptcy under Chapter 7 and has assets below the exeption level, bankruptcy under Chapter 3, or repayent. The second ter denotes the second period utility gain if the household chooses non-response, when the consuer does not have non-exept assets, integrated over the region of non-response. The third ter captures the arginal utility of assets when the consuer files for bankruptcy under Chapter 7, when the consuer does not have non-exept assets, integrated over the region of Chapter 7. The fourth ter equals the arginal utility of assets when the household files for bankruptcy under Chapter 3, ultiplied by the probability that a household files for bankruptcy under Chapter 3. Finally, the last ter denotes the arginal utility of assets when the household chooses repayent, integrated over the region of repayent. x d d d dg G x t dt G x x G x x dt 6 dx,,, dx dx dx Leibnitz rule x x is used to find the first order conditions. x 3

14 B. The Lender s Proble Lenders are assued to be risk neutral. Their zero profit condition is given by: F / g [ g ax A E, ] ax AE, F / g ax A E, D r F ax AE, D r F D r D r f where the first ter represents partial repayent fro non-response, the second ter represents partial repayent under Chapter 7, the third ter represents partial repayent under Chapter 3, the fourth ter represents full repayent and r f is the risk free rate of return. If no interest rate satisfies this equation, lenders will not lend. I assue that lending is liited by D+r <. The lender s first order condition with respect to the interest rate is: d ax A E, dd ax A E,* [ D r ]*[ D r] The first ter represents the lender s gain if the household files under Chapter 7 of the Bankruptcy Code and has non-exept assets. Recall that when a household files under Chapter 7 of the Bankruptcy Code, it ust use its non-exept assets to repay its debts. If a lender increases the interest rate, households will increase their assets; thus, when a household files for bankruptcy under Chapter 7, the lender will receive a higher payent. The second ter represents the lender s gain if the household repays its debts in full. This ter represents several forces. If a lender increases the interest rate, households will decrease their debts. In addition, households are less likely to repay in full. C. Predictions about Assets, Debts and the Interest Rate The first order conditions are totally differentiated and Craer s rule is used to solve for /de. 7 High exeption levels protect assets should the household choose non-response or bankruptcy. Thus, lenders collect less when households choose non-response or if they file for bankruptcy. As a result, lenders will charge a higher interest rate in states with high exeption levels. 7 For ore details, see the Matheatical Appendix. 4

15 Siilarly, after totally differentiating the first order conditions and applying Craer s rule, the odel predicts that /de and dd/de. Thus, in high exeption states, households will siultaneously accuulate assets and debts; Lehnert and Maki refer to this behavior as borrowing to save. High exeption levels protect assets should the household choose nonresponse or bankruptcy. Thus, under either of these options, households can retain a larger aount of assets without having to repay their debts. This creates an incentive for households to accuulate additional assets at the cost of additional units of debts. The odel predicts siilar behavior fro households in high garnishent states /dg and dd/dg. A household is ore likely to file for bankruptcy in a high garnishent state. And because it is preferable to have high assets and high debts when filing, households in high garnishent states will siultaneously accuulate both assets and debts. Furtherore, the odel predicts that /dg. In states with high garnishent rates, lenders can garnish a larger portion of the household s incoe. This creates two opposing effects. Because lenders collect ore when a household chooses non-response, high garnishent rates create an incentive for lenders to lower interest rates. On the other hand, ore households will file for bankruptcy instead of choosing non-response; these households will repay less debt. This creates an incentive for lenders to increase interest rates. This second effect is larger, and, as a result, the interest rate is higher in states with high garnishent rates. D. Predictions About Bankruptcy Again, households file for bankruptcy when in[ in A, E F, A F] A Q, A, D, r or equivalently, when in[ in A, E, A] F A Q, A, D,. Fro this condition, the odel akes r several predictions about the likelihood of bankruptcy. Predictions about the Likelihood of Bankruptcy: 8 High exeption levels encourage high asset households to file for bankruptcy. High garnishent rates encourage low incoe households to file for bankruptcy. The likelihood of bankruptcy is decreasing in A. The likelihood of bankruptcy is increasing in D. 8 For ore details, see the Matheatical Appendix. 5

16 IV. Data Description To test the theoretical odel detailed above, I use a choice-based sapling technique. 9 As data on bankrupt households are difficult to obtain, epirical studies on personal bankruptcy often rely on choice-based sapling. For exaple, Doowitz and Sartain 999 atch a saple of households who filed for bankruptcy in 98 with households fro the 983 Survey of Consuer Finances SCF. Siilarly, Zhu forthcoing augents bankruptcy filing data fro 3 with detailed household and consuption data fro the 4 SCF. The dataset used in this paper coes fro two distinct sources: inforation on bankrupt households was handcollected fro bankruptcy petitions while data on non-bankrupt households cae fro the Panel Study of Incoe Dynaics PSID. It should be noted that the ideal dataset would cobine three saples of households: households who chose non-response, households who chose bankruptcy, and households who chose repayent. Unfortunately, a large dataset on non-responsive households is not readily available. In Section VI.D., I create a saller alternative choice-based dataset, in which households chose between these three options. My findings are robust to this alternative specification. A. Bankruptcy Saple When filing for bankruptcy, debtors ust coplete a bankruptcy petition: a detailed for describing their finances. 3 This petition is then archived on PACER Public Access to Court Electronic Records, the court s centralized registration and billing website. With a subscription to this website, I created a unique dataset of the,694 households that filed for bankruptcy on February, 7. 3,3 9 Choice-based sapling arises when selection into the saple is deterined by the dependent variable. Manski and Leran 977, as discussed in ore detail below, show that choice based sapling techniques can correct for the probles of the endogenous saple and ensure consistent and unbiased estiates. 3 Households provide inforation under penalty of perjury. In addition, the household s attorney is required to verify all financial stateents. 3 As the petitions are only available in a PDF forat, the relevant inforation then had to be extracted into a ore usable forat. This dataset is a nationwide census of households who filed for bankruptcy on February, 7. 3 There are several advantages to collecting data fro February, 7. Because I collected data fro 7, I was able to collect a nationally representative dataset--- prior to 7 any bankruptcy courts did not use PACER s Case Manageent/Electronic Case Files CM/ECF syste. Therefore, it would not have been possible to collect a nationally representative saple of cases electronically. February st was selected as a rando date at the beginning of the year; the bankruptcy petitions collect inforation on the household s incoe fro the previous two calendar years. Thus, for petitions filed on this date, I likely have accurate incoe inforation. 6

17 Bankruptcy petitions provide an abundance of financial inforation. First, households iteize their unsecured debts including credit card debts and edical bills. 33 In addition, households ust record the value of their hoe as well as any secured clais on their hoe. Thus I can calculate the household s hoe equity; because exeption laws protect equity in one s hoe equity, I used hoe equity as y easure of assets. In addition, debtors ust provide coprehensive incoe inforation--- they detail their incoe for the past two calendar years. Finally, court records provide the debtor s adess, arital status, and household size. Other deographic inforation, including age and education are unfortunately, unavailable. However, since I know the filer s adess, I use Census block statistics to estiate these deographic characteristics B. Control Saple I augent the choice-based saple of bankrupt households with households fro the 7 PSID. Because bankrupt households are not identified in this study, I begin by assuing that none of the households in the PSID filed for bankruptcy. This assuption is tested in Section VI.C.. Each of the 6,996 households in the PSID details its incoe fro the previous two calendar years. In addition to annual easures of faily incoe, the PSID provides a detailed inventory of the faily s hoe equity and unsecured debts. Deographic inforation is also provided. Because y data was collected in 7, y saple consists of data fro cases filed after the BAPCPA--- this paper is one of the first to exaine bankrupt households since the Bankruptcy Code was aended in 5. White and Zhu also collected electronic bankruptcy petitions fro households who filed for bankruptcy after 5. Their saple contained 586 households who filed a Chapter 3 bankruptcy in 6 in Delaware. 33 In particular, fro Schedule F, I extract inforation on a household s unsecured non-priority debts. In general, these are the debts which households can discharge in bankruptcy. My results are statistically siilar when I include unsecured priority debts fro Schedule E. Unsecured priority debts include doestic support obligations and unpaid taxes. 34 Prior research suggests that issing deographic data should not be a concern. In fact, earlier works argue that once financial characteristics are controlled for, deographic characteristics are irrelevant. In interviewing 4 households, Stanley and Girth 97 found that deographic considerations did not influence the bankruptcy decision. Doowitz and Eovaldi 993 confired this result; they found that race was statistically insignificant in explaining aggregate filing rates. As shown in the third colun of Table 3, y results do not change when I exclude the iputed deographic variables. 35 Bankruptcy petitions contain a wealth of inforation. For exaple, in addition to the variables entioned above, Agarwal, Chosisengphet, McMenain and Skiba ention that bankruptcy petitions include inforation on onthly expenditures, nuber of creditors, and previous bankruptcy filings. However, as y control saple does not contain these covariates, they are not included in this paper. 7

18 C. Descriptive Statistics Suary statistics, coparing bankrupt and non-bankrupt households, can be found in Table. Several facts are worth noting. While bankrupt households earned nearly $, less than non-bankrupt households in the first year 5, and over $37, less than non-bankrupt households in the second year 6. Not surprisingly, bankrupt households are also characterized by lower levels of hoe equity $9,833 copared to $95,46. And, bankrupt households have higher levels of debt; while households in the PSID have an average $8,7 of unsecured debt bankrupt households have an average $5,876 of unsecured debt. Known deographic variables including faily size and arital status are siilar for households in the bankrupt and suppleental saples. There are two differences between the saples--- although less likely to be arried and ore likely to be single or separated, bankrupt households are larger. The iputed deographic characteristics, based off the bankrupt household s census block, indicate that bankrupt households are younger. V. Estiation A. Probability of Bankruptcy In estiating the ipact of state laws on a household s bankruptcy decision, y priary specification is: 36 ProbBankrupt = α + α High Exeption + α High Exeption*Assets + α 3 High Garnishent + α 4 High Garnishent*Incoe + α 5 Assets + α 6 Debt + α 7 X + α 8 Region + ε it where Bankrupt is a duy variable equal to one if the household files for bankruptcy. High Exeption 37 and High Garnishent 38 denote high exeption levels and high garnishent rates 36 Again, y theoretical odel supports a ultinoial logit regression as households are allowed to choose between non-response, bankruptcy and full repayent. However, y dataset does not distinguish between non-response and full repayent. Therefore y ain epirical analysis eploys a logit regression, estiating the probability that a household files for bankruptcy versus the alternative that they do not file and choose either non-response or repayent in full. For an alternative specification, using a saller dataset, see Section A survey of state exeption laws reveals that one third of states have exeption levels greater than the ean exeption of $5,. Accordingly, the duy variable High Exeption equals one if the household lives in a state with a hoestead exeption greater than $5,. As seen in Table 3, y findings are robust to different definitions. 8

19 respectively. Assets, Debt, and Incoe are a household s assets easured by hoe equity, debts easured by unsecured debts, and incoe. X is a vector of other variables including incoe, change in incoe, and deographic characteristics. Finally, Region is a vector of regional duy variables. Recall, the odel predicts that high exeption laws encourage high asset households to file for bankruptcy α = and α > and that high garnishent rates encourage low incoe households to file for bankruptcy α 3 > and α 4 <. Finally, the odel predicts that the likelihood of bankruptcy is decreasing in assets and increasing in debts α 5 < and α 6 >. If the saple were rando, equation could be estiated using a siple logit regression. However, the saple at hand is not rando; instead it is choice-based. 39 Therefore, as detailed in Manski and Leran 977 equation ust be estiated using weighted exogenous sapling axiu likelihood estiation WESMLE. By controlling for the oversapling of bankrupt households, I a able to obtain consistent estiates. Specifically, I weight each ter by the inverse of the ex-ante probability that an observation is included in the saple. Let Q denote the fraction of the population that is bankrupt and H denote the fraction of the saple that is bankrupt. Then each bankrupt household is weighted by Q / H and each non-bankrupt household is weighted by -Q /-H. Regression coefficients and arginal effects are reported in Table 3. All errors are clustered at the state level. In colun I, I estiate the probability of bankruptcy without any interaction ters. My estiates are saller than earlier works for exaple: Agarwal, Liu, and Mielnicki 3; and Dawsey and Ausubel, 4 because I a exaining behavior after the BAPCPA of As discussed in Section, under federal law, creditors ay garnish twenty five percent of wages. Many states, however, have placed further liits on the garnishent rate. The duy variable High Garnishent equals one in states that use the federal garnishent liits. As seen in Table 3, y findings are robust to alternative specifications. 39 Choice-based sapling arises when selection into the saple is deterined by the dependent variable. Like any surveys, I have oversapled an infrequently ade choice. With such a dataset, logit estiates will be biased and inconsistent Manski and Leran 977. To see the inconsistency of standard binary choice ethods, consider the logit odel when the only regressor is the intercept. Then ΛXi β = Λβ and the logit MLE first-order condition is N-Σi - Λβ=. Thus, ˆ =ln /-. Consistency of ˆ clearly requires a rando saple; oversapling = leads to overestiation of and hence overestiation of ˆ. 4 The BAPCPA increased filing fees and required debtors to coplete additional paperwork. Because it was cheaper and easier to file for bankruptcy prior to 5, households would have been ore inclined to file for bankruptcy. Since the BAPCPA, the nuber of households filing for bankruptcy each year has declined. For exaple, between 994 and, the tie period for the data set utilized by Agarwal, Liu, and Mielnicki 3, on average,,46,85 households filed for bankruptcy each year. By coparison, only 8,59 households filed for bankruptcy in 7. 9