Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
NEW INSIGHTS IN THE DETERMINANTS OF REGIONAL VARIATION IN BANKRUPTCY FILING RATES *
Kelly D. Edmiston† Senior Economist Community Affairs Department Federal Reserve Bank of Kansas City 925 Grand Boulevard Kansas City, MO 64198-0001 USA November, 2005 Abstract Nonbusiness bankruptcy filing rates have increased very rapidly over the last couple of decades. In 1980, roughly 15 of every 10,000 Americans filed for bankruptcy protection. By 2004, that number had reached 54 of every 10,000 Americans. These alarming increases in bankruptcy filing rates over the last decade were largely the impetus for the Bankruptcy Abuse Prevention and Consumer Protection Act, which went into effect in October, 2005. A substantial literature already exists that seeks to determine the causes of personal bankruptcy, but critical holes in the literature remain. In particular, existing studies offer only weak inferences about the role of stigma in explaining the decision to file for bankruptcy or in explaining regional variation in bankruptcy filing rates. I enhance the existing literature by using innovative approaches to measuring the effects of age and geography, traditional proxies for stigma, and by utilizing a novel proxy for stigma, namely, religious adherence. There is also a lack of consensus on the effects of gambling on bankruptcy, with most research finding no statistically significant relationships. I utilize a unique measure of proximity to gambling establishments and subsequently find more definitive results. The existing literature lacks consensus on the effects of homestead exemptions as well, with some finding positive effects, some finding negative effects, and still others finding no effects. I assert that the explanation of these inconsistent results may lie in endogeneity, and therefore I estimate two-stage models that effectively instrument for homestead exemptions. In addition, I explore the effects on bankruptcy filing rates of factors generally left out of existing studies, including small business and self-employment, a full distribution of age and income, more narrow demographic definitions, disability, lack of health insurance, public assistance, housing and vehicle choices, and additional information on debts and debt service. JEL Codes: D1, R2
* I would like to thank Kate Fisher for extremely valuable research assistance. The views expressed in this paper are those of the author and do not necessarily reflect the views of the Federal Reserve Bank of Kansas City or the Federal Reserve System. † Tel: (816) 881-2004; Fax: (816) 881-2704; E-Mail: [email protected].
NEW INSIGHTS IN THE DETERMINANTS OF REGIONAL VARIATION IN BANKRUPTCY FILING RATES
1. Introduction
Nonbusiness bankruptcy filing rates have increased very rapidly over the last couple of
decades. In 1980, roughly 15 of every 10,000 Americans filed for bankruptcy protection (Figure
1).1 By 2004, that number had reached 54 of every 10,000 Americans, for a compound annual
growth rate of 5.6 percent. These alarming increases in bankruptcy filing rates over the last
decade were largely the impetus for the Bankruptcy Abuse Prevention and Consumer Protection
Act, which went into effect in October, 2005.
From 1960 – 1979, nonbusiness bankruptcy filing rates increased at a much slower 2.3
percent annual pace. Bankruptcy reform legislation passed in 1978 often is given the blame for
the acceleration in growth of filing rates in recent decades. However, passage of the legislation
should have lead to a one-time jump in the bankruptcy filing rate (a change in intercept in Figure
1) rather than an increase in the growth rate of bankruptcy filings (a change in slope in Figure 1).
More likely is that a combination of several factors is responsible for accelerating growth in
bankruptcy filing rates in the last couple of decades. Indeed, Congress expressed its concerns
about the underlying causes of bankruptcy in developing the 2005 law.
While bankruptcy filing rates have climbed consistently in the aggregate over time, there
is considerable variation across regions (Figure 2). In 2000, for example, non-business
bankruptcy filing rates in counties with over 50,000 population ranged from only 0.2 per 10,000
people in Bibb County, GA (Macon) to 155 per 10,000 people in Shelby County, TN (Memphis).
The lack of uniformity in bankruptcy filing rates across counties (and states) suggests that
regional factors likely play a key role in the determination of filing rates.
1 Data on bankruptcy filing rates were available from two sources that are unfortunately inconsistent with each other. Both series are presented in Figure 1. I consider the data from the Administrative Office of the U.S. Courts (with Census Bureau population data) to be more reliable and therefore focus on these data when possible.
DRAFT. Please do not distribute without permission from the author.
2
A substantial literature already exists that seeks to determine the causes of personal
bankruptcy, but critical holes in the literature remain. In particular, existing studies offer only
weak inferences about the role of stigma in explaining the decision to file for bankruptcy or in
explaining regional variation in bankruptcy filing rates. I enhance the existing literature by using
innovative approaches to measuring the effects of age and geography, traditional proxies for
stigma, and by utilizing a novel proxy for stigma, namely, religious adherence. There is also a
lack of consensus on the effects of gambling on bankruptcy, with most research finding no
statistically significant relationships. I utilize a unique measure of proximity to gambling
establishments and subsequently find more definitive results. The existing literature lacks
consensus on the effects of homestead exemptions as well, with some finding positive effects,
some finding negative effects, and still others finding no effects. I assert that the explanation of
these inconsistent results may lie in endogeneity, and therefore I estimate two-stage models that
effectively instrument for homestead exemptions. In addition, I explore the effects on
bankruptcy filing rates of factors generally left out of existing studies, including small business
and self-employment, a full distribution of age and income, more narrow demographic
definitions, disability, lack of health insurance, public assistance, housing and vehicle choices,
and additional information on debts and debt service.
Many of the existing studies use data on individuals in an effort to examine the individual
decision to file for bankruptcy. Others utilize geographically aggregated data, such as county-
level or state-level data, to explore the reasons why bankruptcy filing rates differ across the
country. This study falls in the latter category. Specifically, I use a cross-section of all counties
in the United States for the year 2000. Panel data on bankruptcy filings are available through
2003; however, many of the variables included in this analysis are available only for 2000 or are
available only every ten years (including 2000). The goal here is to be as comprehensive as
possible in choosing independent regressors, and hence the benefits of using a large number of
regressors, which is possible only in a cross-section, were deemed to exceed the added benefit of
time-series variation and additional degrees of freedom.
DRAFT. Please do not distribute without permission from the author.
3
2. Model and Data
The basic model employed in this study is one of bankruptcy filing rates per 10,000
households. The units of observation are U.S. counties and data are from the year 2000.
Numerous potential determinants are included in the model and are described below. The
emphasis of the study is on the effects on bankruptcy filing rates of asset exemptions, wage
garnishment laws, proximity to gambling establishments, and proxies for social stigma, but
numerous other regressors are included in the model in an effort to make it comprehensive. Data
descriptions, sources, and simple statistics are presented in Table 1.
Exemption Levels
Each state has established a homestead exemption (EH, which in some cases is zero),
which limits the amount of home equity that must be used to pay unsecured debt under
bankruptcy.2 For a house valued at V, the total amount of home equity to be allocated to pay
unsecured debt would then be V – EH. In the case of a mortgage foreclosure, proceeds from the
sale of the house would be allocated (until exhausted) first to the transactions costs associated
with the foreclosure, then to the first mortgage, then to the second mortgage. Of any remainder,
only the amount in excess of EH would be used to pay unsecured debt.
Homestead exemptions allow bankruptcy filers to have higher post-bankruptcy
consumption than would be the case in the absence of exemptions. Because non-exempt assets
(such as deposit accounts) are easily converted to home equity (e.g., by selling assets and paying
off mortgage debt), high homestead exemptions offer protection for all kinds of wealth, not just
actual homesteads. Empirical evidence suggests that some high asset households move to high
(homestead) exemption states to make such a conversion of assets (Brinig and Buckley, 1996; 2 The homestead exemption generally would apply in a Chapter 7 bankruptcy. However, if there is such significant equity in the house that the creditors of the Debtor would be better off if the Debtor filed a Chapter 7 rather than a Chapter 13, the Debtor could argue that because of his entitlement to a homestead exemption, the equity is of inconsequential value. Further, if the Debtor has equity in his home, sells the home during the case, and the proceeds are required to be paid into his Chapter 13 plan, the Debtor would be entitled to assert his homestead exemption in the proceeds ahead of the Chapter 13 estate's interest.
DRAFT. Please do not distribute without permission from the author.
4
Elul and Subramanian, 2002). Homestead exemptions lower the cost of filing for bankruptcy,
and therefore higher exemption levels should be expected to lead to higher filing rates, all else
equal.
Most states include numerous other asset exemptions besides homestead exemptions.
These are impossible to value in any uniform way across states because the exemptions generally
are for specific items, such as wedding rings, livestock, and clothing. Some states provide both
specific property exemptions and dollar value limits for these items, however, and Dawsey and
Ausubel (2001) take advantage of this information in an attempt to get around the personal
property exemption valuation problem. They use the value limits provided in a limited number
of states to value personal property exemptions in all states. Despite this clever approach,
Dawsey and Ausubel are unable to get “consistent results” for personal property exemptions and
therefore “follow previous studies and focus on the homestead exemption” (17). Personal
property exemptions are not included in the present model because of a lack of confidence in the
valuation. Further, there is some tendency for personal property exemptions and real property
exemptions to move together as one surveys the states. The correlation coefficient in the
Dawsey and Ausubel data is 0.241.
Empirical evidence on the relationship between homestead exemption levels and
bankruptcy filing rates is somewhat limited and mixed in its conclusions.
Peterson and Aoki (1984) created dummy variables based on the generosity of state
exemptions in 1980, representing the following categories: very high, high, low, and very low.
“Low” states had exemptions that were more restrictive than federal exemptions (under the 1978
law), while “high” states had exemptions that were more generous than federal exemptions. The
left out category was for states that did not have their own exemptions, but rather relied on
federal exemption standards.
Surprisingly, Peterson and Aoki found positive coefficients for all of the dummy
variables, three of which were significant, meaning that having state exemption rules increased
bankruptcy filing rates in 1980, regardless of whether the state exemptions were more or less
DRAFT. Please do not distribute without permission from the author.
5
restrictive than the federal standards. The authors suggest that this result may reflect a reverse
causality in that the over-riding of the federal law with more restrictive exemptions was likely a
response to especially elevated bankruptcy filing rates, a notion that is supported by the data.
That is, states with higher bankruptcy filing rates were more likely to impose more restrictive
exemption rules. The positive coefficient on high exemptions can be explained by would-be
filers responding to reduced costs of filing.
Apilado et al. (1978) and Shiers and Williamson (1987) both produce the unusual finding
that higher exemption levels lead to fewer bankruptcies. The explanation given by Shiers and
Williamson is that the amount of credit issued to high risk borrowers is lower in states with high
exemptions. In point of fact, Gropp et al. (1997) find that the size of a state’s bankruptcy
homestead exemption has a (statistically and economically) significant positive effect on the
probability that potential borrowers in the state will be denied credit. Of course, if the results of
Apilado et al. and Shiers and Williamson are reflective of the true relationship between asset
exemptions and bankruptcy filings, the effect of restrictive credit practices would have to more
than offset the stimulus to bankruptcies of lower filing costs. Another possible explanation for a
negative relationship between the level of asset exemptions and bankruptcy filing rates is that the
direction of causality goes the other way. It may well be that states with relatively high
bankruptcy filing rates keep exemptions low to keep the incentives to file for bankruptcy low and
that states with relatively low filing rates are more inclined to be generous with exemptions.
White (1987) finds a positive relationship between homestead exemption levels and
Chapter 7 filing rates, but a negative effect of homestead exemptions on Chapter 13 filings.
Using individual data on credit card holders, Agarwal et al. (2003) find a positive relationship
between homestead exemptions and the decision to file for bankruptcy. In some states,
homestead and personal property exemptions apply uniformly in and outside of formal
bankruptcy (see Woodward, 1982), which may diminish the relationship between homestead
exemptions and bankruptcy.
DRAFT. Please do not distribute without permission from the author.
6
Lin and White (2001) argue that since few states change their exemption levels each year
(11 over the period 1992 – 1997), “individual states’ bankruptcy exemptions can be treated as
exogenous in the model” (151). Supporting this view, in cross-sectional regressions using 1996
data, Posner et al. 2001 were able to explain 23 percent to 68 percent of the variation in
exemption laws across states, depending on the specification of the dependent variable. But the
only variable of any significance in the regressions was “history,” defined as the homestead
exemption in that state in 1920. This result changed little using pooled regressions over the
period 1975 – 1996. Nevertheless, given the potential duality in the direction of causality, it may
be important to instrument for exemptions, which I accomplish with two-stage models that
utilize both county and state-level data.
Information on homestead exemptions (for a married couple) were collected from a
search of the bankruptcy codes for all 50 states and the District of Columbia relevant for the year
2000. A variable was then created for each county to reflect the homestead exemption in its state
(or its predicted value) as a fraction of the median house value in the county. Six states (Florida,
Iowa, Kansas, North Dakota, South Dakota, and Texas) had exemptions that were unlimited in
value (but limited in acreage, e.g., the exemption might be for 40 acres and improvements). The
homestead exemptions in these states are given a value of $1,000,000, which should cover most
homesteads given that bankruptcy filers come predominantly from the bottom half of the income
distribution in their states (White, 2005). Further, a dummy variable (HSD) is created to
represent these states, taking on a value of unity for states with unlimited exemptions and zero
otherwise. For other states, the value of the homestead exemption in 2000 (the data year for this
study) ranged from $2,500 (Maryland) to $200,000 (Minnesota), with the mean exemption of
states with limited exemptions of $43,967.3 In 2005, the average exemption, excluding states
with unlimited exemptions, is $63,527 and ranges from $5,000 (Alabama, Kentucky, and
Maryland) to $500,000 (Massachusetts).
3 This average includes the value of the Federal homestead exemption ($15,000 in 2000) for states where the state exemption is less than the federal exemption, but petitioners have the option of taking the federal exemption.
DRAFT. Please do not distribute without permission from the author.
7
Social Stigma
Declining social stigma often has been put forth as a reason why bankruptcies have
increased at such a rapid rate in the United States over the last several years. In testimony before
Congress, for example, Federal Reserve Chairman Alan Greenspan argued that “personal
bankruptcies are soaring because Americans have lost their sense of shame.”4 The effects of
social stigma on bankruptcy filing propensities are probably impossible to estimate empirically
in any direct way because stigma cannot be directly measured, but the effects often can be
inferred from other empirical results.
Livshits et al. (2005) use an applied general equilibrium to get around the measurement
problem, incorporating a utility cost to bankruptcy in their model. They find that greater stigma
reduces the number of filings, but increases borrowing ex ante and generates “much more
desperate bankrupts” (28). Gross and Souleles (2002) use a panel of individual credit card
accounts to investigate credit card delinquency and the decision to file for bankruptcy. They find
that a credit card holder in 1997 was almost one percentage point more likely to file for
bankruptcy than an individual with similar characteristics in 1995, a result they compare to the
entire population of card holders becoming one standard deviation riskier. Gross and Souleles
suggest that the increased demand for bankruptcy may be attributable to a decline in social
stigma, but they give equal validity to the idea that information costs may have declined.
To the extent that social stigma varies regionally, geographic variables will at least partly
pick up the differences. This is essentially the approach used in Fay et al. (2002). The study
utilizes data on individuals from the PSID, and the authors suggest that the lagged bankruptcy
rate in the state in which a household resides may be an inverse proxy for stigma experienced by
that household. Their results suggest that the probability of filing for bankruptcy decreases with
the degree of social stigma. The authors also note that localization of bankruptcy filing might
4 Cited in Julie Kosterlitz, 1997 (May 3), “'Over the Edge,”' National Journal, 29(18), 871.
DRAFT. Please do not distribute without permission from the author.
8
reflect information flow: “[t]heir friends and relatives will probably tell them that filing for
bankruptcy is quick and easy” (710).
In the present study a number of measures of stigma are used with the hope that the
combination of results will lead to a more definitive answer to the stigma question.
Geographic variables, or more generally, neighborhood effects, are used to explain
regional patterns in bankruptcy filing rates, which at least in part should pick up the effects of
stigma. The simplest way this is accomplished is by including a set of dummy variables for the
eight regions defined by the Bureau of Economic Analysis. A second way is by including a
spatial autoregressive term, a procedure explained in some detail below. As shown in Figure 2,
there are strong patterns in bankruptcy filing rates across counties throughout the United States,
with especially high concentrations in the Southeast and lower Midwest.
Another way I pick up stigma is by including a set of variables reflecting adherence to
various religious groups within each county, under the assumption that different religions have
different views on the acceptability of defaulting on debt payments and filing for bankruptcy.
For example, the repayment of debts is a paramount moral obligation in Islam.5 Further, Islam is
commonly thought to prohibit the charging or paying of interest (see El-Gamal, 2001), which
likely results in less borrowing, and hence fewer bankruptcies. The repayment of debts is a
virtue in the Judeo-Christian tradition as well,6 but at the same time, perhaps the most important
tenet of the Christian religion is forgiveness.7 Similarly, Jewish tradition, as outlined by the Old
Testament, insists on the forgiveness of debts every seven years (Deuteronomy 15:1-2). This
suggests that bankruptcy, while subject to some stigma in the Judeo-Christian traditions, may not
receive the same condemnation by Jews and Christians as by Muslims. A central theme in
many Eastern religions is the avoidance of injury to others. Thus, there is likely to be stigma
attached to any injurious action, including bankruptcy and defaulting on debts. Further, in most 5 Consider the following verses, for example: “O you who believe! Fulfill (your) obligations” (Al-Maa’idah 5:1) and “Verily! Allah commands that you should render back the trusts to whom they are due . . . ” (An-Nisaa 4:58). 6 Consider, for example, the passage from Psalms 37:21 (KJV): “The wicked borroweth, and payeth not again.” 7 The Lord’s Prayer entreats God to “forgive us our debts, as we forgive our debtors” (Matthew 6:12, KJV, emphasis added).
DRAFT. Please do not distribute without permission from the author.
9
Eastern religions, karma has a critical place: all deeds have consequences, or “you reap what you
sew.” A default on debts would likely generate bad karma that would have to be paid in either
this life or a future life. The model includes as one measure of stigma the share of the population
in each county that adheres to Catholicism, an Eastern religion such as Hinduism or Buddhism,
Islam, Judaism, Mormonism (LDS), Orthodox Christianity, Unitarian Universalism, and
Protestantism.
A final measure of stigma included in the model is the age distribution. VISA (1996)
argues that younger age cohorts may view bankruptcy with less stigma than older groups, and
they find a statistically significant positive relationship between bankruptcy filing rates and the
share of the population that is 25 – 44, which supports this view. Flynn and Bermant (2003)
show 25 – 44 to be the “peak years” for filing for bankruptcy as well, and Domowitz and Eovaldi
(1993) also find a positive association between bankruptcy filing rates and the proportion of the
population that is 25 – 44. I include the entire age distribution of each county in the analysis in
part to pick up the effects of stigma.
Of course, there are other ways that age distribution might influence bankruptcy filing
rates as well, including by reflecting income and consumption patterns (VISA, 1996). Although
traditionally younger people have made the bulk of the bankruptcy petitions, the pool is
becoming increasingly older, and middle-aged people make up a significant part of the increase
in filings over the last couple of decades (Figure 3). Fay et al. (2002) find that the probability of
filing for bankruptcy increases with the age of the head-of-household but decreases with squared
age. They suggest that bankruptcy filings first pick up with age because access to credit
increases with age, but the effect of age reverses course later in life when households often have
accumulated wealth and do not need as much access to credit. White (1987) did not find a
significant relationship between bankruptcy and the proportion of the population that is elderly.
Wage Garnishments
When an individual or married couple files for bankruptcy protection, future earnings are
protected from garnishment. The idea behind this exemption is to give bankruptcy filers a “fresh
DRAFT. Please do not distribute without permission from the author.
10
start” and to provide proper work incentives for filers following discharge. Because of this,
wage garnishment often spurs bankruptcy filings. Naturally, the lower the amount of an
individual’s wages that can be legally garnished, the less is his incentive to file for bankruptcy
protection. Hence, bankruptcy filing rates, all else equal, are expected to be negatively related to
the proportion of wages protected from garnished. In this study, we account for wage
garnishment by including in the model the fraction of state household median income protected
from garnishment, keeping in mind that all but the wealthiest working-age individuals derive
virtually all of their income from wages and salaries (Burman and Kobes, 2003).8
Federal law sets a floor on monies collected via wage garnishment. In all states, the
District of Columbia, and U.S. territories, an employer may not garnish anymore than the lesser
of 25 percent of an employee’s earnings or the amount by which an employee’s earnings exceed
30 times the federal minimum wage, which in 2000 was $5.15 per hour. Some states provide for
even greater protection of employee’s wages. For example, Texas, North Carolina, and South
Carolina exempt all wages from garnishment, while Delaware, Illinois, New York, Pennsylvania,
and West Virginia allow individuals to protect 80 percent or more of their disposable income
from garnishment.
Surprisingly, Ararwal et al. (2003) find that the level of garnishment protection has a
negative effect on the probability of a credit card holder becoming delinquent, but they do not
find a statistically significant relationship between garnishment levels and the probability of
formally filing for bankruptcy. Apilado, et al. ( 1978), Sullivan and Worden (1990), and Dawsey
and Ausubel (2001), however, all find a negative association between personal bankruptcies and
the share of income protected from garnishment.
8 In 2000, the year of data used in this study, when capital gains were at an all time high during the late 1990s stock market boom, those making less than $100,000 earned 70 – 80 percent of their income from wages and salaries, with the bulk of the remainder coming from IRA distributions and social security. Thus for working age people in that income class, virtually all income comes in the form of wages and salaries.
DRAFT. Please do not distribute without permission from the author.
11
Legal Gambling
Legalized gambling has grown rapidly in the United States over the last two decades,
spurred largely by the introduction of gaming on Indian reservations in the 1980s. State law at
the time severely restricted most types of gambling, although exemptions were often granted for
state-sponsored gambling (i.e., lotteries) and charitable events (like “casino nights”). Indian
tribes believed that gambling could not be restricted or regulated on their reservations because
state laws generally do not apply to Indian tribes or Indian people within their reservations. The
exception is when the applicability of state laws is expressly authorization by Congress, and
Public Law 280 (1953) allows for state criminal laws to be enforced on reservations in some
states. There was some initial debate about the applicability of state gambling laws under Public
Law 280. A U.S. Supreme Court ruling on the issue (Cabazon v. California, 1987) sided largely
with Indian tribes and subsequently opened the door to widespread Indian gaming.9 Today there
are over 300 Indian casinos across the United States.
Commercial gaming expanded at about the same time as Indian gaming. South Dakota
legalized casino gambling in the historic town of Deadwood in 1989, and the State of Iowa
quickly followed with the first riverboat casinos. The following year, riverboat gambling was
legalized in Illinois and Mississippi. Today eleven states offer either land-based (6) or “floating”
(6) commercial casino gaming.
Casino gaming in some form (Indian or commercial) is allowed in 33 states. In 1999,
approximately $769 billion was wagered in the United States, encompassing all forms of
commercial gaming, leading to gambling losses to consumers (or revenues to the gaming
establishments) of roughly $59.4 billion, or about 0.64 percent of gross domestic product
9 The ruling essentially stated that if states allow any type of gaming, they must allow similar types of gaming on Indian reservations. My interpretation of the basic argument is that in cases where gaming is allowed in some form in a state, restrictions on Indian gaming amount to regulation, while in cases where all types of gaming are illegal, restrictions on Indian gaming are a matter of enforcing state criminal laws, in which case Public Law 280 would seem to apply. The current regulatory structure for Indian gaming is largely embodied in the Indian Gaming Regulatory Act (P.L. 100-497, codified at 25 U.S.C. 2701).
DRAFT. Please do not distribute without permission from the author.
12
(GDP).10 By 2003, consumer gambling losses increased to $72.9 billion, suggesting that the total
wager in that year was around $962 billion,11 an increase of over 25 percent in four years.
As shown in Figure 4, there is substantial variation in the geographic pattern and
concentration of private and Indian casinos across the United States. The heaviest concentrations
of Indian casinos are, unsurprisingly, in the upper Midwest, on the West coast, and in the State of
Oklahoma. The largest concentrations of private casinos are in Atlantic City, Deadwood, SD,
central Colorado, throughout the State of Nevada, and along the Mississippi river. Very little
gaming of any type is found in the eastern third of the country.
The mechanism whereby gambling would lead to a greater likelihood of filing for
bankruptcy is by raising debt levels relative to income. In one of the first formal studies to look
at the relationship between bankruptcy and gambling, Hira (1998) found that average annual net
income is about a third lower for gamblers than for non-gamblers and that the mean total debt of
gamblers is about 19 percent higher than the mean total debt of non-gamblers.
Problem gambling has been shown to increase with the availability of gambling
opportunities. The National Opinion Research Center (NORC) (1999), in a combined patron and
telephone survey, found that the prevalence of “problem and pathological gambling” roughly
doubles within 50 miles of a casino (vs. within 50 and 250 miles) (see also Goodman, 1995)
(ix).12 If bankruptcy is associated with problem gambling, therefore, bankruptcy also should be
associated with proximity to gambling establishments. The NORC report revealed that 19.2
percent of “pathological gamblers” have filed for bankruptcy at least once, compared to 5.5
percent for “low-risk gamblers” and 4.2 percent for non-gamblers. Granted, pathological
gamblers have personal characteristics that make them more likely to file for bankruptcy than
non-gamblers even ignoring their gambling problem, but given those characteristics, the
10 Christiansen Capital Advisors, LLC, “The Gross Annual Wager of the United States 2003.” 11 The total wager value for 2003 was estimated by the author from the 2003 value for consumer gambling losses provided by Christiansen Capital Advisors, LLC. 12 The use of the terms “problem gambling” and “pathological gambling” reflect specifically the criteria set by the American Psychiatric Association. For criteria, see http://www.ncpgambling.org/about_problem/ about_problem_timeline.asp.
DRAFT. Please do not distribute without permission from the author.
13
expected share who have ever filed for bankruptcy was only 10.8 percent, significantly less than
the actual 19.2 percent figure for pathological gamblers. In a study of gambling treatment in the
State of Minnesota, Stinchfield and Winters (1996) found that 21 percent of 944 gamblers
admitted for treatment had filed for bankruptcy.
Empirical results from the existing literature on the relationship between gambling and
personal bankruptcy are mixed. Results from state-level time series regressions for New Jersey,
Mississippi, and Nevada conducted by the U.S. Department of the Treasury (1999) suggest that
“gambling has no measurable impact on statewide bankruptcy rates” (54). The Treasury study
also found no impact when estimating panel regression using county-level data . This result was
echoed in the work of de la Viña and Bernstein (2002), who in a study examining 100 counties in
36 states find no statistically significant evidence that bankruptcy filing rates are higher in casino
and/or pari-mutual betting counties vis-à-vis non-gambling states. Using a sample of 398
counties from the casino states of Illinois, Iowa, Missouri, and Mississippi over a period of eight
years, Thalheimer and Ali (2004) also fail to show a significant relationship between access to
gambling establishments and personal bankruptcy filings. Their measure of access to gambling
for any given county was given by a gravity model: )031.0exp( d−=access , where d is the
distance from the geographic centroid of the county to the relevant gaming establishment, and
the decay parameter is based on the observation from another study (Illinois Gaming Board,
1997) that 21 percent of riverboat casino visits are made by those who live at least 50 miles
away.13
A limited number of studies have found a positive association between gambling and
personal bankruptcy filings. Nichols et al. (2000) compare bankruptcy filing rates in eight
casino counties in Illinois, Iowa, Mississippi, and Missouri with corresponding counties with
similar characteristics (by means of a cluster analysis) and find that the introduction of casino
gambling was associated with a significant increase in bankruptcy rates in five of the counties
13 The authors argue that this implies that accessibility declines from 100 percent to 21 percent as distance increases from 0 to 50 miles, yielding a 3.1 percent compound rate of decline.
DRAFT. Please do not distribute without permission from the author.
14
and a significant decrease in one county. There was no statistically significant relationship
between bankruptcy filing rates and the introduction of casino gambling in the other two
counties, but the relationship was positive. A study by Barron et al. (2002) suggests that the
elimination of casino gambling would have reduced bankruptcy filings by five percent in casino
counties in 1998 and would have reduced the national filing rate by about one percent.
The examination of the relationship between gambling and bankruptcy in this study
improves on the existing literature firstly by including a full set of Indian gaming establishments
while also including all counties in the United States. Thalheimer and Ali (2004) did include
Indian gaming locations in their analysis of 398 counties, but many Indian gaming
establishments were likely missed. For example, Figure 4 shows 44 Indian establishments
spread throughout Oklahoma in 2000, whereas Thalheimer and Ali show 0 for 1997.
Information from the Oklahoma Indian Affairs Commission suggests there were at least 10
compacts with Indian tribes by 1996, the earliest being executed in 1992.14 Further, the National
Indian Gaming Commission asserts there were 36 Indian gambling establishments in Oklahoma
in 1997.15 Thalheimer and Ali likely were considering only class III casinos, which provide all
types of gambling and were not legalized in Oklahoma until 2005. However, the establishments
existing in 1997 (and 2000) provided “fast-style class II gaming,” in which substantial amounts
of money can be wagered.16 If all available gambling establishments are not included in the
analysis, the estimate of the relationship between gambling and bankruptcy rates would be biased
towards zero.
A second improvement upon the existing literature is an accounting for all forms of
gambling. In addition to Indian and commercial casinos, dummy variables are included to reflect
the legality of card rooms and race tracks and the presence of a state lottery.
14 http://www.state.ok.us/~oiac/gaming.htm 15 Personal correspondence on September 14, 2005. 16 Ibid.
DRAFT. Please do not distribute without permission from the author.
15
Finally, a novel formulation of accessibility to casino gambling is included in the study.
Specifically, the variable used to pick up proximity to casinos is for each county the minimum
distance from the geographic center of that county to the geographic center of a casino-
containing zip code. This distance is determined by calculating the distance from the population
center of each county to the center of each zip code that contains at least one casino, and then
taking the minimum of all of those distances for each county. This provides a measure that
allows accessibility to change as distance varies, and the assumption is that the relationship
decays in a linear fashion as distance increases. Given mixed results in the past, it is important to
bring another formulation to bear on the issue.
Entrepreneurship and Small Business
About 70 – 80 percent of new businesses fail in the first year, and of the remainder, half
fail within five years. Given these high failure rates, one might expect that entrepreneurs have
relatively high bankruptcy rates. In calendar year 2004, the latest date for which complete data
are available, only 34,317 of approximately 1.6 million bankruptcy filings, or less than 2.2
percent, were deemed to be business filings by the U.S. Bankruptcy Court.17 However, Lawless
and Warren (2005) find that these official reports significantly understate the actual number of
business filings because self-employed people, small business owners, and independent
contractors are likely to file for personal bankruptcy protection if they have a failed business
undertaking. Official reports put business filings at 2.3 percent of total filings in 2003. In
reality, as much as 18.6 percent of all bankruptcy filings in 2003 may have been due to business
failures (the actual figure is more likely around 13.5 percent). Thus, a significant number of
personal bankruptcy filings (as much as 19 percent) are business filings, and an examination of
the determinants of personal bankruptcy filing rates should account for this.
Fay et al. (2002) find that an individual owning his own business is not any more or less
likely to file for bankruptcy than individuals who do not own their own businesses. As less
17 Data is available at http://www.uscourts.gov/bnkrpctystats/bankrupt_f2table_dec2004.pdf (accessed August 23, 2005).
DRAFT. Please do not distribute without permission from the author.
16
direct evidence of a relationship between entrepreneurship and personal bankruptcy filing rates,
Mathur (2005) finds that the decision to start a new business is positively related to own-state
homestead exemptions and negatively related to neighboring states’ homestead exemptions,
while the decision to close a small business is negatively related to own-state homestead
exemptions and positively related to neighboring states’ homestead exemptions (see also Fan and
White, 2003). The findings suggest that homestead exemptions are an important consideration in
the business start-up/closure decision, and hence also suggest that bankruptcy may be a factor.
I include in the model the fraction of total employment that is made up of self-employed
people. If the hypothesis above is correct, higher self-employment rates, should be associated
with higher bankruptcy filing rates. I also include net firm births, which is total firm births less
total firm deaths, per 10,000 residents as a measure of the environment for entrepreneurship in
local areas. A high number of firm births relative to firm deaths is indicative of an environment
conducive to entrepreneurship and small business and should be associated with lower
bankruptcy filing rates. Net firm births may also reflect the general business climate.
Economic and Social Factors and Healthcare
VISA (1996) finds employment growth rates to be “the single most powerful coefficient
explaining bankruptcy behavior” (13), although the unemployment rate does not have much
predictive value. White (1987) finds that a one percent increase in the county unemployment
rate leads to an increase in Chapter 7 bankruptcy filings of approximately one-half of one
percent, while higher unemployment rates, surprisingly, seem to lower the rate of Chapter 13
filings. In general results seem to be reversed in Chapter 7 and 13 filings in White’s study,
which suggests that the variables affect chapter choice in addition to the decision to file for
bankruptcy. Numerous others have found a positive relationship between nonbusiness
bankruptcy filing rates and the unemployment rate (see, e.g., Domowitz and Eovaldi,1993;
Dawsey and Ausubel, 2001; Agarwal et al., 2003).
Income has been shown to be a critical determinant of bankruptcy in a number of studies,
but the effects are mixed and depend on how income enters the model. Agarwal et al. (2003) and
DRAFT. Please do not distribute without permission from the author.
17
Fay et al. (2002) find a negative relationship between income and bankruptcy, while Dawsey and
Ausubel (2001) find a positive relationship. Domowitz and Sartain (1999) find income not to be
a significant determinant of the probability of filing for bankruptcy, except when combined with
home ownership.
In 1995, for example, 32 percent of U.S. households in the lowest income decile had
credit cards (Yoo, 1998). That number increased steadily to 99 percent for the highest income
decile. The percentage for the U.S. as a whole was 75 percent. Likewise, average credit card
balances (for households with credit cards) increased steadily from $1,358 in the lowest income
decile to $2,551 (roughly 90 percent higher) for the highest income decile. Of course, given that
average income in the top decile is ten times higher than average income in the bottom decile,
lower income people, on average, have larger credit card balances relative to income.
Another potentially important economic factor is the share of the population receiving
public assistance. People receiving public assistance are likely to have much less access to credit
than others, which should be reflected in lower bankruptcy filing rates. Further, the existence of
financial support during times of hardship may reduce bankruptcy filings on the margin.
Domowitz and Eovaldi (1993) find a statistically significant negative relationship between per
capita transfer payments and bankruptcy filing rates.
Hospitals and doctors provide about $45 billion in uncompensated care each year, much
of which is bad debt.18 HCA, the largest hospital operator in the U.S., for example, reserves 12
percent of revenues for bad debt. Tenet Healthcare Corp., the second largest operator, reserves
11 percent. Unsurprisingly, these bad debts often arise from bankruptcy filings.
Warren et al. (2000) evaluate responses from 1,974 survey participants and find that one
out of four debtors identified injury or illness as a reason for filing for bankruptcy. One third (in
total) stated that they had substantial medical bills, defined as $1,000 or more in medical bills not
18 See “More Americans May Have Lost Medical Coverage for Fourth Year,” Bloomberg.com, August 30, 2005. Accessed September 30, 2005 at http://www.bloomberg.com/apps/news?pid=10000103&sid= aQNqUDzLGkRY&refer=us#.
DRAFT. Please do not distribute without permission from the author.
18
covered by insurance. A more recent study by Himmelstein et al. (2005) has very similar
findings. Like the Warren et al. study, the Himmelstein et al. study finds that about one-quarter
of the 1,771 personal bankruptcy filers surveyed named injury or illness as a specific reason for
bankruptcy. An additional quarter were found to have uncovered medical bills of at least $1,000.
Domowitz and Sartain (1999) find “high medical debt” to have the “greatest single impact of any
household condition variables in raising the conditional probability of bankruptcy” (413).
Specifically, they find that households with high medical debt have a probability of filing for
bankruptcy greater than 28 times that of the baseline household.
In an effort to account for the role of medical costs in bankruptcy filing rates, I include in
the model the share of the county population without health insurance coverage. Agarwal et al.
(2003) found a positive association between the lack of health insurance coverage and the
probability of an individual filing for bankruptcy. Also included in the model is the share of the
population aged 21 – 64 that is disabled.
Credit and Debt
Clearly debt levels and access to credit are critical determinants of bankruptcy filing, as
bankruptcy implies liabilities in excess of assets. The proportion of debt that is unsecured is
significantly higher among bankruptcy filers than among the population in general (U.S.
Department of the Treasury, 1999; Sullivan et al., 1989). Revolving debt in particular,
especially credit card debt, seems to be positively correlated with bankruptcy filing (Dawsey and
Ausubel, 2001; Domowitz and Sartain, 1999). Existing research indicates that account balances
and installment debt relative to income also are positively associated with income (Agarwal et
al., 2003; VISA, 1996), although Domowitz and Eovaldi (1993) do not find a statistically
significant relationship between bankruptcy and debt relative to income.
Increased access to credit also has been found to be correlated with bankruptcy filing.
VISA (1996) finds that bankruptcy filing is increasing in the number of credit card accounts, but
they caution that the number of accounts per adult creates a redundancy with revolving debt as a
percent of consumer installment credit, which is also included in the model. On the other hand,
DRAFT. Please do not distribute without permission from the author.
19
Agarwal at al. (2001) finds that bankruptcy filings are decreasing in the overall credit limit.
They also produce the expected result that bankruptcy filings are negatively associated with
credit scores.
Homes and Autos
Conversations with attorneys suggest that excessive expenditures on autos are common
among bankruptcy filings.19 I include variables reflecting the average number of vehicles in a
household and the proportion of total debt that is auto debt, on average, in the county.
Domowitz and Sartain (1999) find that a debtor who is not a home owner is about seven
times more likely to file for bankruptcy than the average homeowner. White (1987) finds a
positive relationship between home ownership and Chapter 13 bankruptcy filings, but does not
reveal a statistically significant relationship between Chapter 7 filings and home ownership.
Mortgage debt may be used to finance consumer spending by refinancing and cashing out
equity. Consumers, if they choose the right lender, can in some cases finance 125 percent of the
value of their home. According to the Freddie Mac website, in the second quarter of 2005, the
latest date for which data are available, “74 percent of Freddie Mac-owned loans that were
refinanced resulted in new mortgages with loan amounts that were at least five percent higher
than the original mortgage balances.”20 Controlling for mortgage debt, as done in this study,
median home value serves as a good proxy for home equity, which should be negatively related
to bankruptcy filing rates, given that home equity can be tapped for emergency spending and that
negative equity often spurs bankruptcy. VISA (1996) finds such a negative relationship
empirically between median home prices and bankruptcy filing rates, but surprisingly, Domowitz
and Eovaldi (1993) do not find a statistically significant relationship between homeowners’
equity and bankruptcy filings.
19 As are excessive expenditures on furniture, but data on furniture expenditures or related debt are not available at the county, or even state levels. 20 Accessed September 13, 2005 at http://www.freddiemac.com/ news/archives/rates/2005/2qupb05.html.
DRAFT. Please do not distribute without permission from the author.
20
In this study we include not only median house value and mortgage debt (as a share of
total debt, which is also included in the model), but also owner-occupancy rate, the vacancy rate
of owner-occupied housing, and the share of owner-occupied dwellings carrying a mortgage. As
a measure of living expenses, we include median rent and median costs of owner-occupied
housing.
Demographics and Other Variables
As in most studies of bankruptcy, demographics variables are included in the models here
as additional controls. Existing literature suggests that these controls are important. White
(1987) finds a positive association between the proportion of the population that is black and
“Spanish” and Chapter 13 filings, but finds no relationship between percentage black and
Chapter 7 filings and a negative relationship among the share that is Spanish and Chapter 7
filings. In the White study it is difficult to distinguish effects of race and ethnicity on bankruptcy
filing and chapter choice; however, Domowitz et al. (1993) find a positive relationship between
bankruptcy filings and the proportion of the population that is nonwhite. I include a range of
race and ethnicity variables in my models to account for potential effects on bankruptcy filings.
Divorce often causes a substantial, immediate, and unanticipated reduction in income.
This is true for women in particular, who are estimated to suffer a 30 percent decline in
economic status in the first year following a divorce (Hoffman and Duncan, 1988).21 Numerous
studies of bankruptcy have found a statistically significant positive relationship between divorce
and bankruptcy (White, 1987; Shiers and Williamson, 1987; Fay et al., 2002). Domowitz and
Eovaldi (1993) found a positive but statistically insignificant relationship between divorce rates
and bankruptcy filing rates. Domowitz and Sartain (1999) find that people who are unmarried
have a higher probability of filing for bankruptcy than people who are married. I include in the
model the shares of the population in each county that are married, separated, divorced, and
widowed, with never married being the left out variable.
21 Economic status is defined as the ratio of average income relative to needs.
DRAFT. Please do not distribute without permission from the author.
21
3. Methodology and Empirical Results
Let bv
be a vector of bankruptcy filing rates, X represent the matrix of explanatory
variables discussed in Section 2, and θv
be a vector of parameters. The basic model (Model 1) is
then given by
(1) εθ vvv+= Xb
where ( ) Ω2σεε =′vvE . The model is solved using least squares and White’s (1980)
heteroscedasticity consistent estimator.
Exemptions
In Model 1, the coefficient on homestead exemption is statistically insignificant, which
suggests that homestead exemptions have no impact on bankruptcy filing rates (Table 2).
Alternatively, the zero coefficient may indicate either a dual causality or countervailing effects.
The presumption underlying the empirical model is that homestead exemptions increase
bankruptcy filing rates by reducing the cost of bankruptcy, which theoretically should produce a
positive coefficient. But it is possible that states with relatively high bankruptcy filing rates
respond by lowering the homestead exemption. If this is the case, the coefficient theoretically
should be negative, which means the two forces may cancel each other out. Further, since
creditors tend to restrict lending to risky individuals in states with generous exemptions (Gropp
et al., 1997; Grant, 2001; Berkowitz and White, 2004), high exemptions may lead to tighter
credit restrictions and hence less borrowing and fewer bankruptcies. In an effort to account for
this potential endogeneity, a two-stage model was also estimated.
The two-stage model begins with an estimation of homestead exemptions in the 50 states
and the District of Columbia. The dependent variable is the level of the homestead exemption in
2000, in dollars. Virtually all of the independent variables included in the model are statistically
significant at the 90 percent confidence level or better (Table 3). Poorer states, as measured by
median household income, the poverty rate, and the share of households receiving some form of
public assistance, tend to have lower homestead exemptions. States with populations that are
DRAFT. Please do not distribute without permission from the author.
22
older and states with larger families and greater shares of males and minorities tend to have more
generous exemptions. Surprisingly, both higher owner occupancy rates and home values tend to
lower exemption levels. The political tendencies of the state do not seem to have a significant
effect on the levels of homestead exemptions. These variables in total explained about 21
percent of the total variation in homestead exemption levels across states. Of course, as noted
above, states do not change exemption levels very often, and the most important determinant of
exemption levels in the Posner et al. (2001) study was the exemption level in 1920.
The predicted values for state homestead exemptions generated in the first stage were
utilized in the second stage. Specifically, for each county, the homestead exemption variable is
equal to the predicted value of the homestead exemption in the state in which it resides divided
by the actual median home value in the county. The values are restricted to be nonzero.
Because a generated regressor is used in the second stage, t-statistics are calculated using
bootstrapped standard errors, following Efron (1979) and as described in Greene (1993). Least
squares estimation of (1) results in an estimated parameter vector θ and a residual vector ε) :
(2) εˆ += θXb
A sample of B = 100 bootstrap estimates Bθθθ ~,,~,~21 L are then obtained by sampling n
observations from X, with replacement, and reestimating the model B times. The estimated
asymptotic covariance is then
(3) [ ][ ]∑ =
′−−
B
b bbB 1ˆ~ˆ~1 θθθθ
The two-stage model is presented in Table 2 as Model 2.
For most variables the results in Models 1 and 2 are virtually identical. For homestead
exemptions, however, the coefficient is now positive and significant at the 90 percent confidence
level. The value of 0.68 suggests that for every one percentage point increase in the share of the
median house value covered by the homestead exemption, bankruptcy filing rates increase by
0.68 per 10,000 households in the county. An exemption of 100 percent of median house value
would thus yield 35 additional bankruptcies per 10,000 households in that county relative to a
DRAFT. Please do not distribute without permission from the author.
23
county in which the state exemption is only 50 percent of the median house value. The average
county had roughly 100 personal bankruptcies per 10,000 households in 2000 (the median had
91), so this would represent a 35 percent increase in the bankruptcy filing rate.
Social Stigma
I attempt to get at the role of stigma in explaining variation in bankruptcy filing rates in
multiple ways. The first way is by accounting for regional tastes for bankruptcy. The inclusion
of a set of dummy variables representing the eight BEA regions reveals a marked regional
pattern (Table 2, Model 3).22 Relative to counties in New England, which is the left-out region
in the model, counties in the Mid-Atlantic, Great Lakes, Southeast, and Rocky Mountain regions
have significantly higher bankruptcy filing rates, all else equal. The bankruptcy filing rates in
counties in the Plains, Southwest, and West regions are not significantly different from those
with similar characteristics in the Northeast. This result suggests that bankruptcy filers may face
greater stigma in these regions and in the Northeast than in the rest of the country.
An additional way I account for geographic differences is with a spatial autoregressive
model. Let D be an NxN matrix where
(4) ( ) 2,/000,000,1, jidjiD = ,
D(i,j) is the row i column j element of D, N is the total number of counties, and di,j is the arc
distance between the geographic centroid of county i and the geographic centroid of county j.
Letting ρ be a scalar, the spatial autoregressive model is given by
(5) ( )IDX 2,0~, σεερθvvvvvv
Nbb ++=
(5) is solved for θv
and ρ by the method of maximum likelihood.
The parameter ρ is a measure of the degree that the bankruptcy filing rate in any county i
depends on the bankruptcy filing rates of all other counties, where the strength of the relationship
between filing rates in any two counties declines quadratic ally with the distance between them.
The spatial autoregressive term estimated in Model 4 (Table 2) is positive and statistically
22 Models 2 – 4 are all two-stage models.
DRAFT. Please do not distribute without permission from the author.
24
significant at the 99 percent confidence level, which is indicative of stigma effects: a given
county’s bankruptcy filing rate is higher the higher is its neighbors’ bankruptcy filing rates.
It has been suggested that relatively high bankruptcy filing rates among younger cohorts
is indicative of declining stigma, the assumption being that social mores are declining over time.
My results show that bankruptcy filing rates are lower the greater the proportion of the
population that is aged 55 – 64, which is consistent with this view. Inconsistent with this view,
however, is that bankruptcy filing rates increase with the share of the population that is aged 75 –
84. These results actually are more consistent with stigma being a generational phenomenon
rather than chronological, which is also consistent with the pattern shown in Figure 3.
Perhaps the best proxy for stigma is the data on religious adherence. My results (Models
1 – 4) suggest that bankruptcy filing rates are lower the greater the concentrations of Muslims,
Eastern religion adherents, and Unitarian Universalists, and bankruptcy filing rates are higher the
greater the concentrations of Jews and Christians, relative to the non-religious share of the
population. These results are entirely consistent with the emphasis on repayment of debts in
Islam, avoidance of injury to others in the Eastern religions, and forgiveness in the Christian
religions, as well as with the Jubilee concept in Judaism.
While clearly not a direct measure of stigma, the confluence of effects of geographic
variables, the age distribution, and data on religious adherence is highly suggestive of a strong
role for stigma in the determination of bankruptcy filing rates.
In the remainder of the section, results from Model 3 are utilized in the discussion,
keeping in mind that for any given variable, coefficient values may differ across specification.
For most variables, these differences in values are negligible.
Wage Garnishment
As expected, the greater the protection given to wages, the lower the bankruptcy filing
rate. Specifically, the number of bankruptcies per 10,000 households decreases roughly one-to-
one for every one percentage point increase in the proportion of state household median income
protected from garnishment.
DRAFT. Please do not distribute without permission from the author.
25
Legal Gambling
Minimum distance to a casino has a statistically significant negative effect on non-
business bankruptcy filing rates, meaning that the further a county centroid is from the nearest
(legal) casino, the lower is the filing rate. Specifically, an additional 100-mile distance from a
casino results in 4.3 fewer bankruptcy filings per 10,000 households, or given that the average
number of bankruptcies per county is 100, roughly four percent of filings in the average county.
Counties in states where card rooms are legal on average have 11 more bankruptcy filings than
counties in states where card rooms are illegal. Surprisingly, the availability of race track
gambling and a state lottery reduced filing rates compared to counties in states without race
tracks and state lotteries. Race track betting is legal in 42 states, while 33 states have state
lotteries, so there was little variation in the data, especially given that these are state-level
variables.
Entrepreneurship and Small Business
Both the self-employment rate and net firm births (firm births less firm deaths) are
negatively related to bankruptcy filing rates. A one percentage point higher self-employment
rate is associated with 4.4 fewer bankruptcy filings per 10,000 households, on average, which
conflicts with the expectation that higher self-employment rates would be associated with
increased bankruptcy filing rates. 10 additional net firm births are associated with roughly 2.6
additional bankruptcy filings per 10,000 households, suggesting, as expected, that greater
business success engenders lower bankruptcy filing rates.
Economic and Social Factors and Health Care
The employment rate is negatively related to bankruptcy filing rates, as expected. A one
percentage point difference in the employment rate is associated with about 0.8 less bankruptcy
filings per 10,000 households, a change of less than one percent for the average county.
Although this effect seems small on the surface, employment rates across counties varied from
23.8 percent to 84.6 percent in 2000. The mean employment rate was 57.4 percent and the
standard deviation was 7.5 percent, which means that a county one standard deviation up from
DRAFT. Please do not distribute without permission from the author.
26
the mean would incur roughly 6 fewer bankruptcies than the county with the mean employment
rate.
Previous studies generally have included the level of income in individual studies or the
level of median or average income in aggregate studies, with a quadratic term in some cases, and
the results have been mixed. I include the entire distribution of income, and the results suggest
that bankruptcy filing rates likely rise with income initially (from very low income), drop
slightly in the middle income range, then spike at the $50,000 – $75,000 range before declining
steadily with higher levels of income (Figure 5). Very low bankruptcy filing rates would be
expected in the lowest income ranges because there is little access to credit, and therefore less
borrowing. At higher income levels, bankruptcy filing rates increase with much better access to
credit but relatively moderate incomes. At some point, at the $50,000 – $75,000 range in this
model, bankruptcy filing rates begin to decline with a reduced need to borrow relative to income.
Of course, these results reflect proportions of the population that are in the various
income ranges. The assumption is that these proportions mirror closely the distribution of filers.
In fact, Rodríguez et al. (2002), in a study using the 1998 Survey of Consumer Finances, show
that the income quintile with the highest bankruptcy filing rate is the third quintile, which in
2000 included households with income in the range of $33,006 to $52,272,23 which is consistent
with this view. Nevertheless, county income distributions may reflect other phenomena besides
the income distribution of bankruptcy filings. For example, any individual low income person
may be more likely to file for bankruptcy if he lives in a high income area than if he lives in a
low income area if there is pressure to “keep up with the Joneses.”
Bankruptcy filing rates are higher the greater the proportion of the population receiving
public assistance. This result is surprising in that public support in times of financial crisis
would presumably forestall bankruptcy filings that might otherwise occur. On the other hand,
controls for income are included in the model, and the receipt of public assistance may for some
23 Data on limits of income quintiles for 2000 are from the U.S. Census Bureau, Historical Income Tables – Households, Table H-1, accessed November 2, 2005 at http://www.census.gov/hhes/income/histinc/h01.html.
DRAFT. Please do not distribute without permission from the author.
27
people (but certainly not all) reflect poor financial decision making and little capacity for
climbing out of financial messes. Moreover, a large proportion of the population receiving
public assistance may be indicative of underlying economic crises in the county.
As expected, bankruptcy filing rates are higher the greater the share of the population
without health insurance and the larger the share that is disabled. Specifically, a one percentage
point higher share of the population without health insurance is associated with 0.9 more
bankruptcies per 10,000 households. These shares ranged from 7.4 percent to 24.2 percent in
2000. A one percentage increase in the share of the 21 – 64 population that is disabled, which
ranged from 5.7 percent to 45.6 percent in 2000, is associated with 1.5 additional bankruptcies
per 10,000 households.
Credit and Debt
Results suggest that debt loads are an important factor in explaining bankruptcy filing
rates, but measures of access to credit, such as the number of new accounts and credit cards per
capita are not statistically significant explanatory factors. Total debt to income is positively
related to bankruptcy filing rates, but the magnitude is quite small at less than 0.2 filings per
10,000 households for every additional point in the debt to income ratio. The coefficient on the
past due index is positive and fairly substantial. A one unit higher past due index, which ranges
from 0.7 to 48.9, is associated with 1.6 additional bankruptcies per 10,000 households.
Homes and Autos
As expected, significant spending on autos is associated with higher bankruptcy filing
rates. A one percentage point higher share of the population with more than two vehicles is
associated with 1.4 additional bankruptcies per 10,000 households, while greater ratios of auto
debt to total debt are related to higher bankruptcy filing rates in roughly one-to-one fashion.
As in existing literature I find that higher rates of home ownership generate lower
bankruptcy filing rates. Specifically, a county with a one percentage point higher owner
occupancy rate has, on average, 0.8 fewer bankruptcy per 10,000 households. However, owner
occupancy rates across counties are highly concentrated around the mean of 74 percent (Table
DRAFT. Please do not distribute without permission from the author.
28
1). Also supporting the literature is that my results suggest that home equity is associated with
lower bankruptcy filing rates. Specifically, median house value has a statistically significant
positive coefficient, while the share of owner-occupied homes with mortgages, controlling for
the owner occupancy rate, is positively correlated with bankruptcy filing rates. The vacancy rate
of owner-occupied dwellings does not appear to affect bankruptcy filing rates. Higher costs of
home ownership are associated with higher bankruptcy rates, but median rent does not seem to
affect bankruptcy filing rates.
Demographics and Other Variables
In terms of race, relative to the share of the population that is white, Black and
Hawaiian/Pacific Islander shares of the population are positively correlated with bankruptcy
filing rates, while proportions that are American Indian, Asian, or “Other Race” are negatively
correlated with bankruptcy filing rates. The results on black share of the population are
consistent with existing literature, which has little to say about other races.
As expected, the share of the population that is divorced is positively correlated with
bankruptcy filing rates. Specifically, a one percentage point higher share of the population that
is divorced is associated with 7.8 more bankruptcies per 10,000 households, which is substantial
even though county divorce rates are highly concentrated around their mean value (Table 1).
Married households also have higher filing rates than never married households. Domowitz and
Sartain (1999) find that people who are unmarried have a higher probability of filing for
bankruptcy than people who are married. My results are not inconsistent with this finding in the
sense that in the Domowitz and Sartain analysis, divorced people were included in the unmarried
category, while the comparative (and left out) category here is never married. Surprisingly, the
share of the population that is separated is negatively associated with bankruptcy filings, even
though separation is known to cause substantial financial strain. A possible explanation for this
unlikely finding is that people in the throes of a marital dissolution are unlikely to want to deal
with bankruptcy inconveniences at such a time in their lives, and those in severe financial
DRAFT. Please do not distribute without permission from the author.
29
distress will put off filing as long as possible. The share of the population that is widowed does
not have a statistically significant impact on bankruptcy filing rates.
4. Conclusions
This paper analyzes a variety of potential determinants of bankruptcy filing rates.
Proximity to gambling establishments was found to be associated with higher bankruptcy filing
rates. The use of an innovative, and I believe much improved, measure of proximity helps to
shed light on a still unresolved issue. Homestead exemptions were found to be a substantial
determinant of bankruptcy filing rates, once controls for endogeneity were included in the model.
This finding suggests that the weak relationship between homestead exemptions and bankruptcy
filing rates found in much of the existing literature may be more of an econometric issue than a
substantive issue. Another important policy variable, the share of wages protected from
garnishment, was shown to be a significant factor in the determination of bankruptcy filing rates
as well, and the results support the conceptual view that wage garnishment spurs bankruptcy
filing. Data on religious adherence was incorporated in the model in an effort to account for the
role of stigma in explaining bankruptcy filing rates. I believe that the religion variables included
in this study are to date the closest proxy to stigma that have been used in bankruptcy studies.
Further, results from regional dummy variable and spatial autoregressive models bring new
evidence to bear on the stigma issue, as well as related spatial issues, and the inclusion of the
entire age distribution, rather than median age, provides more informative results on the role of
age in determining bankruptcy filing rates. Numerous other variables also were included to
provide results from the most comprehensive model of bankruptcy filing rates to date in terms of
the number of potential factors investigated.
The new bankruptcy law likely will have some effect on filing rates by curbing abuses
and forcing more petitioners into a reorganization plan that requires them to pay part of their
debts. This study suggests that exemption and garnishment laws also should be evaluated to
ensure that consumers are offered some protection and a fresh start without being given too
DRAFT. Please do not distribute without permission from the author.
30
much incentive to engage in risky financial behavior that may lead to bankruptcy. But efforts to
change consumer behavior through education likely will have the greatest impact, as many of the
factors shown to affect bankruptcy filing rates, such as gambling, debt composition, home
ownership, and prompt payment of bills, are reflections of financial decisions, and good
decisions require substantial knowledge.
DRAFT. Please do not distribute without permission from the author.
31
REFERENCES Apilado, Vincent P., Joel J. Dauten, and Douglas E. Smith (1978). “Personal Bankruptcies,”
Journal of Legal Studies, 7(2), 371 – 392. Agarwal, Sumit, Chunlin Liu, and Lawrence Mielnicki (2003). “Exemption Laws and Consumer
Delinquency and Bankruptcy Behavior: An Empirical Analysis of Credit Card Data,” The Quarterly Review of Economics and Finance, 43(2), 273 – 289.
Berkowitz, Jeremy and Michelle J. White (2004). “Bankruptcy and Small Firms’ Access to
Credit,” RAND Journal of Economics, 35(1), 69 – 84. Brinig, Margaret F. and F. H. Buckley (1996). “The Market for Deadbeats,” Journal of Legal
Studies, 25(1), 201 – 232. Burman, Leonard E. and Deborah I. Kobes (2003). “Composition of Income Reported on Tax
Returns,” Tax Analysts’ Tax Notes, November 10. Accessed at http://www.taxpolicycenter.org/UploadedPDF/1000639_TaxFacts_111003.pdf on November 1, 2005.
Dawsey, Amanda E. and Lawrence M. Ausubel (2001). “Informal Bankruptcy,” Working Paper,
Department of Economics, University of Maryland, January. de la Viña, Lynda and David Bernstein (2002). “The Impact of Gambling on Personal
Bankruptcy Rates,” Journal of Socio-Economics, 31(5), 503 – 509. Domowitz, Ian and Thomas L. Eovaldi (1993). “The Impact of the Bankruptcy Reform Act of
1978 on Consumer Bankruptcy,” Journal of Law and Economics, 36(2), 803 – 835. Domowitz, Ian and Robert L. Sartain (1999). “Determinants of the Consumer Bankruptcy
Decision,” Journal of Finance, 54(1), 403 – 420. Efron, Bradley (1979). “Bootstrap Methods: Another Look at the Jackknife” The Annals of
Statistics, 7(1), 1 – 26. El-Gamal, Mahmoud A. (2001). “An Economic Explication of the Prohibition of Ribā in
Classical Islamic Jurisprudence,” working paper, Rice University, May. Elul, Ronel and Narayanan Subramanian (2002). “Forum-Shopping and Personal Bankruptcy,”
Journal of Financial Services Research, 21(3), 233 – 255. Fan, Wei and Michelle J. White (2003). “Personal Bankruptcy and the Level of Entrepreneurial
Activity,” Journal of Law and Economics, 46(2), 543 – 567. Fay, Scott. Erik Hurst, and Michelle J. White (2002). “The Household Bankruptcy Decision,”
American Economic Review, 92(3), 706 – 718. Flynn, Ed and Gordon Bermant (2003). “Chapter 7 Debtors – From 19 to 92,” ABI Journal,
December/January. Grant, Charles (2001). “Consumer Bankruptcy Law, Credit Constraints and Insurance: Some
Empirics,” Working Paper, Finance and Consumption Programme, European University Institute, April.
DRAFT. Please do not distribute without permission from the author.
32
Greene, William H. (1993). Econometric Analysis, 3rd Ed. Upper Saddle River, NJ: Prentice-
Hall, Inc. Gropp, Reint, John Karl Scholz, and Michelle J. White (1997). “Personal Bankruptcy and Credit
Supply and Demand,” Quarterly Journal of Economics, 112(1), 217 – 251. Gross, David B. and Nicholas S. Souleles (2002). “An Empirical Analysis of Personal
Bankruptcy and Delinquency,” The Review of Financial Studies. 15(1), 319 – 347. Himmelstein, David U., Elizabeth Warren, Deborah Thorne, and Steffie Woolhandler (2005).
“Illness and Injury as Contributors to Bankruptcy,” Health Affairs, Web Exclusive, February 2. Accessed September 30, 2005 at http://content.heathaffairs.org/cgi/content/
full/hlthaff.w5.63/DCI. Hira, Tahira K. “Bankruptcy and Gambling in Iowa: Is There a Connection?” Paper presented at
the Annual Conference for the National Coalition Against Gambling Expansion, St. Louis. MO.
Hoffman, Saul D. and Greg J. Duncan (1988). “What Are the Economic Consequences of
Divorce?” Demography, 25(4), 641 – 645. Illinois Gaming Board (1997). 1997 Annual Report, Springfield, IL. Lawless, Robert M. and Elizabeth Warren (2005). “The Myth of the Disappearing Business
Bankruptcy,” California Law Review, 93(5), 743. Lin, Emily Y. and Michelle J. White (2001). “Bankruptcy and the Market for Mortgage and
Home Improvement Loans,” Journal of Urban Economics, 50(1), 138 – 162. Livshits, Igor, James MacGee and Michèle Tertilt (2005). “Accounting for the Rise in
Consumer Bankruptcies in Canada and the United States,” working paper, University of Western Ontario, March 9.
Mathur, Aparna (2005). “A Spatial Model of the Impact of State Bankruptcy Exemptions on
Entrepreneurship,” Office of Advocacy, U.S. Small Business Administration, July. National Opinion Research Center (1999). Gambling Impact and Behavior Study: Report to the
National Gambling Impact Study Commission, April. Nichols, Mark W., B. Grant Stitt, and David Giacopassi (2000). “Casino Gambling and
Bankruptcy in New United States Casino Jurisdictions,” Journal of Socio-Economics, 29(3), 247 – 261.
Peterson, Richard L. and Kiyomi Aoki (1984). “Bankruptcy Filings Before and After
Implementation of the Bankruptcy Reform Law,” Journal of Economics and Business, 36(1), 95 – 105.
Posner, Eric A., Richard Hynes, and Anup Malani (2001). “The Political Economy of Property
Exemption Laws,” John M. Olin Law & Economics Working Paper No. 136, The Law School, University of Chicago, September 19.
DRAFT. Please do not distribute without permission from the author.
33
Rodríguez, Santiago Budría, Javier Díaz-Giménez, Vincenzo Quadrini, and José-Victor Ríos-Rull (2002). “Updated Facts on the U.S. Distributions of Earnings, Income, and Wealth,” Federal Reserve Bank of Minneapolis Quarterly Review, 26(3), 2 – 35.
Shiers, Alden F. and Daniel P. Williamson (1987). “Nonbusiness Bankruptcies and the Law:
Some Empirical Results,” Journal of Consumer Affairs, 21(2), 277 – 292. Stinchfield, Randy and Ken C. Winters (1996). “Treatment Effectiveness of Six State-Supported
Compulsive Gambling Treatment Programs in Minnesota,” Presented to the Compulsive Gambling Treatment Program, Minnesota Department of Human Services.
Sullivan, Teresa A., Elizabeth Warren, and Jay Lawrence Westbrook (1989). As We Forgive
Our Debtors: Bankruptcy and Consumer Credit in America, Oxford University Press, Oxford, UK.
Sullivan, A. Charlene and Debra Drecnik Worden (1990). “Rehabilitation or Liquidation:
Consumers’ Choices in Bankruptcy,” Journal of Consumer Affairs, 24(1), 69 – 88. Thalheimer, Richard and Mukhtar M. Ali (2004). “The Relationship of Pari-Mutual Wagering
and Casino Gaming to Personal Bankruptcy,” Contemporary Economic Policy, 22(3), 420 – 432.
U.S. Department of the Treasury (1999). “A Study of the Interaction of Gambling and
Bankruptcy,” Washington, DC, July. VISA USA, Inc. (1996). “Consumer Bankruptcy: Causes and Implications,” VISA Consumer
Bankruptcy Reports, July. Warren, Elizabeth, Teresa Sullivan, and Melissa Jacoby (2000). “Medical Problems and
Bankruptcy Filings,” Working Paper No. 008, Public Law and Legal Theory Working Paper Series, Harvard Law School, April.
White, Michelle J. (1987). “Personal Bankruptcy Under the 1978 Bankruptcy Code: An
Economic Analysis,” Indiana Law Journal, 63(1), 1 – 53. White, Michelle J. (2005). “Economic Analysis of Corporate and Personal Bankruptcy Law,”
National Bureau of Economic Research Working Paper 11536, August. Woodward. William J., Jr. (1982). “Exemptions, Opting Out, and Bankruptcy Reform,” Ohio
State Law Journal, 43(2), 335 – 376.
DRAFT. Please do not distribute without permission from the author.
34
TABLES AND FIGURES Figure 1: Personal Bankruptcies in the U.S., 1980 – 2004
0.0
10.0
20.0
30.0
40.0
50.0
60.0
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Ban
krup
tcie
s/10,
000
Am
eric
ans
Treasury (1999)
Administrative Office of the U.S.Courts
DRAFT. Please do not distribute without permission from the author.
35
Figure 2: Bankruptcy Filing Rates by County, United States, 2000
DRAFT. Please do not distribute without permission from the author.
36
Figure 3: Age Distribution of Bankruptcy Filings, 1991 and 2001
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0Sh
are
of F
Iling
s
< 25 25-34 35-44 45-54 55-64 > 65
Age Group
1991 2001
Source: Consumer Bankruptcy Project, as cited in Suein Hwang, "Mr. Hester Takes a Mall Job," The Wall Street Journal, August 6, 2004.
DRAFT. Please do not distribute without permission from the author.
37
Figure 4: Gambling in the United States, 2000
DRAFT. Please do not distribute without permission from the author.
38
Figure 5: Results on Income Distribution
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
< 15
K
15K
- 25
K
25K
- 35
K
35K
- 50
K
50K
- 75
K
75K
- 10
0K
> 10
0K
Income Range
Coe
ffici
ent V
alue
All are significant at the 99 percentconfidence level.
DRAFT. Please do not distribute without permission from the author.
39
Table 1: Data Sources Variable Source Religion Association of Statisticians of American Religious Bodies (ASARB), Religious
Congregations and Memberships in the United States: 2000, Nashville TN, Glenmary Research Center, 2002.
Education U.S. Census Bureau, Decennial Census Age Distribution U.S. Census Bureau, Decennial Census Race U.S. Census Bureau, Decennial Census Income U.S. Census Bureau, Decennial Census Social Households w/ Public Assistance Disabled Population, 21 – 64 No Health Insurance
U.S. Census Bureau, Decennial Census U.S. Census Bureau, Decennial Census U.S. Census Bureau, Current Population Survey, Annual Social and Economic
Supplement Marital Status U.S. Census Bureau, Current Population Reports Housing Median House Value Median Costs, OOCC Housing Median Rent Owner-Occupany Rate Vacancy Rate, OOCC Housing OOCC Housing, % w/ Mortgage Mortgage Debt / Total Debt
U.S. Census Bureau, Decennial Census U.S. Census Bureau, Decennial Census U.S. Census Bureau, Decennial Census U.S. Census Bureau, Decennial Census U.S. Census Bureau, Decennial Census U.S. Census Bureau, Decennial Census Trendata, TransUnion LLC
Vehicles Number in Households Auto Debt / Total Debt
U.S. Census Bureau, Decennial Census Trendata, TransUnion LLC
Economy Employment Rate Self-Employment Rate Net Firm Births
Annual Statistical Supplement, 2000, U.S. Social Security Administration
(www.ssa.gov/policy/docs/statcomps/supplement/2000, accessed July 1, 2005).
Unpublished data from the U.S. Census Bureau, 1989 – 2000 Business Information Tracking Series.
Debt Management Trendata, Transunion LLC Gaming Minimum Distance to Casino Lottery, Card Rooms, and Race
Tracks
Author’s calculation using data from the National Indian Gaming Commission
(www.nigc.gov), various tribal web sites, calls to tribal contacts, various state gaming agency compliance reports, and unpublished data provided by Casino City Press.
“Gaming Activities,” Internal Revenue Service (www.irs.gov/businesses/ small/industries/article/0,,id=99639,00.html, accessed on June 3, 2005).
State Law Legal codes of the 50 states and the District of Columbia
DRAFT. Please do not distribute without permission from the author.
40
Table 2: Results, Bankruptcy Filing Rates
GLS OLS-IV† OLS-R-IV† SAR-IV† Variable / Model 1 2 3 4
Intercept - 128.1 (- 1.360)
- 145.3 (- 1.688)
- 194.2* (- 1.892)
- 173.3* (- 1.760)
Christian - Catholic 0.050 (0.734)
0.048 (0.624)
0.019 (0.260)
0.146* (1.822)
Eastern Religions - 13.79*** (- 3.609)
- 13.75*** (- 4.016)
- 13.72*** (- 3.498)
- 7.575** (- 2.018)
Islam - 4.308 (- 1.294)
- 4.243** (- 2.129)
- 4.809** (- 2.564)
- 6.106*** (- 3.052)
Jewish 2.420** (2.506)
2.369*** (3.100)
1.976** (2.527)
1.452** (2.106)
Christian - LDS 0.178 (1.505)
0.188 (1.613)
0.129 (0.972)
0.192* (1.945)
Christian - Orthodox 0.195 (0.127)
0.178 (0.126)
0.500 (0.356)
- 0.158 (- 0.109)
Unitarian Universalist - 44.19*** (- 5.390)
- 44.02*** (- 4.580)
- 38.42*** (- 4.231)
- 33.27***
(- 3.474)
REL
IGIO
N (%
)
Christian - Protestant 0.155** (2.341)
0.152*** (2.862)
0.141** (2.038)
0.053 (0.895)
< High School - 0.011 (- 0.044)
- 0.002 (- 0.009)
0.031 (0.124)
- 0.311 (- 1.232)
Some College / Associate Deg
- 0.849*** (- 3.686)
- 0.831*** (- 3.571)
- 0.507** (- 2.212)
- 0.182 (- 0.808)
EDU
C (%
)
Bachelor’s or Higher - 0.004 (- 1.365)
- 0.003 (- 1.176)
- 0.003 (- 1.225)
- 0.005* (- 1.691)
< 15 0.409 (0.413)
0.454 (0.386)
0.947 (0.968)
1.144 (1.049)
15 – 19 0.181 (0.191)
0.146 (0.142)
0.377 (0.332)
0.103 (0.105)
20 – 24 1.016 (0.788)
1.189 (0.905)
1.720 (1.331)
0.547 (0.388)
35 – 44 - 2.452 (- 1.560)
- 2.365 (- 1.582)
- 2.124 (- 1.374)
- 0.682 (- 0.450)
45 – 54 - 0.669 (- 0.586)
- 0.399 (- 0.365)
- 0.196 (- 0.177)
0.472 (0.423)
55 – 59 - 4.277* (- 1.691)
- 4.374* (- 1.960)
- 4.551* (- 1.738)
- 4.685** (- 2.069)
60 – 64 - 4.483* (- 1.671)
- 4.329** (- 2.049)
- 3.842 (- 1.366)
- 4.038* (- 1.773)
65 – 74 - 1.768 (- 1.101
- 1.740 (- 1.106)
- 1.332 (- 0.872)
- 0.248 (- 0.162)
75 – 84 4.497** (2.117)
4.560** (2.492)
3.970* (1.764)
3.768* (1.836)
AG
E D
ISTR
IBU
TIO
N (%
)
> 84 - 2.049 (- 0.919)
- 2.096 (- 0.909)
0.235 (0.093)
- 0.410 (- 0.153)
t-statistics are in parentheses ***, **, and * indicates statistical significance at the 1, 5, and 10 percent levels, respectively. † indicates t-statistics are computed with bootstrapped standard errors. (Continued on next page)
DRAFT. Please do not distribute without permission from the author.
41
Table 2 (Continued)
GLS OLS-IV† OLS-R-IV† SAR-IV† Variable / Model 1 4 5 6
Black 1.134*** (7.708)
1.133*** (9.342)
1.152*** (9.624)
0.871*** (6.987)
American Indian - 0.689*** (- 4.269)
- 0.688*** (- 3.974)
- 0.733*** (- 4.526)
- 0.584*** (- 4.290)
Asian - 0.103 (- 0.190)
- 0.112 (- 0.204)
0.202 (0.358)
- 0.282 (- 0.447)
Hawaiian / Pacific Islander 9.402** (2.560)
9.070*** (2.657)
7.307* (1.810)
5.400 (1.542)
Other Race - 0.770*** (- 3.091)
- 0.760*** (- 3.127)
- 0.480* (- 1.818)
- 0.068 (- 0.275)
RA
CE
(%)
2 or More Races - 0.526 (- 0.495)
- 0.386 (- 0.345)
0.430 (0.384)
0.268 (0.269)
15K – 25K 2.217*** (3.654)
2.199*** (3.776)
1.944*** (3.502)
2.181*** (4.283)
25K – 35K 2.964*** (4.874)
2.989*** (4.707)
2.926*** (4.561)
2.354*** (4.027)
35K – 50K 1.831*** (3.140)
1.874*** (3.737)
1.872*** (3.240)
1.327** (2.318)
50K – 75K 2.873*** (5.312)
2.956*** (6.370)
3.144*** (6.962)
2.192*** (4.070)
75K – 100K 2.073** (2.436)
2.126** (2.510)
2.319*** (3.095)
0.996 (1.227)
INC
OM
E (%
)
> 100K 1.497** (2.355)
1.491** (2.457)
1.582*** (2.629)
0.135 (0.204)
Households w/ Public Assist.
0.922 (1.341)
0.992* (1.834)
1.671** (2.495)
0.785 (1.330)
Disabled Pop. 21-64 1.691***
(4.505) 1.765*** (5.176)
1.621*** (4.702)
1.460*** (4.340)
SOC
IAL
No Health Insurance 1.520*** (4.620)
1.533*** (5.252)
0.986** (2.345)
0.906*** (3.318)
Married 1.739*** (4.459)
1.778*** (4.660)
1.847*** (5.044)
0.902*** (2.692)
Divorced 7.272*** (9.127)
7.212*** (9.315)
7.824*** (11.063)
5.201*** (7.599)
Separated - 6.108*** (- 3.663)
- 5.988*** (- 3.724)
- 7.457*** (- 5.369)
- 5.418*** (- 3.517)
MA
RIT
AL
STA
T
Widowed 3.336*** (2.776)
3.425*** (3.653)
3.464*** (3.744)
1.078 (1.084)
t-statistics are in parentheses ***, **, and * indicates statistical significance at the 1, 5, and 10 percent levels, respectively. † indicates t-statistics are computed with bootstrapped standard errors. (Continued on next page)
DRAFT. Please do not distribute without permission from the author.
42
Table 2 (Continued)
GLS OLS-IV† OLS-R-IV† SAR-IV† Variable / Model 1 4 5 6
Median House Value ($) - 0.00025*** (- 3.677)
- 0.00024*** (- 5.470)
- 0.00020*** (- 3.935)
- 0.00022*** (- 5.760)
Median Costs, Owner-Housing
0.048***
(2.986) 0.047*** (3.261)
0.034** (2.424)
0.055*** (4.324)
Median Rent 0.010 (0.465)
0.009 (0.474)
0.004 (0.187)
0.029 (1.401)
Owner-Occup. Rate - 0.877*** (- 3.700)
- 0.878*** (- 4.082)
- 0.839*** (- 3.516)
- 0.389** (- 2.042)
Owner-Occup. Vacancy Rate
1.068 (1.620)
0.907 (1.219)
1.021 (1.374)
1.133* (1.660)
Owner-Occup. w/ Mortgage
1.078*** (6.397)
1.097*** (6.407)
1.021*** (5.846)
0.859*** (5.714)
HO
USI
NG
Mortgage Debt Share of Total
- 0.010 (- 0.884)
- 0.009 (- 0.593)
- 0.003 (- 0.219)
- 0.013 (- 0.912)
None 0.514 (1.052)
0.498 (1.199)
0.526 (1.387)
- 0.098 (- 0.242)
Two 0.150 (0.264)
0.124 (0.315)
0.057 (0.127)
- 0.248 (- 0.553)
More than Two 1.443*** (4.188)
1.448*** (5.441)
1.446*** (4.971)
0.402 (1.458) V
EHIC
LES
Auto Debt Share of Total 1.101*** (4.546)
1.028*** (4.396)
1.016*** (3.834)
0.828*** (3.351)
Employment Rate - 0.852*** (- 3.292)
- 0.894*** (- 3.682)
- 0.825*** (- 3.406)
- 0.722*** (- 2.972)
Self-Employ Rate - 4.450*** (- 5.429)
- 4.305*** (- 4.915)
- 4.402*** (- 4.261)
- 1.681** (- 2.448)
ECO
NO
MY
Net Firm Births - 0.242** (- 1.965)
- 0.236** (- 2.011)
- 0.257** (- 2.540)
- 0.322*** (- 2.911)
Total Debt to Income 0.190** (2.394)
0.201*** (2.647)
0.157** (2.065)
0.062 (0.787)
Revolving Debt Share of Total
- 0.481* (- 1.941)
- 0.492* (- 1.734)
- 0.593**
(- 2.249) - 0.398
(- 1.503)
Number of Bankcards - 7.691 (- 1.297)
- 8.281 (- 1.515)
- 4.655 (- 0.832)
1.697 (0.350)
Number of New Accounts - 159.5 (- 0.923)
- 161.2 (- 0.989)
- 82.25 (- 0.487)
- 139.88 (- 0.889)
DEB
T M
AN
AG
EMEN
T
Past Due Credit Bills Index
1.555*** (5.524)
1.550*** (6.032)
1.590*** (6.991)
1.272*** (5.531)
Min Distance to a Casino - 0.040*** (- 4.822)
- 0.040*** (- 5.917)
- 0.043*** (- 5.777)
- 0.017*** (- 2.703)
Lottery - 5.357* (- 1.768)
- 4.969** (- 2.054)
- 2.814 (- 1.195)
- 11.821*** (- 5.904)
Card Rooms 8.918** (2.542)
7.831* (1.706)
10.78*** (2.616)
4.885 (1.347) G
AM
ING
Race Tracks - 13.932*** (- 3.900)
- 14.26*** (- 4.846)
- 13.67*** (- 4.835)
- 3.554 (- 1.585)
t-statistics are in parentheses ***, **, and * indicates statistical significance at the 1, 5, and 10 percent levels, respectively. † indicates t-statistics are computed with bootstrapped standard errors. (Continued on next page)
DRAFT. Please do not distribute without permission from the author.
43
Table 2 (Continued)
GLS OLS-IV† OLS-R-IV† SAR-IV† Variable / Model 1 4 5 6
Homestead Exemption 0.119 (0.649)
0.680* (1.765)
0.939** (2.244)
0.253 (0.685)
Unlimited Homest. Exmp. - 3.589 (- 0.981)
- 5.096* (- 1.659)
- 2.354 (- 0.729)
- 4.706 (- 1.610)
STA
TE L
AW
Not Subj Wage Garnishment
- 1.072*** (- 9.962)
- 1.042*** (- 12.113)
- 1.038*** (- 12.200)
- 0.704*** (- 9.554)
Mid Atlantic 25.617*** (5.219)
Great Lakes 6.499 (1.328)
Plains 0.208 (0.044)
Southeast 17.28*** (4.017)
Southwest 6.918 (1.297)
Rocky Mountain 12.24** (2.140) R
EGIO
NA
L D
UM
MIE
S
West/Pacific 4.119 (1.096)
ρ 0.528*** (18.92)
Adj. R2 0.5169 0.5174 0.5244 0.5137 t-statistics are in parentheses ***, **, and * indicates statistical significance at the 1, 5, and 10 percent levels, respectively. † indicates t-statistics are computed with bootstrapped standard errors. (Continued on next page)
DRAFT. Please do not distribute without permission from the author.
44
Table 3: Results, Homestead Exemptions
Variable Parameter Value (t-statistic) Variable Parameter Value
(t-statistic)
Intercept - 7,796,626* (- 1.733) Percent White - 5,853*
(- 1.844)
Poverty Rate - 140,452*** (- 2.891) Average Family Size 1,770,668***
(3.045)
Median House Value - 8.062***
(- 3.059) Household Median Income - 32.52*
(- 1.905) Wage Not Subject to Garnishment
- 4,577** (- 2.166)
Households with Public Assistance
- 238,177**
(- 2.644) Percent of Population that is Male
122,891* (1.811) Owner-Occupany Rate - 21,184*
(- 1.954)
Median Age 76,621*** (2.813) Percent GOP - 14,540
(- 1.406) Adjusted R2 0.2063 No. Observations 51 ***, **, and * indicates statistical significance at the 1, 5, and 10 percent levels, respectively.