Parolee concentration,parolee embeddedness,and the reciprocal relationship with crime rates: A longitudinal study of neighbourhoods and re-entry

Abstract

Drawing on recent scholarship on mass incarceration and prisoner re-entry, this study examines the reciprocal relationship between returning parolees and neighbourhood crime rates in five large cities in Texas. Besides the more common approach of counting the number of people on parole in communities (parolee concentration), we propose a novel approach for measuring people on parole by capturing their exposure in the community as parolee embeddedness (i.e., the cumulative number of days that people on parole resided in the neighbourhood). Results show that parolee concentration has a significant positive effect on both violent and property crime, but parolee embeddedness is significantly associated with reductions in violent and property crime. Our findings detect different effects depending on the measurement of people on parole and their community context, illustrating the need to better understand the dynamics of parolee re-entry in the era of mass incarceration.

Keywords

Embeddedness people on parole neighbourhood crime prisoner re-entry

The process of leaving prison and returning to free society on parole is a long-standing interest among criminologists and sociologists (Clear et al., 2001; Clear et al., 2005; La Vigne & Parthasarathy, 2005; Petersilia, 2003; Visher & Travis, 2003), particularly in the modern era of mass incarceration (Garland, 2001). A consequence of mass incarceration is the unprecedented number of formerly incarcerated people returning to society. As of 2016, an estimated 4.5 million adults, or approximately one in 55 adults, are under community supervision, in which 874,800 or 13.2% of the total U.S. correctional population are on parole (Kaeble & Cowhig, 2016). With the past four decades showing a dramatic increase in people returning to their neighbourhoods on parole, there is a need for research to better understand the neighbourhood patterns and consequences of parole for crime.

People on parole are likely to return to a relatively small number of spatially concentrated neighbourhoods (Kearns et al., 2018; Kirk, 2015), are primarily in impoverished urban areas (Cadora et al., 2003; Kirk, 2015), have many health-related issues (Baer et al., 2006; Western, 2018), and have limited employment prospects (Boessen & Hipp, 2021; Visher et al., 2011). One potential consequence of these patterns is that people coming back into communities may increase neighbourhood crime through recidivism (Hipp et al., 2010). Other research, however, documents that when people on parole go back to a community, it can be beneficial for the neighbourhood in a variety of ways, including family reunification (Braman, 2007), increasing informal social control in neighbourhoods (Hipp & Yates, 2009), and improving economic prospects for families (Braman, 2007). These factors suggest parolee re-entry may reduce crime in the neighbourhoods they are rejoining. Taken as a whole, the consequences of parole for crime patterns are unclear.

Whereas most existing research focuses on parolee concentration – that is, the number of people on parole returning to a neighbourhood – a key contribution of our study is to consider the temporal aspects of people on parole's exposure to neighbourhoods, or what we call parolee embeddedness. Parolee embeddedness in neighbourhoods involves the extent to which people on parole become integrated or (re)connected within the communities where they live. This approach is rooted in Granovetter's concept of “social embeddedness”, which is the idea that economic behaviour and institutions are heavily influenced by ongoing social relations (Granovetter, 1985). One aspect of embeddedness encompasses a person's immersion in the social environment with social relations, sense of belonging, and stability (Singh et al., 2018). An alternative perspective on the embeddedness of people on parole considers the duration of their stay in the neighbourhood, and this is the current study's focus. Our approach is distinct from nearly all prior quantitative research that focuses on the concentration of people on parole by measuring the number (or rate) of people on parole in a neighbourhood when examining the consequences of reintegration for crime or other outcomes; however, a limitation of this common approach is that it does not capture how long people on parole are in the neighbourhood. Given that many positive effects of being in a neighbourhood likely require a certain degree of neighbourhood embeddedness, we consider both where and for how long they stay in the neighbourhood, instead of simply asking where they return. We propose using the average cumulative number of days people on parole reside in the neighbourhood to capture this potential embeddedness and assess whether it affects crime beyond the concentration of people on parole in the neighbourhood.

This study uses a longitudinal design to examine the relationship between parolee concentration, parolee embeddedness, and neighbourhood crime. We use data of all people released on parole to neighbourhoods in the five largest cities of the state of Texas from 2003 to 2011: Austin, Dallas, Fort Worth, Houston, and San Antonio. As of the 2010 Census, these cities represent five of the top 13 largest cities by population in the United States of America. As the second largest state in the United States of America by area and population, Texas is also among the top states for number of people on parole. By the end of year 2016, there were 111,287 people on parole in Texas, which is the highest parolee population across states (Kaeble, 2018). Whether returning parolees affect neighbourhood crime and vice versa is a crucial empirical question for researchers and policymakers, along with the role of parolee embeddedness in these processes.

People on parole and neighbourhood crime

The impact of people on parole for neighbourhood crime

The most direct way that people on parole can impact neighbourhood crime is through recidivism. Numerous studies have shown that people on parole often do not successfully complete parole (Alper et al., 2018; Durose et al., 2014; Lynch & Sabol, 2001). For example, in one study about four in nine (44%) of released prisoners were arrested during the first year following release, and about two-thirds of the prisoners were rearrested within three years (Alper et al., 2018). One possibility is that this implies more offending, as shown in prior research that focuses on concentrated parole re-entry leading to greater recidivism (Chamberlain & Wallace, 2016). Another possibility is that the concentration of people on parole in a community may attract considerable surveillance from criminal justice agents (Seiter & Kadela, 2003), thereby increasing the probability of coming to police attention, resulting in increased chances of being arrested. This pattern might also imply that many people who experienced prison were not helped by their incarceration. This experience of incarceration did not prepare them for re-entry and likely created additional challenges for these individuals and their communities (e.g., see the literature on collateral consequences: National Research Council 2014). The implication is that it is not surprising that recidivism is a challenge post-incarceration, and that recidivism contributes to neighbourhood crime rates. Therefore, when more people return to a neighbourhood on parole, it will likely increase the level of crime (Chamberlain, 2016; Hipp & Yates, 2009; Kovandzic et al., 2004; Livingston et al., 2014; Raphael & Stoll, 2004).

Another possibility for why neighbourhoods with more people on parole might experience higher crime rates stems from the instability associated with the spatially concentrated cycling of people in and out of prison, as well as high rates of residential mobility. Although there is debate about how much residential instability is actually experienced by people on parole – with some scholars finding high levels of mobility (Harding et al., 2013) whereas others do not (La Vigne & Parthasarathy, 2005; Simes, 2019) – they likely experience more residential mobility than residents more generally. For example, people on parole have limited job prospects upon their return from prison (Boessen & Hipp, 2021), and they usually face financial hardship due to high levels of debt from legal expenses (Drakulich et al., 2012). As a result, families also bear a substantial economic burden, but seeking assistance can be challenging. From this perspective, there is reason to suspect that people on parole generally do not remain in a single location for an extended period to establish a stable residence, especially given limited help from families, social ties, or society (Seiter & Kadela, 2003).

As another form of instability, incarceration results in some residents being forcibly removed from their communities to prisons or jails, while others are released back into their communities, and this process is conceptualised as “coercive mobility” (Clear et al., 2001, 2003; Rose & Clear, 1998). This frequent mobility of people on parole presumably reduces informal social control in neighbourhoods, which in turn can lead to more recidivism and neighbourhood crime (Drakulich et al., 2012; Hipp et al., 2010). For example, Drakulich et al. (2012) found that high concentrations of returning prisoners are associated with a reduced capacity for collective efficacy and higher levels of violent crime. Similarly, a study in Tallahassee, Florida found strong evidence that prison cycling has a positive effect on crime in marginalised neighbourhoods, which largely supports the coercive mobility theory (Dhondt, 2012). Thus, a community with many residents cycling in and out of prison impedes neighbourhood stability and also feeds back into higher crime rates (Rose & Clear, 1998).

Most studies use cross-sectional data to study whether crime is more likely to occur in places with larger numbers of people on parole, and only a smaller body of literature has studied people on parole and crime in neighbourhoods in longitudinal designs (Chamberlain, 2016; Hipp & Yates, 2009). A typical strategy measures the number of people on parole released to the neighbourhood over some time period, often annually. However, due to high mobility (Clear et al., 2003) and high recidivism rates (Alper et al., 2018), simply counting the number of people on parole may not accurately capture the degree of neighbourhood exposure, especially for studies using longitudinal annual data.

Failing to account for how long people on parole reside in the neighbourhood results in an imprecise assessment of parolee impact (see also work on mobility and environmental exposure; Browning et al., 2017, 2021; Wikström, 2006). The one study of which we are aware that has addressed this issue was by Chamberlain (2016) who created a measure of the number of days each person on parole spent in the neighbourhood during the year, rather than the simple count of people on parole. Thus, instead of assuming that each person on parole equally “impacts” a neighbourhood, she weighted them by time of exposure. Her strategy addressed this methodological challenge by creating a measure of the number of days of exposure (regardless of the number of people on parole). However, she did not explicitly account for the number of people on parole in the neighbourhood. Therefore, her measure confounds the number of people on parole and the number of days they spend in the neighbourhood, which is an important distinction given our conceptual argument. In our study, we propose this conceptual idea of embeddedness and how this would affect or be affected by neighbourhood crime. Our measure is computed as the average cumulative number of days that parolees spend in the neighbourhood, while also accounting for parolee concentration. Thus, the present study is not a replication of Chamberlain’s (2016) work in Ohio, rather, our study highlights the conceptual distinction of parolee measures, and helps to move the field forward by considering how people on parole are affecting the neighbourhood based on their exposure in the neighbourhood.

Parolee embeddedness in neighbourhoods

People on parole with longer exposure to the neighbourhood are more likely to build social networks and receive support from families and friends. Thus, people on parole who stay in the neighbourhood longer are more likely to be embedded in their community and their embeddedness within it might strengthen families, their social networks, access to resources, and social support (Braman, 2007; Hipp & Yates, 2009). This parolee embeddedness might impact the neighbourhood through changes in informal social control, involvement in community organisations, increased communal solidarity, and neighbouring interactions (Lynch & Sabol, 2004). Nonetheless, it is challenging for people on parole to engage with communities when their presence is transient and of short duration.

As an example, consider a neighbourhood with 20 people on parole in a year, but these people on parole rapidly commit crimes and return to prison. As a result, these people on parole therefore do not spend much time in the neighbourhood. Consider another neighbourhood that also has 20 people on parole return in that year, but these persons all spend the entire year in that neighbourhood. Do the people on parole affect these two neighbourhoods in the same way? Do they contribute to neighbourhood crime similarly even if they stay in the neighbourhood for different lengths of time? The answer would be most likely no. As evident in the coercive mobility process, an increasing number of people on parole face abrupt removal from their neighbourhoods and are sent back to prison, whether due to parole violations or heightened supervision. One consequence is the heightened level of neighbourhood instability, which is more likely to lead to adverse neighbourhood effects, including increased crime and neighbourhood unsafety, as compared to a scenario where the same number of people on parole remain in the neighbourhood for a longer duration. The difference could simply be due to the length of exposure in the neighbourhood or could capture conceptual differences if the persons in the second neighbourhood are better able to adjust to life on the outside and therefore contribute to the neighbourhood context through their embeddedness.

For these reasons, we propose an alternative approach for measuring people on parole that supplements the traditional measure of counting the number of people on parole: the cumulative number of days people on parole are present in the neighbourhood. We refer to this as parolee embeddedness. Thus, whereas existing ecological studies of neighbourhoods measure residential population concentration and residential stability (i.e., based on length of residence), under the presumption that these capture criminal opportunities and informal social control capability, respectively, we view parolee concentration and parolee embeddedness measures analogously, except focused on a particular subpopulation.

An advantage of our parolee embeddedness measure is that it may capture the effect of non-recidivating people on parole who stay in the neighbourhood longer. The longer duration in the neighbourhood to some extent signals that people on parole are better able to adjust to life on the outside. For example, due to coercive mobility, people on parole returning to the community may have more fragmented social networks and concentrate in the same communities, thereby decreasing the resources and services available for reintegration (Fagan et al., 2002). This concentration of people on parole may therefore increase the chances of recidivism (Chamberlain & Wallace, 2016; Stahler et al., 2013). In contrast, parolee embeddedness – people on parole who are embedded in the neighbourhood longer – may indicate reuniting supportive relationships that create more cohesion, which could result in less neighbourhood crime. Whereas work on coercive mobility typically uses the number of people on parole to capture churning (Chamberlain, 2016; Hipp & Yates, 2009), it may be that the cumulative length of time in the community is more important. Even though we claim that people who stay longer in the neighbourhood are better embedded, the operationalisation of our parolee embeddedness measure is still a proxy for the connections that one can have when living in a neighbourhood longer. As such, we argue that time spent in the neighbourhood may be a conceptually interesting process to measure because it captures embeddedness in the neighbourhood, to some extent.¹ Thus, to better capture embeddedness we utilise the novel measure of cumulative number of days people on parole live in the neighbourhood. We assess whether this measure better explains the parolee–crime association through our analyses while controlling for parolee concentration.

Non-linear effect of people on parole

The consequences of parolee concentration and parolee embeddedness for neighbourhood crime are likely non-linear. Regarding parolee concentration, one possibility considers the consistent evidence of greater concentration being associated with more neighbourhood crime. A non-linear possibility suggests diminishing returns. At smaller concentrations the people on parole in a neighbourhood may have an association with more crime, but as the size of this parolee concentration increases, this positive relationship may diminish as the neighbourhood accumulates more social and human capital, strengthening the community's ability and efficacy to self-regulate. The implication is a non-linear relationship between parolee concentration and neighbourhood crime.

Likewise, there is reason to expect a non-linear relationship between parolee embeddedness and crime. Here we expect reductions in crime associated with their initial reintegration where social ties are rekindled upon release, but as people are integrated longer into the community these initial reductions in crime might wane. This implies a non-linear negative association in which the relationship between embeddedness and crime flattens out at higher levels. Accordingly, crime may decrease due to the immediate effect of this group's initial adjustment to their reintegration, but as people on parole become more embedded in their communities, their beneficial effect on crime may flatten out. This is because this embeddedness is expected to increase social ties and cohesion in the neighbourhood initially, but at some point, there is less ability to add additional social ties as one stays longer in the neighbourhood. Thus, we hypothesise a decreasing negative relationship between parolee embeddedness and neighbourhood crime.

Whereas parolee concentration and parolee embeddedness likely independently impact neighbourhood crime, there are theoretical reasons to expect that they may multiplicatively interact with one another in their relationships with crime. Whereas the literature has consistently documented that neighbourhoods with more people on parole are associated with higher crime rates (Chamberlain & Boggess, 2018; Hipp & Yates, 2009), it may be that parolee embeddedness is most important, and most effective, in neighbourhoods with greater parolee concentration, possibly due in part to the benefits of agglomeration and longer term parolees’ connection to other institutions and people on parole (e.g., at nonprofits). Thus, the cohesion in the neighbourhood that parolee embeddedness engenders may be strongest in a neighbourhood with a high parolee concentration. This implies an interaction between parolee embeddedness and parolee concentration in which the strongest negative relationship occurs for parolee embeddedness in neighbourhoods with high parolee concentration.

The impact of neighbourhood crime on parolee re-entry

In the current study, we use a longitudinal study design to explore our research questions. Although our primary focus is not which neighbourhoods people on parole return to, we nonetheless build on work by Chamberlain (2016) and account for this possible endogeneity in our longitudinal models. Whereas Chamberlain (2016) focused on how parolees might impact a broader range of neighbourhood factors – including residential instability, vacancies, and economic disadvantage – we focus only on how people on parole impact crime levels but expand the focus by also accounting for our novel measure of parolee embeddedness. A well-known pattern is that people on parole generally return to neighbourhoods with similar characteristics as their home neighbourhood before prison, and they are likely to be pushed into disadvantaged neighbourhoods with high crime rates (Harding et al., 2013). Although people on parole may wish to avoid high crime areas as they likely experience more criminal justice activities in those spaces (i.e., searches of residences, enhanced supervision; Seiter & Kadela, 2003), they likely have limited economic ability to avoid them. A common phenomenon is that people on parole are predominantly concentrated in impoverished urban areas (Cadora et al., 2003; La Vigne et al., 2005) and reside in disadvantaged neighbourhoods with little support for prisoner re-entry (La Vigne et al., 2004; Seiter & Kadela, 2003). This push of people on parole into high crime areas is in part about being priced out of other neighbourhoods, but also arguably due to the isolation/exclusion of other people outside the neighbourhood.

Data and methodology

Data

The data for this study come from all people on parole released in Texas from 2003 to 2011. Data were obtained directly from the Texas Department of Criminal Justice (TDCJ). The data provide information on when people started on parole and when they ended. People on parole were followed until the end of their parole (revoked or discharged) or until July 2012, which is the date when the data collection ended. The TDCJ also tracked where people on parole resided after release and we geocoded their home addresses using Google and ArcGIS ArcMap 10.6. Nearly all people on parole reported an address, and about 88% of unique addresses were geocoded to an exact X–Y coordinate and joined to the appropriate census block. These 88% of unique addresses are then matched to the five cities with crime data in Texas in the current study. For crime data, we obtained yearly data from the five largest cities in Texas: Austin, Dallas, Fort Worth, Houston, and San Antonio. Annual crime data cover Part I crimes, and we geocode the addresses of the crime events to X–Y coordinates and aggregate them to census blocks. Overall, the geocoding match rate is quite high, ranging from 97.6% in Houston to 99.3% in San Antonio.

The average number of people on parole per year is 14,920 for Austin, 63,737 for Dallas, 36,142 for Fort Worth, 124,611 for Houston, and 44,983 for San Antonio. The percentage of male people on parole is similar for these five cities, but Austin (31.7%) and Fort Worth (30.6%) have the highest percent of white people on parole, Dallas (66.9%) and Houston (64.2%) have the highest percent of Black people on parole, and San Antonio (70.5%) has the highest percent of Latino people on parole. The majority of people on parole fall into the age categories of 35–49 and 50–64 years old, and there is no notable difference between each city. More summary information about the data of those released on parole is shown in Appendix Table A1, including number of people on parole, gender, race, married or not, and different age categories, averaged over the nine year period.²

Existing studies examining prisoner re-entry have focused on the state level (Hannon & DeFina, 2010; Kovandzic et al., 2004) or the tract level (Lee et al., 2017). Studies on the parolee–crime relationship at smaller geographic scales have typically used census-defined tracts (Kubrin & Stewart, 2006) or block groups (Chamberlain, 2016) to measure neighbourhoods. However, a limitation of census-defined geographic units is that they typically do not capture the normal activity space of persons (Golledge & Stimson, 1997). For example, a person living near the boundary of a census-defined unit will typically spend much more time in adjacent units rather than their own tract (Jones & Pebley, 2014), and therefore using census-defined units inappropriately fails to capture these processes. For this reason, the strategy we adopt uses egohoods as our measure of neighbourhoods to better take into account the spatial patterns of local daily routine of people on parole (Hipp & Boessen, 2013). Egohoods are built on the insight that the spatial patterns of individuals’ social lives tend to exhibit a distance decay function, rather than being confined to their own census-defined unit. Furthermore, the egohoods approach captures the spatial patterns of persons’ local daily routine activities and how that can give rise to crime in the built and social environment around a small geographic unit. Indeed, Hipp and Boessen (2013) in their study showed that crime in ecological units was much better explained by aggregating typical measures to egohoods rather than census-defined units. For these reasons, we argue that egohoods are a more effective measure of neighbourhood for understanding the relationship between returning people on parole and neighbourhood crime, and better address the modifiable areal unit problem (Openshaw & Taylor, 1981). Thus, the units of analysis in the current study are egohoods in these cities, where an egohood is a census block and all blocks surrounding it within a quarter mile radius.³ Overall, there are a total of 83,836 egohoods with crime data in the five largest cities of Texas.

To measure neighbourhood characteristics, we combine several datasets with these parole and crime data, and all data are harmonised to egohoods in 2010 census block boundaries. First, we capture business information with ReferenceUSA Historical business data from Infogroup. ReferenceUSA is an annual dataset that contains geographic information allowing us to locate businesses at the address level each year. Second, as community voluntary organisations may help a neighbourhood reintegrate people on parole, we measure the presence of voluntary organisations using annual data come from the National Center for Charitable Statistics, which contains information on exempt organisations from the Internal Revenue Service's Business Master File.⁴ We geocode these organisations based on their provided address and place them into the appropriate census block.⁵ Finally, we use 2000 U.S. census data to capture neighbourhood sociodemographic information.

Outcome measures

Neighbourhood crime rates

We use crime rates per 10,000 population in a quarter-mile egohood per year as our crime measure. Crime events are aggregated into yearly totals for violent crime (homicide, robbery, and aggravated assault) and property crime (burglary, larceny, and motor vehicle theft). All crime rate variables are log-transformed to account for the skewed distribution.

People on parole

To examine the relationship between the presence of people on parole returning to the community and neighbourhood crime, we compute the number of people on parole (i.e., parolee concentration, log-transformed, after adding 1, to address the skewed distribution) residing in a particular egohood in a given year. In this measure, anyone who has stayed in the neighbourhood in that year, regardless of duration, is counted as one parolee. We also construct a measure of parolee embeddedness, the average cumulative parolee days in an egohood until released from parole. This was calculated by taking the cumulative number of days people on parole have resided in a particular egohood based on the start date and end date of people on parole's status and then dividing by the yearly total number of people on parole.⁶ The correlation between parolee concentration and parolee embeddedness varies year by year, ranging from 0.38 to 0.57, which does not pose collinearity issues, especially given the large sample size (see O’Brien, 2007) and highlights the conceptual distinctness of these measures. To capture non-linear effects, we include the squared terms of parolee concentration and parolee embeddedness. We also compute the interaction between parolee concentration and parolee embeddedness and their quadratic measures.

Exogenous variables

Socio-demographic variables

Several measures from the 2000 census are included in the models to capture neighbourhood characteristics. For census measures that are not available in blocks, we first impute them from block groups or tracts to blocks using the synthetic estimation of ecological inference strategy and then construct the egohood measures based on the imputed block values.⁷ Residential stability is measured with a standardised factor score from a factor analysis. Three variables are combined: average length of residence, percent of households that moved into their residence within the last five years (which loads negatively), and percent homeowners. Concentrated disadvantage is also measured as a factor score from a factor analysis using five variables: percent of residents below poverty, percent unemployed, percent single-parent households, average home value, and average household income (these last two have negative loadings).⁸ Racial/ethnic composition of neighbourhoods is measured as percent Black and percent Latino. We capture the racial/ethnic heterogeneity of the neighbourhood with a Herfindahl index of five racial/ethnic groups (Asian, Black, Latino, White, and other race). To capture inequality, we use household income to construct the Gini inequality index.⁹ We also include percent immigrants in the neighbourhood with a measure of percent foreign born. We construct a measure of young people (percent individuals aged 16–29) in the egohood. We also control for population in the egohood, which is implicitly population density given that egohoods have a constant size.

Business and organisation variables

First, research shows that nearby jobs are a key predictor of people on parole's successful reintegration (Boessen & Hipp, 2021), and prior work also suggests that businesses (i.e., land uses) are key predictors of neighbourhood crime (Boessen & Hipp, 2015). Accordingly, we measure the number of (1) total employees in all businesses, (2) retail employees, (3) recreation employees, and (4) food employees, which come from the Reference USA Historical data. We log transform these measures given that this better captures the empirical relationships. Second, voluntary organisations help people on parole reintegrate with the community (Hipp, Petersilia, & Turner 2010), and we measure organisations using data from National Center for Charitable Statistics. Using the National Taxonomy of Exempt Entities codes that are provided by the organisations, we compute the total service voluntary organisations (log transformed, after adding 1), and these organisations include mental health services (e.g., mental health treatment, alcohol/drug abuse treatment), crime (e.g., delinquency prevention, crime prevention), care (e.g., rehabilitation services, transitional care), abuse (e.g., spouse abuse, child abuse), legal (e.g., legal services, public interest law), vocational (e.g., training, job procurement assistance), food (e.g., food banks, nutrition programs), recreational (e.g., community recreation centres, recreation clubs), and neighbourhood (e.g., block clubs, community coalitions).¹⁰ These organisation variables are captured year by year. Note that we are not able to determine which of these organisations provide services specifically to persons on parole, the services they do provide would arguably be of use and importance to such persons.

Analytic strategy

To account for possible endogeneity between neighbourhood crime and people on parole, we employ longitudinal Structural Equation Models (SEM) in our analyses. Specifically, using longitudinal data on neighbourhoods, we estimate a series of cross-lagged equation models, a procedure that allows us to account for temporal autocorrelation in the residuals. The models are estimated using city fixed effects (by centring the data by city), and we calculate robust standard errors. Additionally, we constrain the coefficients of the variables to be equal over waves and test the consequences of this constraint for model fit (Hipp et al., 2009; Hipp & Wickes, 2017). We model one year lags given that crime responding to parolee re-entry likely requires a year to capture the effect. Thus, our model specifies that the presence of people on parole in a prior year affects neighbourhood crime in the current year while taking into account the one year lag of crime and controlling for a variety of additional neighbourhood-level factors. We also assess whether crime in the prior year predicts parolee concentration and parolee embeddedness in the next year, while also controlling for the same neighbourhood factors. We include all these socio-demographic measures at the year 2003 time point of the 2000 Census.¹¹ The theoretical model is depicted in Figure 1.

Figure 1.
Path model depicting the reciprocal relationship between parolees and neighbourhood crime controlling for neighbourhood measures.

For each outcome, the cross-lagged models are estimated using the following equations:
$C r i m e_{it} = β_{10 t} + β_{11 t} C r i m e_{i (t - 1)} + β_{12 t} C o n c e n t r a t i o n m e a s u r e s_{i (t - 1)} + β_{13 t} E m b e d d e d n e s s m e a s u r e s_{i (t - 1)} + β_{14 t} X_{i (t - 1)} + β_{15 t} Z_{i} + ζ_{1 it}$
(1)
$P a r o l e e c o n c e n t r a t i o n_{it} = β_{20 t} + β_{21 t} C r i m e m e a s u r e s_{i (t - 1)} + β_{22 t} P a r o l e e c o n c e n t r a t i o n_{i (t - 1)} + β_{23 t} X_{i (t - 1)} + β_{24 t} Z_{i} + ζ_{2 it}$
(2)
$P a r o l e e e m b e d d e d n e s s_{it} = β_{30 t} + β_{31 t} C r i m e m e a s u r e s_{i (t - 1)} + β_{32 t} P a r o l e e e m b e d d e d n e s s_{i (t - 1)} + β_{33 t} X_{i (t - 1)} + β_{34 t} Z_{i} + ζ_{3 it}$
(3)
where Crime_it is the outcome variable of logged neighbourhood-level violent (or property) crime rate in year t. The outcomes Parolee concentration and Parolee embeddedness are the number (logged) and average cumulative days of people on parole for egohood i which are measured at time t, respectively. In equation (1), $β_{10 t}$ is the intercept, $β_{11 t}$ is a vector of parameters that captures the crime measure from the previous time point (t–1), in which we have constrained the effect for different waves to be equal. We include parolee concentration in egohood i at time point (t–1) with coefficients in the $β_{12 t}$ vector. We also include parolee embeddedness in egohood i at time point (t–1) with coefficients in the $β_{13 t}$ vector. Here, concentration measures include parolee concentration and parolee concentration squared; and embeddedness measures include parolee embeddedness and parolee embeddedness squared. X_i(t−1) is a matrix of organisation variables and the constrained equal coefficients for different waves are captured in $β_{14 t}$ vector. Z_i is a matrix of several neighbourhood-level socio-demographic variables at the first time point of year 2003 with the accompanying coefficients captured in $β_{15 t}$ vector, where the coefficients for different waves are also constrained to be equal.¹² Equations (2) and (3) are the ones with parolee concentration and parolee embeddedness as outcomes; we do not have parolee concentration or parolee embeddedness predict each other. Here crime measures include crime and crime squared. Finally, $ζ$ _1it, $ζ$ _2it and $ζ$ _3it are error terms with assumed normal distributions. We account for temporal autocorrelation by estimating covariances between the errors of crime at time t–1 and time t, and those between people on parole at time t–1 and time t. We also estimate the covariances between the errors of crime and the parole measures at the same time point.

We estimate a series of SEM in Stata 15.0 (StataCorp, College Station, TX) and use full information maximum likelihood to handle missing data (Allison, 2012).¹³ Table 1 presents the summary statistics of the variables used in the analyses.

Table 1.
Summary statistics of variables used in analyses.

Violent crime rate (logged) Property crime rate (logged) Parolee concentration (Number of parolees logged) Parolee embeddedness (cumulative days of parolees/1000)

Mean SD Mean SD Mean SD Mean SD

2003 3.57 1.83 6.04 1.15 0.81 0.75 0.25 0.28

2004 3.55 1.83 6.05 1.11 1.04 0.83 0.37 0.33

2005 3.52 1.83 6.02 1.09 1.17 0.87 0.46 0.38

2006 3.44 1.82 5.95 1.05 1.12 0.86 0.43 0.42

2007 3.48 1.82 6.02 1.04 1.06 0.85 0.41 0.44

2008 3.38 1.81 5.91 1.02 1.07 0.85 0.37 0.44

2009 3.39 1.80 6.03 1.00 1.09 0.87 0.36 0.42

2010 3.35 1.76 5.97 1.01 1.09 0.87 0.35 0.42

2011 3.23 1.77 5.91 1.01 1.08 0.86 0.34 0.43

Average 3.43 1.81 5.99 1.05 1.06 0.84 0.37 0.40

Control variables

Concentrated disadvantage 0.39 10.67

Residential stability 0.02 0.69

Income inequality 0.86 0.12

Racial/ethnic heterogeneity 0.41 0.19

Percent immigrants 0.20 0.13

Percent aged 16–29 0.24 0.10

Percent Black 0.19 0.27

Percent Latino 0.39 0.30

Population density 0.10 0.08

Annual measures – average over years

Total employees (logged) 5.47 1.66

Retail employees (logged) 2.78 1.77

Recreation employees (logged) 0.69 1.18

Food employees (logged) 2.13 2.07

Total voluntary organisations (logged) 0.19 0.40

Note. N = 83,836 egohoods.

Results

Relationship between returning parolees and crime

To examine the cross-lagged relationship between people on parole and crime, Table 2 presents a series of SEMs with outcome variables of violent/property crime and parolee concentration and parolee embeddedness controlling for a variety of socio-demographic and organisation measures. We discuss the result of people on parole's effect on crime – either violent crime or property crime – first, and then discuss the effects of neighbourhood crime on people on parole.

Table 2.
Reciprocal relationship between parolee concentration/embeddedness in egohoods and crime rates (logged) with fixed effects, cross-lagged regression models for five largest cities in Texas, 2003–2011.

Model 1: Violent crime model Model 2: Property crime model

Violent crime Parolee concentration Parolee embeddedness Property crime Parolee concentration Parolee embeddedness

Violent/property crime (logged) 0.865(0.002) 0.015(0.000) 0.003(0.000) 0.932(0.002) 0.002(0.001) 0.001(0.000)

Parolee concentration 0.058(0.002) 0.801(0.002) 0.021(0.005) 0.805(0.002)

Parolee embeddedness −0.032(0.004) 0.771(0.002) −0.044(0.013) 0.771(0.002)

Violent/property crime (logged)² −0.002(0.000) −0.002(0.000) −0.004(0.000) −0.002(0.000)

Parolee concentration² −0.013(0.001) −0.009(0.002)

Parolee embeddedness² 0.022(0.004) 0.036(0.011)

Concentrated disadvantage 0.004(0.000) 0.002(0.000) 0.001(0.000) 0.001(0.000) 0.003(0.000) 0.001(0.000)

Residential stability −0.016(0.001) 0.018(0.001) 0.015(0.001) −0.001(0.001) 0.016(0.001) 0.014(0.001)

Income inequality 0.058(0.006) 0.075(0.004) 0.044(0.003) 0.057(0.004) 0.085(0.004) 0.044(0.003)

Racial/ethnic heterogeneity 0.060(0.004) 0.030(0.003) 0.050(0.002) 0.009(0.003) 0.041(0.003) 0.055(0.002)

Percent immigrants 0.047(0.009) −0.147*(0.006) −0.033***(0.004) −0.013(0.006) −0.144(0.006) −0.034(0.004)

Percent aged 16–29 −0.042(0.010) −0.156(0.007) −0.058(0.004) −0.051(0.007) −0.154(0.007) −0.049(0.004)

Percent Black 0.273(0.007) 0.342(0.004) 0.078(0.003) 0.037(0.004) 0.372(0.004) 0.085(0.003)

Percent Latino 0.167(0.007) 0.221*(0.004) 0.083***(0.003) 0.009(0.004) 0.245(0.004) 0.091(0.003)

Population density −0.179(0.010) 0.619(0.016) 0.121(0.005) −0.195(0.009) 0.593(0.016) 0.124(0.006)

Total employees (logged) 0.014(0.001) 0.000(0.000) 0.002(0.000) 0.000(0.001) 0.002(0.000) 0.002(0.000)

Retail employees (logged) 0.012(0.001) −0.003(0.000) −0.000(0.000) 0.006(0.000) −0.002(0.000) 0.000(0.000)

Recreation employees (logged) 0.007(0.001) 0.001(0.000) 0.001(0.000) 0.004(0.000) 0.002(0.000) 0.001(0.000)

Food employees (logged) 0.019(0.001) −0.002(0.000) 0.001(0.000) 0.008(0.000) 0.001(0.000) 0.001(0.000)

Total voluntary organisation (logged) 0.015(0.001) 0.022(0.001) 0.003(0.001) 0.013(0.001) 0.024(0.001) 0.003(0.001)

Intercept 0.004(0.004) 0.005(0.001) 0.006(0.001) 0.006(0.002) 0.005(0.001) 0.003(0.001)

N 83,836 83,836

Note. Standard errors in parentheses. p < .05. p < .01. p < .001.

Starting with violent neighbourhood crime as an outcome, unsurprisingly, the one year lag of violent crime rates has a strong positive effect on violent crime rates in the following year, indicating considerable stability in the level of crime from year to year. In Model 1 of Table 2, the coefficient of lagged violent crime is 0.865, indicating that a 1% increase in violent crime rates in the prior year is expected to increase violent crime rates in the current year by around 0.87%, holding other variables constant. For ease of interpretation, the non-linear results of parolee concentration and embeddedness are presented in Figures 2 and 3. Note that in these figures, we are plotting parolee concentration or embeddedness from one standard deviation below the mean to one standard deviation above the mean. The fixed effects models centre the variables within each city, and therefore the zero value indicates the average level of parolee concentration (or embeddedness) in a particular city, with negative values showing egohoods with below average levels of parolee concentration (or embeddedness). There is a positive relationship between parolee concentration and violent/property crime. Specifically, we used the plotted expected values from the non-linear relationship to find that for an egohood with no parolee concentration, a one standard deviation increase in parolee concentration is associated with 5.6% more violent crime and 2.3% more property crime the next year; if the egohood had an average parolee concentration level, then a one standard deviation increase is associated with 3.9% more violent crime and 1.1% more property crime the next year. As we can see in Figure 2, the effect of parolee concentration on violent crime is larger than that on property crime. However, in Figure 3, we see that our parolee embeddedness measure yields a different relationship with crime compared to the concentration measure: the cumulative parolee days residing in the egohood in one year has a significant negative relationship with both violent and property crime rates the following year. A one standard deviation increase in parolee embeddedness is associated with about a 1%–2% decline in violent or property crime.

Figure 2.
Effect of parolee concentration on violent and property crime in next year.

Figure 3.
Effect of parolee embeddedness on violent crime and property crime in next year.

Regarding the control variables, we see that a one standard deviation increase in concentrated disadvantage is associated with 4.4% more violent crime and 1.1% more property crime, a similar increase in residential stability is associated with about 1.1% less violent crime, and a similar increase in racial/ethnic heterogeneity is associated with 1.2% and 0.2% more violent and property crime, respectively. Thus, these measures are of a similar magnitude, or even smaller, than the parolee measures.

Turning to the effect of crime on people on parole, the results in Table 2 also highlight strong evidence that the violent or property crime rate affect parolee concentration the following year, and modestly impact parolee embeddedness. A one standard deviation increase in the violent crime rate in an egohood in one year is associated with an approximately 3.3% increase in the number of people on parole the following year (though this is somewhat weaker at higher violent crime rates). A similar increase in property crime has a smaller effect, resulting in a 1.6% increase in people on parole the following year for neighbourhoods increasing from low property crime levels; this effect weakens and reverses at high property crime levels. This suggests that people on parole tend to return to neighbourhoods with more neighbourhood crime, particularly violent crime. There is only a very modest relationship between violent or property crime and parolee embeddedness.

Other neighbourhood characteristics also impact parolee concentration and parolee embeddedness. As expected, a one standard deviation increase in concentrated disadvantage is associated with 2.6% more people on parole in the egohood; however, opposite our expectations, higher concentrated disadvantage results in more parolee embeddedness – about four more days, on average, the following year – holding other variables constant. One of the strongest effects is for percent Black in the egohood: however, although a one standard deviation increase is associated with almost 10% more people on parole the following year in such neighbourhoods, it is also associated with greater parolee embeddedness (about 12 more days spent in the neighbourhood, on average, the following year). As expected, greater residential stability in the neighbourhood has a relatively strong positive relationship with parolee embeddedness the following year. Furthermore, consistent with the expectation that services attract people on parole as residents, there is evidence that service voluntary organisations in the neighbourhood are associated with greater parolee concentration and parolee embeddedness the next year.

Interaction effects

We also observe a substantial moderating effect of parolee concentration and parolee embeddedness for property crime, but not for violent crime. As shown in Figure 4, the negative relationship between parolee embeddedness and property crime is much stronger in egohoods with greater parolee concentration compared to those with few people on parole. Thus, the combination of people on parole along with parolee embeddedness appears particularly beneficial regarding property crime for neighbourhoods with more concentration.

Figure 4.
Effect of parolee concentration and parolee embeddedness on property crime in next year.

Sensitivity analyses

Given that our analyses combined five separate cities, we estimated ancillary models for each city separately to assess the robustness of the results.¹⁴ Among these five cities in Texas, the positive relationship for parolee concentration was present for both violent crime and property crime for all cities with the exception of San Antonio for property crime, which was non-significant. The parolee embeddedness measure showed a consistent non-linear negative relationship with violent crime for all cities, and for property crime in all cities except Austin. There was also a consistent positive relationship between violent crime rates and where people on parole relocated across these cities. Thus, the pattern of results was generally consistent across these cities, although the non-significant relationship between parolee embeddedness and crime in Austin suggests that future research will want to assess this measure across different cities.

Discussion and conclusion

Although crime rates and prison populations have been mostly on a downward trend over the last decade in the United States of America, the accumulated prison population is still massive, and there are increasing numbers of people being released back to communities. Moving beyond the traditional measure of parolee concentration, this study considers parolee embeddedness, which is the amount of time people on parole live in the neighbourhood after returning from prison. We also test whether there are potential non-linear effects of parolee concentration and parolee embeddedness on neighbourhood crime rates in our longitudinal analyses of the five largest cities in Texas. Key findings are discussed below.

Overall, our study provides evidence that parolee concentration contributes to neighbourhood crime: a greater number of people on parole living in a neighbourhood in one year is associated with higher violent and property crime rates the following year. This finding is in line with a longitudinal study in Sacramento with monthly crime rates (Hipp & Yates, 2009), studies in Cincinnati and Columbus (Chamberlain, 2012) and Cleveland (Chamberlain, 2016; Chamberlain & Boggess, 2018) with annual crime rates, cross-sectional studies in Seattle (Drakulich et al., 2012), and Multnomah County, Oregon (Kubrin & Stewart, 2006). This body of literature has focused on neighbourhoods within cities or counties across different states, including California, Ohio, Michigan, Washington, and Oregon. We also find similar evidence in the five Texas cities captured here – Austin, Dallas, Fort Worth, Houston, and San Antonio. Although some research argues that the impact of prison releases on crime differs across various state-level parole systems (Raphael & Stoll, 2004), our study shows a similar pattern compared to other locations. That said, these effects are relatively small, which may suggest that the considerable attention paid to re-entry of people on parole is overblown, particularly in comparison to other issues that communities face.

Nonetheless, a key contribution of this study is introducing a new way to conceptualise the effect of people on parole – the average cumulative number of days that people on parole resided in the neighbourhood – and this measure of parolee embeddedness exhibits interesting results. Neighbourhoods with people on parole who reside longer in their neighbourhood (i.e., longer exposure) experience lower violent and property crime rates. This negative relationship suggests that people returning to and staying in the communities might actually help reduce neighbourhood crime. It is possible that people on parole who stay in the neighbourhood longer are those ex-offenders who have a low recidivism rate. Another possibility is that people on parole who stay in the neighbourhood longer are better able to integrate into the community and return to a prosocial life trajectory, which can help enhance social ties in the neighbourhood and further decrease neighbourhood crime rates (Clear et al., 2003). Still, more research is needed for understanding the dynamic relationship between people on parole and crime.

Future research might not simply examine how long people on parole are embedded into the community, but how various mechanisms of re-entry (access to social support networks; changes in the economy, recidivism, access to resources, whether someone was on parole or not, etc.) connect with length of time spent in the neighbourhood. While we have shown that being in the community longer is typically associated with lower crime rates, it is not clear why this is occurring. Depending on the mechanism, it is also not always clear that longer is better, and there may be differences in the association between parolee embeddedness and crime. As such, this project has been a necessary first step to using an alternative measure to capture re-entry, and future research will want to connect it more explicitly to various mechanisms.

Neighbourhood crime rates are not simply affected by only the concentration of people on parole or the embeddedness of this group of people, rather, both concentration and duration matter. We find evidence of a moderating effect in that parolee embeddedness along with larger numbers of people concentrated on parole particularly benefits neighbourhoods and resulting in greater reductions in property crime. Accordingly, this pattern indicates that embeddedness is most important in communities that have average or even above-average concentrations of people on parole. As such, duration in a hot spot of people on parole suggests that the agglomeration of people, place, and networks of support matter for reducing crime rates and reintegration.

Despite the uniqueness of our data and the importance of our findings, certain limitations deserve to be acknowledged. Although we proposed this novel measure of parolee embeddedness, our measure is primarily a proxy of embeddedness, rather than directly capturing the connections and duration of ties people on parole may have in the neighbourhood. Relatedly, while we argued that egohoods better capture the spatial patterns of residents compared to other neighbourhood measures, they are not able to perfectly capture social interactions. An additional limitation is that we did not have annual measures of the sociodemographic variables, but simply had cross-sectional measures. Given Chamberlain's (2016) work showing that parolees are pushed into disadvantaged neighbourhoods, this is a limitation (Clear, 2009; Morenoff & Harding, 2014). For the present study, people on parole were tracked to their current address after release from prison, however, we do not have information on where people on parole lived before they were sent to prison. Due to this data limitation, we cannot tell whether people on parole cluster in neighbourhoods in which they previously lived, or if they geographically disperse. We also emphasise that our approach here is not designed to encourage extending the periods in which people are under parole supervision, but to instead argue that the measurement of re-entry is important, and that people need time to readjust to life on the outside.

The findings from this study indicate that when people on parole are in their communities longer, there tends to be some reductions in violent and property crime. As such, one implication is that minor infractions or technical violations of parole that disrupt and often remove people from the community (i.e., a stint in jail or prison) may be counter-productive to helping people re-enter into the community. In some ways, this logic is akin to some school discipline policies (i.e., expelling vs. supporting and keeping youth within the school). In other words, polices that suspend or expel people from their community for minor infractions may ultimately hinder people on parole's ability to successfully complete parole because it will likely disrupt their lives and support in considerable ways, and even have negative consequences for the neighbourhood.

In closing, our research advances the understanding of prison re-entry and neighbourhood crime across multiple dimensions. We advance the literature of people on parole and neighbourhood crime by introducing a novel concept – parolee embeddedness – and examining how the cumulative amount of time people on parole reside in the neighbourhood shapes the dynamics of parolee re-entry and neighbourhood crime in an era of mass incarceration. In addition, we extend previous research on the effect of people on parole on neighbourhood crime by also examining the non-linear effect of parolee concentration and parolee embeddedness patterns. Our new approach to measuring people on parole augments the strategy of simply counting the number of people on parole in a neighbourhood, and we believe it may provide new insights for researchers interested in neighbourhood and parole processes.

	Violent crime rate (logged)	Property crime rate (logged)	Parolee concentration (Number of parolees logged)	Parolee embeddedness (cumulative days of parolees/1000)
2003	3.57	1.83	6.04	1.15	0.81	0.75	0.25	0.28
2004	3.55	1.83	6.05	1.11	1.04	0.83	0.37	0.33
2005	3.52	1.83	6.02	1.09	1.17	0.87	0.46	0.38
2006	3.44	1.82	5.95	1.05	1.12	0.86	0.43	0.42
2007	3.48	1.82	6.02	1.04	1.06	0.85	0.41	0.44
2008	3.38	1.81	5.91	1.02	1.07	0.85	0.37	0.44
2009	3.39	1.80	6.03	1.00	1.09	0.87	0.36	0.42
2010	3.35	1.76	5.97	1.01	1.09	0.87	0.35	0.42
2011	3.23	1.77	5.91	1.01	1.08	0.86	0.34	0.43
Average	3.43	1.81	5.99	1.05	1.06	0.84	0.37	0.40
Control variables
Concentrated disadvantage	0.39	10.67
Residential stability	0.02	0.69
Income inequality	0.86	0.12
Racial/ethnic heterogeneity	0.41	0.19
Percent immigrants	0.20	0.13
Percent aged 16–29	0.24	0.10
Percent Black	0.19	0.27
Percent Latino	0.39	0.30
Population density	0.10	0.08
Annual measures – average over years
Total employees (logged)	5.47	1.66
Retail employees (logged)	2.78	1.77
Recreation employees (logged)	0.69	1.18
Food employees (logged)	2.13	2.07
Total voluntary organisations (logged)	0.19	0.40

	Model 1: Violent crime model	Model 2: Property crime model
Violent/property crime (logged)	0.865***(0.002)	0.015***(0.000)	0.003***(0.000)	0.932***(0.002)	0.002**(0.001)	0.001(0.000)
Parolee concentration	0.058***(0.002)	0.801***(0.002)		0.021***(0.005)	0.805***(0.002)
Parolee embeddedness	−0.032***(0.004)		0.771***(0.002)	−0.044***(0.013)		0.771***(0.002)
Violent/property crime (logged)²		−0.002***(0.000)	−0.002***(0.000)		−0.004***(0.000)	−0.002***(0.000)
Parolee concentration²	−0.013***(0.001)			−0.009***(0.002)
Parolee embeddedness²	0.022***(0.004)			0.036***(0.011)
Concentrated disadvantage	0.004***(0.000)	0.002***(0.000)	0.001***(0.000)	0.001***(0.000)	0.003***(0.000)	0.001***(0.000)
Residential stability	−0.016***(0.001)	0.018***(0.001)	0.015***(0.001)	−0.001(0.001)	0.016***(0.001)	0.014***(0.001)
Income inequality	0.058***(0.006)	0.075***(0.004)	0.044***(0.003)	0.057***(0.004)	0.085***(0.004)	0.044***(0.003)
Racial/ethnic heterogeneity	0.060***(0.004)	0.030***(0.003)	0.050***(0.002)	0.009***(0.003)	0.041***(0.003)	0.055***(0.002)
Percent immigrants	0.047***(0.009)	−0.147***(0.006)	−0.033***(0.004)	−0.013*(0.006)	−0.144***(0.006)	−0.034***(0.004)
Percent aged 16–29	−0.042***(0.010)	−0.156***(0.007)	−0.058***(0.004)	−0.051***(0.007)	−0.154***(0.007)	−0.049***(0.004)
Percent Black	0.273***(0.007)	0.342***(0.004)	0.078***(0.003)	0.037***(0.004)	0.372***(0.004)	0.085***(0.003)
Percent Latino	0.167***(0.007)	0.221***(0.004)	0.083***(0.003)	0.009*(0.004)	0.245***(0.004)	0.091***(0.003)
Population density	−0.179***(0.010)	0.619***(0.016)	0.121***(0.005)	−0.195***(0.009)	0.593***(0.016)	0.124***(0.006)
Total employees (logged)	0.014***(0.001)	0.000(0.000)	0.002***(0.000)	0.000(0.001)	0.002***(0.000)	0.002***(0.000)
Retail employees (logged)	0.012***(0.001)	−0.003***(0.000)	−0.000(0.000)	0.006***(0.000)	−0.002***(0.000)	0.000(0.000)
Recreation employees (logged)	0.007***(0.001)	0.001(0.000)	0.001***(0.000)	0.004***(0.000)	0.002***(0.000)	0.001***(0.000)
Food employees (logged)	0.019***(0.001)	−0.002***(0.000)	0.001***(0.000)	0.008***(0.000)	0.001**(0.000)	0.001***(0.000)
Total voluntary organisation (logged)	0.015***(0.001)	0.022***(0.001)	0.003***(0.001)	0.013***(0.001)	0.024***(0.001)	0.003***(0.001)
Intercept	0.004(0.004)	0.005***(0.001)	0.006***(0.001)	0.006***(0.002)	0.005***(0.001)	0.003***(0.001)
N		83,836			83,836

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Xiaoshuang Iris Luo

Notes

Appendix

Table A1.

Summary statistics of people on parole (averaged from 2003 to 2011).

	Austin	Dallas	Fort Worth	Houston	San Antonio
Number	14,920	63,737	36,142	124,611	44,983
Male	87.7%	88.3%	86.2%	89.0%	90.1%
White	31.7%	16.3%	30.6%	15.0%	14.6%
Black	38.2%	66.9%	50.7%	64.2%	14.6%
Latino	29.7%	16.4%	18.3%	20.4%	70.5%
Married	15.5%	17.5%	18.3%	15.7%	18.9%
Age 0–24	1.6%	1.1%	1.5%	1.9%	1.4%
Age 25–34	21.6%	18.9%	19.0%	22.5%	21.1%
Age 35–49	41.6%	43.0%	42.6%	41.1%	42.5%
Age 50–64	30.7%	31.6%	31.5%	30.0%	30.4%
Age 65 and up	4.6%	5.4%	5.4%	4.4%	4.6%

References

Allison

P. D.

(2012). Handling missing data by maximum likelihood (Working Paper No. 312-2012). SAS Global Forum. https://statisticalhorizons.com/wp-content/uploads/MissingDataByML.pdf

Alper

Durose

M. R.

Markman

(2018). 2018 update on prisoner recidivism: A 9-year follow-up period (2005–2014). US Department of Justice, Office of Justice Programs, Bureau of Justice.

Baer

Bhati

Brooks

Castro

La Vigne

Mallik-Kane

Roman

(2006). Understanding the challenges of prisoner reentry: Research findings from the Urban Institute’s prisoner reentry portfolio. The Urban Institute.

Boessen

Hipp

J. R.

(2015). Close-ups and the scale of ecology: Land uses and the geography of social context and crime. Criminology; An Interdisciplinary Journal, 53(3), 399–426. https://doi.org/10.1111/1745-9125.12074

Boessen

Hipp

J. R.

(2021). The network of neighborhoods and geographic space: Implications for joblessness while on parole. Journal of Quantitative Criminology, 38(3), 597–636. https://doi.org/10.1007/s10940-021-09510-z

Braman

(2007). Doing time on the outside: Incarceration and family life in urban America. University of Michigan Press.

Browning

C. R.

Calder

C. A.

Boettner

Smith

(2017). Ecological networks and urban crime: The structure of shared routine activity locations and neighborhood-level informal control capacity. Criminology: An Interdisciplinary Journal, 55(4), 754–778. https://doi.org/10.1111/1745-9125.12152

Browning

C. R.

Pinchak

N. P.

Calder

C. A.

(2021). Human mobility and crime: Theoretical approaches and novel data collection strategies. Annual Review of Criminology, 4, 99–123. https://doi.org/10.1146/annurev-criminol-061020-021551

Cadora

Swartz

Gordon

(2003). Criminal Justice and Health and Human Services: An Exploration of Overlapping Needs, Resources, and Interests in Brooklyn Neighborhoods (From Prisoners Once Removed: The Impact of Incarceration and Reentry on Children, Families, and Communities, P 285-311, 2003, Jeremy Travis and Michelle Waul, eds).

10.

Chamberlain

A. W.

(2012). Community change and recidivism: The interrelationship between neighborhood ecology and prisoner reintegration. University of California.

11.

Chamberlain

A. W.

(2016). From prison to the community: Assessing the direct, reciprocal, and indirect effects of parolees on neighborhood structure and crime. Crime & Delinquency, 64(2), 166–200. https://doi.org/10.1177/0011128716678194

12.

Chamberlain

A. W.

Boggess

L. N.

(2018). Parolee concentration, risk of recidivism, and the consequences for neighborhood crime. Deviant Behavior, 40(12), 1522–1542. https://doi.org/10.1080/01639625.2018.1540574

13.

Chamberlain

A. W.

Wallace

(2016). Mass reentry, neighborhood context and recidivism: Examining how the distribution of parolees within and across neighborhoods impacts recidivism. Justice Quarterly, 33(5), 912–941. https://doi.org/10.1080/07418825.2015.1012095

14.

Clear

T. R.

(2009). Imprisoning communities: How mass incarceration makes disadvantaged neighborhoods worse. Oxford University Press.

15.

Clear

T. R.

Rose

D. R.

Ryder

J. A.

(2001). Incarceration and the community: The problem of removing and returning offenders. Crime & Delinquency, 47(3), 335–351. https://doi.org/10.1177/0011128701047003003

16.

Clear

T. R.

Rose

D. R.

Waring

Scully

(2003). Coercive mobility and crime: A preliminary examination of concentrated incarceration and social disorganization. Justice Quarterly, 20(1), 33–64. https://doi.org/10.1080/07418820300095451

17.

Clear

T. R.

Waring

Scully

(2005). Communities and reentry: Concentrated reentry cycling. In Travis

Visher

(Eds.). Prisoner reentry and crime in America, 179–208. Cambridge University Press. https://doi.org/10.1017/CBO9780511813580.007

18.

Cohen

M. L.

Zhang

X. D.

(1988). The difficulty of improving statistical synthetic estimation (Statistical Research Division Report Series). U.S. Census Bureau.

19.

Dhondt

(2012). The bluntness of incarceration: Crime and punishment in Tallahassee neighborhoods, 1995 to 2002. Crime, Law and Social Change, 57(5), 521–538. https://doi.org/10.1007/s10611-012-9376-z

20.

Drakulich

K. M.

Crutchfield

R. D.

Matsueda

R. L.

Rose

(2012). Instability, informal control, and criminogenic situations: Community effects of returning prisoners. Crime, Law and Social Change, 57(5), 493–519. https://doi.org/10.1007/s10611-012-9375-0

21.

Durose

M. R.

Cooper

A. D.

Snyder

H. N.

(2014). Recidivism of prisoners released in 30 states in 2005: Patterns from 2005 to 2010. US Department of Justice, Office of Justice Programs, Bureau of Justice.

22.

Fagan

West

Hollan

(2002). Reciprocal effects of crime and incarceration in New York city neighborhoods. Fordham Urban Law Review, 30(5), 1551–1602. https://heinonline.org/HOL/P?h=hein.journals/frdurb30&i=1593 https://heinonline.org/HOL/PrintRequest?handle=hein.journals/frdurb30&collection=journals&div=67&id=1593&print=section&sction=67

23.

Garland

(2001). The culture of control: Crime and social order in contemporary society. Oxford University Press.

24.

Golledge

R. G.

Stimson

R. J.

(1997). Spatial behavior: A geographic perspective. Guilford Press.

25.

Granovetter

(1985). Economic action and social structure: The problem of embeddedness. American Journal of Sociology, 91(3), 481–510. https://doi.org/10.1086/228311

26.

Hannon

Defina

(2010). The state of the economy and the relationship between prisoner reentry and crime. Social Problems, 57(4), 611–629. https://doi.org/10.1525/sp.2010.57.4.611

27.

Harding

D. J.

Morenoff

J. D.

Herbert

C. W.

(2013). Home is hard to find: Neighborhoods, institutions, and the residential trajectories of returning prisoners. The ANNALS of the American Academy of Political and Social Science, 647(1), 214–236. https://doi.org/10.1177/0002716213477070

28.

Hipp

J. R.

Boessen

(2013). Egohoods as waves washing across the city: A new measure of “neighborhoods”. Criminology, 51(2), 287–327. https://doi.org/10.1111/1745-9125.12006

29.

Hipp

J. R.

Petersilia

Turner

(2010). Parolee recidivism in California: The effect of neighborhood context and social service agency characteristics. Criminology, 48(4), 947–979. https://doi.org/10.1111/j.1745-9125.2010.00209.x

30.

Hipp

J. R.

Tita

G. E.

Greenbaum

R. T.

(2009). Drive-bys and trade-ups: Examining the directionality of the crime and residential instability relationship. Social Forces, 87(4), 1778–1812. https://doi.org/10.1353/sof.0.0184

31.

Hipp

J. R.

Wickes

(2017). Violence in urban neighborhoods: A longitudinal study of collective efficacy and violent crime. Journal of Quantitative Criminology, 33(4), 783–808. https://doi.org/10.1007/s10940-016-9311-z

32.

Hipp

J. R.

Yates

D. K.

(2009). Do returning parolees affect neighborhood crime? A case study of Sacramento. Criminology, 47(3), 619–656. https://doi.org/10.1111/j.1745-9125.2009.00166.x

33.

Jones

Pebley

A. R.

(2014). Redefining neighborhoods using common destinations: Social characteristics of activity spaces and home census tracts compared. Demography, 51(3), 727–752. https://doi.org/10.1007/s13524-014-0283-z

34.

Kaeble

(2018). Probation and parole in the United States, 2016 (NCJ 251148). Bureau of Justice Statistics.

35.

Kaeble

Cowhig

(2016). Correctional populations in the United States, 2016 (NCJ-251211). Bureau of Justice Statistics.

36.

Kearns

Livingston

Galster

Bannister

(2018). The effects of neighbourhood offender concentrations on the number, type and location of crimes committed by resident offenders. The British Journal of Criminology, 59(3), 653–673. https://doi.org/10.1093/bjc/azy065

37.

Kirk

D. S.

(2015). A natural experiment of the consequences of concentrating former prisoners in the same neighborhoods. Proceedings of the National Academy of Sciences, 112(22), 6943–6948. https://doi.org/10.1073/pnas.1501987112

38.

Kovandzic

T. V.

Marvell

T. B.

Vieraitis

L. M.

Moody

C. E.

(2004). When prisoners get out: The impact of prison releases on homicide rates, 1975–1999. Criminal Justice Policy Review, 15(2), 212–228. https://doi.org/10.1177/0887403404263443

39.

Kubrin

C. E.

Stewart

E. A.

(2006). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44(1), 165–197. https://doi.org/10.1111/j.1745-9125.2006.00046.x

40.

La Vigne

Visher

Yahner

(2005). Chicago communities and prisoner reentry. Urban Institute.

41.

La Vigne

N. G.

Parthasarathy

(2005). Returning home Illinois policy brief: Prisoner reentry and residential mobility. Justice Policy Center, Urban Institute.

42.

La Vigne

N. G.

Visher

Yahner

(2004). Chicago prisoners’ experiences returning home. Urban Institute.

43.

Lee

Harding

D. J.

Morenoff

J. D.

(2017). Trajectories of neighborhood attainment after prison. Social Science Research, 66, 211–233. https://doi.org/10.1016/j.ssresearch.2016.12.004

44.

Livingston

Galster

Kearns

Bannister

(2014). Criminal neighbourhoods: Does the density of prior offenders in an area encourage others to commit crime? Environment and Planning A: Economy and Space, 46(10), 2469–2488. https://doi.org/10.1068/a140180p

45.

Lynch

J. P.

Sabol

W. J.

(2001). Prisoner reentry in perspective. Urban Institute.

46.

Lynch

J. P.

Sabol

W. J.

(2004). Effects of incarceration on informal social control in communities. In Pattillo

Weiman

Western

(Eds.), Imprisoning America: The social effects of mass incarceration (pp. 135–164). Russell Sage Foundation.

47.

Morenoff

J. D.

Harding

D. J.

(2014). Incarceration, prisoner reentry, and communities. Annual Review of Sociology, 40(1), 411–429. https://doi.org/10.1146/annurev-soc-071811-145511

48.

O’Brien

R. M.

(2007). A caution regarding rules of thumb for variance inflation factors. Quality & Quantity, 41(5), 673–690. https://doi.org/10.1007/s11135-006-9018-6

49.

Openshaw

Taylor

P. J.

(1981). The modifiable areal unit problem. In Wrigley

Bennett

R. J.

(Eds.), Quantitative geography: A British view (pp. 60–69). Routledge and Kegan Paul.

50.

Petersilia

(2003). When prisoners come home: Parole and prisoner reentry. Oxford University Press.

51.

Raphael

Stoll

M. A.

(2004). The effect of prison releases on regional crime rates. In Brookings-Wharton papers on urban affairs (Vol. 2004, pp. 207–255). https://doi.org/10.1353/urb.2004.0011

52.

Rose

D. R.

Clear

T. R.

(1998). Incarceration, social capital, and crime: Implications for social disorganization theory. Criminology, 36(3), 441–480. https://doi.org/10.1111/j.1745-9125.1998.tb01255.x

53.

Seiter

R. P.

Kadela

K. R.

(2003). Prisoner reentry: What works, what does not, and what is promising. Crime & Delinquency, 49(3), 360–388. https://doi.org/10.1177/0011128703049003002

54.

Simes

J. T.

(2019). Place after prison: Neighborhood attainment and attachment during reentry. Journal of Urban Affairs, 41(4), 443–463. https://doi.org/10.1080/07352166.2018.1495041

55.

Singh

Shaffer

M. A.

Selvarajan

(2018). Antecedents of organizational and community embeddedness: The roles of support, psychological safety, and need to belong. Journal of Organizational Behavior, 39(3), 339–354. https://doi.org/10.1002/job.2223

56.

Stahler

G. J.

Mennis

Belenko

Welsh

W. N.

Hiller

M. L.

Zajac

(2013). Predicting recidivism for released state prison offenders: Examining the influence of individual and neighborhood characteristics and spatial contagion on the likelihood of reincarceration. Criminal Justice and Behavior, 40(6), 690–711. https://doi.org/10.1177/0093854812469609

57.

Steinberg

(1979). Synthetic estimates for small areas: Statistical workshop papers and discussion (Vol. 24). National Institute on Drug Abuse.

58.

Visher

C. A.

Debus-Sherrill

S. A.

Yahner

(2011). Employment after prison: A longitudinal study of former prisoners. Justice Quarterly, 28(5), 698–718. https://doi.org/10.1080/07418825.2010.535553

59.

Visher

C. A.

Travis

(2003). Transitions from prison to community: Understanding individual pathways. Annual Review of Sociology, 29(1), 89–113. https://doi.org/10.1146/annurev.soc.29.010202.095931

60.

von Hippel

P. T.

Hunter

D. J.

Drown

(2017). Better estimates from binned income data: Interpolated CDFs and mean-matching. Sociological Science, 4(26), 641–655. https://doi.org/10.15195/v4.a26

61.

Western

(2018). Homeward: Life in the year after prison. Russell Sage Foundation.

62.

Wikström

(2006). Individuals, settings, and acts of crime: Situational mechanisms and the explanation of crime. In Wikström

P. H.

Sampson

R. J.

(Eds.), The explanation of crime: Context, mechanisms, and development (pp. 61–107). Cambridge University Press.

	Violent crime rate (logged)		Property crime rate (logged)		Parolee concentration (Number of parolees logged)		Parolee embeddedness (cumulative days of parolees/1000)
	Mean	SD	Mean	SD	Mean	SD	Mean	SD
2003	3.57	1.83	6.04	1.15	0.81	0.75	0.25	0.28
2004	3.55	1.83	6.05	1.11	1.04	0.83	0.37	0.33
2005	3.52	1.83	6.02	1.09	1.17	0.87	0.46	0.38
2006	3.44	1.82	5.95	1.05	1.12	0.86	0.43	0.42
2007	3.48	1.82	6.02	1.04	1.06	0.85	0.41	0.44
2008	3.38	1.81	5.91	1.02	1.07	0.85	0.37	0.44
2009	3.39	1.80	6.03	1.00	1.09	0.87	0.36	0.42
2010	3.35	1.76	5.97	1.01	1.09	0.87	0.35	0.42
2011	3.23	1.77	5.91	1.01	1.08	0.86	0.34	0.43
Average	3.43	1.81	5.99	1.05	1.06	0.84	0.37	0.40
Control variables
Concentrated disadvantage					0.39	10.67
Residential stability					0.02	0.69
Income inequality					0.86	0.12
Racial/ethnic heterogeneity					0.41	0.19
Percent immigrants					0.20	0.13
Percent aged 16–29					0.24	0.10
Percent Black					0.19	0.27
Percent Latino					0.39	0.30
Population density					0.10	0.08
Annual measures – average over years
Total employees (logged)					5.47	1.66
Retail employees (logged)					2.78	1.77
Recreation employees (logged)					0.69	1.18
Food employees (logged)					2.13	2.07
Total voluntary organisations (logged)					0.19	0.40

	Model 1: Violent crime model			Model 2: Property crime model
	Violent crime	Parolee concentration	Parolee embeddedness	Property crime	Parolee concentration	Parolee embeddedness
Violent/property crime (logged)	0.865***(0.002)	0.015***(0.000)	0.003***(0.000)	0.932***(0.002)	0.002**(0.001)	0.001(0.000)
Parolee concentration	0.058***(0.002)	0.801***(0.002)		0.021***(0.005)	0.805***(0.002)
Parolee embeddedness	−0.032***(0.004)		0.771***(0.002)	−0.044***(0.013)		0.771***(0.002)
Violent/property crime (logged)²		−0.002***(0.000)	−0.002***(0.000)		−0.004***(0.000)	−0.002***(0.000)
Parolee concentration²	−0.013***(0.001)			−0.009***(0.002)
Parolee embeddedness²	0.022***(0.004)			0.036***(0.011)
Concentrated disadvantage	0.004***(0.000)	0.002***(0.000)	0.001***(0.000)	0.001***(0.000)	0.003***(0.000)	0.001***(0.000)
Residential stability	−0.016***(0.001)	0.018***(0.001)	0.015***(0.001)	−0.001(0.001)	0.016***(0.001)	0.014***(0.001)
Income inequality	0.058***(0.006)	0.075***(0.004)	0.044***(0.003)	0.057***(0.004)	0.085***(0.004)	0.044***(0.003)
Racial/ethnic heterogeneity	0.060***(0.004)	0.030***(0.003)	0.050***(0.002)	0.009***(0.003)	0.041***(0.003)	0.055***(0.002)
Percent immigrants	0.047***(0.009)	−0.147***(0.006)	−0.033***(0.004)	−0.013*(0.006)	−0.144***(0.006)	−0.034***(0.004)
Percent aged 16–29	−0.042***(0.010)	−0.156***(0.007)	−0.058***(0.004)	−0.051***(0.007)	−0.154***(0.007)	−0.049***(0.004)
Percent Black	0.273***(0.007)	0.342***(0.004)	0.078***(0.003)	0.037***(0.004)	0.372***(0.004)	0.085***(0.003)
Percent Latino	0.167***(0.007)	0.221***(0.004)	0.083***(0.003)	0.009*(0.004)	0.245***(0.004)	0.091***(0.003)
Population density	−0.179***(0.010)	0.619***(0.016)	0.121***(0.005)	−0.195***(0.009)	0.593***(0.016)	0.124***(0.006)
Total employees (logged)	0.014***(0.001)	0.000(0.000)	0.002***(0.000)	0.000(0.001)	0.002***(0.000)	0.002***(0.000)
Retail employees (logged)	0.012***(0.001)	−0.003***(0.000)	−0.000(0.000)	0.006***(0.000)	−0.002***(0.000)	0.000(0.000)
Recreation employees (logged)	0.007***(0.001)	0.001(0.000)	0.001***(0.000)	0.004***(0.000)	0.002***(0.000)	0.001***(0.000)
Food employees (logged)	0.019***(0.001)	−0.002***(0.000)	0.001***(0.000)	0.008***(0.000)	0.001**(0.000)	0.001***(0.000)
Total voluntary organisation (logged)	0.015***(0.001)	0.022***(0.001)	0.003***(0.001)	0.013***(0.001)	0.024***(0.001)	0.003***(0.001)
Intercept	0.004(0.004)	0.005***(0.001)	0.006***(0.001)	0.006***(0.002)	0.005***(0.001)	0.003***(0.001)
N		83,836			83,836