Crime, Law and Social Change

, Volume 57, Issue 5, pp 493–519

Instability, informal control, and criminogenic situations: community effects of returning prisoners

Authors

    • School of Criminology and Criminal JusticeNortheastern University
  • Robert D. Crutchfield
    • Department of SociologyUniversity of Washington
  • Ross L. Matsueda
    • Department of SociologyUniversity of Washington
  • Kristin Rose
    • School of Criminology and Criminal JusticeNortheastern University
Article

DOI: 10.1007/s10611-012-9375-0

Cite this article as:
Drakulich, K.M., Crutchfield, R.D., Matsueda, R.L. et al. Crime Law Soc Change (2012) 57: 493. doi:10.1007/s10611-012-9375-0

Abstract

Incarceration, whose putative goal is the reduction of crime, may at higher concentrations actually increase crime by overwhelming neighborhoods with limited resources. The present research poses and provides initial support for an explanation of this paradoxical consequence of a crime control strategy. Specifically, we draw on two different lines of theoretical work to suggest that large numbers of returning prisoners may negatively impact a community’s economic and residential stability, limiting a community’s capacity for informal social control and resulting in labor market conditions conducive to criminal behavior. This study combines data on local social organization processes from a large survey of Seattle residents with contextual, crime, and incarceration data from the US Census, Seattle Police Department, and Washington State Department of Corrections. The results suggest that high concentrations of returning prisoners are associated with a reduced capacity for collective efficacy; the fostering of social situations conducive to criminal behavior; and higher levels of violent crime. The impact of incarceration on these neighborhood processes, however, appears to be largely indirect through the turmoil that concentrations of incarceration create in a neighborhood’s labor and housing markets. We conclude with a call for greater scrutiny of the goals and actual outcomes of incarceration policy.

The number of incarcerated persons in the United States has grown at an alarming rate over the last 30 years. The number of inmates in state and federal prisons has increased nearly eight-fold from less than 200,000 in 1970 to more than 1.5 million by 2009, far exceeding the 50 % increase in population over this same period.1 As of 2008, more than one percent of the U.S. adult population was behind bars.2

In the face of this historically and globally unprecedented use of incarceration, questions about both the goals and the actual outcomes of incarceration have taken on new urgency. If the goal of incarceration is crime control, then we must be sure of the effect of incarceration on future crimes, and in particular that this effect holds regardless of the scale of imprisonment. In other words, we must be concerned about whether potential gains from incarceration through incapacitation [16, 76] or deterrence [45] are being offset by negative consequences of imprisonment that are unique to this magnitude of its usage. Even if the goal of imprisonment is retribution, we must be concerned whether we are decreasing public safety in its pursuit.

In particular, we should be concerned about consequences stemming from the incredible concentration of this incarceration both demographically and spatially. For instance, as of midyear 2009 more than one in ten African-American males aged 25–29 was in prison (9.7 %) versus 1 in 25 Hispanic males (4.0 %) and 1 in 86 white males (1.2 %) in the same age group.3 The intersection of race and class produces even higher imprisonment risks. Comparing blacks and whites in the U.S., Pettit and Western find that “by the end of the 1990s, 21 % of young black poorly educated men were in state or federal prison compared to an imprisonment rate of 2.9 % for young white male dropouts” ([55], p. 160). Lifetime risks were even more disturbing: “a black male dropout, born 1965–1969, had nearly a 60 % chance of serving time in prison by the end of the 1990s” ([55], p. 161).

A combination of historical economic forces and residential segregation have clustered impoverished racial and ethnic minorities in specific urban neighborhoods, resulting in “concentration effects” of compounded disadvantageous conditions which may have uniquely deleterious consequences [84]. As incarceration has skyrocketed, these neighborhoods have increasingly been exposed to a new concentration effect: incarceration. Consequently, incarceration, whose putative goal is the reduction of crime, may at higher magnitudes in particular places actually increase crime by overwhelming neighborhoods with limited resources. Recent research has revealed that both releases from prison and admissions to prison—at least beyond a certain threshold—can increase community crime rates [15, 61]. Less is known about the specific mechanisms by which incarceration raises the risk of crime.

The present research poses—and provides initial support—for an updated explanation of this paradoxical consequence of a crime control strategy. Specifically, we focus on the collateral consequences of incarceration for two major realms of community life: the labor market and the social organization of the neighborhood. In doing so, we seek to integrate and expand on two lines of research on the consequences of mass incarceration. One line of research, drawing on a social disorganization mechanism, has focused on incarceration as a churning process disrupting local networks and thereby impairing local social processes relevant to informal social control [33, 64]. In this way, an overreliance on formal controls, especially incarceration, may be undermining less coercive sources of control, including families and communities.

A second line of research has sought to investigate the devastating economic consequences of incarceration for the incarcerated individual directly, for their families, and for the communities in which they live [50, 57, 81, 83]. We draw on Crutchfield’s [18] work on labor stratification to develop expectations about the potential of concentrated incarceration to increase crime through labor market mechanisms. Specifically, if concentrations of incarceration increase concentrations of unemployed and underemployed persons in specific geographic spaces, this may result in ‘situations of company’ conducive to crime—situations in which congregations of people with lowered stakes in conformity provide potential pools of both offenders and victims.

The present work seeks to integrate these perspectives by framing concentrated incarceration as the source of a more general instability that manifests in the local labor and housing markets. This instability, in turn, has consequences for the social organization of the neighborhood—in particular limiting a community’s capacity for informal social control and resulting in labor market conditions conducive to criminal behavior. In other words, this instability discourages the kinds of associations that foster social capital or collective efficacy while encouraging the kinds of associations likely to lead to trouble.

The following section elaborates on this premise and details our proposed explanation. After a general discussion of the myriad difficulties faced by released prisoners and the families and communities they return to, we focus in particular on the consequences for local social processes when large numbers of prisoners return to communities. First, we describe the idea of coercive mobility and its potential to disrupt local capacities for informal social control. Second, we discuss the idea of labor instability and its potential to cultivate criminogenic social circumstances. Finally, we describe how these community consequences, may, in turn, increase crime within the community.

Community consequences of returning prisoners

Rose and Clear characterize concentrated incarceration—including both admissions from and releases into a neighborhood—as a combined churning process that creates turmoil within neighborhoods [64]. The present work focuses on one aspect of this process: the consequences of returning prisoners to neighborhoods. Prior work suggests that the consequences of admissions to prison, especially at higher rates, may play a similar and related role [15].4 However, after of decades of prison admissions outnumbering releases—a consequence both of the accelerated use of incarceration and a trend toward longer prison sentences—a recent leveling off of prison admissions has resulted in nearly as many prison releases as admissions in 2009 (see Fig. 1). In other words, the consequences of prison releases have become a critical topic.
https://static-content.springer.com/image/art%3A10.1007%2Fs10611-012-9375-0/MediaObjects/10611_2012_9375_Fig1_HTML.gif
Fig. 1

Number of admissions to and releases from state prisons per 1,000 Americans. ([22, 44, 2, 67], U.S. Census Population Estimates)

Released prisoners have always faced substantial challenges upon their reentry into free society, as have their families and the communities that they return to. In recent years, these challenges have been exponentially exacerbated by the historically and globally unprecedented increase in incarceration as well as reentry programs that are both overwhelmed and unnecessarily inhibited by current policies and practices.

Returning prisoners face a parole system that has become increasingly centered on surveillance and punitiveness over discretion, reintegration, and assistance [1, 51, 74], legal barriers making it more difficult to reunite with their families [24], severely limited employment prospects [50], and overwhelming legal fees, especially relative to their limited earning potential [25, 62]. These challenges have been compounded by the sheer volume of releases in recent years, resulting in state services being increasing overwhelmed [79]. Petersilia reports that just as both the total number and the complexity of the problems of returning prisoners—including drug abuse, mental illness, and diseases such as HIV—have increased, we have responded by increasingly cutting services for them [52].

Releases are also having devastating effects on the communities experiencing them at the highest concentrations, in part because all of these negative social and economic impacts of reentry on individuals and their families have become compounded [78]. In the following sections we develop specific expectations for the effect of concentrations of returning prisoners on two dimensions of community life: a community’s capacity to address local problems and the degree to which a community is socially organized in ways conducive to crime. By harming neighborhoods along these dimensions, we suggest, incarceration may be indirectly increasing the likelihood of future crimes.

Coercive mobility and collective efficacy

Collective efficacy refers to the differential ability of neighborhoods to realize a common goal—in this case, safety—and to act collectively to achieve that goal [73]. Drawing on notions from systemic social disorganization theory [8, 30, 31], Sampson, Morenoff and Earls suggest that this differential capacity is rooted in differential social-structural conditions of neighborhoods [72]. Specifically, they find collective efficacy to be more likely in affluent and residentially stable neighborhoods, and less likely in neighborhoods experiencing concentrations of disadvantageous economic conditions. When incarceration is concentrated at high levels in particular neighborhoods, it may have a deleterious effect on the neighborhood social-structural conditions relevant to collective efficacy.

Concentrations of incarceration are likely to have both direct and indirect effects on the residential stability of a neighborhood. Most directly, incarceration forcibly removes some residents to prison while returning others, resulting in a kind of coercive mobility [64]. The effect is not necessarily insubstantial: Rose and Clear note that in some neighborhoods 15  % of residents are incarcerated every year [63]. Recent changes in parole placing a greater emphasis on surveillance and sanctions for technical violations have exacerbated this coercive mobility by increasing the likelihood that the recently returned will be sent back to prison for violating parole conditions—a phenomenon Lynch and Sabol refer to as “churning” [34]. Incarceration may also increase residential instability when families are unable to afford their current residence either because of the lost income of the person heading to prison or by the economic burden of supporting someone with limited job prospects upon their return from prison.

By reducing the human capital of those returning from prison and placing the burden of their care on families and communities, concentrations of incarceration are likely to have substantially detrimental effects on the economic conditions of a neighborhood, a phenomenon we discuss in greater detail in the following section. Rose and Clear suggest that financial hardship can reduce civic participation, cohesion, and trust among neighbors, all of which are elements conducive to collective efficacy [63]. In addition to decreasing the chances of finding employment, incarceration can also hurt family formation by decreasing the likelihood of getting married among those who have been to prison [28] and disrupting families by removing one or both parents to prison. At the county level, Sabol and Lynch find that high levels of incarceration are associated with an increase in the number of single-mother households, at least for African-Americans [66]. To the degree that two-parent families can provide better social supports and greater social control over youth, neighborhoods with greater numbers of disrupted families will have greater difficulty keeping local youth out of trouble.

Direct evidence for the effect of incarceration on collective efficacy or informal social control has been limited. Several qualitative studies focusing on neighborhoods with high concentrations of incarceration in Tallahassee have suggested that one major impact is the disruption of community relationships that may be important in the production of social capital or informal social control [14, 63]. The stigma associated with imprisonment may also increase isolation and discourage the kinds of connections among local residents useful for social capital or social control [6, 14, 63]. Survey-based research on Baltimore found a small positive direct effect of neighborhood-level changes in prison admissions on informal social control, but also larger negative indirect effects through the social processes on which informal social control depends: participation in voluntary associations, feelings of community solidarity, and neighboring activities [36, 37]. This research, however, focused only on prison admissions rather than releases. Additionally, the research examines an individual’s willingness to participate in informal social control, while collective efficacy represents an emergent property of neighborhoods—shared expectations about the control of local delinquency—that theoretically would not be captured by the simple aggregation of reports of individuals’ willingness to engage in such control.

Our expectation is that concentrations of released prisoners will have negative consequences for neighborhoods’ capacities for collective efficacy, primarily through a worsening of the social-structural conditions this social process is rooted in.

Hypothesis 1

High neighborhood rates of prison releases will be associated with a decreased capacity for collective efficacy; this effect will largely be indirect through the social-structural consequences of released prisoners: labor and housing instability, decreased affluence, and single mother households.

Labor market instability and criminogenic situations

Crutchfield, Matsueda, and Drakulich find that neighborhoods characterized by labor instability—those populated by large numbers of people with little attachment to or investment in their jobs—are more likely than others to foster particular social situations that are conducive to crime: groups of people gathering on street corners or in residences, generally making noise and causing trouble [20].5 Not unlike the social disorganization model described above, this work suggests that the structural context of a neighborhood has consequences for the social organization of the neighborhood; in this case that poor labor market conditions foster criminogenic situations of company. When incarceration is concentrated at high levels in particular neighborhoods, it may have a deleterious effect on the neighborhood labor-market conditions relevant to these social situations.

The labor market consequences of incarceration have been well documented. Most individuals who are incarcerated will eventually be released, and recent work has suggested that incarceration substantially damages the chances of employment upon return. For instance, incarceration significantly reduces the chances of a callback for an entry-level job, an effect that appears to be compounded by race [49] (see also [48, 50]). Even employers who said they would be more willing to hire ex-prisoners were no more likely to actually hire them [47]. When employment opportunities are available to returning prisoners, they exist almost exclusively in the secondary labor market: jobs characterized by low wages, poor conditions, and instability [5, 56]. A combination of missed time during incarceration and poor employment prospects upon return has significant consequences for wage mobility and overall levels of inequality [82, 83]. These employment struggles are compounded by high levels of debt from legal expenses among returning prisoners [25].

Recent research has also investigated the neighborhood-level labor market consequences of high incarceration rates. High incarceration appears to be associated with higher unemployment, lower median incomes, and lower human capital at the group-level, especially for racial and ethnic minorities [57, 66]. Interviews with residents of high incarceration neighborhoods in Tallahassee suggested that the financial impacts of incarceration are perhaps the most consequential [14, 63]. Prisoners return with limited resources, substantial needs, and minimal labor market opportunities, placing a burden on their families and communities (also see [6]). Like Crutchfield [18], Rose and Clear suggest that high unemployment rates result in a kind of social disorder in which large groups congregate in public areas [63].

Our expectation is that concentrations of released prisoners will worsen neighborhood labor conditions, resulting in an increase in social situations conducive to crime.

Hypothesis 2

High neighborhood rates of prison releases will be associated with an increase in social situations conducive to crime; this effect will largely be indirect through the social-structural consequences of released prisoners: labor and housing instability, decreased affluence, and single mother households.

Concentrated incarceration, instability, social organization, and crime

Recent research has explored the direct impacts of incarceration practices on crime rates. Clear et al. find a curvilinear relationship, or a tipping point, between incarceration and crime: low levels of a neighborhood’s incarceration rate decrease the crime rate, higher levels increase it [15]. Lynch and Sabol, however, suggest that this research fails to distinguish the effect of prior crime from the effect of prison admissions [36]. Using the drug arrest rate as an instrument for prison admissions, Lynch and Sabol suggest that prison admissions may have a small negative effect on future crime, though they do not report on the potential for a curvilinear relationship between admissions and crime [36].6 Renaur et al. [61] replicate Clear et al.’s [15] study using new data from Portland, OR and generally support the finding of a curvilinear relationship between prison admissions and future crime, though multicollinearity prevents them from simultaneously estimating an effect for prison releases. Prison releases, as opposed to admission, appear to have a direct linear positive effect on future crime [15, 26].

Prior work has also suggested relationships between the proposed mechanisms and crime. Collective efficacy appears to reduce crime [43, 73], while labor instability [1820] and criminogenic situations of company [20] appear to increase crime. Familial instability—in part a consequence of both incarceration [28, 66] and labor instability [68]—may also play a mediating role [68]. However, collective efficacy’s potential role in explaining the relationship between incarceration and crime, as suggested by Rose and Clear [64], has not yet been directly investigated. Lynch and Sabol investigate the effect of incarceration on both informal social control and crime, but not the effect of informal social control on crime [35, 36]. Similarly, the potential role of labor instability and criminogenic social situations in mediating the relationship between incarceration and crime has also not been directly investigated.

Our basic expectation is that returning prisoners—and likely concentrated incarceration more generally—create economic and residential instability in the neighborhoods to which they return. These two sources of instability act in concert to simultaneously create conditions conducive to crime while weakening controls against it. Economic instability creates situations ripe for crime: concentrations of the un- and under-employed with little to lose and an abundance of free time to get into trouble—in other words, ready reserves of both suitable victims and offenders. At the same time, residential instability impairs a neighborhood’s ability to address such criminogenic situations as they arise—in essence diminishing a neighborhood’s capacity for guardianship. In sum, we expect that the resulting instability discourages the kinds of associations that foster social capital or collective efficacy while encouraging the kinds of associations likely to lead to trouble.

Hypothesis 3

High neighborhood rates of prison releases will be associated with increases in violent crime rates; this effect will largely be indirect through the social-structural consequences of released prisoners—labor and housing instability, decreased affluence, and single mother households—as well as the social organizational consequences of incarceration: increased criminogenic situations and decreased collective efficacy.

Data, models, and measures

Data to test these propositions come from the Seattle Neighborhood Crime Survey (SNCS), a survey of Seattle residents which included measures tapping into communal trust and cohesion and a collective capacity for informal social control [40]. This survey was stratified over census tracts, allowing the integration of contextual data including information on prisoner reentry from the Washington State Department of Corrections, general information on the local context from the 2000 U.S. Census, and crime counts at the census-tract level from the Seattle Police Department. The benefit of this design is that it allows the construction of models that control for potential response bias associated with individual-level characteristics while simultaneously modeling neighborhood-level variation in collective efficacy and criminogenic situations as consequences of contextual conditions.

The Seattle neighborhoods and crime survey

The SNCS is a survey of 5,812 residents of 123 Seattle census tracts conducted in 2002 and early 2003.,78 Four separate sampling schemes were employed in data collection. The first was a simple random sample stratified by tract. The second was designed to be comparable to Miethe’s survey of Seattle residents [42].9 The third sample sought to increase the number of ethnic and racial minorities by disproportionately targeting blocks with the largest proportions of ethnic and racial minorities. Each of the first three samples was drawn from a “white pages” residential telephone directory and contacted via telephone (though respondents were given the option of receiving a paper survey by mail if they preferred it). This method of selecting a sample excludes residences with no telephone, no listed telephone number, or only cell phones (which are generally not listed). To attempt to re-include these residences in the sample, we constructed a list of residential addresses with no listed telephone numbers and drew a new stratified random sample from this to match the first random sample in design. We then sent teams of trained research assistants to the addresses with paper copies of the survey and materials to return the completed survey by mail. To capture mean differences between the samples, controls for the three non-random samples were added to the regressions at the individual level, as was a control for all those who completed the survey on paper rather than by telephone.10 Respondents who completed our survey were somewhat more likely to be white, to have a greater number of years of education, a higher salary, and to own their own home than the average member of the population our survey was designed to represent.

Models

Both collective efficacy and the presence of social situations conducive to crime are neighborhood-level phenomena measured from the reports of individual neighborhood residents. Individuals with particular demographic or biographical attributes may be biased in their reports, and if these attributes are differentially distributed over neighborhoods then neighborhood-level estimates of collective efficacy or criminogenic situations will be biased. For instance, if those who have recently moved to a new neighborhood consistently underestimate their new neighbors’ capacity for informal control, then the capacity for control in neighborhoods with many new residents will be underestimated relative to other neighborhoods. To account for this, a series of individual-level characteristics that may be associated with report bias are controlled for.11 Respondents are clustered in neighborhoods and report on the same neighborhood conditions, so simple random intercept multi-level models are employed to control for individual-level response bias while investigating the neighborhood-level relationship between returning prisoners and other neighborhood characteristics and the presence of collective efficacy and criminogenic social situations. To assess whether neighborhood variation in collective efficacy or criminogenic situations can help explain the association between concentrations of incarceration and future crime, a two-stage process is used employing empirical Bayes residuals (derived from a multi-level model controlling for the person-level measures) as a neighborhood-level predictor.

It is also possible that high levels of collective efficacy, criminogenic conditions, or crime in one neighborhood may be related to the level of collective efficacy, criminogenic conditions, or crime in surrounding neighborhoods. This may result either from a substantive spatial processes like diffusion, or from a mismatch between the administratively-defined census tract boundaries and the behavioral boundaries in which individuals truly experience their local area. To investigate this possibility, we look for spatial dependence in the residuals from neighborhood-level models predicting each outcome.12

Measures

Table 1 presents descriptive information for the survey measures of local social organization as well a series of individual controls. Following Sampson et al. [73], collective efficacy is comprised of two theoretical dimensions, each captured by a number of survey measures. As in Chicago, the two components—trust/cohesion and informal social control—are highly correlated (0.88), so they are combined into a single construct (Table 1). Following Crutchfield et al. [20], criminogenic socialsituations are captured by asking respondents to report on the presence of two kinds of local problems in their neighborhood: groups of teens hanging around and the presence of trouble-causing or noisy neighbors. Finally, Table 1 also describes a number of person-level attributes that may be associated with bias in the reporting of neighborhood levels of collective efficacy or criminogenic social situations.
Table 1

Individual-level measures

 

Mean

S.D.

Collective efficacy (alpha = 0.80)

Average of non-missing responses

2.98

0.53

 Neighbors would respond: kids hanging out

2.61

0.91

 Neighbors would respond: kids graffiti

3.36

0.74

 Neighbors would respond: kids disrespecting adult

2.49

0.83

 Neighbors would respond: kids fighting

3.10

0.80

 In this neighborhood: people willing to help neighbors

3.18

0.57

 In this neighborhood: people can be trusted

3.15

0.63

Criminogenic situations of company (alpha = 0.80)

Average of non-missing responses

1.46

0.54

 Local problem: groups of teens hanging around the street

1.42

0.64

 Local problem: neighbors who cause trouble/make noise

1.49

0.64

Demographics, socio-economics, and residential status

Female

0.50

0.50

Age

48.87

16.25

Married/cohabitating

0.54

0.50

Number of children living at home

0.41

0.82

Years of education completed

15.92

2.56

Household income (mean-replaced)

$67,749

44,426

Income missing flag

0.12

0.33

Number of years at current address

11.83

12.75

Homeowner

0.68

0.47

Asian

0.08

0.27

African-American

0.06

0.23

Latino

0.05

0.21

Foreign born

0.13

0.34

Controls for non-random samples

Member of the Miethe replication sample

0.31

0.46

Member of the ethnic over-sample

0.22

0.41

Member of the sample of unlisted numbers or no phone

0.05

0.22

Completed a mail-back written questionnaire

0.08

0.27

N = 5,382. Note: in subsequent models age is portrayed in tens of years and income in thousands of dollars

Table 2 presents descriptive information for a variety of measures of neighborhood context. Prior work has found that neighborhood structural characteristics are differentially distributed over neighborhoods based on the racial and ethnic composition of the neighborhood, likely related to historical and contemporary discrimination in both the housing and labor market [39, 84]. 2Other work has suggested the racial composition of the neighborhood is associated with differential policing practices [29, 54], and differential likelihoods of the concentration of incarceration [12], likely due at least in part to race and ethnic discrimination in the criminal justice system (see review in [65]). We include controls for the racial and ethnic composition of the neighborhood in part to illustrate these disparities and in part to control for the possible role of unobserved forces such as discrimination in the labor market, housing market, and criminal justice system. In Seattle, Asians are the largest immigrant group and the percentage Asian and the percentage foreign born are correlated very highly at the neighborhood level (0.94), so we use a combined measure for Asian-Americans and immigrants.
Table 2

Neighborhood-level measures

 

Mean

S.D.

Race/ethnicity/immigration

 Proportion African-American

0.08

0.10

 Proportion Latino

0.05

0.04

Asian/foreign born (average of z-scores)

0.00

0.99

 Proportion Asian

0.12

0.12

 Proportion foreign born

0.16

0.11

High-incarceration-risk demographic group

 Proportion males ages 15 to 24

0.07

0.04

Labor and housing instability (alpha = 0.79)

Average of z-scores

0.00

1.37

 Proportion in poverty

0.12

0.09

 Proportion unemployed

0.05

0.04

 Proportion employed in secondary sector

0.14

0.06

 Proportion renters

0.50

0.23

 Proportion not in same residence 5 years ago

0.56

0.13

Concentrated affluence (alpha = 0.90)

Average of z-scores

0.00

0.92

 Proportion households with income >$100 K

0.16

0.10

 Proportion college graduates

0.47

0.17

 Proportion managerial or professional occupations

0.48

0.13

Family structure

 Proportion single-mother households

0.09

0.07

Returning ex-prisoners

 Reentries (released 1998 to 2002) per 100 residents

0.41

0.70

Crime rates

 Average yearly violent crimes per 1 K population 1996–98

9.62

15.50

 Average yearly violent crimes per 1 K population 2003–05

8.21

12.01

 Total violent crimes 2003–05

81.98

92.35

N = 123

Young males are at substantially higher risk of committing crimes than other demographic groups, so we control for the possibility that crime rates are a product of the disproportionate distribution of members of this high-risk group by including a measure of the proportion of the neighborhood which are males between the ages of 15 and 24.

High levels of multicollinearity between measures of neighborhood characteristics necessitated the creation of several indices to capture relevant dimensions of neighborhood context. Instability inthe laborand housingmarkets is captured by the proportion in poverty, the proportion unemployed, the proportion employed in unstable secondary sector jobs, the proportion who rent versus own their home, and the proportion who did not live in the same residence 5 years earlier.13Concentrated affluence, reflecting the idea that extremely advantageous neighborhood conditions are differentially distributed and as important as disadvantageous conditions [7, 38, 72], is captured by the proportion of wealthy households, college graduates, and those employed in managerial or professional occupations. Finally, the proportion ofsingle-mother households is included to capture the differential distribution of family structures over neighborhoods.

To create a relatively stable indicator of returning prisoners measured prior to the survey and crime outcomes and temporally comparable to the other exogenous measures of structure from the census, we used the population of prisoners released by the Washington State Department of Corrections between 1998 and 2002, geo-coded to census tracts, adjusted by population, and multiplied by 100 to produce easier to interpret coefficients (Table 2). Of these ex-prisoners, 41 % were convicted of at least one violent offense, while 40 % committed only drug-related offenses. However, there were not substantial differences by offense type in the neighborhoods prisoners returned to. In general, large numbers of prisoners return to the downtown area, the industrial corridor stretching south from downtown, and the Central District—a historically African-American neighborhood to the east of downtown. Relatively few prisoners return to the whiter and more affluent areas in the northern half of the city or along any of the city’s more affluent waterfront areas.

The violent crime outcome is measured as the number of violent crimes over the three-year period of 2003 to 2005.14 The control for prior crime is measured as the average yearly violent crime rate over the three-year period of 1996 to 1998. To ease comparison with prior crime, Table 2 also reports descriptive statistics for the 2003–2005 average yearly crime rate. Violent crimes include murders, rapes, robberies, and aggravated assaults.

Results

This paper is interested in investigating direct and indirect effects of returning prisoners on neighborhood levels of violent crime. We begin an investigation of potential indirect effects by examining the relationship between returning prisoners and two aspects of local social organization: a capacity for collective efficacy and the presence of social situations conducive to crime. Specifically, we explore whether concentrated incarceration reduces collective efficacy and increases criminogenic social conditions by undermining the local social structural context in which these processes are rooted. Finally, we explore the roles of returning prisoners, social-structural conditions, and social-organizational characteristics in the production of neighborhood violence.

Consequences of returning prisoners for collective efficacy

Table 3 presents coefficients from three models predicting collective efficacy. In each, the person-level covariates represent attributes potentially associated with response bias in the respondent’s report of the presence of collective efficacy in the neighborhood. The results suggest that older persons, cohabitating persons, those with children, homeowners, those with higher incomes, Latinos, and foreign-born persons all tend to report higher levels of collective efficacy relative to their respective comparisons groups, while Asian respondents tend to report lower levels of collective efficacy relative to non-Hispanic whites.
Table 3

Individual and neighborhood predictors of collective efficacy

 

β

S.E.

St.β

β

S.E.

St.β

β

S.E.

St.β

Person-level

 Female

0.03

0.01

0.02

0.02

0.01

0.02

0.02

0.01

0.02

 Agea

0.04***

0.01

0.10

0.04***

0.01

0.10

0.04***

0.01

0.10

 Married/cohabitating

0.08***

0.01

0.06

0.07***

0.01

0.05

0.07***

0.01

0.05

 # of children in home

0.04***

0.01

0.04

0.03***

0.01

0.04

0.03***

0.01

0.04

 Years of education

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

 Incomeb

0.01***

0.00

0.04

0.01***

0.00

0.04

0.01***

0.00

0.04

 Years in neighborhood

0.00

0.00

0.01

0.00

0.00

0.00

0.00

0.00

0.00

 Homeowner

0.13***

0.02

0.09

0.12***

0.02

0.08

0.12***

0.02

0.08

 Asian

−0.08**

0.03

−0.03

−0.08**

0.03

−0.03

−0.08**

0.03

−0.03

 African-American

−0.02

0.03

−0.01

−0.03

0.03

−0.01

−0.02

0.03

−0.01

 Latino

0.08*

0.03

0.02

0.07*

0.03

0.02

0.07*

0.03

0.02

 Foreign born

0.06**

0.02

0.03

0.06**

0.02

0.03

0.06**

0.02

0.03

Neighborhood-level

 Asian/foreign born

−0.04**

0.01

−0.18

0.00

0.01

0.01

0.00

0.01

0.02

 Prop. African-American

−0.41**

0.12

−0.19

−0.44**

0.14

−0.21

−0.47**

0.14

−0.22

 Proportion Latino

−1.01***

0.29

−0.19

−0.51

0.29

−0.10

−0.44

0.29

−0.08

 Proportion males 15–24

−1.34***

0.24

−0.28

−0.60*

0.25

−0.12

−0.65**

0.25

−0.14

 Reentries per 100 residents

−0.06**

0.02

−0.19

−0.01

0.02

−0.03

0.03

0.03

0.11

 Labor/housing instability

   

−0.05***

0.01

−0.32

−0.04**

0.01

−0.25

 Concentrated affluence

   

0.06***

0.02

0.28

0.07***

0.02

0.29

 Single mother households

   

0.37

0.22

0.11

0.32

0.22

0.10

 Violent crime rate’96–‘98

      

−0.03*

0.01

−0.21

 Intercept

3.20***

0.02

 

3.07***

0.03

 

3.08***

0.03

 

 Var. Expl. (w/i, b/w)

0.06, 0.83

0.06, 0.90

0.06, 0.91

*p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed)

N = 5,382 persons, 123 neighborhoods. I.C.C.: 0.16

Model also includes controls for inclusion in non-random sample and missing income

a in tens of years; b in thousands of dollars

The first column of results examines the association between returning prisoners and collective efficacy controlling for the racial composition of the neighborhood and the degree to which the neighborhood’s demographic composition puts it at especially high risk for crime. Conditional on these basic neighborhood-level controls, high rates of ex-prisoners are associated with lower levels of collective efficacy.

The second column of results adds measures of structural conditions of the neighborhood that are likely to be affected by large numbers of returning prisoners. Neighborhoods experiencing high levels of labor and housing instability are substantially less likely to have the capacity for collective efficacy,15 while those experiencing high concentrations of affluent conditions are substantially more likely to act collectively towards a goal of neighborhood safety. Controlling for labor and housing instability and concentrations of affluence appear to account for the effect of released prisoners on collective efficacy, potentially suggesting a mediation effect: that returning prisoners are relevant to levels of collective efficacy because they undermine the neighborhood structural conditions in which it is rooted.

Accounting for labor and housing instability and concentrations of affluence also appear to account for much of the association between the racial/ethnic composition and collective efficacy—only neighborhoods with a larger proportion of African-American residents have significantly lower levels of collective efficacy relative to neighborhoods with greater numbers of non-Hispanic whites.16 Having a large number of young males in the neighborhood remains a significant impediment to the production of collective efficacy.

Finally, the last column of results adds a control for the violent crime rate of the neighborhood, investigating the possibility that high crime rates are themselves an impediment to the production of collective efficacy. However, because the model does not contain a measure of prison admissions, and because violent crimes are more likely than others to result in an incarceration, this measure may also be capturing some of the effect of prison admissions on collective efficacy, as would be expected under a coercive mobility perspective [33, 64]. In either case, a high number of violent crimes are associated with lower levels of collective efficacy. After controlling for the violent crime rate, both neighborhood labor and housing instability and concentrations of affluence remain relevant to the production of collective efficacy.

Consequences of returning prisoners for criminogenic situations

Table 4 presents models exploring the distribution of criminogenic situations of company over neighborhoods similar to those presented for collective efficacy above. As with collective efficacy, the person-level covariates represent attributes potentially associated with response bias in the respondent’s report of the presence of criminogenic social conditions in the neighborhood. Females, younger people, the less educated, the less wealthy, renters, and those not born in the U.S. were all significantly more likely than their reference groups to report the presence of criminogenic situations of company in the neighborhood.
Table 4

Individual and neighborhood predictors of criminogenic situations of company

 

β

S.E.

St.β

β

S.E.

St.β

β

S.E.

St.β

Person-level

 Female

0.05***

0.01

0.04

0.05***

0.01

0.04

0.05***

0.01

0.04

 Age

−0.03***

0.01

−0.07

−0.03***

0.01

−0.07

−0.03***

0.01

−0.07

 Married/cohabitating

0.02

0.02

0.01

0.02

0.02

0.02

0.02

0.02

0.02

 # of children in home

0.00

0.01

0.00

0.00

0.01

0.00

0.00

0.01

0.00

 Years of education

−0.01**

0.00

−0.03

−0.01**

0.00

−0.03

−0.01**

0.00

−0.03

 Income

−0.01**

0.00

−0.04

−0.01**

0.00

−0.03

−0.01**

0.00

−0.03

 Years in neighborhood

−0.00

0.00

−0.01

−0.00

0.00

−0.01

−0.00

0.00

−0.01

 Homeowner

−0.04*

0.02

−0.03

−0.03

0.02

−0.02

−0.03

0.02

−0.02

 Asian

0.04

0.03

0.02

0.04

0.03

0.02

0.04

0.03

0.02

 African-American

0.00

0.03

0.00

0.00

0.03

0.00

0.00

0.03

0.00

 Latino

0.00

0.03

0.00

0.00

0.03

0.00

0.00

0.03

0.00

 Foreign born

0.08***

0.02

0.04

0.07**

0.02

0.04

0.07**

0.02

0.04

Neighborhood-level

 Asian/foreign born

0.02

0.01

0.08

−0.01

0.02

−0.05

−0.01

0.02

−0.06

 Prop. African-American

0.54***

0.14

0.27

0.49**

0.18

0.24

0.54**

0.17

0.27

 Proportion Latino

0.95**

0.34

0.19

0.55

0.37

0.11

0.40

0.35

0.08

 Proportion males 15-24

1.37***

0.28

0.30

0.75*

0.31

0.16

0.88**

0.29

0.19

 Reentries per 100 residents

0.07***

0.02

0.25

0.04

0.02

0.13

−0.05

0.03

−0.19

 Labor/housing instability

   

0.04**

0.01

0.30

0.02

0.02

0.12

 Concentrated affluence

   

−0.04

0.02

−0.16

−0.04*

0.02

−0.20

 Single mother households

   

−0.08

0.28

−0.03

0.03

0.27

0.01

 Violent crime rate’96–‘98

      

0.06***

0.02

0.48

 Intercept

1.22***

0.03

 

1.32***

0.04

 

1.28***

0.04

 

 Var. Expl. (w/i, b/w)

0.02, 0.73

0.02, 0.77

0.03, 0.80

*p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed)

N = 5,382 persons, 123 neighborhoods. I.C.C.: 0.15

Model also includes controls for inclusion in non-random sample and missing income

The results presented in the first column suggest that, conditional on the race composition and demographic environment of the neighborhood, prisoner reentries are associated with a significantly higher chance of the presence of social conditions conducive to criminal behavior. As expected, such conditions are also highly more likely in neighborhoods with large numbers of young males.

The second column of results adds measures of structural conditions of the neighborhood that are likely to be impacted by large numbers of returning prisoners to explore possible mediation. Neighborhoods experiencing high levels of labor and housing instability are significantly more likely to foster social situations conducive to crime.17 Controlling for labor and housing instability results in a reduced and non-significant effect for the number of returning prisoners, potentially suggesting that returning prisoners influence criminogenic conditions by exacerbating disadvantageous labor and housing conditions. Accounting for the effect of labor and housing instability also appears to account for much of the association between the racial composition of the neighborhood and criminogenic situations, with the exception of a remaining association for the proportion African-American.

Finally, the last column of results adds a control for the violent crime rate of the neighborhood, which, as for collective efficacy, might also reflect a role for prison admissions. Not surprisingly, higher numbers of violent crimes in the past make the presence of criminogenic situations of company more likely in the present. Accounting for the prior crime rate results in a reduction in the direct effect of labor and housing instability on criminogenic conditions, but an increase in the role of concentrations of affluence in suppressing such conditions.

Spatial dependence in collective efficacy and criminogenic situations

It is possible that high levels of collective efficacy or criminogenic conditions in one neighborhood are associated with higher levels of these social processes in surrounding neighborhoods, potentially indicating that a process such as diffusion is at work.18 Diagnostics using the residuals of neighborhood-level models predicting the two social processes confirmed the presence of spatial dependence, and tests revealed a spatial lag model would be an appropriate method with which to account for this dependence.19

Table 5 presents coefficients from spatial lag models for both collective efficacy and criminogenic situations of company. The first column of results suggests a positive effect of being proximate to neighborhoods with high levels of collective efficacy on the level of collective efficacy in one’s own neighborhood. When surrounding neighborhoods have a greater capacity for collective action in the interest of neighborhood safety, this may help foster similar efforts in one’s own neighborhood. On the other hand, being surrounded by neighborhoods with a collective efficacy deficit may be a drag on one’s own control efforts. As in the multilevel model, released prisoners are not directly significantly associated with collective efficacy, though unstable labor and housing market conditions are still associated with significant reductions and concentrations of affluence are still associated with significant increases in collective efficacy.
Table 5

Maximum likelihood spatial lag regression: collective efficacy

 

Collective efficacy

Criminogenic situations

Neighborhood-level

β

S.E.

St.β

β

S.E.

St.β

 Asian/foreign born

0.01

0.01

0.05

−0.01

0.01

−0.07

 Prop. African-American

−0.23*

0.11

−0.16

0.35*

0.15

0.23

 Proportion Latino

−0.37

0.22

−0.10

0.29

0.28

0.07

 Proportion males 15–24

−0.51**

0.18

−0.16

0.65**

0.22

0.19

 Reentries per 100 residents

0.02

0.02

0.10

−0.04

0.02

−0.20

 Labor/housing instability

−0.02*

0.01

−0.20

0.01

0.01

0.09

 Concentrated affluence

0.04**

0.01

0.28

−0.03

0.02

−0.18

 Single mother households

0.14

0.17

0.06

0.11

0.22

0.05

 Violent crime rate’96–‘98

−0.01

0.01

−0.14

0.04***

0.01

0.44

 Spatial lag of outcome

0.37***

0.09

0.26

0.22***

0.10

0.15

 Intercept

0.06**

0.02

 

−0.12***

0.03

 

 Variance explaineda

0.74

0.65

*p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed). N = 123

a Squared correlation between fitted and observed values

The second column of results reveals a positive effect of criminogenic social situation in one neighborhood on such situations in surrounding neighborhoods. It appears that public gatherings of trouble-causing young men in surrounding neighborhoods may either spill over into or otherwise inspire similar gatherings in one’s own neighborhood. Accounting for this spatial dependence reduces the effect of concentrations of affluence below conventional levels of significance (the effect is still significant at p < 0.1). However, caution is recommended in interpreting this change, as a high degree of multicollinearity exists between concentrated affluence, labor and housing instability, and the spatial lag.20

Consequences of returning prisoners for violent crime

Table 6 presents results exploring the role of returning prisoners, neighborhood structural conditions, collective efficacy, and criminogenic situations of company in the distribution of violent crime. As violent crimes are relatively rare events, using a crime rate as an outcome may result in neighborhoods with smaller populations having larger errors, violating the assumption of constant error variance (see [46]). As such, we present results for violent crime counts using a negative binomial model and account for differences in tract population by controlling for the natural log of the population.21 Notably, no spatial dependence was found in the residuals of models that control for prior crime, so spatial models were not estimated.
Table 6

Negative binomial regression: 2003–2005 violent crime counts

 

1

2

3

4

5

Neighborhood-level

β

S.E.

β

S.E.

β

S.E.

β

S.E.

β

S.E.

exp

Logged population

0.31

0.16

0.60***

0.13

0.57***

0.12

0.59***

0.12

0.68***

0.11

1.98

Asian/foreign born

0.13

0.07

−0.08

0.07

−0.07

0.07

−0.07

0.07

−0.07

0.06

0.94

Prop. African-American

2.23**

0.73

2.37**

0.76

1.89*

0.75

1.47*

0.74

1.76*

0.70

5.81

Proportion Latino

4.82**

1.71

2.52

1.49

1.79

1.46

1.36

1.41

1.28

1.32

3.58

Proportion males 15–24

6.01***

1.40

−0.05

1.24

−1.03

1.22

−1.42

1.19

−0.62

1.13

0.54

Reentries per 100 residents

0.61***

0.10

0.25**

0.09

0.18*

0.09

0.21*

0.08

−0.10

0.11

0.91

Labor/housing instability

  

0.44***

0.06

0.39***

0.06

0.33***

0.06

0.26***

0.06

1.29

Concentrated affluence

  

−0.26**

0.10

−0.25**

0.10

−0.19*

0.10

−0.22*

0.09

0.80

Single mother households

  

−2.42*

1.19

−2.50*

1.15

−1.92

1.11

−1.35

1.05

0.26

Criminogenic situations

    

1.32**

0.44

0.66

0.47

0.09

0.47

1.09

Collective efficacy

      

−1.69**

0.60

−1.49**

0.57

0.22

Violent crime rate’96–‘98

        

0.02***

0.01

1.02

Intercept

0.40

1.35

−1.23

1.06

−0.75

1.03

−0.95

1.00

−1.96*

0.96

 

Dispersion (θ)

2.30***

0.29

4.10***

0.55

4.41***

0.60

4.81***

0.67

5.51***

0.78

 

LR test p value

<0.001a

 

<0.001b

 

<0.004b

 

<0.004b

 

<0.001b

  

*p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed). N = 123 neighborhoods

a versus empty model. b versus model in column to left

The first column presents results from a model assessing the relationship between prisoner reentries and violent crime controlling for the demographic and racial composition of the neighborhood. Neighborhoods with larger populations of young men, not surprisingly, are likely to have higher violent crime rates. Similarly, neighborhoods with larger numbers of African-Americans and Latinos are likely to have higher violent crime rates. Conditional on these controls, higher rates of prison returnees are significantly related to increases in violent crime.

The second column suggests that much of the effect of prisoner reentries on violent crime is potentially indirect through increased disadvantageous neighborhood conditions like labor and housing instability, a dearth of affluence, and the presence of single-mother households.22 Accounting for these social structural conditions also explains much of the association between violent crime and both the presence of young males as well as the proportion Latino.

The third column includes the presence of criminogenic social situations in the neighborhood.23 When such situations are present, violent crimes are significantly more likely. Accounting for the presence of criminogenic situations appears to account for some of the relationships between crime and prison reentries, labor and housing instability, and the proportion African-American.

The fourth column adds the relative capacity for collective efficacy, which is associated with a significant reduction in the level of violent crime. Adding a measure of collective efficacy appears to mediate the direct effect of criminogenic social conditions on crime. In other words, the relationship between criminogenic conditions and crime appears to be explained by a neighborhood’s capacity to control such situations. The relationship between criminogenic conditions and a capacity to control criminogenic conditions is somewhat tautological, so the lack of an effect for criminogenic conditions when the capacity to control such conditions is accounted for does not necessarily mean that criminogenic conditions are irrelevant, but rather that they are relevant because some neighborhoods are better equipped to cope with them than others.

The fifth and final column adds a control for the neighborhood prior violent crime rate. Crime rates are relatively stable in neighborhoods over time, as reflected in the strong effect of prior crime on future crime. That prior crime rate, of course, may be a product of the social structure of and social processes within neighborhoods. Once prior crime is controlled for, the model only attempts the explanation of short-term variations in crime, a conservative test of the other substantive predictors.24 In a model investigating a deleterious role for incarceration, the omission of a measure of prison admissions further complicates the interpretation of prior crime, by creating an omitted variable problem in which the effect of prison admissions on future crime may be captured by other predictors with which prison admissions are associated: the prior violent crime rate that will lead to prison admissions, the rate of returning prisoners which prior research has found to be strongly correlated with admissions [15], as well as the other conditions of neighborhoods predictive of crime and incarceration.

Controlling for prior crime appears to explain away the direct effect of prison reentries on future crime, though labor and housing instability, concentrated affluence, and collective efficacy all retain significant effects. Exponentiated coefficients representing log odds are included in the final column to facilitate interpretation of the magnitude of the effects on crime, though the relative magnitude of these relationships is more clearly illustrated in Fig. 2, which plots means and 95 % confidence intervals for expected changes in the incidence of violence for standard deviation changes in each of the predictors.
https://static-content.springer.com/image/art%3A10.1007%2Fs10611-012-9375-0/MediaObjects/10611_2012_9375_Fig2_HTML.gif
Fig. 2

Each line represents the expected difference (and 95 % confidence interval) in the number of violent crimes between a neighborhood with low levels of the independent variable versus one with high levels (specifically −1 S.D. versus +1 S.D)

The figure suggests that conditional on the other covariates, a neighborhood one standard deviation above the mean level of labor instability in Seattle would have had 39 more violent crimes over the 3 year period than a neighborhood which is one standard deviation below the mean. To give some context, this represents the difference, roughly, between a neighborhood in which 3 % of residents are in poverty, 9 % are out of work or marginally employed, and 27 % rent rather than own, and a neighborhood in which 20 % are in poverty, 27 % are marginally or unemployed, and 73 % rent rather than own. A similar difference in prior violent crime rates results in 40 more crimes over the 3 year period.

A change from one standard deviation below the mean to one above the mean for concentrated affluence was worth a 22 crime reduction. This represents, roughly, the difference between a neighborhood in which 6 % of residents have incomes above a hundred thousand dollars, 30 % are college graduates, and 35 % work in managerial or professional occupations, and a neighborhood in which 26 % have incomes above a hundred thousand dollars, 65 % are college graduates, and 62 % work in managerial or professional occupations. A similar difference in levels of collective efficacy nets a 23 crime reduction.

Discussion

This research contributes to a growing understanding of the consequences of high levels of incarceration for communities. The results suggest that concentrations of returning prisoners are associated with a reduced capacity for collective efficacy, the fostering of social situations conducive to criminal behavior, and higher levels of violent crime. The impact of incarceration on these neighborhood processes, however, appears to be largely indirect through the turmoil that concentrations of incarceration create in a neighborhood’s labor and housing markets.

These findings confirm prior work on the consequences of incarceration in that high levels of returning prisoners seem to have relatively unambiguously negative consequences for neighborhoods [15, 26]. The consequences of prison admissions for neighborhoods, however, may have the ability to lower crimes rates, at least when used sparingly [15, 36, 61]. Prison admissions may also have small positive consequences for local informal social control efforts, though these are likely overwhelmed by negative indirect effects [36, 37].

Thus it is possible that the present work, in not being able to include a measure of prison admissions, misses a potential positive consequence of incarceration for neighborhoods—that at low levels prison admissions may successfully achieve their purported goal of “addition by subtraction” [11]. However, even if a strategically frugal use of prison admissions has the potential to reduce crime, the case for a net benefit of incarceration for neighborhoods faces two obstacles. First, it is clear that incarceration is not currently being used in such a parsimonious fashion. Instead, some urban neighborhoods are being exposed to extraordinarily high levels of incarceration [12]. When incarceration is employed at this magnitude, a combination of weakened deterrent effects and costly collateral consequences make it likely that “the crime-reducing aspects of incarceration on communities are considerably negated by the crime-enhancing ones” ([10], p. 1). Second, even when used sparingly, prison admissions will still eventually lead to prison releases in the vast majority of cases, and these releases, at least as they currently are handled, appear to have negative consequences for the individual being released, and, as this research suggests, for the community that they return to.

Conclusion

Over the last century, the U.S. has begun using prison as a common response to an increasingly wide variety of behaviors. If the aim of incarceration is a social good—the incapacitation of those who might do harm to others or some kind of moral retribution for the harm they have already done—then the social costs of incarceration must be included when assessing whether a net good is really being achieved. Prior work has suggested that sending people to prison may have substantial social costs [12], and the present work emphasizes in particular the costs for communities when prisoners are released—especially at high volume and without adequate preparation or support. In this sense prison punishes not just an offender and their family, but also the communities they come from and return to.

Criminologists may not find these conclusions surprising. Many of our sociological theories of individual involvement in crime are not optimistic about the prospects of incarceration reducing crime, especially when employed at extraordinarily high levels. Social control theories emphasize the importance of conventional bonds to work and family [27, 69] that are likely to be broken by prison [28, 50]. Differential association emphasizes relationships with people who convey favorable definitions of crime as well as the skills and knowledge to commit them [77], and prison potentially provides the opportunity to dramatically increase the number of associations with pro-crime views and skills. Similarly, as prison appears to block conventional routes to success [50, 83], individuals may respond to the strain resulting from the mismatch between cultural expectations of success and these limited opportunities by increasingly choosing adaptations that may result in crime [41]. The present work adds to a growing body of literature suggesting that group-level sociological theories of crime, too, are pessimistic about the prospects of incarceration reducing crime. Specifically, by damaging local labor and housing market conditions, incarceration undermines less coercive efforts to control crime and fosters situations likely to lead to future crime.

If the current level and trend of incarceration is doing more harm than good, then the formulation of an exit plan may require a better understanding of the forces that brought us to this point. Our policies and practices surrounding release have been moving away from any focus on discretion, reintegration, and assistance and instead toward surveillance and punitiveness, which may be decreasing the chances that reentries will be successful [1, 51, 74]. This shift may be part of a larger political agenda to “replace social welfare with social control as the principle of state policy” ([4], p. 10), especially with regard to the governance of marginal social populations [3]. This shift has been exacerbated by the increasing number of private prisons for which high incarceration rates are good business. The private prison industry can encourage high incarceration rates by engaging in a kind of marketing that acclimates the public to see high incarceration rates as inevitable and to view criminals as commodities to be managed in a risk market [9].

Though the overall rate of incarceration in the country is staggeringly high, it has been concentrated among specific demographic populations and within specific geographic spaces. Some suggest this may not have been accidental. Racial antipathy appears to have played a role in the genesis of current drug policies [58] as well as the larger “tough on crime” movement [4, 75, 80]. Incarceration is both rooted in existing race and class disparities but simultaneously acts to worsen these inequalities [83], in part through processes described in this paper.

This self-reinforcing process plays out not just on a broad scale, but also within neighborhoods. Fagan, West, and Holland [21] find that neighborhoods with high incarceration rates are also subject to higher rates of aggressive policing, increasing the likelihood that incarceration will continue to rise in these places. The present research illustrates some of the ways that returning offenders with reduced human capital damage the social capital of the neighborhood. When incarceration damages human and social capital within neighborhoods, it also increases the social isolation of these neighborhoods from mainstream society and institutions [13], making the process for their repair even more daunting. The most troubling implication of the present work, then, is that as a cycle of disadvantage, crime, and incarceration begin to spin some neighborhoods out of control, current criminal justice policies and practices appear to be accelerating rather than reversing this momentum.

Footnotes
1

http://bjs.ojp.usdoj.gov/content/glance/tables/corr2tab.cfm, population estimates from www.census.gov, both accessed on 8/25/2011.

 
3

Incarceration estimates from http://bjs.ojp.usdoj.gov/content/pub/pdf/pim09st.pdf; population estimates from http://www.census.gov/popest/national/asrh/NC-EST2009-asrh.html, both accessed on 8/25/2010.

 
4

The implications of focusing on releases over admissions for the present analysis are elaborated in the Results and Discussion section, where we also raise questions for future research.

 
5

In this sense Crutchfield et al. [20] avoid the “materialistic fallacy” that economic strife is limited to the production of economic motivations for crime [70], instead drawing on control [27] and routine activities [17] perspectives to argue that such economic conditions cluster those with reduced attachments and commitments to conventional labor market institutions, and that such clusters provide ample supplies of offenders with diminished constraints as well as suitable victims in environments likely bereft of capable guardians.

 
6

For a longer discussion of issues related to simultaneity and endogeneity, see chapter 7 of Clear’s book [12].

 
7

One census tract, entirely composed of the University of Washington campus, was not included in the survey. For this tract, census data are unreliable (the population turns over yearly), Seattle Police data are incomplete (the University’s police force handles most of the crime), and we collected no survey information.

 
8

The cooperation rate for the survey was over 97 %. The AAPOR response rate was over 50 %.

 
9

Miethe targeted 600 city blocks of 100 of Seattle’s then 121 census tracts. In each tract, three blocks with at least one burglary reported to the police in 1989 were selected, along with one adjoining “control” block for each [42]. The SNCS targeted these same street segments, but in many cases had to extend the block in one direction or the other to achieve the requisite number of completed surveys.

 
10

Comparison of responses from phone surveys, mail-back surveys, and questionnaires from households with unlisted telephone numbers revealed similar distributions on key variables.

 
11

At the individual level, the models ask how personal characteristics are associated with perceptions of collective efficacy and situations of company. This paper, however, is interested in relative differences between neighborhoods in the overall presence of collective efficacy and situations of company. If, for instance, female respondents are more likely to report the presence of collective efficacy than their male counterparts in the same neighborhood, then regardless of who is “more correct” about the presence of collective efficacy, a neighborhood where we happen to interview a larger number of men will appear to have less collective efficacy than a neighborhood where we happen to interview a large number of females even if the “true” levels of collective efficacy are identical. Thus, we control for individual characteristics not because we are interested substantively in the influence of these characteristics on perceptions of collective efficacy or situations of company, but to create comparable measures of neighborhood-level values of these qualities to assess whether neighborhood characteristics are associated with relative differences in collective efficacy and informal social control.

 
12

Proximity is measured using a “queen” type contiguity-based spatial weight matrix, which defines all tracts sharing a common border or at least a common corner with a tract as neighbors. Contiguity was restricted across major waterways separating census tracts in Seattle (the Duwamish River and the Lake Washington Ship Canal).

 
13

Ideally labor and housing instability would be examined separately, as the expectation is that collective efficacy will be particularly strongly related to residential instability while criminogenic situations will be particularly strongly related to labor instability. In practice, however, the two sources of instability overlap so highly across Seattle neighborhoods—a trend that should not be surprising as they are likely both products of similar macro-social forces [53, 84]—that it is not possible to empirically distinguish their effects. In the presentation of results I also report on additional exploratory models in which only labor or residential instability is included.

 
14

A 3-year period is used to reduce the influence of single-year aberrations in the number of crimes in a neighborhood.

 
15

Exploratory analyses in which residential and labor instability were considered independently in consecutive models suggest that residential instability plays a much more important role in the production of collective efficacy than labor instability—at least in models that also control for affluence, the presence or absence of which appears to be the more relevant economic dimension for this social process.

 
16

Or, more precisely: perceptions of collective efficacy—controlling for differences based on individual race, education and other characteristics—are lower in neighborhoods with greater numbers of African-Americans. It is possible that some form of racial stigma drives respondents to perceive neighborhoods with greater numbers of African-Americans as having lower capacities for collective action than they truly do—prior work has suggested that residents perceive the presence of local disorder [71] and safety more generally [59] through a racial lens. If this stigma is not captured by the individual-level controls then the association between the racial composition and collective efficacy may reflect perceptual rather than true differences. This caveat also applies to the interpretation of the association between the racial composition and situations of company.

 
17

Exploratory analyses in which residential and labor instability were considered independently in consecutive models suggest that both residential and labor instability are strongly related to criminogenic conditions when considered on their own.

 
18

Though the present model is not able to identify the source of this dependence as being a substantive process like diffusion versus measurement error resulting from a mismatch between administratively-defined census tract boundaries and the behavioral boundaries in which individuals truly experience their local area. At minimum, accounting for spatial dependence addresses the omitted variable problem which exists in models that do not account for existing spatial dependence.

 
19

These models employ empirical Bayes residuals derived from a multi-level model which includes controls for the various sources of potential response bias at the person-level. Spatial dependence is assessed using a Moran’s I test for spatial autocorrelation in residuals from a linear model predicting the empirical Bayes residuals.

 
20

Variance inflation factors are quite high when instability is included and its removal results in a larger significant negative role for concentrated affluence.

 
21

A Poisson model, also used for counts, assumes that the mean and variance are equal. However, as the likelihood ratio tests for dispersion (included in Table 6) confirm, a negative binomial model which accounts for overdispersion (where the variance exceeds the mean) is more appropriate.

 
22

Exploratory analyses in which residential and labor instability were considered independently in consecutive models suggest that both residential and labor instability each are strongly related to crime when considered on their own.

 
23

The measures of both criminogenic conditions and collective efficacy are empirical Bayes residuals derived from a multi-level model that includes controls for the various sources of potential response bias at the person-level. Empirical Bayes residuals are equal to the least-squares residuals shrunk towards zero by a factor equal to their unreliability ([60], p. 48).

 
24

Controlling for prior crime is also conservative in the sense that some of the observed correlation between prior and future crime is likely the product of a strong error correlation resulting from similar systematic biases in the reporting of crime in each time period. Exploratory analyses suggest evidence for this in a weaker effect for a different measure of prior crime constructed from survey victimization reports—though such a measure is subject to its own measurement issues (e.g. [23, 32]). In fact, a logged version of prior crime (which matches the functional form of the crime outcome induced by the log-link function) is so strongly related to future crime that other substantive predictors do not have significant effects. Alternatively, column 4 can be viewed as a basic cross-sectional model.

 

Acknowledgements

This research is supported by grants from the Open Society Institute, the Jeht Foundation, the National Science Foundation (SES-0004324, SES-0966662), and the National Consortium on Violence Research (SBR-9513040). An early version of this work was presented at the American Society of Criminology Annual Meeting in Los Angeles, November 2006. Thanks to Northeastern University’s School of Criminology and Criminal Justice Faculty Writing Group as well as the anonymous reviewers for comments and guidance on an earlier draft of this paper.

Copyright information

© Springer Science+Business Media B.V. 2012