Social Networks and Personal Bankruptcy

• Date: 01 June 2015
• Published date: 01 June 2015
• Author: Michelle M. Miller
• DOI: http://doi.org/10.1111/jels.12073

Social Networks and Personal Bankruptcy

Michelle M. Miller*

This article examines the role of social networks in a household’s bankruptcy decision.

Social networks may affect a household’s bankruptcy decision in many ways: they could

provide information about the required paperwork, recommend an attorney, reduce the

stigma associated with bankruptcy, or increase awareness of its benefits. Using data from the

Panel Study of Income Dynamics (PSID), I exploit county and racial variation to identify

network effects. My empirical strategy asks whether being surrounded by others of the same

race increases bankruptcy use more for those in racial groups with high filing rates. This

methodology allows me to include both county-year and racial-group fixed effects in my

regressions. The results strongly confirm the importance of networks in a household’s

bankruptcy decision.

I. Introduction

A growing literature documents the impact of social networks on individual behavior.

Prominent work in this area shows, for example, that social networks impact welfare

participation (Bertrand et al. 2000; Gee & Giuntella 2011; Fertado & Theodoropoulos

2013), publicly-funded prenatal care (Aizer & Currie 2004), health-care utilization (Deri

2005), education (Calvo-Armengol et al. 2009; Aaronson 1998), employment (Beaman

2012; Topa 2001), and investment decisions (Li 2014; Duflo & Saez, 2003). In this article,

I examine the role of social networks in a household’s bankruptcy decision.

In 2013, nearly 1.1 million people filed for bankruptcy. Yet despite the large number

of filers, economists and legal scholars still have a limited understanding of an individual’s

bankruptcy decision. The majority of the empirical literature has focused on the financial

factors that contribute to an individual’s bankruptcy decision (e.g., White 2007; Fay et al.

2002; Gross & Souleles 2002; Domowitz & Sartain 1999; Buckley 1994; White 1987). A

smaller literature has examined the demographic factors (Lefgren & McIntyre 2009) and

legal factors (Dawsey & Ausubel 2004; Miller 2010) that contribute to bankruptcy. To date,

however, there has been little convincing empirical work to show that social networks

contribute to bankruptcy.

*Assistant Professor, Department of Economics, Loyola Marymount University, 1 LMU Dr., Rm. 4216, Los Angeles,

CA 90045; email: Michelle.Miller@lmu.edu.

I thank the editors, two anonymous referees, Mary E. Hansen, Frank McIntyre, and conference participants at the

Conference on Empirical Legal Studies and the American Economic Association Annual Meetings, as well as seminar

participants at the Committee on the Status of Women in the Economics Profession CeMENT workshop for their

thoughtful comments.

bs_bs_banner

Journal of Empirical Legal Studies

Volume 12, Issue 2, 289–310, June 2015

289

To be sure, descriptive studies (e.g., Jacob 1969; Braucher 1993) have hypothesized

that social networks may impact bankruptcy filings. They have noted that social networks

could provide information about the required paperwork, recommend an attorney, reduce

the stigma associated with bankruptcy, or increase awareness of its benefits. And while

numerous empirical studies (Fay et al. 2002; Han & Li 2004; Cohen-Cole & Duygan-Bump

2008; Scholnick 2013) have found a positive correlation between a household’s bankruptcy

decision and the lagged filing rate in the household’s area, these positive correlations may

reflect more than local network effects. For example, as will be discussed in more detail

below, these positive correlations may result from unobserved time-varying neighborhood

characteristics such as the number of advertisements for bankruptcy attorneys. Of addi-

tional concern is the broad definition of “neighborhood” used in many of these papers; Fay

et al. (2002) and Han and Li (2004), for example, define neighborhoods at the district and

state level, respectively.

Using a methodology similar to Bertrand et al. (2000), Aizer and Currie (2004), Deri

(2005), Gee and Giuntella (2011), and Fertado and Theodoropoulos (2013), this article is

better able to estimate the impact of social networks on personal bankruptcy. Using data

from the 1991–1995 Panel Study of Income Dynamics (PSID), I define networks using a

household’s county and racial/ethnic group.1Specifically, my empirical strategy asks

whether being surrounded by others of the same racial group increases bankruptcy use

more for those in racial groups with high filing rates. Because my regressions include

county-year and racial-group fixed effects, I am able to eliminate omitted variable bias

caused by unobserved time-varying neighborhood characteristics and unobserved house-

hold characteristics that are correlated with race. To test for remaining omitted variable

bias I (1) use an instrumental variables approach, (2) explore the effect of dropping

covariates, and (3) include household fixed effects. My results withstand all three tests.

By using a household’s racial group to define its network, this article supplements the

growing literature on race and bankruptcy. Most notably, recent works have shown that

upon filing for bankruptcy, African Americans are more likely to file under Chapter 13 of

the Bankruptcy Code (which requires greater repayment of debt, costs more in legal fees,

and takes longer) as compared to debtors of other races (Braucher et al. 2012; Dickerson

2012; Doherty 2012; Eisenberg 2012). Moreover, once they file their Chapter 13 plans,

African-American households have worse outcomes than other groups, with higher dismis-

sal rates (Braucher et al. 2012; Agarwal etal. 2010). This article uses race to examine a

broader question: What causes a household to file for bankruptcy in the first place?

The presence of social networks may help explain several phenomena observed in

the bankruptcy data. First, social networks, or the lack thereof, may at least partially explain

1Both theoretical and empirical works justify a definition of networks along racial/ethnic lines. For example, Alba

(1990) studied U.S.- born white ethnics and found that half of all nonrelated childhood friends belonged to the same

ethnic group. Additionally, Aizer and Currie (2004) found that there is a higher correlation of takeup of publicly-

funded prenatal care within ethnic groups in an area than across ethnic groups in an area, further justifying the

definition of network based along ethnic lines. Finally, Banerjee (1992) and Bikhchandani et al. (1992)

provide theoretical models on information cascades and social learning that provide justification for this empirical

specification.

290 Miller