Doing time and the unemployment line: The impact of incarceration on ex-inmates’ employment outcomes

Abstract

This study measures the influence multiple incarcerations and age at first incarceration have on the lengths of time ex-inmates are not employed and the amount of time ex-inmates spend looking for employment. Fixed effects analyses of longitudinal data from the Rochester Youth Development Study (RYDS) finds a relationship between incarceration at younger ages and longer non-employment experiences, but no association between incarcerations between 23 – 32 years old and non-employment lengths. Meanwhile, these individuals who experience incarceration younger spend equivalent time looking for employment as their never-incarcerated peers, despite having nonequivalent periods without employment.

More than 600,000 inmates are released on a yearly basis from prisons and jails across the United States (Carson & Golinelli, 2013). To ease reintegration of ex-inmates into society and avert unnecessary societal costs, it is important to understand the extent of unintended adverse effects of imprisonment on offender life. A series of theories and studies suggest incarceration inhibits participation in conventional activates and stigmatizes a growing population of individuals (Apel & Sweeten, 2010; Farrington, Gallagher, Morley, St. Ledger, & West, 1986; Petersilia, 2003; Sampson & Laub, 1993; Western, 2006).

This study expands on current literature by disaggregating incarceration experiences to focus on two aspects of life course theory – cumulative disadvantage and the age at which ex-inmates experience their first incarceration. More specifically, this study considers how first incarceration, multiple incarcerations, and age at first incarceration each impact ex-inmates’ employment outcomes by estimating fixed effects analyses. I discern that individuals with multiple incarceration experiences and those who experience incarcerations at younger ages (who are often the same) demonstrate longer periods of non-employment and spend less time looking for employment. However, these same individuals demonstrate longer periods of non-employment and shorter periods looking for employment prior to their most recent incarceration. These findings support Kerley and Copes’s (2004) conclusion that incarcerations at younger ages lead to cumulative disadvantage due to interrupted life course transitions, and lends credence to Apel and Sweeten’s (2010) concern that ex-offenders voluntarily opt out of the labor market.

Theory & Literature

Applying Sampson and Laub’s (1993) life course theory to employment and incarceration suggests incarceration interrupts key life transitions and social bonds, resulting in negative employment outcomes upon release. Two elements of life course theory deserve further attention in exploring the relationship between incarceration and employment: cumulative disadvantage and the age at which incarceration interrupts life courses.

Sampson and Laub’s (1993) life course theory suggests incarcerations interrupt and divert life transitions and trajectories. More specifically, they propose incarceration has a negative impact on offenders’ lives by interrupting key life transitions and disrupting social bonds necessary for informal social control to occur (Sampson & Laub, 2004). By disrupting or redirecting the employment trajectories of ex-inmates, incarceration prevents individuals from obtaining the same jobs, wages, and wage mobility demonstrated by their never-incarcerated peers. Incorporating aspects of labeling (Becker, 1963; Tannenbaum, 1938) and social bond theories (Hirschi, 1969), Sampson and Laub (1990, 1997) suggest incarceration weakens social ties that help ex-inmates find employment, while stigmatization inhibits new bonds and employment opportunities.

Sampson and Laub (1997) adopt aspects of social embeddedness (Granovetter, 1992; Hagan, 1993) to explain cumulative disadvantage in the life course. They state contact with the criminal justice system triggers an avalanche of negative employment outcomes building upon themselves, hindering successful employment. Hagan (1993) explains that initial jobs and contacts act as springboards for networking and future job mobility. Weak or late labor market entry inhibits the accumulation of employment contacts, as difficulty builds on previous failures, and “poor matching of one’s abilities to jobs will make it difficult for one’s actual skills to become widely known” (Granovetter, 1992, p. 243). Ultimately, these disadvantages leave individuals to trail the employment trajectories of their never-incarcerated peers if they ever move past entry-level jobs. Unfortunately, few studies explore the impact of incarceration interacted with age.

Research

The majority of research – with notable exceptions discussed below – suggests arrest, conviction, and incarceration negatively influence ex-inmates’ employment outcomes (Apel & Sweeten, 2010; Bushway, 1998; Freeman, 1992; Grogger, 1995; Holzer, 1996; Hunter & Borland, 1997, 1999; Kerley & Copes, 2004; Lopes et al., 2012; Nagin & Waldfogel, 1998, 1995; Pager, 2003; Pager, Western, & Sugie, 2009; Pettit & Lyons, 2009; Sampson & Laub, 1993; Waldfogel, 1994; Western, 2002; Western & Beckett, 1999; Witte & Reid, 1980). For example, in a study of males incarcerated as juveniles, Sampson and Laub (1993) find incarceration between ages 17 and 23 negatively correlates with work commitment, employment, and continuity of employment at ages 25 to 32. Numerous studies demonstrate contact with the criminal justice system at earlier ages is more harmful to employment and economic outcomes (Hagan, 1991, 1993; Hagan & McCarthy, 1998; Kerley & Copes, 2004; Thornberry & Christenson, 1984). Meanwhile, Apel and Sweeten (2010) find ex-inmates spend less time looking for employment (when non-employed) than their non-employed, never-incarcerated peers, raising concern that initial or anticipated rejections during pursuits of employment may lead ex-inmates to voluntarily opt out of participating in the labor market.

However, a few studies do not find incarceration negatively influence employment outcomes. Kling (2006) and Lalonde and Cho (2008) perform fixed effects analyses on administrative data and find that immediately following incarceration, employment rates increase for offenders compared to their own pre-incarceration employment histories. However, in the long term, employment rates drop to mirror pre-incarceration rates (Kling, 2006; Lalonde & Cho, 2008). Similarly, Ramakers et al. (2012) find ex-inmates have more success finding employment than individuals unemployed and incarcerated at later points. They conclude differences in employment outcomes result partly from ex-inmates demonstrating poor employment histories even prior to incarceration (Ramakers et al., 2012; Ramakers, van Wilsem, Nieuwbeerta, & Dirkzwager, 2014).

It is important to note, in a natural experiment where criminal defendants are randomly assigned to judges demonstrating statistically significant imprisonment sentencing disparities, Loeffler (2013) finds employment rates of ex-inmates five years after indictment fail to be statistically significantly different. Loeffler (2013) suggests his results address two large concerns in incarceration research: isolation of the impact of incarceration on offenders and selection bias. Some scholars argue that any stage of the criminal justice process can instigate an impact on employment, making it difficult to separate out the effects of incarceration from the effects of previous criminal justice experiences. Additionally, numerous scholars suggest individuals who demonstrate preexisting characteristics of disadvantage (minority, urban poor, unemployment, insufficient education, etc.) are more likely to be arrested, convicted, and imprisoned (Loeffler, 2013; Manski & Nagin, 1998; Nagin, Cullen, & Jonson, 2009; Smith & Paternoster, 1990). As such, the negative life outcomes demonstrated by ex-inmates may reflect systemic social selection and stagnation of negative attributes during incarceration. Loeffler (2013) argues that random assignment to judges who demonstrate statistically significant imprisonment patterns controls for incarceration selection bias. Thus, his finding that incarceration fails to impact employment may suggest employment differences observed in other research reflect preexisting disadvantage among individuals who experience incarceration, but incarceration itself does not add additional disadvantage.

Building on Previous Literature

While these studies provide valuable insight into the relationships between criminal justice system contact and future employment outcomes, it is possible to explore life course theory further by disaggregating age of incarceration and the cumulative disadvantage of multiple incarceration experiences. Previous literature has employed different sampling restrictions. Numerous studies limit their samples to concentrate on the impact of first-time incarceration on employment (Apel & Sweeten, 2010; Kerley & Copes, 2004; Lalonde & Cho, 2008).ⁱ Meanwhile, other studies consider incarceration regardless of whether the experience is the first or one of multiple incarcerations (Freeman, 1992; Kling, 2006; Loeffler, 2013; Lyons & Pettit, 2008; Pettit & Lyons, 2009; Western, 2002, 2006; Western & Pettit, 2000). The disadvantage of limiting samples and populations in these ways is that neither addresses whether ex-inmates experience cumulative disadvantages. Studies of first-time ex-inmates only address employment after the first incarceration, and studies of ex-inmates, regardless of the number of incarcerations experienced, deliver estimates of the average employment outcome across all ex-inmates.

While a few studies address life course concerns by limiting their samples to specific age groups (see Freeman, 1992; Sampson & Laub, 1993), without the ability to compare the employment outcomes of individuals who experience their first incarceration at a young age to those who experience it at older ages, it is impossible to ascertain whether employment outcomes differ based on age of interruptions in the life course.

Two studies have considered the impact of age on employment outcomes. Pettit and Lyons (2009) find little evidence that the effects of incarceration on post-release employment and wages vary by age at admission. By comparison, Witte and Reid (1980) measure the impact of age of release from incarceration on employment outcomes. Yet, there is equal, if not greater, significance in measuring the age individuals begin their first incarcerations. Age at first incarceration indicates a crucial life course interruption, signifying the moment human capital attainment stops, slows, or stalls. Kerley and Copes (2004) emphasize that contact with the criminal justice system at younger ages is especially devastating, as harm accumulates as individuals age. Therefore, age at incarceration may predict ex-inmate employment outcomes more successfully than measures of admission age or release age.

This study expands on previous literature by focusing on the cumulative employment disadvantages experienced when individuals incur multiple incarceration experiences, and how employment outcomes differ based on individuals’ first incarceration interruption in the life course. By measuring the impact of any incarceration experience and first incarcerations on employment outcomes, this study replicates previous studies’ samples and establishes baseline employment outcomes with RYDS data. The study then expands on previous literature by disaggregating incarceration to study individuals with multiple incarceration experiences and age at first incarceration. Life-course theory’s discussion of cumulative disadvantage and early life-course interruptions leads to this study’s the hypothesis that employment outcomes worsen with additional incarcerations and incarceration at earlier ages more negatively impact employment outcomes than later incarcerations. The null hypothesis is employment outcomes are equivalent or improve with additional incarcerations or younger initial incarcerations.

Methods

Data

To examine the relationship between incarceration and employment, I utilize data from the Rochester Youth Development Study (RYDS), an ongoing longitudinal study of antisocial and delinquent behaviors. Beginning in 1988 with 1,000 middle school students in the Rochester (New York) Public School System, RYDS has interviewed students from early adolescence through adulthood.

The original RYDS sample is stratified on two dimensions to target subjects at high risk for violence and serious delinquency. First, males are oversampled (75% versus 25%), as they are more likely than females to engage in serious and violent offenses (Blumstein, Cohen, Roth, & Visher, 1986; Huizinga, Morse, & Elliott, 1992). Second, students living in areas of Rochester that experience high adult residential arrest rates are oversampled under the assumption that adolescents living in these areas are at greater risk for offending than adolescents living in areas with lower adult residential arrest rates. Researchers identified high residential offender areas based on the proportion of each census tracts’ total adult population arrested by Rochester police in 1986. The RYDS samples students at random within tracts at frequencies proportionate to the rates of offenders living in each tract.

From Waves 2 to 14, the RYDS experiences less than 2% attrition per year. To maximize sample retention, the RYDS utilizes participant-tracking techniques and provides survey incentives. Similarly, the RYDS attempts to track and interview all subjects who move from Rochester and conduct in-institution surveys of participants institutionalized during data collection periods. At Wave 14, the subject panel is 68.9% black, 16.2% Hispanic, and 14.9% white, and 79% of the original sample is still participating. A formal test of subject attrition within RYDS reveals subjects retained do not significantly differ from those not retained on multiple dimensions, including gender, social class, family structure, drug use, delinquency, property crime, and violent crime (see Krohn & Thornberry, 1999). None of the difference tests reached statistical significance (p < .05).

Current study

This study uses data from the first waves RYDS surveys inquire into participants’ employment activities in enough detail to discern cumulative employment or the length of time participants spend looking for employment since last interviewed. During these interviews, interviewers ask subjects (who are approximately 29- to 32-years-old) to complete a life history calendar (LHC) in which they report on life events (including employment, non-employment, incarceration, arrest, and probation) that occurred since they were last interviewed at approximately 23-years-old.ⁱⁱ These questions include start and end dates of these experiences and follow-up questions regarding event details. Respondents are also able to report multiple and concurrent events. The study includes two dependent measures of employment outcomes based on participants’ responses to LHC questions. Data collected from the Rochester Police Department (RPD) and the New York Division of Criminal Justice Services (DCJS) enable identification of participants incarcerated prior to age 23.ⁱⁱⁱ Figure 1 demonstrates the study design employed.

Figure 1.

Study Design

*Contact occurs at different points in the LHC period for each respondent. Pre- and post-contact observation length is measured for each participant to control for differences in observation lengths.

To measure the impact of incarceration on employment outcomes, the “experimental” group includes 184 LHC participants who experience incarceration during the LHC. The comparison group includes all 66 respondents who experience an arrest or conviction to probation during the LHC, but do not experience incarceration during the period.^iv Three structural aspects of the RYDS data make it particularly well suited for this study. First, the RYDS’s oversampling of at-risk youth is advantageous because the comparison group consists of individuals truly “at-risk” of incarceration. Second, the RYDS’s longitudinal panel design makes fixed-effects comparisons of employment history before and after incarceration possible. Third, the RYDS’s self-report structure includes unreported (“off-the-books”) work and quantifies time participants spend looking for employment.

While individuals who experience incarceration necessarily experience an arrest at some point prior to incarceration, I only observe these individuals pre- and post-incarceration. In addition, individuals included in the arrest/probation subsample do not demonstrate an experience of incarceration before or during the LHC period. As such, each individual is included in only one of the subsamples (incarceration or arrest/probation), with no overlap between subsamples.

Table 1 outlines demographic characteristic descriptive statistics for the arrest/probation comparison subsample (columns 4 and 5), individuals experiencing their first incarcerations (columns 6 and 7), and individuals with multiple incarceration experiences (columns 8 and 9). The comparison group is dissimilar from the incarceration subsamples in many regards, including gender and race demographics, and cumulative delinquency throughout the life course. However, the arrest/probation comparison group proves to be a closer fit to the incarceration samples than are RYDS participants excluded from the study sample (see columns 2 and 3). Thus, the comparison group serves as the best possible answer to what the incarceration subsample’s employment outcomes might look like if they had contact with the criminal justice system but did not experience incarceration.

Table 1.

Sample Demographic Characteristics

	RYDS Sample Excluded from Study		Study Sample
			Arrest/Probation Subsample		First Incarceration Subsample		Multiple Incarceration Subsample
	Mean/%	SD	Mean/%	SD	Mean/%	SD	Mean/%	SD
Gender
Female	35.4		21.0		13.0		10.4
Male	64.6		79.0		86.8		89.6
Race/Ethnicity
Black	66.1		72.6		75.4		74.8
Hispanic	17.1		11.3		13.0		15.7
White	16.8		16.1		11.6		9.6
Have a Partner	67.4		68.0		71.7		61.4
Have Children	57.6		53.2		60.9		53.0
Education
High school graduate	51.7		50.0		55.7		48.0
Employment
Percent of time non-employed between ages 19–21	23.0	28.4	22.3	29.2	24.6	30.9	44.0	33.3
Delinquency
General ever-variety	3.8	3.9	4.6	3.4	6.1	4.1	8.1	4.7
General incidence	165.7	357.9	171.2	249.6	409.8	580.9	509.6	714.6
Serious ever-variety	0.7	1.2	0.6	1.1	1.4	1.6	2.0	1.6
Serious incidence	4.0	14.1	3.5	11.6	6.5	14.4	12.4	23.7
Violence ever-variety	1.2	1.3	1.5	1.2	2.2	1.4	2.6	1.4
Violence incidence	7.4	19.9	9.4	20.3	15.6	31.2	21.3	29.7
Socioeconomic Level
Lower class	62.3		63.6		63.8		69.2
Middle class	37.7		36.4		36.2		30.8
N	540		66		70		114

Measurement

Dependent measures

A common concern in the literature is the comparative rates or lengths of non-employment for individuals who do and do not experience incarceration. The first dependent variable, non-employment, addresses this concern by measuring the amount of time individuals report being non-employed in days.^v To establish whether incarceration affects individuals’ non-employment lengths, it is necessary to measure non-employment for a control period and then for a period following contact with the criminal justice system. The control period, or pre-contact period, measures the cumulative length of non-employment individuals experience before they are either arrested, convicted to probation, or incarcerated, and serves as the baseline measure. Then, non-employment for a post-contact period is measured following individuals’ contact with the criminal justice system to determine whether contacts affect lengths of non-employment.

Table 2 shows the means, standard deviations, and ranges of all study measures by subsample. In the pre-contact period, the mean non-employment length is 183 days for the arrest/probation subsample, 205 days for the first incarceration subsample, and 356 days for the incarceration subsample. The mean lengths of non-employment in the post-contact period are slightly shorter, averaging 121 days for the arrest/probation subsample, 147 days for the first incarceration subsample, and 263 days for the multiple incarceration subsample. With means of 183, 204, and 356 days in the pre-contact period, and larger standard deviations of 437, 343, and 550 days, non-employment demonstrates a large skew. Skew proves to be an issue with both dependent measures. I correct for this by logging the dependent variables.

Table 2.

Variable Descriptive Statistics

	Arrest/Probation Subsample
	Pre-contact period			Post-contact period
	Mean	SD	Range	Mean	SD	Range
Non-employment	183.0	437.0	2038	121.0	263.7	1095
Looking for employment	66.7	306.1	2038	22.1	68.4	334
Observation length	1482.5	750.7	2771	1564.2	748.5	2799
	First Incarceration Subsample
	Pre-contact period			Post-contact period
	Mean	SD	Range	Mean	SD	Range
Non-employment	204.6	343.4	1249	147.2	355.6	2220
Looking for employment	109.9	274.0	1186	49.5	147.3	728
Observation length	1426.7	755.8	2772	1335.0	882.9	2891
	Multiple Incarceration Subsample
	Pre-contact period			Post-contact period
	Mean	SD	Range	Mean	SD	Range
Non-employment	356.0	550.0	2191	262.5	356.4	1461
Looking for employment	189.5	439.1	2191	57.4	192.3	1218
Observation length	1265.5	883.7	3200	1107.8	823.1	2891

The measure looking for employment attempts to determine whether individuals who experience incarceration have more difficulty finding employment than their never-incarcerated peers. In the RYDS survey, respondents report periods of time when they were and were not employed. When respondents report a period of non-employment, follow-up questions inquire as to whether they actively looked for employment during the period.^vi Looking for employment serves as a count measure of the number of days respondents report being non-employed and looking for employment during the pre- and post-contact periods separately. The mean lengths of time looking for employment during the pre- and post-contact periods are 67 and 22 days respectively for the arrest/probation subsample, 110 and 50 days respectively for the first incarceration subsample, and 190 and 57 days respectively for the multiple incarceration subsample. Looking for employment is logged to account for skew.

Independent measures

I use police data and LHC responses to construct four incarceration measures.^vii The first incarceration variable (incarceration) is a dichotomous measure of whether individuals in the sample experience incarceration during the LHC observation period. This measure is important in distinguishing between the never-incarcerated subsample and the incarcerated subsample, and is included to enable result comparisons to previous studies that use aggregated incarceration samples.

Two dichotomous variables subdivide the incarceration subsample based on number of incarcerations. Bifurcating the sample allows investigation of cumulative disadvantage by analyzing whether individuals experiencing their first incarcerations and those with previous incarceration experiences demonstrate different employment histories from the never incarcerated subsample, and from each other. First incarceration identifies individuals in the incarceration subsample whose LHC incarceration is the first incarceration they have ever experienced. Multiple incarcerations similarly identifies individuals in the incarceration subsample who experienced at least one incarceration prior to their incarceration during the LHC period.^viii Individuals in the first incarceration and multiple incarcerations subsamples make up 28.8 and 44.8% of the overall sample respectively.

Age at first incarceration measures how old individuals in the incarceration subsample are/were at their first experience of incarceration. This measure serves to test life-course theory’s proposition that incarceration differentially impacts ex-inmates’ employment outcomes based on when incarcerations begin. The average age at first incarceration in the incarceration subsample is 23-years-old, and ranges from 15 to 32 years of age.

Control measures

Since incarceration, arrest, or conviction to probation occur at different times in the LHC period for each respondent, pre- and post-contact period lengths vary for participants. Thus, it is necessary to control for the observation length of each period for each respondent. Pre-contact observation length measures the amount of time in days between a respondent’s last interview and their first incarceration, arrest, or probation during the LHC period. Average observation lengths are comparable for the arrest/probation subsample and incarceration subsamples. For example, the average pre-contact period for the arrest/probation subsample is 1483 days (4.1 years), is 1427 days (3.9 years) for the first incarceration subsample, and is 1266 days (3.5 years) for the multiple incarceration sample.^ix However, observation length ranges from 28 to 3228 days (8.8 years) during this period.^x Post-contact observation length measures the time in days between a respondent’s experience of arrest, probation, or release from incarceration and either the end of the LHC, or a second occurrence of the experience (incarceration, arrest, or probation) during the LHC. The average observation length in the post-contact period is 1564 days (4.3 years) for the arrest/probation subsample, 1335 days (3.7 years) for the first incarceration subsample, and 1108 days (3 years) for the multiple incarceration subsample.

Analyses

While the never-incarcerated comparison group acts as a counterfactual to the incarcerated experimental group, the judicial decision to incarcerate some individuals participating in crime while not others can represent the existence of underlying differences between the groups. To address existing selection biases that result from nonrandom assignment of respondents to incarceration, this study employs fixed effects analyses in a manner akin to difference-in-differences analyses.^xi Within person analyses compare respondents’ employment outcomes prior to (“pre-contact”) and following (“post-contact”) contact with the criminal justice system. Differences observed between the pre- and post-contact periods are then compared for respondents who experience incarceration and individuals in the comparison group who experience arrest or conviction to probation without incarceration,^xii thus establishing between person comparisons. Each individual acts as their own control, while the measurement technique holds time-invariant unobservables constant.

Models

For both dependent variables, I discuss five models, each with different independent measures of incarceration. To correct the previously discussed over-dispersion of both dependent measures (see Table 3), the dependent variables are log transformed.^xiii

Table 3.

Logged Fixed Effects Estimates for Non-Employment Outcome

	Model 1	Model 2	Model 3	Model 4	Model 5
Observation length	0.001*** (0.000)	0.001*** (0.000)	0.001*** (0.000)	0.001*** (0.000)	0.001*** (0.000)
Post-contact	−0.256 (0.458)	−0.230 (0.428)	−0.261 (0.473)	−0.105 (0.432)	0.957 (1.889)
Incarceration	6.686** (2.395)
Post-contact* Incarceration	0.448 (0.532)
First incarceration		−0.052 (2.243)
Post-contact* First incarceration		−0.005 (0.575)
Multiple incarcerations			6.584** (2.478)	6.714** (2.424)
Post-contact* Multiple incarcerations			0.762 (0.602)	0.527 (0.557)
Age at first incarceration					−0.552* (0.213)
Post-contact* Age at first incarceration					−0.038 (0.081)
Intercept	−1.350 (1.714)	−0.932 (1.615)	−1.446 (1.781)	−1.613 (1.739)	13.510** (4.621)
R2	.708	.705	.703	.685	.689
N	243	131	171	155	141

^*p < .05

^**p < .01

^***p < .001

Each of the five models follow the form:

$$\begin{array}{l}\ln {\left(\mathit{employment\; outcome}\right)}_{it}\\ ={\alpha }_i+ {\beta }_1{\left(\mathit{observation\; length}\right)}_{it}+{\beta }_2{\left(\mathit{post}-\mathit{contact}\right)}_i\\ +{\beta }_3{\left(\mathit{incarceration\; measure}\right)}_{it}+{\beta }_4{\left(\mathit{post}-\mathit{contact}\ast \mathit{incarceration\; measure}\right)}_{it}\end{array}$$

where t signifies comparison by pre- and post-contact periods, and i denotes individual level comparison.^xiv Observation length controls for the amount of time observed in the pre- and post-contact periods separately, and a dichotomous variable (post-contact) distinguishes between the pre- and post-contacts periods (0 = pre-contact, 1 = post-contact). The equations include one of the four independent measures of incarceration discussed above, as well as an interaction measure between post-contact and the independent measure of incarceration (post-contact*incarceration measure). Interaction measures between post-contact and incarceration measures isolate the impact of incarceration on employment outcomes during the post-contact period.

Since it is only possible to measure age at first incarceration for individuals who experience an incarceration, the model exploring age at first incarceration excludes respondents who do not experience incarceration. As such, the fixed effects model for age at first incarceration does not compare slope estimates between the incarcerated and never-incarcerated subsamples. Instead, this model only compares the slopes of the incarcerated subsample based on age at first incarceration.

Results

Non-Employment

Table 4 shows estimates of the impact of incarceration on ex-inmates’ cumulative lengths of non-employment.^xv As stated previously, the purpose of the variable observation length is to hold constant the length of time respondents are observed in the pre- and post-contact periods. Observation length also demonstrates that when respondents have longer pre- and post-contact periods, there is increased chance they will experience slightly more non-employment. This proves true in the models using incarceration, first incarceration, and multiple incarcerations. However, this finding is trivial due to its extremely small influence (0.001%) on non-employment length. Similarly, the variable post-contact controls for whether non-employment occurs during the pre- or post-contact period. Post-contact fails to predict length of non-employment in all five models. Thus, the entire sample demonstrates constant non-employment rates across the full observation period.^xvi

Table 4.

Logged Fixed Effects Estimates for Looking for Employment Outcome

	Model 1	Model 2	Model 3	Model 4	Model 5
Observation length	0.001*** (0.000)	0.001** (0.000)	0.001** (0.000)	0.001** (0.000)	0.001*** (0.000)
Post-contact	−0.098 (0.373)	−0.090 (0.338)	−0.098 (0.390)	−0.546 (0.361)	−1.910 (1.480)
Incarceration	0.902 (1.972)
Post-contact* Incarceration	−0.664 (0.435)
First incarceration		0.318 (1.788)
Post-contact* First incarceration		−0.558 (0.455)
Multiple incarcerations			0.952 (2.064)	0.770 (2.039)
Post-contact* Multiple incarcerations			−0.774 (0.500)	−0.331 (0.467)
Age at first incarceration					−0.086 (0.171)
Post-contact* Age at first incarceration					0.053 (0.064)
Intercept	−0.974 (1.412)	−0.785 (1.287)	−0.965 (1.484)	−0.856 (1.462)	1.227 (3.720)
R2	.660	.668	.641	.646	.667
N	243	131	171	155	141

^*p < .05

^**p < .01

^***p < .001

Model 1 demonstrates the impact of incarceration – regardless of whether it is the first incarceration or following a history of previous incarcerations – on non-employment length. As such, Model 1 acts as a baseline for later models and replicates previous studies exploring aggregate incarceration samples. The model finds incarceration statistically significant. Individuals who experience incarceration are non-employed for eight times longer than their never-incarcerated peers. However, the incarceration interaction term is not statistically significant. Thus, individuals in the incarceration subsample do not spend a statistically different length of time non-employed post-contact compared to their never-incarcerated peers. In other words, incarcerated individuals demonstrate longer non-employment spells than their never-incarcerated peers, but the incarcerations observed during this study are not the cause.

Models 2 and 3 explore cumulative disadvantage by analyzing whether individuals with one or multiple incarceration experience(s) demonstrate negative employment outcomes compared to never incarcerated peers. Model 2 establishes that individuals experiencing their first incarceration between ages 23 and 32 have employment histories equivalent to their never-incarcerated peers. This proves true even after they experience incarceration. However, Model 3 establishes that individuals with prior histories of incarceration(s) (multiple incarcerations) report longer periods of non-employment (over seven times longer) than individuals who never experience incarceration. Yet, they do not report an increase in non-employment length following an additional incarceration experience. This suggests that incarcerated individuals with previous incarceration experience(s) drive the difference observed in Model 1 between the incarcerated and never-incarcerated subsamples.

Model 4 compares non-employment cumulative disadvantage between the first incarceration and multiple incarcerations subsamples. As one might expect based on the outcomes of Models 2 and 3, Model 4 shows that individuals with prior incarceration experiences (multiple incarcerations) report being non-employed over eight times longer than those experiencing their first incarceration after age 23. Yet again, participants do not report an increase in non-employment length following additional incarceration experience. This suggests individuals with previous incarceration experience(s) demonstrate different employment histories than their peers incarcerated later in life.

Model 5 tests life-course theory’s proposition that incarceration differentially impacts ex-inmates’ employment outcomes based on when incarcerations first interrupt the life-course. Model 5 finds age at first incarceration to have a statistically significant influence on non-employment length. As age at first incarceration increases, non-employment length decreases. In other words, individuals who experience their first incarceration at young ages (for example, 16-years-old) experience longer cumulative non-employment than individuals who are older when they experience their first incarcerations. Figure 2 demonstrates the relationship between age at first incarceration and non-employment length. While individuals who experience incarceration at age 16 demonstrate over 13 times longer non-employment over the observation period, individuals incarcerated at 26-years-old report 4% longer non-employment (holding observation length constant at average length). Interestingly, non-employment lengths do not change between the pre- and post-incarceration periods when age at first incarceration is the independent measure. This suggests ex-inmates’ post-incarceration employment patterns mirror their employment histories prior to incarceration.

Figure 2.

Non-Employment Length by Age at First Incarceration

Looking for Employment

In Table 4, the control measure observation length successfully predicts the length of time individuals spend looking for employment in each of the five models. The impact of observation length is very small (0.001%) in each of the models, to the point of being trivial.^xvii

In Model 1, both the incarceration measure and the interaction term between post-contact and incarceration fail to predict the length of time individuals look for employment. Thus, experiencing incarceration during the observation period does not result in more time spent looking for employment. Models 2, 3, 4, and 5 demonstrate similar findings; first incarceration, multiple incarcerations, and age at first incarceration do not predict time spent looking for employment.

These findings suggest incarceration does not result in differing lengths of time spent looking for employment. Of note, despite spending more time non-employed, individuals with multiple incarcerations and those incarcerated at younger ages do not spend more time looking for employment.

Discussion

The findings of this study unravel a complicated story regarding incarceration and employment. Incarceration between ages 23 and 32 does not cause longer periods of non-employment or looking for employment among study participants. However, not all ex-inmates’ employment histories mirror those of their never-incarcerated peers. Ex-inmates with multiple and earlier incarceration experiences – who are the same participants in this study – demonstrate longer periods of non-employment than their never-incarcerated peers. This suggests that participants’ employment histories differ prior to their most recent incarceration.

Prior research (Bushway, 1998; Freeman, 1992; Grogger, 1995; Hunter & Borland, 1997; Lopes et al., 2012; Manski & Nagin, 1998; Nagin, Cullen, & Jonson, 2009; Smith & Paternoster, 1990; Waldfogel, 1994) finds employment disparities precede incarceration, and Loeffler (2013) suggests previous studies finding incarceration effects on employment may capture and attribute effects from previous criminal justice contact with incarceration itself. This study’s finding that age at first incarceration corresponds to length of non-employment may supplement these findings, suggesting cumulative disadvantage or employment stagnation may result from incarceration at younger ages. It seems likely that younger first-time inmates are incarcerated before they develop continuous or stable employment, and return to employment outlooks that emulate their pre-incarceration employment. If this is the case, incarcerating adolescents and young adults has repercussions on their later employment outcomes and should be avoided whenever possible. Replication of previous studies with added control measures and sample limitations would aid in exploring this hypothesis.

Also of note, is the discovery that despite having nonequivalent lengths of non-employment, participants generally spend equal lengths of time looking for employment. This mirrors Apel and Sweeten’s (2010) observation that ex-inmates appear to opt out of the labor market, and may reflect ex-inmates’ fatigue after legal and institutional barriers inhibit finding employment, or concession once ex-inmates realize the barriers they face against competing applicants. Regardless, based on these findings, it is imperative that reentry programs work with individuals with multiple incarceration experiences and those incarcerated at younger ages to build the skills and confidence necessary to pursue labor market participation post-incarceration.

Two limitations of this study are sample size and age range. The sample size used in this study is considered appropriately large based on Cohen’s (1992) statistical power analysis, however, the complexities of the techniques used likely limit the statistical power of the models. As such, the null findings for the first incarceration models (Models 2) may reflect model confinement due to sample size.^xviii Also, it is not possible to disentangle the relationship between employment, age of incarceration, and number of incarcerations in this study, as age at first incarceration and number of incarcerations are highly correlated. This correlation is likely to exist in most studies, as repeated incarcerations take time, frequently requiring first incarcerations to occur at younger ages. However, studies covering wider age ranges or older samples may escape this limitation. Future research using a larger sample and wider age range would be valuable.

With the understanding that individuals incarcerated at younger ages experience more negative employment outcomes, it is possible to shape policies to aid these populations. Avoiding incarceration of youths, particularly those still in school, could alleviate the employment disadvantages demonstrated by these individuals later in life. Alternatively, skills training and education during incarceration, and reentry assistance following release would likely improve the employment outlooks of individuals incarcerated at younger ages. Future research should explore the relationship between age at first incarceration, employment outcomes, and other life outcomes such as education and family formation. This knowledge can help in informing public policy during sentencing and within prisons and jails.