Examining parolees in their communities: Poverty, rurality, and criminal justice resources
vii TABLE OF CONTENTS
LIST OF TABLES ix LIST OF FIGURES x ACKNOWLEDGEMENTS xi CHAPTER 1: INTRODUCTION 1 CHAPTER 2: LITERATURE REVIEW 7 Defining Recidivism 7 Individual-Level Risk Factors 8 Criminal Record 8 Demographic Variables 9 Criminogenic Need 10 Social Bonds 10 Effects of Prison 11 Community-Level Studies 12 Criticisms of Neighborhood Effects Studies 15 Parolees’ Communities 18 Research on Poverty 20 Concentrated Disadvantage 21 Extreme Poverty 26 Relative Deprivation 28 Racial Inequality 30 Poximity to Poverty 33 Conclusions from the Poverty Discussions 35 Urban/ Rural Continuum 37 Criminal Justice Resources 40 CHAPTER 3: RESEARCH GOALS AND HYPOTHESES 43 Concentrated Disadvantage and Recidivism 43 Extreme Poverty and Recidivism 44 Relative Deprivation and Recidivism 45 Racial Inequality and Recidivism 46 Proximity to Poverty and Recidivism 48 Rurality and Recidivism 48 Criminal Justice Resources and Recidivism 49 CHAPTER 4: METHODOLOGY AND DATA 52 The Study Site 52 Data Preparation and Measurement 54 County-Level Data: Unit of Analysis 56 Measurement of Variables 58 Dependent Variable: Recidivism 58 Independent Variables: Poverty Variables 60 Concentrated Disadvantage 60 Extreme Poverty 62 Relative Deprivation 62 Racial Inequality 63 Spatial Poverty 63
viii Rural/ Urban Measures 65 Criminal Justice Resources 66 Parole Offices and Spatial Measures 66 UCR Crime Data 66 Individual Parolee Demography 67 CHAPTER 5: DESCRIPTIVE RESULTS 70 Descriptive Statistics 70 Maps of Parolees and Recidivism in Georgia 75 Bivaiate Correlations 77 Statistical Methods 81 Missing Data Techniques 81 Multilevel Analyses 82 CHAPTER 6: MULTIVARIATE RESULTS 84 Hierarchical Logistic Regession Models 84 Equation Models 85 Unconditional Models 88 Individual-Level Model: Parolees’ Backgrounds and Characteristics 89 Community-Level Models 92 Poverty Models: Concentrated Disadvantage 92 Poverty Models: Extreme Poverty 95 Poverty Models: Relative Deprivation 99 Poverty Models: Racial Inequality 100 Poverty Models: Spatial Proximity to Poverty 105 Rural Models: The Effects of Rural and Sububan Communities 108 Criminal Justice Resource Models: Parole Offices and Crime Rates 110 Summary of Main Findings 112 CHAPTER 7: DISCUSSIONS AND CONCLUSIONS 114 Findings 114 Communities and Recidivism 114 Race and Recidivism 118 Race, Risk Scores, and Recidivism 120 Community Poverty Measures 122 Aggregate-Level Recidivism 123 The Relative Strength of Individual-Level Predictors 124 Limitations 125 Conclusions 128 REFERENCES 131 APPENDIX A 156 Maps of the Independent Variables 156 APPENDIX B 159 Additional Variance Components Analyses 159 APPENDIX C 162 Examining the Effects of Poverty Upon Aggregate-Level Recidivism 162 APPENDIX D 165 Examining the Effects of Racial Inequality on White Parolees 165
LIST OF TABLES
Table 1: Comparing Parolees in Georgia to Parolees across the United States 53 Table 2: Description of Variables 55 Table 3: Reliabilities and Factor Loadings on Social Disorganization Constructs Using Principal Components Analysis with Varimax Rotation 61 Table 4: Descriptive Statistics 72 Table 5: Correlation Matrix of Dependent and Individual-Level Independent Variables 78 Table 6: Correlation Matrix of Community-Level Variables 79 Table 7: Hierarchical Logistic Regression Models Predicting Parolee Recidivism 91 Table 8: Concentrated Disadvantage Hierarchical Logistic Regression Models Predicting Parolee Recidivism 93 Table 9: Extreme Poverty Hierarchical Logistic Regression Models Predicting Parolee Recidivism 97 Table 10: Relative Deprivation Hierarchical Logistic Regression Models Predicting Parolee Recidivism 100 Table 11: Racial Inequality Hierarchical Logistic Regression Models Predicting Parolee Recidivism 101 Table 12: Racial Inequality Hierarchical Logistic Regression Models Predicting Minority Parolee Recidivism 104 Table 13: Spatial Measures of Proximity to Poverty Hierarchical Logistic Regression Models Predicting Parolee Recidivism 107 Table 14: Rural Hierarchical Logistic Regression Models Predicting Parolee Recidivism 109 Table 15: Criminal Justice Resources Hierarchical Logistic Regression Models Pedicting Parolee Recidivism 111 Table 16: Re-examining the Proportions of Explained Variance 161 Table 17: Poverty Measures Hierarchical Poission Regression Models Predicting Aggregate Parolee Recidivism 164 Table 18: Racial Inequality Hierarchical Logistic Regression Models Predicting White Parolee Recidivism 166
LIST OF FIGURES
Figure 1: Major Cities in the State of Georgia 76 Figure 2: Percentage of Parolees in the Population 76 Figure 3: Recidivism Rates of Parolees in Georgia 77 Figure 4: The Effects of Concentrated Disadvantage by Race of Parolees on Recidivism 94 Figure 5: The Effects of Extreme Poverty by Race of Parolees on Recidivism 98 Figure 6: The Effects of Racial Inequality by Risk Scores of Minority Parolees on Recidivism 103 Figure 7: Percentage of County Residents Who Receive Public Assistance 156 Figure 8: Percentage of Residents Who Live Below the Poverty Line 157 Figure 9: Percentage of Rural Counties in Georgia 157 Figure 10: Number of Parole Offices in Georgia 158 Figure 11: County Crime Rates in Georgia 158
First of all, I would like to thank my advisor, Barry Ruback. It has been an honor and a pleasure to work with such a supportive mentor. Whatever I achieve in the future will be due in no small part to his mentorship, guidance, and the experience of rewrites. I would also like to thank Wayne Osgood who introduced me to the world of evaluation research. I little realized at the time that those early days evaluating a drug and alcohol abatement program would bring me to my career path. I would like to also thank Barry Lee and Alex Klippel, who graciously agreed to serve on my dissertation committee. I would like to thank my friends who have made graduate school so enjoyable. I had the lucky fortune of coming to Penn State with one of the most amazing cohorts of friends – Gretchen Ruth Cusick, Keri Buchfeld, Alison Cares and her husband Todd, Arnold Alexander, and Karla Haber. You all made State College so much more than just a rural college town. Also, I would like to thank two of my mentors in GIS and spatial analysis, Karen Hayslett-McCall and Michelle Zeiders. I still love what I get to do for a living and I owe my thanks to you both for ushering me into this strange world. Finally, but most importantly, I would like to thank my parents, Valerie K. and I. Townsend Burden; my sister, Virginia K. Burden; my grandmother, Virginia Knauer; and the rest of my family for their unfailing guidance, love, and support. I never could have accomplished all that I have without them and I truly appreciate their valiant attempts to stay interested in a subject so tangential to their interests. I would also like to thank my husband William Pate who endured two years of my dissertation. I am so thankful that you find women who do statistics appealing.
CHAPTER 1: INTRODUCTION
Almost 70 percent of parolees are arrested for a new offense within three years of their initial release (Langan and Levine, 2002). Parolees have such high recidivism rates because they have risk factors that predispose them to crimes (e.g., youth, broken family, uneven employment history), and their incarceration weakened such protective factors as health and family relations. This study focuses on a third risk factor for parolees -- the communities to which they are released. Research indicates that certain communities are linked to negative outcomes, such as lower educational achievement and higher crime rates. These “risky” communities are often precisely those places to which parolees return. Using the entire parolee population in Georgia, this study examines how community contextual factors are related to parolees’ recidivism. Unlike most of the urban sociology literature, which focuses attention on cities, and sometimes only the slums within these cities, this study compares extremes, including poor to affluent areas, rural to urban areas, and areas with greater criminal justice resources to those with fewer resources. This study addresses three questions. First, this study examines how community poverty (measured in different ways to tap into selected theories of poverty) affects individual parolees. Secondly, this study analyzes recidivism rates and whether they vary across urban and rural areas. Finally, this study looks at how the presence of parole offices affects recidivism among parolees.
2 With regard to the first component, this study examines five theoretical perspectives on how poverty might be related to recidivism: concentrated disadvantage, extreme poverty, relative deprivation, racial inequality, and proximity to poverty. Concentrated disadvantage assesses the level of structural poverty in communities and is one of three variables from social disorganization theory. Social disorganization studies generally measure concentrated disadvantage as a composite of structural characteristics, such as female-headed households, rates of welfare recipients, unemployment rates, and percentages of the population living below the poverty line. Generally, concentrated disadvantage has consistently predicted crime rates (Sampson and Lauritsen, 1994) and in two studies concentrated disadvantage has been shown to increase parolee recidivism (Kubrin and Stewart, 2006; Mears et al., 2008). William Julius Wilson (1987, 1996) suggested that it was not poverty that affects crime rates and other social problems, but the extreme poverty faced by many inner- cities. Communities with extreme poverty are indeed so poor that there are few or no employment opportunities and often no positive middleclass influences (Wilson, 1987, 1996). It is also likely that extreme poverty communities have negative effects on a parolee’s likelihood of staying out of prison, as these communities have few positive influences (e.g., employment opportunities, positive social bonds) and multitudinous negative influences (e.g., high crime rate, unreliable public transportation). Aside from concentrated disadvantage and extreme poverty, crime rates are driven by economic inequalities and the “relative deprivation” that individuals feel as they compare their lot in life with others. According to this theory, individuals commit crime either to supplement their needs (Merton, 1938) or to vent their frustration (Blau
3 and Blau, 1982). Consistent with this theory, studies have found that relative deprivation is generally linked to crime rates (Land et al., 1990; Blau and Blau, 1982; Loftin and Hill, 1974), and one study in particular found a positive relationship between relative deprivation and recidivism (Kubrin and Stewart, 2006). While relative deprivation measures the economic deprivation one faces, racial inequality attempts to measure the deprivation one race faces compared to members of another race. A recent article linked community racial inequality to higher recidivism rates for African American parolees (Reisig et al., 2007). This finding is in line with other racial inequality research that links high rates of racial inequality with black interracial homicide (Parker & McCall, 1999), the number of African Americans killed by police, and the number police officers killed (Jacobs, 1998). Therefore, one would predict that high levels of racial inequality would also predict higher rates of recidivism in Georgia, particularly for minority parolees. Finally, proximity to poverty is also linked to crime rates. From this diffusion perspective, crimes are committed in and around areas where poverty rates are high. Generally, crimes cannot be explained by the presence of structural factors alone (Baller et al., 2001); rather, they cluster in space (Sherman et al., 1989). Recently, studies have suggested that community poverty clusters also predict certain types of crimes (Stretesky et al., 2004; Mears and Bhati, 2006). These studies suggest that impoverished communities have a stronger negative effect upon residents when these residents are surrounded by other impoverished communities (Krivo and Peterson, 1996). This study examines the effects clusters of poverty and their effects upon parolee recidivism.
4 This study uses these five theoretical perspectives on the role of poverty in predicting parolees’ recidivism. Each of these theories has a different prediction. Concentrated disadvantage and the extreme poverty perspectives suggest that poor communities have high recidivism rates, although each theory may identify different poor communities. Relative deprivation suggests that wealthy areas situated near poor areas have high recidivism rates, due to people from poor areas traveling to an attractive crime opportunity. Racial inequality indicates that counties with large differences in economic achievements between African American and Caucasian residents have high rates of recidivism, whereas counties in which African Americans and Caucasians are economically similar have low levels of recidivism. Finally, the proximity perspective suggests that communities, whether poor or wealthy, have higher crime rates if situated near a poor community. The second component of this study examines how rurality affects recidivism. Almost all research on criminals has been conducted in urban areas (e.g., Shaw and McKay, 1942), despite the fact that 59.6 percent of the population in the United States lives in rural and suburban areas (Census, 2000). The issue is theoretically important because evidence suggests that crime rates and the correlates of crime (e.g., poverty and informal social control) differ greatly between urban, rural, and suburban areas. Thus, it may be incorrect to assume that the causes of crime are invariant across the spectrum of rural and urban areas. Finally, the third component of this study explores the relationship between the presence of parole offices in counties and the recidivism rates for those counties. The ability of a parole officer to monitor parolees is likely to be affected by both the officer’s
5 caseload (i.e., the number of parolees he or she needs to supervise) and the characteristics of the officer’s domain (e.g., long distances between parolees might translate into less supervision). By examining the 53 parole offices across Georgia, this study attempts to understand how criminal justice agencies and their geographical distribution affect recidivism rates among parolees. Georgia is an excellent study area for this current investigation as it has one of the largest per capita incarceration rates nationally (Hughes et al., 2001). Additionally, Georgia offers a diverse range of structural characteristics including urban areas (e.g., Atlanta), several counties that are exclusively rural, a sizeable minority population, and severe pockets of both rural and urban poverty. Moreover, Georgia has invested heavily in data collection and these computerized records are accessible to researchers.
In sum, this study addresses three broad questions: (1) Of the five theoretical perspectives on poverty (concentrated disadvantage, extreme poverty, relative deprivation, racial inequality, and proximity to poverty), which is the best predictor of recidivism?; (2) Is a parolee’s successful reintegration into society affected by living in a rural area?; and (3) Is there a relationship between presence of criminal justice resources and higher recidivism rates? The next chapter summarizes the literature on parolees and the three components of the current study. First, individual-level risk factors for parolees and their communities are examined. Then, community-level theories on poverty, rural areas, and the distribution of criminal justice resources in Georgia are discussed. This literature review is followed by Chapter 3, which describes the seven hypotheses underlying this
6 research. Chapter 4 describes the methodology and data. The results are described in Chapter 5 (Descriptive Results) and Chapter 6 (Multivariate Results). Chapter 7 contains the discussion and conclusions.
CHAPTER 2: LITERATURE REVIEW
Parolee recidivism rates have risen dramatically over the past three decades (Petersilia, 2003; Lurigio, 2001), and there is also evidence that offenders are recidivating more quickly and for more serious crimes (Petersilia, 2003). Overall, about 30 percent of parolees are rearrested within the first six months, 44 percent of parolees are rearrested within their first year out of prison, and 68 percent are rearrested during their first three years out of prison (Langan and Levin, 2002). These high rates of recidivism reflect the high risks that parolees face. This chapter examines the literature on how recidivism has been defined and the individual-level risk factors of parolees. Attention is paid to the field of community research, some of the more important criticisms of this area of research, and the current research on communities in which parolees live. As this study is particularly interested in examining how parolees are affected by their community poverty, rurality, and criminal justice resources, this section of the study focuses on the current state of research knowledge in these three areas.
Defining Recidivism: Recidivism occurs when offenders, who have been released from community supervision or prison after serving their sentence, commit new crimes. One problem with measuring recidivism is that researchers must rely on official records (Travis and Visher, 2005). Self-report data from parolees on their criminal activities would be a superior measure of recidivism, yet the high costs of this form of data collection makes this option
8 unlikely. Therefore, most recidivism studies measure recidivism using official records of parolee rearrest, reconviction, probation or parole revocation, and reimprisonment (Claggion, 2008). This study measures recidivism using return to prison, which is the most conservative measure but which also contains the least measurement error (Langan and Levin, 2002). Parolees today face a number of individual and community factors that impede their successful reintegration into society. The next section discusses research on individual-level factors that make parolees more likely to recidivate during parole.
Individual-Level Risk Factors: Generally, parolees’ likelihood of returning to prison can be estimated by calculating a risk score based on such factors as criminal history, demographic characteristics (e.g., gender and race), criminogenic need (e.g., drug or alcohol addiction), social bonds (e.g., family ties), and the effects of having served time in prison (e.g., lower employment opportunities). The following section summarizes research on risk factors that have been shown to increase recidivism.
Criminal Record : The length of an offender’s criminal record and the type of conviction offense are strong predictors of recidivism (Gottfredson and Gottfredson, 1994). Specifically, parolees with longer criminal records are more likely to recidivate than parolees with shorter criminal records (Langan and Levine, 2002; Gendreau et al., 1996), although length of time served in prison is not related to recidivism (Langan and Levine, 2002). The type of crime parolees initially commit also affects their later
9 probability of recidivism. In the most recent large-scale study, conducted by the Bureau of Justice Statistics (BJS), 73.8 percent of all property offenders were rearrested within three years, compared to 61.7 percent of violent offenders and 62.2 percent of public order offenders. Drug offenders were also rearrested at high rates; within three years, 66.7 percent of drug offenders were rearrested (Langan and Levine, 2002).
Demographic Variables: There are differences in the recidivism rates of parolees by gender, race, and age. Specifically, in the BJS study, men (53 percent) were returned to prison at higher rates than women (39.4 percent), and African Americans (54.2 percent) were reimprisoned at higher rates than whites (49.9 percent) (Langan and Levine, 2002). However, it is important to note that in predictive modeling, demographic variables, particularly race and gender, are not always consistent statistical predictors of recidivism (Gottfredson and Gottfredson, 1994). In the BJS study, younger parolees were reimprisoned at higher rates than older parolees (Langan and Levine, 2002; Beck and Shipley, 1989). However, in predictive modeling, the effect of age is lessened and sometimes even nullified with the addition of other variables (Gottfredson and Gottfredson, 1994). On the other hand, the age at first official involvement in delinquency is a strong and consistent predictor of recidivism (Ashford and LeCroy, 1990). This finding suggests that while younger parolees offend more often than older parolees, their youth is not necessarily the reason for continued criminal activity.
10 Criminogenic Need: There are two types of risk factors, those that are static (e.g., demographic factors) and those that are mutable (e.g., criminogenic need). Researchers have suggested that criminogenic needs should be targeted so as to decrease recidivism (Andrews and Bonta, 1994). Gendreau’s meta-analysis (1996) found that criminogenic needs, particularly substance abuse history, are important in predicting recidivism. Moreover, the study found that criminogenic needs, rather than static characteristics (e.g., age, gender, race), were much stronger predictors of recidivism among parolees. Several studies have also supported the important role that substance abuse can play in determining success or failure in reintegrating into society (Sampson and Laub, 1993; Gottfredson and Gottfredson, 1994). It is also important to note that drug and alcohol dependency issues are widespread among prison populations. In one study of inmates, 52 percent of the inmates surveyed reported that they were under the influence of drugs or alcohol at the time they committed the crimes for which they were incarcerated (Mumola, 1999). Additionally, among first-time offenders, 40 percent reported a substance abuse problem, but among high rate repeat offenders (5 or more prior convictions), 80 percent reported a substance abuse problem (Petersilia, 2003). This finding suggests that substance abuse is not only widespread among the offender population, but that it poses a continued risk factor in determining future recidivism throughout parolees’ lifetimes.
Social Bonds: Former criminals who establish social bonds, such as marriage, are less likely than those who do not to reoffend (Fagan, 1989; Laub and Sampson, 2001). But, fewer than half of all prisoners are married. Across the general prison population,
11 just 17 percent of state prisoners and 30 percent of federal prisoners are married (Petersilia, 2003). Marriage can also provide parolees with a place to live upon their release from prison and these supportive social bonds can also help parolees find work (Solomon et al., 2001). Because of their criminal history, former inmates are often disadvantaged in their attempts to forge social bonds (Laub et al., 1998). Having served time in prison makes ex-offenders less able to form future social bonds through marriage or cohabitation (Western and McLanahan, 2000), and repeat offenders are at high risk for separation and divorce (Laub et al., 1998).
Effects of Prison: Having served time in prison increases parolees’ risk for future offending by further increasing their risk factors. For instance, studies show that serving time in prison worsens employment opportunities (Pager, 2003), chances for steady employment (Crutchfield and Pitchfork, 1997; Western and Beckett, 1999), and prospects for higher lifetime earnings (Needles, 1996). In addition, prison aggravates health and mental illness problems (Hammett et al., 2001; Lurigio, 2001), and has been linked with early death (Binswanger et al., 2007). Prison is also problematic for family relations, which suffer when a loved one is imprisoned (Clear et al., 2001), and is especially difficult when the loved one is the primary caregiver (Hagan and Coleman, 2001).
The next section examines community-level factors and how they relate to parolees. This discussion begins with the historical background of community studies
12 and how community studies are defined and then proceeds to examine some of the strengths and weaknesses of these studies. Finally, this section concludes with a discussion on where parolees live upon release from prison.
Community-Level Studies: Social scientists have long been interested in understanding the ways in which society affects individuals and in the last century, many social scientists have focused on smaller units of society, or neighborhoods (Sampson, 1987a). At the turn of the twentieth century, American cities were becoming more industrialized and urbanized and inside these cities, communities were becoming more spatially segregated by nationality, economic status, and the physical condition of neighborhood structures. The School of Human Ecology at the University of Chicago was at the forefront of community-level research (Park et al., 1925; Shaw and McKay, 1942; Wirth, 1938). Human ecology conceptualized cities as ecosystems, with people interacting with other people in their environment in much the same way animals and plants interact in their natural habitats. These human ecosystems were referred to as “natural areas” (e.g., Little Italy, downtown), or communities in which every individual plays a role in the social processes of “invasion, dominance, and secession.” More specifically, new groups of people “invade” a city area, come to “dominate” that area in numbers and cultural influence, and the old group moves out of that area, or “secedes.” These social processes of invasion, dominance, and secession tended to occur in radiating concentric zones, originating from the city center (Park et al., 1925). Neighborhood structural characteristics between these concentric zones varied considerably, especially in poverty,
13 population heterogeneity, residential instability (Shaw and McKay, 1942), and density (Wirth, 1938). By its nature, ecological research measures the effect that a geographical place has upon individuals, but the measurement of ecology can vary substantially. Neighborhood studies tend to be place-based, meaning that these studies measure the impact that a geographically bounded area has upon an individual. Communities, which can include neighborhoods, also include social boundaries. Thus, communities imply connections or a combination of shared beliefs, circumstances, priorities, relationships, and concerns (Chaskin, 1994). This study assesses the influence that communities, or counties, have upon parolees. Counties were chosen as the geographical unit of analysis because counties represent political, social, and governmental entities that can influence their residents. Additionally, counties were chosen because they are the unit of geography best suited to rural analysis (Osgood & Chambers, 2000). Although the current study examines parolees and their communities at the county-level, many of the neighborhood-level studies cited here are appropriate as they contribute greatly to our understanding of how place affects individuals. In the past twenty-five years, there has been a resurgence of community studies, most searching for elusive “neighborhood effects.” Neighborhood effects studies attempt to quantify the effect that living in a particular community has over individuals. Often, neighborhood effects are referred to as contextual effects, or direct causal relationships that communities have on their residents, although many neighborhood effects studies are also interested in social effects (e.g., social ties, collective efficacy) that mediate the relationship between communities and their residents. Statistically, neighborhood effects