Domestic Violence, Poverty, and Social
Services: Does Location Matter?n
Andrea Hetling, Rutgers, The State University of New Jersey
Haiyan Zhang, Rutgers, The State University of New Jersey
Objective. This study investigates whether or not domestic violence agencies are
located in areas of need. Recent research indicates that community economic disadvantage
is a risk factor for intimate partner violence, but related questions regarding
the geographic location of social service agencies have not been
investigated. Methods. Using Connecticut as a case study, we analyze the relationship
of agency location and police-reported domestic violence incidents and assaults
using OLS regression and correcting for spatial autocorrelation. Results. The presence
of an agency within a town has no relationship with the rates of domestic
violence. However, regional patterns are evident. Conclusion. Findings indicate
that programs are not geographically mismatched with need, but neither are programs
located in towns with higher rates of incidents or assaults. Future research
and planning efforts should consider the geographic location of agencies.
Numerous studies have documented the overrepresentation of domestic
violence in poor households relative to middle- and upper-income households
and the associated risk of poverty (Benson and Fox, 2004; Carlson
et al., 2003; Hotaling and Sugarman, 1990; Tjaden and Thoennes, 2000).
According to the U.S. Department of Justice, the average annual female
intimate partner victimization rate per 1,000 persons between 2001 and
2005 was 12.7 for women residing in households earning less the $7,500
annually compared to 2.0 for women with annual household incomes over
$50,000 (Catalano, 2007). The relationship between poverty and domestic
violence is complex. In some situations, poverty may exacerbate the likelihood
of experiencing domestic violence. With fewer options for economic
self-sufficiency and social support systems with little ability to offer financial
help, poor women may feel more trapped in unhealthy relationships
n
Direct correspondence to Andrea Hetling, Ph.D., Assistant Professor, Bloustein School of
Planning and Public Policy, Rutgers, The State University of New Jersey, 33 Livingston Ave.,
New Brunswick, NJ 08901 hahetling@rutgers.edui. The corresponding author is willing to
share all data and coding information with anyone wishing to replicate the study. This
research was partially funded by the University of Connecticut Faculty Large Grant Program.
The authors thank Scott Allard and Judy Postmus for their very helpful comments and Shane
Van Housen and Yan Zhao for research assistance. Gary Lopez of the Connecticut Department
of Public Safety and Linda Blozie, Lisa Holden, and Gerrie Wilde of the Connecticut
Coalition Against Domestic Violence facilitated access to the data and patiently answered
questions. All omissions and errors are the authors’ own.
SOCIAL SCIENCE QUARTERLY, Volume 91, Number 5, December 2010
r 2010 by the Southwestern Social Science Association
(Tolman and Raphael, 2000). In other situations, domestic violence may
lead to poverty for women in previously financially sound situations as
women who escape violent relationships are often left with no financial
resources (Davis, 1999). Moreover, victims of domestic violence often have
related difficulties such as limited or inconsistent work experience, poor
education, physical and mental health problems, and substance abuse issues
(Lloyd and Taluc, 1999; Logan et al., 2007; Tolman and Raphael, 2000;
Williams and Mickelson, 2004). Such problems pose serious barriers in
escaping violence and achieving self-sufficiency and present a challenging
situation for social service programs (Renzetti, 2009).
Survivors of domestic violence report that successful transitions to violence-free
and independent lives are often made with the assistance of social
services such as shelters, counseling, and public cash assistance (Lyon, Lane,
and Menard, 2008; Postmus et al., 2009). Such services are particularly
important for low-income women who likely encounter barriers when attempting
to leave an abusive relationship and often turn to public assistance
and services for help (Lyon, 2000). Services are only useful, however, if they
are accessible to and utilized by those in need. In designing and delivering
services to domestic violence victims, policymakers and program administrators
are faced with a dilemma. On one hand, research on geography and
social services aimed at the poor indicates that spatial proximity is a strong
predictor of service uptake (Allard, Tolman, and Rosen, 2003). On the
other hand, research on domestic violence indicates that contact with agencies
outside a woman’s neighborhood is a critical link to accessing services
(Warrington, 2001).
The current study investigates the relationship between poverty and domestic
violence on a town level and explores where domestic violence programs
are located as related to domestic violence assaults and incidents. The
central research question asks whether or not the location of services at the
town level is related to rates of domestic violence. In other words, if social
services aimed at domestic violence victims are to be accessible to victims,
one should see a geographic pattern of domestic violence related to the
location of shelters and agencies after controlling for other environmental
risk factors such as poverty. We hypothesize that because domestic violence
agencies developed out of grassroots and community activism, the geographic
location of agencies is not related to rates of domestic violence or
other community risk factors.
Place and Social Policy
Academic research and policy reports on the causes of poverty and efficacy
of anti-poverty programs have adopted a strong focus on the importance of
place and community (e.g., Blank, 2004). Policy scholars have shown that
the proximity of services to the target population is an important aspect of
Domestic Violence, Poverty, and Social Services 1145
service uptake because the location of social services greatly affects the ability
of those in need to access and utilize services and programs (Allard, 2004;
Allard, Tolman, and Rosen, 2003). In other words, place matters and the
closer services are placed to individuals in need, the more likely these individuals
will utilize these services. This finding makes intuitive sense, as
services located close to one’s home would be more convenient. However,
recent research demonstrates a mismatch between where providers are located
and the communities where individuals in need reside (Allard, 2009).
Community Risk Factors and Domestic Violence
Building on findings related to individual poverty and domestic abuse,
researchers in geography, criminology, and public health have begun examining
the relationship between community context and domestic violence.
Qualitative studies have utilized concept mapping to frame how
abused women perceive the relationship between community factors and
interpersonal violence (O’Campo et al., 2005; Burke, O’Campo, and Peak,
2006). A growing number of quantitative studies empirically document
these relationships. Miles-Doan (1998), in a case study of Duval County,
Florida, found that neighborhoods with larger proportions of unemployed
males, female-headed households, and residents in poverty also had higher
rates of domestic violence than other neighborhoods. Studies using the
National Survey of Families and Households support the hypotheses that
economic disadvantage on the community level as well as the household
level affects the likelihood of experiencing domestic violence (Benson et al.,
2003; Fox and Benson, 2006). Grana (2001) explores the relationship of
state-level characteristics and domestic femicide and concludes that the association
between state size and homicide rates may be related to other
variables such as poverty and public services. International case studies in
Australia (DiBartolo, 2001) and Bangladesh (Koenig et al., 2003) also conclude
that domestic violence is related to community contextual variables.
Some of these empirical studies, along with related research investigating
teenage sexual activity, crime involvement, and health outcomes (e.g.,
Browning, Leventhal, and Brooks-Gunn, 2005; Dembo et al., 2009), have
found evidence of a concentrated disadvantage effect on individual-level
outcomes. Wilson (1987) introduced the theory that communities with a
very high concentration of poverty or disadvantage have different effects on
residents than does poverty alone. The seminal work of Sampson, Raudenbush,
and Earls (1997) operationalized concentrated disadvantage as a
combination of the proportion of poverty, public assistance receipt, femaleheaded
households, unemployment, children under 18, and African Americans;
their work demonstrated the effect of residing in a very disadvantaged
community on stranger violence. Other domestic violence studies have investigated
the possible similarities and differences between stranger violence
1146 Social Science Quarterly
and domestic violence. Although community socioeconomic characteristics
are less powerful predictors of domestic violence compared to stranger violence
(Frye et al., 2008), increased community violence is associated with
higher rates of domestic violence (Raghavan et al., 2006), further bolstering
the case for a consideration of community characteristics and concentrated
disadvantage as risk factors for individual domestic violence.
Domestic Violence Services and Program Goals
Social services designed to assist victims of domestic violence provide a
wide range of services at the community level. The first emergency shelters
were founded in the 1970s in response to a growing grassroots movement
focused on the rights and needs of battered women and their children,
primarily their immediate safety (Schechter, 1982). Over the past 40 years,
the movement has grown and evolved to a professionalized, institutionalized
system supported in part by public funding (Reinelt, 1995). The goal of the
movement has also evolved, from a focus on coaching women to leave
relationships to a goal of empowerment (Arizona Coalition Against Domestic
Violence, 2000). Empowerment service models aim to provide the
necessary support, encouragement, and resources to enable a woman to
make her own choices and decisions about herself and her life. Thus, many
agencies have vastly expanded their services to include longer-term housing
along with nonresidential services as well as support groups aimed at different
subpopulations and financial literacy programs (Sullivan and Gillum,
2001).
A growing body of literature examines the efficacy of this system. Studies
provide strong support that domestic violence services, including shelter,
advocacy, legal services, and counseling, have positive effects on the ability of
women to escape and remain free from abusive relationships as well as
enhance their well-being (Sullivan and Bybee, 1999; Tan et al., 1995; Tutty,
Weaver, and Rothery, 1999; Weisz, Tolman, and Bennett, 1998). However,
poor women’s ability to access services is often compromised by the difficulties
they face in escaping an abusive relationship (Davis, 1999). In a
national study of 20 domestic violence programs, Zweig, Schlichter, and
Burt (2002) found that agencies consistently cited poverty as a barrier to
obtaining and receiving services. It seems reasonable to conclude that nonresidential
services and thus their location are especially important for this
at-risk population.
Location and Domestic Violence Services
Research on the structure of domestic violence services has focused on
integrated community responses and the importance of cooperation and
Domestic Violence, Poverty, and Social Services 1147
communication among service providers (Clark et al., 1996). However, little
research has been completed on the question of place and location. Mainstream
geographic literature focuses on stranger violence, particularly fear of
crime outside the home and in public spaces (for a review, see Pain, 2000).
Feminist geographic studies have expanded this view and have much to
contribute to the question of space and safety for domestic violence victims
(Pain, 2001; Whitzman, 2007). Warrington (2001) provides a framework
for how women conceptualize restricted spaces and theorized that shelters
exist in a space distant and isolated from the home regardless of measured
distance.
Thus, unlike social service provision literature, the findings from this
literature seem to indicate that distance or geographic accessibility or proximity
may not be as important as availability or the level of services. Perhaps
findings based on geographic location and service uptake of other types of
social services are not generalizable to domestic-violence-related services.
However, intuitively, some measure of geographic distance may be too far,
especially for poor women. If certain communities are at an increased risk of
experiencing a higher rate of domestic violence, it seems to make sense that
more services should be made available to them, regardless of physical location.
In a study of a rural community in New Zealand, Panelli, Little, and
Kraack (2004) caution that domestic violence service providers should consider
the cultural and spatial construction of violence and safety in designing
appropriate services.
If community and individual poverty both increase one’s risk of domestic
violence and decrease one’s ability to obtain services, then perhaps the current
construction that Panelli, Little, and Kraack (2004) are emphasizing is
such that an investigation of agency location is merited. The limited amount
of empirical research in this area is problematic in a pragmatic sense as
domestic violence agencies, in particular nonresidential services, are structured
to deliver services to geographic areas. We hypothesize that because the
history of agency formation is based on grassroots and community awareness
and involvement, agency location is likely not related to town-level need.
Study State and Policy Environment: Connecticut
We use Connecticut as a case study to examine our hypotheses and focus
on services and need at the town level. Connecticut is a small state encompassing
4,844 square miles and was home to 3,405,565 residents in 2000.
The state contains rural in additional to urban areas, but does not have the
difficulty of remote or isolated regions due to its relatively small size. Although
it is possible to drive across the state, from the southwest corner to
the northeast corner, in less than two-and-a-half hours, a commute of this
length is not conducive to service uptake. Connecticut’s relative wealth and
high cost of living also present formidable challenges to low-income women
1148 Social Science Quarterly
struggling to achieve self-sufficiency and find safe and affordable housing,
making social services critical.
Connecticut is a useful and manageable case study site for a number of
reasons. Connecticut’s Department of Public Safety has been collecting
family violence data since 1986 and maintains an extensive electronic database
of this information, making data easily accessible for research purposes.
The community of domestic violence advocates also has a strong
history in Connecticut, with the first emergency shelter founded in 1975,
following closely the history of the movement on the national level. Like
many other states, Connecticut has a state coalition, the Connecticut Coalition
Against Domestic Violence, which serves as an umbrella organization.
The state currently has 18 domestic violence agencies, that are active
members of the Coalition and all receive at least part of their funding from
the state. The Department of Social Services, the Judicial Branch via the
Office of Victim Services, and the Office of Policy and Management provide
support focused on different aspects of service provision. All these sources
are partially comprised, to differing degrees, of federal pass-through dollars.
These 18 funded agencies provide a range of support services to victims of
domestic violence and their children. Shelter, legal services, counseling, and
public education are the most common services. All agencies provide some
type of emergency shelter to victims. Sixteen of the 18 agencies run shelters
with between 12 and 20 beds. The remaining two agencies arrange for host
homes.
As seen in Figure 1, the domestic violence agencies are not disbursed
evenly throughout the state. Most agencies are clustered in the western half
of the state, with five of the 18 agencies located in the southwest corner.
Only four agencies are located east of the Connecticut River, which divides
the state in half. Many of the agencies are located near the capital of Hartford
in the center of the state or in the southwestern corner in the suburbs of
New York City. Measuring from town center to town center, the average
distance from a town without a domestic violence agency to the closest town
with an agency is 8.6 miles, with the furthest distance at 19.5 miles.
Methods
Sample and Dependent Variable Definitions
The sample for the research project consists of the 169 Connecticut towns
and cities. Two measures of domestic violence arrests in the year 2000—
domestic violence incidents and domestic violence assaults per 1,000 residents—serve
as the dependent variables for the analyses. Data were gathered
by the Connecticut State Police, Department of Public Safety. According to
the Department, family violence is defined as physical abuse or a violent
threat that causes fear of imminent danger. Verbal and emotional abuse does
Domestic Violence, Poverty, and Social Services 1149
not constitute family violence per the definition. Data are collected by police
officers who fill out a one-page, three-ply, offense report when the incident
to which they are responding is a family violence offense. The original is sent
to the Crime Analysis Unit of the Department of Public Safety, and the data
are entered into the Family Violence Reporting Program by Family Violence
Data Entry Persons. The second copy is sent to the State’s Attorney of the
appropriate court, and the third copy is retained in the local police files. The
report includes information on date, time, and town location of the offense;
the type of offense; the type of weapon used; the extent of injuries; whether
or not drugs or alcohol were involved; and who was involved in the incident
and their relationship to the victim. There is also a box for optional remarks.
Our first dependent variable is incidents per 1,000 residents. Incidents
include all occurrences of family violence that result in at least one arrest.
These crimes include breach of peace and disorderly conduct (the formal
charges associated with threats, harassment, and endangerment) as well as
the less common charges of assault, homicide, and kidnapping. The second
FIGURE 1
Domestic Violence Agencies in Connecticut
NOTE: Map shows the town location of the 18 domestic violence agencies in
Connecticut.
SOURCE: Connecticut Coalition Against Domestic Violence.
1150 Social Science Quarterly
dependent variable, assaults, serves as a measure of more serious instances of
family violence in which physical violence such as punching or kicking
results in injuries. Homicides and kidnappings were prohibitively small to
allow for analyses.
Town Characteristics and Data Sources
Variables measuring domestic violence services and socioeconomic-related
characteristics are also from the year 2000 and come from two additional
data sources. Data on the sociodemographic and economic characteristics
come from the 2000 U.S. Decennial Census. Measures of community
characteristics include the proportion of the population below the poverty
rate, the proportion of population below 18 years old, population density,
the log of the median house value, local crime rate per 1,000 residents, and a
composite risk indicator. Due to multicollinearity among a number of other
risk-related community variables of interest and the theoretical importance
of concentrated disadvantage, we created a risk indicator to reflect the
combination of multiple poor social-economic characteristics of a certain
town or city. The town variables used in creating the composite risk indicator
were: (1) proportion of population receiving public assistance, (2)
proportion of female-headed households, (3) unemployment rate, and (4)
proportion of population below poverty level. The towns falling in the top
quarter of the distribution for every variable were categorized as towns at risk
for an increased level of domestic violence. Thus, the risk indicator is a
dummy variable with a value of 1 for towns meeting this criteria (N 5 23)
and zero for all other towns (N 5 146).1
This variable also reflects the theory
of concentration effects or concentrated disadvantage, which says that a
nonlinear relationship exists between crime and neighborhood disadvantage
and that only over a particular threshold do risk characteristics become
important (Sampson and Wilson, 1995). Lastly, we created a dummy variable,
CONNECTICUT RIVER, that equals 1 if the town is located west of the
river and 0 if the town is east of the river. This variable is used to correct for
spatial autocorrelation as discussed later and is based on the observation of
the clustering of agencies in the western portion of the state.2
1
We also calculated the risk indicator without including the poverty variable, since this
variable is already included in the model. The result was the same in terms of which towns
were coded for concentrated disadvantage.
2
We investigated the possibility that two dummy variables measuring clustering around
Hartford and New York City, the two areas with noticeably more agencies, might be a better
fit. We set the New York City variable equal to 1 if a town was within 25 miles from the
center of Greenwich and the Hartford variable equal to 1 if a town was within 10 miles from
the center of that city (in order to capture areas of similar geographic size). Neither variable
was statistically significant, but they did correct for the spatial autocorrelation—similar to the
correction achieved by the river variable.
Domestic Violence, Poverty, and Social Services 1151
Domestic Violence Agency Variables and Data Source
The Connecticut Coalition Against Domestic Violence provided data on
the number and location of domestic violence agencies in the state. In 2000,
18 domestic violence agencies were located in Connecticut. We constructed
three measures of agency location. AGENCY is a dummy variable that indicates
whether or not an agency is located in the town. AGENCY BOUNDARY is
also a dummy variable, which equals 1 when the town borders at least one
town with a domestic violence agency. The third measure, DISTANCE, is the
distance in miles to a town with an agency, measured from town centroid to
centroid. Models were run with the first two measures and separately with
the third measure. Because the results were very similar and because the
distance measure is only an approximation of the actual distances, we do not
present the models with the distance measure. Similarly, we had originally
planned to include the number of beds in the model, but upon further
investigation into the realities of service provision, we decided not to include
the variable. We discovered that the state hotline facilitates temporary
placements and moves among the agencies and that programs are also willing
to use other arrangements to meet overflow demands. Moreover, because
83 percent (15 out of 18 agencies) of the agencies have between 12 and 16
beds, there is little variation in the on-paper capacity as well.
The present analysis focuses on the location of agencies exclusively designed,
funded, and implemented for serving victims of domestic violence.
Of course, other more general social service agencies also offer programs for
domestic violence victims such as emergency funds or legal counseling. The
present analysis focuses on domestic violence agencies because they offer
more extensive and comprehensive services, are more permanent and stable,
and are more visible to the community through public education and advocacy
programs.
Analyses and Spatial Autocorrelation Tests
We examined the relationship between rates of domestic violence and
poverty using bivariate correlations and spatial patterns. We then analyzed
the influence of community indicators and domestic violence services on
domestic violence rates using multivariate and spatial statistics. Specifically,
we investigated variations of an OLS model and tested for spatial autocorrelation.
The model was run for the two dependent variables, INCIDENCE
RATE and ASSAULT RATE.
Tests for spatial autocorrelation were run to determine if the models were
biased. Town characteristics, such as rates of domestic violence arrests, are
usually spatially clustered. In extant studies, few consider the spatial autocorrelation
issue in regression analyses of domestic violence (Miles-Doan,
1998). Although the inclusion of other independent variables that have
1152 Social Science Quarterly
similar spatial patterns can explain some of the spatial error of the dependent
variable, the regression error can still be spatially biased. If additional variables
cannot explain the spatial error, geographically weighted regression is
one frequently used spatial regression technique to address this issue.
Moran’s I test is used as an index of spatial autocorrelation based on
feature location and attribute values. It is widely used to test for the presence
of spatial dependence such as clusters and dispersion (Hongfei, Calder, and
Cressie, 2007). The null hypothesis is that the data are random in their
spatial distribution. The test compares the difference in values for neighboring
features and the difference in value of all features. The values are
clustered if the average of the differences in values of adjacent features is less
than between all features. The Moran’s I index ranges from À 1 to 11, with
values close to 0 indicating the lack of spatial dependence. A Moran’s I value
indicates spatial clustering when its value is near11.0 and spatial dispersity
when it is near À 1.0. Moran’s I indices of domestic violence incidence rates
and assault rates indicate that domestic violence rates and assault rates are
significantly clustered in Connecticut. In our final model, a variable indicating
whether or not an agency was located west of Connecticut River is
added to adjust the spatial autocorrelation bias. This variable corrects the
bias and weighted regression models were deemed unnecessary.
Findings
Correlation Between Poverty and Domestic Violence
The maps in Figure 2 present the percent of the population living below
the federal poverty line and the rates of domestic violence incidents and
assaults for the 169 towns and cities of Connecticut. The geographic patterns
of each variable are noticeably similar. Higher rates of poverty and
domestic violence seem to cluster around the more urban areas of the state,
including the area surrounding Hartford in the center of the state and New
Haven in the south.
Table 1 contains correlation coefficients between domestic violence rates
and other indicators of socioeconomic distress. Similar to the literature on
neighborhood indicators, we find a high correlation between these variables.
Specifically, these values are all over 0.5, with most in the 0.7 to
0.85 range and all correlations are statistically significant at the 0.01 level.
The relationships between community risk characteristics and assault rates
are slightly stronger than the relationships between these variables and the
rate of domestic violence incidents. For example, the correlation between
poverty rates and domestic violence assaults is nearly 0.8, more than onetenth
greater than the correlation with domestic violence incidents
(0.683).
Domestic Violence, Poverty, and Social Services 1153
FIGURE 2
Connecticut Town-Level Maps, 2000; Poverty Rates by Town, 2000
1154 Social Science Quarterly
Agency Location and Community Characteristics
The overlap between domestic violence rates and measures of community
disadvantage demonstrates geographic areas of service need. Moreover, comparing
the maps in Figure 2 to the map of agencies in Figure 1, it seems at a
glance that at least some agencies are located in towns with great need. Table
2 presents a comparison of socioeconomic characteristics of towns with and
without a domestic violence agency within its borders. On average, agencies
are located in more disadvantaged towns and cities and have higher rates of
general crime, domestic violence incidence, and domestic violence assaults.
Agency Location and Domestic Violence Incidents
OLS regression was used to analyze the relationship between domestic
violence and agency location controlling for other community characteristics.
FIGURE 2–CONTINUED
NOTES: Maps are at the town level; shading indicates the level of poverty, domestic
violence incidents, and domestic violence assaults, respectively.
SOURCES: The poverty rate is from the 2000 Census, U.S. Bureau of the Census.
Measures of domestic violence come from the Connecticut Department of Public Safety.
Domestic Violence, Poverty, and Social Services 1155
Regression results examining the relationship between domestic violence
incidence rates and key town characteristics are presented in the first three
columns of Table 3. Results in Column 1 shows that without controlling for
other variables, the presence of a domestic violence agency in the town
TABLE 1
Correlation Between Reported Domestic Violence per 1,000 Residents and
Socioeconomic Indicators, 2000
Domestic Violence
Incidents
Domestic Violence
Assaults
Percent below poverty line 0.683n n
0.797n n
Percent receiving public
assistance
0.775n n
0.849n n
Percent female-headed
households
0.782n n
0.841n n
Percent unemployed 0.514n n
0.581n n
n n
po0.01.
NOTE: N 5 169 Connecticut towns and cities.
SOURCES: U.S. Bureau of the Census and the Connecticut Department of Public Safety.
TABLE 2
Socioeconomic Indicators of Towns With and Without Domestic Violence
Agencies, 2000
Towns With an
Agency
(N 5 18)
Towns Without an
Agency
(N 5 151)
Percent below poverty linen n
12.21 3.94
(7.27) (2.18)
Percent receiving public assistancen n
5.7 1.66
(3.82) (1.07)
Percent female-headed householdsn n
15.19 8.12
(7.73) (2.4)
Percent unemployedn n
6.89 3.43
(3.74) (1.81)
Crime rate per 1,000 residentsn n
52.48 19.38
(34.66) (10.92)
Domestic violence incidences per 1,000
residentsn n
8.49 3.51
(3.76) (2.35)
Domestic violence assaults per 1,000
residentsn n
3.08 0.94
(1.95) (0.73)
n n
po0.01.
NOTES: N 5 169 Connecticut towns and cities; t test of means, standard errors in parentheses.
SOURCES: U.S. Bureau of the Census, Connecticut Department of Public Safety, Connecticut
Coalition Against Domestic Violence.
1156 Social Science Quarterly
relates to an increase of 5.3 arrests per 1,000 residents, but the presence of an
agency in a neighboring town has no effect.
The model presented in Column 2 of Table 3 has additional independent
variables to control for town characteristics. Similar to our hypothesis, the
proportion of population below the poverty line and the composite risk
indicator have a significant positive impact on the domestic violence rate.
A 1 percent increase in the proportion of population below the poverty
line relates to an increase of about 1 arrest for every 10,000 town residents. In
towns with a combination of high poverty rates, unemployment
rates, and proportions of female-headed households and public assistance
recipients, the number of arrests is greater by about 2.3 per 1,000 residents.
Also, the domestic violence rate of incidents is lower in towns with higher
median income, at a statistically significant level. No longer significant is the
variable of interest, location of a domestic violence agency, indicating that the
rate of domestic violence incidents is not related to the location of an agency.
When the models in the first two columns of Table 3 are tested for spatial
autocorrelation, the Moran’s I test indicates cause for concern. Although the
TABLE 3
OLS Regression Models Explaining Domestic Violence Incidents and Assaults per
1,000 Residents at the Town Level, 2000
Predictor
Domestic Violence Incidents Domestic Violence Assaults
Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Agency in town 5.262n n n
0.735 0.657 2.167n n n
0.052 0.032
(0.662) (0.732) (0.726) (0.107) (0.227) (0.226)
Agency in adjacent town 0.557 0.092 0.024 0.057 À 0.151 À 0.168+
(0.411) (0.319) (0.318) (0.151) (0.099) (0.099)
% below poverty line 0.117+
0.145n
0.090n n n
0.097n n n
(0.069) (0.070) (0.021) (0.022)
Risk indicator 2.267n n
2.253n n
0.460n
0.457n
(0.658) (0.651) (0.204) (0.203)
Log of median house
value
À 2.054n n n
À 2.438n n n
À 0.546n n n
À 0.646n n n
(0.458) (0.480) (0.142) (0.152)
% less than 18 years old À 0.028 À 0.011 0.025 0.030+
(0.057) (0.057) (0.018) (0.018)
Population density 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)
Crime rate per 1,000
residents
0.015 0.012 0.004 0.004
(0.016) ((0.016) (0.005) (0.005)
Located west of CT River 0.750n
0.195+
(0.365) (0.113)
Moran’s I 0.19n n
0.09+
0.07 0.22n n
À 0.02 À 0.02
Adjusted R2
0.271 0.584 0.593 0.331 0.731 0.733
+
po0.10; n
po0.05; n n
po0.01; n n n
po0.001.
NOTES: N 5 169 Connecticut towns and cities. The above are OLS models with standard errors
shown in parentheses. INCIDENTS is a continuous variable, ranging from 0 to 15.28, with a mean
of 4.043. ASSAULTS is also continuous, ranging from 0 to 7.41, with a mean of 1.172.
Domestic Violence, Poverty, and Social Services 1157
additional socioeconomic variables in the second model explained some of
the spatial autocorrelation error, it is still significant at the 10 percent level. In
Model 3, the geographic variable, LOCATED WEST OF CT RIVER, adjusts the
spatial error, and spatial autocorrelation of the regression error is not statistically
significant. In this final model, the three statistically significant
variables of the previous model, percent below poverty line, the risk indicator,
and the log of the median house value, retain their significance. No
previously insignificant variables gain statistical significance. The new geographic
variable, however, has a statistically significant impact on incidents;
towns located west of the Connecticut River (where 14 out of the 18 agencies
are located) have higher incidence rates, holding other variables constant,
indicating that this clustering of agencies is related to areas of need.
Agency Location and Domestic Violence Assaults
Regression results examining the relationship between assault per 1,000
residents and key town characteristics are presented in the last three columns
of Table 3. The dependent variable in this case is limited to domesticviolence-related
arrests in which a charge of assault was made. This measure
of domestic violence is stricter and more narrow than the dependent variable
in the previous models. Without controlling for other influences, again, the
location of a domestic violence agency within the town limits has a significantly
positive relationship with the rate of assaults in a town, but the
location of an agency in a neighboring town does not.
Model 2, presented in the fifth column of Table 3, contains additional
independent variables measuring other town characteristics. Similar to the
models explaining domestic violence incidents, the proportion of the population
below the poverty line and the risk indicator have significantly
positive impacts on rates of domestic violence assaults. The log of the median
house value has a negative impact on domestic violence assaults.
Testing these models for spatial autocorrelation, the Moran’s I test indicates
cause for concern in only the first model. When the social-economic
town variables are added to Model 2, the error is adjusted and the spatial
autocorrelation of the regression error is not statistically significant. Model 2
is not spatially biased. However, we decided to add the geographic variable
measuring location in relation to the Connecticut River in the third model
because the variable was statistically significant in the model explaining
incidents. Similar to the results in Table 2, towns located to the west of the
Connecticut River tend to have higher domestic violence assault rates,
holding other variables constant. Moreover, the addition of this variable
changes the effect of the variable measuring the presence of an agency in an
adjacent town. When the CT RIVER variable is added to the model, the
adjacent town variable gains statistical significance. The findings indicate
that although towns west of the Connecticut River have higher rates of
1158 Social Science Quarterly
assaults, holding other variables constant, towns with an agency in a neighboring
town have lower rates of assaults by 0.168 assaults per 1,000 res-
idents.
Discussion
Our findings offer a beginning answer to our research question asked in
the introduction to this article. Is the location of domestic-violence-related
services at the town level related to rates of domestic violence? We hypothesized
that the data would show a geographic pattern of domestic violence
rates unrelated to the location of agencies after controlling for other environmental
risk factors such as poverty. On one hand, findings indicate no
relationship between the presence or lack of a domestic violence agency
within a town or in an adjacent town and the rate of police-reported domestic
violence incidents. On the other hand, the results of our second set of
models demonstrate that towns with agencies located in adjacent towns have
lower rates of domestic violence assaults, albeit this effect is very small.
Importantly, in both cases, we found that the clustering of agencies in the
western half of the state is related to higher rates of both domestic violence
incidents and domestic violence assaults. Furthermore, although the location
of domestic violence agencies at the town level does not match town-level
needs, neither were agencies overrepresented in affluent or privileged towns.
In fact, we find that domestic violence rates are correlated with other community
measures of disadvantage and agencies are overrepresented in these
areas.
These findings are bound by the limitations of the project. Most importantly,
the findings do not address the question of causal direction. We
analyze cross-sectional data from 2000 and thus cannot examine the direction
of causality. Second, the analysis unit is the town and the data are at the
polygon level. Given the sensitive and confidential nature of the topic,
available data on both arrest and service locations were limited and point
data were unavailable.
Third, our sample is limited to domestic abuse experiences that were
reported to the police and resulted in at least one arrest. Domestic violence is
an underreported crime and we likely vastly undercount the extent of the
issue using police data. Additionally, the data measure only physical assaults
and threats and do not capture emotional or economic violence and perhaps
sexual coercion as well. To the extent, however, that domestic violence
agencies serve women who are in the process of escaping the abuse, police
data are not an unreasonable data source. Agencies would not serve women
who have not yet recognized the situation and are not ready to seek help. It
is also possible that certain subpopulations of women such as immigrants or
rural residents may be less likely to reach out to either law enforcement or
shelter providers. Ideally, we could either match police calls with service
Domestic Violence, Poverty, and Social Services 1159
utilization or complement them with service utilization rates. While domestic
violence agencies do keep track of the number of women and children
served, these data had no town information. Thus, we were unable to
determine in what towns clients resided or even if they were Connecticut
State residents and could not include service utilization rates in our analyses.
Finally, the project is limited to an investigation of domestic violence
agencies. As noted in the methodology section of the article, this limits the
scope of our research in that victims also report seeking services from other
organizations such as churches and synagogues, health-care agencies, and
community organizations. Other work has shown that faith-based organizations
are more likely than other social service agencies to locate in areas of
great need (Allard, 2009). Our findings may be different if we were to
include other agencies. However, because domestic violence agencies are
designed particularly for this population and receive public funding to serve
them, their services are less volatile and vulnerable. Thus, we feel they merit
particular attention.
Conclusion
Although limited in scope, our analyses push current research into a new
realm. Research on domestic violence services should examine geography as well
as capacity and community coordination. The efficacy of service provision
should consider those for whom services are unavailable or inaccessible in
addition to the outcomes of those served by the programs. More qualitative
research into the influence of community sentiments regarding domestic violence
and service provision would also further our understanding. Similarly,
qualitative research on the program decisions of agencies themselves would add
valuable insight into the factors that agencies consider when locating or expanding
programs. Service providers already consider many place-related factors
such as security, confidentiality, and legal and police policies. An understanding
of how agencies also investigate and consider geographic need is lacking.
Future quantitative research can build on the modest findings presented here.
The limited scope of our research leaves a number of other questions unanswered,
such as the role of more general social service organizations and investigations
with other measure of domestic violence. If data on the point
location of organizations and incidences were to become available, research with
distance measures could be completed and complemented with investigations of
transportation accessibility and cost. Research in other states and with other
data sets could serve to bolster or refine our findings. Finally, and most obviously,
longitudinal studies could illuminate questions regarding causality and
the possible interconnection among service availability, propensity to report and
seek services, positive resident outcomes, and public awareness of the issue.
From a policy and planning perspective, the spatial pattern of domestic
violence and community characteristics should be to taken into consider-
1160 Social Science Quarterly
ation in designing the location and services of agencies. Our findings indicate
that established domestic violence agencies are not necessarily located
in areas of great need. Similar to other domestic violence research, however,
this study provides evidence that a regional perspective may be warranted
and that services just outside one’s immediate town may be appropriate.
Town borders, after all, are very real on paper, but may not pose barriers to
service update. It is possible that the closest agency is not on the other side of
one’s own town, but just down the block and in the neighboring town.
Future conversations about services should question scope, capacity, and
location at the town and regional level.
REFERENCES
Allard, S. 2004. Access to Social Services: The Changing Urban Geography of Poverty and Service
Provision. Washington, DC: Brookings Institute Survey Series.
———. 2009. Out of Reach: Place, Poverty, and the New American Welfare State. New Haven,
CT: Yale University Press.
Allard, S., R. Tolman, and D. Rosen. 2003. ‘‘Proximity to Service Providers and Service
Utilizations Among Welfare Recipients: The Interaction of Place and Race.’’ Journal of Policy
Analysis and Management 22(4):599–613.
Arizona Coalition Against Domestic Violence. 2000. Best Practices Manual for Domestic
Violence Programs. Phoenix, AZ: ACADV.
Benson, M. L., and G. L. Fox. 2004. When Violence Hits Home: How Economics and Neighborhood
Play a Role. Washington, DC: U.S. Department of Justice, National Institute of
Justice.
Benson, M. L., G. L. Fox, A. DeMaris, and J. VanWyk. 2003. ‘‘Neighborhood Disadvantage,
Individual Economic Distress and Violence Against Women in Intimate Relationships.’’
Journal of Quantitative Criminology 19(3):207–35.
Blank, R. M. 2004. Poverty, Policy and Place: How Poverty and Policies to Alleviate Poverty are
Shaped by Local Characteristics. Working Paper 04-02. Rural Poverty Research Center University
of Missouri.
Browning, C. R., T. Leventhal, and J. Brooks-Gunn. 2005. ‘‘Sexual Initiation in Early
Adolescence: The Nexus of Parental and Community Control.’’ American Sociological Review
70:758–78.
Burke, J. G., P. O’Campo, and G. L. Peak. 2006. ‘‘Neighborhood Influences and Intimate
Partner Violence: Does Geographic Setting Matter?’’ Journal of Urban Health 83(2):182–94.
Carlson, B. E., A. P. Worden, M. van Ryn, and R. Bachman. 2003. Violence Against Women:
Synthesis of Research for Practioners. Document 199577. Report to the U.S. Department of
Justice.
Catalano, S. 2007. Intimate Partner Violence in the United States. Washington, DC: U.S.
Department of Justice. Available at hhttp://bjs.ojp.usdoj.gov/index.cfm?ty=pbde-
tail&iid=1000i.
Clark, S. J., M. R. Burt, M. M. Schulte, and K. E. Maguire. 1996. Coordinated Community
Responses to Domestic Violence in Six Communities: Beyond the Justice System. Washington,
DC: Urban Institute Research Paper.
Domestic Violence, Poverty, and Social Services 1161
Davis, M. 1999. ‘‘The Economics of Abuse: How Violence Perpetuates Women’s Poverty.’’
Pp. 17–30 in R. Brandwein, ed., Battered Women, Children, and Welfare Reform: The Ties
that Bind. Thousand Oaks, CA: SAGE Publications.
Dembo, R., S. Belenko, K. Childs, J. Wareham, and J. Schmeidler. 2009. ‘‘Individual and
Community Risk Factors and Sexually Transmitted Diseases Among Arrested Youths: A Two
Level Analysis.’’ Journal of Behavioral Medicine 32:303–16.
DiBartolo, L. 2001. ‘‘The Geography of Reported Domestic Violence in Brisbane: A Social
Justice Perspective.’’ Australian Geographer 32(3):321–41.
Fox, G. L., and M. L. Benson. 2006. ‘‘Household and Neighborhood Contexts of Intimate
Partner Violence.’’ Public Health Reports 121:419–27.
Frye, V., S. Galea, M. Tracy, A. Bucciarelli, S. Putnam, and S. Wilt. 2008. ‘‘The Role of
Neighborhood Environment and Risk of Intimate Partner Femicide in a Large Urban Area.’’
Research and Practice 98(8):1473–79.
Grana, S. J. 2001. ‘‘Sociostructural Considerations of Domestic Femicide.’’ Journal of Family
Violence 16:421–35.
Hongfei, L., C. A. Calder, and N. Cressie. 2007. ‘‘Beyond Moran’s I: Testing for
Spatial Dependence Based on the Spatial Autoregressive Model.’’ Geographical Analysis
39(4):357–75.
Hotaling, G. T., and D. B. Sugarman. 1990. ‘‘A Risk Marker Analysis of Assaulted Wives.’’
Journal of Family Violence 5(1):1–13.
Koenig, M. A., A. Saifuddin, M. B. Hossain, and A. B. M. K. A. Mozumder. 2003.
‘‘Women’s Status and Domestic Violence in Rural Bangladesh: Individual and Community
Effects.’’ Demography 40(2):269–88.
Lloyd, S., and N. Taluc. 1999. ‘‘The Impact of Recent Partner Violence on Women’s
Capacity to Maintain Work.’’ Violence Against Women 5(4):370–92.
Logan, T. K., L. Shannon, J. Cole, and J. Swanberg. 2007. ‘‘Partner Stalking and Implications
for Women’s Employment.’’ Journal of Interpersonal Violence 22:268–91.
Lyon, E. 2000. Welfare, Poverty and Abused Women: New Research and its Implications.
Policy and Practice Paper 0. Harrisburg, PA: National Resource Center of Domestic
Violence.
Lyon, E., S. Lane, and A. Menard. 2008. Domestic Violence Shelters: Survivors’ Experiences.
Washington, DC: National Institute of Justice.
Miles-Doan, R. 1998. ‘‘Violence Between Spouses and Intimates: Does Neighborhood
Context Matter?’’ Social Forces 77(2):623–45.
O’Campo, P., J. Burke, G. L. Peak, K. A. McDonnell, and A. C. Gielen. 2005. ‘‘Uncovering
Neighborhood Influences on Intimate Partner Violence Using Concept Mapping.’’ Journal of
Epidemiology and Community Health 59(7):603–08.
Pain, R. 2000. ‘‘Place, Social Relations and the Fear of Crime: A Review.’’ Progress in Human
Geography 24(3):365–87.
———. 2001. ‘‘Gender, Race, Age and Fear in the City.’’ Urban Studies 38:899–913.
Panelli, R., J. Little, and A. Kraack. 2004. ‘‘A Community Issue? Rural Women’s Feelings of
Safety and Fear in New Zealand.’’ Gender, Place, and Culture 11(3):445–67.
Postmus, J. L., M. Severson, M. Berry, and J. A. Yoo. 2009. ‘‘Women’s Experiences of
Violence and Seeking Help.’’ Violence Against Women 15:852–68.
1162 Social Science Quarterly
Raghavan, C., A. Mennerich, E. Sexton, and S. E. James. 2006. ‘‘Community Violence and
its Direct, Indirect, and Mediating Effects on Intimate Partner Violence.’’ Violence Against
Women 12(12):1132–49.
Reinelt, C. 1995. ‘‘Moving onto the Terrain of the State: The Battered Women’s Movement
and the Politics of Engagement.’’ Pp. 88–104 in M. M. Ferree and P. Y. Martin, eds.,
Feminist Organizations: Harvest of the New Women’s Movement. Philadelphia, PA: Temple
University.
Renzetti, C. M. 2009. Economic Stress and Domestic Violence. Applied Research Forum,
National Online Resource Center on Violence Against Women. Available at hhttp://
www.vawnet.org/category/Main_Doc.php?docid=2187i.
Sampson, R. J., S. W. Raudenbush, and F. Earls. 1997. ‘‘Neighborhoods and Violent Crime: A
Multilevel Study of Collective Efficacy.’’ Science 277:918–24.
Sampson, R. J., and W. J. Wilson. 1995. ‘‘Toward a Theory of Race, Crime, and Urban
Inequality.’’ Pp. 37–54 in J. Hagan and R. D. Peterson, eds., Crime and Inequality. Stanford,
CA: Stanford University Press.
Schechter, S. 1982. Women and Male Violence. Boston, MA: South End Press.
Sullivan, C. M., and D. I. Bybee. 1999. ‘‘Reducing Violence Using Community-Based
Advocacy for Women with Abusive Partners.’’ Journal of Consulting & Clinical Psychology
67(1):43–53.
Sullivan, C. M., and T. Gillum. 2001. ‘‘Shelters and Other Community-Based Services for
Battered Women and Their Children.’’ Pp. 247–60 in C. M. L. Renzetti, J. L. Edleson, and
R. K. Bergen, eds., Sourcebook on Violence Against Women. Thousand Oaks, CA: Sage.
Tan, C., J. Basta, C. M. Sullivan, and W. S. Davidson. 1995. ‘‘The Role of Social Support in
the Lives of Women Exiting Domestic Violence Shelters: An Experimental Study.’’ Journal of
Interpersonal Violence 10(4):437–51.
Tjaden, P., and N. Thoennes. 2000. Extent, Nature and Consequences of Violence Against
Women: Findings from the National Violence Against Women Survey. Washington, DC:
National Institute of Justice and the Centers for Disease Control and Prevention.
Tolman, R., and J. Raphael. 2000. ‘‘A Review of Research on Welfare and Domestic
Violence.’’ Journal of Social Issues 56(4):655–82.
Tutty, L. M., G. Weaver, and M. A. Rothery. 1999. ‘‘Residents’ Views of the Efficacy of
Shelter Services for Assaulted Women.’’ Violence Against Women 5(8):898–925.
Warrington, M. 2001. ‘‘‘I Must Get Out’: The Geographies of Domestic Violence.’’ Transactions
of the Institute of British Geographers 26:365–82.
Weisz, A. R., R. M. Tolman, and L. Bennett. 1998. ‘‘An Ecological Study of Nonresidential
Services for Battered Women Within a Comprehensive Community Protocol for Domestic
Violence.’’ Journal of Family Violence 13(4):395–415.
Whitzman, C. 2007. ‘‘Stuck at the Front Door: Gender, Fear of Crime and the Challenge of
Creating Safer Space.’’ Environment and Planning 38(11):2715–32.
Williams, S. L., and K. D. Mickelson. 2004. ‘‘The Nexus of Domestic Violence and Poverty:
Resilience in Women’s Anxiety.’’ Violence Against Women 10:283–93.
Wilson, W. J. 1987. The Truly Disadvantaged: The Inner City, The Underclass, and Public
Policy. Chicago, IL: University of Chicago Press.
Zweig, J. M., K. A. Schlichter, and M. R. Burt. 2002. ‘‘Assisting Women Victims
of Violence Who Experience Multiple Barriers to Service.’’ Violence Against Women 8(2):
162–80.
Domestic Violence, Poverty, and Social Services 1163
Copyright of Social Science Quarterly (Blackwell Publishing Limited) is the property of Wiley-Blackwell and
its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.