Personal Bankruptcy, Asset Risk, and. Entrepreneurship: Evidence from Tenancy by the. Entirety Laws. Jerey Traczynski. November 20th, 2014.

Transcription

1 Personal Bankruptcy, Asset Risk, and Entrepreneurship: Evidence from Tenancy by the Entirety Laws Jerey Traczynski November 20th, 2014 Abstract Personal bankruptcy law aects entrepreneurship decisions and credit markets for small businesses. I show that personal bankruptcy law impacts rm debt and equity sources, indicating that personal bankruptcy law is immediately salient to small business owners. I show that levels of personal asset protection aect small business decisions by exploiting variation in tenancy by the entirety laws, a form of bankruptcy exemption available only to married people, to create within-state variation in bankruptcy exemptions. I nd that owners value unlimited asset protection more than the mean level provided by homestead exemptions at more than $16,000 per year. I also nd that owners reduce labor supply between 3 and 6 hours per week compared to mean exemptions. However, I do not nd evidence of a statistically signicant impact of tenancy by the entirety laws on spending on risky projects. Preliminary and incomplete. Do not cite. Certain data included herein are derived from the Kauman Firm Survey restricted access data le. Any opinions, ndings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reect the views of the Ewing Marion Kauman Foundation. Department of Economics, University of Hawaii at Manoa, Honolulu, HI 96822; jtraczyn@hawaii.edu.

2 1 Introduction Small businesses are a large part of the U.S. economy, accounting for approximately 43% of total payroll and 50% of employment in the U.S. and generating 65% of net new jobs in the U.S. over the past 17 years (U.S. Small Business Administration, 2011). Small businesses also have very high turnover rates: in 2009, an estimated 552,600 new rms opened and 660,900 rms closed. This turnover led to 60,847 business bankruptcies led in 2009 as rm owners attempted to discharge debts accrued to their businesses. However, the number of bankruptcies due to business closures is likely underreported because rm debts are frequently personal liabilities of the rm's owners for non-corporate rms, so owners often choose to le for personal bankruptcy instead to eliminate both business and personal debts. 1 Lenders may require that the owner guarantee business loans even for small corporate rms, making personal bankruptcy law relevant to a wide range of small businesses. 2 The bankruptcy system may also lead to moral hazard problems in the operation of small businesses. In personal bankruptcy, exemptions allow the debtor to keep some property as part of the debtor's post-bankruptcy fresh start. These exemptions provide wealth insurance to individuals and oer protection against negative personal and business asset shocks for small business owners. The insurance eect of exemptions may inuence an owner's decisions outside of the bankruptcy system, as the level of exemptions aects the amount of risk an owner faces from negative shocks. Under generous bankruptcy exemptions, owners may choose to engage in more risky investments, such as expensive research and development projects with uncertain returns, or spend less time and energy on work knowing that the bankruptcy system will cushion a business failure. To the extent that the bankruptcy system distorts a rm owner's decisions, the ineciencies created may be large. This paper quanties the value of personal bankruptcy exemptions to rm owners and documents the eects on the operation of small businesses, particularly the labor supply decisions of 1 Sullivan et al. (1999) and Lawless and Warren (2005) estimate that approximately 20% of all personal bankruptcy lings involve the discharge of business debts. 2 See Berkowitz and White (2004) for further discussion. 1

3 entrepreneurs. Previous research analyzes the moral hazard problems posed by the bankruptcy system in several ways. One strand focuses on an individual's decision to start a business, showing that high state bankruptcy exemptions encourage self-employment through higher insurance and cause interest rates for entrepreneurs to rise, with the positive insurance eects empirically dominant (Fan and White, 2003; Berkowitz and White, 2004; Jia, 2010). In contrast, this paper studies the decisions made after the creation of the business to see how the personal bankruptcy system aects the amount of eort exerted by the entrepreneurs. Another strand of research analyzes the eects of wage garnishments on individual labor supply after ling for bankruptcy theoretically at either the individual (Wang and White, 2000; White, 2005) or macroeconomic level, with implications for the design of bankruptcy policy. 3 Among empirical papers, Han and Li (2007), Chen (2011), and Dobbie and Song (2013) estimate the impact of bankruptcy on post-ling labor supply. This paper presents empirical estimates of the eects of the bankruptcy system's implicit wealth insurance on small business owners regardless of whether they actually le for bankruptcy. To the best of my knowledge, this paper is the rst to examine the labor supply eects of bankruptcy law on all small businesses. A key challenge in assessing the importance of personal asset protections to entrepreneurs is that the ability to exempt assets in bankruptcy is always available to debtors except in cases of fraud. Similarly, rm owners may become more interested in bankruptcy protections when economic conditions are poor, making it dicult to nd eects of bankruptcy law through the life of the rm. Since business owners do not need to take any specic actions to use exemptions when ling for personal bankruptcy, it can be dicult to determine if owners' awareness of bankruptcy law is the cause of dierent business operation decisions or how valuable bankruptcy protections are to entrepreneurs. This problem explains the focus in the prior literature on the one time decision to start a business rather than decisions that can change over time, such as labor supply or investment decisions. 3 See Athreya (2005) for a survey of macroeconomic equilibrium models of personal bankruptcy. 2

4 To address these issues, I use variation in tenancy by the entirety (TBE) laws across states. TBE laws allow debtors in some states to exempt property owned jointly by a husband and wife from the debts of only one spouse. A married entrepreneur can enjoy the exemption oered by TBE laws only if the spouse has no role in nancing the business. Thus, loans must be made in the owner's name only and business assets must not be jointly owned by husband and wife for TBE laws to apply. Eectively, TBE laws create bankruptcy exemptions that a debtor can choose to contract around through the types of debts the owner acquires at any point in the life of the rm, unlike all other bankruptcy exemptions. A married small business owner has the opportunity to accumulate substantial individual debts, which TBE property cannot be used to repay. Coupled with the high turnover rate of small businesses, owners have both strong incentives and a clear opportunity to use TBE laws to shield assets from creditors. This paper documents evidence of owners changing the nancing of their businesses to take advantage of bankruptcy exemptions through TBE ownership, showing that bankruptcy laws aect the behavior of rm owners throughout the rm's existence. I investigate the eect of personal asset protections on small business decisions using several complementary sources of individual and rm level data. The empirical analysis uses a dierence-in-dierence approach, exploiting cross-state variations in exemption levels and TBE laws. My results show that rm owners are aware of the protections oered by TBE laws and arrange the nancing of their businesses to maximize the protection of personal assets. This eect is mildly stronger in states with stronger TBE laws and weaker in states with large bankruptcy homestead exemptions, indicating that TBE laws and homestead exemptions are substitutes. These results establish that rm owners consider the level of personal asset risk when making business decisions throughout the life of the rm, not only when closing the rm or considering ling for bankruptcy. Firm owners have lower revenues when utilizing TBE laws, indicating that rm owners are willing to surrender over $16,000 per year in prots to obtain these asset protections. 3

5 I nd that when rm owners face less personal asset risk from business failure, they devote less eort to their business by working fewer hours. This moral hazard eect is large: a married rm owner in a TBE state works between 3 and 6 fewer hours per week than a married rm owner in an average exemption state, a decrease in labor supply of 8-14%. However, I nd no statistically signicant eects on the probability of a rm engaging in and spending on research and development projects with uncertain returns. This suggests that labor supply is a primary channel for the moral hazard created by bankruptcy law to aect small businesses. 2 Background 2.1 Bankruptcy When ling for personal bankruptcy, debtors have a choice between Chapter 7 and Chapter 13 bankruptcy. Chapter 7 bankruptcy oers a complete discharge of debts, allowing the debtor to keep only assets that can be held exempt from creditors. Chapter 13 requires debtors repay some debts before receiving a discharge, but allows debtors to keep more of their property. Businesses may le a Chapter 11 reorganization bankruptcy that allows the business to restructure contracts and retain assets while paying o creditors, though high ling costs and long negotiations with creditors make Chapter 11 unattractive to small business owners relative to Chapters 7 and The Bankruptcy Abuse Prevention and Consumer Protection Act of 2005 (BAPCPA) made many changes to the personal bankruptcy system in the United States. Most signicantly, BAPCPA introduced a means test whereby debtors with suciently high income are under a presumption of abuse if they le for Chapter 7 bankruptcy. 5 The means test makes it harder to le Chapter 7, pushing lers towards Chapter 13 and repaying some of their 4 See Levin and Ranney-Marinelli (2005) for a discussion of changes to Chapter 11 as part of BAPCPA that make Chapter 11 more dicult for businesses U.S.C. Ÿ707(b)(2)(B), U.S.C. Ÿ707(b)(2)(A) contains details on the means test. 4

6 debts. However, the means test for Chapter 7 bankruptcy applies only to a debtor whose debts are primarily consumer debts, which has been interpreted by courts to mean that debtors with primarily business debts are not subject to the means test. 6 As a result, the BAPCPA changes in ling Chapter 7 bankruptcy were smaller for business owners than for consumers, and business owners retained most of the protection of personal bankruptcy law that they had before Paik (2013) nds that BAPCPA did not change the relationship between bankruptcy laws and entrepreneurship. Both states and the federal government oer debtors a variety of exemptions to use in bankruptcy. The homestead exemption is designed to shield housing equity and has the greatest dollar value in most states. Personal property, such as cars, tools of trade, furniture, and jewelry are covered by smaller exemptions. Some states oer wildcard exemptions that debtors can use for any type of property up to a certain dollar amount. Wildcard exemptions are sometimes available to non-homeowners to use in place of the homestead exemption, though not of the same dollar value. Previous work investigates the link between personal bankruptcy law and small business. Fan and White (2003) show that individuals are more likely to choose self-employment in states with high personal bankruptcy exemptions, while Berkowitz and White (2004) nd that small businesses in states with high personal bankruptcy exemptions are more likely to be denied credit, receive smaller loans, and pay higher interest rates. These papers focus on rm entry and access to credit, while this study examines the eect of the insurance oered by exemptions on the business decisions of rm owners. On bankruptcy and labor supply, Wang and White (2000) calculate optimal rates for wage garnishment of a debtor's post-bankruptcy labor income through simulations. White (2005) develops a theoretical model that considers the relationship between labor supply and bankruptcy ling decisions in the context of optimal bankruptcy policy, including both 6 11 U.S.C. Ÿ707(b)(1), For interpretation of the clause, see Wedo (2005) and In re Kinnee, Case No (Bankr. E.D. Wis, 2006) (unpublished decision available at gov/opinions/files/pdfs/in_re_kinnee,_ pdf). In re Kinnee asserts that debt is primarily consumer if more than 50% of the amount is consumer debt. 5

7 exemption levels and wage garnishments. Han and Li (2007) nd no eect of bankruptcy ling on post-bankruptcy labor supply using PSID data, while Chen (2011) nds a positive eect using NLSY79 data. Dobbie and Song (2013) exploit random assignment of rst time Chapter 13 lers to bankruptcy judges and nd that ling increases earnings. In contrast, this paper focuses on the eect that exemption levels have on labor supply regardless of ling status by providing wealth insurance. 2.2 Tenancy by the Entirety Tenancy by the entirety is a form of joint ownership in which a husband and wife both own the undivided whole of a piece of property. This concept of ownership reects the idea that a husband and wife are a single entity. An individual spouse cannot unilaterally give away, partition, or sell his or her interest in TBE property, as the property is owned by the union of husband and wife rather than either of the individuals. When an individual debtor les for bankruptcy, the bankruptcy estate must include all legal or equitable interests of the debtor in property. 7 Property owned by an entirety is not property of an individual debtor and is thus exempt from creditors with no dollar limit on the value of the TBE property. A married debtor is therefore able to exempt property held as TBE from creditors with claims only against the debtor. If a debtor has joint debts with a spouse such as a home equity loan on property that both partners own, or if a married couple les for bankruptcy jointly, then property held as TBE is part of the bankruptcy estate and may be sold to pay creditors. 8 The protections of TBE are maximized when a married individual can accumulate debt in the individual's name only. An important feature of TBE laws is that married rm owners can choose whether or not to use the protection of TBE laws through the debt structure of the rm and may choose to forgo the protection of TBE laws by incurring a joint debt or allowing both spouses to 7 11 U.S.C. Ÿ541(a)(1), After United States vs. Craft, 535 U.S. 274 (2002), TBE property can be subject to federal tax liens against an individual spouse. Since all data used in this paper are from after 2002, this decision does not directly aect my empirical ndings. 6

8 have an equity interest in the rm at any time. This is dierent from all other bankruptcy exemptions, which a debtor cannot agree to give up. Since avoiding joint debts precludes the use of common funding sources for small businesses, most notably home equity for a married couple that jointly owns their home, qualifying for TBE protections requires careful planning. I therefore test whether bankruptcy exemptions inuence small business operation decisions by evaluating if married entrepreneurs structure the debts and ownership of the business so as to shelter assets under TBE laws. Tenancy by the entirety as a form of ownership rst appeared in England in the 1200s (Carrozzo, 2001). Under the doctrine of coverture in English common law, TBE gave the husband complete control of all property owned by a married couple, as the wife's right to own property was minimal. Phipps (1951, p. 24) describes this early form of TBE existing into the 19th century as man and wife were one and the one was male... marriage amounted to an absolute gift of all the wife's personal property to the husband. In the United States, Married Women's Property Acts passed by states in the mid-1800's allowed wives to own and control property separately from their husbands, creating a conict with traditional TBE. State court interpretations of the relationship between the Property Acts and TBE created variation in the strength of TBE laws across states. While some states eliminated the protections of TBE and some kept it largely intact, other states opted for a middle ground where some TBE property could be held to satisfy debts, subject to various rights of the non-debtor spouse. Franke (2009) provides a description of TBE laws across the 25 states and District of Columbia that recognize TBE in some form. Franke (2009) also categorizes states as either full or modied TBE, where full TBE states do not allow creditors of an individual spouse to make any claims against TBE property. 9 Table 1 summarizes TBE laws across states in Though TBE is regarded in both the economics and law literatures as a powerful form of asset protection due to its unlimited dollar value, this paper is the 9 Older categorizations of TBE laws may be found in Phipps (1951) and Sawada v. Endo, 561 P.2d 1291, (Haw. 1977). 7

9 rst empirical evaluation of the eects of TBE laws on businesses and the rst to exploit variation in TBE across states Empirical Model and Data To estimate the eect of the exemptions on small business operation decisions, I use a dierence-in-dierence model to compare decisions made by married business owners and single business owners in states with and without TBE laws. As a baseline, I estimate the statistical model Y is = α + β 1 married i + β 2 T BE s + β 3 married i T BE s + π X is + ɛ is (1) where Y is is the outcome of interest, T BE s is a dummy variable indicating whether state s recognizes TBE ownership in any form, married i indicates if the owner of rm i is married, and X is are other control variables. When Y is is a binary variable, the regression model is a logistic specication. X is consists of the rm owner's years of work experience, age and age squared, as well as dummy variables for the owner's education level, race, ethnicity, and gender. I also include dummy variables for the legal status of the rm and the 2-digit NAICS code for the rm's industry. For rms with multiple owners, I dene the primary owner as the owner who holds the largest percentage of the rm. If two or more owners hold the same percentage, this tie is broken in favor of the owner with a greater number of hours worked, level of education, age, and years of work experience in order, following Robb and Robinson (2013). In equation (1), β 3 is the estimate of how the dierence between married and single rm owners in the outcome variable diers across states with and without TBE laws. Married individuals in TBE states receive the treatment of an unlimited bankruptcy exemption for 10 See Kalevitch (1986), Concannon (1990), Dickerson (1998), Carrozzo (2001), Hynes et al. (2004), Hynes (2004), and White (2007). Hynes et al. (2004) use TBE laws as an outcome variable in studying determinants of property exemptions across states. 8

10 TBE property against the debts of only one spouse, while single people do not. Unmarried rm owners are exposed to many of the same regulations and economics conditions in each state as married rm owners, making them a plausible control group that is unaected by TBE laws. Identication of the eect of TBE laws relies on the assumption that there are no other dierences between TBE and non-tbe states that aect the relative outcomes of married and single rm owners. I examine some supporting evidence for this assumption below. The main dataset for analysis is the Kauman Firm Survey (KFS), a longitudinal rm level survey of companies founded in The KFS collects data from rm owners on characteristics ranging from basic demographics and hours worked to types of debts and equity investments yearly. I use the KFS data over the period , so the sample consists of rms that have remained in business for at least 5 years. The dierence-in-dierence analysis uses the 2009 data for demographics and rm characteristics, while previous years reveal if a rm ever used a particular source of either debt or equity funding. The KFS was created from a random sample of Dun & Bradstreet's 2004 listing of new businesses and oversamples rms from industries with a high industry-wide level of employees performing research and development. All results presented from the KFS data use the provided sample weights due to this sampling structure. I use the restricted access version of the data to obtain information on the state in which each rm is located. I report summary statistics in Table 2. Nearly 47% of rms in this sample are in states with some form of TBE law, making the asset protections of TBE laws relevant to a large fraction of rm owners across the U.S. To obtain a broader sample of rms, I supplement this analysis with the 2007 Survey of Business Owners Public Use Microdata (SBO) from the U.S. Census Bureau. 12 The data include all nonfarm businesses in the U.S. ling IRS tax forms with receipts of $1,000 or 11 More information about the KFS can be found at aspx. 12 More information about the SBO can be found at 9

11 more in a tax year and are weighted to be nationally representative. All results presented use the provided sample weights. The SBO provides the state in which each rm is located, though some states are grouped together for disclosure purposes. I keep data from these grouped states only if all states in a group have the same TBE laws. Summary statistics are reported in Table 2. The rm owners in the SBO sample have lower levels of education than those in the KFS sample, but the samples appear otherwise similar. I use the state identiers in both the KFS and SBO to match rms' locations to data on TBE laws from Franke (2009) summarized in Table 1. Since TBE laws oer unlimited exemptions, I also use data on state bankruptcy exemptions to determine the eect of TBE laws beyond the regularly available state exemption levels. Intuitively, TBE laws do not oer much additional protection in bankruptcy if the state's bankruptcy exemptions are already high or unlimited, while they may have a much greater eect on decision making of rm owners in states with very low exemption levels. Since TBE laws apply to property jointly owned by a husband and wife, the marital home is likely the most valuable jointly owned asset and thus the homestead exemption is the most relevant comparison for the extent to which TBE laws provide additional wealth insurance. Table 1 lists the available homestead exemptions for married couples in all states in For states with a dened homestead exemption, the correlation between the homestead exemption level and whether or not the state recognizes any form of TBE ownership is , indicating that states with TBE laws tend to have lower homestead exemptions. This correlation suggests that TBE laws may serve as a substitute for generous bankruptcy exemptions. This negative relationship between homestead exemption size and TBE laws also holds true when states with unlimited homestead exemptions are added to the sample. I assign a value of $550,000 to the homestead exemption in states with unlimited exemptions, matching the largest dened homestead exemption. 13 After including the unlimited 13 This is consistent with Berkowitz and Hynes (1999) and Traczynski (2011), both of which use a value of $500,000 for states with unlimited homestead exemptions. Both papers, however, use older samples where no state had a dened homestead exemption larger than $500,000. I therefore increase the value assigned to unlimited exemption states so that unlimited exemptions remain the largest exemptions available. The 10

12 exemption states, the correlation between the homestead exemption and recognition of TBE laws is Interestingly, this negative relationship appears to be driven by states with full TBE laws rather than modied TBE laws. The correlation between the homestead exemption size and having a modied TBE law is among states with dened exemptions and when including all states, while the correlation between homestead exemption size and having full TBE protections is across dened exemption states and across all states. The dierential negative relationship between homestead exemption levels and types of TBE laws motivates the heterogeneity analysis below. Table 3 compares the observable characteristics of rm owners in states with and without TBE laws. The only statistically signicant dierence in means at conventional levels across these characteristics of rms and owners is that non-tbe states appear to have a slightly higher percentage of Hispanic rm owners. This suggests that rm owners in states with and without TBE laws are similar, so there is no general sorting across states correlated with TBE status. Since TBE laws aect only married people, Table 4 looks specically at whether married entrepreneurs in TBE states or their rms dier in their observables. Each regression in Table 4 uses the observables variable at top as the dependent variable, with only the Married, TBE, and married T BE dummies as explanatory variables. 14 The results show that married individuals in TBE states are more likely to be Asian than married rm owners in non-tbe states, a result statistically signicant at the 10% level. However, this is the only observable dierence between married rms in TBE and non-tbe states, and nding only one of these 16 regressions to have an interaction term statistically signicant at the 10% level is consistent with expected rates of Type I error. Overall, it appears that there is little evidence of dierences in observables across TBE and non-tbe states for all rm owners and for married rm owners, supporting the claim that TBE laws are not correlated results presented below are not sensitive to the exemption value chosen for unlimited homestead exemption states. 14 Other Legal Form is excluded as a rm characteristic because of the small number of rms in the data with this form, none of which are owned by a married entrepreneur in a TBE state. Columns (1) and (2) of Table 4 report coecients from weighted least squares regressions, while all other columns report marginal eects from logistic regressions. 11

13 with other observable dierences that may confound estimates. 4 Results 4.1 Do Owners Take Advantage of TBE Protections? As discussed above, rm owners may only exempt property owned as TBE from the claims of creditors of one spouse. If a creditor has a claim against both spouses, then the TBE property may be used to satisfy the debt. Married rm owners who value the protection oered by TBE laws should therefore not use home equity loans to nance the business, as both spouses approving the mortgage would make the house vulnerable to seizure in bankruptcy. There should also be less joint ownership of rms by married partners in TBE states, as business debts for which both spouses are liable similarly expose any TBE property to collection. I test the eect of TBE laws using data on personal debts of rm owners in the KFS data using equation (1). I dene a binary variable equal to 1 if a rm owner has ever used a particular source of credit for business purposes and present results in Table 5. Summary statistics for these variables are in Table 2. All standard errors are clustered at the state level. Columns (1)-(4) report logistic regression results for whether or not a rm owner has ever used personal loans from a bank including home equity loans or mortgages, business credit cards issued in the owner's name, personal credit cards, or personal loans from family or friends, respectively. The results show that married rm owners in TBE states are less likely to use mortgages or home equity loans and more likely to use business credit cards issued in an owner's name than married rm owners in non-tbe states. Business credit cards in the owner's name separate business and personal debts, keeping debt in the name of only one individual. By contrast, married rm owners' personal credit cards may be held jointly with a spouse, creating joint debts that expose TBE property, or may be held in only the rm owner's name. The data do not reveal if a personal credit card is joint, so the insignicant 12

14 result here is not surprising. Personal loans from friends and family are less likely than the other three sources of credit to go through a formal market or use a contract and therefore serve as a falsication test. Since providing collateral and the ability to seize assets in bankruptcy is likely less of a factor in obtaining loans from family and friends than in formal credit markets, TBE laws should not aect the use of this source of credit. Column (4) shows that TBE laws do not have a statistically signicant impact on the prevalence of personal loans from friends and family. Also, no regression reveals a statistically signicant dierence in the level of the dependent variable for single individuals across TBE and non-tbe states, another falsication exercise as TBE laws should not impact single rm owners. The eects on home equity loans, mortgages, and business credit card use are also economically signicant. The dierence between married and single rm owners using home equity loans or mortgage debt to nance a business is 10.9 percentage points smaller in TBE states. Since 32.7% of rms report ever using these loans, this eect represents a decrease in the use of home equity loans and mortgages as a means of business nance of approximately one-third. For business credit cards, the 7.26 percentage point increase in usage corresponds to a 10.7% increase in the use of this form of nancing. To determine if TBE laws aect sources of equity investments in the business, I dene a binary variable equal to 1 if a rm has ever received an equity investment from that source. I again estimate the eect of TBE laws using equation (1) and present results in columns (5)-(7) of Table 5. I focus on two types of potential equity holders, government agencies and spouses who are not also owners. Government agencies refers to Small Business Investment Companies (SBICs), privately owned companies backed by the U.S. Small Business Administration that can make equity or debt investments in small businesses. 15 SBICs receive guarantees on loans up to a certain dollar amount from the Small Business Administration, so these groups should show little sensitivity to risk. The results in column (5) of Table 5 conrm this intuition, as married rm owners in TBE states show no statistically signicant 15 See for more information on the program. 13

15 dierence in the likelihood of equity investment from government agencies. I also nd no dierence in the likelihood of equity investments for single rm owners across TBE and non- TBE states. When limiting the sample to married rm owners in column (6), I nd that the rm owner's spouse is less likely to own an equity stake the rm in TBE states. Since the KFS does not indicate the relationships between rm owners, I turn to the SBO data to see if there are fewer instances of spouses jointly owning a business in TBE states. Column (7) of Table 5 shows that the percentage of businesses jointly owned by married couples is lower in TBE states. The results in columns (6) and (7) provide further evidence of rm owners attempting to maximize the protections of TBE laws by excluding a spouse from owning any share of the business. 4.2 Heterogeneity in TBE Laws I investigate whether rm owners are more likely to change their loan types or receive dierent equity investors when TBE laws oer stronger protections. I replace the dummy for TBE laws in equation (1) with separate dummies for states with modied and full TBE laws, following the categorization of variation in TBE laws given by Franke (2009). In states with full TBE laws, the protections for TBE property against creditors of an individual spouse are very strong, while states with modied TBE laws may allow creditors to attach liens to TBE property or make other claims against it, subject to a variety of conditions. 16 In Table 6, I estimate the eects of these dierent types of TBE laws using the same dependent variables as in Table 5. In columns (1)-(4) of Table 6, the dierence between the coecients on the interaction terms F ull T BE Married and Modified T BE Married is never statistically signicant, suggesting that the modications made to TBE laws have not led rm owners to take out dierent types of loans. Despite the dierences in TBE laws across states, rm owners still try to take advantage of TBE protections by using business credit cards and not using home equity loans or mortgages to nance their business spending 16 Franke (2009) contains a thorough discussion of the dierent types of modications made across states. 14

16 in all TBE states. Among types of equity investors, only column (6) shows a dierential eect of types of TBE laws, with full TBE states showing larger negative eect on the probability of a rm having an equity investment from an owner's spouse. As a whole, these results show that the type of TBE law does not have a large eect on loan types or the identity of rm equity investors, though what dierence exists suggests that there is a larger behavioral response when TBE laws oer stronger protection from creditors. 4.3 TBE Laws and Bankruptcy Exemptions To the extent that TBE laws function as large bankruptcy exemptions for married people, married rm owners in TBE states that also have high homestead exemptions may gain minimal additional asset protection from TBE laws. I interact a state's homestead exemption for married couples (in $10,000s) with TBE laws, yielding Y is = α + β 1 married i + β 2 T BE s + β 3 exempt s + β 4 T BE s exempt s +β 5 T BE s married i + β 6 married i exempt s (2) +β 7 T BE s married exempt s + π X is + ɛ is where exempt s is the state's homestead exemption. I focus on the homestead exemption because housing equity is likely to be a married couple's largest jointly owned asset and is therefore potentially aected by both TBE laws and a state's homestead exemption. I show results in Table 7 for the full sample as well as a sample of states with a dened homestead exemption to explore the sensitivity of the results to the $550,000 exemption amount assigned to states with an unlimited homestead exemption. Specications for the sample of all states also include a dummy variable for whether a state has an unlimited homestead exemption to capture any eects of an unlimited homestead exemption beyond the dollar amount assigned For states that allow the use of federal exemptions, I replace the state homestead exemption with the federal homestead exemption amount if the federal exemption is greater. 15

17 The results in Table 7 conrm the intuition that TBE laws are substitutes for the wealth insurance oered by homestead exemptions. For loans, the dierence-in-dierence estimates of Table 5 show that TBE laws reduce the probability of using a mortgage or home equity loan to nance a small business and increase the likelihood of using a business credit card. Columns (1) and (2) of Table 7 have a positive coecient on the triple interaction term, indicating that married rm owners in TBE states are more likely to use a mortgage or home equity loan when exemption levels are high. This result shows that when exemptions are high, rm owners are less likely to take steps to preserve the protections of TBE laws. Columns (3) and (4) repeat this analysis for the probability of using business credit cards and show that the triple interaction term has the expected negative sign, though the term is statistically insignicant. In non-tbe states with larger homestead exemptions, married rm owners are more likely to use business credit cards. However, this is not true in TBE states, where there is no signicant dierence in business credit card usage for married rm owners when the homestead exemption is larger. Finally, columns (5) and (6) show the same pattern for spousal equity investment. Though TBE laws lower the probability of a spouse having an equity investment in a rm, the probability rises in TBE states with higher homestead exemption levels as the asset protection of the high homestead exemption replaces that of the TBE laws. Overall, these results show that TBE laws and bankruptcy homestead exemptions function as substitutes, with rm owners less likely to structure a rm's debts and ownership to take advantage of TBE asset protections when a state already provides generous bankruptcy exemptions. 4.4 Owner Heterogeneity Taking full advantage of the protections of TBE laws requires the owner to be aware of TBE laws and plan out the debt structure of the rm accordingly. More experienced or sophisticated owners may be more likely to use TBE laws to shelter assets. To investigate this, I interact various owner characteristics with TBE laws in equation (1) and present 16

18 results in Table 8. I use the owner's years of experience as a rm owner, whether the owner owns another rm in the same industry, whether the owner owns any other rm, and whether the owner has completed a 4 year college or graduate degree as proxies for the owner's likelihood of knowing about TBE laws and how to use them. I focus on the rm owner's decision to take out a bank loan or mortgage due to the consistent responsiveness of this variable to TBE laws in previous results and because the need to avoid joint debts with a spouse is a central feature of TBE laws. Table 8 shows that in all regressions, the triple interaction term is negative, indicating that a married owner in a TBE state with more work experience, ownership of other rms, or higher education is less likely to use a personal bank loan as part of business nancing. However, this dierential eect is only statistically signicant for years of work experience. Column (1) shows that each additional year of experience for the rm owner reduces the probability of ever taking out a personal bank loan by 0.76 percentage points. As 32.7% of rm owners report using a personal bank loan, this result implies that one additional year of work experience decreases the use of personal bank loans by approximately 2%. The results in Table 8 oer some evidence that owners with more experience are more likely to take advantage of TBE protections, a sensible result given the requirements of TBE laws. 4.5 Eects of Asset Protections on Business Operations The above results establish that rm owners make nancial decisions for the rm to take advantage of the personal asset protections oered by TBE laws. I now turn to how these asset protections aect rm outcomes and other business operation decisions such as labor supply and spending on risky projects. Firm prots may be aected through several channels. If rms are credit constrained because owners gained asset protection at the cost of losing access to housing equity as capital for business use, then prots may fall because the business is smaller. The magnitude of a fall in prots provides a measure of how much rm owners are willing to give up in exchange 17

19 for asset protections. I estimate eects using equation (2) to control for the potentially important eect of the homestead exemption level on credit markets. Columns (1) and (2) of Table 9 show that rm revenues are lower for married owners in TBE states, but there is no statistically signicant eect on rm expenditures. 18 Firms in states with larger homestead exemptions see less of a negative eect of TBE laws on revenues and expenditures, further supporting the substitutability of TBE laws for exemption levels as shown in Table 7. These results are economically large: in a state with the mean homestead exemption level, the estimates from column (1) imply that revenues for married rm owner in a TBE state are 20.1% lower while column (2) shows that expenditures are 8.9% lower. To translate these gures into an estimate of the eect on prots, I rst note that the denition of total expenditures in the KFS data includes amounts spent on wages, salaries, interest on loans, capital leases, and materials. This may not include all expenditures of the rm, partially explaining the dierence between the reported values of total revenues, total expenditures, and rm prots. I therefore calculate the reduction of the gap between revenues and expenditures associated with TBE laws, and apply this percentage to the mean level of prots. Using the average values of revenues and expenditures, a 20.1% reduction in revenues and a 8.9% reduction in expenditures means that the dierence between revenues and expenditures falls from approximately $332,000 to $209,000, reducing the gap to 63% of its previous level. As the mean value of prots is $14,505, such a reduction implies that prots would fall by $5383. The results in Table 5 imply that roughly one-third of rm owners take advantage of TBE laws by not taking out a loan against home equity. Scaling up this estimate appropriately, I nd that rm owners are willing to give up approximately $16,150 in yearly prots in exchange for the additional asset protections of TBE laws above what their states already protect through homestead exemptions. Columns (3)-(6) explore some of the possible mechanisms behind the changes in revenues and expenditures. Column (3) shows that there is no statistically signicant change in the 18 A stacked regression shows that the dierence between the T BE Married coecients in columns (1) and (2) is statistically signicant at the 10% level. 18

20 number of employees hired when asset protections are greater, suggesting that rms do not change in size by this measure. Instead, changes in behavior other than hiring must lead to the observed declines in revenue. Columns (4) and (5) show that conditional on obtaining personal or business loans, TBE laws do not have a statistically signicant eect on the total size of loans, and column (6) conrms this result for home equity loans and mortgages. This result indicates that any eects of limited credit for married rm owners in TBE states must arise from not obtaining loans rather than the loan amounts. This is consistent with using TBE laws as asset protection, as only the existence of a loan and not the loan amount is relevant for making an asset vulnerable to seizure by creditors. To further explore the mechanisms behind these observed changes in rm protability, I examine whether changes in revenue are related to changes in owner eort, and whether changes in costs can be explained by decisions to take on riskier projects. The question of how wealth insurance aects labor supply is an open one in the literature. 19 If greater work eort decreases the probability of business failure and loss to the owner, then wealth insurance and work hours are substitutes and may be negatively related. Similarly, greater insurance may limit the downside risk of failed projects, encouraging rm owners to have more projects with uncertain returns. I again use equation (1) to determine if TBE laws aect the hours worked by a rm's owner and present results in Table 10. In column (1), the dierence-in-dierence approach shows a statistically insignicant negative eect of the asset protections of TBE laws on owner hours worked. This estimate implies that at the mean homestead exemption level, the availability of TBE protections causes a 14.7% reduction on the intensive margin of hours worked. In column (2), I control for the possible confounding eects of state homestead exemptions by adding interactions as in equation (2). I nd that in non-tbe states, married rm owners work 22.3% more hours than their single counterparts, a dierence of approximately 9 hours per week. However, 19 See Krueger and Meyer (2002) for an overview of the eects of social insurance programs (particularly income maintenance programs) on labor supply and White (2005) for a theoretical model of labor supply response to post-bankruptcy wage garnishments. Athreya and Simpson (2006) study the interaction between bankruptcy systems and unemployment insurance. 19