Race and the Geography of Opportunity in the Post-Prison Labor Market

Abstract

Research on racial disparities in post-prison employment has primarily focused on the differential effects of stigma on blacks and whites, but we otherwise know little about racial differences. This paper examines racial differences in post-prison employment by industry and geography. We find that the formerly incarcerated are most likely to find work in a small number of “felon-friendly” industries with formerly incarcerated whites having higher employment rates than blacks. Whites are more likely to be employed in felon-friendly industries associated with the primary labor market, particularly construction and manufacturing, which have higher wages and more job stability. To explain these racial differences, we investigate the degree to which employment among the formerly incarcerated is related to where felon-friendly employers are located and where individuals who work in felon-friendly industries live. We find that post-prison employment is associated more with proximity to workers in felon-friendly industries than with proximity to employers. Because formerly incarcerated whites are more likely to live near current workers in felon-friendly industries, the geography of opportunity in the post-prison labor market contributes to the racial disparity in post-prison employment.

Racial inequities in incarceration rates are often identified as a key contributor to employment differences between blacks and whites, but even among the formerly incarcerated, there are racial inequalities in post-prison employment (Western 2006). While we know that stigma and its differential effects on blacks and whites plays an important role in these differences (Holzer, Raphael, and Stoll 2004, 2006; Pager 2003), we know less about how racial differences in where the formerly incarcerated live post-prison influences their employment.

Much like where blacks and whites live pre-prison, formerly incarcerated blacks live in neighborhoods with much larger disadvantages than their white counterparts (Massoglia, Firebaugh, and Warner 2013). Given this, employment outcomes of the formerly incarcerated are likely influenced by the same geography of access that affects other low-wage, low-skill workers (Dawkins 2016; Kain 1968; Lens 2014; Mouw 2000; O’Regan and Quigley 1996; Parks 2004; Wilson 1996). However, geography may be especially important for employment opportunities among the formerly incarcerated, because the jobs for which they are eligible are limited by their human capital deficits and the stigma of a criminal record (Holzer, Raphael, and Stoll 2004, 2006; Pager 2003; Schmitt and Warner 2010; Visher and Travis 2003).

The prior literature on location and job search suggests that where the formerly incarcerated live is likely to influence employment outcomes in two ways: by providing proximity to jobs and by providing access to networks with employment information (Fernandez and Su 2004; Kain 1968; Mouw 2002; Wilson 1996). Workers can be located closer to or further away from their potential employers, which affects access to employment opportunities. Additionally, where a person lives affects the composition of their social networks, particularly for low-income Americans who are more likely to have close ties in their neighborhood than those in higher income groups are (Campbell and Lee 1992; Chaskin 1997). These factors could produce racial differences not only in rates of employment, but also in access to “good” jobs in the primary labor market (Bushway, Stoll, and Weiman 2007; Kalleberg 2013; Western 2002). Yet, the frequency of employment in the primary and secondary labor markets among the formerly incarcerated has never been directly assessed in a large representative sample.

Using an administrative dataset for a large, randomly sampled cohort of formerly incarcerated people in Michigan, we investigate the relationship between the geographic location of employers and workers, and the employment outcomes of the formerly incarcerated. Our data show that the formerly incarcerated are most likely to find work in construction, manufacturing, food services, retail, and temporary labor.¹ We term these “felon-friendly” industries and divide them into primary labor market jobs associated with “relatively high wages, good working conditions, [and] chances of advancement” (e.g., construction and manufacturing), secondary labor market jobs associated with “low-paying, with poorer working conditions, [and] little chance of advancement” (e.g., retail and food services), and jobs in the temporary labor market (Piore 1972:2). Our results show that not only are formerly incarcerated whites more likely to be employed than formerly incarcerated blacks, but whites are more likely to be employed in felon-friendly industries associated with the primary labor market.

Our subsequent analyses seek to understand the role of geography in producing post-prison black-white labor market inequalities. While racial differences in employment are well documented in the prior research (Bushway et al. 2007; Western 2002, 2006), the role racial differences in spatial location might play in explaining these post-incarceration outcomes has not previously been examined. We investigate the degree to which employment among the formerly incarcerated is impacted by where felon-friendly employers are located and where workers in felon-friendly industries live.

The findings demonstrate that living near felon-friendly employers and near current workers increases the chances of finding employment after incarceration, but differentially by race and industry. The strongest results are for proximity to current workers. Net of local unemployment rates, post-prison employment is associated with living close to workers in certain felon-friendly industries. Whites benefit from living in areas with more residents employed in construction and manufacturing, while blacks only see benefits from living in areas with a large number of manufacturing workers. We also find that proximity to jobs increases the chances of finding employment in retail, but only for whites. In fact, blacks are more likely to be geographically proximate to employers and workers of non-felon friendly industries that are less likely to hire them, which may contribute to their lower rates of post-prison employment.

These findings demonstrate the importance of the geography of the low-wage labor market for the outcomes of the formerly incarcerated. Although the results reflect challenges faced by low-wage workers in general, the formerly incarcerated represent a hard-to-employ population that faces more severe restrictions on employment. Proximity to workers who can connect them to jobs into which they can be hired, and to a lesser extent proximity to jobs, play a role in whether they find employment post-release. The limits of their employment opportunities and the role that residential racial segregation plays in structuring those opportunities illustrate how spatial patterns of inequality exacerbate other racial inequalities.

POST-PRISON EMPLOYMENT, THE LOW-SKILL LABOR MARKET, AND THE GEOGRAPHY OF OPPORTUNITY

Individuals with criminal records are a particularly hard to employ population. Employment rates for the formerly incarcerated are lower than those for other low-skilled, low-education groups (Schmitt and Warner 2010; Western 2006). These lower rates are due in part to employers’ general unwillingness to hire someone with a record (Holzer, Raphael, and Stoll 2004), but also because the formerly incarcerated often face restrictions on the type of work they can do (Petersilia 2003; Travis 2005). In addition to the racially stratified challenges of finding employment, prior studies show that non-whites with criminal records face lower rates of employment and have less stable employment post-release (Grogger 1992; Visher et al. 2010).

When the formerly incarcerated do find work, they are limited to employers who are both willing and able to hire individuals with criminal records. Prior evaluation studies with non-random samples suggest that these industries are construction, manufacturing, food services, retail, and temporary work (Leshnick et al. 2012; Visher et al. 2010). These industries vary in their stability, pay, and benefits, which could lead some formerly incarcerated to hold “good” jobs – or jobs in the primary labor market – while others work in “bad” jobs in the secondary labor market (Bushway et al. 2007; Kalleberg 2013; Reich, Gordon, and Edwards 1973; Western 2002). Construction and manufacturing tend to provide primary labor market job opportunities that offer higher wages, benefits, and, if not year round employment opportunities, the promise of rehiring after seasonal breaks. Food services and retail represent secondary job market options with lower wages, no benefits, and often seasonal employment. Although typically considered to be part of the secondary labor market, the temporary labor industry provides even less stability than retail and food services. However, temporary labor can sometimes be a route to more stable employment as it has been increasingly used as a way for employers to test out “risky” workers, including those with criminal records, a lack of prior work history, and no prior experience in the industry of work (Houseman, Kalleberg, and Erickcek 2003).

Geography and Access to Employers and Job Networks

The aforementioned studies focused on individual-level characteristics (e.g., education and work experience) to explain employment patterns among the formerly incarcerated. In contrast, this paper examines whether residential location matters for employment outcomes, expanding the literature on the role of geography. Prior research has studied the relationship between geographic patterns of access to jobs and broader crime patterns (Gould, Weinberg, and Mustard 2002; Hannon and DeFina 2010; Wang and Minor 2002) and the relationship between employment opportunities and recidivism (Bellair and Kowalski 2011; Raphael and Weiman 2007; Wang, Mears, and Bales 2010). These studies assume that geography affects employment and that employment affects recidivism, but they do not specifically examine these relationships, nor do they address racial differences in employment among the formerly incarcerated. Only one study of which we are aware examines geographically-based labor market conditions and post-prison employment. Sabol (2007) found that county level unemployment rates were negatively associated with employment outcomes for formerly incarcerated individuals in Ohio. Although this study controlled for race, it did not directly address racial differences in either county level unemployment or its effect on employment.

Where individuals reside immediately after prison might lead to racial differences in their employment through neighborhood racial segregation (Charles 2003; Massey and Denton 1993; Mouw 2002). In addition to returning to racially segregated neighborhoods, recent studies suggest that formerly incarcerated blacks might be doubly disadvantaged because they are more likely than formerly incarcerated whites to return to the most disadvantaged neighborhoods (Harding, Morenoff, and Herbert 2013; Kubrin and Stewart 2006; Massoglia et al. 2013). Formerly incarcerated blacks are exposed to residential neighborhoods with substantially different socio-spatial characteristics than formerly incarcerated whites, which has the potential to impact their post-prison outcomes, including their ability to find employment.

Patterns of residential segregation and socio-economic disadvantage may lead to racial differences in employment among ex-offenders through two main processes. First, the literature on spatial isolation suggests that employers followed white flight to suburbs, leaving predominately black inner-city areas with fewer jobs in many U.S. cities (Kain 1968; Wilson 2011). Spatial isolation studies suggest that lower rates of employment in low-income, inner-city neighborhoods are due in part to the physical distance between the low-wage worker labor force and low-wage jobs (Glaeser and Kahn 2001; Lens 2014; O’Regan and Quigley 1996; Parks 2004; Stoll, Holzer, and Ihlanfeldt 1999; Stoll and Raphael 2000; Wilson 2011).

But geographic segregation and residential patterns by race may also matter for employment outcomes for a second reason: the location of potential network ties for referrals and information during a job search (e.g., Bayer, Ross, and Topa 2008). Because low-income people tend to have geographically constrained social ties (Campbell and Lee 1992; Chaskin 1997), their area of residence has implications for their access to networks. If the neighborhoods to which formerly incarcerated individuals return are also home to many workers of a particular industry, we expect those potential neighborhood social networks to influence their employment prospects in that industry (Granovetter 1995; Mouw 2002). In addition, the prior literature suggests that racial differences in using network ties for job referrals and information may also contribute to racial differences in post-prison employment. This research finds that blacks are less likely to provide referrals to friends or family members than whites (Royster 2003; Smith 2007).²

In sum, spatial patterns of residence are likely to affect the employment opportunities and potential information networks to which the formerly incarcerated have access post-release. Given spatial patterns of racial segregation in the United States and racial differences in the use of job networks, these factors are likely to vary by race. However, we also have reason to believe that the role of job information networks will vary by industry, given industry-based variation in whether employers use referrals (Rees 1966; Rees and Shultz 1970), as described further below.

Social Networks in the Primary, Secondary, and Temporary Labor Markets

Potential connections to worker networks may matter less for some jobs than for others due to differences in hiring practices (Baron, Jennings, and Dobbin 1988). Secondary labor market jobs, such as those in retail and food service, are ubiquitous. That retail and food service jobs are an option for employment, where they are located, whether they are hiring, and that they generally accept applications continuously is common knowledge. While evidence suggests that social networks can facilitate hiring in the secondary labor market (Heneman and Berkley 1999; Pettinger 2005), their effects are likely to be substantially smaller than in the primary job market, whose hiring practices differ dramatically and with specific implications for our hypothesis that potential worker networks affect employment of the formerly incarcerated.

While it may be common knowledge that primary labor market industries offer potential jobs, when and how to apply for those jobs may require specialized knowledge that social networks can provide. Primary market firms typically engage in more selective hiring practices, which can begin before the application process, by limiting who has the opportunity to submit an application and when. For example, construction and manufacturing firms are most likely to accept applications during temporally narrow hiring periods dictated by projected and actual market demand for the goods they produce. Additionally, some firms do not advertise when they start hiring, and accept applications from the general public only if positions remain unfilled after informal referrals (Granovetter 1995; Marsden and Gorman 2001; Ozga 1960).

Some felon-friendly industries in the primary labor market, particularly construction, predominately hire whites, in part because they hire within current employees’ social networks (Alexander, Entwisle, and Olson 2014; Royster 2003; Waldinger 1995; Waldinger and Bailey 1991). Given the racial segregation of friend networks in the United States (Jackman and Crane 1986; McPherson, Smith-Lovin, and Cook 2001; Smith, McPherson, and Smith-Lovin 2014), formerly incarcerated blacks are thus less likely to hear about some of the limited primary job market options available to them post-release, simply because their social networks are less likely to include white workers.

Hiring practices among temporary firms appear similar to both the secondary and primary labor markets. Temporary firms are similar to secondary labor market firms in that knowledge of their existence is common and that they accept applications continuously. However, like the primary labor market, temporary job placement can be a highly selective process, due in part to temporary firms’ responsibility for the quality of workers they represent (Kalleberg 2000). In contrast to hiring processes in both the primary and secondary job markets, however, potential social networks may subvert, rather than facilitate, employment in the temporary workforce. Temporary workers have a disincentive to refer others to their firms because of internal competition for placements. They may also attempt to protect their own job prospects by actively discouraging other workers from seeking temporary employment, a practice employers are likely to encourage by highlighting their job insecurity (Elcioglu 2010; Smith 1998).

In sum, we expect the contribution of geographically proximate social networks to post-prison employment to vary by industry. We expect the association between employment and proximity to workers to be strongest and positive in primary labor market jobs, where information regarding the availability of jobs and how to get them may not be public knowledge. We expect that association to be diminished in the secondary labor market, where information about available jobs is not restricted. In the temporary market, we expect that access to other temporary workers may actually be associated with decreased employment in temporary work.

DATA AND METHODS

This study investigates whether racial differences in the geographically-based availability of felon-friendly jobs and workers in felon-friendly industries contributes to black-white disparities in employment post-incarceration, and how those effects vary across the primary, secondary, and temporary labor markets. We rely on three data sources. Individual-level records come from the Michigan Department of Corrections (MDOC). This includes a cohort of 11,064 Michigan prisoners who were paroled during 2003.³ This study is limited to a random one-third sample of 3,689 individuals for whom detailed residential data were collected.⁴ Individual-level data include records from the Michigan Unemployment Insurance (UI) system on formal employment and gross earnings; MDOC databases on pre-prison demographic, social, and criminal-history characteristics, and post-release residential data for the time each individual was on parole; and Michigan State Police records on arrests.⁵ Zip codes’ social, demographic, and economic characteristics come from the U.S. Census Bureau Decennial Census of 2000 and the American Community Survey (ACS) data from 2007-2011 and 2008-2012.⁶ Finally, data on the presence and type of employers by zip code come from the Business Register’s County Business Patterns (CBP) dataset from 2004 through 2010.⁷ Data across sources were merged by the zip code of residence and year. The combined dataset is organized by individual and quarter, with multiple observations per individual as we had data for each individual for 1 to 33 quarters. The final sample size includes 3,504 individuals and 27,675 person-quarters.⁸

Multinomial Logistic Regression Models

We estimated multinomial logistic discrete time event history models. Event history models predict the probability of an “event,” meaning a transition from one condition or state to another. In these analyses, we predicted the transition to employment, measured by employment in any felon-friendly industry or employment in a specific felon-friendly industry. In each quarter, a parolee can 1) achieve employment in any felon-friendly industry or a specific felon-friendly industry, 2) achieve employment in another industry, 3) remain unemployed, or 4) no longer be eligible for employment due to a return to prison, a move out of the state of Michigan, or death. “Episodes” of unemployment begin at release from prison or when a parolee becomes “at risk” of employment again starting with a quarter of unemployment. People paroled to institutions (e.g., hospitals, in-patient treatment centers, or county jails) do not begin their first episode until they move to a non-institutional address.⁹ Episodes end when the individual was employed or censoring occurs due to imprisonment, death, or moving out of state. If individuals returned to prison, their observations are excluded from the analysis until they were re-released.

More formally, in order to model employment in any industry while also accommodating censoring due to imprisonment, death, or moving out of state, we define a four-category outcome variable in quarter t for person j in episode i:

$$y_{\mathit{tij}}=\left\{\begin{array}{c}0, \mathrm{n}\mathrm{o} \mathrm{e}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{t} (\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{u}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{u}\mathrm{n}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{y}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t})\\ 1, \mathrm{e}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{y}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{i}\mathrm{n} \mathrm{a} \mathrm{f}\mathrm{e}\mathrm{l}\mathrm{o}\mathrm{n}‐\mathrm{f}\mathrm{r}\mathrm{i}\mathrm{e}\mathrm{n}\mathrm{d}\mathrm{l}\mathrm{y} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y} \mathrm{o}\mathrm{r} \mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c} \mathrm{f}\mathrm{e}\mathrm{l}\mathrm{o}\mathrm{n}‐\mathrm{f}\mathrm{r}\mathrm{i}\mathrm{e}\mathrm{n}\mathrm{d}\mathrm{l}\mathrm{y} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}\\ 2, \mathrm{e}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{y}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{i}\mathrm{n} \mathrm{o}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}\\ 3, \mathrm{c}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{e}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{s}\end{array}\right.$$

The discrete-time hazard function can then be defined as the probability of event r during quarter t (for person j in episode i) relative to remaining unemployed, given that no event has occurred before quarter t:

$$p_{\mathit{tij}}^{\left(r\right)}=\mathrm{Pr}\left(y_{\mathit{tij}}=r|y_{t-1,ij}=0\right) \text{for}\mathrm{ }r=\mathrm{1,2}.$$

We estimate this hazard function using a multinomial logit model that compares the probability of experiencing an event of type r to the probability of having no event,${ p}_{\mathit{tij}}^{\left(0\right)}$:

$$\log \left(\frac{p_{\mathit{tij}}^{\left(r\right)}}{p_{\mathit{tij}}^{\left(0\right)}}\right)={\alpha }^{\left(r\right)}{\mathit{D}}_{\mathit{tij}}^{\left(\mathit{r}\right)}+{\beta }^{\left(r\right)}{\mathit{x}}_{\mathit{tij}}^{\left(\mathit{r}\right)},\mathrm{ }\text{for}\mathrm{ }r=\mathrm{1,2}.$$

where ${\mathit{D}}_{\mathit{tji}}$ is a vector specifying a function of the cumulative duration of quarters within the episode at week t, with coefficients $\alpha$, and ${\mathit{x}}_{\mathit{tij}}$ is a vector of covariates that include both time-varying and time invariant characteristics, with coefficients $\beta$ (Steele 2008). This is also known as a discrete-time model for recurrent events with competing risks (Steele 2008).

To estimate the baseline logit-hazard functions,$\alpha {\mathit{D}}_{\mathit{ti}}$, we employed a quartic specification of time. We also included a linear term indexing episodes because few individuals had more than two episodes. The $\beta$ coefficients represent associations between the covariates and the log-odds of the event during a given quarter. The results of these models can be interpreted in terms of episode duration because having a higher probability of employment in any given quarter also means that the episode is likely to end sooner. Positive coefficients indicate associations with shorter durations of unemployment and negative coefficients longer durations. When interpreting results from these models, we focus on the comparison between finding employment and remaining unemployed. Full models are provided in the appendix.

Each model predicts employment based on the number of jobs in local firms per resident in the zip code (jobs per capita) and the percent of the local labor force (those living in the immediate zip code) working in the industry specified. The first independent variable was calculated based on CBP data about the sizes of firms, which is a range of the number of workers employed (e.g., 10–30). For each category of firm size, we multiplied the number of firms by the midpoint of the range of the number of workers. Since the largest size category was top coded, we multiplied the number of firms in that category by 1.5 times the value provided. Finally, we divided by the total population in the zip code based on Census data. The second independent variable of interest, the percent of the local labor force employed in the industry specified, came directly from Census and American Community Survey data.

Finally, each model includes controls for individual and zip code level characteristics that may affect one’s ability to find work.¹⁰ Individual characteristics include both baseline measures and time varying measures. Baseline measures include age, age-squared, race (black or white),¹¹ gender, education, marital status, whether the individual has dependents, the most serious offense category from their most recent felony sentence (assaultive, non-assaultive, or drug), whether the individual was released with electronic monitoring,¹² the number of prior prison spells, the number of years in prison during the last spell, discharge from parole, substance abuse history (any self-reported history of substance abuse pre-prison), release year, and whether the individual was released to a correctional center.¹³ Time-varying measures include positive substance abuse test results in the prior quarter while on parole, the cumulative number of post-release formal employment spells, and controls for the quarter and year of data. Zip code level controls include the average employer size, disadvantage score,¹⁴ the disadvantage score of the zip code the parolee resided in pre-prison, population density, and parolee density based on the entire 2003 parole cohort. These zip code controls ensure that our measures of the effects of employers and workers reflect the specific zip code characteristics for which we control, rather than the general social and economic characteristics of the zip code.

RESULTS

Racial Differences in Post-Prison Employment

Table 1 compares the industries of employment for workers in the reentry cohort who were employed in any quarter in 2004 with Michigan workers in general.¹⁵ The formerly incarcerated workers were predominately employed in different industries than other Michigan workers, regardless of race, which mainly consisted of five felon-friendly industries: construction, manufacturing, retail, food services, and temporary employment. In fact, only about 30 percent of the formerly incarcerated worked outside of those five industries, whereas the majority of Michigan workers (58 percent) worked in non-felon-friendly industries.

Table 1.

Employment Outcomes by Race, 2004

	Reentry Cohort			State of Michigan
Percent employed ina	All	Black	White	All	Black	White
Construction	10.94	3.81	18.06	3.20	1.21	5.18
Manufacturing	24.64	20.21	29.07	15.09	13.55	16.63
Retail	12.01	11.68	12.33	11.01	9.48	12.53
Food services	23.71	23.62	23.79	8.22	8.67	7.77
Temporary employment	35.25	45.93	24.56	4.01	5.22	2.81
All other industries	29.92	29.00	30.84	58.47	61.87	55.07
Number of individuals	1,670	762	908	3,928,062	481,313	3,446,749

Source: Reentry cohort data. State of Michigan data were retrieved from the Quarterly Workforce Indicators for 2004.

Note: See Appendix Table 1 for full NAICS codes.

^aReentry cohort data will add up to more than 100 percent because some reentry cohort members worked in more than one industry within the same quarter.

Although the formerly incarcerated were more likely to be employed in felon-friendly industries, we found racial variation in the felon-friendly industries of employment. Whites were more likely to be employed in primary labor market jobs in construction or manufacturing than their black counterparts. In contrast, blacks were more likely to work in temporary labor. There was no appreciable difference in employment in retail or food services for blacks and whites.

As shown in Table 2, black-white differences in employment among the formerly incarcerated extended beyond industry of employment. The first panel shows that whites were 15 percentage points more likely than blacks to have held a job in any industry at some point between 2001 and 2009. The largest black-white difference was in the construction industry where whites were 13 percentage points more likely to have ever held a job than blacks. Blacks were only more likely than whites to ever be employed in temporary work (6 percentage points).

Table 2.

Characteristics of Employment for Reentry Cohort, 2001‐2009

	All	Black	White
Percent ever employed
In any industry	58.60	51.71	66.41 ***
In construction	9.12	3.16	15.88 ***
In manufacturing	16.32	11.98	21.24 ***
In retail	8.93	7.42	10.65 ***
In food services	16.57	14.11	19.35 ***
In temporary employment	25.61	28.32	22.53 ***
All other industries	22.08	18.46	26.18 ***
Probability of finding a job conditional on being unemployed
In any industry	0.112	0.090	0.142 ***
In construction	0.012	0.002	0.025 *
In manufacturing	0.018	0.012	0.027
In retail	0.010	0.008	0.013
In food services	0.022	0.018	0.027
In temporary employment	0.034	0.036	0.031
All other industries	0.026	0.019	0.036
Number of quarters to first employmenta
In any industry	2.97	3.23	2.74 ***
In construction	4.54	5.90	4.23 **
In manufacturing	4.24	4.32	4.19
In retail	4.50	4.60	4.41
In food services	4.09	4.34	3.89
In temporary employment	4.52	4.43	4.64
All other industries	4.85	4.92	4.79
Number of individuals	3,627	1,927	1,700

Source: Reentry cohort data.

Note: Reentry cohort members can be employed in multiple sub‐industries across quarters. Figures presented here exclude quarters when reentry cohort members were in prison. Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

^aCalculation for reentry cohort members who experienced employment.

When unemployed, whites were 1.6 times more likely than blacks to find employment in any industry as shown in the second panel of Table 2. Within felon-friendly industries, blacks were never more likely than whites to find employment. While the probability of finding employment in the other felon-friendly industries was consistently 1.5 to 2.0 times higher for whites than for blacks, the largest racial discrepancy was in the construction industry, where whites were more than 10 times more likely than blacks to find a construction job.

Finally, the third panel shows that blacks who found employment searched longer to find their first job. Whites found their first job in 8 months (2.74 quarters) on average, whereas blacks spent almost 10 months (3.23 quarters) unemployed on average before finding their first job. Quarters to first employment were otherwise equal for blacks and whites with the exception of employment in construction, where whites found employment faster than blacks.

While formerly incarcerated individuals were generally disadvantaged in their employment outcomes in comparison to the average worker as shown in higher probabilities of employment in the secondary labor market and lower earnings than the average Michigan worker, formerly incarcerated whites had better outcomes than their black peers. Whites were more likely to be employed in primary labor market industries, had higher quarterly earnings, were more likely to be employed at any point in time, and found work faster than blacks. The racial disparity in employment in this already difficult-to-employ sample may reflect patterns of residential racial segregation that create differential access to jobs and workers.

The Racial Geography of Prisoner Reentry

The residential geography of the formerly incarcerated blacks and whites in our sample reflects broader patterns of residential racial segregation across the state of Michigan. Because where one lives affects exposure to employers and workers (Wilson 1996), blacks and whites in our sample were exposed to different concentrations of felon-friendly jobs and different concentrations of workers in felon-friendly industries.

Where the Formerly Incarcerated Live Relative to Employers and Workers

We found distinct residential patterns among formerly incarcerated blacks and whites that expose them to different employers and workers. Whites were fairly evenly dispersed across the state, as shown in the map to the left in Figure 1, and throughout urban, suburban, and rural areas. In contrast, blacks were largely concentrated around urban areas in the state including Detroit, Grand Rapids, Muskegon, Lansing, and Flint. These residential patterns were consistent with those in the general Michigan population.

Figure 1.

Number of Formely Incarcerated Individuals Residing in Zip Code by Race, 2004

The more urban location of black sample members compared with more suburban and rural locations of white sample members means that formerly incarcerated blacks and whites also lived in areas with distinct characteristics. As shown in Table 3, formerly incarcerated blacks lived in zip codes with more jobs per resident than whites. Formerly incarcerated whites lived in zip codes with 0.32 jobs per resident, while blacks lived in zip codes with 0.54 jobs per resident. This advantage also applied to felon-friendly industries, where there were 0.22 jobs per resident in zip codes where blacks lived, as compared to 0.17 jobs per resident where whites lived.

Table 3.

Characteristics of the Zip Codes Where Reentry Cohort Lives by Race

	All	Black	White
Per capita jobs at local firms
All employers	0.40	0.54	0.32
Felon‐friendly employers	0.19	0.22	0.17
Construction	0.02	0.02	0.02
Manufacturing	0.06	0.07	0.06
Retail	0.05	0.06	0.05
Food services	0.04	0.05	0.03
Temporary employment	0.02	0.02	0.02
All other industries	0.21	0.32	0.14
Concentration of parolees (# per sq. mile)	0.47	0.74	0.31
Population density (# per sq. mile)	3,170	4,355	1,799

Source: Reentry cohort data and Business Register's County Business Patterns.

Note: Data from quarter 1 of 2004

However, formerly incarcerated blacks were also more likely to live in zip codes with a higher concentration of individuals on parole. Blacks lived in zip codes with 2.5 times more formerly incarcerated residents per square mile than whites and almost 2.5 times more people per square mile. Thus, while blacks had a higher exposure to felon-friendly employers due to living in higher density, urban areas, they also lived in areas that potentially had more competition for felon-friendly jobs both due to living in areas with higher population densities and higher concentrations of formerly incarcerated individuals.

While the geographic distribution of jobs relative to where the formerly incarcerated lived appears to favor blacks over whites, the uneven geographic distribution of felon-friendly industries across the state suggests that where formerly incarcerated blacks and whites lived has the potential to explain their differential employment outcomes. Figure 2 shows the ratio of the number of formerly incarcerated individuals to the number of felon-friendly jobs in each zip code with darker colors representing more felon-friendly jobs per parolee in a zip code.¹⁶ This ratio provides a measure of the density of the formerly incarcerated relative to job availability.

Figure 2.

Ratio of Total Felon-Friendly Jobs to Formerly Incarcerated

These maps suggest that formerly incarcerated whites were generally exposed to less competition for felon-friendly jobs than blacks. Overall, whites lived in areas where the ratio of felon-friendly jobs to the formerly incarcerated was higher, as shown by darker shading. Furthermore, the areas with more job availability per formerly incarcerated whites were clustered, suggesting that whites were living in regions with lower competition. In contrast, blacks tended to live in areas with moderate or small parolee-to-job ratios as indicated by lighter shades of gray, indicating that there was more competition for felon-friendly jobs.

The average number of jobs in the zip codes where formerly incarcerated blacks and whites lived is consistent with these findings, as shown in Table 4. Within the zip code where the average black lived, there were significantly more jobs than where the average white lived. However, most of these jobs were predominately in non-felon-friendly industries. Formerly incarcerated whites had exposure to more felon-friendly jobs in all five categories. These racial differences in level of access to felon-friendly jobs suggests that employer location may contribute to the black-white disparities in employment post-incarceration.

Table 4.

Average Employee Zip Code Composition by Race, 2004

	All	Black	White
Number of workers by industry	11,054	11,355	10,705 *
Construction	436	399	479 ***
Manufacturing	1,753	1,661	1,861 **
Retail	1,422	1,320	1,540 ***
Food services	949	898	1,008 ***
Temporary employment	462	439	492 *
All other industries	6,074	6,650	5,409 ***
Percent of labor force employeda	86.57	83.63	89.97 ***
Construction	6.79	5.59	8.18 ***
Manufacturing	10.59	10.31	10.91 ***
Retail	8.76	8.05	9.57 ***
Food services	5.47	5.45	5.49
Temporary employment	12.96	12.95	12.96
All other industries	51.47	53.23	49.42 ***

Source: U.S. Census.

Notes: Data from quarter 1 of 2004. Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

^aThe percent of residents employed is based on a population 16 years of age or older.

Where the Formerly Incarcerated Live Relative to Workers in Different Industries

The potential to associate with others who have the ability to connect formerly incarcerated individuals to jobs also varied by race, as shown in the bottom panel of Table 4. Formerly incarcerated whites lived in zip codes with higher employment rates than blacks on average. Whites also had higher exposure to individuals employed in felon-friendly industries than blacks, particularly for construction, manufacturing, and retail. In contrast, blacks generally lived in zip codes with more residents employed in non-felon-friendly industries. The two exceptions to these general findings are that blacks and whites lived in zip codes with similar percentages of residents employed in food services and temporary work.

Based on the concentration of individuals employed in each industry, formerly incarcerated blacks seem to be at a slight disadvantage for finding employment because residents in their zip codes were less likely to work in felon-friendly industries. Figure 3 shows the ratio of formerly incarcerated individuals to the number of residents working in felon-friendly industries. As shown in the left map, whites tended to live in zip codes and regions with higher rates of exposure to workers in felon-friendly industries than blacks. In contrast, blacks lived in areas with moderate or small ratios as indicated by lighter shades of gray.

Figure 3.

Ratio of Employees of Felon-Friendly Industries to Formerly Incarcerated

Racial differences in residential patterns among the formerly incarcerated exposed whites and blacks to different proximities to post-prison employment opportunities and employment information networks, particularly for felon-friendly industries. Although proximity to employers may expand opportunities for blacks, limited access to job opportunities with those employers could reduce their influence on employment outcomes. The following section examines whether and how employment opportunities and employment information networks in area of residence contribute to actual employment outcomes by race.

When Does Location Matter for Employment Outcomes and for Whom?

To predict the probability of finding a job, we estimated a series of discrete time event history models using multinomial logistic regression to account for the competing risks discussed above. We estimated six models that predict the probability of finding employment in any felon-friendly industry and each of the five specific felon-friendly industries, which we ran separately by race.¹⁷ Our two primary independent variables were the number of local jobs per capita and the percentage of local residents employed in felon-friendly industries. Jobs per capita capture the geographic proximity to jobs themselves, whereas residents employed is a measure of potential employment information networks. For industry-specific models, the corresponding variables are the number of jobs per capita and the percentage of local residents employed in that industry. Full results from the multinomial logistic models are provided in the Appendix.

Table 5 displays the marginal effects for our primary variables: jobs in local firms and residents employed. The marginal effects are calculated with all other variables fixed at their race-specific means. Marginal effects should be interpreted as the effect of a one-unit change in the predictor on the probability of finding a job in a given calendar quarter. Although these marginal effects may appear small, they are actually quite substantial when considered relative to the overall probability of an individual in our sample finding a job (or finding a job in a specific industry) in a given calendar quarter, provided that the individual was not employed in the prior quarter (see the second panel of Table 2). We compare the marginal effects of these independent variables to those for education level to contextualize the size of the effect.¹⁸

Table 5.

Marginal Effects from Multinomial Regression: Probability of Finding Employment When Unemployed, by Race

Variable	White	Black
Model 1 (outcome: employed in felon‐friendly industry)
Jobs in felon‐friendly local firms per capita	0.0125	‐0.0363
	(‐.0401, .0656)	(‐.0727, .0001)
Residents employed in felon‐friendly industries (%)	0.0020**	0.0004
	(.0005, 0034)	(‐.0007, .0014)
Model 2 (outcome: employed in construction)
Jobs in construction industries per capita	‐0.0143	‐0.0081
	(‐.1957, .1670).	(‐0252, .0090)
Residents employed in construction (%)	0.0020*	‐0.0000
	(.0005, .0035)	(‐.0008, .0008)
Model 3 (outcome: employed in manufacturing)
Jobs in manufacturing industries per capita	‐0.0297	‐0.0185
	(‐.0181, .0776)	(‐.0602, .0231)
Residents employed in manufacturing (%)	0.0016**	0.0009**
	(.0005, .0026)	(.0003, .0015)
Model 4 (outcome: employed in retail)
Jobs in retail industries per capita	0.0750**	0.0132
	(.0195, .1306)	(‐.0288, .0551)
Residents employed in retail (%)	‐0.0007	0.0001
	(‐.0022, .0008)	(‐.0017, .0015)
Model 5 (outcome: employed in food services)
Jobs in food services industries per capita	‐0.0103	‐0.0407
	(‐.1148, .0942)	(‐.0904, .0090)
Residents employed in food services (%)	‐0.0005	0.0012
	(‐.0036, .0025)	(‐.0005, .0029)
Model 6 (outcome: employed in temporary work)
Jobs in temporary work industries per capita	‐0.1870	‐0.0145
	(‐.3770, ‐.0030)	(‐.1203, .0912)
Residents employed in temporary work (%)	‐0.0000	‐0.0018*
	(‐.00019, .0020	(‐.0035, ‐.0002)

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: The 95% confidence interval for each coefficient is displayed in parentheses. Bold indicates a statistically significant difference results for black and white parolees. Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Model 1 shows that living near people employed in felon-friendly industries slightly increased employment for whites. Whites were 0.0020 times more likely to find a job in a felon-friendly industry with a one percentage point increase in the resident felon-friendly industry workers in their zip code. In contrast, those with less than a high school diploma were significantly less likely to find work in a felon-friendly industry than their high school educated counterparts (0.0214 less likely among whites and 0.0162 less among blacks). In addition, blacks with a GED were 0.0128 less likely to find a job in a felon-friendly industry than their counterparts with a high school diploma.

Within the felon-friendly industries, there were some industry-specific associations between geography and finding a job. In the construction model (Model 2), for every percentage point increase in resident construction workers in the zip code, the probability of being employed in construction increased by 0.0020 for formerly incarcerated whites, compared to a decrease of 0.0096 for whites with less than a high school diploma rather than a diploma. For blacks, an increase in resident construction workers had no effect, and there were no differences by educational attainment. Compounded across multiple quarters of job search, this association is substantial in magnitude and substantively relevant. The probability among whites of finding a construction job is 0.025 in any quarter (Table 2). A two percentage point increase in construction workers in the zip code (approximately a one standard deviation shift in that variable) increases that probability to 0.029 (0.025 + 2 × 0.002). Cumulative over 3 years (12 quarters) that difference corresponds to a 3.6 percentage point difference in the probability of ever finding a construction job (29.8 vs. 26.2).

For manufacturing (Model 3), the number of jobs per capita did not change the probability of finding a job. However, the percent of resident manufacturing workers was associated with an increased probability of finding employment in manufacturing among both whites and blacks. As with construction, there were larger benefits for whites. The results indicate an increase in finding employment in manufacturing of 0.0009 for blacks and 0.0016 for whites. Relative to the baseline probability of finding a manufacturing job of 0.027 for whites and 0.012 for blacks, these marginal effect estimates are also substantial in magnitude. Increasing the proportion of manufacturing workers in the residential zip code by two percentage points increases the probability of finding a manufacturing job in three years by 1.9 percentage points among blacks (15.4 vs. 13.5) and by 2.7 percentage points among whites (30.7 vs, 28.0). In contrast, none of the education variables had a statistically significant effect on the likelihood of finding a job in manufacturing for whites, while holding a GED reduced the likelihood of employment in manufacturing by 0.0053 for blacks.

The only location variable that affected employment in retail (Model 4) was the number of jobs per capita for whites. One additional retail job per capita was associated with an increase in employment in retail of 0.0750 for whites. A one standard deviation change in retail jobs per capita (0.06) corresponds to an increase in the probability of employment in retail in a calendar quarter of 0.0046. Educational attainment also mattered, and its effects were about twice as large. Whites with less than a high school diploma were less likely to be employed in retail than those with a high school diploma (−0.0072).

There were no statistically significant associations between finding work in food services (Model 5) and proximity to more food services jobs or more food services workers. Educational attainment was a stronger predictor. Having less than a high school diploma decreased the likelihood of employment in food services by 0.0083 for whites and 0.0061 for blacks. Formerly incarcerated whites with a GED were also 0.0093 less likely to be employed in food services than their high school educated counterparts. The findings for food services and retail suggest that finding employment in the secondary labor market, where jobs are advertised openly, may be less sensitive to social network effects than finding employment in the primary labor market. In contrast, we found a negative association between the total number of temporary workers and finding a temporary labor job among blacks, and no influence of educational attainment (Model 6). For each percentage point increase in resident temporary labor workers, the probability of finding employment in the temporary work industry decreased by 0.0018 for blacks. This decrease translates into a 2.8 percentage point difference in a three year job search (38.7 vs. 35.9). The negative effects of living near temporary workers on formerly incarcerated blacks’ probability of finding a job in temporary labor is consistent with the worker competition hypothesis, which suggests that temporary workers have a disincentive to refer others to the same jobs for which they themselves are competing.

These results suggest that residential proximity to potential employment networks is associated with an increased probability of finding employment after prison in manufacturing and construction, the two felon-friendly industries associated with the primary labor market. However, there are substantive racial differences in the strength of these potential network effects. Whites appear to benefit much more than blacks. In contrast, employment in retail increased with more exposure to retail jobs, but for whites only.

Can Potential Worker Networks Explain Racial Differences in Post-Prison Employment?

The black-white gap in employment depends on two factors: the extent to which formerly incarcerated blacks and whites live in areas with different concentrations of workers (or jobs) and the extent to which residential proximity to workers (or jobs) affects the probability of being hired. To determine how much each contributes, we separated our estimates into these two components using Blinder-Oaxaca decomposition (Blinder 1973; Oaxaca 1973). The first reflects differences in characteristics, here the baseline black-white differences in proximity to workers or jobs (termed “endowments” in the decomposition literature). The second component reflects differences in the coefficients, or black-white differences in how much proximity to workers or jobs influences employment outcomes (i.e., the returns to proximity). The decomposition process can be understood as a counterfactual thought experiment that asks how the differences in the probability of finding a job between blacks and whites would grow or shrink if one group had either the coefficients or the characteristics of the other group.¹⁹ As described further below, we found that the black-white gap in employment is driven by differences in the characteristics of where formerly incarcerated blacks and whites live.

Table 6 shows the results of the decompositions for our core predictors of interest, proximity to jobs and proximity to workers. The “Differences in Characteristics” rows show the percent of the black-white gap accounted for by mean differences in the proximity variables by race, assuming the association between proximity and employment among whites for both races. The “Differences in Coefficients” rows show the percent of the black-white gap accounted for differential returns to proximity, assuming the white mean number of jobs and percent residents employed in felon-friendly firms for both races. We focus our discussion on the outcomes of employment in any felon-friendly industry (Model 1), and the construction, manufacturing, and retail sectors (Models 2, 3, and 4) due to consistent and statistically significant associations in the regression models and the decomposition analysis.

Table 6.

Percent of Racial Difference in Employment Probability Attributable to Racial Differences in Means and Coefficients (Selected Variables)

Variable	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Felon‐Friendly	Construction	Manufacturing	Retail	Food	Temp
Jobs in local firms per capita
Difference in characteristics (%)	0.69	‐0.43	3.44	31.47 **	‐0.40	0.27
Difference in coefficients (%)	13.97	0.01	5.24	36.08	31.13	457.59
Residents employed in industries
Difference in characteristics (%)	15.94 **	13.38 **	7.75 **	‐22.10	‐0.30	‐1.61
Difference in coefficients (%)	93.56	0.15	‐3.21	‐68.90	‐214.77	‐2737.20

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Note: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

For felon-friendly industries, 16 percent of the black-white employment gap was due to living in residential proximity to people employed in those industries, capturing that formerly incarcerated whites lived in areas with more residents working in felon-friendly industries than blacks. In the primary labor market, the industry-specific results highlight the importance of proximity to workers over returns to that proximity. Differential returns to proximity to workers do not seem to account for racial differences in either construction or manufacturing employment. Yet proximity to workers in each of those industries does. That whites live in areas with higher concentrations of construction workers than blacks do accounts for about 13 percent of the black-white gap in the probability of finding a construction job. This was driven by whites’ closer proximity to construction workers than blacks. Similarly, about 8 percent of the black-white gap in the probability of finding manufacturing employment after prison is accounted for by black-white differences in residential proximity to manufacturing workers.

Finally, there were differential returns to proximity to more jobs per capita in retail. While blacks and whites had similar exposure to jobs in local retail firms per capita, 31 percent of the difference between black and white employment in retail is explained by differences in the number of retail jobs per capita. The negative marginal effects for temporary work shown in Table 5 did not produce significant results in the decomposition analysis.

Differences in black-white employment in primary market jobs was driven by differences in their potential connections with workers in construction and manufacturing and not by racial differences in the effects that those potential connections have. Because formerly incarcerated blacks and whites returned to different communities with different resident compositions, they had exposure to distinct social networks. For formerly incarcerated whites, these networks had connections to felon-friendly jobs in the primary labor market, which contributed to their higher likelihood to find employment in construction and manufacturing.

DISCUSSION AND CONCLUSION

Among a large randomly sampled cohort of formerly incarcerated individuals in Michigan, we found that residential patterns by race produced racial differences in employment outcomes in the primary and secondary labor markets. Although the formerly incarcerated individuals in our sample generally experienced poor employment outcomes, whites were more likely than blacks to be employed and more likely to find employment in the primary labor market. Whites lived in areas with more jobs and workers from the felon-friendly industries in which the formerly incarcerated were more likely to find work, while blacks lived in urban areas with more proximity to employers and workers, but greater competition for felon-friendly jobs. Our analyses demonstrate that the geography of access to workers in felon-friendly industries contributed to these black-white differences in post-prison employment. Residential proximity to potential worker network ties was associated with an increased probability of finding post-prison employment in manufacturing and construction, mainly for whites. Our decomposition analysis confirmed that differential exposure to workers in the primary labor market, and not racial differences in responses to that exposure, helps to explain the black-white gap in construction and manufacturing employment. With regard to the secondary labor market, proximity to retail jobs increased employment in retail, but only for whites.

These findings support the literature on employment patterns among low-income people by race. Studies of networks and employment among low-income populations have found that whites benefit more from employed social ties than blacks, particularly in the primary labor market (Alexander et al. 2014; Royster 2003; Smith 2018; Waldinger 1995; Waldinger and Bailey 1991). Royster (2003) finds that whites seeking jobs in construction were more likely to receive referrals for jobs than their black counterparts. While these prior studies are qualitative and thus have small samples, we found similar racial differences in the effect of potential exposure to employed neighbors for a large, random sample. Racial differences in employment emerged in part because the predominate jobs and workers to which blacks are exposed are not in felon-friendly industries, whereas whites are more exposed to workers in felon-friendly industries.

Given residential patterns by race in the State of Michigan, it is not surprising that formerly incarcerated blacks are unable to find employment in industries in which mainly whites are employed. Blacks and whites live in such distinct areas of Michigan that they are exposed to different configurations of job opportunities and workers, a pattern that is likely to be similar in other highly segregated states. These findings suggest that higher rates of employment in felon-friendly industries among whites, coupled with their higher job referral rates, ease the transition to employment for formerly incarcerated whites while exacerbating racial disparity in post-prison employment outcomes. This pattern is likely true across low-income groups, not only the formerly incarcerated, suggesting that policies aimed at increasing employment and educational outcomes for low-income groups are also likely to benefit the formerly incarcerated.

Finally, we further research on networks and employment by providing evidence that the processes by which low-income populations secure employment differs by whether the job is in the primary, secondary, or temporary labor market. Contrary to prior work, we find that finding employment in the secondary market is less susceptible to social network effects than in the primary market (Heneman and Berkley 1999; Pettinger 2005). Neither proximity to jobs, nor to workers, explained the black-white gap in food services employment, although proximity to jobs did explain some of the black-white gap in retail employment. Additionally, in the temporary labor market, proximity to workers negatively affected employment outcomes for blacks, but not for whites, suggesting that patterns of temp workers discouraging other workers from seeking temporary employment may vary by race (Elcioglu 2010; Smith 1998).

The racial inequity we found could be addressed with place-based employment programs. For example, Cook et al. (2014) found that parolees who received subsidized employment and job placement services had an employment rate 20 percentage points higher than parolees who did not. Introducing this type of program in neighborhoods with high concentrations of parolees might address racial inequities given that blacks in our sample lived in areas with higher concentrations of parolees than their white counterparts. Providing formerly incarcerated blacks “resume building” jobs and job placement services could reduce the effects of racial differences in finding work through social networks by providing those with less social capital access to information about better jobs. Programs could ensure that this happens by creating connections to felon-friendly employers to create a pipeline for employment. There is also evidence from studies following people affected by Hurricane Katrina that relocating the formerly incarcerated from their former neighborhoods can reduce recidivism (Kirk 2009). Offering the formerly incarcerated the option to move to a neighborhood with a lower density of parolees could offer another avenue to address the black-white disparities found here. However, such a program would need to address the shortcomings found in prior evaluations including pre-identifying housing with willing landlords, providing transportation assistance, and connecting parolees who re-located with local social and employment services (Comey, Briggs, and Weismann 2008).

Although this study provides new insight into racial inequities in post-prison and low-wage employment, we are limited by the nature of our data sources. The MDOC data consists of a random sample and we use extensive controls in our analyses. However, the formerly incarcerated in our sample are not randomly assigned to zip codes. The nonrandom geographic concentration of the formerly incarcerated in economically challenged areas (Lee, Harding, and Morenoff 2017) prohibits us from making strong causal claims about the relationship between geography and post-prison employment. Our use of administrative records also means we may have omitted variable bias around individual-level factors such as motivation. This could overstate or understate our findings depending on the magnitude and direction of the effects of these unmeasured characteristics. In addition, the white sample here may include a small proportion of white Hispanics, which might mean our models underestimate black-white differences. The employment data are also limited in that they only capture formal employment and cannot account for whether sample members are working full- or part-time, each of which may underestimate employment. Additionally, our use of Census data and administrative records limits our ability to measure actual job networks. Our proxy measure, the percentage of residents employed in a particular industry, therefore, imperfectly reflects the potential for neighborhood social networks to connect the formerly incarcerated to jobs because we do not know who is connected to whom. We assume that every additional resident employed in a felon-friendly industry contributes to the network ties of the formerly incarcerated living in that zip code, thus our results could overstate the role of networks. Finally, our analysis is limited to the State of Michigan in the mid-2000s. Michigan had higher rates of segregation and racial inequality, and an economy with a higher proportion of jobs in manufacturing than most of the United States at that time. States with lower rates of segregation and less racial inequality may experience less pronounced racial inequality in opportunity. Furthermore, states that have fewer manufacturing jobs may also have fewer felon-friendly jobs, which would increase competition and could lower rates of employment among the formerly incarcerated.

Our findings suggest several potential avenues for future sociological research on studies of post-prison employment outcomes and the use of social networks in job search. Further research in the area of post-prison employment should explore racial variation including outcomes for Hispanics in the effect of spatial patterns of residence like parolee density on employment outcomes, including employment rates, wages, and full- or part-time employment. Additionally, future studies exploring the use of social networks in job searches should investigate the relationship between network composition and spatial proximity to understand how network exposure and general exposure to workers affects employment outcomes. Finally, we need more applied research aimed at addressing racial disparities in social networks.

APPENDIX

Table A1.

Employment by Industry

	Reentry Cohort Data		State of Michigan
	Blacks	Whites	Blacks	Whites
Agriculture, forestry, fishing, and hunting	0.16	0.81	0.14	0.48
Mining, quarrying, and oil and gas extraction	0.00	0.21	0.02	0.19
Construction	3.27	14.30	1.21	5.18
Whole sale trade	2.39	3.36	2.10	4.38
Information	0.46	0.62	1.91	1.60
Finance and insurance	0.46	0.27	3.53	3.71
Real estate and rental and leasin	0.68	0.99	1.24	1.36
Professional, scientific, and technical services	1.73	1.92	4.10	5.87
Mangement of companies and enterprises	0.08	0.12	1.79	1.65
Administrative and support and waste managemenet and remediation services	2.70	1.65	9.78	5.38
Educational services	1.05	0.40	8.93	9.19
Health care and social assistance	6.89	2.01	20.20	12.14
Arts, entertainment, and recreation	0.51	1.22	1.29	1.42
Accomodation and food services	3.06	2.70	8.67	7.77
Other services (except public administration)	1.59	1.27	2.59	3.39
Public administration	0.60	0.24	6.08	4.13
Manufacturing	8.33	10.70	13.55	16.63
Retail	6.90	8.35	9.48	12.53
Transportation and warehousing	1.50	2.04	2.85	2.49

Source: Reentry cohort data. State of Michigan data were retrieved from the Quarterly Workforce Indicators for 2004.

Table A2.

Descriptive Statistics of Regression Variables

	All		Blacks		Whites
Variable	Mean	Standard deviation	Mean	Standard deviation	Mean	Standard deviation
Number of jobs in zip code of residence per capita	0.47	0.87	0.51	1.07	0.43	0.54
Felon‐friendly industries	0.19	0.20	0.17	0.23	0.21	0.16
Construction	0.02	0.03	0.02	0.03	0.02	0.02
Manufacturing	0.06	0.08	0.05	0.09	0.07	0.07
Retail services	0.05	0.05	0.04	0.05	0.06	0.06
Food services	0.04	0.06	0.04	0.08	0.04	0.04
Temporary work	0.02	0.04	0.02	0.04	0.02	0.03
Proportion of residents employed in zip code of residence:
Construction	6.79	2.60	5.59	1.95	8.18	2.55
Manufacturing	3.97	1.37	4.41	1.34	3.47	1.23
Retail services	8.76	1.89	8.05	1.62	9.57	1.85
Food services	5.47	1.41	5.45	1.37	5.49	1.46
Temporary work	10.59	3.53	10.31	3.23	10.91	3.83
Average number of jobs in zip code of residence	18.21	7.05	19.47	7.16	16.75	6.63
Proportion unemployed	13.43	6.14	16.37	6.04	10.03	4.20
Disadvantage score	1.74	1.82	2.80	1.60	0.50	1.15
Population density	3,170.20	2,690.84	4,354.66	2,664.03	1,799.29	1,976.36
Density of ex‐offenders	0.76	0.75	1.00	0.80	0.47	0.58
Age at release	35.32	9.67	35.58	9.46	35.02	9.91
Proportion white, non‐Latino	0.46	0.50	0.00	0.00	1.00	0.00
Proportion female	0.08	0.27	0.08	0.27	0.08	0.26
Proportion less than high school diploma	0.42	0.49	0.47	0.50	0.37	0.48
Proportion GED	0.30	0.46	0.28	0.45	0.32	0.47
Proportion employed in year prior to incarceration	0.14	0.35	0.12	0.32	0.17	0.38
Proportion single	0.68	0.47	0.75	0.43	0.60	0.49
Proportion divorced and widowed	0.20	0.40	0.13	0.34	0.27	0.44
Proportion with dependents	0.62	0.48	0.69	0.46	0.55	0.50
Proportion with sex offender offense	0.07	0.26	0.06	0.23	0.10	0.29
Proportion with assault offense	0.28	0.45	0.29	0.46	0.27	0.44
Proportion with a drug offense	0.27	0.44	0.39	0.49	0.13	0.34
Proportion under electronic monitoring	0.01	0.12	0.01	0.10	0.02	0.14
Number of prior prison spells	0.99	1.37	1.19	1.47	0.75	1.21
Years in prison prior to 2003	2.95	3.07	3.18	3.31	2.68	2.73
Proportion with any substance abuse test history	0.48	0.50	0.49	0.50	0.47	0.50
Proportion with positive substance abuse test	0.19	0.62	0.23	0.70	0.13	0.50
Year of release
2000	0.03	NA	0.00	NA	0.07	NA
2001	1.03	NA	0.85	NA	1.24	NA
2002	6.17	NA	4.85	NA	7.70	NA
2003	92.77	NA	94.30	NA	90.99	NA
Proportion released to a center	0.10	0.31	0.09	0.28	0.12	0.33

Source: Reentry cohort data.

Table A3.

Event History Results for All Outcomes

Variable	All	White	Black
Employed
Jobs per capita	−0.1792***	−0.2138**	−0.1520**
Employed residents	0.0000	−0.0055	0.0056
Zip code characteristics
Average size employer	0.0007	−0.0008	0.0017
Disadvantage score	−0.0807	−0.1010	−0.0705
Disadvantage score of preprison neighborhood	−0.0122	0.0163	−0.0255
Population density	0.0000	0.0000	0.0000
Parolee density	0.0733	0.0309	0.0749
Parolee characteristics
Relative age	0.0263	−0.0012	0.0845***
Relative age, squared	−0.0007**	−0.0004	−0.0014***
White, non−Hispanic (ref: black, non−Hispanic)	0.2590***	NA	NA
Female (ref: male)	0.1356	0.1138	0.1295
Less than a high school diploma (ref: high school diploma)	−0.2307***	−0.1730*	−0.3017***
GED	−0.1217*	−0.0202	−0.2186**
Employed in year or quarter before incarceration	0.2808***	0.2233**	0.3587***
Single, not married (ref: married)	−0.0792	−0.1085	−0.0559
Divorced or widowed	−0.0360	−0.1293	0.1597
Has dependents	−0.0530	−0.0636	−0.0210
Charged for sex offense	0.0655	0.1306	0.0016
Charged for assault crime	0.0501	0.0248	0.0747
Electronic monitoring sentence	0.1319	0.1744	0.0013
Charged for drug offense	−0.0642	−0.0262	−0.0677
Number of prior prison spells	−0.0124	−0.0355	−0.0017
Number of years in prison	0.0400***	0.0212	0.0491***
Completed parole	0.1704***	0.0636	0.2792***
Any history of substance abuse	−0.1062*	−0.1790**	−0.0434
Positive substance abuse test in quarter	0.0249	0.1041	−0.0209
Release year	−0.0592	−0.0454	−0.1100
Release to a center	−0.0520	−0.0630	0.0163
Employment spell number	−0.1213*	−0.3081***	0.0344
Employment spell quarter	0.5935***	0.5071***	0.7542***
Employment spell quarter, squared	−0.1630***	−0.1442***	−0.2012***
Employment spell quarter, cubed	0.0122***	0.0103***	0.0158***
Employment spell quarter, quartic	−0.0003***	−0.0002***	−0.0004***
Quarter 1 (ref: Quarter 4)	−0.2154***	−0.2257**	−0.2275*
Quarter 2	−0.5804***	−0.7230***	−0.4425***
Quarter 3	−0.2114***	−0.2724***	−0.1476
2001 (ref: 2004)	−16.1110***	−14.9319***	−16.1060***
2002	−4.3031***	−3.9466***	−15.7886***
2003	0.2141***	0.1831*	0.2269**
2005	−0.0450	0.1545	−0.2163*
2006	−0.0642	0.1253	−0.2423
2007	−0.0235	0.2790	−0.3078
2008	−0.0801	0.3271	−0.4615*
2009	−0.4381*	0.0830	−1.0146**
Constant	116.8177	90.5979	216.6958
Returned to prison or dead
Jobs per capita	−0.1473**	−0.2273*	−0.0972
Employed residents	0.0203	−0.0111	0.0353*
Zip code characteristics
Average size employer	0.0021	0.0041*	−0.0001
Disadvantage score	−0.0730	−0.0215	−0.0977
Disadvantage score of preprison neighborhood	0.0554	0.1090*	0.0393
Population density	0.0000	0.0000	0.0000
Parolee density	0.1356*	0.2074	0.0673
Parolee characteristics
Relative age	0.0669*	0.0070	0.1189**
Relative age, squared	−0.0010**	−0.0003	−0.0016**
White, non−Hispanic (ref: black, non−Hispanic)	0.0414	NA	NA
Female (ref: male)	0.0722	−0.0853	0.1238
Less than a high school diploma (ref: high school diploma)	−0.0558	0.1060	−0.1846
GED	−0.0037	0.1021	−0.0949
Employed in year or quarter before incarceration	0.0397	−0.0923	0.1289
Single, not married (ref: married)	0.1860	0.0716	0.2127
Divorced or widowed	0.1457	0.0549	0.2115
Has dependents	−0.0776	−0.0174	−0.0834
Charged for sex offense	−0.3311*	−0.1817	−0.4567*
Charged for assault crime	−0.3793***	−0.3453*	−0.4034***
Electronic monitoring sentence	−0.0216	−0.2550	0.0908
Charged for drug offense	−0.3001***	−0.2218	−0.3029**
Number of prior prison spells	0.1025***	0.0780	0.1023**
Number of years in prison	−0.0735***	−0.1133**	−0.0578**
Completed parole	−4.7684***	−4.6893***	−4.8497***
Any history of substance abuse	−0.0517	−0.0681	−0.0532
Positive substance abuse test in quarter	0.1243***	0.1298	0.1194**
Release year	−0.1671	0.2625	−0.6125**
Release to a center	−0.5637*	−0.2861	−0.7588***
Employment spell number	−0.0016	0.0521	−0.0653
Employment spell quarter	0.5317***	0.5181**	0.5794***
Employment spell quarter, squared	−0.0933***	−0.0969**	−0.0973***
Employment spell quarter, cubed	0.0058***	0.0062*	0.0059***
Employment spell quarter, quartic	−0.0001***	−0.0001*	−0.0001***
Quarter 1 (ref: Quarter 4)	−0.1071	−0.1815	−0.0786
Quarter 2	−0.0628	−0.0137	−0.0950
Quarter 3	0.0130	0.1690	−0.0823
2001 (ref: 2004)	−15.3730***	−13.7541***	−15.8773***
2002	−1.4484**	−2.1153*	−1.3282*
2003	−0.6108***	−0.7785***	−0.5017***
2005	0.2753**	0.3480*	0.2318
2006	−0.0929	−0.1486	−0.0716
2007	−0.3393	−0.6610	−0.1896
2008	−0.7790**	−1.0508*	−0.6246
2009	−0.9782**	−2.0876***	−0.5281
Constant	330.5229	−528.5148	1221.6193**

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Table A4.

Event History Results for All Outcomes, Felon‐Friendly Industries

Variable	All	White	Black
Employed in felon‐friendly industries
Jobs in felon‐friendy industries per capita	−0.3126	0.0999	−0.5893*
Residents employed in felon‐friendly industries	0.0122*	0.0236**	0.0049
Zip code characteristics
Average size employer	0.0007	−0.0008	0.0016
Disadvantage score	−0.0575*	−0.1302***	−0.0153
Disadvantage score of preprison neighborhood	−0.0237	0.0102	−0.0480
Population density	0.0000	0.0000	0.0000
Parolee density	−0.0524	−0.0058	−0.0651
Parolee characteristics
Relative age	0.0347	−0.0051	0.1028***
Relative age, squared	−0.0008**	−0.0004	−0.0017***
White, non‐Hispanic (ref: black, non‐Hispanic)	0.1423*	NA	NA
Female (ref: male)	0.0657	0.1063	0.0117
Less than a high school diploma (ref: high school diploma)	−0.2406***	−0.2218*	−0.2770***
GED	−0.1339*	−0.0520	−0.2148*
Employed in year or quarter before incarceration	0.2458***	0.0891	0.4157***
Single, not married (ref: married)	−0.0977	−0.1347	−0.0710
Divorced or widowed	0.0208	−0.0807	0.2015
Has dependents	−0.0739	−0.0389	−0.0682
Charged for sex offense	0.1188	0.1523	0.1351
Charged for assault crime	0.0241	0.0122	0.0123
Electronic monitoring sentence	0.0167	0.1407	−0.2415
Charged for drug offense	−0.0873	−0.0632	−0.1075
Number of prior prison spells	−0.0105	−0.0494	0.0063
Number of years in prison	0.0337***	0.0263*	0.0382***
Completed parole	0.1144*	−0.0508	0.2841***
Any history of substance abuse	−0.0888	−0.1363	−0.0476
Positive substance abuse test in quarter	0.0373	0.1283*	−0.0227
Release year	−0.1176	−0.1367	−0.1070
Release to a center	−0.0686	−0.0483	0.0101
Employment spell number	−0.0153	−0.2248**	0.1551*
Employment spell quarter	0.5057***	0.4732**	0.5812***
Employment spell quarter, squared	−0.1450***	−0.1441***	−0.1546***
Employment spell quarter, cubed	0.0110***	0.0108**	0.0119***
Employment spell quarter, quartic	−0.0003***	−0.0002*	−0.0003***
Quarter 1 (ref: Quarter 4)	−0.2182**	−0.2764**	−0.1752
Quarter 2	−0.5647***	−0.7732***	−0.3622***
Quarter 3	−0.1887**	−0.3155***	−0.0623
2001 (ref: 2004)	−17.9741***	−16.8060***	−16.0339***
2002	−17.6284***	−16.4292***	−15.6712***
2003	0.2234***	0.1636	0.2752**
2005	−0.0956	0.1280	−0.2914**
2006	−0.1382	0.0830	−0.3384*
2007	−0.2385	0.0053	−0.4415*
2008	−0.2580	0.1542	−0.6553**
2009	−0.7206**	−0.1052	−1.4169***
Constant	233.0069	272.1105	210.3604
Employed in other industry
Jobs in felon‐friendy industries per capita	−0.4202	−0.2298	−0.5206
Residents employed in felon‐friendly industries	0.0040	0.0068	0.0000
Zip code characteristics
Average size employer	0.0003	−0.0033	0.0041
Disadvantage score	−0.1030*	−0.1215	−0.1098
Disadvantage score of preprison neighborhood	−0.0008	−0.0051	0.0070
Population density	0.0000	0.0000	0.0000
Parolee density	0.1497	−0.1087	0.2423*
Parolee characteristics
Relative age	0.0150	0.0189	0.0473
Relative age, squared	−0.0003	−0.0005	−0.0007
White, non‐Hispanic (ref: black, non‐Hispanic)	0.5139***	NA	NA
Female (ref: male)	0.2548	0.1540	0.4075
Less than a high school diploma (ref: high school diploma)	−0.0964	0.1587	−0.3751*
GED	0.0056	0.2188	−0.2321
Employed in year or quarter before incarceration	0.3306**	0.4947***	0.0425
Single, not married (ref: married)	−0.1051	−0.1109	−0.0816
Divorced or widowed	−0.2592	−0.3930	0.0874
Has dependents	−0.0316	−0.2031	0.1533
Charged for sex offense	−0.1784	0.1329	−1.0091*
Charged for assault crime	0.2057	−0.0213	0.5417**
Electronic monitoring sentence	0.5463	0.3063	0.8721
Charged for drug offense	0.1230	0.0158	0.3752
Number of prior prison spells	−0.0251	−0.0101	−0.0285
Number of years in prison	0.0672***	0.0459	0.0836***
Completed parole	0.4441***	0.5387***	0.2630
Any history of substance abuse	−0.2346*	−0.3796**	−0.1060
Positive substance abuse test in quarter	−0.0214	0.0172	−0.0171
Release year	0.1478	0.0419	0.1620
Release to a center	0.1408	−0.1558	0.3763
Employment spell number	−0.2266*	−0.3435**	−0.1515
Employment spell quarter	1.9417***	1.7557***	2.3580***
Employment spell quarter, squared	−0.4123***	−0.3476***	−0.5443***
Employment spell quarter, cubed	0.0290***	0.0230***	0.0414***
Employment spell quarter, quartic	−0.0007***	−0.0005***	−0.0010***
Quarter 1 (ref: Quarter 4)	−0.3219*	−0.1975	−0.5037*
Quarter 2	−0.4343***	−0.3156	−0.5999**
Quarter 3	−0.2402*	−0.1305	−0.3807*
2001 (ref: 2004)	−17.2445***	−15.5110***	−16.0005***
2002	−17.2706***	−15.5388***	−15.8999***
2003	0.2167	0.4542*	−0.0784
2005	0.3071*	0.4547*	0.1998
2006	0.2652	0.4458	0.1150
2007	0.6498**	1.0991***	0.0583
2008	0.5138	0.9608*	0.0412
2009	0.4974	0.8726	0.0786
Constant	−302.1549	−89.4931	−331.7835
Returned to prison or dead
Jobs in felon‐friendy industries per capita	−0.2792	−0.4776	−0.0047
Residents employed in felon‐friendly industries	0.0093	0.0355**	−0.0109
Zip code characteristics
Average size employer	0.0020	0.0049**	−0.0007
Disadvantage score	0.0335	−0.0566	0.0686
Disadvantage score of preprison neighborhood	0.0556	0.1129*	0.0342
Population density	0.0000	0.0000	0.0000
Parolee density	0.0450	0.2028	−0.0751
Parolee characteristics
Relative age	0.0680*	−0.0054	0.1185**
Relative age, squared	−0.0010**	−0.0001	−0.0016**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0407	NA	NA
Female (ref: male)	0.0402	−0.0977	0.0906
Less than a high school diploma (ref: high school diploma)	−0.0597	0.0836	−0.1931
GED	−0.0093	0.0801	−0.0921
Employed in year or quarter before incarceration	0.0488	−0.0845	0.1455
Single, not married (ref: married)	0.1881	0.0689	0.2155
Divorced or widowed	0.1495	0.0380	0.2104
Has dependents	−0.0872	−0.0163	−0.0808
Charged for sex offense	−0.3496*	−0.2011	−0.4770*
Charged for assault crime	−0.3809***	−0.3808*	−0.3925**
Electronic monitoring sentence	−0.0349	−0.2558	0.0887
Charged for drug offense	−0.2958***	−0.2054	−0.2863**
Number of prior prison spells	0.1034***	0.0908*	0.1049**
Number of years in prison	−0.0745***	−0.1119**	−0.0599**
Completed parole	−4.7679***	−4.6884***	−4.8565***
Any history of substance abuse	−0.0529	−0.0572	−0.0585
Positive substance abuse test in quarter	0.1174**	0.1081	0.1142*
Release year	−0.1513	0.2530	−0.5734**
Release to a center	−0.5433*	−0.2791	−0.7068***
Employment spell number	−0.0289	0.0830	−0.1271
Employment spell quarter	0.4823***	0.4540**	0.5438***
Employment spell quarter, squared	−0.0824***	−0.0807*	−0.0894***
Employment spell quarter, cubed	0.0051***	0.0051*	0.0053***
Employment spell quarter, quartic	−0.0001***	−0.0001	−0.0001**
Quarter 1 (ref: Quarter 4)	−0.1078	−0.1559	−0.0927
Quarter 2	−0.0699	−0.0124	−0.1108
Quarter 3	0.0101	0.1672	−0.0896
2001 (ref: 2004)	−17.2033***	−15.4485***	−15.9545***
2002	−1.5082**	−2.1768*	−1.3811*
2003	−0.6516***	−0.7950***	−0.5421***
2005	0.3008**	0.3374*	0.2686*
2006	−0.0257	−0.1649	0.0243
2007	−0.2097	−0.6469	−0.0234
2008	−0.5582*	−0.9668*	−0.3484
2009	−0.7269*	−1.9738***	−0.2226
Constant	298.6783	−511.0736	1144.2477**

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Table A5.

Event History Results for All Outcomes, Construction

Variable	All	White	Black
Employed in construction
Jobs in construction per capita	−0.9331	−0.9537	−3.8830
Residents employed in construction	0.0672*	0.0892**	−0.0174
Zip code characteristics
Average size employer	−0.0018	0.0074	−0.0084
Disadvantage score	−0.3142***	−0.4298***	0.0508
Disadvantage score of preprison neighborhood	0.0049	0.0560	−0.2310
Population density	0.0001	0.0001	0.0000
Parolee density	0.0315	−0.0486	0.2822
Parolee characteristics
Relative age	0.2488***	0.2290***	0.5336*
Relative age, squared	−0.0039***	−0.0038***	−0.0072**
White, non‐Hispanic (ref: black, non‐Hispanic)	1.7850***	NA	NA
Female (ref: male)	−1.5527**	−1.4111**	−20.5203***
Less than a high school diploma (ref: high school diploma)	−0.4270*	−0.4097*	−0.3419
GED	−0.1070	−0.1043	0.1395
Employed in year or quarter before incarceration	0.2104	0.2632	−0.1382
Single, not married (ref: married)	0.1837	0.1942	0.0664
Divorced or widowed	0.6434*	0.6861*	0.3182
Has dependents	0.0697	0.0540	0.3771
Charged for sex offense	−0.0750	−0.0619	0.3398
Charged for assault crime	0.0362	−0.0844	0.9218
Electronic monitoring sentence	0.3467	0.2782	0.7979
Charged for drug offense	0.3299	0.3849	0.6064
Number of prior prison spells	−0.2211**	−0.2279**	−0.1997
Number of years in prison	0.0209	0.0128	0.0074
Completed parole	−0.0115	−0.0864	0.2068
Any history of substance abuse	0.1511	0.1680	−0.0439
Positive substance abuse test in quarter	−0.0365	0.0775	−19.4122***
Release year	−0.3700	−0.6094*	0.9408
Release to a center	−0.5976	−0.9570**	1.1547*
Employment spell number	0.2711	0.2830	0.1569
Employment spell quarter	1.0007***	1.0957***	6.5574**
Employment spell quarter, squared	−0.2585***	−0.2830***	−2.8228**
Employment spell quarter, cubed	0.0200***	0.0222***	0.4551**
Employment spell quarter, quartic	−0.0005***	−0.0005***	−0.0246**
Quarter 1 (ref: Quarter 4)	−0.3465	−0.2468	−0.9441
Quarter 2	−0.3307	−0.3236	−0.4210
Quarter 3	−0.0525	0.0524	−0.8870
2001 (ref: 2004)	−16.9063***	−17.6368***	−21.3915***
2002	−16.4388***	−16.9928***	−21.5371***
2003	0.3465*	0.4437*	−0.4381
2005	0.2001	0.2287	0.0453
2006	−0.2147	−0.2069	−0.1832
2007	−0.5917	−0.6215	−0.5429
2008	−0.3728	−0.5361	0.6164
2009	−0.5496	−0.4906	−15.1656***
Constant	730.5982	1211.9127*	−1903.8104
Employed in other industry
Jobs in construction per capita	−1.5119	−0.6415	−2.6212
Residents employed in construction	0.0188	0.0339*	0.0095
Zip code characteristics
Average size employer	0.0005	0.0120	0.0002
Disadvantage score	−0.0157	−0.0432	−0.0127
Disadvantage score of preprison neighborhood	−0.0218	0.0090	−0.0404
Population density	0.0000	0.0000	0.0000
Parolee density	−0.0777	−0.1679	−0.0265
Parolee characteristics
Relative age	0.0069	−0.0204	0.0638*
Relative age, squared	−0.0004	0.0000	−0.0011**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0495	NA	NA
Female (ref: male)	0.1743	0.2393	0.0816
Less than a high school diploma (ref: high school diploma)	−0.1930**	−0.0985	−0.2724**
GED	−0.1210	0.0823	−0.3030**
Employed in year or quarter before incarceration	0.2654***	0.2128*	0.3188**
Single, not married (ref: married)	−0.1486	−0.1021	−0.1808
Divorced or widowed	−0.1273	−0.2942*	0.1610
Has dependents	−0.0478	−0.0414	−0.0296
Charged for sex offense	0.1287	0.2538	−0.0124
Charged for assault crime	0.0659	0.0277	0.1074
Electronic monitoring sentence	−0.0587	0.1112	−0.3761
Charged for drug offense	−0.0452	−0.0920	−0.0089
Number of prior prison spells	−0.0017	−0.0315	0.0069
Number of years in prison	0.0486***	0.0349*	0.0561***
Completed parole	0.2521***	0.2220*	0.2819**
Any history of substance abuse	−0.1477**	−0.2291**	−0.0756
Positive substance abuse test in quarter	0.0249	0.1377*	−0.0290
Release year	−0.0408	0.0626	−0.2104
Release to a center	0.2644*	0.3123*	0.2112
Employment spell number	0.0236	−0.1006	0.1481
Employment spell quarter	1.6328***	1.6614***	1.6393***
Employment spell quarter, squared	−0.3208***	−0.3225***	−0.3266***
Employment spell quarter, cubed	0.0209***	0.0206***	0.0219***
Employment spell quarter, quartic	−0.0004***	−0.0004***	−0.0005***
Quarter 1 (ref: Quarter 4)	−0.4030***	−0.3272**	−0.4759***
Quarter 2	−0.3631***	−0.4333***	−0.3046**
Quarter 3	−0.1716*	−0.2151*	−0.1384
2001 (ref: 2004)	−16.8664***	−17.2740***	−22.1624***
2002	−16.7823***	−17.2261***	−21.9462***
2003	0.2568**	0.3423**	0.1659
2005	0.1989*	0.3589**	0.0560
2006	0.3057*	0.3411*	0.2202
2007	0.2248	0.3427	0.0411
2008	−0.0142	0.2386	−0.2860
2009	−0.3758	−0.1761	−0.5981
Constant	77.7757	−128.9665	416.2883
Returned to prison or dead
Jobs in construction per capita	−3.0022*	−5.5192*	−2.1788
Residents employed in construction	0.0074	0.0475*	−0.0274
Zip code characteristics
Average size employer	0.0012	0.0218*	−0.0006
Disadvantage score	0.0253	−0.0861	0.0623
Disadvantage score of preprison neighborhood	0.0607	0.0822	0.0585
Population density	0.0000	0.0000	0.0000
Parolee density	0.0253	0.0842	−0.0229
Parolee characteristics
Relative age	0.0651*	0.0025	0.1110**
Relative age, squared	−0.0010**	−0.0003	−0.0016**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0647	NA	NA
Female (ref: male)	0.0532	−0.0458	0.0782
Less than a high school diploma (ref: high school diploma)	−0.0665	0.0509	−0.2126
GED	−0.0057	0.1060	−0.1089
Employed in year or quarter before incarceration	−0.0152	−0.0761	0.0240
Single, not married (ref: married)	0.2266	0.1487	0.2248
Divorced or widowed	0.1945	0.1169	0.2377
Has dependents	−0.0913	−0.0280	−0.0792
Charged for sex offense	−0.2943	−0.1235	−0.4379*
Charged for assault crime	−0.3613***	−0.3501*	−0.3918**
Electronic monitoring sentence	−0.0898	−0.2610	−0.0116
Charged for drug offense	−0.3025***	−0.2310	−0.3082**
Number of prior prison spells	0.1043***	0.0657	0.1130***
Number of years in prison	−0.0717***	−0.1079**	−0.0566**
Completed parole	−4.9934***	−5.3385***	−4.7867***
Any history of substance abuse	−0.0316	−0.0898	−0.0274
Positive substance abuse test in quarter	0.1153**	0.1342	0.1110*
Release year	−0.1367	0.3571	−0.6369**
Release to a center	−0.4740	−0.2183	−0.6564**
Employment spell number	−0.1176	0.0520	−0.3626*
Employment spell quarter	0.6080***	0.5801***	0.6548***
Employment spell quarter, squared	−0.1054***	−0.1012**	−0.1117***
Employment spell quarter, cubed	0.0065***	0.0064*	0.0066***
Employment spell quarter, quartic	−0.0001***	−0.0001*	−0.0001***
Quarter 1 (ref: Quarter 4)	−0.0889	−0.1278	−0.1049
Quarter 2	−0.0272	0.0648	−0.0975
Quarter 3	0.0335	0.2241	−0.0887
2001 (ref: 2004)	−16.4178***	−16.4266***	−22.1299***
2002	−1.4297**	−2.0105	−1.3819*
2003	−0.5780***	−0.6949***	−0.5196***
2005	0.3436***	0.3534*	0.3618**
2006	0.1330	−0.0990	0.3277
2007	−0.1346	−0.7286*	0.2696
2008	−0.5157*	−1.0964**	−0.0307
2009	−0.6848*	−2.0514***	0.1313
Constant	269.6995	−718.5729	1271.2302**

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Table A6.

Event History Results for All Outcomes, Manufacturing

Variable	All	White	Black
Employed in manufacturing
Jobs in manufacturing per capita	−0.4126	1.2838	−1.5698
Residents employed in manufacturing	0.0674***	0.0652**	0.0746**
Zip code characteristics
Average size employer	−0.0004	−0.0036	0.0009
Disadvantage score	−0.0361	−0.0405	−0.0408
Disadvantage score of preprison neighborhood	−0.0061	−0.0589	0.0839
Population density	0.0000	0.0000	0.0000
Parolee density	0.0114	−0.0314	0.0289
Parolee characteristics
Relative age	0.0655	0.0118	0.1960*
Relative age, squared	−0.0011*	−0.0003	−0.0030**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.5695***	NA	NA
Female (ref: male)	−0.6736**	−0.6623	−0.6699
Less than a high school diploma (ref: high school diploma)	−0.1385	0.0342	−0.2941
GED	−0.2207	−0.0613	−0.4687*
Employed in year or quarter before incarceration	0.2425	0.1618	0.3538
Single, not married (ref: married)	−0.3065*	−0.1756	−0.5417*
Divorced or widowed	−0.1550	−0.2937	0.1914
Has dependents	0.0092	0.0985	−0.0458
Charged for sex offense	−0.0515	−0.0596	0.0467
Charged for assault crime	0.0367	−0.0196	0.1270
Electronic monitoring sentence	0.1248	0.1032	0.2196
Charged for drug offense	0.0860	−0.3138	0.3390
Number of prior prison spells	0.0549	0.1078	−0.0381
Number of years in prison	0.0138	0.0285	−0.0094
Completed parole	0.1829	−0.0325	0.5034**
Any history of substance abuse	−0.0980	−0.1107	−0.0957
Positive substance abuse test in quarter	−0.1048	0.0973	−0.4575*
Release year	−0.3761	0.0150	−1.1036*
Release to a center	−0.2338	0.0402	−0.9240
Employment spell number	0.1848	0.2144	−0.0067
Employment spell quarter	0.0331	0.0625	0.4038
Employment spell quarter, squared	−0.0335	−0.0295	−0.1676
Employment spell quarter, cubed	0.0021	0.0017	0.0160
Employment spell quarter, quartic	0.0000	0.0000	−0.0005
Quarter 1 (ref: Quarter 4)	−0.1528	−0.0716	−0.3213
Quarter 2	−0.6342***	−0.8128***	−0.4121
Quarter 3	0.0422	−0.1533	0.3025
2001 (ref: 2004)	−18.3922***	−15.7348***	−17.9869***
2002	−17.8443***	−15.4885***	−16.6962***
2003	0.2840*	0.4243*	0.0762
2005	−0.4483**	−0.3347	−0.5490*
2006	−0.4024	−0.2309	−0.7915
2007	−0.8686*	−1.1986*	−0.3599
2008	−0.4589	−1.1237	0.4474
2009	−1.5200*	−2.0829*	−0.6975
Constant	748.4179	−33.7870	2203.6321*
Employed in other industry
Jobs in manufacturing per capita	0.2177	0.9760	−0.3982
Residents employed in manufacturing	0.0054	0.0057	0.0032
Zip code characteristics
Average size employer	0.0003	0.0000	0.0003
Disadvantage score	−0.0811**	−0.1692***	−0.0127
Disadvantage score of preprison neighborhood	−0.0070	0.0603	−0.0537
Population density	0.0000	0.0000	0.0000
Parolee density	−0.0558	−0.1096	−0.0432
Parolee characteristics
Relative age	0.0249	0.0097	0.0638*
Relative age, squared	−0.0006*	−0.0005	−0.0011*
White, non‐Hispanic (ref: black, non‐Hispanic)	0.1484*	NA	NA
Female (ref: male)	0.1889	0.2193	0.1253
Less than a high school diploma (ref: high school diploma)	−0.1800**	−0.1763	−0.1840*
GED	−0.0799	0.0225	−0.1726
Employed in year or quarter before incarceration	0.3004***	0.3013***	0.3286**
Single, not married (ref: married)	−0.1176	−0.1248	−0.1320
Divorced or widowed	−0.0525	−0.1213	0.1019
Has dependents	−0.0906	−0.1040	−0.0402
Charged for sex offense	0.0482	0.1445	−0.0102
Charged for assault crime	0.0728	0.0258	0.0918
Electronic monitoring sentence	−0.0764	−0.0164	−0.1779
Charged for drug offense	−0.0704	0.0652	−0.1144
Number of prior prison spells	−0.0615*	−0.1369**	−0.0209
Number of years in prison	0.0486***	0.0280	0.0617***
Completed parole	0.2151***	0.1603	0.2475**
Any history of substance abuse	−0.1254*	−0.2124**	−0.0580
Positive substance abuse test in quarter	0.0164	0.1314	−0.0444
Release year	0.0530	0.0594	0.0525
Release to a center	0.2258*	0.1274	0.4015*
Employment spell number	0.1125	0.0827	0.1554
Employment spell quarter	1.3215***	1.2625***	1.5884***
Employment spell quarter, squared	−0.2452***	−0.2281***	−0.3121***
Employment spell quarter, cubed	0.0147**	0.0131**	0.0205***
Employment spell quarter, quartic	−0.0003*	−0.0002*	−0.0004***
Quarter 1 (ref: Quarter 4)	−0.4328***	−0.3851***	−0.4868***
Quarter 2	−0.3608***	−0.3559***	−0.3710***
Quarter 3	−0.1906**	−0.1454	−0.2379*
2001 (ref: 2004)	−17.2449***	−15.0895***	−15.7198***
2002	−17.2277***	−15.0880***	−15.6762***
2003	0.2751***	0.4070***	0.1518
2005	0.1929*	0.3398**	0.0900
2006	0.2525*	0.1774	0.2951
2007	0.2867	0.3187	0.1413
2008	−0.1098	0.1143	−0.4000
2009	−0.5312	−0.4161	−0.5874
Constant	−110.2488	−122.4952	−110.2992
Returned to prison or dead
Jobs in manufacturing per capita	0.7082	0.0359	0.8757
Residents employed in manufacturing	0.0163	0.0554**	−0.0170
Zip code characteristics
Average size employer	0.0005	0.0018	0.0001
Disadvantage score	0.0213	−0.1342*	0.0938*
Disadvantage score of preprison neighborhood	0.0663	0.1140*	0.0439
Population density	0.0000	0.0001	0.0000
Parolee density	−0.0079	*0.1046	−0.1021
Parolee characteristics
Relative age	0.0607*	−0.0017	0.1033*
Relative age, squared	−0.0009*	−0.0002	−0.0014**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0653	NA	NA
Female (ref: male)	0.0458	−0.0745	0.0923
Less than a high school diploma (ref: high school diploma)	−0.0649	0.1071	−0.2210
GED	−0.0129	0.0738	−0.1050
Employed in year or quarter before incarceration	−0.0269	−0.1520	0.0544
Single, not married (ref: married)	0.1931	0.1030	0.2021
Divorced or widowed	0.1479	0.0722	0.1809
Has dependents	−0.1034	−0.0601	−0.0808
Charged for sex offense	−0.2848	−0.1472	−0.3995
Charged for assault crime	−0.3936***	−0.4126*	−0.4079**
Electronic monitoring sentence	−0.1006	−0.3066	0.0228
Charged for drug offense	−0.2961**	−0.2080	−0.3019**
Number of prior prison spells	0.1080***	0.0801	0.1139***
Number of years in prison	−0.0710***	−0.1007**	−0.0592**
Completed parole	−4.9804***	−5.3173***	−4.7819***
Any history of substance abuse	−0.0360	−0.0387	−0.0569
Positive substance abuse test in quarter	0.1039**	0.0836	0.1060*
Release year	−0.1337	0.3484	−0.6061**
Release to a center	−0.4683	−0.1689	−0.6773***
Employment spell number	−0.1311	−0.0535	−0.2337
Employment spell quarter	0.6283***	0.5889***	0.6740***
Employment spell quarter, squared	−0.1062***	−0.0987**	−0.1135***
Employment spell quarter, cubed	0.0064***	0.0060*	0.0067***
Employment spell quarter, quartic	−0.0001***	−0.0001*	−0.0001***
Quarter 1 (ref: Quarter 4)	−0.0715	−0.1344	−0.0526
Quarter 2	−0.0426	−0.0111	−0.0687
Quarter 3	0.0504	0.2178	−0.0547
2001 (ref: 2004)	−17.0364***	−14.5283***	−15.9410***
2002	−1.4159**	−2.0264	−1.3165*
2003	−0.5738***	−0.7082***	−0.4851**
2005	0.3595***	0.3927**	0.3462**
2006	0.1638	0.0137	0.2488
2007	−0.0732	−0.4769	0.1362
2008	−0.4316	−0.7800*	−0.1971
2009	−0.7716*	−1.9799***	−0.1924
Constant	263.6011	−701.3390	1209.4569**

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Table A7.

Event History Results for All Outcomes, Retail

Variable	All	White	Black
Employed in retail
Jobs in retail per capita	4.0203*	5.9267*	1.6751
Residents employed in retail	−0.0525	−0.0539	−0.0127
Zip code characteristics
Average size employer	−0.0160	−0.0333	0.0269
Disadvantage score	−0.2253**	−0.3497**	−0.0566
Disadvantage score of preprison neighborhood	−0.0482	−0.0161	−0.0984
Population density	0.0001**	0.0000	0.0002**
Parolee density	−0.2181	−0.0588	−0.1904
Parolee characteristics
Relative age	−0.0147	−0.0678	0.1320
Relative age, squared	−0.0003	0.0004	−0.0022
White, non‐Hispanic (ref: black, non‐Hispanic)	−0.0089	NA	NA
Female (ref: male)	0.7631***	0.8221**	0.7463*
Less than a high school diploma (ref: high school diploma)	−0.6315***	−0.6042*	−0.6859***
GED	−0.4756*	−0.0948	−1.0245***
Employed in year or quarter before incarceration	0.0258	−0.0751	0.2003
Single, not married (ref: married)	−0.1946	−0.1259	−0.2667
Divorced or widowed	−0.0368	0.0805	−0.1013
Has dependents	−0.2841	−0.3182	−0.2343
Charged for sex offense	−0.2589	−0.8782	0.3134
Charged for assault crime	−0.0856	−0.0832	−0.0375
Electronic monitoring sentence	−0.2933	−0.0682	−0.7418
Charged for drug offense	−0.4550*	−0.5585	−0.3975
Number of prior prison spells	−0.1153	−0.2515*	−0.0569
Number of years in prison	0.0369	0.0674*	0.0128
Completed parole	0.3294*	0.1312	0.5307*
Any history of substance abuse	−0.1296	−0.4151	0.1476
Positive substance abuse test in quarter	0.0270	0.1197	−0.0174
Release year	0.0885	0.3006	−0.0504
Release to a center	−0.0028	−0.1341	0.2764
Employment spell number	−0.1050	0.0827	−0.4815
Employment spell quarter	0.2740	0.1317	0.4967
Employment spell quarter, squared	−0.1232	−0.1052	−0.1625
Employment spell quarter, cubed	0.0123	0.0133	0.0130
Employment spell quarter, quartic	−0.0004	−0.0005	−0.0003
Quarter 1 (ref: Quarter 4)	−0.3388	−0.2077	−0.4977
Quarter 2	−1.1176***	−1.0102***	−1.2739***
Quarter 3	−0.9667***	−0.9262***	−1.0485***
2001 (ref: 2004)	−16.5060***	−15.8812***	−16.1076***
2002	−16.3931***	−16.0351***	−15.8424***
2003	0.4337*	0.5079*	0.3384
2005	−0.3053	−0.4012	−0.1347
2006	−0.3620	−1.0851	0.4927
2007	−0.3714	−0.5272	−0.0291
2008	−0.3602	−1.1698	0.7455
2009	−0.4142	−1.1738	0.8100
Constant	−178.7961	−602.5637	95.8041
Employed in other industry
Jobs in retail per capita	−1.0678	−1.7962	0.0259
Residents employed in retail	0.0037	0.0278	−0.0216
Zip code characteristics
Average size employer	0.0019	0.0042	−0.0087
Disadvantage score	−0.0366	−0.0685	−0.0237
Disadvantage score of preprison neighborhood	−0.0167	0.0295	−0.0419
Population density	0.0000	0.0000	0.0000
Parolee density	−0.0725	−0.1226	−0.0910
Parolee characteristics
Relative age	0.0306	0.0139	0.0750*
Relative age, squared	−0.0007**	−0.0005	−0.0012**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.2333**	NA	NA
Female (ref: male)	−0.0470	−0.0386	−0.0906
Less than a high school diploma (ref: high school diploma)	−0.1686**	−0.0985	−0.2354*
GED	−0.0722	0.0560	−0.1993
Employed in year or quarter before incarceration	0.2760***	0.2433**	0.3258**
Single, not married (ref: married)	−0.1234	−0.1387	−0.1387
Divorced or widowed	−0.0253	−0.1680	0.2246
Has dependents	−0.0432	−0.0418	−0.0194
Charged for sex offense	0.1117	0.2195	0.0000
Charged for assault crime	0.0722	0.0349	0.1025
Electronic monitoring sentence	0.0000	0.0988	−0.2024
Charged for drug offense	0.0082	0.0409	0.0184
Number of prior prison spells	−0.0192	−0.0450	−0.0065
Number of years in prison	0.0414***	0.0195	0.0543***
Completed parole	0.2237***	0.1579	0.2818**
Any history of substance abuse	−0.1014	−0.1500*	−0.0642
Positive substance abuse test in quarter	−0.0082	0.0972	−0.0698
Release year	−0.1209	−0.0929	−0.1895
Release to a center	0.1207	0.0675	0.1999
Employment spell number	0.1317	0.0621	0.1782
Employment spell quarter	1.4337***	1.4800***	1.5585***
Employment spell quarter, squared	−0.2705***	−0.2740***	−0.3090***
Employment spell quarter, cubed	0.0166**	0.0162**	0.0206***
Employment spell quarter, quartic	−0.0003*	−0.0003*	−0.0004***
Quarter 1 (ref: Quarter 4)	−0.4083***	−0.3553**	−0.4734***
Quarter 2	−0.2785***	−0.3300***	−0.2343*
Quarter 3	−0.0749	−0.0587	−0.0921
2001 (ref: 2004)	−16.8799***	−16.7344***	−16.2904***
2002	−16.7693***	−16.6388***	−16.0945***
2003	0.2643***	0.4071***	0.1075
2005	0.1920*	0.4000***	0.0188
2006	0.2484*	0.3464*	0.1185
2007	0.1624	0.2296	−0.0089
2008	−0.1275	0.1018	−0.3227
2009	−0.6095*	−0.4044	−0.7679
Constant	237.8941	181.9863	374.6359
Returned to prison or dead
Jobs in retail per capita	−0.1306	−0.7634	0.5665
Residents employed in retail	−0.0539*	−0.0737	−0.0442
Zip code characteristics
Average size employer	0.0209*	0.0327**	−0.0012
Disadvantage score	0.0432	−0.0578	0.0595
Disadvantage score of preprison neighborhood	0.0721*	0.1115*	0.0582
Population density	0.0000	0.0001	0.0000
Parolee density	−0.0496	0.0177	−0.1096
Parolee characteristics
Relative age	0.0665*	0.0073	0.1131**
Relative age, squared	−0.0010**	−0.0003	−0.0016**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.1172	NA	NA
Female (ref: male)	0.0566	−0.0573	0.0804
Less than a high school diploma (ref: high school diploma)	−0.0752	0.0709	−0.2111
GED	−0.0238	0.0747	−0.1126
Employed in year or quarter before incarceration	−0.0303	−0.1295	0.0391
Single, not married (ref: married)	0.2050	0.1028	0.2227
Divorced or widowed	0.1864	0.0989	0.2462
Has dependents	−0.1100	−0.0686	−0.0897
Charged for sex offense	−0.3124	−0.0971	−0.4455*
Charged for assault crime	−0.3776***	−0.3827*	−0.3914**
Electronic monitoring sentence	−0.1472	−0.3450	−0.0228
Charged for drug offense	−0.2905**	−0.2383	−0.2973**
Number of prior prison spells	0.1004***	0.0566	0.1083**
Number of years in prison	−0.0701***	−0.1089**	−0.0571**
Completed parole	−4.9906***	−5.3431***	−4.7809***
Any history of substance abuse	−0.0350	−0.0684	−0.0341
Positive substance abuse test in quarter	0.1124**	0.1067	0.1087*
Release year	−0.2045	0.2397	−0.6439**
Release to a center	−0.4603	−0.2101	−0.6425**
Employment spell number	−0.3017*	−0.1968	−0.4158*
Employment spell quarter	0.5922***	0.6074***	0.6084***
Employment spell quarter, squared	−0.1013***	−0.1057**	−0.1021***
Employment spell quarter, cubed	0.0060***	0.0065**	0.0059***
Employment spell quarter, quartic	−0.0001***	−0.0001*	−0.0001***
Quarter 1 (ref: Quarter 4)	−0.1316	−0.1680	−0.1303
Quarter 2	−0.0653	0.0018	−0.1088
Quarter 3	0.0190	0.2079	−0.0950
2001 (ref: 2004)	−16.6422***	−16.2127***	−16.3719***
2002	−1.5772**	−2.1925*	−1.4425**
2003	−0.6165***	−0.7297***	−0.5529***
2005	0.4056***	0.4514**	0.3809**
2006	0.2779	0.0930	0.3801
2007	0.1238	−0.4270	0.3984
2008	−0.1744	−0.6749	0.1469
2009	−0.3661	−1.5905**	0.2365
Constant	406.0350	−482.6147	1285.6840**

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Table A8.

Event History Results for All Outcomes, Food Services

Variable	All	White	Black
Employed in food services
Jobs in food services per capita	−2.3551	−0.9640	−2.6739
Residents employed in food services	0.0284	−0.0184	0.0708
Zip code characteristics
Average size employer	−0.0005	−0.0015	−0.0035
Disadvantage score	−0.1355**	−0.2732***	−0.0273
Disadvantage score of preprison neighborhood	0.0459	0.1980**	−0.0475
Population density	0.0000	0.0000	0.0000
Parolee density	−0.0355	0.1880	−0.2592
Parolee characteristics
Relative age	−0.0043	−0.0438	0.1016
Relative age, squared	−0.0006	−0.0003	−0.0018
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0917	NA	NA
Female (ref: male)	0.8871***	1.0677***	0.6850***
Less than a high school diploma (ref: high school diploma)	−0.3700**	−0.3339*	−0.3940*
GED	−0.3525**	−0.3557	−0.3024
Employed in year or quarter before incarceration	0.2255	0.0296	0.4040*
Single, not married (ref: married)	0.1146	−0.0606	0.3004
Divorced or widowed	0.0385	−0.2634	0.5074
Has dependents	−0.1518	−0.1983	−0.0674
Charged for sex offense	0.1817	0.4803*	−0.3450
Charged for assault crime	0.1973	0.2882	0.0291
Electronic monitoring sentence	0.1037	0.2405	−0.0603
Charged for drug offense	−0.0966	−0.2598	−0.1609
Number of prior prison spells	0.0045	−0.0660	−0.0029
Number of years in prison	0.0075	0.0273	−0.0156
Completed parole	−0.1021	−0.3304	0.1457
Any history of substance abuse	−0.0497	−0.1935	0.0945
Positive substance abuse test in quarter	0.0826	0.2041*	0.0098
Release year	−0.1519	−0.1843	−0.0954
Release to a center	−0.0457	0.1322	−0.1773
Employment spell number	0.1182	−0.3207	0.6130***
Employment spell quarter	−0.0555	−0.1825	0.1636
Employment spell quarter, squared	−0.0366	−0.0237	−0.0633
Employment spell quarter, cubed	0.0042	0.0035	0.0059
Employment spell quarter, quartic	−0.0001	−0.0001	−0.0002
Quarter 1 (ref: Quarter 4)	−0.1676	−0.3058	0.0210
Quarter 2	−0.7511***	−0.9684***	−0.4750**
Quarter 3	−0.2536*	−0.4377**	−0.0569
2001 (ref: 2004)	−18.0340***	−17.2827***	−16.6607***
2002	−17.5539***	−16.5754***	−16.3419***
2003	0.3250**	0.1448	0.6087***
2005	−0.4038**	−0.1067	−0.7665***
2006	−0.3067	−0.0060	−0.7135*
2007	−0.4616	0.0720	−1.0707**
2008	−0.4719	−0.2079	−0.9997*
2009	−1.3737*	−0.2972	−2.7741**
Constant	302.2796	369.5143	185.7416
Employed in other industry
Jobs in food services per capita	−2.0855*	−3.5662*	−1.7145
Residents employed in food services	0.0078	0.0069	0.0094
Zip code characteristics
Average size employer	−0.0019	−0.0015	−0.0016
Disadvantage score	−0.0604*	−0.1058*	−0.0280
Disadvantage score of preprison neighborhood	−0.0325	0.0051	−0.0539
Population density	0.0000	0.0000	0.0000
Parolee density	0.0552	−0.0379	0.0882
Parolee characteristics
Relative age	0.0390	0.0172	0.0828**
Relative age, squared	−0.0008**	−0.0005	−0.0013**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.2461**	NA	NA
Female (ref: male)	−0.2067	−0.2096	−0.2216
Less than a high school diploma (ref: high school diploma)	−0.1752**	−0.0664	−0.2863**
GED	−0.0564	0.1223	−0.2575*
Employed in year or quarter before incarceration	0.2806***	0.3416***	0.2417*
Single, not married (ref: married)	−0.1810*	−0.1557	−0.2221
Divorced or widowed	−0.0429	−0.1183	0.1108
Has dependents	−0.0422	−0.0441	−0.0197
Charged for sex offense	0.0620	0.0907	0.0923
Charged for assault crime	0.0440	−0.0767	0.2006
Electronic monitoring sentence	−0.0543	0.1075	−0.3927
Charged for drug offense	0.0214	0.0600	0.0862
Number of prior prison spells	−0.0216	−0.0443	−0.0007
Number of years in prison	0.0512***	0.0236	0.0649***
Completed parole	0.2906***	0.2325**	0.3230***
Any history of substance abuse	−0.1194*	−0.1624*	−0.0828
Positive substance abuse test in quarter	0.0205	0.0980	−0.0202
Release year	−0.0246	−0.0467	−0.0654
Release to a center	0.1759	0.0160	0.3406*
Employment spell number	0.0298	−0.0577	0.0977
Employment spell quarter	1.4074***	1.3659***	1.7546***
Employment spell quarter, squared	−0.2632***	−0.2472***	−0.3567***
Employment spell quarter, cubed	0.0158***	0.0141**	0.0243***
Employment spell quarter, quartic	−0.0003*	−0.0003*	−0.0005***
Quarter 1 (ref: Quarter 4)	−0.3928***	−0.3240**	−0.4851***
Quarter 2	−0.3016***	−0.3076**	−0.3101**
Quarter 3	−0.0983	−0.0473	−0.1574
2001 (ref: 2004)	−17.2985***	−16.0662***	−16.5441***
2002	−17.2922***	−16.0604***	−16.4612***
2003	0.2461**	0.4207***	0.0463
2005	0.2106*	0.3814***	0.1080
2006	0.2703*	0.3717*	0.1710
2007	0.2958	0.3939	0.0772
2008	0.0098	0.3068	−0.2397
2009	−0.4394	−0.2142	−0.5110
Constant	44.8094	89.8547	125.2881
Returned to prison or dead
Jobs in food services per capita	−2.8126***	−3.3125	−2.2883**
Residents employed in food services	0.0000	0.0364	−0.0272
Zip code characteristics
Average size employer	0.0116*	0.0090	0.0103
Disadvantage score	0.0206	−0.0732	0.0654
Disadvantage score of preprison neighborhood	0.0709*	0.1167*	0.0553
Population density	0.0000	0.0000	0.0000
Parolee density	0.1158*	0.1835	0.0751
Parolee characteristics
Relative age	0.0606*	0.0009	0.1081**
Relative age, squared	−0.0010*	−0.0002	−0.0015**
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0715	NA	NA
Female (ref: male)	0.0485	−0.1093	0.1177
Less than a high school diploma (ref: high school diploma)	−0.0564	0.1258	−0.1997
GED	−0.0022	0.1263	−0.0915
Employed in year or quarter before incarceration	0.0008	−0.0721	0.0566
Single, not married (ref: married)	0.1911	0.0666	0.2366
Divorced or widowed	0.1814	0.0747	0.2710
Has dependents	−0.0772	−0.0238	−0.0768
Charged for sex offense	−0.3172	−0.1657	−0.4288
Charged for assault crime	−0.3783***	−0.3783*	−0.3926**
Electronic monitoring sentence	−0.0787	−0.3512	0.0420
Charged for drug offense	−0.3148***	−0.2182	−0.3196**
Number of prior prison spells	0.1019***	0.0658	0.1102***
Number of years in prison	−0.0720***	−0.1104**	−0.0589**
Completed parole	−4.9989***	−5.3544***	−4.7880***
Any history of substance abuse	−0.0276	−0.0780	−0.0179
Positive substance abuse test in quarter	0.1078**	0.1036	0.1070*
Release year	−0.1810	0.2596	−0.6077**
Release to a center	−0.5012	−0.2103	−0.6918***
Employment spell number	−0.1397	−0.1330	−0.1582
Employment spell quarter	0.5649***	0.5998***	0.5626***
Employment spell quarter, squared	−0.0960***	−0.1044**	−0.0928***
Employment spell quarter, cubed	0.0058***	0.0063*	0.0055***
Employment spell quarter, quartic	−0.0001***	−0.0001*	−0.0001**
Quarter 1 (ref: Quarter 4)	−0.0967	−0.1844	−0.0593
Quarter 2	−0.0511	−0.0201	−0.0715
Quarter 3	0.0261	0.2009	−0.0774
2001 (ref: 2004)	−17.0994***	−15.5072***	−16.7692***
2002	−1.4768**	−2.0783*	−1.3757*
2003	−0.6070***	−0.7201***	−0.5348***
2005	0.3357***	0.4277**	0.2787*
2006	0.0992	0.0458	0.1066
2007	−0.1336	−0.4413	−0.0042
2008	−0.5154*	−0.7248	−0.3723
2009	−0.7253*	−1.6886**	−0.3105
Constant	358.6354	−522.8403	1212.5503**

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.

Table A9.

Event History Results for All Outcomes, Temporary Employment

Variable	All	White	Black
Employed in temporary work
Jobs in temporary work per capita	−1.2524	−6.3386*	−0.6932
Residents employed in temporary work	−0.0340	0.0049	−0.0518*
Zip code characteristics
Average size employer	0.0004	−0.0003	0.0006
Disadvantage score	0.0016	0.0934	−0.0447
Disadvantage score of preprison neighborhood	−0.0846	−0.0797	−0.0764
Population density	0.0000	0.0001	0.0000
Parolee density	−0.2084**	−0.1516	−0.2310**
Parolee characteristics
Relative age	0.0296	−0.0372	0.0728
Relative age, squared	−0.0006	0.0003	−0.0012*
White, non‐Hispanic (ref: black, non‐Hispanic)	−0.2991**	NA	NA
Female (ref: male)	−0.2598	0.0394	−0.5703*
Less than a high school diploma (ref: high school diploma)	−0.1007	−0.2872	−0.0034
GED	0.0498	0.0570	0.0734
Employed in year or quarter before incarceration	0.3169**	0.2059	0.4113**
Single, not married (ref: married)	−0.3028*	−0.6344**	−0.1263
Divorced or widowed	−0.2897	−0.5924**	−0.0776
Has dependents	0.0554	0.0345	0.0889
Charged for sex offense	0.1686	0.1449	0.1576
Charged for assault crime	−0.0586	−0.2056	0.0273
Electronic monitoring sentence	0.0101	0.2795	−0.3640
Charged for drug offense	−0.1611	−0.0503	−0.1777
Number of prior prison spells	0.0198	−0.0188	0.0469
Number of years in prison	0.0587***	0.0139	0.0718***
Completed parole	0.0612	−0.0701	0.1755
Any history of substance abuse	−0.1319	−0.2655	−0.0822
Positive substance abuse test in quarter	0.0535	0.1656	0.0072
Release year	0.1899	0.3023	0.0828
Release to a center	0.1361	0.1222	0.1492
Employment spell number	0.2115*	0.3219*	0.0794
Employment spell quarter	0.2926	0.5005*	0.2585
Employment spell quarter, squared	−0.0867*	−0.1462*	−0.0775
Employment spell quarter, cubed	0.0065	0.0112*	0.0060
Employment spell quarter, quartic	−0.0002	−0.0003	−0.0002
Quarter 1 (ref: Quarter 4)	−0.2626*	−0.1020	−0.3803*
Quarter 2	−0.5732***	−0.9117***	−0.4043**
Quarter 3	−0.2032*	−0.3884*	−0.0985
2001 (ref: 2004)	−17.3563***	‐15.8944***	−16.1870***
2002	−17.1229***	−16.0335***	−15.8756***
2003	−0.0325	0.0359	−0.0870
2005	−0.0889	−0.1210	−0.0722
2006	−0.2643	−0.2113	−0.2553
2007	−0.2788	−0.6583	−0.0393
2008	−1.0171**	−0.7005	−1.4898**
2009	−1.3240**	−1.6818*	−1.0078
Constant	−382.7863	−607.3564	−169.0301
Employed in other industry
Jobs in temporary work per capita	−2.0577	0.2148	−3.5421*
Residents employed in temporary work	0.0155	0.0035	0.0304
Zip code characteristics
Average size employer	0.0003	0.0000	0.0006
Disadvantage score	−0.1008**	−0.1572***	−0.0570
Disadvantage score of preprison neighborhood	−0.0201	0.0019	−0.0230
Population density	0.0000	0.0000	0.0000
Parolee density	0.0224	−0.0729	0.0815
Parolee characteristics
Relative age	0.0539*	0.0584	0.0775
Relative age, squared	−0.0010**	−0.0011*	−0.0013*
White, non‐Hispanic (ref: black, non‐Hispanic)	0.4319***	NA	NA
Female (ref: male)	0.2678*	0.0584	0.4906**
Less than a high school diploma (ref: high school diploma)	−0.2471**	−0.0518	−0.4698***
GED	−0.1633	0.0107	−0.3919**
Employed in year or quarter before incarceration	0.2292**	0.2289*	0.2394
Single, not married (ref: married)	−0.1045	0.0605	−0.2949
Divorced or widowed	0.0132	0.0285	0.1449
Has dependents	−0.0321	−0.0066	−0.0230
Charged for sex offense	−0.0618	0.0414	−0.1613
Charged for assault crime	0.1536	0.1250	0.1968
Electronic monitoring sentence	0.1880	0.1884	0.1154
Charged for drug offense	0.0136	−0.0312	0.1025
Number of prior prison spells	−0.0499	−0.0608	−0.0478
Number of years in prison	0.0475***	0.0370*	0.0560***
Completed parole	0.2861***	0.2251*	0.3386**
Any history of substance abuse	−0.0964	−0.1251	−0.0611
Positive substance abuse test in quarter	−0.0438	0.0662	−0.1107
Release year	−0.1791	−0.0696	−0.4056
Release to a center	0.0932	0.1327	0.0472
Employment spell number	0.0361	−0.0054	0.0564
Employment spell quarter	1.3673***	1.2197***	2.4575***
Employment spell quarter, squared	−0.2535**	−0.2067***	−0.5523***
Employment spell quarter, cubed	0.0153*	0.0113**	0.0422***
Employment spell quarter, quartic	−0.0003	−0.0002*	−0.0010***
Quarter 1 (ref: Quarter 4)	−0.4996***	−0.3932**	−0.6539***
Quarter 2	−0.3693***	−0.3954***	−0.3641**
Quarter 3	−0.0981	0.0076	−0.2276
2001 (ref: 2004)	−17.5120***	−15.9012***	−17.0063***
2002	−17.3619***	−15.8075***	−16.5381***
2003	0.4210***	0.5441***	0.2917
2005	0.1020	0.3353*	−0.0221
2006	0.0069	0.0592	−0.0233
2007	0.1389	0.3241	−0.2777
2008	−0.1354	−0.0754	−0.0180
2009	−0.5647	−0.6195	−0.3764
Constant	353.5869	134.5858	805.7167
Returned to prison or dead
Jobs in temporary work per capita	−2.0725	−4.6133*	−0.8992
Residents employed in temporary work	0.0318	0.0647	0.0203
Zip code characteristics
Average size employer	0.0001	0.0011*	−0.0007
Disadvantage score	0.0484	0.0111	0.0716
Disadvantage score of preprison neighborhood	0.0508	0.0874	0.0469
Population density	0.0000	0.0000	0.0000
Parolee density	0.0149	0.1230	−0.0479
Parolee characteristics
Relative age	0.0551	−0.0223	0.1038*
Relative age, squared	−0.0008	0.0002	−0.0014*
White, non‐Hispanic (ref: black, non‐Hispanic)	0.0672	NA	NA
Female (ref: male)	0.0959	−0.0921	0.1486
Less than a high school diploma (ref: high school diploma)	−0.0172	0.1824	−0.1565
GED	0.0440	0.1306	−0.0193
Employed in year or quarter before incarceration	−0.0067	−0.1321	0.0855
Single, not married (ref: married)	0.2247	0.0719	0.2790
Divorced or widowed	0.1501	0.0715	0.1758
Has dependents	−0.0610	−0.0489	−0.0470
Charged for sex offense	−0.2825	−0.1254	−0.4125
Charged for assault crime	−0.4198***	−0.4629*	−0.4148**
Electronic monitoring sentence	0.0737	−0.1020	0.1711
Charged for drug offense	−0.3152**	−0.2458	−0.3179**
Number of prior prison spells	0.1165***	0.0826	0.1247***
Number of years in prison	−0.0658***	−0.1167**	−0.0515*
Completed parole	−4.5370***	−4.4917***	−4.6519***
Any history of substance abuse	−0.0524	−0.0492	−0.0677
Positive substance abuse test in quarter	0.1055*	0.0335	0.1171*
Release year	−0.0426	0.3618	−0.4265
Release to a center	−0.4309	−0.1635	−0.6210**
Employment spell number	0.0649	0.2294	−0.0529
Employment spell quarter	0.5431***	0.4372**	0.6459***
Employment spell quarter, squared	−0.0843***	−0.0649*	−0.1011***
Employment spell quarter, cubed	0.0049***	0.0038	0.0058***
Employment spell quarter, quartic	−0.0001**	−0.0001	−0.0001**
Quarter 1 (ref: Quarter 4)	−0.0578	−0.1153	−0.0423
Quarter 2	−0.0630	−0.0421	−0.0813
Quarter 3	0.0076	0.1636	−0.0838
2001 (ref: 2004)	−16.7787***	−15.0301***	−16.1640***
2002	−1.2675**	−1.9147	−1.1218*
2003	−0.5052***	−0.5634**	−0.4551**
2005	0.2382*	0.2712	0.2305
2006	−0.1146	−0.2721	−0.0182
2007	−0.3282	−0.9883*	−0.0083
2008	−0.7996**	−1.3377**	−0.4910
2009	−1.2389***	−2.7768***	−0.5086
Constant	80.7914	−728.1507	849.1519

Source: Reentry cohort data, U.S. Census, American Community Survey, and Business Register's County Business Patterns.

Notes: Stars of significance indicate differences between blacks and whites with “***” representing a statistically significant difference at the .001 level, “**” .01 level, and “*” .05 level.