Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

836075

research-article2019
SCUXXX10.1177/2329496519836075Social CurrentsDollar et al.

Article
Social Currents

Joblessness, Poverty, and


2019, Vol. 6(4) 343­–360
© The Southern Sociological Society 2019
Article reuse guidelines:
Neighborhood Crime: sagepub.com/journals-permissions
DOI: 10.1177/2329496519836075
https://doi.org/10.1177/2329496519836075

Testing Wilson’s Assertions journals.sagepub.com/home/scu

of Jobless Poverty

Cindy Brooks Dollar1, Ellen A. Donnelly2,


and Karen F. Parker2

Abstract
Political and public rhetoric often tout the detrimental effects of poverty and the stabilizing
effects of jobs when discussing issues of crime and disorder. Several decades ago, William
Julius Wilson proposed that while poverty has been a long-standing problem across U.S. cities,
a “new urban poverty” associated with the massive disappearance of jobs may better explain
neighborhood crime and violence. Although scholarship continues to theorize about the unique
criminogenic effects of area-level poverty and joblessness, criminological research has not
systematically examined the distinct effects of poverty and jobless poverty on neighborhood
crime. In the present analysis, we use data from the National Neighborhood Crime Study
(NNCS) to provide the first empirical investigation of how joblessness, separately from poverty,
influences rates of homicide and other reported violence across 6,406 neighborhoods in 53
U.S. metropolitan areas. A spatially adjusted, multilevel regression analysis reveals that places
with poverty and joblessness are distinct from areas with poverty and no marked joblessness.
Analysis by neighborhood further reveals that joblessness has a stronger effect on violent crime
than poverty itself in severely impoverished areas. As such, we find support for Wilson’s distinct,
criminogenic notion of jobless poverty.

Keywords
joblessness, poverty, jobless poverty, neighborhood crime, homicide

Evidence of political and economic divisive- upward mobility starting in the 1970s. Adding
ness in the United States has grown. Rising to that, the economy is an increasingly
inequality and gaps in achievement, wages, and bifurcated reality, with low-income families
income are well documented (Deaton 2013;
Piketty and Saez 2003), especially if one exam- 1
ines educational outcomes among children University of North Carolina at Greensboro, USA
2
University of Delaware, Newark, USA
(Mayer 2001; L. McCall 2000) or differences
across race and ethnic groups (Iceland 2004; Corresponding Author:
Tienda 1989). While substantial progress was Cindy Brooks Dollar, Department of Sociology,
University of North Carolina at Greensboro, 337 Frank
made in the 1950s and 1960s to reduce inequal- Porter Graham Building, P.O. Box 26170, Greensboro,
ity in education, health, and earnings (Piketty NC 27402-6170, USA.
and Saez 2003), trends shifted away from Email: cbdollar@uncg.edu
344 Social Currents 6(4)

experiencing more rapid declines relative to dissolution, weakened community control, and
high-income families (Kopczuk, Saez, and crime (Wilson 1996). Wilson concluded that
Song 2010). Indeed, the latter part of the twen- prior to the 1970s, a substantial portion of high
tieth century was a period of economic reces- poverty areas had working populations, but by
sion and of prosperity, where the healthy the 1990s, available jobs in accessible areas
economy of the 1960s met the recession of the had largely disappeared.
1970s (Morris and Western 1999). The present study offers the first empirical
William Julius Wilson’s work brought investigation of how joblessness differentially
scholarly attention to the economic transfor- influences rates of violent crime in U.S. neigh-
mation heightening the concentration of pov- borhoods when it is separated from poverty. In
erty and joblessness from the 1970s through doing so, we return to Wilson’s emphasis that
2000s (Wilson 1987, 1996). This industrial places in which poverty exists are distinct from
restructuring (Kasarda 1989, 1992; Wilson places where poverty and joblessness exist.
1987, 1991) resulted in a massive exodus of all We also evaluate Wilson’s hypothesis that job-
major industries and businesses from U.S. cit- lessness is more criminogenic than poverty,
ies, eroding structural conditions of many especially in communities with severe eco-
urban centers and leaving residents in margin- nomic deprivation. Using neighborhood and
alized positions within the labor markets or metropolitan area-level data from the U.S
jobless (Krivo and Peterson 1996, 2000; Parker Census Bureau and the National Neighborhood
2004, 2008). Evidence of the impact of indus- Crime Study (NNCS), we estimate a series of
trial restructuring on U.S. cities has been well spatially adjusted, nonlinear multilevel models
documented (Harrington and Levinson 1985; to separately investigate the effects of poverty
Massey and Denton 1993; Massey, Gross, and and joblessness on neighborhood homicide
Eggers 1991; Waldinger 1996; Wilson 2009), and other violent crime rates. We further
including a sizable drop in the percentage of explore the roles of joblessness to differences
persons employed in manufacturing jobs in violent crime across neighborhoods with
between 1970 and 2000, where some studies differing levels of poverty, testing Wilson’s
suggested the decline exceeds 40 percent projection that high poverty areas are most
(McCall, Parker, and MacDonald 2008; Parker affected by “new urban poverty.”
2008).
Most relevant to this article is capturing Poverty, Joblessness, and
Wilson’s depiction of a “new urban poverty”—
a jobless poverty that is associated with rising
Crime: The Differences
unemployment and joblessness. With the pub- Wilson (1996) theorized joblessness as a new
lication of When Work Disappears over 20 phenomenon that led to socio-political isola-
years ago, Wilson (1996) argued that areas tion, thus reproducing and concentrating dis-
where joblessness and poverty coexist produce advantage and resulting in crime-related
a unique social reality as compared with pov- outcomes. While criminological theory has
erty-stricken areas with available jobs. Since historically linked poverty to higher crime
this publication, Wilson (2003:1103) has fur- rates (e.g., Shaw and McKay 1942), scholars
ther clarified his position by writing “the con- have long debated about the link between pov-
sequences of high neighborhood joblessness erty and crime. These debates have focused on
are more devastating than those of high neigh- the criminogenic factors associated with pov-
borhood poverty.” Wilson, then, has consis- erty (absolute deprivation) versus income
tently described poverty and joblessness as inequality (relative deprivation) (Bailey 1984;
distinct social features. While poverty has been Blau and Blau 1982; Messner 1982), the
a chronic condition of many inner cities, he importance of unemployment or labor stratifi-
conceptualizes jobless poverty as a new, isolat- cation over poverty (Chiricos 1987; Crutchfield
ing neighborhood characteristic that increased 1989; Levitt 2001), and arguments that it is the
various social disorders, including family “concentration” of disadvantage in areas over
Dollar et al. 345

specific structural features, such as poverty and declines in low-skill jobs in American cit-
(Land, McCall, and Cohen 1990; Krivo and ies (Wilson 1996, 1999). Wilson (1999) identi-
Peterson 1996; Sampson and Wilson 1995; fies several, interrelated bases of joblessness.
Wilson 1978). Moreover, Crutchfield adds that He contends that the growth of international
marginal or unstable employment is as relevant trading agreements, economic restructuring,
as the absence of employment, where the types technological changes, and increasing college
of jobs (e.g., part-time or low skilled service enrollment all resulted in a decreased need for
work) located in neighborhoods are linked to low-skilled labor historically common in poor
crime (Crutchfield 2014; Crutchfield and areas. As a result of these changes, many jobs
Pitchford 1997). moved from the inner-city to the suburbs, and
While some ecological studies on crime the few jobs that remained were extremely low
suggest support for the claim that poverty is wage. In response to the job loss, persons with
one of several characteristics of communities enough resources moved from their inner-city
that lead to crime (McLanahan, Garfinkel, and residences, which left urban neighborhoods
Watson 1988; Messner and Tardiff 1986; and areas surrounding the cities with a dearth
Patterson 1991), other research finds inconsis- of social and economic resources, including a
tent connections between poverty and crime lack of mentors promoting conventional or
across aggregated units (e.g., Kovandzic, pro-social routines. The enduring inner-city
Vieraitis, and Yeisley 1998; Land et al. 1990; population, hence, not only remained poor but
Pratt and Cohen 2005; Warner and Rountree also became socially and politically estranged.
1997). For example, Krivo and Peterson (1996) Through this isolation, community support
found that extremely disadvantaged neighbor- waned, and the stability of community institu-
hoods have unusually high rates of crime; tions and controls weakened. As such, neigh-
however, Hipp and Yates (2011) found that borhood relationships were further disrupted,
crime does not rise sharply in extreme-poverty and familial and community conflict erupted.
neighborhoods. Although studies often exam- The detrimental consequences of jobless-
ine different neighborhood locales at different ness have been recorded by numerous scholars,
time points, inconsistencies in poverty-crime particularly with regard to criminological out-
relationships across studies are nonetheless comes. For example, Crutchfield, Matsueda,
notable. and Drakulich (2006) concluded that high rates
In investigating poverty’s effect on crime, of neighborhood disorder and violence were
researchers generally conceptualize poverty as related to socioeconomic disadvantage result-
some official measure of economic deprivation ing from labor instability. In addition, Anderson
(i.e., percentage of the population living below (1999) revealed that poor inner-city communi-
the federal poverty threshold). Unlike poverty, ties marked with joblessness became socially
joblessness is theorized to produce social mar- isolated and turned to violence as a cultural
ginalization from “mainstream” or dominant maladaptation. In his earlier work, Anderson
society. Wilson (1987), in fact, characterizes (1989) claimed that a lack of legitimate jobs
jobs as a conventional mechanism of social was the primary reason many inner-city resi-
control, an opportunity for economic and dents turn to sell drugs. Bourgois (1995) simi-
social stability, and a means of upward mobil- larly found that working in the illegal drug
ity. Areas marked by a state of joblessness, market was one way to obtain economic
then, lack conventional controls and opportu- resources in an area where jobs were scarce,
nities, consequently producing socio-political unstable, and low paid. Given the illegal nature
marginalization. In turn, joblessness increases of the work, this job also allowed men to dem-
participation in illegitimate labor markets and onstrate their masculinity through perfor-
violence (Wilson 1978). mances of aggression, risk-taking, and strength.
The causes of joblessness are multifaceted. In support for the criminogenic capacity of
In brief, joblessness is caused by macroeco- weakened community controls, Pattillo (1999)
nomic changes, including deindustrialization argued that “criminal networks” can overpower
346 Social Currents 6(4)

conventional or “pro-social” ones in areas had missing crime data (e.g., nonzero, n =
where pro-social resources are unavailable or 303), more than 50 percent of the tract’s popu-
largely limited, which results in high rates of lation was housed in group quarters (n = 164),
violent crime. or less than 300 people lived in the tract (n =
In sum, it is a “new urban poverty” or a job- 623). This process reduced the total number of
less poverty that stands in considerable con- metropolitan areas by two. We further subset
trast to the poverty of the past (Wilson 2003). our data to observations with complete infor-
As When Work Disappears so clearly illus- mation on key variables. Due to missing data
trates, neighborhoods in which people are poor on these covariates, our final analytic sample
and jobless are different from neighborhoods features 6,406 tracts nested within 53 metro-
marked with poverty but with available jobs. politan areas throughout the United States. Our
Wilson (2003:1103) states “. . . many of today’s analytic sample then retains more than two
problems in the inner-city ghetto neighbor- thirds of NNCS observations.1
hoods—crime, family dissolution, welfare,
low levels of social organization and so on—
Measures
are fundamentally a consequence of the disap-
pearance of work.” It is this larger process, a Outcomes. Two outcomes of interest are the
process where joblessness affects violence in number of homicides and the number of vio-
urban neighborhoods separately from the lent crime incidents at the tract level from 1999
influences of poverty that we pursue in the to 2001. The NNCS defines homicide as offi-
present empirical investigation. Below, we cial reports of murder and nonnegligent man-
describe the data and methodological strategy slaughter. We define nonlethal, violent crime
used to capture the unique links between pov- (hereafter violent crime) as official reports of
erty and joblessness on neighborhood levels of robbery, aggravated assault, and forcible rape.
violence. The definition follows NNCS’s measure of
violent crime, but it excludes homicide inci-
dents. We use homicide and violent crime as
Data and Method outcomes because these crimes are more overt
This study uses data from the NNCS to under- and more likely to be measured reliably than
stand the impacts of joblessness as a “new pov- property crimes (Browning et al. 2010).
erty” on violent crime. Exceptional in its
compilation of crime and sociodemographic Tract-level explanatory variables. Our primary
data at the neighborhood level, the dataset independent variables are joblessness and pov-
gathers official tract-level crime data on the erty rates at the tract level. We measure jobless-
Federal Bureau of Investigation’s (FBI) Index ness as an index. Our joblessness index
crimes. It also includes sociodemographic combines measures of the percentage of the
information collected from the 2000 U.S. civilian workforce age 16 and over who are
Census of Population and Housing for each unemployed (unemployment rate), the percent-
census tract and the metropolitan statistical age of the civilian workforce age 16 to 64 who
area (hereafter metropolitan area) for which are unemployed or not in the labor force (job-
the tract is located. Observations at the tract less rate), the proportion of the employed civil-
and metropolitan area levels allow us to exam- ian population employed in the six occupations
ine what social conditions affect crime patterns with the lowest mean incomes (secondary sec-
at two geographic levels. tor workers),2 and the proportion of the
The NNCS includes a representative sample employed civilian population age 16 and over in
of large U.S. metropolitan areas and their cor- management, professional, and related occupa-
responding census tracts. The original NNCS tions (percentage professionals and managers,
sample features 9,593 census tracts across 64 reverse coded) (Cronbach’s α = .85). Our defi-
metropolitan areas. We then removed observa- nition of joblessness attempts to approximate
tions from this original dataset if a census tract Wilson’s conceptualization of “new
Dollar et al. 347

urban poverty” by taking into consideration years. Percentage divorced measures the pro-
unemployment, labor force attachment, and portion of men aged 15 and over who are
the absence of high-paying jobs. Likewise, the divorced.
index extends previous empirical conceptions Prior theory and research further suggest
of joblessness as a measure of labor force that racial-ethnic heterogeneity may positively
participation (Krivo and Peterson 2000). Our influence crime rates (e.g., Sampson and Graif
joblessness index is constructed as a regression- 2009). Wilson’s work focuses primarily on
based factor score that has a mean of 0 and stan- African American neighborhoods. However,
dard deviation of 1 for the entire sample. Like a the United States has seen a significant increase
z-score, positive, higher values on the jobless- in other minority populations, especially from
ness index indicate that a tract experiences more Asia and Latin America. We approximate
problems of joblessness while negative, lower racial and ethnic heterogeneity in an area by
values suggest that a tract has fewer issues incorporating a tract-level measure of the per-
(DiStefano, Zhu, and Mindrila 2009). centage of the population that identifies as
We define poverty as the percentage of the non-Latino, black; non-Latino, Asian; or non-
population at or below the official poverty level Latino and of some Other race. To comprehen-
in 1999. We first use poverty as an explanatory sively examine the role of immigration, we
variable in our models. We also use this opera- computed an immigration index that includes
tionalization of poverty to separate neighbor- four variables: percentage Hispanic, percent-
hoods affected by the longstanding problems of age foreign-born, percentage new immigrants,
economic scarcity. Prior ecological studies and percentage linguistically isolated.
define high poverty areas as those areas where Finally, given recent research on the rela-
40 or more percent of the population lives within tionships between youth institutional engage-
the official poverty guidelines (e.g., Jargowsky ment and homicide, we include tract-level
and Bane 1990; Ricketts and Sawhill 1988; measures of these variables (e.g., Dollar et al.
Wacquant and Wilson 1989). Areas of moderate 2017; P. L. McCall et al. 2013). Consistent
poverty are marked by 20 to 39 percent of the with these studies, we measure the engage-
population living in poverty, and areas of low ment of young adults using two variables: the
poverty are identified by less than 20 percent of percentage of persons enrolled in college and
the population being impoverished (Krivo and the percentage of persons age 15 to 29 who are
Peterson 1996). married with a spouse present. These measures
capture dimensions of engagement with edu-
Tract-level covariates.  We control for other rel- cational and family institutions.3
evant factors that influence neighborhood
crime levels. We first introduce two demo- Metropolitan-area-level controls.  Our models also
graphic measures: age structure and sex ratio. adjust for the characteristics of metropolitan
Age structure is measured as the percentage of areas. We add in controls for the log of the area
the population that is aged 15 to 29. The sex of the metropolitan area in square meters and its
ratio of neighborhoods is defined as the ratio of logged total population. Prior ecological studies
males to females living in the area (Glick, suggest socioeconomic and racial inequality in a
Beresford, and Heer 1963; Messner and Samp- metropolitan are influences social relations at
son 1991; Schoen 1981; Wilson 1987). lower levels of aggregation, especially in urban
We then take into account residential and crime (e.g., Hipp 2011; Massey and Denton
family stability within neighborhoods that may 1993; Wilson 1987). We first introduce metro-
affect crime (e.g., Dollar et al. 2017; Sampson politan area Gini Index to estimate household
1987; Sampson and Groves 1989; Shaw and income inequality. This index measures the dis-
McKay 1942; Wilson 1996). Residential insta- tribution of household income among the popu-
bility calculates the standardized average score lation of the metropolitan area. We also adjust
of the percentage of rentals and the percentage for segregation between racial and ethnic groups.
of the population that has moved in the last 5 Ethnoracial segregation is operationalized using
348 Social Currents 6(4)

two dissimilarity indices that indicate the even- Second, we estimate another set of spatially
ness in the distribution of non-Hispanic, whites adjusted, multilevel models to assess the
and blacks and non-Hispanic, whites and His- impacts of joblessness on violent crime in
panic/Latinos across census tracts within met- areas of low, moderate, and severe problems of
ropolitan areas (Peterson and Krivo 2000). The poverty. We anticipate that joblessness may be
dissimilarity indices are calculated for two more criminogenic in high poverty neighbor-
groups at a time and estimate how much one hoods relative to neighborhoods with less pov-
group (i.e., Hispanic/Latinos or blacks) would erty due to Wilson’s articulation of jobless
have to relocate to have the same proportional poverty and its devastating consequences.
representation in a neighborhood as the other Our multilevel models estimate the share in
group (i.e., white) (Iceland and Weinberg the variance of violent crime rates related to
2002:119). neighborhood conditions, such as joblessness
We then introduce three variables to capture and poverty. The models take geographical
the characteristics of the labor force in a metro- hierarchies into consideration, such that tracts
politan area. We first estimate the availability as the unit of analysis (Level 1) are nested
of jobs in and around the area. An employment within metropolitan areas as aggregated con-
change ratio identifies the change in the num- textual units (Level 2). We specifically intro-
ber of employees in a metropolitan area duce a generalized linear mixed model
between 1990 and 2000. A suburban employ- (GLMM) that adjusts for metropolitan-level
ment ratio estimates the percentage of workers effects by allowing the intercepts of Level 1
who work in the suburban ring of the metro- factors to vary while keeping slopes of Level 1
politan area divided by the percentage of the factors fixed. As a class of models, GLMM
area’s total population that lives in this subur- extends the linear mixed effects approach by
ban ring in 2000. We also account for the adjusting for different distributions of one’s
diversity of jobs in a metropolitan area. outcome variables (West, Welch, and Galecki
Industrial diversity is measured by a dissimi- 2014). Our GLMMs use the negative binomial
larity index examining employment across 20 distribution because our outcomes involve
industries within a metropolitan area relative counts, and our data display overdispersion
to employment in these industries throughout (i.e., the variance of our response variables
the nation. Last, we construct a categorical exceeds the mean) that violates assumptions of
variable to identify the region of the country in the Poisson distribution (Raudenbush and
which a metropolitan area is located. We spec- Bryk 2002). Correction for overdispersion
ify whether the metropolitan area is located in through the negative binomial distribution
the U.S. South, North, West, or Midwest. may then provide more accurate estimates of
regression coefficients and standard errors
(Ismail and Jemain 2007). We finally adjust
Analytic Strategy
our negative binomial regression models with
Given the importance of capturing Wilson’s an exposure variable of total population in the
claims concerning the influence of joblessness census tract.
on crime, our analytical approach has two Our GLMMs can further take into account
stages. First, we introduce a multilevel statisti- the possibility of spatial autocorrelation.
cal model that adjusts for all neighborhood and Neighborhoods have the characteristic of
metropolitan-area-level conditions related to being more socially and culturally similar to
homicide and violent crime.4 Our model also units in close proximity to them. As such,
takes into account the possibility of spatial models may be misspecified, and estimates of
autocorrelation, or the nonrandom distribution predictor effects can be biased if spatial pat-
of crime patterns given the proximity of tracts terns are not taken into consideration (Anselin
to one another. This model is used to assess et al. 2000; Baller et al. 2001; Ludwig et al.
whether joblessness has a distinct role from 2008; Sampson 2008). We use the spdep pack-
poverty in contributing crime outcomes. age in R to examine spatial autocorrelation in
Dollar et al. 349

our data (Bivand, Pebesma, and Gómez-Rubio the same specification to predict homicide and
2008). Spatial autocorrelation is detected by other violent crimes in all tracts and then in
identifying the nearest first neighbor of a cen- tracts with varying levels of poverty.
sus tract among all census tracts from all met-
ropolitan areas. We identify neighbors based
Results
on the centroids of the census tract and require
neighbors to be found between 0 and two times Table 1 provides summary statistics of our tract-
the minimum distance that all centroids would level and metropolitan area-level variables. The
have a distance-based neighbor. This upper- first column shows descriptive measures for the
bound of interpoint distance eliminates any no- entire analytic sample (n = 6,406) and the
neighborhood census tracts from our national remaining columns present measures for all
sample (Bivand 2018). variables for tracts with low, moderate, or high
After locating nearest distance neighbor- levels of poverty. These three subsamples fea-
hoods, we inspect the homicide and violent ture 3,914 tracts with less than 20 percent living
offenses for spatial autocorrelation. We first below the poverty line (low), 1,972 tracts with
complete a global Moran’s I test that evaluates 20 to 39 percent impoverished population
deviations from the mean outcome and then (moderate), and 520 tracts with 40 percent or
spatially weights such deviations based on more impoverished population (high). Variation
knowledge of neighboring tracts in our sam- in crime rates is evident across samples.
ple. Evidence of spatial autocorrelation was Neighborhoods with high levels of poverty have
significant (global Moran’s I = 0.105, p < .01 an average of 3.16 homicides over a three-year
for homicide offenses; 0.144, p < .01 for other period.5 The number of homicides is lower in
violent offenses), indicating that crime is not neighborhoods with less poverty, as mid-range
randomly distributed across tracts. We supple- and low poverty areas report an average of 2.62
ment this global test with local Moran’s I per- and 0.73 homicide incidents during this period,
mutation tests for each census tract based on a respectively. The average number of homicides
spatially weighted matrix of neighbors. is 1.51 homicides per tract across all neighbor-
Approximately 42 percent of tracts (n = 2,738) hoods. Similar patterns are seen in nonfatal vio-
showed spatial dependence in violent crime lent crime. High poverty rates have the highest
incidents and about a quarter of tracts (n = number of violent crimes (200 incidents per
1,581) displayed spatial patterns in homicide tract) compared with mid-range (180 incidents)
incidents (i.e., local Moran’s I values were sta- and low (73 incidents) poverty tracts.
tistically significant at p < .05). These local Tract-level conditions also vary by poverty
tests mean that crime outcomes in census tracts level. Joblessness is most pronounced in tracts
are affected by incidents in neighboring tracts. with the highest poverty rates. Joblessness is
These tests highlight the need to adjust for spa- still common in moderate poverty areas (as
tial relationships in our models of fatal and seen in an average joblessness index score
nonfatal violent crime. above 0), but it is less so in low poverty areas
Given the evidence of spatial autocorrela- (as seen in an average joblessness index score
tion, we construct spatially lagged dependent below 0). Tracts with 40 percent or more of
variables based on weighted average of counts residents living in poverty have younger and
of homicide or violent crime incidents in neigh- more female populations. More disadvantaged
boring tracts. We add these spatial weights to neighborhoods also show greater problems of
the full set of control variables to reduce poten- residential instability. Nonwhite populations
tial biases in our estimates from spatial autocor- are largest in neighborhoods with high poverty,
relation. This approach allows us to effectively but immigration is more common in areas of
address local and global forms of spatial moderate rates of poverty. Less impoverished
autocorrelation, as well as provide a more inter- neighborhoods have higher rates of college
pretable spatial lag process (e.g., Lambert, enrollment and marriage among younger
Brown, and Florax 2010). We run GLMMs with populations.
350 Social Currents 6(4)

The conditions of metropolitan areas are neighborhood turnover. Greater proportions of


also presented in Table 1. The size and popula- college enrollment are linked to reduced homi-
tion of metropolitan areas are similar across the cide and violent crime incidents in neighbor-
three subsamples of tracts. These metropolitan hoods. At the metropolitan area level, large
areas show comparable levels of income populations are linked to more lethal and non-
inequality. Black-white and Hispanic/Latino- lethal incidents. Metropolitan areas with lower
white segregation is slightly higher in metro- suburban employment ratios appear to show
politan areas containing high poverty tracts. All lower rates of violent crime, though this asso-
metropolitan areas have similar suburban ciation is weak relative to other relationships (p
employment ratios and experience changes in < .05).
the size of their employed populations between
1990 and 2000, but areas containing high pov-
erty tracts have less diversity in their industries.
Violent Crime by Neighborhoods’
Metropolitan areas are distributed similarly Poverty Level
across regions of the United States. In all three Because high poverty neighborhoods experi-
subsamples, about 30 percent of the metropoli- ence more homicides and violent crime inci-
tan areas are located in the South, 15 percent in dents relative to moderate and low poverty
the North, and 26 percent in the West with the neighborhoods, a priority is to understand the
remainder being in the Midwest. effects of joblessness on crime rates across
poverty levels. We introduce separate spatially
Multilevel Spatial Analysis of adjusted, multilevel negative binomial regres-
sions for our three subsamples of tracts to
Homicide and Violent Crime Levels understand if high poverty areas differ from
Table 2 reports the estimated effects of poverty moderate and low poverty areas in crime pat-
and joblessness on homicide and other violent terns. Table 3 first reports estimates predicting
crime from spatially adjusted, mixed effects homicide count for three types of neighbor-
negative binomial regression models. Estimates hoods differentiated by overall level of pov-
from these multilevel models indicate that job- erty. Across all poverty levels, joblessness and
lessness is positively related to both violent homicide incidents are positively related. The
crime outcomes (p < .01). Poverty is also con- effect of joblessness is similar in magnitude
sistently associated with fatal and nonfatal vio- across models. Conversely, poverty has less
lent crimes (p < .01). The impacts of joblessness consistent effects on homicide among the three
on homicide rates are stronger than those of neighborhood types. Within low poverty areas,
poverty (z-score comparison of coefficients for an increase in poverty rates is associated with
joblessness and poverty = 7.90). Likewise, significant increases in homicides. This rela-
joblessness is more strongly related to robbery, tionship also appears in moderate poverty
assault, and rape incidents than poverty itself areas, but this effect of poverty is less strong
(z-score comparison of coefficients for jobless- compared with the impact estimated for low
ness and poverty = 9.31). poverty areas (z-score comparison of coeffi-
Sociodemographic conditions related to cients for poverty rate = −4.13). The associa-
population characteristics, family, and neigh- tion between poverty and homicide is likewise
borhood control generally show expected asso- weaker in high poverty areas.
ciation with violent crime. Having younger, Table 4 presents spatially adjusted, multilevel
more male, racially diverse populations and negative binomial regression estimates of the
high divorce rates at the tract level is associated effect of joblessness, poverty, and other controls
with increases in both types of violent crime on nonfatal violent crime by poverty level. Our
incidents. The effects of residential instability estimates again show joblessness is a key factor
are more mixed across outcomes, though non- associated with violent crime for all neighbor-
fatal violent crime is positively related to hoods. For all neighborhoods, joblessness has a
Table 1.  Descriptive Statistics of Variables by Level of Poverty in Tracts.

All areas Low poverty areas Moderate poverty areas High poverty areas

Standard Standard Standard Standard


Variable Mean deviation Mean deviation Mean deviation Mean deviation
Tract level
  Number of Homicides 1.51 2.46 0.73 1.39 2.62 3.10 3.16 3.42
  Number of Violent Crimes 116.07 116.39 72.82 75.47 179.71 126.15 200.18 165.02
  Joblessness Index 0.00 1.00 −0.59 0.63 0.73 0.63 1.62 0.74
  Poverty (%) 18.61 13.78 9.51 5.29 28.56 5.53 49.33 8.46
  Age 15–29 Population (%) 23.30 8.23 21.55 6.88 25.74 7.85 27.18 13.61
  Sex Ratio 1.02 0.25 1.02 0.23 1.04 0.25 0.93 0.35
  Residential Instability 0.02 0.87 −0.21 0.84 0.30 0.77 0.74 0.71
  Divorced Males (%) 9.80 4.33 9.57 3.79 10.17 4.82 10.07 5.84
  Nonwhite Population (%) 33.05 30.76 22.74 22.99 46.33 33.64 60.35 34.24
  Immigration Index 44.56 48.05 33.39 31.52 63.95 60.81 55.16 66.93
  Enrolled in College (%) 9.06 7.89 9.01 4.93 8.86 8.99 10.22 16.64
  Age 15–24 Population Married (%) 8.27 7.01 8.48 7.01 8.36 6.97 6.43 6.85
N 6,406 3,914 1,972 520  
Metropolitan Area Level
  Area (log square meters) 13.98 0.92 13.98 0.92 13.99 0.93 14.14 0.83
  Population (log) 22.55 0.91 22.55 0.91 22.57 0.93 22.70 0.85
  White/Black Dissimilarity Index 58.21 12.86 58.21 12.86 58.38 12.98 59.64 12.85
  White/Hispanic Dissimilarity Index 44.57 11.20 44.57 11.20 44.59 11.16 45.09 11.24
  Gini Index 0.45 0.04 0.45 0.04 0.45 0.04 0.44 0.02
  Suburban Employment Ratio 0.71 0.12 0.71 0.12 0.71 0.13 0.72 0.12
  Industrial Diversity Index 8.90 3.25 8.90 3.25 8.88 3.28 8.61 2.96
  Employment Change 25.15 16.37 25.15 16.37 25.15 16.60 25.19 17.35
 Region
  South 30.19 30.19 29.41 31.81  
  West 26.42 26.41 25.49 27.27  
  North 15.09 15.09 15.86 15.90  
N 53 53 51 44  

351
Note. Violent crime includes aggravated assault, forcible rape, and robbery. Poverty levels are classified at the tract level. Low poverty area = less than 20 percent impoverished.
Moderate poverty area = 20 to 39 percent impoverished. High poverty area = 40 percent or more impoverished.
Table 2.  Multilevel Negative Binomial Regression Estimates of Homicide and Violent Crime.

352
Homicide Violent crime

95% confidence 95% confidence


Variable B Standard error interval β Standard error interval
Tract level
  Joblessness Index 0.306** 0.036 [0.235, 0.377] 0.189** 0.018 [0.153, 0.224]
  Poverty (%) 0.021** 0.002 [0.016, 0.025] 0.021** 0.001 [0.018, 0.023]
  Age 15–29 Population (%) 0.026** 0.005 [0.017, 0.035] 0.013** 0.002 [0.009, 0.017]
  Sex Ratio 0.402** 0.066 [0.273, 0.531] 0.770** 0.045 [0.681, 0.860]
  Residential Instability −0.104** 0.030 [−0.163,–0.046] 0.119** 0.014 [0.092, 0.146]
  Divorced Males (%) 0.045** 0.004 [0.036, 0.054] 0.047** 0.002 [0.043, 0.051]
  Nonwhite Population (%) 0.014** 0.001 [0.013, 0.016] 0.009** 0.0004 [0.008, 0.010]
  Immigration Index 0.003** 0.001 [0.002, 0.004] 0.001** 0.0003 [0.000, 0.001]
  Enrolled in College (%) −0.027** 0.004 [−0.034,–0.019] −0.016** 0.002 [−0.020,–0.013]
  Age 15–24 Population Married (%) 0.002 0.003 [−0.004, 0.008] −0.002 0.001 [−0.005, 0.000]
  Spatial Weight −0.098 0.097 [−0.288, 0.093] −0.0004 0.002 [−0.004, 0.004]
Metropolitan Area Level
  Area (log square meters) 0.001 0.100 [−0.196, 0.197] 0.173 0.117 [−0.056, 0.402]
  Population (log) 0.321** 0.085 [0.154, 0.487] 0.196* 0.090 [0.019, 0.372]
  White/Black Dissimilarity Index 0.017* 0.007 [0.002, 0.031] 0.005 0.008 [−0.010, 0.020]
  White/Hispanic Dissimilarity Index 0.001 0.005 [−0.009, 0.011] 0.001 0.006 [−0.010, 0.013]
  Gini Index −4.977** 1.877 [−8.657,–1.298] −0.716 1.658 [−3.965, 2.534]
  Suburban Employment Ratio −1.393* 0.640 [−2.648,–0.138] −1.667* 0.744 [−3.124,–0.209]
  Industrial Diversity Index 0.014 0.016 [−0.018, 0.047] −0.023 0.019 [−0.060, 0.015]
  Employment Change −0.009* 0.004 [−0.017,–0.001] −0.007 0.005 [−0.016, 0.002]
 Region
  South −0.084 0.157 [−0.391, 0.224] −0.047 0.170 [−0.381, 0.287]
  West 0.020 0.207 [−0.385, 0.426] −0.164 0.225 [−0.604, 0.277]
  North −0.128 0.190 [−0.500, 0.244] 0.082 0.203 [−0.315, 0.479]
Overdispersion −0.799** 0.052 −1.183** 0.019  
Variance Components 0.070** 0.023 0.113** 0.041  
Tract-level N 6,406 6,406  
Metropolitan-area-level N 53 53  

Note. Violent crime includes aggravated assault, forcible rape, and robbery.
*p value = .05, **p value = .01.
Table 3.  Multilevel Negative Binomial Regression Estimates of Homicide by Poverty Level.

Low poverty area Moderate poverty area High poverty area

Standard 95% confidence Standard 95% confidence Standard 95% confidence


Variable B error interval β error interval β error interval
Tract Level  
  Joblessness Index 0.288** 0.075 [0.140, 0.435] 0.271** 0.053 [0.167, 0.375] 0.257** 0.073 [0.114, 0.400]
  Poverty (%) 0.057** 0.008 [0.041, 0.072] 0.018** 0.005 [0.009, 0.028] 0.012 0.006 [−0.001, 0.024]
  Age 15–29 Population (%) 0.017* 0.008 [0.002, 0.033] 0.012 0.008 [−0.003, 0.027] 0.00003 0.011 [−0.022, 0.022]
  Sex Ratio 0.401** 0.095 [0.215, 0.586] 0.284** 0.108 [0.072, 0.496] 0.585** 0.148 [0.295, 0.874]
  Residential Instability −0.078 0.056 [−0.189, 0.032] −0.077 0.043 [−0.161, 0.007] −0.238** 0.083 [−0.401,–0.075]
  Divorced Males (%) 0.056** 0.008 [0.040,0.073] 0.018** 0.007 [0.005, 0.031] 0.012 0.01 [−0.007, 0.030]
  Nonwhite Population (%) 0.014** 0.001 [0.012, 0.017] 0.009** 0.001 [0.006, 0.011] 0.008** 0.003 [0.002, 0.014]
  Immigration Index −0.0005 0.001 [−0.003,0.002] 0.0004 0.001 [−0.001, 0.002] −0.002 0.002 [−0.006, 0.002]
  Enrolled in College (%) −0.023** 0.008 [−0.040,–0.007] −0.028** 0.006 [−0.040,–0.016] −0.01 0.009 [−0.026, 0.007]
  Age 15–24 Population Married (%) 0.001 0.005 [−0.008, 0.010] −0.002 0.005 [−0.011, 0.007] 0.002 0.008 [−0.014, 0.018]
  Spatial Weight 0.006 0.088 [−0.167, 0.178] 0.0003 0.084 [−0.164, 0.165] 0.234* 0.101 [0.035, 0.432]
Metropolitan Area Level
  Area (log square meters) 0.115 0.108 [−0.096, 0.327] −0.004 0.109 [−0.218, 0.209] −0.036 0.165 [−0.359, 0.287]
  Population (log) 0.210* 0.088 [0.038, 0.382] 0.302** 0.098 [0.109, 0.494] 0.201 0.158 [−0.108, 0.510]
  White/Black Dissimilarity Index 0.005 0.007 [−0.009, 0.019] 0.016 0.009 [−0.001, 0.034] −0.009 0.014 [−0.037, 0.019]
  White/Hispanic Dissimilarity Index 0.004 0.005 [−0.006, 0.015] −0.001 0.005 [−0.012, 0.009] −0.001 0.007 [−0.014, 0.013]
  Gini Index −1.890 2.016 [−5.842, 2.062] −7.110** 2.52 [−12.050,–2.171] −1.951 4.13 [−10.046, 6.144]
  Suburban Employment Ratio −1.1996 0.695 [−2.562, 0.163] −1.390* 0.681 [−2.725,–0.054] 1.446 1.117 [−0.743, 3.636]
  Industrial Diversity Index 0.012 0.016 [−0.020, 0.044] 0.018 0.017 [−0.016, 0.052] −0.005 0.023 [−0.051, 0.040]
  Employment Change −0.003 0.004 [−0.011, 0.004] −0.008 0.004 [−0.016, 0.000] −0.0002 0.006 [−0.011, 0.011]
 Region
  South −0.062 0.167 [−0.389, 0.265] −0.065 0.166 [−0.391, 0.260] 0.183 0.256 [−0.319, 0.685]
  West −0.057 0.211 [−0.470, 0.357] 0.086 0.22 [−0.346, 0.518] 0.261 0.345 [−0.415, 0.936]
  North −0.297 0.215 [−0.719, 0.125] −0.022 0.183 [−0.380, 0.337] −0.158 0.257 [−0.662, 0.346]
Overdispersion −0.497** 0.088 −1.038** 0.076 −1.137** 0.147  
Variance Components 0.046* 0.021 0.051** 0.020 0.032 0.025  
Tract-level N 3,914 1,972 520  
Metropolitan-area-level N 53 51 44  

Note. Low poverty area = less than 20 percent impoverished. Moderate poverty area = 20 to 39 percent impoverished. High poverty area = 40 percent or more impoverished.

353
*p value = .05. **p value = .01.
Table 4.  Multilevel Negative Binomial Regression Estimates of Violent Crime by Poverty Level.

354
Low poverty area Moderate poverty area High poverty area

Standard 95% confidence Standard 95% confidence Standard 95% confidence


Variable B error interval β error interval β error interval
Tract Level
  Joblessness Index 0.231** 0.029 [0.173, 0.288] 0.099** 0.025 [0.050, 0.149] 0.156** 0.042 [0.074, 0.237]
  Poverty (%) 0.048** 0.003 [0.042, 0.054] 0.026** 0.002 [0.021, 0.030] 0.002 0.004 [−0.005, 0.009]
  Age 15–29 Population (%) 0.012** 0.003 [0.006, 0.018] −0.015** 0.003 [−0.022,–0.008] −0.002 0.006 [−0.014, 0.010]
  Sex Ratio 0.538** 0.080 [0.381, 0.696] 0.725** 0.058 [0.612, 0.839] 0.578** 0.089 [0.403, 0.752]
  Residential Instability 0.105** 0.020 [0.065, 0.144] 0.131** 0.021 [0.090, 0.172] 0.148** 0.047 [0.055, 0.240]
  Divorced Males (%) 0.047** 0.003 [0.041, 0.053] 0.017** 0.003 [0.011, 0.023] 0.018** 0.005 [0.008, 0.028]
  Nonwhite Population (%) 0.008** 0.001 [0.007, 0.009] 0.004** 0.001 [0.003, 0.006] 0.001 0.002 [−0.002, 0.004]
  Immigration Index −0.001* 0.001 [−0.002,–0.000] −0.001 0.000 [−0.002, 0.000] −0.004** 0.001 [−0.006,–0.002]
  Enrolled in College (%) −0.015** 0.003 [−0.021,–0.009] −0.004 0.003 [−0.009, 0.001] −0.016** 0.005 [−0.026,–0.007]
  Age 15–24 Population Married (%) −0.002 0.002 [−0.005, 0.002] −0.007** 0.002 [−0.011,–0.003] −0.001 0.005 [−0.010, 0.008]
  Spatial Weight 0.003** 0.001 [0.001, 0.005] 0.004** 0.001 [0.002, 0.006] 0.004** 0.001 [0.003, 0.006]
Metropolitan Area Level
  Area (log square meters) 0.162 0.090 [−0.014, 0.337] 0.009 0.086 [−0.159, 0.177] 0.062 0.093 [−0.121, 0.245]
  Population (log) 0.116 0.072 [−0.025, 0.258] 0.158* 0.069 [0.022, 0.294] 0.082 0.088 [−0.090, 0.255]
  White/Black Dissimilarity Index 0.002 0.006 [−0.010, 0.014] −0.0003 0.006 [−0.012, 0.011] −0.0004 0.008 [−0.016, 0.015]
  White/Hispanic Dissimilarity Index −0.001 0.005 [−0.009, 0.008] 0.0003 0.004 [−0.008, 0.009] 0.002 0.004 [−0.006, 0.009]
  Gini Index −0.677 1.337 [−3.297, 1.943] −0.677 1.427 [−3.474, 2.121] 0.424 2.290 [−4.064, 4.913]
  Suburban Employment Ratio −1.319* 0.575 [−2.445,–0.192] −0.556 0.530 [−1.594, 0.483] −0.204 0.635 [−1.448, 1.040]
  Industrial Diversity Index −0.007 0.015 [−0.036, 0.022] −0.009 0.013 [−0.035, 0.017] 0.0004 0.014 [−0.026, 0.027]
  Employment Change −0.004 0.003 [−0.011, 0.003] −0.00004 0.003 [−0.006, 0.006] 0.003 0.003 [−0.004, 0.009]
 Region
  South −0.034 0.137 [−0.301, 0.234] −0.047 0.126 [−0.294, 0.201] −0.092 0.146 [−0.378, 0.193]
  West −0.126 0.179 [−0.478, 0.225] −0.106 0.168 [−0.434, 0.223] −0.005 0.193 [−0.383, 0.374]
  North −0.037 0.165 [−0.360, 0.285] 0.121 0.142 [−0.157, 0.400] −0.071 0.146 [−0.357, 0.214]
Overdispersion −1.107** 0.024 −1.588** 0.033 −1.450** 0.065  
Variance Components 0.067** 0.017 0.049** 0.014 0.013 0.009  
Tract-level N 3,914 1,972 520  
Metropolitan-area-level N 53 51 44  

Note. Violent crime includes aggravated assault, forcible rape, and robbery. Low poverty area = less than 20 percent impoverished. Moderate poverty area = 20 to 39 percent
impoverished. High poverty area = 40 percent or more impoverished.
*p value = .05. **p value = .01.
Dollar et al. 355

strong, independent association with nonfatal We further estimated the impacts on job-
violent offenses. This connection is greater than lessness on crime across neighborhoods with
poverty alone across neighborhoods. The data varying degrees of poverty. Our results indi-
indicate that poverty does not elevate violent cate that joblessness consistently and posi-
crimes after taking into account joblessness and tively contributes to more homicides and
other conditions in the most severely impover- violent crime in communities. Poverty rates
ished areas. Joblessness plays a similar role in are less influential than joblessness in shaping
shaping violent crime in high poverty areas rela- violent crime in all neighborhoods. Moreover,
tive to moderate poverty areas (z-score compari- poverty alone does not significantly change
son of coefficients for joblessness = 1.16) and violent crime incidents in high poverty areas
relative to low poverty areas (−1.47 z-score after adjusting for joblessness and other condi-
comparison of coefficients for joblessness). tions. These results then provide evidence that
joblessness primarily drives differences in
homicide and violent crime outcomes in highly
Discussion and Conclusion
impoverished neighborhoods, relative to less
Wilson’s (1996, 1999) scholarship identifies a impoverished poverty communities.
“new urban poverty” associated with chang- Notwithstanding its contribution, the current
ing, enduring joblessness. What made this new examination has some limitations. First, our
poverty distinct from earlier conditions of pov- findings depend on modeling assumptions,
erty in urban communities was the distinctive such as the proper measurement of variables
influence of joblessness, or the presence of and correct specifications of regression models.
jobless poverty. Wilson (1996), in fact, argued Spatial dependence may be addressed through
that while poverty had long been a damaging other statistical strategies, such as the correc-
feature of many inner cities, growing levels of tion of correlational structures, and yield differ-
joblessness resulted in an isolating existence ent results from incorporating spatial lags as
and led to various social disorders, including control variables into GLMMs (Lambert et al.
violent crime. 2010). Our measure of joblessness focuses on
This study offers the first empirical investi- employment and labor force attachment and
gation of how joblessness influences homicide does not capture all aspects of jobless poverty.
and nonfatal violent incidents in U.S. neighbor- For instance, Wilson (1987, 1996) argues that
hoods, separately from poverty and other envi- the disappearance of work intensifies social
ronmental conditions. The data indicate several isolation, resulting in a shift of cultural values
noteworthy findings. First, consistent with whereby poor, jobless communities replace
other studies on neighborhood crime, we find dominant norms of achieving success with
nonrandom distributions of violent offenses, “subcultural” norms that pursue alternative
thus indicating crime has spillover effects on forms of status and respect. Cultural responses
neighboring communities. Second, and perhaps to these joblessness conditions shape neighbor-
more importantly, we find support for Wilson’s hood organization and may increase criminal
(1996, 1999, 2003) hypotheses that joblessness involvement (e.g., Anderson 1999; Stewart and
has a strong relationship with crime relative to Simons 2006; Young 2004). Our estimated
poverty alone. Employing spatially adjusted, effects of joblessness on crime may then be
multilevel regression analyses, we find jobless- biased and likely underestimated. Likewise,
ness has an independent effect on homicide and our research may understate the contributions
violent crime after taking into account varying of suburbanization processes to social prob-
levels of poverty, relevant community char- lems in urban neighborhoods. The confluence
acteristics, and metropolitan-area-level predic- of land use and housing regulations, new high-
tors. Importantly, joblessness has a greater ways and infrastructure, and exodus of the mid-
impact on crime than poverty alone. These dle class from urban centers created
results empirically support Wilson’s character- unprecedented development in the suburbs
ization of joblessness as a “new poverty” that (Squires and Kubrin 2005). An empirical study
contributes to crime. by Browning and colleagues (2010)
356 Social Currents 6(4)

underscores that commercial and residential of financial stability but also as a crucial inte-
density affects violent crime outcomes in com- grating structure given its ability to provide
munities around Columbus, Ohio. Land use individuals with consistent routines, personal
regulations vary across urban centers, however, and social discipline, and self-efficacy. As
and our data lack information on commercial such, high areal levels of poverty and a lack of
and residential density to further test these land labor market engagement opportunities influ-
use claims. Our analysis also may not capture ence quality of life from a structural, cultural,
other restructurings of the American economy, and psychological standpoint. Wilson (1999)
such as the relocation of businesses from cities posits that solutions to the poverty-joblessness
to lower-cost areas in the United States or the nexus on crime must involve increasing liv-
outsourcing of operations to other countries able-wage jobs, especially among low-skilled,
(Bluestone and Harrison 2001). inner-city residents. In other words, crimino-
Our study points to several opportunities genic cultural and psychological maladapta-
for additional research. First, our results derive tions stem from ecological conditions. The
from a snapshot of neighborhood and metro- present analysis underscores that efforts to pri-
politan area conditions in the 2000s. The oritize structural resolutions must consider the
impacts of joblessness may vary in more recent unique effects that poverty and joblessness
times when crime rates have decreased. Future may have as each appears to distinctively
research may also explore how the availability influence violence.
of jobs affects different types of crime, as vio-
lent crime may be motivated by expressive Declaration of Conflicting Interests
concerns (e.g., Logan and Messner 1987), The author(s) declared no potential conflicts of
while property crime may be driven by more interest with respect to the research, authorship,
instrumental reasons (e.g., Agnew 2001). In and/or publication of this article.
addition, the effects of poverty and joblessness
on crime may vary by race. Although Wilson Funding
argues that deindustrialization disproportion- The author(s) received no financial support for the
ately affected blacks by eradicating upwardly research, authorship, and/or publication of this
mobile opportunities and life chances, he article.
maintained that poverty, joblessness, and its
relation to crime should not be conceptualized Notes
or measured as a race-specific issue (e.g., 1. Tracts with nonmissing data differ from those
Wilson 1987, 1996, 1999). However, subse- with missing data. For instance, our analytic
quent research on poverty and joblessness sample has higher homicide rates than the
might further disaggregate criminal outcomes full sample; however, we find similar counts
by race and ethnicity to identify and under- of other violent crime. Nonmissing informa-
stand any differences across racial and ethnic tion tracts also have larger, more divorced, and
populations. racially diverse populations. Missing informa-
The present article finds that area-level tion tracts have less college enrollment. Both
types of tracts have similar sex ratios, levels
poverty and joblessness are consistently yet
of residential instability, and marriage rates
uniquely related to neighborhood crime. Over among younger populations. At the metro-
half a century ago, a “war” on poverty was politan area level, there are no statistically
declared. Data from the U.S. Census Bureau significant differences in the population, size,
indicate that since that time, the national pov- segregation, or labor force characteristics
erty rate has declined by only 5 percent. What of the full (i.e., 62 metropolitans areas) and
is more, well over half of the persons currently analytic sample (i.e., 53 metropolitan areas).
living in poverty fall between the ages of 18 to Missing metropolitan areas are less likely to be
64, which are crucial working ages (Desilver in the North or West, and more likely to be in
2014). As discussed earlier, Wilson views the the South, however. As a result, the findings
formal labor market not only as a viable means presented herein may better generalize to tracts
Dollar et al. 357

with more severe problems of homicide and Bailey, William C. 1984. “Poverty, Inequality and
disadvantage. City Homicide Rates: Some Not So Unexpected
2. These six categories include health care sup- Findings.” Criminology 22(4):531–50.
port; food preparation and serving-related Baller, Robert D., Luc Anselin, Stefen F. Messner,
occupations; building and grounds cleaning Glenn Deane, and Darnell F. Hawkins.
and maintenance; personal care and service; 2001. “Structural Covariates of U.S. County
farming, fishing, and forestry; and material Homicide Rates: Incorporating Spatial Effects.”
moving (see Peterson and Krivo 2000). Criminology 39:561–90.
3. We exclude youth employment and participa- Bivand, Roger. 2018. “Creating Neighbours.”
tion in the armed forces as forms of institutional (https://cran.r-project.org/web/packages/spdep/
engagement to avoid repetitive measurement vignettes/nb.pdf).
of labor force activity of this population. Bivand, Roger, Edzer J. Pebesma, and Virgilio
4. As a robustness check, we ran a set of multi- Gómez-Rubio. 2008. Applied Spatial Data
level models with cities as our Level 2 unit of Analysis with R. New York: Springer Science
analysis. These models can account for racial and Business Media.
and socioeconomic inequalities within cit- Blau, Judith and Peter Blau. 1982. “Metropolitan
ies, but they cannot control for employment Structure and Violent Crime.” American
change or the suburban employment ratio as Sociological Review 47:114–28.
metropolitan area conditions. Our substantive Bluestone, Barry and Bennett Harrison. 2001.
findings do not change depending on whether Growing Prosperity: The Battle for Growth
we use cities or metropolitan areas as a higher with Equity in the Twenty-first Century.
level of aggregation. Still, we prefer using met- Berkeley, CA: University of California Press.
ropolitan areas as our Level 2 unit of analysis Bourgois, Philippe. 1995. In Search of Respect:
because we can account for industry character- Selling Crack in El Barrio. Cambridge, MA:
istics outside of the city itself and, thus, take Cambridge University Press.
into account Wilson’s observations on the exo- Browning, Christopher R., Reginald A. Byron,
dus of jobs to suburban areas. Catherine A. Calder, Lauren J. Krivo, Mei-Po
5. Approximately 48 percent of all census tracts Kwan, Jae-Yong Lee, and Ruth D. Peterson. 2010.
had no reported homicides. Less than 1 percent “Commercial Density, Residential Concentration,
(0.31 percent) of census tract had zero-reported and Crime: Land Use Patterns and Violence in
aggravated assault, forcible rape, or robbery Neighborhood Context.” Journal of Research in
incidents. Crime and Delinquency 47(3):329–57.
Chiricos, Theodore G. 1987. “Rates of Crime and
Unemployment: An Analysis of Aggregate Research
References Evidence.” Social Problems 34(2):187–212.
Agnew, Robert. 2001. “Building on the Foundation Crutchfield, Robert. 1989. “Labor Stratification and
of General Strain Theory: Specifying the Types Violent Crime.” Social Forces 68(2):489–512.
of Strain Most Likely to Lead to Crime and Crutchfield, Robert. 2014. Get a Job: Labor
Delinquency.” Journal of Research in Crime Markets, Economic Opportunity and Crime.
and Delinquency 38(4):319–61. New York: New York University Press.
Anderson, Elijah. 1989. “Sex Codes and Family Crutchfield, Robert, Ross Matsueda, and Kevin
Life among Poor Inner-city Youths.” The Drakulich. 2006. “Race, Labor Markets, and
ANNALS of the American Academy of Political Neighborhood Violence.” Pp. 199–220 in The
and Social Science 501(1):59–78. Many Colors of Crime: Inequalities of Race,
Anderson, Elijah. 1999. Code of the Street: Decency, Ethnicity, and Crime in America, edited by R.
Violence, and the Moral Life of the Inner City. Peterson, L. Krivo, and J. Hagan. New York:
New York: W.W. Norton. New York University Press.
Anselin, Luc, Jacqueline Cohen, David Cook, Crutchfield, Robert and Susan Pitchford. 1997.
Wilpen Gorr, and George Tita. 2000. “Spatial “Work and Crime: The Effects of Labor
Analyses of Crime.” Pp. 213–62 in Criminal Stratification.” Social Forces 76(1):93–118.
Justice: Vol. 4: Measurement and Analysis Deaton, Angus. 2013. The Great Escape: Health,
of Crime and Justice, edited by D. Duffee. Wealth, and the Origins of Inequality. Princeton,
Washington, DC: National Institute of Justice. NJ: Princeton University Press.
358 Social Currents 6(4)

Desilver, Drew. 2014. “Who’s Poor in America? the American Academy of Political and Social
50 Years into the ‘War on Poverty,’ a Data Science 501:26–47.
Portrait.” Fact Tank, January 13. Retrieved Kasarda, John D. 1992. “The Severely Distressed
February 25, 2019 (http://www.pewresearch. in Economically Transforming Cities.” Pp.
org/fact-tank/2014/01/13/whos-poor-in-amer- 45–97 in Drugs, Crime, and Social Isolation:
ica-50-years-into-the-war-on-poverty-a-data- Barriers to Urban Opportunity, edited by A. V.
portrait/). Harrell and G. E. Peterson. Washington, DC:
DiStefano, Christine, Min Zhu, and Diana Mindrila. The Urban Institute.
2009. “Understanding and Using Factor Scores: Kopczuk, Wojciech., Emmanuel Saez, and Jae
Considerations for the Applied Researcher.” Song. 2010. “Earnings Inequality and Mobility
Practical Assessment, Research & Evaluation in the United States: Evidence from Social
14(20):1–11. Security Data since 1937.” Quarterly Journal
Dollar, Cindy B., Patricia L. McCall, Kenneth C. of Economics 125(1):91–128.
Land, and Joshua Fink. 2017. “Age Structure Kovandzic, Tomislav V., Lynne M. Vieraitis,
and Neighborhood Homicide: A Test and and Mark R. Yeisley. 1998. “The Structural
Elaboration of the Differential Institutional Covariates of Urban Homicide: Reassessing the
Engagement Hypothesis.” Homicide Studies Impact of Income Inequality and Poverty in the
21(4):243–66. Post-Reagan Era.” Criminology 36(3):569–600.
Glick, P. C., J. C. Beresford, and D. M. Heer. 1963. Krivo, Lauren J. and Ruth D. Peterson. 1996.
“Family Formation and Family Composition: “Extremely Disadvantaged Neighborhoods and
Trends and Prospects.” Pp 30–39 in Sourcebook Urban Crime.” Social Forces 75(2):619–48.
in Marriage and the Family, edited by M. B. Krivo, Lauren J. and Ruth D. Peterson. 2000. “The
Sussman. Boston, MA: Houghton Mifflin. Structural Context of Homicide: Accounting
Harrington, Michael and Mark Levinson. 1985. “The for Racial Differences in Process.” American
Perils of a Dual Economy.” Dissent 32:417–26. Sociological Review 65:547–59.
Hipp, John R. 2011. “Spreading the wealth: The effect Lambert, Dayton M., Jason P. Brown, and Raymond
of the distribution of income and race/ethnicity J. G. M. Florax. 2010. “A Two-step Estimator for
across households and neighborhoods on city a Spatial Lag Model of Counts: Theory, Small
crime trajectories.” Criminology 49: 631–665. Sample Performance and an Application.” Regional
Hipp, John and Daniel Yates. 2011. “Ghettos, Science & Urban Economics 40(4):241–52.
Thresholds, and Crime: Does Concentrated Land, Kenneth C., Patricia L. McCall, and Lawrence
Poverty Really Have an Accelerating Increasing E. Cohen. 1990. “Structural Covariates of
Effect on Crime?” Criminology 49(4):955–90. Homicide Rates: Are There Any Invariances
Iceland, John. 2004. “More Pay, More Inequality? across Time and Social Space?” American
The Influence of Average Wage Levels and Journal of Sociology 95:922–63.
the Racial Composition of Jobs on the Black- Levitt, Steven D. 2001. “Alternative Strategies for
White Wage Gap.” Social Science Review Identifying the Link between Unemployment
33:498–520. and Crime.” Journal of Quantitative
Iceland, John and Daniel H. Weinberg. 2002. Criminology 17(4):377–90.
Racial and Ethnic Residential Segregation in Logan, John and Steven Messner. 1987. “Racial
the United States 1980-2000. Washington, DC: Residential Segregation and Suburban Violent
U.S. Census Bureau. Crime.” Social Science Quarterly 68(3):510–27.
Ismail, Noriszura and Abdul A. Jemain. 2007. Ludwig, Jens, Jeffrey B. Liebman, Jeffrey R. Kling,
“Handling Overdispersion with Negative Greg J. Duncan, Lawrence F. Katz, Ronald C.
Binomial and Generalized Poisson Regression Kessler, and Lisa Sanbonmatsu. 2008. “What
Models.” Casualty Actuarial Society Forum can we learn about neighborhood effects
(7):103–58. from the moving to opportunity experiment?”
Jargowsky, Paul and Bane, Mary Jo. 1990. “Ghetto American Journal of Sociology 114: 114–188.
poverty: Basic questions.” In L.E. Lynn, Jr. Massey, Douglas S. and Nancy A. Denton. 1993.
and M.G.H. McGeary, ed. Inner-city poverty American Apartheid: Segregation and the
in the United States. Washington, DC: National Making of the Underclass. Cambridge, MA:
Academy Press. Harvard University Press.
Kasarda, John D. 1989. “Urban Industrial Massey, Douglas S., Andrew B. Gross, and Mitchell
Transition and the Underclass.” The Annuals of L. Eggers. 1991. “Segregation, the Concentration
Dollar et al. 359

of Poverty, and the Life Chances of Individuals.” Piketty, Thomas and Emmanuel Saez. 2003.
Social Science Research 20:397–420. “Income Inequality in the United States: 1913-
Mayer, Susan. 2001. “How Did the Increase in 1998.” Quarterly Journal of Economics 1:1–39.
Economic Inequality between 1970 and 1990 Pratt, Travis C., and Francis T. Cullen. 2005.
Affect Children’s Educational Attainment?” “Assessing Macro-Level Predictors and Theories
American Journal of Sociology 107(1):1–32. of Crime: A Meta-Analysis.” Crime and Justice
McCall, Leslie. 2000. “Gender and the New Crime and Justice 32: 373–450.
Inequality: Explaining the College/Non-college Raudenbush, Stephen W. and Anthony S. Bryk.
Wage Gap.” American Sociological Review 2002. Hierarchical Linear Models: Applications
65(2):234–55. and Data Analysis Methods. 2nd ed. Thousand
McCall, Patricia L., Kenneth C. Land, Cindy B. Oaks, CA: Sage.
Dollar, and Karen F. Parker. 2013. “The Age Ricketts, Erol R. and Isabel V. Sawhill. 1988. “Defining
Structure-crime Rate Relationship: Solving a and Measuring the Underclass.” Journal of Policy
Long-standing Puzzle.” Journal of Quantitative Analysis and Management 7(2):316–25.
Criminology 29(2):167–90. Sampson, Robert J. 1987. “Urban Black Violence:
McCall, Patricia L., Karen F. Parker, and John M. The Effect of Male Joblessness and Family
Macdonald. 2008. “The Dynamic Relationship Disruption.” American Journal of Sociology
between Homicide Rates and Social, Economic, 93:348–82.
and Political Factors from 1970 to 2000.” Sampson, Robert J. 2008. “Moving to Inequality:
Social Science Research 37(3):721–35. Neighborhood Effects and Experiments
McLanahan, Sare, Irwin Garfinkel, and Dorothy Meet Social Structure.” American Journal of
Watson. 1988. “Family Structure, Poverty, and Sociology 114:189–231.
the Underclass.” Pp. 102–47 in Urban Change Sampson, Robert J. and Corina Graif. 2009.
and Poverty, edited by M. G. H. McGeary and “Neighborhood Social Capital as Differential
L. E. Lynn, Jr. Washington, DC: The National Social Organization: Resident and Leadership
Academies Press. Dimensions.” American Behavioral Scientist
Messner, Steven F. 1982. “Poverty, Inequality, and 52(11):1579–605.
the Urban Homicide Rate: Some Unexpected Sampson, Robert J. and W. Byron Groves. 1989.
Findings.” Criminology 20(1):103–14. “Community Structure and Crime: Testing
Messner, Steven F. and Robert J. Sampson. 1991. Social-disorganization Theory.” American
“Sex Ratio, Family Disruption, and Rates of Journal of Sociology 94:774–802.
Violent Crime: The Paradox of Demographic Sampson, Robert J. and William J. Wilson. 1995.
Structure.” Social Forces 69(3):693–713. “Toward a Theory of Race, Crime, and Urban
Messner, Steven F. and Kenneth Tardiff. 1986. Inequality.” Pp. 37–54 in Crime and Inequality,
“Economic Inequality and Levels of Homicide: edited by H. J. Peterson. Stanford, CA: Stanford
An Analysis of Urban Neighborhoods.” University Press.
Criminology 24(2):297–316. Schoen, Robert. 1981. “The Harmonic Mean as the
Morris, Martina and Bruce Western. 1999. Basis of a Realistic Two-sex Marriage Model.”
“Inequality in Earnings at the Close of the Demography 18:201–16.
Twentieth Century.” Annual Review of Shaw, Clifford R. and Henry D. McKay. 1942.
Sociology 25:623–57. Juvenile Delinquency and Urban Areas: A Study
Parker, Karen. 2004. “Polarized Labor Markets, of Rates of Delinquents in Relation to Differential
Industrial Restructuring and Urban Violence: A Characteristics of Local Communities in
Dynamic Model of the Economic Transformation American Cities. Chicago, IL: The University of
and Urban Violence.” Criminology 42(2):619–45. Chicago Press.
Parker, Karen. 2008. Unequal Crime Decline: Squires, Gregory D. and Charis E. Kubrin. 2005.
Theorizing Race, Urban Inequality, and Criminal “Privileged Places: Race, Uneven Development
Violence. New York: New York University Press. and the Geography of Opportunity in Urban
Patterson, E. B. 1991. “Poverty, Income Inequality, America.” Urban Studies 42(1):47–68.
and Community Crime Rates.” Criminology Stewart, Eri and Ronald Simons. 2006. “Structure
29:755–76. and Culture in African-American Adolescent
Pattillo, Mary. 1999. Black Picket Fences. Chicago, Violence: A Partial Test of the Code of the
IL: The University of Chicago Press. Street Thesis.” Justice Quarterly 23:1–33.
360 Social Currents 6(4)

Tienda, Marta. 1989. “Race, Ethnicity and the and Public Policy. Chicago, IL: The University
Portrait of Inequality: Approaching the 1990s.” of Chicago Press.
Sociological Spectrum 9(1):23–52. Wilson, William Julius. 1991. “Studying Inner-City
Wacquant, Loïc J. D. and William J. Wilson. 1989. Social Dislocations: The Challenge of Public
“The Cost of Racial and Class Exclusion in Agenda Research.” American Sociological
the Inner City.” The ANNALS of the American Association 56(1): 1–14.
Academy of Political and Social Science Wilson, William Julius. 1996. When Work
501(1):8–25. Disappears: The World of the New Urban
Waldinger, Roger D. 1996. Still the Promised City? Poor. New York: Alfred A. Knopf.
African Americans and New Immigrants in Wilson, William Julius. 1999. “Jobless Poverty:
Postindustrial New York. Boston, MA: Harvard A New Form of Social Dislocation in the
University Press. Inner-city Ghetto.” Pp. 142–152 in A Nation
Warner, Barbara D. and Pamela W. Rountree. 1997. Divided: Diversity, Inequality, and Community
“Local Social Ties in a Community and Crime in American Society, edited by P. Moen, D.
Model: Questioning the Systemic Nature of Dempster-McClain, and H. A. Walker. New
Informal Social Control.” Social Problems York: Cornell University Press.
44(4):520–36. Wilson, William Julius. 2003. “Race, Class and
West, Brady T., Kathleen B. Welch, and Andrzej Urban Poverty: A Rejoinder.” Ethnic and
T. Galecki. 2014. Linear Mixed Models: A Racial Studies 26(6):1096–114.
Practical Guide Using Statistical Software. 2nd Wilson, William Julius. 2009. More than Just Race:
ed. Boca Raton, FL: CRC Press. Being Black and Poor in the Inner City. New
Wilson, William Julius. 1978. The Declining York: W.W. Norton.
Significance of Race: Blacks and Changing Young, Alford A. Jr. 2004. The Minds of
American Institutions. Chicago, IL: The University Marginalized Black Men: Making Sense
of Chicago Press. of Mobility, Opportunity, and Future Life
Wilson, William Julius. 1987. The Truly Chances. Princeton, NJ: Princeton University
Disadvantaged: The Inner City, the Underclass Press.

You might also like