ITH I

 

 

\ _ UBRARY
13"?) Micnigan State
\ \ University

 

This is to certify that the
thesis entitled

STRAIN, STATUS OR STEREOTYPE?
THE NEXUS OF ECONOMIC DISTRESS AND FAMILY VIOLENCE

presented by

Rebecca Jane Grainger Stone

has been accepted towards fulfillment
of the requirements for the

MS. degree in Criminal Justice

 

 

 

\1 ' Majo‘r Pro’fessor’s Signature
/Q/O?7kuxfl?

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Aoc&Pres/ClRC/DateDue.indd

 

STRAIN, STATUS OR STEREOTYPE?
THE NEXUS OF ECONOMIC DISTRESS AND FAMILY VIOLENCE

by

Rebecca Jane Grain ger Stone

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Criminal Justice

2009

ABSTRACT

STRAIN, STATUS OR STEREOTYPE?
THE NEXUS OF ECONOMIC DISTRESS AND FAMILY VIOLENCE

by

Rebecca Jane Grainger Stone

The current study is an attempt to reproduce and extend upon prior research by
Fox, Benson, DeMaris and Wyk (2002) examining the impact of economic distress on the
odds of intimate partner violence. While there are many theoretical approaches to this
topic, some in direct contradiction to each other, the results of the logistic regression
analysis support an integrated theory of strain, informal social control, and gendered
power-control theories. The results are relevant in light of the 2008-9 economic crisis and
what might be expected as the nation experiences massive rates of home foreclosure and
high unemployment rates. Finally, there is discussion of the difficulties of reproducing
social science from the information published in scholarly journals and suggestions for
future publications, given the importance of regularity of findings when making informed

policy decisions.

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... 4
CHAPTER I: INTRODUCTION .................................................................................... 1
CHAPTER H: THEORETICAL LINKS ........................................................................ 2
Employment ........................................................................................................ 2
Income ................................................................................................................. 9
Status Incompatibility ....................................................................................... 11
CHAPTER HI: CURRENT RELEVANCE .................................................................. 13
CHAPTER IV: CURRENT STUDY ............................................................................ 17
Weakness of Prior Research ............................................................................. 17
Decision to Reproduce a Prior Study ................................................................ 21
Methodology ..................................................................................................... 24
Replication and Reproduction Issues ................................................................ 27
CHAPTER V: RESULTS ............................................................................................. 31
Descriptive Statistics ......................................................................................... 32
Employment ...................................................................................................... 34
Job Strain .......................................................................................................... 35
Household Indicators ........................................................................................ 37
Change in Household Indicators ....................................................................... 39
Final Model ....................................................................................................... 4O
Foreclosure-Specific Model .............................................................................. 42
CHAPTER VI: DISCUSSION ..................................................................................... 43
APPENDICES ........................................ v ...................................................................... 51
REFERENCES ............................................................................................................. 54

iii

LIST OF TABLES

Univariate Statistics ...................................................................................................... 31
Table 1: Regression of Employment Measures on Male-to-Female Violence ............. 33
Table 1a: Regression of Employment Measures on Female-to—Male Violence ........... 33
Table 2: Regression of Job Strain Measures on Male-to-Female Violence .................. 35
Table 2a: Regression of Job Strain Measures on Female-to-Male Violence ................ 35
Table 3: Regression of Household Indicators on Male-to—Female Violence ................ 36
Table 3a: Regression of Household Indicators on Female-to-Male Violence .............. 37

Table 4: Regression of Changes in Household Indicators on Male-to-Female

Violence ........................................................................................................................ 39
Table 4a: Regression of Changes in Household Indicators on Female-to-Male

Violence ........................................................................................................................ 39
Table 5: Final Model ..................................................................................................... 41
Table 6: Regression of Foreclosure Indicators on Male-to-Female Violence .............. 42

iv

INTRODUCTION

Domestic violence advocates frequently claim that violence against intimates is a
“classless” problem. They argue that assaults of “women and children occur in all social
class, across occupation types, in all racial and ethnic groups in all types of
neighborhoods, in cites, and rural areas” (Fagan, 1993: 209). A wealth of research
published over the past forty years has measured domestic violence among families in
every social class, supporting the anecdotal evidence provided by advocates for this claim
(Gelles, 1993). For instance, analysts of both the 1985 and the 1995 National Family
Violence Survey report finding violence in every social class (Straus et a]. 1980; Fagan
and Browne 1994). Nevertheless, a claim that the rate of family violence is independent
of several socially constructed categories “runs counter to empirical evidence about the
social epidemiology of violence between spouses and other intimates” (Fagan 1993: 209).
Gelles (1993) argues that family violence is more likely found among the poor and
unemployed or those holding low-prestige jobs (c.f., Finkelhor et a1. 2005). This paper
will seek to unravel the connections between income, employment, financial stress and
family violence so that we may better understand the interaction between these variables
and address them with more potency. I begin with an overview of the many theoretical
approaches to linking economic concepts to domestic violence, then move to a
description of a 2002 study that I seek to reproduce and the extensions I wish to make. I
then detail the difficulties I encountered while attempting to reproduce the original
analysis. The results of my re-analysis are followed by a comparison between my results
and the original findings, and I close with a discussion of both the relevance of the

findings to this field of study and the difficulty inherent in reproducing social science.

1

THEORETICAL LINKS

For the past forty years of systematic family violence research, researchers
seeking to explain the epidemiology of family violence or the patterns of violence
recidivism following interventions have included several forms of economic capital.
Generally, the choice of which of these measures to model seems more often determined

by their availability than by a theoretical suggestion or favoring of one over another.

Emplognent I will begin by assessing the question about the link between
employment status and family violence. Of the many theoretical arguments available, the
most simplistic links employment and domestic violence through a change in either or
both family member’s routine activities (Cohen and Felson 1979). Out-of-the-home
employment of one or both spouses removes the presence of both perpetrator and target
from each others’ presence for at least some hours of the day. If both spouses are in the
home throughout the day due to unemployment, frustration may build as each feels their
space is invaded. During working hours, neighbors who normally serve as capable
guardians (those who make noise complaints or are likely to alert police) may be at their
own workplaces or alternatively addressing their own home foreclosure and therefore are
unavailable to intervene or reduce the likeliness of violence disputes. This matter is of no
small importance for preventing domestic violence prevention since about fifty percent of
calls to police for domestic violence are made by neighbors. Similarly, children are at

,9

school during the day and not able to play the role of the “capable guardian. Thus, any
connection between employment and violence may have nothing to do with most

stresses, less resources, or symbolism, but purely opportunity: more employment equates

2

to less contact. On the flip-side of this theory, while working long hours reduces contact
(and the opportunity for violence) and increases household income, it can also increase
stress on the employee and his family such that when the partners are in contact, the
likelihood of violence is increased (Fox et al. 2002). Thus, theories and polices linking
more employment to less violence may find their antithesis in overworked individuals.
Previous scholars have also argued that an individual’s “stake in conformity” may
suppress battering behavior (Sherman 1992; Fagan 1993). While the "stakes in
conformity" theory is not developed well enough to specifically identify which sorts of
' measures ought to be good or bad indicators of a person’s stakes or even the exact nature
of the interaction, most of the scholarship in the area of family violence connecting stakes
to violence have relied on employment as the key proxy. Employment is consistently
viewed as an overt measure of a person’s stakes since it is generally accepted that
individuals wish to avoid crime and formal punishment so as not to reduce their chances
of establishing or retaining employment. More specifically under the social control
theme, the conditional deterrence hypothesis asserts that formal sanctions can only deter
offenders if they will also suffer social costs because of their bond to conventional
commitments or norms. Employment as well as home ownership serves as one of several
social bonds (Hirschi 1969), insuring a person’s commitment and attachment to their
family, neighborhood and community. In fact, more so than length of residence,
homeownership is a critical determinant of one’s bond or attachment to their
neighborhood (Brisson & User 2007; Oh 2004). Thus, just like losing a job, losing a

home should weaken a person’s bond that may have deterred them from using violence.

In the context of formal social controls, the effectiveness of sanctions such as
arrest or mandated treatment is believed to be influenced by the reinforcing stimuli of
other informal controls. The hypothesis would therefore predict that the addition of a
criminal sanction should have no deterrent effect among those that have no concurrent
conventional "stakes", and only deter among those that do have some "stakes".
Alternatively, the social control replacement hypothesis argues that criminal sanctions are
only noticeably effective when other informal controls are missing. For those with social
controls, the presence of a criminal sanction adds little to the overall cost already
imposed by other social control mechanisms. In other words, those without informal
controls can be deterred by the direct cost of the criminal sanction (e. g., a night in jail or
a monetary fine), while others are deterred by other costs (e.g., lost of a job or breakup of
a marriage) arising solely from informal social controls. Finally, the additive hypothesis
claims that both informal and formal controls independently deter future deviance, and
that more of either type of control results in failures. As such, those with the highest
degree of social control should have less frequent offending, while those with the least
control should have the highest level of failing.

The “stakes” hypothesis would therefore predict that family violence should go up
as men transition from an employed to an unemployed status, even if the system applies
more sanctions to compensate for the reduced quantity of inform social controls. In fact,
Sherman (1992) suggest the application of more sanctions may make the situation worse
or backfire, which may be particularly true if the men believe their lay-offs are due to the
economic crisis rather than their employee performance. This process may damage a

person’s belief in a fair and just economy and lower their sense of procedural justice (i.e.,

4

anomie). Conversely,-in times of economic hardship, when any employment is highly
valued, many more men than usual may be unwilling to risk unemployment by engaging
in violent behavior. Thus, any increases in violence among those with lower stakes
because they lost their jobs, may in the aggregate, be offset by an increased fear felt
among those still (but now marginally) employed.

Changes in employment status loss of a home status may also represent a ‘turning
point’, as in and Laub and Sampson’s (1993) life-course model. As individuals proceed
along their life-course trajectory, life changes can alter the path of the trajectory and lead
individuals in a different direction, including toward or away from a criminal lifestyle. A
sudden home foreclosure caused by unemployment can be a negative tuming-point,
directing otherwise law-abiding individuals towards criminal behavior. When the
subject’s place of employment fosters frequency, duration and intensity of pro—social
bonds, sudden unemployment may result in the severing of these bonds and the fortnation
of new, less positive attachments (Sutherland 1947). If a person was previously mixing
socially with other employed peers, he may have been discouraged from offending as it
would threaten these social bonds. After a lay-off, the same individual may form bonds
with other unemployed peers who may be pursuing illegitimate sources of income. This
leads to the normalization of criminal behavior, allowing the individual to integrate
antisocial norms into his cognitive schemata and rationalize behavior he would
previously have rejected. Bowker (1984) suggested that these types of bonds (through
reinforcement) that can lead to violence or make it more difficult to stop.

Another casual process that is often discussed in criminology to explain

delinquency, though not discussed among family violence researchers, is Agnew’s

5

revision of strain theory (1992). Of special importance to researchers is Agnew’s notion
of self-generated norms, as this may tie strain theory to theories of social learning and
environment. In this context, Agnew suggests that economic capital such as
employment is a source of positive stimulus and that its removal (unemployment) will
trigger anger, frustration, and criminal behavior. This process would certainly also exist
when someone loses their home to a foreclosure regardless of whether they lost
employment as well. The resulting anger would surely arise when a home foreclosure is
considered unfair and unrelated to an individual’s ability to pay prior monthly balances
but can no longer keep-up because of an adjustment of the interests, as is the case with
many subprime loans.

Closely related to this theoretical tangent is the frustration-aggression hypothesis
(Dollard et al. 1939). When applied to layoffs and violence, the frustration-aggression
hypothesis predicts that men who are laid-off will have an increased risk for violence
while men who fear lay-offs will have a reduced risk. A direct test of
provocation/inhibition effects (and subsequent study replication) supported this approach
(Catalano et al. 2002). Layoffs have been found to increase angry, irritable and insulting
behaviors (Vinokur and Price 1996). One study found that men who were laid off
between the first and second study interviews were six times more likely to report poor
psychiatric condition versus subjects who had retained employment (Catalano and
Dooley 1993). Interestingly, men who were employed in relatively unstable fields (for
example, contract work as opposed to stable industries) were twenty percent less likely to
engage in violent behavior. Both findings remained significant when controlling for age,

gender, socioeconomic status, marital status, psychiatric and alcohol disorders. Thus, it

6

appears that the effect of employment on violence may be illustrated by a mediated rather
than a linear model. Men who fear unemployment may practice greater self-control, to
which Gottfredson and Hirschi (1990) attribute most desistance from crime.

Employment may also serve as a status or resource indicator in the relationship
between a working partner and a non-working spouse. Feminist theories of male
dominance and traditional gender roles would suggest that unemployment strips the male
of one method of dominance over his spouse, leaving him only force or its threat (Goode
1971). If the female partner maintains employment, the male may feel pressured to exert
his dominance through using physical aggression (Macmillan & Gartner 1999). Financial
stress and ambiguity of gender roles may damage a normally loving and supportive
relationship, removing the otherwise positive effects of this bond. This tension could be
exacerbated by women wanting their partners to work longer hours (presumably to
increase household income) (Fox et al. 2002). Comparative resource theory suggests that
the interaction effect of men’s and women’s employment is “as salient to violence as the
independent effect of each alone” (Fox et al. 2002). This theory may effectively cover
many types of family violence as we detect imbalances between parents and children,
parents and their elders, and both same-sex and opposite-sex couples.

As an alternative to the causal mechanism suggested by strain theory or the
mediating model proposed by the “stakes” proponents, employment may also viewed as
simple a proxy for self-control. Punctuality, compliance, and overall work ethos may all
symbolize an individual commitment and attachment to one’s workplace and greater
environment. Alternatively, these employee characteristics may simply provide evidence

of Gottfredson and Hirschi’s (1990) “self-control”. As this example illustrates, the

7

relationships between microeconornic variables and intimate violence can be dangerously
spurious. While not well studied within the context of family violence, some research
implies that intimate or farme violence, including violence directed at older family
members, has more to do with personality characteristics like low self control than stress
(Pillemer and Finkelhor 1989; Sellers 2006).

Certainly, some of these theoretical augments are more supported than others.
Employment has been found to deter or prevent violence, functioning as a protective
factor while unemployment is a risk factor (Bachman 2000; Bowlus & Seitz 2006; Mistry
et al. 2008; O’Donnell et al. 2002). In a German study, researchers found that men who
were unemployed or expected to be unemployed in the near future self-reported an
increase in their violent behavior, though this relationship was reduced by the presence of
greater “self-awareness” (Fischer et al. 2008), a possible connection to what Gottfredson
and Hirschi (1990) label “self-control”. Tauchen and Witte (2001) also found that
employed men were less likely to be violent than those who were unemployed. When
combined with the earlier findings of Catalano and colleagues (2002), this suggests that
employment serve as a social bond that moderates aggressive behavior. When
employment disappears, the resulting frustration manifests as violence.

There is also substantial evidence to support gendered theories of domestic
violence. When comparing employment as a protective factor for men and women, it has
been found that employment is a stronger deterrent for men than for women (Magdol et
al. 1997). When considered alone, male and female unemployment did not produce
significant changes in violence (Macmillan & Gartner 1999); this relationship was found

to hinge upon marital equality rather than individual employment. Female employment

8

functioned as a protective factor only when her partner was also employed; if an
employed woman was partnered with an unemployed spouse, her risk of violence
increased. This certainly suggests, as feminist theories would claim, that male violence is
a function of a masculine need for dominance in the relationship. This concept will be

revisited in the further discussion of status incompatibility.

Income. Theoretical connections between income and the perpetration of violence
are largely similar to those mentioned for employment, and the variables of income and
employment are often used in place of each other when including demographics in a
model. We can consider income an important financial variable for detecting more subtle
differences than employment or unemployment. The greater the income, for example, the
greater the prestige of the occupation, in reference to the afore-mentioned studies
highlighting discrepancies in occupational prestige between spouses. The greater the
income, the greater the stakes in conformity — a man of high status with a high-earning
position has more to lose than a lower-class man if his abuse is publicized. However,
income is often inextricably entangled with other socioeconomic factors and may serve as
a proxy for age, race, education or gender. In a culture where men have a higher average
income than women, we could argue that “batterers have higher incomes than their
victims” and simply be detecting gender differences rather than income effects. Finally,
income is the predominant factor in socioeconomic status, and a favorite predictor of
violence in the cadre of “drunken bum” (Kantor and Straus 1987) theories, other than

alcohol use.

Historically, income has been measured at the household level. Women of lower-
income households have been found to have a higher risk of abuse (Smith et al. 2002;
Bachman 2000). While some may take this to mean that income is a proxy for general
socioeconomic status, income has been found to have a significant negative relationship
with risk of abuse when controlling for other socioeconomic variables (O'Donnell et al.
2002). Based upon data from the National Family Violence Survey and the National
Crime Survey, Fagan (1993) argued that income’s effects are confounded with race and
urbanism. In central cities, spousal assault rates were highest for low-income groups.
However, regardless of household location, spousal assault rates were consistently over
thirty percent for the poorest of African American families. These rates do decline as
income increases, but only in suburban and rural areas. Rates remain high in central
cities, independent of income. For white couples, spouse-assault rates are inversely
associated with income in central cities but not elsewhere. Low income is also associated
with high rates of child abuse, though the strength of the association depends on the
composition of the family (Berger 2005).

It may be important, however, to separate female and male income in order to
better detect their relationship with risk of abuse. Exploration of female income as a
protective or risk factor has produced mixed results. Research has found that marital
dissatisfaction and wives’ proportion of household income has a relationship resembling
an inverted U-shape, with divorce rates being highest when wives contribute 40-50
percent of household income. When wives’ incomes is measured in dollars, the
relationship appears to be linear - the higher the wife’s income, the higher the likelihood

of divorce (Rogers 2004). Higher income and educational attainment appears to be a

10

protective factor against emotional abuse (Kaukinen 2004). However, in a meta-analysis
of common risk factors for spousal abuse, Hotaling and Sugarman (1986) found only
three studies measuring female income, none of which found a significant relationship
with husband-to-wife violence. Of four studies measuring male income, three found a
negative relationship between income and violence, and one found no significant
relationship.

Income also appears to be closely tied to women’s cycle of abuse. Researchers
have identified that the consequences of abuse and the presence of dependent children
impact a woman’s ability to earn, often resulting in her dependence on welfare. Welfare
reforms enacted in 1996 require that welfare beneficiaries work to receive support, a
demand that can seem impossible to meet when there is a lack of court-ordered child
support and few affordable childcare options (Riger & Staggs 2004). When evidence
suggests that economic dependency significantly influences women to remain with an
abusive partner (Barnett 2000; Scott et al. 2002), it seems unreasonable to enact welfare
reforms that increase a woman’s dependency on a breadwinning partner rather than
decrease dependency by providing childcare and empowering women to find and

maintain employment.

Status Incompatibility. Current evidence points to income discrepancy and status
incompatibility being a more likely risk factor for domestic violence than employment or
income alone. Women appear to be at a higher risk of abuse when they are employed and
their partners are not, a relationship that is not significant when employment of either

spouse is considered separately (Macmillan & Gartner 1999). Studies on status

11

relationships and marital satisfaction have found that a woman's greater status in relation
to her husband is associated with the husband's marital dissatisfaction (Homung and
McCullough 1981). The cultural depiction of the husband as breadwinner has supported
the greater rewards accorded to men in the workplace, legitimized male power within the
family, and provided men with a resource for demonstrating their masculinity (Ferree
1990; Stark and Flitcraft 1996). When this resource is removed, men may resort to
violence instead. Female employment functioned as a protective factor only when her
partner was also employed; if an employed woman was partnered with an unemployed
spouse, her risk of violence increased. This certainly suggests, as feminist theories would
assert, that male violence is a function of a masculine need for dominance in the
relationship. When economic dominance (through more prestigious employment) is no
longer possible, men may resort to physical violence and assert their dominance through
strength rather than employment status. Raphael (2001) argues that male dependency (a
product of male unemployment and female employment, in this case) causes shame,
which can then fuel violence. This argument has been supported with findings that
women who can at least equalize their ‘occupational prestige’ with their husbands (no
more, no less) may find a safe balance between being a threat to masculinity and being
economically dependent (Lambert & Firestone 2000). Research shows that women
making more money than their spouse are at a higher risk for abuse (Atkinson et al.
2005), and this risk increases as the portion of household income contributed by the
woman increases (Fox et al. 2002). Anderson (1997) found that the odds of male-
perpetrated domestic violence are approximately 40 percent lower when their female

partners earn less than 31 percent of the couple's total income. When women earn slightly

12

more (between 55 %-69%) of the couple's earnings, men have approximately 3.5 times
greater odds of perpetrating violence than men with earnings similar to their female
partners. This risk increases to over 5.5 times greater odds of male violence when women
earn 70 percent or more of the couple's income.

In today’s economic climate, where traditionally male-dominated manufacturing
jobs are disappearing, we may see women retaining employment where men do not,
increasing status incompatibility and ultimately heightening the risk of domestic
violence. In light of the above evidence, however, it is interesting to note that Farmer and
Tiefenthaler (1997) find that external and independent sources of income for women
(child support, government assistance, lawsuit payouts, family money, etc.) reduce
women’s risk and that greater economic equality for women will lead to a decrease in
violence. Perhaps female equality must strike a delicate balance between “not enough”

and “too much” when considering the contribution to household income.

CURRENT RELEVANCE

The 2008-09 economic recession has caused, and is causing, profound
consequences in the US. and beyond. Housing values have plummeted across most
western societies, by nearly 30 percent in one year in some localities (National Home-
Value 2009). Many global equities are now less than half their worth when compared to
just one year ago (World Markets 2009), and the US. unemployment rate has increased
nearly 45 percent in just two years, now exceeding the 1983 rate (United States
Unemployment 2009). Aside from these quantitative measures, many scholars and others

are now discussing how this recession is qualitatively different from recent dips and its

13

forthcoming contagious, harmful effects. Some of these contagious consequences have
already appeared, while others may not materialize for years or even decades (Robb
2008) (e.g., Elder and Liker 1982). Arguments are appearing in the news and elsewhere
contemplating that this recession is unlike any we have experienced since the “Great
Depression”, suggesting that this downturn will usher in structural changes in the global
job market that will last for decades, if not longer (Goodman & Healy 2009). Others have
argued that because this recession started with a decline in home values and an upswing
in their foreclosures, the one notable consequence to watch for is the decline -- not just
now, but for decades -- in the ability of inner-city communities and neighborhoods to
foster collective efficacy (Foreclosure Help Needed 2009). The concentrated, mass
foreclosures and the resulting systemic unoccupied housing has not only hit inner-city
neighborhoods, many of which have not dealt well with decades of decline in
manufacturing jobs, but also begun to negatively affect suburban and wealthy rural areas
(Wilson & Paulsen 2008). These latter areas are now beginning to wrestle with the
consequence of the rapid contraction of the white-collar job market (Hahn 2009). If these
areas begin to demonstrate patterns of high unoccupied housing with broken windows,
sidewalks, depleted streets and playgrounds, and an influx of new residents from
disparate cultural backgrounds, these communities may too demonstrate a rise in the sale
and use of controlled substance, incivilities, and violent crime that has long victimized
inner-city neighborhoods across the United States (Tuthill 2008). These factors are in
turn linked to increased risks for childhood depression, anxiety disorders, and aggression

(Gorman and Tolan 1998).

14

As of March 2009, the Internet and other widespread media sources are replete
with stores about the link between home foreclosures and the economic downturn and
incidents of family violence. There are numerous heartbreaking stories documenting
how the 2008-09 economic slump has not only caused families to lose their homes, but is
also implicated as a cause of familiar violence and the breakup of families. For example,
there is the 2008 suicide of an older husband and wife in Oregon that followed their home
foreclosure (Armour 2008). In Los Angeles, California, an unemployed man who once
worked for Pricewaterhouse and Sony Pictures murdered his wife, three sons and his
mother-in-law before turning the gun on himself. He left a suicide note saying that he was
in deep financial trouble and had considered killing only himself, but decided that it was
more honorable to kill his entire family before committing suicide (Karthik Rajaram Kills
2008). Just recently, Michigan’s I-696 highway experienced eight hours of stand-still
traffic while police negotiated with a suicidal man who was reportedly devastated by job
loss and family stress (Satyanarayana & Dinh 2009; Louwers 2009).

These anecdotes are likely not unique and may represent a pattern of increased
violence among family members. Across the United States (from California to Florida
and up to Massachusetts), family and mental health counselors are reporting substantial
increases in complaints related to money and mortgages and the resulting family violence
(Armour 2008). The Women Center in Stockton, California reported that during 2008
domestic violence reports increased 12 percent and “that the majority occurred in families
that are losing their homes through foreclosures” (Banks 2008). “Foreclosure, or the
threat of it, can destroy families, says the group’s executive director, Joelle Gomez. ‘The

housing market is still bottoming out and we are really bracing to see far more clients

15

utilizing both our shelter services and our help lines and counseling, she says” (Jensen
2008). Similarly, the National Domestic Violence Hotline (NDVH) reported that calls
were up 21 percent in September 2008 when compared to the September 2007 count.
They also reported that between mid-November 2008 and the end of the year, among
7,868 callers who agreed to participate in their study, 54 percent said that there was
change in their household’s financial situation in the past year, and 64 percent also
believed that “abusive behavior has increased in the past year” (Increased Financial
Stress 2009). This increased demand for help extends beyond domestic violence hotlines,
as reported by the Polk County, Florida crisis hotline in late September 2008. Their crisis
phone line reported a 30 percent increase in calls, with many of their callers complaining
that because of lost jobs they can no longer care for their children (Down Economy
Means 2008). Likewise, the San Diego Police Chief claims that because of the bad
economy, calls for domestic and alcohol-related crimes are increasing. He also reports
that that there are more cases of "identity theft, mortgage fraud, senior abuse, too —
people taking advantage of seniors, trying to get to their money" (Sullivan 2008). One
advocate has speculated that if the economic downturn is restricting once-functioning
couples’ ability to separate, we may now face a period of rising and often dangerous
relationships that more frequently turn to abuse rather than uncoupling (Bad Economy
Makes 2009; Scheerer 2009; At Risk of 2008).

In other locations, domestic violence shelters are feeling a surge in their housing
requests. In Spring Hillsborough, Florida, the number of women seeking shelter
increased by more than 100 percent, from 90-95 per month to 196 in October 2008

(Danielson 2009). Olvera Lighter, the Spring's president, believes that what is “driving

16

this [increases] is [the] increased stress over the economy," Lighter maintains that
"because domestic battery is about power and control, when you're standing on a
shrinking iceberg of what you can control, some people might tend to lash out"
(Danielson 2009). This increased demand for use and stay at shelters hurts in two ways:
while demand for their services is up, funding is down (Danielson 2009). This secondary
stress may be reduced somewhat by the $50 million added to the Federal Violence
Against Women Act (VAWA) Transitional Housing program through the 2009 American
Recovery and Reinvestment Act (i.e., President Obama’s Stimulus plan) (VAWA

Transitional Housing 2009).

WEAKNESS OF PRIOR RESEARCH

The first methodological weakness is that the vast majority of the studies linking
economics to family violence use cross-sectional designs which ask about violence and
income at the same time. It is possible that repeated violence may cause some
respondents to report less income than they once could because the violence led to lower
wages, fewer work hours and/or more physical or mental health problems. Another
possibility, as indirectly suggested by Gottfredson and Hirshi (1990), is that the negative
correlation between social capital and violence that is detected in the cross-sectional
studies is more likely driven by a third variable, self-control. Accordingly, the more self-
control one possesses, the more likely he or she is to be employed or have higher income,
less likely to take on a risky subprime home loan, and less prone to use violence to

achieve his or her goals.

17

A second reason for delaying acceptance of the economic-violence casual process
is that the few longitudinal studies that link variables such as employment to violence are
weak because they rely on stagnant employment data (c.f., (Catalano and Dooley 1993;
Fox et al. 2002). In the literature uncovered for this paper, just two studies have tested
what happens in households or in the relationships when there is a change in one or both
partners’ employment statuses or incomes. If a man transitions from employment to
unemployment, does this lead him to begin using violence or to use it more frequently
than before? Alternatively, does an unemployed subject who becomes employed during
the study report less frequent use of violence, or do they completely desist? This latter
scenario is particularly important because some suggest that for women, an improvement
in employment status should reduce the likelihood of violence (e. g., (Farmer &
Tiefenthaler 2003). Some claim that as women gain resources through their employment,
they should become less dependent on their male partners, who should in turn fear more
that the women will leave them if they try to use violence to control the relationship.
These latter two scenarios are also consistent with predictions arising out of the “stakes in
conformity” and “the turning-point” models we reviewed earlier. Both theories suggest
that the male’s employment status is linked to legitimate or elicit bonds that in turn
protect against or promote the use of violence.

When considering economic variables, it appears to be extremely important to
measure not at the household or family level, but at the individual level, as status
discrepancies between spouses may be more powerful predictors of violence than income
or employment alone. We must also make a greater effort to control for other

demographic variables that confound the effects of income and employment, such as

18

race, age, and education. Finally, the body of literature reviewed herein offers no
evidence about the connection between economics and male intimate violence
victimization; though there is substantial evidence to suggest this victimization often
occurs.

Besides these methodological flaws and aforementioned sampling biases, today’s
employment environment may well represent something qualitatively different than it did
in 2001 and is certainly different when compared to the years before the 19803; there is a
great likelihood of generating false positives by generalizing from past research to today.
While it is difficult for us to set aside the painful stories noted earlier, there are several
contemporary employment facts that make the risk of generalizing older studies quite
clear: for the most part, women are participating in today’s workforce more so than they
ever did before, and less of them are unemployed when compared to men (Global
Employment Trends 2008). This upward employment trend began in the early 19805
(Howe 1990), and by the end of 2008 women represented 49 percent of the entire
workforce (Mulligan 2009). Today, nearly 20 percent of married women earn more than
their husbands (Labor Force Participation 2008). Furthermore, it is expected that, in the
near future, more women will work than men. This gap is the product of a shift in the US
economy away from manufacturing jobs, traditionally filled by men, to a service
economy in which women play a much larger role (Gavin 2008). In February 2009, the
female unemployment rate had risen to 6.7 percent, but this rate is about 20 percent lower
than the male unemployment rate (6.7% vs. 8.1%) (Kirchoff 2009). This is due to men
accounting for 82 percent of job losses in the first 13 months of the recession, starting

December 2007 (Rampell 2009).
19

This structural employment shift offers both positive opportunities for women and
the potential for negative consequences. While women in 2009 generally produce and
control more resources than ever before, they still do not make as much money as men,
and therefore must continue to manage the burden that this gap places on them. For
instance, women are now more likely than ever before to be the sole breadwinner in the
family, yet their ‘bread’ is not as sizeable, nor does it come with as many fringe benefits
as what their husband’s jobs once provided (Warren 2006). Given both the symbolic
meaning of this tum-of-events, and the practical problems caused by women continuing
to contribute a disproportionate amount of household labor, this period could cause more
stress and strain than previously experienced by families. At the same time, their
experience in the labor force may also bring them a better understanding of their
husbands’ unemployment plight and help them to adjust to their new role, a process that
is in stark contrast to what happened in families facing unemployment of the male partner
as recently as the early 19905 (Nordheimer 1990). For example, a woman interviewed in
2009 by the New York Times, who started working after the her husband lost his job for
the first time in the early 20005, claims that “[t]hings are not happy in the house ifI
blame him all the time, so I don’t do any of that anymore, I know he is doing his best”
(Rampell 2009). At the same time, men are reporting that they are responsible for more
household tasks. The combination of these two changes may lessen today’s strains when

compared to those once felt when women became the sole income producer.

20

DECISION TO REPRODUCE A PRIOR STUDY

The proposed study is a reproduction and up—to-date extension upon a previous
study. In 2002, Fox, Benson, De Maris and van Wyk used data from the 1990 Census and
Waves 1 and 2 of the National Survey of Families and Households to produce their
article, “Economic distress and intimate violence: Testing family stress and resources
theories”. They, as I do now, sought to explore the relationship between measures of
economic status and the incidence and frequency of family violence. This study is one of
very few to measure the effect of financial strain on rates of violence, as opposed to
simply using static measures of employment and income levels. The variable of financial
strain is particularly important, as it is a result of the balance between resources and
needs, which will vary within families and within socioeconomic classes. For example,
while the breadwinner of a family may be employed, his earnings may not satisfy the
needs of his family, producing stress within the family unit and heightening the
likelihood of violence. Fox and colleagues attempted to explore this relationship by
modeling not only income and employment variables, but also measures of job
satisfaction and status incompatibility within the family (e. g. spouse’s happiness with
partner’s work hours, job satisfaction, etc.). Given the rarity of longitudinal studies on
this topic and the high relevance of the study to today’s economic climate, I felt it
worthwhile to attempt to reproduce and extend upon their results.

The Fox et al. (2002) analysis revealed that each spouse’s employment alone had
no significant impact on violence, nor was there an interaction effect, in contrast to prior
studies (Macmillan & Gartner 1999; Kalmuss & Straus 1982). The researchers did find

that female partners’ desire for their husbands to work more hours increased their risk of

21

man-to-woman violence by 133%; when males desire that their female partners work
more hours, the risk of violence is almost doubled. The risk of violence was also higher
for women whose work leaves them irritable and tired, who feel that not working is not a
viable option, and who work in lower—level, blue-collar jobs. While the analysis found
that this was also true of men, the researchers did not find any effect when considering
each partner’s contributed share of household resources. As mentioned in the preceding
paragraph, measures of income-to-needs ratios showed that violence is decreased when
the income-to-needs ratio is high (odds ratio = .929), debts are lower (odds ratio = 1.273),
and when the male earns a larger share of household income (odds ratio = .670), as is
supposedly common in higher-income families. Fox and colleagues report that for each
unit increase in a woman’s sense of financial well-being, her risk of violence falls 36
percent.

To extend upon the original analysis, I will add a second dependent variable
measuring female—to-male violence. Studies measuring female violence attract much
controversy from scholars who point out that the overwhelming majority of domestic
violence victims (measured by their presence at clinics, shelters, and hospitals) are female
(Berk et al. 1983; Dobash et al. 1992). Others cite evidence from nation-wide surveys
that females are just as likely to engage in violence as their male partners (Gelles &
Straus 1988; Steinmetz 1977-78; Straus et al. 1980). This debate has largely centered on
the use of the Conflict Tactics Scale (CTS). Critics argue that the CTS does not provide
information about the context in which violent acts occur (initiation, intention, history, or
pattern of violence) and therefore risks misrepresenting the nature of intimate partner

violence (Flood 1999).
22

In the National Survey of Families and Households, the questions about
intimate partner violence resemble the CTS but are not nearly as thorough. There is no
measure of psychological, verbal or emotional abuse. The questions are “How often in
the past year did you and your partner end up hitting or throwing things at each other?”
and “Sometimes arguments between partners become physical. During the past year has
this happened in arguments between you and your partner?” The variables in this file
allow for measurements of male reports of perpetration, male reports of victimization,
female reports of perpetration, and female reports of victimization. For the dichotomous
dependent variables, “male to female violence” is comprised of the maximum value of
0/1 responses from males admitting perpetration and females reporting victimization.
“Female to male violence” is the inverse — the maximum value of on responses for
females admitting perpetration and males reporting victimization.

I chose to include the direction of violence in my extension for two reasons.
Firstly, I am interested in the presence of any violence at all. Whether female violence is
retaliatory or not, its presence contributes to an atmosphere of violence within the home.
Secondly, the theoretical perspectives I am exploring are not limited to the feminist
perspective. While issues of gender roles and status incompatibility certainly invoke
feminist theory, other economic and strain theories make no distinction between men and
women, presuming that both sexes are vulnerable to the economic forces that may
instigate or increase violence. Feminist and constructionist theorists may identify this
lack of distinction as a fault of these theories, as employment, income and financial strain
may have different implications for men and women. Unfortunately, the NSFH data does

not allow for an analysis of these complex issues.

23

METHODOLOGY

The Fox and Benson study used data from the National Survey of Families and
Households (N SFH). The following description of the methodology was provided by the
University of Wisconsin-Madison Center for Demography and Ecology, whose website
hosts the N SFH data.

The NSFH was first conducted during 1987-1988 and includes interviews with a
probability sample of 13,017 respondents. The sample includes a main cross-sectional
sample of 9,643 households plus a double sampling of blacks, Puerto Ricans, Mexican
Americans, single-parent families and families with stepchildren, cohabiting couples and
recently married persons. One adult per household was randomly selected as the primary
respondent. Several portions of the main interview were self-administered to facilitate the
collection of sensitive information and to ease the flow of the interview. The average
interview lasted one hour and forty minutes. In addition, a shorter self-administered
questionnaire was given to the spouse or cohabiting partner of the primary respondent.
Individuals, rather than families or households, are the units of observation.

An introductory letter was sent to each sample address. The letter provided
potential respondents with information about the survey and alerted the household that an
interviewer would visit their home. The interviewer visited the home and completed a
household screening form. Interviewers were instructed to conduct the screening
interview with a knowledgeable adult member of the household. Self-enumerated
questionnaires were provided to the respondents at some points throughout the interview,
facilitating the collection of data on sensitive topics and the use of multi-item scales. A

respondent payment of $10 was instituted on July 29, 1987 (about halfway through the
24

interviewing). N 0 payment was offered to secondary or tertiary respondents.
Coordinators could authorize interviewers to offer a payment of $15 to respondents who
had previously refused the interview. Interviewers were also given an incentive payment
of $10 for converting a refusal.

All data from the main interview were entered using the Direct Data Entry facility
of the CASES program. This program utilized the skip patterns in the interview and
numerous programmed checks to ensure that the data entry process properly captured all
data. The self-administered questionnaires were entered directly from the respondents'
answers.

The study was originally designed to be longitudinal, with hope for funding for
a follow-up after five years. Wave II of the N SFH received funding and data was
collected from 1992-1994. Technology at the second data collection period allowed for
the use of Computer Assisted Personal Interviewing (CAPI) technology with laptop
computers. This approach was combined with prior Computer Assisted Telephone
Interviewing (CATI) technology.

The CAPI sample included main respondents, 1987-88 spouse/partners, and
new spouses or partners of our main respondents. The main respondents were the 13,014
respondents who were interviewed in person during the original survey. Of the 1987-88
spouse/partners, 4,508 were still living with and/or married to the N SFH-I main
respondents. Another 791 were ex-spouses or partners, and 390 were deceased.

In addition to the NSFH and N SFH-II data, Fox and colleagues incorporated
tract-level data from the 1990 national census. The final sample consisted of Wave 1

couples (5,493), Wave 2 couples (4,538), and couples who were present at both waves

25

(3,680 of total). Domestic violence was operationalized by comparing responses by both
males and females. Fox and colleagues recoded the N SFH results into a dichotomous
measure where '0' indicated no violence and '1' indicated that either the man reported
being physically violent with the woman, the woman reported that the man was
physically violent with her, or both. In cases where partners' responses were
contradictory, the authors assigned a code of 1 (violence present).

The independent variables in the Fox paper are grouped into concepts of
employment experience, financial adequacy, and control variables. Employment includes
a dichotomous measure of current employment (1 = currently employed), job ranking
according to a 5-item census occupation classification scale (5 = executive/managerial, 1
= laborers/unskilled operatives), number of hours currently worked compared to the
number of hours the corresponding spouse desires, and job strain (exhausted, tense and
irritable after work). Financial well-being was measured by satisfaction with financial
situation, frequency of worrying about meeting expenses with current finances, and
personal debt (bank and personal loans, installment loans, credit card debt). The measure
of debt did not include mortgages or car payments.

Control variables include measures of race, age, and household composition.
Fox and colleagues also included measures of neighborhood concentrated disadvantage
by matching census tract variables to each case. Concentrated disadvantage was
measured with a scale including the percentage of single parents, percentage of non—white
residents, percentage unemployed, percentage of families on public assistance, and
percentage below poverty line. A correlation matrix of these variables can be found in

Appendix A.
26

REPLICATION AND REPRODUCTION ISSUES

The original plan for this study was to extend upon the Fox & Benson (2000)
effort by appending Wave IH of the National Survey of Families and Households (NSFH)
data to their combined Wave I and Wave II file. Upon beginning the necessary cleanup
and coding, however, it became apparent that the Wave III data was not suitable for my
purposes.

The Wave III data file is plagued with missing data, especially in response to
questions necessary for my analysis. All interviews were conducted over the telephone
using CATI (computer-assisted telephone interviewing) technology. The CATI system
used by the Survey Center is CASES. In the CASES CATI system, the text of the survey
appears question by question on a computer screen for the interviewer to read to the
respondent. Routing through the interview is based on skip logic pre-programmed into
the computer. While this method may be extremely useful, it is poorly translated into
codebooks. For Wave HI, there is no map or record layout file as was provided for Waves
I and H. This forces users to make assumptions about the positioning of screening or
gateway questions that limit the possible number of valid responses to each question. For
example, a simple crosstab of the question “Have you and your partner had a physical
argument in the past year?” reveals very few valid responses and a lot of missing data. To
be asked this question, respondents had to first answer in the affirmative to a previous
question about relationship status and then be divided into married vs. non-married. This
is just one example; the data is rife with confusing jumps and presumed filter questions

that leave the user full of doubt as to the reliability of the answer set.

27

For some questions, it is difficult to determine if there was a screening question or
if the question was asked at all. The question “Are you currently employed?” seems
straightforward and is vital to my analysis, and yet there are zero valid responses.
Elsewhere in the data set one can find measures of income and variables pertaining to
occupational status, retirement status, hours worked, and so on, but there is no
information suitable for a dichotomous measure of current employment. To simply
determine if a respondent is'employed or not, a user would have to combine the
dichotomous “Do you consider yourself retired?”, the scale “Number of hours worked in
the past week”, the interval “Reason you were absent from work last week”, the
continuous “Amount earned from employment this year”, and other vaguely
operationalized concepts. Though I attempted to combine these. measures into a single
measure of employment, I felt neither confident nor comfortable that my final estimate
was truly a reflection of the employment status of the respondents in the sample.

The flaws in the Wave HI data set may be attributable to both the pitfalls of a
longitudinal study design and to a lack of funding. The Wave IH Field Report reveals
that, due to “budgetary restrictions” only a subset of the original sample is included in the
third wave of data. Only parents of young adult children and respondents in mid to late-
life were included in Wave ID. The mid-to-later life sample was comprised of main
respondents who did not have eligible focal children but who were 45 years and older at
time 3. The Wave HI sample did not include new spouses or partners currently living
with the main respondent if different from the time 1 spouse or partner. These facts have
serious implications for my study. The exclusion of new cohabitating partners prevents

me from tracking a respondent’s violence over the life course as influenced by changing

28

economic factors. The advanced age of the sample decreases the prevalence of violence,
further reducing the number of valid cases available for analysis. Given these issues with

the data, I have made the decision to omit it from this study.

CURRENT STUDY

The data for this replication is available in two locations. The sections of data
used in the prior study are available on the Inter-university Consortium for Political and
Social Research (ICPSR). From the NSFH, the researchers abstracted data on conflict and
violence among couples, as well as data on their economic resources and well-being, the
composition of the household in which the couple lived, and a large number of socio-
demographic characteristics of the sample respondents. From the 1990 Census, the
original researchers abstracted tract-level data on the characteristics of the census tracts in
which the NSFH respondents lived. Demographic information contains each respondent's
race, sex, age, education, income, relationship status at Wave 1, marital status at Wave 1,
cohabitation status, and number of children under 18. Using variables abstracted from
both Wave 1 and Wave 2 of the NSFH and the 1990 Census, the researchers constructed
new variables, including degree of financial worry and satisfaction for males and females,
number of job strains, number of debts, changes in debts between Wave 1 and Wave 2,
changes in income between Wave 1 and Wave 2, if there were drinking and drug
problems in the household, if the female was injured, number of times the female was
victimized, the seriousness of the violence, if the respondent at Wave 2 was still at the

Wave 1 address, and levels of community disadvantage.

29

Upon downloading the data set and beginning my initial overview of the
information, it became apparent that the file available from ICPSR did not contain key
variables necessary for a reproduction of the associated paper. The file I downloaded,
ICPSR #3410, did not contain a dichotomous measure of employment or the age of each
partner. The file did contain several measure for other variables and it was not indicated
which of these had been used in the statistical models. For example, the “Concentrated
Disadvantage” measure occurred in continuous, trichotomous and dichotomous forms.
Swapping these different measures of the same concept in and out of the model produced
different odds ratios for victimization. I was able to contact Dr. Fox to ask about the
missing variables, and she was kind enough to send me a working data file that she had
saved during the original analysis. While the file did contain some of the variables I
needed, saving me from building them from the NSFH file, this new file had cryptic
variable names, no codebook, and no information on how these measures had been
created. In most cases, it was not clear which variables from the file had been included in
each model; often, there were multiple variables for one concept and no variables for
another. The results reported below reflect my trial-and-error experimentation to find the
variables that produced the most similar findings.

Finally, there was some confusion about the sample size, as I could not seem to
capture the same N used in the original models. For example, consider the variable of
“Male age” used in the Fox models. The file I downloaded did not have such a variable,
only “Age of Respondent”. As mentioned in the methodology section, the main
respondent for each household was a randomly selected adult, and could thus be male or

female. To construct a variable measuring male age, I returned to the original N SFH data

30

and captured the measures of respondent age, respondent sex, partner age, and partner
sex. I then created several “if-then statements” to convert the ages of respondent and
partner to the ages of males and females. The ‘Age of Respondent’ variable from the Fox
file has 3,682 cases aged 65 and under. My constructed measure of male age has 3,344
cases aged 65 and under. Already, there is a discrepancy of more than 300 cases. This
kind of discrepancy would continue to be a problem throughout the analysis, resulting in
sample sizes that often differ by several hundred cases. The implications of this

difference are discussed in the conclusion of this paper.

 

 

Results of Univariate Analysis as Reported in Fox et
al., 2002

Wave 1 Wave 2

N = 5,493 N = 4,538
Age (Male) . 36.2 43.7
Age (Female) 34.3 41.2
Years of Education (Male) 13.4 13.2
Years of Education (Female) 13.2 13.0
% employed (Male) 90 90
% employed (Female) 64 71
% Black 12 12
% Hispanic 6.6 6.6
Income-to-Needs Ratio 4.45 4.69

(SD 4. 72) (SD 4.52)

% Male-to-Female Violence 9.1 6.2

 

 

 

MULTIVARIATE DIAGNOSTICS
Collinearity statistics were assessed for all multivariate models to search for

potentially confounding interrelationships among the independent variables. N o
problematic variance inflation factors (VIFs) were present in any model, as they ranged
between 1.00 and 1.77. The original article states only that multicollinearity diagnostics

were performed and that the results were not problematic.

31

RESULTS
The following tables display my reproduced findings alongside the original

findings by Fox and colleagues. The secondary tables (marked by an ‘a’, e. g. 2a) show
the results of the analysis when considering female-to-male violence as the dependent
variable. The original and secondary tables are displayed together to facilitate comparison
of the similarities and differences in the correlates of intimate partner violence when

controlling for direction of violence.

32

 

 

Her work
tatus
is work
tatus
er work
tatus x his
ork status
er
reference
hat he work
ore
is preference
hat she work
ore
er work
tatus x his
reference
hat she work
ore
is work
tatus x her
preference
that he work
more
-2LL

Model X2

Block X2
N

Table l: Logistic Regression of Male to Female Violence onto Employment Status Predictors, Waves 1 and 2

 

Fox et al. 2002 Current Study
Wave 1 Wave 2 Wave 1 Wave 2

Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Mode12

0.97 1.14 1.23 1.70 2.06 0.74 1.71 1.59

0.81 1.81 0.7 0.93 1.56 0.84 1.73 1.79

1.56 - 0.95 - 034* - 0.79 -

- 2.33*** - 1.91*** - 1.17 - 1.76

- 1.33* - 1.95*** - 0.79 - 1.21

— 0.78 - 0.74 - 1.49 - 1.08

- 0.60 - 1.34 - 1.19 - 1.45
3067.99 2712.32 1665.54 1528.84 1450.8 1217.86 1109.87 1095.49

5.40 45.88*** 4.02 33.52*** 0.073 5.36 7.34 21.73***

5.40 45.88** 4.02 33.52*** 0.073 5.36 7.34 21.73***

4940 4377 3801 3557 3423 3423 3180 3180

 

 

 

*pS.05,**pS.01,***pS.001

 

 

Table la. Logistic Regression of Female to Male Violence onto Employment

Status Predictors, Wave 1. Only

Her work status

His work status

Her work status x his work status
Her preference that he work more
His preference that she work more

 

Her work status x his preference that she work more -

His work status x her preference that he work more

-2LL
Model X2
Block X2
N

Wave 1
Model 1 Model 2
1.06 .98
1.09 1.07
.91 -
- 1.25
- 1.26
1.17
- .85
1450.80 1437.98
.07 4.38
.07 4.38
3423 3398

 

 

* pf .05, ** p s .01, *** p5 .001

 

 

33

 

Employment (See Tables 1 and 1a)
I failed to find the same levels of significance as reported in the Fox & Benson

study. In my reproduction, only the interaction effect of female work status and male
work status was significant in Wave 1. My Wave 1 model odds ratios vary vastly from
the original study. This suggests that perhaps the variables labeled in the data file as
measures of employment status were not the ones used in the original analysis. I also
failed to reproduce the original sample size, which I attribute to missing data. In the
original article, the authors do not address their treatment of missing data, which is a
major issue when using the N SFH. I chose to drop cases with missing answers, resulting
in smaller sample sizes in most cases. This model was the most difficult to reproduce.
When observing the effects for the same variables when measuring female
violence, there were no significant associations. It was only possible to analyze Wave 1
for this model, as there were apparently no cases with valid responses for both the
measure of female employment status and perpetration of violence. This would prove to

be a common problem throughout my extension.

34

 

 

 

 

 

 

 

 

Table 2. Male to Female Violence, Wave 2
Fox et a1 (2002) Current Study
Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Herjob strain l.22*** - 1.21 1.l6** - 1.13*
She agrees both must 1.23 ** - 1.19 1.20* - 1.14
work
Her share of couple 1.43 - - 0.66 - -
earnings
Her job ranking 0.87 * - 0.92 0.87 * - 0.84 **
Hisjob strain - 1.21*** 1.25 - 1.15** 1.15*
He agrees both must - 1.13 * 0.93 - 1.11 0.99
work
His share of couple - 0.74 - - 0.54 -
earnings
His job ranking - 0.87 *** 0.92 - 0.85 ** na
-2LL 772.64 1598.98 603.67 878.51 1066.82 778.28
Model X2 31.98 *** 48.82 *** 34.90 ** 24.81 *** 30.91 *** 27.75 ***
Block X2 31.98 *** 48.82 *** 34.90 ** 24.81 *** 30.91*** 27.75 ***
N 1519 3666 1300 2153 2722 1896
* p S .05, ** p S .01, *** p S .001, na Too few cases for analysis
able 2a. Female to Male Violence, Wave 2
Model 1 Model 2 Model 3
Her job strain 166* - 1.53
She agrees both must work 0.90 - 1.19
Her share of couple earnings 0.91 - -
Herjob ranking 0.81 - O.57*
Hisjob strain - 1.1 1 0.92
He agrees both must work - 0.75 0.64
His share of couple earnings - 0.18 -
His job ranking - 064* no
-2LL 108.14 113.74 81.11
lModeI X2 8.42 7.05 1 129*
Block X2 8.42 7.05 1 1.29*
N 2153 2722 1896

 

 

* p S .05, ** p S .01, *** p S .001, na Too few cases for analysis to run (n = 9)

 

Job Strain (See Tables 2 and Za)
The odds ratios reproduced here were much closer to those in the original

analysis, with only a minor decrease in significance. In this set of models, my closest

35

 

approximation of sample size was occasionally larger for Models 1 and 3, but smaller
than the 2002 study for Model 2. As discussed above, trial-and-error manipulation of the
sample did not remedy this problem.

The effects of male and female job strain remain significant in the final model for
both male and female violence, with higher strain associated with greater prevalence of
violence. Job ranking is also a significant predictor of violence, with less violence
occurring when male and female job rankings are high. Percentage of household income
contributed was not significant.

For female-to-male violence, only female job ranking remained significant in the
final model. As female job ranking increases, the prevalence of female-perpetrated

violence decreases. This is true for both male- and female—perpetrated violence.

 

Table 3. Logistic Regression of Male-to-Female Violence onto Household Financial Status Predictors

 

Fox et al. 2002 Current Study
Wave 1 Wave 2 Wave 1 Wave 2

Model 1 Model 2 Model 1 Model 2
Income/Needs Ratio 0.93*** 0.87*** 0.93* 1.00* 0.89*** 0.95
Debts l.27*** 1.15** 1.06 1.04 1.19** 1.14
His share of couple 0.67* 0.60* 0.69 0.96 0.71 0.76
earnings
Her financial well- - - 0.64*** - - 0.71**
being
His financial well- - - 0.81* - - 0.81
being
-2LL 2686.09 1849.12 1802.67 1230.7 1255.81 1 167.96
Model X2 58.07*** 48.7*** 95.15*** 4.57 26.34*** 27.67***
Block X2 58.07*** 48.7*** 46.45** 4.57 26.34*** 27.67***
N 4467 3885 3885 3464 3199 2987

 

 

*pS.05,**pS.0|,***pS.OOI

 

 

36

 

 

 

Table 3a. Logistic Regression of Female-to-Male Violence onto Household Financial Status Predictorsl
Wave 1 Wave'2
Model 1 Model 2

Income/Needs Ratio 1.00 0.72* 0.81
Debts 1.38*** 1.25 1.13
His share of couple earnings 0.90 - 0.43 0.54
Her financial well-being - - 0.82
His financial well-being - - 0.54
r2114 1446.35 148.50 144.18
Model X2 26.23*** 7.87* 4.32
Block X2 26.23*** 7.87* 4.32
N 3464 2987 2987
* p S .05, ** p S .01, *** p S .001

 

 

 

Household Financial Status (See Tables 3 and 3a)
While the odds ratio for the Income/Needs Ratio measure is greater in my Wave 2

full model than in the original analysis, it fails to retain significance. While the same
measure is significant in Wave 1, it has an odds ratio of 1.00, indicating absolutely no
increase or decrease in violence as a couple’s income/needs ratio changes. The number of
debts measure also loses significance in the Wave 2 full model. The odds ratios are very
close to those in the original study. While the male’s share of household earnings was
significant in the original study and not in my reproduction, my odds ratios for this
measure are actually larger. I attribute this loss of significance to the difference in sample
size. Female financial wellbeing was a significant predictor in both the original and
reproduced models, with greater scores on this measure leading to lower prevalence of
violence. Male financial wellbeing was significant only in the original study, despite
having exactly the same odds ratio in my reproduction. Again, this may be attributable to

the difference in sample size. My sample size is smaller across all models.

37

Wave 1 female-to-male violence was significantly more prevalent in couples with
more debt, though this relationship was not found in the Wave 2 models. The full model

for Wave 2 did not reveal any significant relationships.

38

 

 

 

 

 

 

 

 

 

Table 4. Logistic Regression of Male-to-Female Violence onto Change in
Household Economic Status Predictors for Continuing Couples
Fox et al. 2002 Current Study
Model 1 Model 2 Model 1 Model 2

Wave 1 Violence (1 = yes) 11.54 *** 11.87 *** 2.17 ** 2.13 **
His share of couple earnings is smaller in - 1.31 * - 1.15
Wave 2

His job spells between waves — 1.48 *** - 1.44 ***
Her job spells between waves - 0.90 - 1.02
Increase in debt load at Wave 2 - 1.02 - 1.12
Income/needs ratio smaller at Wave 2 - 0.68 * - 0.67 *
-2LL 1 136.19 1007.67 1314.74 1293.33
Model X2 148.07 *** 176.59 *** 5.87 * 27.28
Block X2 148.07 *** 28.52 *** 5.87 * 21.42
N 3006 3006 3491 3464
*ps.05, **pS .01, ***pS.001

Table 4a. Logistic Regression of Female-to-Male Violence onto Change in

Household Economic Status Predictors for Continuing Couples

Current Study
Model 1 Model 2

Wave 1 Violence (1 = yes) 6.23*** 9.85***

His share of couple earnings is smaller in Wave 2 - 0.47

His job spells between waves - 1.75**

Her job spells between waves - 0.82

Increase in debt load at Wave 2 - 1.34
Income/needs ratio smaller at Wave 2 - 0.65

-2LL 287.39 178.11

Model X2 12.06*** 2590*“
Block X2 12.06*** 25.90***

N 3682 3464

* p S .05, **p S .01, *** p E .001

 

Change in Household Status (See Tables 4 and 4a)

These results are the most relevant to the current economic crisis and the shift in

status I seek to explore. As expected, Wave 1 violence is the best predictor of Wave 2

violence, though my reproduced model shows a vastly lowered odds ratio (11.87 vs.

2.13). The number of male job spells (periods of unemployment greater than six months)

between Waves 1 and 2 was a very significant predictor of violence. Counter-intuitively,

both the original study and my reproduction found that a smaller income-to-needs ratio

39

 

 

(indicative of a tighter budget) decreased the odds of violence. The odds ratios for this
measure differed by only 0.01 when reproduced.

Female-to-male violence in Wave 2 was again predicted by Wave 1 violence,
with an odds ratio much more similar to that in the original analysis of male-to—female
violence. Chronic job spells also significantly increased the odds of female-to-male
violence (odds ratio = 1.75).

Final Model (Table 5, next page)

Only in the final models did Fox and colleagues include the traditional control
variables of race and class (using education as a proxy for social class). The final model
shows that neighborhood disadvantage, multiple children in the household, low male age,
his preference that she work more, and female financial well-being retain significance
when controlling for individual, household and dynamic predictors. Neither the final
model nor the foreclosure model (next section) could be conducted with male
victimization as the dependent variable, as the number of male victims was far too small
for any meaningful analysis. In the reproduction of the final model, only neighborhood
socioeconomic disadvantage, male age, and the number of children in the household
remained significant. Curiously, none of these variables were mentioned directly in the

theoretical overview of either this paper or the original study.

40

 

 

 

0_.0_ ** 0w.n_ .1..." 0—80 5.2.. 5”: Nod “In." Suva 3:." 3.3“ “I”... 00.05 *3. ON: *3. 5.:
“I..." omdo. *5." 5000 .12.. hmdw "I" wfifl Nod .22.. 5M”: *3. 33:»— 1:." $0.:— uli thm .12.. 2.:
no.0; 3 . .Na vodmo m0.§~ 0m.w_0_ _ _.NON_ owé _ m. 0mémm— omdmv— 00va
and mwd
05.0 *1, 00.0
00.0 :0
m0._ 00.0
00.— N0.—
whfi mvm. OWN 2N
no; .30 mwd 2.0
cm; 05— “.23.. n0; .3. 00.N
wvd vmd * mm; mo;
mmd 0m.0 ovd Nvd
0*”.— vc; * E..— *** no;
“In." no.0 "I...” No.0 3.3.. no.0 .12.. no.0 0.3.3." no.0 “3.3.. no.0
.3.” mN._ "In NN._ *2; *3. w_._ flue; *o_.—
c0.0.0 00.0 no.0 V0.0 * No.0 in 00.0
00.0 2.0 00; v0.— 30.— m0.—
* E..— ** 05— ** hm; as; 00.N “I..." WON 0.3.3." O_.N .3: CNN “Ii hNN
find nmd and #50 m _ ._ 00.0 5.0 no.0 No.0 om.—
ho._ 0_._ ON; m_._ * mm; _N._ am; _m._ mm; is." Edd
n 3002 v BUG—2 m Etc—z N 3002 _ 3002 m Eco—2 v 3002 m BUG—2 N 3002 _ 3002
$th H 20 beam E00030 ANONM H 20 NOON .E 00 xom

 

_00.WQ is." ._0.WQ** .w0.WQ*
NR «.35

NR 3ch

EN.

9069:?» 3050:: mi
MERE—03 30:2: .0:

$5500 03:00 .8 0.8.7. 0E
050n—

ocfi $005088:—

058 0:03
0; 000000.003 00; x 0380 0:03 mi

0.5:. 0:03

0% 00500003 mi x 0550. 0:03 .03
05E 0:03 0:0 005 08000005 mi
05.: 0:03 0: 005 00:830.:— 003
0380 0:03 mi

0230 x53 00m

own mi

23030: E 5:220 00 095.2
5:00:00 0:.—

=ocmosv0 00$

0mmEm>ua£D

0_Eo=00020om voofionzwfiz
2:805

x005

001on N 0235 08 £20605 289.com 0:0 28:50 2.5 8:065 208099-202 .00 550%“;— 0:m_w3 .m 030E

 

 

41

Tables 6 and 6a: Foreclosure-Specific Models
In light of the 2008-9 economic crisis and the rising foreclosure rate, I have

developed a model using variables specifically related to current events in hopes of better

understanding the future consequences of the mounting financial strain.

 

 

 

 

 

Table 6: Logistic Regression of Male-to-Female Violence onto Foreclosure—Specific Predictors

Model 1 Model 2 Model 3
Wave 1 Violence 2.06* 2.02* 1.92*
(SE) .32 .32 .32
Change in Residence 1.48* 1.36
(SE) .17 .17
Increased Neighborhood Disadv. 1.35”” 1.20
(SE) .10 .10
More Debts 1.21 1.06
(SE) .18 .18
Smaller Income/Needs Ratio 0.70 066*
(SE) .20 .20
F—Iis financial stress - 080*
(SE) .12
His job spells 1.30“
(SE) .09
Her financial stress 0.67***
(SE) .12
Her job spells 0.98
(SE) .1 I
~2LL 1212.9 1191.91 1147.76
Block X2 4.44* 2099*“ 4415*”
Model X2 4.44* 25.43*** 69.58***
*ps.05, ** p501, ***pS.001

 

The results of the foreclosure model are far more relevant to the current economic crisis
than those in the final model. Here, while changes in residence and neighborhood
disadvantage (consistent with moving into a new area after home foreclosure) are
significant in Model 2, these variables lost significance upon the introduction of

individual-level measures of job spells and subjective financial stress. Counter-

42

intuitively, more stress predicted less violence. As in previous models, Wave 1 violence

remained the best predictor of Wave 2 violence.

DISCUSSION

While this study was originally intended to reproduce the analysis of Fox and
colleagues, it has become equal parts a reproduction and a study of scientific
reproduction itself. As these are both important and currently relevant issues, I will first
discuss the findings of the reproduction and then continue on to discuss the difficulties of
reproducing a study from the information published in a scholarly journal article.

As mentioned in the overview of results, very few variables relating to
employment status and job strain were found to significantly impact the likelihood of
violence. Instead, the majority of significant variables were related to household and
neighborhood measures of economic distress and disadvantage. Unlike the original final
model, the reproduced final model (Table 5) shows only neighborhood disadvantage,
household composition, and male age to be significant violence predictors. None of the
correlates of male-to-female domestic violence were significantly different from those of
female-to-male violence, suggesting that the characteristics of violent households are the
same regardless of who perpetrates the violence. These results suggest that interventions
should target low-income homes in disadvantaged neighborhoods by helping to alleviate
worries about meeting necessary financial burdens. For example, public assistance for
food stamps, education and children’s material needs could reduce parental stress over
meeting their families’ needs with an insufficient income. This is especially necessary for

young families with multiple children, as higher numbers of children and lower male age

43

were both significantly related to increased risk of violence. When considering the effects
of change in economic status on the likelihood of domestic violence, it appears that high
individual subjective strain can act as a protective factor, perhaps encouraging couples to
cling together and suppress violence to procure or retain financial benefits. This finding is
counter—intuitive to many theoretical approaches that would predict that strain and tension
would increase partner violence. Finally, chronic male unemployment was a significant
predictor of violence in both the Fox study and my reproduction, including my
foreclosure-specific model. Unfortunately, this is difficult to address through public
policy, as lobbyists and advocates may be reluctant to offer jobs (perceived as a reward)
to batterers. In the current economic crisis, many jobs are not available at all, even to
highly qualified candidates. It is difficult to imagine a solution to the ups and downs of
the capitalist economy, despite their apparent influence on family violence.

The variables available for this reproduction allowed for only partial tests of the
many theoretical approaches reviewed in the first chapter of this paper. Rather than
address each theory by name, I will group them into themes for this discussion: ‘stakes’
theories, ‘strain’ theories, and ‘power/gendered’ theories. With the small amount of
information available, it appears that an integration of these three themes would be the
best approach to studying the effects of economic distress on family violence. The
imbalance in power (usually assigned by traditional gender roles) and financial strain
increase the likelihood of violence, while higher stakes in conformity reduce the risk.
Employment alone did not account for significant increases in violence, though this may
be explained by confounding effects of the frustration-aggression hypothesis. Employed

individuals may restrain themselves from violence out of fear of unemployment, or they

44

may be experiencing job strain that increases their violence perpetration. This argument
is partially supported in Table 2, where job strain increases the odds of violence and
higher occupational prestige acts as a protector. While we may associate higher
occupational prestige with longer hours worked and increased pressure on the job, there I
is likely a tipping point somewhere on the scale where higher occupational prestige
equates to less strain, less pressure, and less need for long hours. Finally, an unfavorable
change in household indicators (as we expect to see in an economic crisis) supported the
strain hypothesis, with some contradictory findings. While increased job spells and
moving to a poorer neighborhood increased the odds of violence, a smaller income-to-
needs ratio (indicating less income to cover basic costs) reduced the odds of violence.
The significance of this finding alone should prompt further research on the intricate
relationship between strain and violence: Are there some forms of strain that encourage
couples to ‘make it work’ and stay together? This finding would certainly suggest 50.
Overall, the evidence certainly supports strain theories, with stakes in conformity acting
as modifiers of behavior, and some gendered effects on economic variables that prevent
scholars from making blanket statements about the effects of income and employment on
all parties. These variables must be studied separately for each partner, and then as an
interaction.

Of some concern is the consistent significance of neighborhood measures of
disadvantage. Fox and colleagues utilized logistic regression as their method of analysis,
while I argue that a hierarchical approach would be better suited to this study. The final
model mixes individual (financial well-being, job strain, age), household (debts, number

of children, income-to-needs ratio) and neighborhood (concentrated disadvantage) into a

45

single regression equation. It would be preferable to perform a nested analysis of
individuals within homes within census tracts in order to better understand and untangle
the co-morbid effects of living in economically distressed areas. It is difficult, then, to
make inferences about the true effect of neighborhood disadvantage, as the current
regression models do not appropriately treat the different units of analysis. However, in
many cases hierarchical modeling is not feasible. It would be a worthwhile effort to
attempt such an analysis of this data set.

Future inquiries would benefit from a more specific, sensitive methodology.
While national surveys are invaluable for their breadth and representativeness, their very
size and cost means that they must be of use to researchers studying many different
topics. In the interest of keeping the survey manageable and timely, there is a focus on
breadth rather than depth. Using the NSFH to answer questions about economics and
domestic violence forces the researcher to depend on very few measures of these
concepts, and there is some concern that the presentation of these concepts is too
simplistic and may conceal valuable information. Consider, for example, the concept of
employment. Dichotomous measures of employment (employed/unemployed) are
valuable if a theory supposes that the very state of being unemployed increases one’s
aggressive tendencies. This measure conceals the subject’s true feelings about their job
status: is the individual unhappily unemployed? Is this an individual who feels it is
important and socially desirable to be employed? The effect of employment may be
greater than we have seen in prior studies, because those who desire to be employed are
included in the ‘unemployed’ category alongside those who do not place much emphasis

on employment or conformity. Is it possible that a subject’s desires and notions of what

46

should be are more important than his current situation? Only a more sensitive survey
instrument can tell. A survey designed to specifically test the theories presented in this
paper would likely be far smaller in size than the N SFH (due to time and budget
constraints), but could better detect the subtle nuances of status and social norms.

If’we measure the success of a reproduction by obtaining similar odds ratios and
consistent directionality, then this reproduction has been only partially successful. For
many models, the odds ratios reported by Fox and colleagues were reproducible. When
not exactly reproducible, the findings at least tend to be in the same direction (either
increasing or decreasing odds of violence). However, there have been some troublesome
results, especially in the first set of models. The variables measuring the respondent's
preference that his/her spouse work more hours had vastly different effects in my
reproduction. For example, in Table 1, Model 2 for waves 1 and 2 (and again in the final
model, Table 5), Fox and colleagues found that one partner's desire for the other to work
more hours significantly increased the odds of violence. My reproduction, in contrast,
found that at Wave 1, his preference that she works more actually decreased the odds of
violence, though this finding (as with others related to this set of variables) was not
significant. In other places, I found higher odds ratios that were not significant where Fox
and colleagues found lower ratios with very high significance.

It seems that the answer to these discrepancies lies in the sample size of each
model. As mentioned above, reproducing the sample size proved to be one of the most
difficult aspects of this analysis, often leading me to (trial-and-error fishing expeditions
using the very scant information provided in the journal article. Too often, we place great

emphasis on results that test significant at the .05, .01 or .001 levels. These are the results

47

that we look for, remember, and hold on to as we add to the body of knowledge. When
the result is not significant, researchers conclude that there is no difference between
groups. When the results are significant at some level, we consider the research question
answered and then proceed to use this significance to argue for the superiority of one
theory over another. In the social sciences, this threshold is set at .05, regardless of
sample size. As Royall (1986) illustrates, interpretation of statistical significance is not so
simple. He cites Peto et al. (1976), who argue that a given P-value in a large trial is
stronger evidence of true difference than the same P-value in a smaller trial, and Lindley
& Scott (1984), who argue the opposite: the P-value is "more indicative of the falsity of
the null hypothesis with a small sample than with a large one" (p. 3). Royall concludes
that both statements can be true. Often, significance test results are presented simply as
superscript symbols relating to values <05, <01, and <.001. Royall finds that the
assertion by Peto et al. is only true when considering the P value as a "zero-one variable
that simply shows whether P < .05 or not", just as we do when reviewing tables of results
in journal articles. Otherwise, the Lindley & Scott argument that the P-value is indicative
of stronger evidence of difference when using a smaller sample. In the reproduction of
social science, it would be preferable to supplement the significance test with attention to
the regularity of findings in support of a given theory.

In this study, differences in sample size may be attributed to the treatment of
missing data. Fox and colleagues make no mention of missing data in their journal article,
and yet missing data was another difficult part of this reproduction. In a survey as large
as the NSFH, missing data is inevitable, and its treatment can have a significant impact

on the analysis results. In many models, I believe that my sample size is smaller because I

48

conducted complete-case analysis, discarding cases with missing values for variables of
interest. While this approach has been criticized as an unnecessary waste of information,
complete—case analysis is preferable to methods that use incomplete cases improperly
(Little & Schenker, 1994). If Fox and colleagues replaced missing values with
unconditional sample means, the result would be an inflated sample size. Inferences
based on unconditional mean replacement are "seriously distorted" (Little & Schenker,
1994: 45) by bias and overstate precision. In this case, I prefer complete—case analysis to
unconditional mean replacement. The initial NSFH sample is sufficiently large to allow
for some ‘wastage’ by discarding incomplete cases, leaving a complete sample of several
thousand cases contributing to less-distorted inferences.

In conclusion, we must exercise immense caution when making inference and
advising policy from the results of quantitative analysis such as is presented here. As
demonstrated, seemingly minor adjustments can cause major changes in significance
levels and odds ratios. Before using published analyses as guidance for policy decisions,
efforts should be made to reproduce and replicate the relevant results. Using
representative samples of varying sizes, one would expect to find at least similar odds
ratios and directionality. Here, however, there were several instances where ratios and
direction could not be reproduced. This should cast serious doubt on the utility of these
findings and encourage further research on this topic before we consider the question of
economic distress and family violence to be satisfactorily answered. I advise using
hierarchical modeling techniques to study the effects of individual, household and
neighborhood factors and greater theoretical attention to the importance of gender and

social norms, occupational prestige (e. g. perhaps lower-ranking jobs have higher

49

turnover, leading to more job spells), and interaction between high strain and stakes in

conformity.

50

 

 

2.0 00.0- m00-
02m N_0m 000m
000- 0m0- mm0
mncm owNm mwNm
:0- :0 :0
mhcm 0wNm MNNm
:0 E0 :0
mN-cm 0wNm MbNm
Nm0- 30 0m0
whom owNm thm
00.0 00. N00
mhom owNm MNNm
00 V00. m00-
amom ENm omNm
00. no.0- 0_.0-
whom owNm MNNm
0.0- 00.0- 00.
mncm owNm thm
0.0 0.0 _0.0-
00.0- 00.0 .00
00- N00- 00.0
mO< 152m 02m
m5<2 m 2

$0 00. _
mbom wwom
v0.0 00
vvmm vcmm
0N0- _m.0-
34mm 3mm
:0 :0
30mm vow-m
C .0- o __ .0-
Eumm 00mm
00.0 .00
31mm 00mm
N00- 000-
N _ mm Nmmm
m _ .0- 0 _, .0-
3mm 00mm
>00- 00.
#4me 3mm
0_ .0- N _ .0-
0_ .0- N _ .0-
.00- _0.0-
ZME-m ZMFm
.ZE .ZE
mq<2 .EmE

03m
00.—

00 _ m
0.0-

02m
50.

03m
90

mo _ m
00.0

09 m
N00

00 _ m
M00.

02 m
N00

00;.
N_.0-

N00-

_00.

mm<mm $.me Orb/xx M: meZD .UDQm

m.Z<_2

chm
00. _

Nwom
Z _0-

chm
0 _ .0

chm
00

N03
N00-

bvcm
0.0

chm
«0.0

chm
>00

>00

.00

chm
00. _

chm
5N0-

chm
00.0

Nwom
N00-

$0.0m
0N0-

chm
_ _.0-

chm
00.0-

0.0.

00.0

E2007:

chm
00.—

chm
N00

Nwom
_00-

Sucr-
V00

N09”
_ _.0

chm
_0.0

000

0.0-

chm
00.—

Nxom
_0.0-

540m
_00

chm
00.0

Nwom
_00

.00-

_0.0-

Nwom
00. _

500m
000

N03.
00

chm
N00

50-

_00-

.UDQm Nm><>> mmHIv—U/‘Jm
mammz mag $52 mace—Z mézmm .D<m_D

Doom
00. _

53m
VNO

540m
wm0

00.0

0.0

0200

chm
00. _

Nwom
M00.

_00

_00-

chm
00. _

3.0

5.0

.toU
c8300
Z

.toU
50800
Z

.toU
.5030
Z

.toU
ace-80m
Z

.toU
.5030;
Z

.0000
c8300
Z

.toU
ace-$00
Z

.toU
.5030
Z

.toO
50.000
Z

.toU
500000
.toU
E00300
.toU
.5030;

.Em .ZE .Emm

mm<Im m .232.

N m><3 mHmm—Q #

N3 Orr/xx
mamm—ZEEOUZZ

f me75

ZmED-ZEU $52

.UDQm mill)—

.UDQm mix—2mm

N m><3
Q<m~D 0200

.09:

MU<Am
MUZm—AOSV
N m><>>
mUZmAO—xw
_ m><3

 

 

Nay—BE ZO—h<qm~—~—OU u< N—szmm<

51

 

 

mhcm
00. _

RNm
: 0-

obNm
0m0-

0wNm
00.—

0w _ m
0N0

001m
00.

am;

thm
00. _

ELM
_00

mim-

30mm
00.—

MVNm

vow-m

Z

.toU
280000
Z

.toO
50.800
Z

.toU
.5000
Z

.toO
com-.000
Z

mO< m3<2

.15sz .2mm

Jam—2m m5<2

.ZMC-m .ZE m3<2

 

 

52

 

$005 00.3. N 2 000.50 0002. N0

 

 

 

00.0 No.0- 03 N_, .N- 3.?“ seem Essa-.0 2m:

. . - . . - . 3.0050 0001—. N 9 .0005 082. N0
S o No 0 mm a S N 43 m $25 3235 2300

. V . . peas .0205. S 9 .525 oz. 8
N0 _ S m S N Nee N :0ch 02 2a:

. . . . . 059$ Hmonwfl-m. Og OH .593 OZ. Ov
N N am a. 8 S 8 o N8 N 505 a: sees-0
:0 ad m 0 N03 23% £2. 032
a; :3 m 0 Sam 28... 8.. 22E
Z: 00 006 00%. 03:0. 20:00:05 E E00220 # 00 00203 502500 0wSEU
~N.O m0.0 OOA O0.0 #900 ooze—OCI N o>m>>
ONO VOO OOA O0.0 ~0v£ 00:20; m 0.33
one 5.0 8.0 8.0 EN sees a as; 500 ”2.52
va MOO OOA O0.0 mhfim x003 9 020: 50m ”Edema
$0 mNd 00; 00.0 31m 0508 08 0:5: 0:03 058 000603 0.02
wmd £0 00.0 00.0 Sim 005.80 00.. 0.50: 0.83 088 SEED ”0.0800
son .00 :82 a: 52 z

 

 

833...; 050590:— 05 050503 0.. 800.030 203.58: "m 02:200.?

53

REFERENCES

Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency.
Criminology, 30(1), 47-87.

Anderson, K. L. (1997). Gender, status, and domestic violence: An integration of feminist
and family violence approaches. Journal of Marriage and the Family, 59, 655-

669.

Armour, S. (2008, May 16). Foreclosures take an emotional toll on many homeowners
(Money, Economy). Retrieved March 12, 2009, from USA Today:
<http://www.usatoday.com/money/economy/housing/2008-05-14-mortgage-
foreclosures-mental-health_N.htm>.

Armour, S. (2009, 3 Feb). More families move in together; Pooling resources can help
but can also cause friction. USA Today, sec. Money, p. 7A.

At risk of foreclosure? Help may be on the way. (2008, 30/ 10). Retrieved 12/03/2009,
from WowOWow.com: http://www.wowowow.com/post/risk—foreclosure-help-
may-be-way-132290.

Atkinson, M. P., Greenstein, T. N ., & Lang, M. M. (2005). For women, breadwinning
can be dangerous: Gendered resource theory and wide abuse. Journal of Marriage
and the Family, 67(5), 1137.

Bachman, R. (2000). A comparison of annual incidence rates and contextual
characteristics of intimate-partner violence against women from the National
Crime Victimization Survey (NCJ VS) and the National Violence Against Women
Survey (NVAWS). Violence Against Women, 6(8), 839-67.

Bachman, R., & Meloy, M. (2008). The epidemiology of violence against the elderly:
Implications for primary and secondary prevention. Journal of Contemporary
Criminal Justice, 24, 186.

Bad economy makes love go wrong - Domestic violence cases sharply increase. (2009,
25/03). Retrieved 12/03/2009, from E-Releases:
http://www.ereleases.com/pr/bad-economy-love-wrong-domestic-violence-cases-
sharply-increase-15978.

Baird, G. (2007, 21/ 12). Homeowners in foreclosure need free legal help, Ohio 's top
judge says. Retrieved 12/03/2009, from The Plain Dealer:

54

http://blog.cleveland.com/pdworld/2007/12/homeowners_in_foreclosure_need.ht
ml.

Banks, J. (2008, May 22). Foreclosure crisis causing domestic violence (World).
Retrieved March 12, 2009, from NowPublic:
<http://www.nowpublic.com/world/foreclosure-crisis-causing-domestic-
violence>.

Barnett, O. W. (2000). Why battered women do not leave, Part 1: External inhibiting
factors within society. Trauma, Violence & Abuse, 1(4), 343-372.

Barnett, O. W., & LaViolette, A. D. (1993). It could happen to anyone: Why battered
women stay.. Newbury Park, CA: Sage.

Berger, L. M. (2005). Income, family characteristics, and physical violence toward
children. Child Abuse & Neglect, 29(1), 107-133.

Berk, R.A., Berk, S.F., Loseke, D.R. & Rauma, D. (1983) Mutual combat and other
family violence myths. In D. Finkelhor, R.J. Gelles, G.T. Hotaling, & M.A. Straus
(Eds.), The dark side of families: Current family violence research (pp. 197-212).
Newbury Park, CA: Sage.

Bowker, L. H. (1984). Coping with wife abuse: Personal and social networks. In A. R.
Roberts (Ed.), Battered Women and Their Families. New York, NY: Springer.

Bowlus, A. J ., & Seitz, S. (2006). Domestic violence, employment, and divorce.
International Economic Review, 47(4), 1 113-1 149.

Brisson, D., & User, C. L. (2007). The effects of informal neighborhood bonding social
capital and neighborhood context on homeownership for families living in
poverty. Journal of Urban Affairs, 29(1), 65-75.

Catalano, R., Dooley, D., Novaco, R., Wilson, 6., & Hough, R. (1993). Using ECA
survey data to examine the effect of job layoffs on violent behavior. Hospital &
Community Psychiatry, 44(9), 874-878.

Catalano, R., Novaco, R. W., & McConnell, W. (2002). Layoffs and violence revisited.
Aggressive Behavior, 28(3), 233-247.

Cohen, L. E., & Felson, M. (1979). Social change and crime rate trends: A routine
activity approach. American Sociological Review, 44(4), 588-608.

55

Danielson, R. (2008, Dec 7). Hard times pack Tampa Bay domestic violence shelters. St.
Petersberg Times. Retrieved 12/03/2009, from
th/lwww.tampabaycom/news/publicsafetL/crime/article928959.ece.

Dobash, R.P., Dobash, R.E., Wilson, M., & Daly, M. (1992) The myth of sexual
symmetry in marital violence. Social Problems, 39, 71-91.

Dollard, J ., Dobb, L., Miller, N. E., Mowrer, O. H., & Sears, R. (1939). Frustration and
Aggression. New Haven: Yale University Press.

Down economy means crisis calls go up. (2008, 26/09). Retrieved March 12, 2009, from
MyFox, Tampa Bay.:
http://www.myfoxtampabay.com/dpp/news/Down__economy_means_crisis__calls_
go_up.

Elder Jr., G. H., & Liker, J. K. (1982, Sep). Hard times in women's lives: Historical
influences across forty years. The American Journal of Sociology, 88(2), 241—69.

Fagan, J. A., & Browne, A. (1994). Violence against spouses and intimates: Physical
aggression between women and men in intimate relationships. In A. J. Reiss Jr. &
J. A. Roth (Eds.), Social Influence. Vol. 3: Understanding and Preventing
Violence (pp. 115-292) [Panel on theUnderstanding and Control of Violent
Behavior, Committee on Law and Justice, Commission on Behavioral and Social
Science and Education, National Research Council]. Washington, DC: National
Academy Press.

Fagan, J. (1993). Social structure and spouse assault. In Brian Forst (Ed.), The
Socioeconomics of Crime and Justice (pp. 209-237). Armonk, NY: ME. Sharpe.

Farmer, A., & Tiefenthaler, J. (1997). An economic analysis of domestic violence.
Review of Social Economy, 55(3), 337-358.

Farmer, A., & Tiefenthaler, J. (2003a, Apr). The employment effects of domestic
violence. Journal of Labor Research.

Farmer, A., & Tiefenthaler, J. (2003b, Apr). Explaining the recent decline in domestic
violence. Contemporary Economic Policy, 21(2), 158-72.

Ferree, M. (1990). Beyond separate spheres: Feminism and family research. Journal of
Marriage and the Family, 52, 866-884.

Finkelhor, D., Ormrod, R., Turner, H., & Hamby, S. (2005, Feb). The victimization of
children and youth: A comprehensive, national survey. Child Maltreatment,
10(1), 5-25.
56

Fischer, P., Greitemeyer, T., & Frey, D. (2008). Unemployment and aggressions: The
moderating role of self-awareness on the effect of unemployment on aggression.
Aggressive Behavior, 34(1), 34-45.

Flood, Michael. (1999). Claims about husband battering. DVIRC Newsletter, Summer.
Melbourne: Domestic Violence and Incest Resource Centre, pp. 3-8.

Foreclosure help needed. (2009, 23/02). Retrieved 12/03/2009, from Baltimore Sun:
http://www.baltimoresun.com/news/opinion/editorial/bal-
ed.foreclose22feb23,0,4555649.story.

Fox, G. L., Benson, M. L., DeMaris, A. A., & Van Wyk, J. (2002). Economic distress
and intimate violence: Testing family stress and resource theories. Journal of
Marriage and the Family, 64(3), 793-807.

Gavin, R. (2008, 5/12). Losing jobs in unequal numbers (Jobs). Retrieved March 12,
2009, from The Boston Globe:
http://www.boston.com/jobs/news/articles/2008/12/05/losingJobs_in_unequal_n
umbers.

Gelles, R. J. (1993). Through a sociological lenses: Social structure and family violence.
In R. Gelles & D. Loseke (Eds.), Current Controversies on Family Violence (pp.
31-46). Newbury Park, CA: Sage Publications.

Gelles, R.J. & Straus, M.A. (1988). Intimate violence. New York: Simon and Schuster.

International Labour Office. (2008, Mar). Global Employment Trends for Women.
Retrieved 12/03/2009, from http://www.ilo.org/wcmspS/groups/public/—--
dgreports/---dcomm/documents/publication/wcms_091225.pdf.

Goode, W. J. (1971, Nov). Force and violence in the family. Journal of Marriage and the
Family, 33(4), 624-36.

Goodman, P. S., & Healy, J. (2009, 6/03). Job losses hint at vast remaking of US.
economy (Economy). Retrieved 12/03/2009, from New York Times:
http://www.nytimes.com/2009/03/O7/business/economy/07jobs.html?_r=2&ref=b
usiness.

Gorman-Smith, D., Tolan, P., Loeber, R., & Henry, D. B. (1998). Relation of family
problems to patterns of delinquent involvement among urban youth. Journal of
Abnormal Child Psychology, 26(5), 319-333.

Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime. Palo Alto, CA:
Stanford University Press.
57

Greenberg, J ., McKibben, M., & Raymond, J. (1990). Dependent adult children and elder
abuse. Journal of Elder Abuse & Neglect, 24(1/2), 73.

Hahn, M. (2009, 11/03). Foreclosures hit home. Retrieved 12/03/2009, from The
Women's International Perspective (The WIP):

http://www.thewip.net/contributors/2009/02/foreclosures_hit_home_a_microc.ht
ml.

Hall, P. (1989). Elder maltreatment items, subgroups and types: Policy and practice
implications. International Journal of Aging and Human Development, 28(3),
191-205.

Hirschi, T. (1969). Causes of delinquency. Berkeley: University of California Press.

Homung, C., McCullough, C., & Sugimoto, T. (1981). Status relationships in marriage:
Risk factors in spouse abuse. Journal of Marriage and the Family, 43(3), 675-
692.

Hotaling, G. T., & Sugarman, D. B. (1986). An analysis of risk markers in husband to
wife violence: The current state of knowledge. Violence and Victims, 1(2), 101-
124.

Howe, W. J. (1990, Nov). Labor market dynamics and trends in male and female
unemployment. Monthly Labor Review, 113(1 1).

Increased financial stress affects domestic violence victims. (2009, 30/01) (Hotline
News). Retrieved 12/03/2009, from National Domestic Violence Hotline:
http://www.ndvh.org/2009/0llincreased-financial-stress-affects-domestic-
violence-victimsl.

Jensen, B. (2008, 28/03). Foundations and charities struggle to deal with housing-
foreclosure crisis (News Updates). Retrieved 12/03/2009, from The Chronicle of
Philanthropy: http://phi1anthropy.com/news/updates/index.php?id=4059.

Kalmuss, D., S., & Straus, M., A. (1982). Wife's marital dependency and wife abuse.
Journal of Marriage and the Family, 44(2), 277-286. Minneapolis, MN.

Kantor, G. K., & Straus, M. A. (1987). The "drunken bum" theory of wife beating. Social
Problems, 34, 213-321.

Karthik Rajaram kills family and self in Porter Ranch murder-suicide. (2008, Oct 6)
(Across the Nation). Retrieved March 12, 2009, from Bitten & Bound:
http://www.bittenandbound.com/2008/ 10/06/karthik-raj aram-kills-family-and-

self-in-porter-ranch-murder-suicide/.
58

Kaukinen, C. (2004). Status compatibility, physical violence, and emotional abuse in
intimate relationships. Journal of Marriage and the Family, 66(2), 452-471.

Kirchhoff, Sue. (2009, Mar 6). Jobless rate jumps to 8.1% as 65 1,000 jobs slashed. USA
Today, Online. Retrieved November 10, 2009:
http:l/www.usatoday.com/money/economy/2009-O3-06-jobs_N.htm

Labor force participation rates, 1975-2008. (2008). Retrieved March 12, 2009, from US.
Bureau of Labor Statistics: http://www.bls.gov/opub/working/page3b.htm.

Lambert, L. C., & Firestone, J. M. (2000). Economic context and multiple abuse
techniques. Violence Against Women, 65 ( 1), 49-67.

Laub, J. H., & Sampson, R. J. (1993). Turning points in the life course: Why change
matters to the study of crime. Criminology, 31(3), 301-325.

Lindley, D.V., and Scott, W.F. (1984). New Cambridge Statistical Tables. Cambridge,
UK: Cambridge University Press.

Little, R]. & Schenker, N. (1994) Chapter 2: Missing Data. In Arminger, G., Clogg, C. &
Sobel, M. (Eds.), Handbook of statistical modeling for the social and behavioral
sciences, pp.39-69. Plenum Press: New York, NY.

Littwin, S. (1995). A call for help! The untold story of elder abuse today. New Choices of
Retirement Living, 35, 3.

Louwers, B. C. (2009, 8 May). Distraught man attempts suicide along I-696. C & G
Newspapers (Warren, MI), Online edition.

Macmillan, R., & Gartner, R. (1999). When she brings home the bacon: Labor-force
participation and the risk of spousal violence against women. Journal of Marriage
and the Family, 61, 947-958.

Magdol, L., Moffitt, T., Caspi, A., Newman, D., Fagan, J ., & Silva, P. (1997). Gender
differences in partner violence in a birth cohort of 21-year-olds: Bridging the gap
between clinical and epidemiological approaches. Journal of Consulting and
Clinical Psychology, 65(1), 68-78.

Mistry, R. S., Low, E. D., Benner, A. D., & Chien, N. (2008). Expanding the Family
Economic Stress Model: Insights from a mixed-methods approach. Journal of
Marriage and the Family, 70(1), 196-209.

59

Mulligan, C. B. (2009, Jan 14). A milestone for working women? (Economix). Retrieved
March 12, 2009, from New York Times: ,
htLpzl/economix.blogs.nytimes.com/2009/01/l4/a-milestone-for-women-workersl.

National Conference of State Legislatures. (2008, Jul). States that permit victims of
domestic violence to receive unemployment insurance. Retrieved March 12, 2009,
from http://www.ncsl.org/programs/employ/ui-domviol.htm.

National home-value drop accelerated in fourth quarter. (2009, Feb 27). Retrieved from
Freddie Mac:
httpz/lwww.freddiemac.com/news/archives/rates/2009/4thi08.html.

Nordheimer, J. (1990, 20/12). A paycheck gone, a family in crisis. The New York Times,
Home & Garden. Retrieved 22/03/2009, from
http://www.nytimes.com/ l 990/ 12/20/garden/a-paycheck-gone-a-family-in-
crisis.html.

O'Donnell, C. J., Smith, A., & Madison, J. R. (2002). Using demographic risk factors to
explain variations in the incidence of violence against women. Journal of I
Interpersonal Violence, 17(12), 1239-1262.

Oh, J .-H. (2004). Race/ethnicity, homeownership and neighborhood attachment. Race
and Society, 7, 63-77.

Payne, B. K. (2000). Crime and elder abuse: An integrated perspective. Springfield, IL:
Charles C. Thomas.

Peto, R., Pike, M. C., Armitage, P., Breslow, N.E., Cox, D.R., Howard, S.V., Mantel, N.,
McPherson, K., Peto, J ., & Smith, PG. (1976). “Design and analysis of

randomized clinical trials requiring prolonged observation of each patient, I:
Introduction and design”. British Journal of Cancer, 34, 585-612.

Pillemer, K. A., & Finkelhor, D. (1989, Apr). Causes of elder abuse: Caregiver stress
versus problem relatives. American Journal of Orthopsychiatry, 59(2), 179-87.

Rampell, C. (2009, 05/02). As layoffs surge, women may pass men in job force
(Business). Retrieved March 12, 2009, from New York Times:
httpz/lwww.nytimes.com/2009/O2/O6/business/06women.html?pagewanted=1 &fta

2y.

Raphael, J. (2001). Getting the message out: Using the media to change social norms on
abuse. In C. M. Renzetti, J. Edleson & R. Bergen (Eds.), Sourcebook on Violence
Against Women. Sage Publications.

60

Riger, S., & Staggs, S. L. (2004). Welfare Reform, Domestic Violence, and Employment:
What Do We Know and What Do We Need to Know? Violence Against Women,
10, 961.

Robb, J. (2008, Jun 10). Foreclosures and violent crime. In Global Guerillas Journal.
Retrieved March 12, 2009, from
<http://globalguerrillas.typepad.com/globalguerrillas/2008/06/joumal-
foreclo.html?cid=1 1845 l 896>.

Rogers, S. (2004). Dollars, dependency and divorce: Four perspectives on the role of
wives' income. Journal of Marriage and the Family, 66(1), 59-74.

Royall, RM. (1986). The effect of sample size on the meaning of significance tests. The
American Statistician, 40(4), 313—3 15.

Satyanarayana, M., & Dinh, E. C. (2009, 8 May). I-696 overpass drama resonates with
people seeking, giving help. Detroit Free Press (Detroit, MI), Online ed.

Scheerer, M. (2009, 16/03). NY domestic violence rising with joblessness, home
foreclosures. Retrieved 12/03/2009, from Public News Service:
httpzl/www.publicnewsservice.org/index .php?/content/article/799 l- 1.

Scott, E., London, A., & Myers, N. (2002). Dangerous dependencies: The intersection of
welfare reform and domestic violence. Gender & Society, 16, 878.

Sellers, C. S. (2006, Mar). Self-control and intimate violence: An examination of the
scope and specificaiton of the general theory of crime. Criminology, 37(2), 375-
404.

Sherman, L. W. (1992). The influence of criminology on criminal law: Evaluating arrests
for misdemeanor domestic violence. Journal of Criminal Law & Criminology,

83(1), 1-45.

Smith, P. H., Thornton, G. E., DeVellis, R., Earp, J ., & Coker, A. L. (2002). A
population-based study of the prevalence and distinctiveness of battering, physical
assault, and sexual assault in intimate relationships. Violence Against Women,
8(10), 1208-1232.

Stark, E., & Flitcraft, A. H. (1996). Women at risk: Domestic violence and women 's
health.. Thousand Oaks, CA: Sage.

Steinmetz, S.K. (1977-78). The battered husband syndrome. Victimology: An
International Journal, 2, 499-509.

61

Stop foreclosure in Virginia Beach. (2009). Retrieved 12/03/2009, from 877-You-Keep:
http://www.877youkeep.com/stop-foreclosure-in-Virginia-Beach.

Straus, M. A., Gelles, R., & Steinmetz, S. (1980). Behind closed doors: Violence in the
American family. New York: Doubleday/Anchor Books.

Sullivan, L. (2008, Nov 20). Experts: Bad economies don 't cause crime waves
(Economy). Retrieved March 12, 2009, from NPR:
<http://www.npr.org/temp]ates/story/storyphp?storyId=97234406>.

Sutherland, E. H. (1947). Principles of Criminology (4th ed.). Chicago: J. B. Lippincott
Co.

Tauchen, H., & Witte, A. D. (2001). The dynamics of domestic violence. Efi‘ect of
Institutions on Economic Behavior, 85, 414-418.

Tuthill, L. (2008). Breaking new windows: Examining the subprime mortgage crisis
using the broken windows theory. Geography & Public Safety, 1(3), 9-10.

United States Unemployment Rate. (2009, Mar 6). Retrieved 12/03/2009, from
TradingEconomics.com:
http://www.tradingeconomics.com/Economics/Unemployment-
rate.aspx?Symbol=USD.

National Network to End Domestic Violence (NNEDV). (2009). VA WA Transitional
Housing Assistance Grants for Domestic Violence, Dating Violence, Sexual
Assault and Stalking Victims. Retrieved 12/03/2009, from
http://nnedv.org/docs/Policy/thousing_in_arra.doc.

Vinokur, A., Price, R., & Caplan, R. (1996). Hard times and hurtful partners: How
financial strain affects depression and relationship satisfaction of unemployed
persons and their spouses. Journal of Personality and Social Psychology, 71, 166-

179.

Warren, E. (2006, 07/2006). Rewritting the rules: Families, money and risk. In (Calhoun,
C., Ed.) The Privatization of Risk. Retrieved 22/03/2009, from Social Science
Research Council: http://privatizationofrisk.ssrc.org/Warren/.

Williams, J. C. (1998). Domestic violence and poverty: The narratives of homeless

women. Frontiers,
http://findarti cles. c0m/p/articles/mi_qa3 68 7/is_1 99801 /a i_n8 772075.

Wilson, R. E., & Paulsen, D. J. (2008, Oct). Foreclosures and crime: A geographical
perspective. Geography & Public Safety, I (3), 1-2.
62

World Markets (2009). New York Times. Retrieved March 12, 2009, from
http://markets.on.nytimes.com/research/markets/worldmarkets/worldmarkets.asp.

63