WORKING PAPER NO. 13-38/R DEBT COLLECTION AGENCIES AND THE SUPPLY OF CONSUMER CREDIT

Viktar Fedaseyeu Bocconi University and Visiting Scholar, Federal Reserve Bank of Philadelphia

August 2014

Debt Collection Agencies and the Supply of Consumer Credit

Viktar Fedaseyeu∗ August 25, 2014

Abstract I examine contract enforcement in consumer credit markets by studying third-party debt collection. I construct a state-level index of the tightness of debt collection laws and find that stricter regulations of third-party debt collection lead to fewer openings of revolving lines of credit. This effect appears to be the result of lower recovery rates due to fewer debt collectors per capita when debt collection laws are tightened. Less stringent debt collection laws are associated with a larger and riskier pool of borrowers, consistent with the view that effective debt collection enables creditors to expand lending to new and risky borrowers.

Keywords: household finance, consumer credit, lender protection, creditor rights, contract enforcement, debt collection, law and finance

∗ Assistant Professor of Finance, Bocconi University. Email: [email protected]. I am deeply grateful to Phil Strahan for his unwavering support and encouragement and for the extensive feedback he has provided. I thank members of my dissertation committee, Tom Chemmanur, Darren Kisgen, Alan Marcus, Jonathan Reuter, Ronnie Sadka, and Hassan Tehranian, for their guidance. I benefited from helpful comments by Pierluigi Balduzzi, David Chapman, Ethan Cohen-Cole, Cliff Holderness, Edith Hotchkiss, Bob Hunt, Rich Hynes, Miles Kimball, Jeff Pontiff, Jun (QJ) Qian, Dubravka Ritter, Antoinette Schoar, Peter Tufano, Stephanie Wilshusen, seminar participants at Bocconi University, Boston College, Norwegian School of Economics and Business Administration (NHH), and the Federal Reserve Bank of Philadelphia, as well as conference participants at the Household Finance Workshop of the 2010 NBER Summer Institute, the 2010 Financial Management Association meetings, and the 2011 Western Finance Association meetings. Aliaksandra Shelestava provided assistance with legal issues. Access to TransUnion’s Trend Data solution for this project was provided through the Payment Cards Center at the Federal Reserve Bank of Philadelphia. The views expressed in this paper are not necessarily those of the Federal Reserve Bank of Philadelphia or the Federal Reserve System. All errors are my sole responsibility. This paper is available free of charge at www.philadelphiafed.org/research-and-data/publications/working-papers/.

1.

Introduction

Third-party debt collectors play an active role in retail credit markets: The proportion of American consumers with at least one account in third-party collections has not fallen below 10% since 2004 (see Figure 1) and stood at 14% at the end of 2013. Despite the visible presence of third-party debt collectors in the lives of borrowers, however, little is known about their impact on credit provision. The goal of this paper is to evaluate this impact. [INSERT FIGURE 1 ABOUT HERE] Debt collectors try to obtain repayment from borrowers who defaulted on their debts and thus ensure that defaulted debts will not go away easily. In effect, debt collectors enforce contracts in consumer credit markets. The importance of contract enforcement is well understood. However, most research on contract enforcement has focused on nonconsumer debt, despite the fact that consumer credit markets are large (with $2.779 trillion of consumer nonmortgage debt outstanding at the end of 2012)1 and have characteristics that make contract enforcement in these markets different from corporate credit markets. First, because of consumer protection laws, creditors in retail credit markets can never obtain control rights over debtors and do not have full access to debtors’ assets, especially their most valuable asset — human capital. In addition, enforcing contracts over a large number of individual accounts with relatively small balances requires a technology different from that used to enforce contracts over fewer and relatively larger corporate borrowers. Debt collection is not the only relevant factor that affects the ability of creditors to enforce contracts in consumer credit markets. This ability is also affected by consumer bankruptcy 1 Source: http://www.federalreserve.gov/Releases/z1, Z.1 release dated March 7, 2013, Table D.3. This figure includes about $1 trillion in student loans. While third-party debt collectors are arguably less important for the provision of student loans (because of the special status of student debt in bankruptcy and the political process involved in student loan provision), they are not irrelevant: The U.S. Department of Education does not collect on defaulted federal student loans itself; instead, it hires third-party debt collection agencies.

1

law and restrictions on garnishment (the right of creditors to deduct money directly from borrowers’ monetary compensation). Prior literature has shown that garnishment affects consumers’ propensity to file for bankruptcy (e.g., Dawsey and Ausubel (2004) and Dawsey, Hynes, and Ausubel (2013)) and that bankruptcy affects credit supply and demand (e.g., Gropp, Scholz, and White (1997), Berkowitz and Hynes (1999), and Lin and White (2001)). While bankruptcy undoubtedly plays a role in credit provision, Dawsey, Hynes, and Ausubel (2013) note that the majority of defaulting consumers never file for bankruptcy and that most debt collection takes place outside of the courtroom, where restrictions on debt collection practices appear to affect the ability of creditors to obtain repayment. It is therefore likely that such restrictions have an impact on credit provision. To identify the effect of debt collectors on credit provision, I use variation in state laws and construct an index of debt collection restrictions. Stricter debt collection regulations, which make it more difficult for debt collectors to operate, should result in less effective contract enforcement and should therefore lower credit supply. Consistent with this hypothesis, I find that stricter regulations of third-party debt collectors reduce the number of new revolving lines of credit: Increasing the value of the index of debt collection restrictions from its first quartile to its maximum value in the sample increases the number of new revolving account openings by 11%, which is comparable with the effect of bankruptcy exemptions on mortgage denial rates documented in Berkowitz and Hynes (1999) and Lin and White (2001). The impact of state debt collection restrictions on consumer credit is also comparable with the impact of other contract enforcement mechanisms on credit provision (e.g., Jappelli, Pagano, and Bianco (2005)) and is therefore consistent with the general message of a broader literature on law and finance, which finds that investor rights strongly affect the development of financial markets (e.g., La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998)).

2

I also investigate the economic mechanism behind the impact of debt collectors on credit provision and find that stricter debt collection laws decrease recovery rates on charged-off unsecured credit card loans and are associated with fewer debt collectors per capita. Thus, stricter debt collection laws reduce the number of debt collectors, who can therefore exert less pressure on debtors. This reduces recovery rates and makes lenders less willing to provide credit in the first place. Additionally, I find evidence that less stringent debt collection laws are associated with a larger and riskier pool of borrowers, which is consistent with the view that effective debt collection enables lenders to provide credit to new, riskier applicants who would otherwise be rationed out of the credit market. To the extent that access to credit helps borrowers smooth their consumption, effective debt collection increases their welfare. At the same time, less stringent debt collection laws may increase the aggressiveness of debt collectors and therefore impose a utility cost on borrowers. Thus, more research is needed to ascertain the impact of debt collection on consumer welfare. As with any study of credit provision, separating demand effects from supply effects is a challenge. Since I observe the net, general equilibrium effect of debt collection laws on the amount of credit, my results represent the net impact of both the supply and the demand responses. The fact that the direction of my estimates is consistent with a supply-side response may indicate that borrowers do not adjust their demand for credit in response to changes in debt collection laws. Alternatively, my results could be driven by demand-side variation if stricter debt collection laws reduce demand for consumer credit. However, this seems implausible. On the contrary, stricter debt collection regulations should, if anything, increase demand, all else being equal, because they lower consumers’ indirect costs of obtaining credit by limiting the ability of debt collectors to pursue borrowers. Further, in my tests, I control for consumer credit scores and include the number of loan applications that consumers have 3

made (by counting the average number of credit inquiries) to proxy for credit demand. Another concern with my analysis is that changes in debt collection laws may be driven by general economic conditions that are correlated with the credit cycle. Controlling for statelevel economic conditions should mitigate this concern but cannot eliminate it completely. To address this alternative explanation more directly, I use several placebo outcomes as well as the amount of secured credit as additional dependent variables. Since debt collectors are usually employed to collect unsecured debts, they should be less relevant for the provision of secured credit. I find that debt collection laws have little effect on the amount of secured credit and on distantly related placebo variables (state-level unemployment, high school graduation, and infant mortality rates). It is therefore unlikely that my results are driven by some time-varying unobservable factors that affect the business cycle. The rest of this paper is organized as follows. Section 2 reviews related literature. Section 3 develops empirical hypotheses. Section 4 provides details about the regulation of debt collection, describes related laws, and develops the index of debt collection restrictions. Section 5 describes the data, estimation strategy, and empirical results as well as some robustness tests. Section 6 concludes. Appendix A describes the institutional details of the debt collection process. Appendix B describes the procedure used to identify changes in debt collection regulations and related laws, while Appendix C contains a brief summary of changes in debt collection statutes.

2.

Relation to existing literature

This paper studies debt collection in consumer credit markets and therefore complements the large body of corporate finance literature on investor and creditor rights that followed La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998). Extant work on debt collection in consumer credit markets mostly focuses on institutional details and on consumers’ propensity 4

to file for bankruptcy. Hunt (2007) gives an overview of the debt collection industry and provides details about its institutional structure and regulatory environment. Hynes (2008) examines the process of debt collection in state courts and finds that consumers who are sued by creditors or debt collectors are drawn from low-income areas. He also finds that these consumers are not likely to file for bankruptcy. Dawsey, Hynes, and Ausubel (2013) show that states with antiharassment statutes that apply to creditors collecting their own debts (as well as to third-party debt collectors) have lower bankruptcy filing rates, but borrowers living in these states are more likely to default without filing for bankruptcy. My focus on the regulation of debt collection is similar to Dawsey, Hynes, and Ausubel (2013). However, they consider only one aspect of debt collection laws (whether state antiharassment statutes apply to original creditors) and study its effect on consumers’ choice between formal and informal bankruptcy, while I construct an index of the overall restrictiveness of debt collection laws and study its effect on credit provision. A new strand of theoretical research is emerging that starts to exploit unique features of contract enforcement in consumer credit markets to model unsecured consumer credit. Fedaseyeu and Hunt (2013) propose a model of the debt collection industry to explain existing empirical facts and to study welfare implications of outsourcing debt collection to thirdparty agencies. Drozd and Serrano-Padial (2013) show that enforcement of consumer credit contracts via debt collection can help explain the rapid expansion of credit card borrowing in the U.S. in the 1980s and 1990s. Athreya, Sanchez, Tam, and Young (2013) introduce a model of unsecured consumer credit in the presence of both formal bankruptcy and informal default. This paper belongs to the growing body of literature on household finance. Campbell (2006) delineates the field. He finds that many households make effective investment decisions, while a less educated minority make significant mistakes. Tufano (2009) gives a 5

recent overview of household finance research and proposes the functional definition of this field. An active area of research in household finance focuses on consumers’ access to credit and, in particular, on the demand for short-term high-interest loans such as payday loans. Melzer (2009) finds that access to payday loans does not seem to alleviate financial hardship, while Morse (2011) provides evidence that payday lending mitigates individual financial distress. The current paper complements this literature by studying a mechanism that enables traditional financial services providers to extend credit to risky borrowers. This paper also complements a large body of literature on personal bankruptcy, which focuses on explaining the rising rates of personal bankruptcy filings over the last two decades, as well as on the effect of bankruptcy law on credit availability. Gropp, Scholz, and White (1997) find that generous state-level personal bankruptcy exemptions increase the amount of credit held by high-asset households and reduce the availability of credit for low-asset households. Domowitz and Sartain (1999) and Fay, Hurst, and White (2002) find support for the strategic model of bankruptcy, which predicts that households are likely to file when their financial benefit from doing so is high. Gross and Souleles (2002) document that propensity to file for bankruptcy significantly increased from 1995 to 1997, even after controlling for a variety of personal risk characteristics, and they interpret this result as an increase in the borrowers’ willingness to default. Dick and Lehnert (2010) show that the expansion of credit supply over time is responsible for rising personal bankruptcy rates, an explanation that was suggested by White (2007). Scott and Smith (1986) document that the Bankruptcy Reform Act of 1978, which made personal bankruptcy more pro-debtor, led to an increase in the contract interest rates on small business loans. Debt collectors, the focus of this paper, provide a creditor protection mechanism, which complements bankruptcy as a consumer protection mechanism (at least in the U.S.).

6

3.

Hypotheses

The willingness of creditors to lend depends on the likelihood that the debt will be repaid. If a borrower defaults, this likelihood is determined by creditors’ ability to recover the debts owed to them through a process called debt collection. There are two types of debt collection: in-house debt collection (in which creditors try to collect the debt on their own) and thirdparty debt collection (in which creditors outsource debt collection to third-party agencies). Initially, creditors start collecting on their delinquent accounts internally and, if unsuccessful, later transfer these accounts to third-party agencies. The focus of this paper is on the impact of such agencies on credit supply. Since third-party debt collection represents most of the debt collection activity in the U.S.,2 this is a natural starting point in the analysis of debt collection in general. It is also likely that the factors that affect the effectiveness of thirdparty debt collection will have a similar effect on credit provision as the factors that affect in-house debt collection. The effectiveness of debt collection will have an impact on both the supply of credit and the demand for it. Consider first the supply response. A creditor’s decision whether to outsource debt collection to third-party agencies is likely based on the ability of the latter to use more aggressive collection practices while protecting the original creditor from reputational damage (Coffman (2011), Fedaseyeu and Hunt (2013)). Thus, the ability to hire third-party debt collection agencies may enable creditors to recover debts they would not be able to recover otherwise, which should increase their willingness to lend, all else being equal. Similarly, restrictions on third-party debt collection will reduce its effectiveness and may force some of the debt collection back in-house, which will decrease the overall 2 According to the Bureau of Labor Statistics’ Occupational Employment Statistics, in 2012, third-party debt collectors outnumbered debt collectors employed directly by financial institutions. Third-party debt collectors also appear to be using harsher debt collection practices than in-house debt collectors are using (see Fedaseyeu and Hunt (2013) for details).

7

effectiveness of the debt collection process. This will lower the credit supply. Now consider the demand response. More effective debt collection increases the probability that debts will have to be repaid even by borrowers in adverse circumstances (and may also impose a utility cost because of aggressive debt collection practices). Thus, more effective debt collection will reduce the demand for credit, all else being equal. The impact of third-party debt collection on credit provision is the net effect of its impact on credit supply and demand. If the supply effect dominates the demand effect, then restrictions on third-party debt collection will have a negative impact on the amount of credit provided. In principle, however, the demand effect could dominate the supply effect. In this case, debt collection restrictions will have a positive impact on the amount of credit provided. The ability of creditors to pursue delinquent debtors is limited by not only the effectiveness of debt collection but also by bankruptcy law and garnishment law. Garnishment is a legal order that enables creditors to collect a proportion of the debtor’s property in the possession of a third party. The most common form of garnishment is wage garnishment, which refers to the process of deducting funds directly from a person’s monetary compensation to satisfy his/her creditors. Restrictions on garnishment thus limit the ability of creditors (and debt collectors) to recover debts owed to them and will therefore lower credit supply. Since restrictions of garnishment protect a portion of borrowers’ earnings, they may also increase credit demand. Because the demand and supply responses go in opposite directions in this case, the ex-ante effect of garnishment restrictions on credit provision is unclear. Personal bankruptcy allows individuals to discharge their unsecured debts in return for giving up their nonexempt assets (under Chapter 7) or to retain their assets but repay a portion of their debts over a three- to five-year period (under Chapter 13). As with garnishment restrictions, the ex-ante effect of bankruptcy exemptions on credit provision is unclear. In my empirical 8

analysis, I will control for the generosity of bankruptcy law with the level of bankruptcy exemptions available to debtors and for garnishment restrictions with the amount of weekly earnings exempt from garnishment. While the main hypothesis I analyze in this paper is whether more effective debt collection increases credit supply, the ultimate goal of research on debt collection is the analysis of consumer welfare. The effect of debt collection on welfare is ambiguous. On the one hand, it enables lenders to increase credit supply and therefore provide loans to more borrowers. On the other hand, more effective debt collection increases borrowers’ disutility from being collected upon and may therefore negatively affect their welfare. These effects will, of course, depend on the availability of debt relief via the bankruptcy system. A formal welfare analysis is beyond the scope of the current paper. However, I will provide some preliminary results in this regard and will test whether more effective debt collection enables creditors to increase the pool of borrowers and to provide loans to riskier applicants.

4.

Legal framework

Debt collection in the United States is regulated by a federal law, the Fair Debt Collection Practices Act of 1977 (FDCPA). Unlike many other federal statutes, the FDCPA permits states to adopt their own regulations if they provide greater protection to the consumer than the federal law. The FDCPA therefore establishes a floor on consumer protection from debt collectors. Note that, while the state of incorporation governs the regulation of interest rates that banks with a national charter can offer (which is the primary reason many banks have moved to Delaware and South Dakota, the two states with the most lenient banking regulations in the U.S.), the relevant jurisdiction for creditor remedies and debt collection laws is the state in which the consumer resides (or resided when he or she opened the account). 9

Forty-three states have their own laws that regulate collection practices. Many of these statutes provide consumers with protections similar to those found in the FDCPA. However, state laws differ in some important respects that limit the operations of third-party debt collectors. Some states (Arizona, for example) require third-party debt collection agencies to obtain a license, while others (California, for example) do not. Some states (Arkansas, for example) require third-party debt collection agencies to post bonds with state regulators before commencing debt collection activities, while others (Iowa, for example) do not. States also differ in the responsibilities they assign to state debt collection regulators and in the powers those regulators are granted. For example, some states (e.g., Florida) allow attorneys general or special debt collection regulatory bodies to impose civil penalties on violators of debt collection laws. Some states also put limits on consumer civil remedies: Virginia statutes, for example, do not contain a private right of action for consumers aggrieved by debt collectors. State debt collection laws have changed over time: I have been able to identify 33 such changes in 22 states since 1999, of which six changes loosened restrictions on debt collectors, and 27 changes tightened restrictions on debt collectors.3 I use those changes to construct an index of debt collection restrictions that enables me to quantify the tightness of debt collection laws. Initially, I assign to each state a value that is the sum of the following six indicator variables that represent broad restrictions on debt collection activities this state had in 1998: 1) whether the state had a special board or commission that regulated debt collection activities; 2) whether the state imposed licensing requirements on third-party debt collectors; 3) whether the state imposed bonding requirements on third-party debt collectors; 4) whether the state declared certain abusive debt collection practices unlawful; 3 I code changes in debt collection laws from 1999 because my sample starts in 2000 (this is the first year for which all relevant variables are available). See Appendix B for a description of the procedure I used to identify relevant changes in state laws and Appendix C for a summary of those changes.

10

5) whether the state granted consumers a private right of action against debt collectors; and 6) whether the state made violations of debt collection laws a criminal offense. Then, starting with 1999, I identify all nontechnical changes in the six dimensions of debt collection laws described above,4 and I add 1 to the state index if the change can be interpreted as a tightening of debt collection laws and subtract 1 if the change can be interpreted as a loosening of debt collection laws. As a result, a higher value of the index implies a more restrictive environment for third-party debt collectors. Although giving each restriction equal importance in constructing the index does not accurately reflect the relative impacts of these regulations, it has the advantage of being transparent and easily reproducible.5 In addition, it does not require any subjective judgment about the relative strength of each restriction. Consider Colorado, for example. As of 1998, it had all six of the broad restrictions on debt collection activities previously mentioned. Therefore, the initial value of the index for Colorado is 6. In 2000, Colorado repealed the requirement that every individual debt collector has to be licensed (it retained the requirement that debt collection agencies need to be licensed) and shortened the statute of limitations for violations of debt collection laws from two years to one year. I interpret this change as a loosening of debt collection regulations and subtract 1 from the initial value of the index for Colorado. Thus, in 2000, the value of the index for Colorado is 5. It remains 5 until 2003, when Colorado limited the applicability of private remedies (violations of regulations are subject only to administrative enforcement). I interpret this change as another loosening of debt collection regulations and subtract 1 from the 2002 value of the index. Thus, in 2003, the value of the index for 4 I disregard technical changes because they are unlikely to have any material impact on the operations of debt collectors. For example, Florida replaced the Department of Financial Regulation with the Office of Financial Regulation in 2003. 5 States did not change their laws uniformly, which makes it problematic to determine the precise impact of each regulation. Consider the following three examples of the tightening of debt collection laws. In 2004, Georgia allowed class action lawsuits against unlicensed debt collection activity. In 2010, Florida authorized its attorney general to take action against third-party debt collectors and increased the amount of administrative fines from $1,000 to $10,000. In 1999, Oregon made violations of debt collection laws a criminal offense. It is fairly straightforward to see that each of these changes made it more difficult for debt collectors to operate since it increased their potential losses. However, it is unclear whether administrative fines in Florida should have a smaller or larger impact than class action lawsuits in Georgia or criminal punishment in Oregon.

11

Colorado is 4. There were no other changes in debt collection laws in Colorado after 2003; hence the value of the index remains 4 until 2012. Apart from debt collection statutes, other types of laws may affect the effectiveness of debt collection and credit supply and demand. As previously mentioned, two types of legal restrictions are particularly relevant for my analysis: restrictions on garnishment and bankruptcy exemptions. To control for the influence of bankruptcy exemptions, I obtain the level of both homestead exemption and nonhome personal property exemption for each state during my sample period (following the procedure described in Berkowitz and Hynes (1999) and Lin and White (2001)). I then include the dollar amount of the combined bankruptcy exemption (home and nonhome) as a control variable in my analysis. Some states allow their married residents who file jointly to double certain bankruptcy exemptions. Since I don’t have information on the marital status of borrowers, I use the exemptions available to singles. Some states offer unlimited homestead exemptions to their residents. In such cases, I code the value of the homestead exemption as $1 million, which is the amount used in Berkowitz and Hynes (1999) and in some specifications in Lin and White (2001).6 Some states allow their residents to choose between a uniform federal bankruptcy exemption and the state’s exemption level. In these cases, I assign the highest of the two exemption values. In total, there have been 122 changes in the level of bankruptcy exemptions in my sample: 63 changes in the amount of the homestead exemption, 57 changes in the amount of the nonhome personal property exemption, and two cases in which states allowed their residents to choose between the state’s exemptions and the uniform federal bankruptcy exemptions (Kentucky on March 18, 2005, and New York on January 21, 2011). The ability of creditors to garnish wages is limited by federal and state law. Federal law limits wage garnishment to the lesser of 25% of disposable income or the amount by which 6 The

amount of $1 million is greater than the maximum limited homestead exemption in the sample ($550,000).

12

disposable income exceeds 30 times the federal minimum wage, but some states offer greater protection to their residents. Prior literature has generally used the proportion of income exempt from garnishment to construct measures of the strictness of garnishment law (e.g., Dawsey and Ausubel (2004), and Dawsey, Hynes, and Ausubel (2013)). During my sample period, only one state (Massachusetts in 2011) changed the percentage of income exempt from garnishment, which makes the percentage of income exempt from garnishment almost perfectly collinear with state fixed effects. Therefore, I code restrictions on garnishment as the dollar amount of weekly income that is exempt from garnishment. In the states that prohibit garnishment, I assign the value of $1,500 per week, which is greater than the maximum limited garnishment exemption in the sample ($750).7 In total, there have been seven changes in the level of garnishment exemptions during my sample period.

5. 5.1.

Empirical analysis Data and variables description

The variables I use in my analysis come from three main sources: the Trend Data database (compiled by TransUnion), County Business Patterns surveys by the U.S. Census Bureau, and credit union Call Reports. TransUnion, which is one of the three largest consumer reporting agencies in the United States, collects data on, among other things, the amount of various types of consumer credit and on delinquency rates in each state. These data are provided in part via a solution called Trend Data, a database built from a series of large random samples of U.S. consumer credit histories. Each quarter, TransUnion draws a nationally representative random sample that contains 10% of consumer credit histories on file with TransUnion in that quarter. Each credit history contains variables on the amount 7 My

results are robust to alternative values of this amount, such as $1,000 or $2,000.

13

of revolving, installment, auto, and mortgage borrowing, as well as consumer repayment behavior and credit scores. TransUnion then aggregates these variables at the county, MSA, state, and national levels (I use the state-level data set because my main explanatory variable is the index of state laws). I convert all Trend Data variables from quarterly to annual frequency by calculating the average of the four quarterly observations every year for each Trend Data variable I use in my analysis.8 I use annual frequency in my main analysis for two main reasons. First, my main explanatory variable (the index state debt collection restrictions) is annual. Second, using annual frequency provides consistency since the dependent variables that are not obtained from Trend Data are annual.9 My sample period covers 2000–2012 (this is the longest period for which all relevant control variables are available) and contains observations for 48 states. I exclude Delaware and South Dakota because these two states have the most favorable banking laws in the U.S. and are therefore home to the vast majority of national credit card banks. I exclude the District of Columbia because the data on third-party debt collectors there are missing in all years. I keep the years in which debt collection laws changed if the effective date of the change fell in the month of January (there are four such changes). Otherwise, I exclude the years in which debt collection laws changed (there are 29 such changes). Further, I exclude 23 state-years in which data on the number of third-party debt collectors are unavailable, which leaves 572 observations in the main sample. All variables and their sources are listed in Table 1, and summary statistics are provided in Table 2. 8 An alternative would be to sum the four quarterly observations every year for each Trend Data variable. However, in this case, many of the variables, such as consumer credit scores and the proportion of borrowers with low credit scores, would lose their economic interpretation. In untabulated tests, I perform the same analysis after summing up the quarterly observations for the number of new loans while averaging the other Trend Data variables. The results remain qualitatively unchanged. 9 Another reason to use annual frequency is that credit accounts are reported to credit bureaus (including TransUnion) with a lag that ranges from one to three months. Using quarterly data may therefore introduce a measurement error in the dependent variable, which will reduce the efficiency of the estimates (but will not bias them). In untabulated tests, I perform the same analysis using quarterly observations for my main dependent variable. The results remain qualitatively unchanged.

14

[INSERT TABLE 1 ABOUT HERE] [INSERT TABLE 2 ABOUT HERE] My analysis requires dependent variables that reflect current credit conditions in each state and, in particular, the lenders’ willingness to extend credit. Such variables are available from Trend Data, which contains the number of new revolving lines of credit, the number of new auto loans, and the number of new mortgages, all normalized by the number of consumers with a credit report. Trend Data also reports average balances on these newly opened accounts. Revolving debt comprises accounts that are conventionally known as credit cards:10 A credit card allows multiple advances up to a predetermined credit limit and repayment amounts largely at the discretion of the cardholder. Once they pay off the balance, cardholders may borrow this amount again. The average number of new revolving accounts (per thousand borrowers) in my sample is 119.41. Auto loans are loans secured by motor vehicles, while mortgage loans are loans secured by real estate. The average number of new auto and mortgage loans (per thousand borrowers) in my sample is 6.91 and 9.20, respectively. In addition to variables on the number and balance of various loans by type of credit, Trend Data contains variables that reflect debtors’ riskiness and their demand for credit. Riskiness can be measured by consumer credit scores, which are a widely used metric of borrowers’ default probability and represent a rank ordering of consumers’ creditworthiness at a point in time. The average credit score in my sample is 660. Further, Trend Data provides some information about the distribution of credit scores in the population. In particular, Trend Data tracks the proportion of borrowers with credit scores below 700 (700 was the median credit score in TransUnion’s validation sample used to construct credit score 10 In Trend Data, revolving debt also includes some small home equity lines of credit. However, according to TransUnion, non-credit-card debt constitutes less than 10% of the total reported amount of revolving debt.

15

variables). Demand for credit can be proxied by the number of credit inquiries: Whenever a consumer applies for a loan, the creditor initiates what is called a “hard pull” on the consumer’s credit report (regardless of whether a loan is subsequently extended or not).11 By counting the number of hard pulls, one can create a measure of how often consumers apply for credit, which is a proxy for credit demand. The average number of credit inquiries in my sample is 105.3.12 Trend Data does not contain information on credit pricing or recovery rates. To obtain this information, I supplement Trend Data with Call Reports. The difficulty with using Call Reports is that they do not disaggregate data by state. Further, Call Reports from commercial banks do not contain data on the interest rates they charge — the information on interest rates charged by commercial banks is available only at the national level from the Quarterly Report of Credit Card Interest Rates conducted by the Federal Reserve Board. To circumvent both of these problems, I use credit union Call Reports. By law, credit unions are allowed to lend only to their members, who must have a well-defined common bond (employer, location, or profession). Hence, credit unions are more likely than commercial banks to be local credit providers and therefore lend within state borders.13 Further, credit union Call Reports contain data on interest rates charged on credit card loans. During my sample period, credit unions provided about 8% as much revolving credit as commercial banks.14 This means that credit unions generate a substantial volume of credit 11 TransUnion uses all hard pulls from consumers’ credit reports in constructing respective Trend Data variables, regardless of whether they are used in the calculation of consumer credit scores. Generally, hard pulls are used in the calculation of consumer credit scores. However, there is an exception to this practice when consumers engage in “rate shopping.” That is, when a consumer is looking for a mortgage, auto, or credit card loan and more than one lender requests his or her credit report, the calculation of the consumer’s credit score excludes these inquiries made within 30 days of scoring. Note that not all credit inquiries go to TransUnion: Many lenders pull a credit report from only a single credit bureau when evaluating a consumer credit application, and the distribution of inquiries across credit bureaus is not necessarily uniform. However, this distribution is determined by competition in the credit reporting industry and should be unrelated to state debt collection laws. 12 The Trend Data variable that reflects the number of hard pulls is constructed as an index relative to the validation sample: The average number of hard pulls in the validation sample is assigned the value of 100. 13 I exclude the Pentagon Federal Credit Union and the Navy Federal Credit Union because they provide credit across state lines. 14 Source: Board of Governors of the Federal Reserve System, G.19 series — Consumer Credit

16

card lending (even though less than commercial banks): The current total amount of credit card loans outstanding is about $840 billion, and the nature of revolving debt implies that much of this amount has been rolled over multiple times. The average interest rate charged by credit unions on their credit cards was 10.71% (see Table 2), while commercial banks charged on average 13.12%.15 Credit unions had an average recovery rate on charged-off credit card debt (calculated as the amount of credit card loans recovered divided by the amount of credit card loans charged off) of 11.41% during my sample period and an average charge-off rate (calculated as the amount of credit card loans charged off divided by the amount of credit card loans outstanding) of 2.44%. To provide a meaningful comparison, I calculate the corresponding values for small banks (defined as banks with less than $1 billion in total assets), which are more likely than large banks to be local credit providers. During my sample period, the average recovery rate on charged-off credit card debt and the chargeoff rate for small banks were 16.93% and 4.73%, respectively. Thus, credit unions charge their members lower interest rates and have lower default rates than banks, as one would expect given credit unions’ closer relationship with their borrowers. The fact that credit unions have lower recovery rates conditional on default is consistent with the view that they engage in less aggressive collection practices than commercial banks (either internally or via third-party debt collectors). It is therefore likely that the effectiveness of third-party debt collection is at least as important for banks as it is for credit unions. In addition to credit variables, credit union Call Reports contain data on credit union membership. To become a member, a person needs to purchase shares in a credit union by making a deposit in a share savings account (which is the equivalent of a savings account at a commercial bank). Credit unions can open accounts only for customers who satisfy their field of membership requirements (i.e., share a well-defined common bond with other credit 15 Source:

Board of Governors of the Federal Reserve System, G.19 Series – Consumer Credit.

17

union members, such as employer, location, or profession). By law, credit unions can lend only to their members, so the size of their membership can serve as a proxy for the size of the pool of potential borrowers. I calculate the total number of credit union members in a state by adding up the membership of all credit unions located in that state and then divide this number by the state’s population to compute the number of credit union members per thousand people. In my sample, the average number of credit union members per thousand people was 306.14.16 Data on third-party debt collectors (the number of debt collection establishments and their employees) are available from the Census Bureau’s County Business Patterns survey.17 On average, there were 423.81 debt collectors per million people in my sample, with about 11% of them employed by establishments with fewer than 10 employees. Data on personal income come from the Bureau of Economic Analysis. On average, personal income was $38,420 (in 2010 dollars) during my sample period. The state-level unemployment rate is obtained from the Bureau of Labor Statistics and averages 5.93% during my sample period. I use the high school graduation rate (from the National Center for Education Statistics) and the infant mortality rate (from the Centers for Disease Control and Prevention) as placebo outcomes in some of my robustness tests. The midyear population estimates come from the Census Bureau, and the Consumer Price Index (CPI) is obtained from the Bureau of Labor Statistics. 16 Note that this does not imply that 30.6% of people had an account at a credit union, because the same individual can have accounts at several credit unions. 17 The County Business Patterns survey tracks debt collectors under the NAICS code 561440, which comprises firms engaged in collecting payments and then remitting the payments collected to their clients. This survey does not track debt collectors employed by original creditors. Further, note that a single debt collection agency can have several establishments in one or several states, but the survey does not aggregate information at the agency (firm) level.

18

5.2.

The impact of debt collection laws on the amount and pricing of revolving credit

I estimate the effect of debt collection laws on the amount and pricing of revolving credit with the following model: Yi,t = αi + γt + βIndexi,t + η 0 Controlsi,t + εi,t ,

(1)

where Yi,t is the value of the dependent variable in state i in year t, and Indexi,t is the corresponding value of the index of debt collection restrictions. The following controls are included: unemployment rate and real income per capita (to control for general economic conditions), three lags of real per capita income growth (to account for the local business cycle), mean credit score (to control for the riskiness of the pool of borrowers), number of credit inquiries (to account for demand-driven variation), the amount exempt from garnishment, and the combined bankruptcy exemption (to account for state-level protections offered to consumers who default). Time fixed effects are included to remove macro-level trends, while state fixed effects eliminate unobservable time-invariant heterogeneity across states. Standard errors are clustered by state and all nominal variables are converted to 2010 dollars using the CPI. The estimates of the impact of debt collection restrictions on the amount and pricing of revolving credit are presented in Table 3. In the first two columns of Table 3, the dependent variable is the number of new revolving lines of credit. In the next two columns, the dependent variable is the average balance on new revolving lines of credit, while in the last two columns the dependent variable is the average interest rate charged by credit unions on their unsecured credit cards. [INSERT TABLE 3 ABOUT HERE] 19

The estimated effect of debt collection laws on the number of new revolving lines of credit is negative and statistically significant. A 1 point increase in the the value of the index of debt collection restrictions reduces the number of new revolving lines of credit per thousand consumers by roughly 2.1, or about 2% of the sample mean. The negative sign indicates that the supply effect dominates the demand effect: Stricter debt collection laws decrease creditors’ willingness to lend by more than they increase consumers’ willingness to borrow. In terms of its economic magnitude, increasing the value of the index from 2 (the lowest quartile) to 8 (the maximum value of the index in the sample) will decrease the number of new revolving lines of credit by approximately 11%. This effect is comparable with the effect of bankruptcy exemptions on mortgage denial rates documented by prior literature. Berkowitz and Hynes (1999) find that (after accounting for state fixed effects) quadrupling the homestead bankruptcy exemption would lead to a decrease in the probability of mortgage denial by about 10 basis points in the states with low exemption levels and by about 2 percentage points in the states with high exemption levels. These changes represent 0.6% and 13% of their sample average denial rate, respectively. Lin and White (2001) find that applicants who live in the states with unlimited bankruptcy exemptions are 2 percentage points more likely to be denied a home purchase loan and 5 percentage points more likely to be denied a home improvement loan than applicants who live in the states whose exemptions are in the lowest quartile of the exemption distribution. These changes represent 13% and 17% of the average sample denial rate for home purchase and home improvement loans in Lin and White (2001), respectively. The impact of state debt collection restrictions is also comparable with the impact of other contract enforcement mechanisms on credit provision. For example, Jappelli, Pagano, and Bianco (2005) find that decreasing the stock of pending trials per capita (which is one of their measures of judicial inefficiency) by about one-third of its sample mean will increase the ratio of credit to GDP by 1.5 percentage points, or 4% 20

of their sample average ratio of credit to GDP. Overall, the fact that stricter debt collection laws reduce the availability of credit is consistent with the general message of a broader literature on law and finance, which finds that investor rights are a strong determinant of capital markets development (e.g., La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998)). The validity of my estimates relies on the parallel trends assumption, which requires that states that introduced changes in their laws followed similar trends (at the time of such changes) as the states that did not change their laws. In particular, my estimates can be biased upward if the states that tightened their debt collection laws followed a downward trend (relative to other states) in the number of new account openings before the change. To provide graphical evidence on the validity of the parallel trends assumptions, I plot time trends in the number of new revolving accounts around the adoption of debt collection laws (in Figure 2). [INSERT FIGURE 2 ABOUT HERE] Figure 2 plots, in event time, the average number of new revolving lines of credit for treated states (the states that changed their laws at any point during my sample period) and untreated states (the states that did not change their laws during my sample period), with the year of the change omitted from the figure. Black squares represent values for treated states and blue triangles represent values for untreated states. Panel (a) shows time trends around tightenings of debt collection laws, while panel (b) shows time trends around loosenings of debt collection laws. Consider panel (a) first. Treated and untreated states do appear to follow very similar trends three years prior to law changes. Prior to the change, the states that tightened their debt collection laws had, on average, a larger number of revolving account openings than the states that did not change their laws. After the change, however, the number of revolving account openings in the states that tightened their laws decreased

21

to the level below that of the untreated states and followed a downward trend for at least three years after the change. Similar conclusions can be drawn from panel (b), although the changes there are less pronounced. States that loosened their debt collection laws had fewer revolving account openings prior to such laws changes and increased the number of revolving account openings almost to the level of untreated states after the change. It also appears that changes in debt collection restrictions have a lasting, rather than just temporary, effect on credit provision. As is the case with any regulatory change, debt collection laws are enacted after a period during which possible regulatory changes are proposed and then discussed. Furthermore, laws generally do not become effective immediately after passage, which creates an additional time lag during which creditors and debt collectors can adjust their behavior in anticipation of changes in the regulatory environment. Such adjustment is likely to bias my results downward, because some of the changes in credit availability attributable to changes in debt collection laws will occur prior to their implementation. However, in Figure 2, there is no evidence that such adjustment takes place. Other coefficients in Table 3 are reasonable. Unemployment negatively affects the number of new revolving lines of credit, consistent with the idea that creditors are unwilling to lend when borrowers’ job prospects are poor. As expected, higher credit scores and more credit inquiries lead to a greater number of new revolving accounts. The coefficients on garnishment and bankruptcy exemptions have negative signs (albeit statistically insignificant). The somewhat surprising negative sign of the coefficient on income per capita is likely driven by less credit demand, after income growth has been controlled for. Besides the number of new accounts, debt collection restrictions can also affect the size of the loan. As with the number of accounts, the net effect of debt collection restrictions on the size of the loan depends on the relative strength of the supply and demand responses. I 22

explore the impact of debt collection laws on the size of the loan in columns (3) and (4) of Table 3. I do not find any statistically distinguishable effect of debt collection restrictions on loan balances. This may be due to the nature of revolving debt, in which the creditor determines the borrowing limit (which I do not observe), while the debtor decides how much to borrow (up to this prespecified credit limit). Since stricter debt collection laws reduce the number of new account openings (as previously noted), it is plausible that consumers will try to borrow more per account, conditional on obtaining it. Thus, the demand effect here is likely to be stronger than in the case of the number of new accounts, which will attenuate my estimates toward zero. However, the point estimates of the effect of debt collection laws on loan balances are negative and comparable in terms of their economic magnitude to the estimates for the number of new revolving accounts, which suggests that the supply effect (which will manifest itself in lower credit limits) is not entirely offset by the demand effect. Apart from the amount of revolving credit, the efficacy of debt collection may influence its pricing. The ex-ante effect of debt collection laws on pricing is ambiguous. On the one hand, the expansion of credit supply due to better debt collection may lead to lower interest rates. On the other hand, lenders may be willing to expand the pool of borrowers by extending credit to riskier applicants when debt collection is more effective. In this case, the average equilibrium interest rate may go up because these new borrowers will be charged higher interest rates commensurate with their risk characteristics.18 I investigate the effect of debt collection laws on credit card interest rates (charged by credit unions) in the last two columns of Table 3. I find no significant effect of state debt collection laws on the pricing of unsecured credit 18 Slightly more formally, assume that r is the interest rate charged to the safe borrowers, and r is the interest rate charged s r to the risky borrowers (rs < rr ). If debt collection is ineffective and only safe borrowers obtain credit, the equilibrium average interest rate is rs . When debt collection is effective and both types of borrowers obtain credit, the equilibrium average interest rate is rs ω + rr (1 − ω), where ω is the share of credit obtained by safe borrowers. It is immediate that rs < rs ω + rr (1 − ω).

23

card loans. The economic magnitude of the effect is small (around 5 to 7 basis points). One possible explanation is that the nature of credit card contracts dampens the effect of debt collection laws on interest rates. In particular, credit card issuers are more likely to adjust interest rates on new credit card offers than on existing accounts. The latter generally happens after an adverse event on behalf of the borrower (such as a missed payment).19 Thus, any changes in interest rates on existing credit card balances are likely to manifest themselves in penalty rates and late fees on delinquent credit card accounts. Since such accounts represent a small share of total accounts, it is likely that my regressions here lack sufficient statistical power. 5.3.

Transmission mechanism: recovery rates and the number of debt collectors

There are two potential channels by which debt collection laws can influence credit supply. The first possible channel is changes in recovery rates conditional on debtors’ default. Stricter debt collection laws lead to weaker enforcement of consumer credit contracts and should thus reduce recoveries, which, in turn, should make lenders less willing to extend credit. The second channel is changes in debtors’ likelihood of default. Weaker enforcement of consumer credit contracts may prompt debtors to default more often, and a higher likelihood of default should make lenders less willing to extend credit in the first place. I explore these channels in Table 4 (in columns (1) through (4)). [INSERT TABLE 4 ABOUT HERE] The effect of debt collection laws on recovery rates is shown in the first two columns of Table 4.20 Stricter debt collection laws do appear to lower recovery rates on charged-off 19 After the passage of the Credit Card Accountability Responsibility and Disclosure Act of 2009, credit card issuers generally are prohibited from increasing the interest rate on an existing balance unless the cardholder has missed two consecutive payments. Prior to the passage of the Act, lenders could raise interest rates for other reasons related to changes in the borrower’s credit risk (see Furletti (2003)). 20 Note that columns (1) and (2) of Table 4 use data from credit union Call Reports, columns (3) and (4) use Trend Data, and columns (5) and (6) use data from the County Business Patterns surveys.

24

credit card loans by credit unions. The corresponding coefficient is negative, statistically significant, and economically large: A 1 point increase in the value of the index of debt collection restrictions reduces recovery rates by about 1.4 percentage points, or 12% of the sample mean. Thus, debt collection restrictions have a larger economic impact on recovery rates than on the number of accounts. This is not surprising, given that recovery rates post-default are not the only determinant of credit availability. To evaluate the validity of the parallel trends assumption for recovery rates, I plot time trends around law changes in Figure 3. [INSERT FIGURE 3 ABOUT HERE] Panel (a) of Figure 3 shows the history of average recovery rates around tightenings of debt collection laws. Prior to law changes, the states that tightened their laws and the other states followed similar trends. After the change, the states that tightened their debt collection laws saw a pronounced drop in the recovery rates on credit card loans, while the states that did not change their laws experienced a slight increase in recovery rates. Notice also that most of the drop in recovery rates for the treated states takes place around the second year after the implementation of the law. This is in contrast with the number of new revolving lines of credit in Figure 2 and suggests that lenders reduce credit supply in anticipation of lower recovery rates. Panel (b) of Figure 3 shows recovery rates around loosenings of debt collection laws and leads to similar conclusions, although the changes here are less pronounced. The effect of debt collection laws on delinquency rates is shown in columns (3) and (4) of Table 4. The dependent variable here is the number of revolving borrowers 90 days or more past due (per thousand consumers). I find no statistically distinguishable effect of debt collection laws on the likelihood that debtors will default on their revolving loans.

25

However, the point estimate is positive and economically large (a 1 point increase in the value of the index of debt collection restrictions is associated with an increase in the number of delinquent borrowers of about 5% of the sample mean). This suggests that borrowers default more often when debt collection laws are more stringent (and the enforcement of credit contracts is therefore weaker). Nonetheless, the main effect of debt collection laws on credit supply appears to result from changes in recovery rates after default. Debt collection restrictions do not affect recovery rates directly. Rather, they influence the efficacy of debt collection activity (and thus the strength of contract enforcement), which, in turn, affects recovery rates. Changes in the efficacy of debt collection activity may come from two sources: changes in the number of debt collectors and changes in the debt collection practices they use. Debt collection laws affect both of these sources. On the one hand, they restrict entry in the debt collection industry by imposing bonding and licensing requirements. On the other hand, they restrict debt collection practices by defining unlawful conduct and prescribing remedies against violations of debt collection laws. Since there are no public data about the practices of debt collectors, I can only investigate the effect of debt collection laws on the number of debt collectors. The results are presented in the last two columns of Table 4. As expected, a more stringent debt collection environment (reflected in a higher value of the index of debt collection restrictions) leads to a lower number of debt collectors per capita. The coefficient is statistically and economically significant in three out of four specifications: A 1 point increase in the value of the index lowers the number of debt collectors per million people by about 11% of the sample mean. Figure 4 presents graphical evidence on the validity of the parallel trends assumption with respect to the number of debt collectors. [INSERT FIGURE 4 ABOUT HERE]

26

Prior to law changes, time trends in the states that changed their laws appear to be similar to time trends in the other states. After the change, the states that tightened their laws experienced a pronounced drop in the number of debt collectors per capita, unlike the other states, in which debt collector density remained virtually unchanged. The states that loosened their laws saw an increase in the number of debt collectors per million people. However, this increase was most pronounced around the third year after the change. A possible concern with the results in Table 4 is that debt collectors located in one state may sometimes collect on debtors who reside in a different state. While the extent of such cross-state collection is impossible to ascertain with the data I have, it is likely to bias my estimates downward, since it would dampen the effect of debt collection laws on the employment of debt collectors. By the same token, the existence of cross-state debt collection will also lessen the effect of debt collection laws on recovery rates and credit supply: If location didn’t matter for the effectiveness of debt collection, then changes in debt collection laws would simply shift debt collection activity across states without changing the “effective” level of debt collection activity in any given state. The fact that changes in debt collection laws do appear to have an impact on debt collection employment and on credit supply suggests that the redistribution of debt collection activity across states does not occur instantly and/or is imperfect. One reason for this is that state licensing requirements often specify that only debt collection agencies registered in the state are allowed to collect debts owed by the residents of that state. Another reason is that knowledge of local economic conditions is important in the process of debt collection, which requires proximity of debt collectors to borrowers and limits the ability of debt collection agencies to relocate across state borders.

27

5.4.

Welfare implications

The focus of this paper is the effect of third-party debt collection on the provision of unsecured consumer credit. While this is a natural starting point, the ultimate goal of research on debt collection is the analysis of consumer welfare, and this section provides some preliminary evidence in this respect. One aspect of welfare that can be affected by debt collection laws is the size of the pool of borrowers and its composition. Effective debt collection can increase lenders’ willingness to expand the pool of potential borrowers. To the extent that credit access helps borrowers smooth their consumption over time, effective debt collection will increase welfare.21 Further, since debt collection is more relevant for borrowers with a high default probability, lenders should be willing to provide credit to riskier borrowers when debt collection laws are lax and should withdraw credit from such borrowers when debt collection laws are strict. I estimate the effect of debt collection laws on the size and composition of the pool of borrowers in Table 5. [INSERT TABLE 5 ABOUT HERE] I use credit union membership (the number of credit union members per thousand people) as a proxy for the size of the pool of borrowers. Since credit unions are allowed to lend only to their members, a larger credit union membership indicates a greater willingness of credit unions to provide credit. The effect of debt collection laws on credit union membership is negative (columns (1) and (2) of Table 5). However, it is just shy of 10% statistical significance when garnishment laws and bankruptcy exemptions are included in the estimation. In terms of economic magnitude, a 1 point increase in the strictness of debt collection laws reduces the number of credit union members per thousand people by about 5.6, or 1.8% of 21 Of

course, this could also indicate that borrowers do not fully take into account the harshness of collection practices they will face upon default. More research is needed to shed light on this issue, which is beyond the scope of this paper.

28

the sample mean. The magnitude of this effect is similar to the impact of debt collection laws on the number of new revolving lines of credit, suggesting that a significant share of variation in credit supply in response to changes in debt collection laws comes from lenders’ ability to offer credit to new borrowers. Columns (3) and (4) of Table 5 show the estimated impact of debt collection laws on the riskiness of the pool of borrowers. As expected, tighter debt collection laws are associated with a less risky pool of borrowers, presumably because lenders are unwilling to extend credit to borrowers with high default probabilities when the likelihood of recovering delinquent debt goes down. However, the economic magnitude of this effect is small: The share of borrowers with credit scores below 700 goes down by about 10 basis points (0.2% of the sample mean) for a 1 point increase in the value of the index of debt collection restrictions. A larger pool of borrowers indicates that effective debt collection expands credit access and may therefore increase welfare. However, this positive effect can be offset by the fact that effective debt collection may be associated with harsher debt collection practices, which may lower borrowers’ utility. Fedaseyeu and Hunt (2013) suggest that smaller debt collection agencies may be better able to escape regulatory scrutiny and, therefore, may be more likely than large debt collection agencies to use harsh debt collection practices. Thus, a larger share of small debt collection establishments may indicate harsher debt collection practices, on average. Columns (5) and (6) of Table 5 show that the share of debt collection employment by small firms relative to total debt collection employment increases when debt collection laws are more stringent. This supports the idea that the lower regulatory scrutiny that small debt collection agencies enjoy reduces the expected costs imposed by additional regulations (relative to bigger debt collection agencies). This size effect is also consistent with the view that the penalties associated with new regulations sometimes make exit preferable to 29

paying the penalties and continuing operations. Since such exit is plausibly less costly for smaller firms, new regulations may reduce the average firm size. While the evidence above is indirect, it indicates that stricter debt collection laws, while reducing the total number of debt collectors, may in fact favor those debt collectors that are willing to use harsh debt collection practices. Thus, stricter debt collection laws might lower welfare by both reducing credit supply and increasing the propensity of debt collectors (who remain in operation) to use harsh debt collection practices. Another potential impact of debt collection laws on the harshness of debt collection practices may come from changing proportions of third-party debt collectors relative to in-house debt collectors. It may be, for example, that additional restrictions on third-party debt collectors, while decreasing their employment (as shown in Table 4), either increase or don’t change the employment of in-house debt collectors. Since the latter are less likely to use harsh debt collection practices, a larger share of such collectors may increase borrower welfare. However, state-level data on the employment of in-house debt collectors are currently unavailable; therefore, addressing this issue is a topic for future research. 5.5.

Robustness checks

The index of debt collection restrictions used throughout this paper puts equal weight on each restriction. This has the benefit of making the index transparent and easily reproducible and does not require any subjective judgment about the relative strength of each restriction. However, some restrictions are likely to be more important than others, and one would expect the relatively stronger restrictions to have a more pronounced impact on credit availability. To investigate this issue while being minimally ad hoc, I split all restrictions into two categories: restrictions on entry and restrictions on practices. The rationale behind this distinction is the interaction between state and federal law. While debt collection

30

practices are regulated by both federal and state law, only state law places restrictions on entry (such as licensing and bonding requirements). Federal law is uniform across states and binds in the states in which it is stricter than state law. For such states, variation in state laws that regulate debt collection practices is irrelevant (since it is the stronger federal law that operates there). It is therefore plausible that state-level restrictions on debt collection practices will have a smaller impact on credit provision than state-level restrictions on entry. This does not mean that restrictions on debt collection practices are unimportant. Rather, it means that their impact on credit provision is more difficult to identify from state-level variation. Following this logic, I split my index into two subindices. The first subindex includes restrictions on entry, and the second subindex includes restrictions on practices. I then regress the number of new revolving lines of credit, the average recovery rates on charged-off credit card loans by credit unions, and third-party debt collector density on these subindices. The results are reported in Table 6. For comparison, columns (1), (3), and (5) of Table 6 replicate the results from column (2) of Table 3 and columns (2) and (6) of Table 4, respectively. Columns (2), (4), and (6) of Table 6 show the results for the two subindices. [INSERT TABLE 6 ABOUT HERE] The point estimates for both subindices have the expected sign: Their effect on the variables of interest is always negative. However, restrictions on entry uniformly have a larger impact on the outcome variables. For example, the effect of restrictions on entry on the number of new revolving lines of credit is four times greater than the corresponding effect of restrictions on practices. Again, it is important to note that even though restrictions on debt collection practices at the state level appear to be less important than restrictions on entry (as expected), this does not imply that restrictions on debt collection practices have

31

little effect on debt collection activities or on credit provision. Rather, it supports the view that in some states, it is the federal and not the state law that puts the most stringent limits on debt collection practices. It further implies that the results reported in this paper possibly underestimate the effect of debt collection on credit availability because they do not fully account for the impact of restrictions on debt collection practices (since the identification comes primarily from the restrictions on entry). So far in my analysis, I have focused on the impact of third-party debt collection on unsecured credit. This is because debt collectors are primarily engaged in collecting unsecured debts, since in the case of secured credit, the lender can repossess the underlying collateral. Debt collectors are sometimes hired to collect on deficiency judgments (the remaining balance less the value of the collateral seized by the creditor). These judgments, however, are generally small relative to the initial loan amount. Thus, debt collection laws should have little direct effect on the provision of secured credit. However, debt collection can have an indirect effect on secured credit because secured and unsecured credit are linked by household behavior.22 Mitman (2012) shows that higher bankruptcy exemptions may result in higher interest rates on unsecured debt (because households’ propensity to file for bankruptcy increases) and can therefore prompt households to decrease their demand for unsecured credit and increase their demand for secured credit. By a similar logic, more stringent debt collection laws may increase the demand for unsecured credit (because borrowers’ disutility from debt collection goes down) and may reduce the demand for secured credit. As the analysis in this paper demonstrates, the supply of unsecured credit reacts to changes in debt collection laws more strongly than demand. However, the demand effect could dominate the supply effect in the case of secured credit for two reasons. 22 Secured and unsecured credit are also linked by their treatment in the bankruptcy system. For example, write-offs of unsecured debt in bankruptcy can improve repayment prospects on mortgages. However, since debt collection occurs prior to bankruptcy, the prospect of write-offs of unsecured debt to keep making mortgage payments should play little role here.

32

First, as mentioned previously, debt collection has little direct effect on the supply of secured credit (apart from deficiency judgments). Second, the supply of secured credit is likely to be restricted by the availability of collateral. To investigate the effect of debt collection laws on secured credit, I regress the number of new secured auto loans and mortgages on the index of debt collection restrictions, with the results reported in the first two columns of Table 7. [INSERT TABLE 7 ABOUT HERE] I find no statistically significant effect of debt collection laws on either auto loans or mortgages. However, the point estimates in the first two columns of Table 7 are negative and comparable in terms of their economic magnitude with the effect of debt collection laws on unsecured credit. Thus, it may be the case that debt collection laws have an indirect affect on the provision of secured credit by decreasing demand (indirectly). Another possible interpretation of the negative association between debt collection restrictions and the amount of secured debt is that changes in the index of debt collection laws are correlated with the general economic conditions that affect the supply of both secured and unsecured credit. Controlling for unemployment, income per capita, and lags of income growth should mitigate this concern. To address it more directly, I perform several placebo tests and report the results in the last three columns of Table 7, in which I regress state-year unemployment, high school graduation, and infant mortality rates on the index of debt collection restrictions.23 One would expect that debt collection should have little effect on such distantly related outcome variables, which is what I find. None of the estimates are significant. Another concern with my analysis is the influence of outliers. In particular, my results can be driven by individual states that experienced a very rapid growth in the amount of 23 I use the average freshman graduation rate as the dependent variable in the regressions of high school graduation rates on the index of debt collection restrictions. Beginning with the school year 2010–2011, the Department of Education started calculating a more accurate measure of high school completion that takes into account changes in the composition of the student body (called the four-year adjusted cohort graduation rate) and released the first data in April 2014. However, this new rate is not available for prior years, which is why I use the older measure in my analysis.

33

revolving debt after loosening their debt collection laws or a very rapid decline in the amount of revolving debt after tightening their debt collection laws. To investigate this possibility, I run the same regressions as before, but I exclude states that changed their debt collection laws from the analysis, one by one. The results of these regressions are presented in Table 8. Each row in this table presents the coefficients from three regressions, after excluding the corresponding state (specified on the left). The first is the regression of the number of new revolving lines of credit on the index of debt collection restrictions and all controls. The second is the regression of average recovery rates on the index of debt collection restrictions and all controls. The third is the regression of third-party debt collector density on the index of debt collection restrictions and all controls. I report only the coefficient on the index of debt collection restrictions from each of those regressions (but include all control variables in the estimation). [INSERT TABLE 8 ABOUT HERE] The coefficients reported in Table 8 are in line with those reported above in terms of their magnitude and statistical significance. Thus, it is unlikely that my results are driven by outlier states. I perform a similar test by excluding each year from the analysis, one by one. The results, reported in Table 9, demonstrate that the effect of debt collection laws on unsecured credit is unlikely to be driven by outlier years either. [INSERT TABLE 9 ABOUT HERE] I perform several additional robustness checks and report the results in Table 10. Each row of Table 10 presents the results from three separate regressions on restricted subsamples described in the leftmost column. The first is the regression of the number of new revolving lines of credit on the index of debt collection restrictions and all controls. The second is the regression of average recovery rates on the index of debt collection restrictions and all 34

controls. The third is the regression of third-party debt collector density on the index of debt collection restrictions and all controls. I report the coefficient on the index of debt collection restrictions from each of those regressions (but include all control variables in the estimation). For comparison, in the first row of Table 10, I show the results for the full sample reported above. [INSERT TABLE 10 ABOUT HERE] I start with additional tests to gauge the robustness of my findings to the period of analysis. First, I exclude the recent financial crisis from the analysis (by dropping years between 2007 and 2009), with the resulting estimates being in line with the estimates for the full sample. Second, I retain only the sample period before the enactment of the Bankruptcy Abuse Prevention and Consumer Protection Act of 2005 (BAPCPA). The BAPCPA made it more costly for consumers to file for bankruptcy by requiring them to undergo mandatory credit counseling before filing and by limiting their ability to choose between Chapter 7 and Chapter 13. Thus, the BAPCPA made bankruptcy law less debtor friendly and therefore increased the ability of debt collectors to pursue consumers prior to bankruptcy. Consistent with this view, the results reported in Table 10 demonstrate that the impact of debt collection laws on credit provision was smaller before the BAPCPA was enacted: The estimates retain their negative sign but go down in economic magnitude relative to the full sample and lose their statistical significance (likely due to the much shorter time series). In the final robustness check on the period of analysis that I perform, I retain only the year before and after each law change. This reduces the statistical power of my tests (due to the very small number of observations), but my estimates retain their signs and economic significance. As a further test of robustness of the parallel trends assumption, I reestimate my results by including linear state-specific time trends in the analysis. The impact of debt collection

35

laws on the number of new revolving lines of credit and on recovery rates remains virtually unchanged after including parametric time trends, which lends credence to the parallel trends assumption. However, the coefficient on the debt collector density, while retaining its negative sign, falls in economic magnitude and loses statistical significance. This may be due to the fact that, as depicted in panel (b) of Figure 4, the effect of debt collection laws on the number of debt collectors grows gradually (after loosenings of debt collection laws). Thus, it may be more difficult to pick up with a parametric time trend. Another possible concern is the generalizability of my results on recovery rates, since the relevant data come from credit union Call Reports. Two issues arise here. First, credit unions may not be representative of banks, which provide the majority of revolving credit in the U.S. Second, credit unions sometimes define their membership requirements quite broadly. For example, they may permit people to join not only if they live in a particular area but also if they work there. To address the first issue, I calculate the recovery rates on credit card loans for small banks (defined as banks with less than $1 billion in assets).24 Small banks are more likely than large banks to be local credit providers and are therefore more likely to be sensitive to state-level debt collection restrictions. To address the second issue, I exclude credit unions located in border counties from the analysis. Since the sample in this case consists only of credit unions from inner counties, this should minimize the effect of commuters who live in one area but may borrow from credit unions in a different area. The results of these additional tests are reported in the last two rows of Table 10. While the point estimates for both credit unions from inner counties and small banks are similar to the estimate for the full sample, the estimate for small banks is statistically insignificant. This is unsurprising, given that even small banks are more likely than credit unions to lend across state lines. As such, while the overall effectiveness of debt collection may be at least 24 I

use each bank’s headquarters to assign it to a particular state and then aggregate the data from bank Call Reports at the state level.

36

as important for banks as for credit unions, state-level debt collection restrictions are likely to play a smaller role in banks’ lending operations than in credit unions’ lending operations.

6.

Conclusion

I examine contract enforcement in the consumer credit market by studying the role of thirdparty debt collectors. I construct a state-level index of the tightness of debt collection laws and find that stricter regulations of third-party debt collectors are associated with fewer openings of revolving lines of credit: Increasing the value of the index from its first quartile to its maximum value in the sample increases the number of new revolving account openings by 11%, which is similar to the effect of bankruptcy exemptions on mortgage loans documented in prior literature. I also find that stricter debt collection laws reduce recovery rates and the number of third-party debt collectors per capita, which suggests that the impact of stricter debt collection laws on credit provision results from lower recovery rates due to fewer debt collectors and therefore less pressure on borrowers to repay their debts. Further, I document that effective debt collection results in a larger and riskier pool of borrowers, suggesting creditors’ willingness to lend to new, high-risk applicants who would otherwise be rationed out of the credit market. However, more research is needed to ascertain the impact of debt collection on consumer welfare. Although other factors such as social norms and the stigma associated with default surely play an important role, robust contract enforcement can help explain the existence of large and active retail credit markets and contribute to our understanding of how these markets function. In terms of policy implications, my results indicate that financial regulation that institutes strong consumer protection must be balanced with creditor rights in order for the latter to extend consumer credit in the first place.

37

References K. Athreya, J. M. Sanchez, X. Tam, and E. R. Young. Bankruptcy and Delinquency in a Model of Unsecured Debt. Working Paper 2012-042B, Federal Reserve Bank of St. Louis, 2013. J. Berkowitz and R. Hynes. Bankruptcy Exemptions and the Market for Mortgage Loans. Journal of Law and Economics, 42(2):809–830, 1999. J. Y. Campbell. Household finance. Journal of Finance, 61(4):1553–1604, 2006. L. C. Coffman. Intermediation Reduces Punishment (and Reward). American Economic Journal: Microeconomics, 3(4):77–106, 2011. A. E. Dawsey and L. M. Ausubel. Informal Bankruptcy. Working Paper, University of North Carolina Greensboro and University of Maryland, 2004. A. E. Dawsey, R. M. Hynes, and L. M. Ausubel. Non-Judicial Debt Collection and the Consumer’s Choice Among Repayment, Bankruptcy and Informal Bankruptcy. American Bankruptcy Law Journal, 87:1–26, 2013. A. A. Dick and A. Lehnert. Personal Bankruptcy and Credit Market Competition. Journal of Finance, 65(2):655–686, 2010. I. Domowitz and R. Sartain. Determinants of the Consumer Bankruptcy Decision. Journal of Finance, 54(1):403–420, 1999. L. A. Drozd and R. Serrano-Padial. Modeling the Credit Card Revolution: The Role of Debt Collection and Informal Bankruptcy. Working Paper 13-12, Federal Reserve Bank of Philadelphia, April 2013.

38

Ernst & Young. The Impact of Third-Party Debt Collection on the National and State Economies. Technical Report, Ernst & Young, 2014. S. Fay, E. Hurst, and M. White. The Household Bankruptcy Decision. American Economic Review, 92(3):706–718, 2002. V. Fedaseyeu and R. Hunt. The Economics of Debt Collection: Enforcement of Consumer Credit Contracts. Working Paper, Bocconi University and Federal Reserve Bank of Philadelphia, May 2013. M. Furletti. Credit Card Pricing Developments and Their Disclosure. Working Paper 03-02, Federal Reserve Bank of Philadelphia, 2003. R. Gropp, J. K. Scholz, and M. J. White. Personal Bankruptcy and Credit Supply and Demand. Quarterly Journal of Economics, 112(1):217–251, February 1997. D. Gross and N. Souleles. An Empirical Analysis of Personal Bankruptcy and Delinquency. Review of Financial Studies, 15(1):319, 2002. R. Hunt. Collecting Consumer Debt in America. Business Review, Federal Reserve Bank of Philadelphia, Q2:11–24, 2007. R. Hynes. Broke but Not Bankrupt: Consumer Debt Collection in State Courts. Florida Law Review, 60, 2008. T. Jappelli, M. Pagano, and M. Bianco. Courts and Banks: Effects of Judicial Enforcement on Credit Markets. Journal of Money, Credit and Banking, 37(2):223–244, 2005. R. La Porta, F. Lopez-de-Silanes, A. Shleifer, and R. Vishny. Law and Finance. Journal of Political Economy, 106(6):1113–1155, 1998.

39

E. Y. Lin and M. J. White. Bankruptcy and the Market for Mortgage and Home Improvement Loans. Journal of Urban Economics, 50(1):138–162, 2001. B. Melzer. The Real Costs of Credit Access: Evidence from the Payday Lending Market. Working Paper, Kellogg School of Management, Northwestern University, January 2009. K. Mitman. Macroeconomic Effects of Bankruptcy and Foreclosure Policies. Working Paper, University of Pennsylvania, 2012. A. Morse. Payday Lenders: Heroes or Villains? Journal of Financial Economics, 102(1): 28–44, 2011. J. Scott and T. Smith. The Effect of the Bankruptcy Reform Act of 1978 on Small Business Loan Pricing. Journal of Financial Economics, 16(1):119–140, 1986. P. Tufano. Consumer Finance. Annual Review of Financial Economics, 1(1):227–247, 2009. M. J. White. Bankruptcy Reform and Credit Cards. Journal of Economic Perspectives, 21 (4):175–200, 2007.

40

Figure 1: Proportion of Consumers with Accounts in Collection 15%

10%

Proportion of consumers with accounts in third−party collection 5%

0

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

This figure depicts the percentage of U.S. consumers with at least one account reported as being collected by third-party debt collectors. Source: The Quarterly Report on Household Debt and Credit, Federal Reserve Bank of New York, various editions.

41

42

Trend Trend Trend Trend Trend Trend Trend Trend

Number Average Number Number Number Average Average Number

Bank Call Reports (prior to 2001: account riad4263 divided by account riad4262; after 2001: account riadb515 divided by account riadb514) Bureau of Economic Analysis Bureau of Labor Statistics National Center for Education Statistics Centers for Disease Control and Prevention (National Vital Statistics Reports)

Average recovery rate on charged-off unsecured credit card loans by small banks (with less than $1 billion in total assets)

Personal income per capita Unemployment rate High school graduation rate Infant mortality rate (per 1,000 live births)

For each variable, this table indicates its data source and the way it was constructed. The longest time period for which all control variables are available is 2000–2012.

Credit union Call Reports (account 681 divided by account 680) Credit union Call Reports (account 521) Credit union Call Reports (account 521)

Average recovery rate on charged-off unsecured credit card loans by credit unions Average interest rate charged by credit unions on unsecured credit card loans Number of credit union members

(data item rennc) (data item reabn) (data item bannc) (data item mtnnc) (data item repb90m) (data item itoinq180) (data item tmmean) (sum of data items tram1 through tram4)

County Business Patterns County Business Patterns

Number of debt collectors Share of debt collectors employed by establishments with fewer than 10 employees Data Data Data Data Data Data Data Data

State session laws State session laws State session laws

Index of debt collection restrictions Bankruptcy exemptions Restrictions on garnishment

of new revolving lines of credit balance of new revolving lines of credit of new bank auto loans of new mortgage loans of revolving borrowers 90 days or more past due number of credit inquiries over 180 days TransUnion credit score of borrowers with TransUnion credit scores below 700

Source

VARIABLES

Table 1: Variables Construction

43

572 572 572 572 241 572 572 572 572 572 572 572 572 572 572 572 556 572 572 572 543 481

Index of debt collection restrictions Combined bankruptcy exemption ($10,000) Amount exempt from garnishment ($100 per week) Debt collectors per million people Share of debt collectors employed by establishments with fewer than 10 employees (%) Number of new revolving lines of credit per thousand consumers Average balance of new revolving lines of credit Number of new bank auto loans per thousand consumers Number of new mortgage loans per thousand consumers Number of revolving borrowers 90 days or more past due per thousand borrowers Average number of credit inquiries over 180 days Average TransUnion credit score Share of borrowers with TransUnion credit scores below 700 (%) Average recovery rate on charged-off unsecured credit card loans by credit unions (%) Average interest rate charged by credit unions on unsecured credit card loans (%) Number of credit union members per thousand people Average recovery rate on charged-off unsecured credit card loans by small banks (%) Real personal income per capita ($1,000) Growth rate of real personal income per capita (in %) Unemployment rate (%) High school graduation rate (%) Infant mortality rate (per 1,000 live births)

3.42 19.60 3.42 423.81 11.26 119.41 1953.53 6.91 9.20 15.49 105.30 660.12 53.50 11.41 10.71 306.14 16.93 38.42 0.89 5.93 77.15 6.79

Mean 4.00 4.97 2.00 389.51 9.12 119.84 1755.56 6.53 8.38 13.63 103.36 664.15 52.38 10.80 10.72 287.22 15.86 37.08 1.00 5.40 78.00 6.78

Median

1.98 33.03 4.06 217.74 7.71 24.66 782.85 2.59 4.16 6.64 24.23 22.89 6.34 4.63 1.10 120.13 9.96 5.80 2.44 2.12 7.57 1.39

SD

All variables are as described in the text and in Table 1. The main sample includes 572 observations for 48 states from 2000 to 2012, with 29 state-years excluded because of midyear law changes and another 23 state-years excluded because of missing data on debt collectors. Some variables have fewer than 572 observations because of additional missing data for those variables.

N

VARIABLES

Table 2: Summary Statistics

44 YES YES

-2.112*** (0.687) -1.018** (0.478) -0.869*** (0.303) 0.164 (0.178) 0.278 (0.176) 0.242 (0.162) 0.086*** (0.029) 0.680*** (0.136) -0.420 (0.599) -0.043 (0.074)

(2)

572 572 0.959 0.967 Standard errors clustered by state in parentheses *** p