Incarceration and the Legitimate Labor Market: Examining Age-Graded Effects on Employment and Wages

Abstract

Over the past 30 years, the U.S. inmate population has increased dramatically, and the penal system has acquired growing attention in accounts of recent trends in economic stratification. As the prison system has expanded, its population has aged; incarceration rates have risen sharpest among older age groups. A large body of research documents differences in criminal offending and incarceration over the life course, but little attention has been paid to how the effects of spending time in prison depend on the timing of incarceration in the life course. Using state administrative data that provide significant variance in the age of offenders, this article investigates how the timing of incarceration in the life course influences its effects on post-release employment and wages. We do not find consistent evidence that incarceration effects vary by age at admission. Instead, incarceration appears to have important consequences for employment and wage outcomes regardless of when individuals are admitted to prison. Even the most motivated offenders suffer sizeable and significant wage penalties and, over time, decreased likelihood of employment. These findings underscore the relevance of legal and institutional shifts associated with carceral expansion and the aging of the inmate population for life course theories of criminal desistance, accounts of labor market inequality, and prisoner reentry programs.

Over the past 30 years, the U.S. inmate population has increased dramatically. In 1980, just over 500,000 adults were incarcerated in prisons and jails. By 2005, that number had grown to more than 2.1 million, representing an increase of more than 330 percent (Sourcebook of Criminal Justice Statistics 2006: n.p.). Decades of carceral expansion have resulted in 1 in 100 American adults now residing behind bars (Pew 2008:1) and 16 million felons and ex-felons living in the United States (Manza & Uggen 2006:9). Spending time in prison has become a normative experience for low-skilled black men in particular: on any given day more than 10 percent of U.S. black men are in prison or jail, and nearly 60 percent of black men without a high school diploma can expect to spend time in a state or federal prison (Pew 2008:6; Pettit & Western 2004:164).

Law and society scholars have long been interested in the role of law in ameliorating or exacerbating social inequality, and recently the criminal justice system has emerged as a key suspect in accounts of economic, political, and health inequality. Research has documented significant effects of spending time in prison on employment and wages (Pager 2003; Western 2002, 2006), political participation (Behrens et al. 2003; Fellner & Mauer 1998; Manza & Uggen 2006; Uggen & Manza 2002), and health (Binswanger et al. 2007; London & Myers 2006; Massoglia 2008a, 2008b; Patterson 2007; Schnittker & John 2007). There is evidence that incarceration fuels inequality over the life course by setting men on a “trajectory of disadvantage” (London & Myers 2006). However, little research has examined specifically how age at incarceration conditions its economic effects (Uggen 2000; Nagin & Waldfogel 1998), despite the observation that spending time in prison or jail is no longer reserved for youthful indiscretions or serious offenders.

Longer prison sentences and mandatory parole revocations associated with the punitive turn in American criminal justice have fuelled the aging of the incarcerated population. Fully half of male state and federal inmates are over age 35 (Sabol et al. 2007:23). Moreover, even though first-time incarceration rates peak in the early twenties, the contemporary prison expansion has generated the steepest increase in first-time incarceration rates for men in their early thirties (Bonczar 2003).

Rising arrest and incarceration rates among older men call into question long-held beliefs about the “age-crime curve,” which suggests that criminal activity peaks in the late teens, though most offenders desist from crime in adolescence and nearly all offenders desist by their middle to late thirties (Hirschi & Gottfredson 1983; Gottfredson & Hirschi 1990). While criminal offending historically has been concentrated in the early adult years, increases in incarceration—including new commitments—among older men raise questions about criminal desistance over the life course. Furthermore, to the extent that legitimate labor market opportunities influence criminal desistance, shifts in the age composition of inmates also raise empirical questions about whether and how the effects of spending time in prison depend on the age at admission.

Life course criminology emphasizes how involvement in crime—and incarceration—changes over the life course and in relation to bonds of conformity (e.g., Sampson & Laub 1990; Uggen 2000; Shover 1996). This perspective implies that orientations toward crime and work in the legitimate labor market vary by age, suggesting that age at admission may condition the effect of incarceration on post-release labor market outcomes. By contrast, research on the effects of “mass incarceration” emphasizes the salience of a criminal record for economic opportunities (e.g., Pager 2003, 2007; Holzer 1996). This perspective implies that the stigma of a criminal record should not vary by age. Instead, employer preferences for noninmates lead to enduring incarceration effects regardless of its timing in the life course.

Using administrative panel data on inmates with significant variation on age of admission to prison, this article examines the age-graded effects of incarceration on post-release integration into society with a focus on employment and wages. That is, does the effect of incarceration vary by the timing of imprisonment during the life course? Or are the effects of incarceration generalizable across all age groups?

Examining the economic implications of incarceration at different ages is important for both theoretical and political reasons. Age-graded theories of crime and desistance are premised on the idea that as inmates age they shift their orientations to crime in relation to other opportunities associated with work, schooling, or engagement with their families. Although previous research suggests that older inmates are particularly responsive to work programs and incentives associated with work in the legitimate labor market (e.g., Uggen 2000), it is unclear whether one would observe similar patterns in the current era of prison expansion.¹ Recent research contends that incarceration carries with it an enduring stigma endorsed by legal prohibitions on the employment of ex-inmates in certain types of jobs and reinforced by employers who exhibit strong preferences against hiring ex-inmates (Pager 2007).

In addition, estimates suggest that close to 700,000 inmates will be released from state and federal prisons this year (West & Sabol 2008:3). If the past is any guide, almost half of them will be readmitted to prison or jail within three years (Langan & Levin 2002). Stable employment is one of the key elements of successful reintegration (Visher & Travis 2003; Yahner & Visher 2008), and many prisoner reentry programs—including those supported by the Second Chance Act of 2007—are designed to facilitate the employment of recently released inmates. Greater knowledge about variability in post-release employment experiences might be used to target resources more effectively.

Insofar as age at admission conditions the economic impact of incarceration, age-specific correctional and reentry programs to foster the integration of ex-inmates into the legitimate labor market may be worthwhile. Interventions targeted at individuals highly motivated to avoid re-incarceration may be particularly effective. In contrast, evidence that the effects of incarceration endure for all age groups would draw attention to the relevance of the prison system, regardless of age, in accounts of economic inequality. From the latter perspective, more wide-reaching or age-neutral policies aimed at reducing the size of the incarcerated population or providing legal employment protections for formerly incarcerated individuals might reduce the inequality-generating effects of incarceration. In either case, a better understanding of the age-related differences in the effects of incarceration on employment and wages would help clarify the role of the prison system in accounts of stratification and shed light on how criminal law and social policy associated with penal expansion and prisoner reentry may ameliorate or exacerbate social inequality.

Aging Inmates

The dramatic growth of the U.S. prison population since the mid-1970s is now well established. Perhaps less known is the concomitant aging within the inmate population. While the age-crime curve suggests that criminal activity peaks in the late teens, growth of the U.S. prison population has been accompanied by a significant aging of the prison population (Bushway & Tsao 2009; Bonczar 2003).

It is quite striking, though increasingly clear, that penal growth and the aging of the correctional population have not been due to large-scale changes in crime or criminality. Instead, a host of changes at the local, state, and federal levels with respect to law enforcement and penal policy are implicated in the expansion of the prison system. Law enforcement agencies have stepped up policing, prosecutors have more actively pursued convictions, and there have been myriad changes in sentencing policy that now mandate jail or prison time for nonviolent property and drug crimes (Mauer 1999; Tonry 1995; Western 2006).

Increased sentence lengths, parole revocations, and rapid increases in new commitments among older men—including commitments for the first time—are likely contributors to an aging population within state and federal prisons. Existing research has demonstrated compelling evidence of the aging of the inmate population (Bushway & Tsao 2009; Bonczar 2003) yet research has not established a definitive explanation for the aging of the prison population. Bushway and Tsao (2009) speculate that the aging of the prison population, while partially attributable to increases in sentence lengths, is primarily due to changes in the age structure of those who get sent to prison (2009:3).

Prison growth has been accompanied by increased attention—in recent years—to prisoner reentry. Ninety-five percent of all state and federal inmates are eventually released to the community (Peterselia 2003:3). Of them, nearly half are over age 35 (Freeman 2003:17). Evidence indicates that the reentry process differs significantly for older inmates and younger inmates: Older inmates face a lower risk of reincarceration (Freeman 2003) and are more responsive to work programs after release (Uggen 2000). However, surprisingly little research has examined how the labor market effects of incarceration vary by age at admission to prison despite alternative theoretical expectations and growing policy interest in the reentry process.

Life Course Conceptualizations of Incarceration Effects

A life course approach to criminal involvement emphasizes the importance of age-graded social control for understanding the age-crime curve (Sampson & Laub 1993). This argument suggests that the salience of social bonds—often associated with age—affects assessments of the risks and rewards of crime. Laub and Sampson (2001) identify aging, marriage, legal work, and a “reorientation of the costs and benefits of crime” (2001:3) as key predictors of criminal desistance. Research on criminal desistance finds substantial support for the importance of social bonds as determinants of criminal desistance. For instance, several studies indicate that aspects of marriage encourage desistance (Sampson & Laub 1993; Farrington & West 1995; Horney et al. 1995). Although marriage itself does not seem to affect desistance, marital stability, attachment, and involvement seem to decrease offending. In a similar vein, education and employment are useful in explaining desistance to the extent that they help establish bonds to conformity and shift the risks and rewards associated with criminal activity (Sampson & Laub 1993).

Related research on criminal desistance implies that bonds to conformity may vary by age quite apart from connections to institutions of marriage, schooling, and work. Shover argues that aging influences subjective contingencies, or what he calls “orientational, resolve-enhancing contingencies” (1996:130). Shover and colleagues (Shover 1996; Shover & Thompson 1992; Shover & Henderson 1995) contend that changes in offending over the life course are linked to the changing calculus of decisionmaking that occurs as people age. Just as nonoffenders experience age-related changes in their lives and perspectives on risk-taking, as offenders age they become more risk-averse and thus more likely to desist from criminal activity. For a number of reasons, including the development of social bonds and increased stakes in conformity or strengthened resolve to abandon crime, older men may view criminal activity as more costly than younger men. Shover (1996) argues that older ex-inmates have stronger preferences for not returning to prison than younger men, and the prospect of returning to prison is perceived as increasingly costly with age.

This explanation may help clarify not only why attraction to crime may change over the life course and as people age, but also how the attractiveness of work in the legitimate labor market may change. As criminal activity becomes more costly and people are willing to go to greater lengths to avoid spending time in prison, participation in the legitimate labor market may become more attractive. In this way, older offenders may be more willing than younger men to accept employment—even at low wages—in exchange for the legitimacy attached to working in the legal labor market. Older ex-inmates may have stronger preferences for conventional work than younger men who drop out of the conventional labor market to seek illegitimate gains. At the same time, weak employment prospects for young ex-inmates may reinforce beliefs and expectations about the costs and benefits of crime versus work in the legitimate labor market (e.g., Levitt & Venkatesh 2001).

Enduring Disadvantage of Criminal Justice Contact

Involvement with the criminal justice system through incarceration, arrest, or criminal conviction may also affect labor market outcomes through stigma. The stigma hypothesis is derived from the central tenets of labeling theory, which holds that individuals can be typed or labeled as “essentially deviant” by formal agents of the criminal justice system. Others often react to this label as a signal of a “master status” that precedes the existence of any secondary or more positive characteristics of an individual. Incarceration may operate as a “scarlet letter” (Nagin 1998), publicly labeling—and stigmatizing—ex-inmates as a social class. The general consequence of being labeled essentially deviant means that one is viewed as less trustworthy and rule-abiding (Becker 1963).

Research on the prevalence of social stigma and the labor market suggests that at least some employers use criminal history records when making hiring decisions. Holzer (1996) reports that between 30 and 40 percent of employers sampled in a survey of five major U.S. cities checked the criminal history records of their most recently hired employee, and about 65 percent of employers expressed that they “would not knowingly hire an ex-offender,” regardless of offense, preferring instead to hire other marginalized workers such as welfare recipients (1996:59). Experimental studies using criminal and noncriminal job applications also suggest that employers discriminate against individuals with criminal records (Boshier & Johnson 1974; Schwartz & Skolnick 1962; Pager 2003).

This explanation implies that the effects of incarceration on employment and wages may be invariant over the life course. Unexplained absences from the paid labor force, work experience gained in prison, and references from corrections or parole officers provide clear signals to employers of ex-inmate status. To the extent that employers view criminal activity and incarceration as markers of untrustworthiness or low productivity, they may be particularly reluctant to hire an ex-inmate. Inmates may face poor job prospects due to a generalized form of discrimination against those who fit a certain image. If employers do hire ex-inmates, wages may be low in anticipation of low productivity or high turnover.

Previous Empirical Work on the Economic Implications of Incarceration

Despite methodological challenges in isolating the treatment effect of incarceration (Bushway et al. 2007), existing research generally shows negative effects of incarceration on earnings, though the effects of incarceration on employment are mixed. Research shows that ex-inmates face earnings penalties of between 10 and 30 percent (Western et al. 2001), and results from survey and administrative data sources are generally consistent. And while employers clearly express a reluctance to hire ex-inmates (Pager 2003; Holzer 1996), recent research using administrative data finds sizeable, yet transitory, boosts in post-release employment levels (Sabol 2007; see also Tyler & Kling 2007). Although existing data cannot fully explain temporary post-release employment gains, findings are consistent with theories that emphasize the deterrent effects of incarceration, ex-inmates' heightened motivation to work in the legitimate labor market upon release from prison, or programmatic links to employment opportunities provided by corrections institutions.

Much previous research, especially work using National Longitudinal Survey of Youth (NLSY) survey data (Western 2002; Raphael 2007), has been unable to explore age-graded effects of incarceration adequately because of samples restricted to mostly young inmates. Those studies able to consider directly the timing of incarceration reveal some evidence of age-graded effects, yet they are limited in important ways. Uggen (2000) uses information from a quasi-experimental study of ex-inmates' involvement in post-release employment and finds that older ex-inmates are more responsive to work opportunities than younger ex-inmates. While Uggen's study powerfully demonstrates age differences in responsiveness to supported work programs, it is unclear how the study generalizes to the employment experiences of ex-inmates entering the open labor market. In addition, the data employed in Uggen's study were collected more than 30 years ago. Shifts in criminal justice policy, political context, and demographics in the last three decades may have changed how age conditions the relationship between incarceration and employment. Nagin and Waldfogel (1998) find that first-time conviction has a positive effect on income for young offenders, but a negative effect for offenders over age 30. They also find that subsequent convictions reduce income at all ages and argue that first-time convictions move workers off of career earnings profiles into jobs with flatter, and arguably lower, wage profiles (Nagin & Waldfogel 1998). Since their study focuses only on income from all sources, and not employment income per se, it is difficult to pinpoint the nature of incarceration effects (i.e., employment, hours worked, or wages).

In this study we explicitly examine whether the effects of incarceration on employment and wages vary in relation to age of incarceration. Theories of life course criminology lead us to expect that we should find variation in the economic effects of incarceration by age, as older inmates would be more likely to accept employment in the legitimate labor market. Theories emphasizing the stigma of incarceration lead us to expect age-invariant incarceration effects.

Data and Measures

We used administrative data, which included a wide age distribution of offenders, to examine how the economic effects of incarceration depend on the age at admission. We combined administrative data from the Washington State Department of Corrections and unemployment insurance (UI) records to investigate the consequences of incarceration on employment and wages. While existing research generally shows negative effects of incarceration on employment and earnings, previous estimates of incarceration effects that rely on survey data have not always been sensitive to challenges of causal inference. Past studies have often been unable to isolate the effects of spending time in prison from other factors that jointly affect the probability of incarceration and poor labor market outcomes. Recent policy changes in the state of Washington have led to the collection of a rich array of covariate information including information on age at admission, conditions of confinement, and prior work history that allow for a closer examination of alternative explanations for post-release effects on employment and wages.

We begin with data on men admitted to and released from state correctional facilities in Washington between 1990 and 2000. Data on offenders came from the Department of Corrections Offender Based Tracking System (OBTS). We linked these data with quarterly employment, hours, and earnings data from UI records using a unique identifier contained in both databases. More than 85 percent of inmates were located in available UI records. We collected all available UI employment and wage data between the first quarter of 1988 and the first quarter of 2002, which provided at least two years of labor market data before and after incarceration. Using these data, we constructed criminal and work history files for all inmates beginning in 1988, or when the inmate was 18 years old. This generated a dataset including 19,184 individuals who spent time in Washington State prisons in the 1990s and were observed, at least once, working in UI-covered employment. Each inmate had up to 57 quarterly observations, although at least some of those quarters were spent incarcerated in state prison. We limited our analysis to inmates who were at least 20 years old at the time of their first observed incarceration so that we could record their involvement in UI-covered employment as adults for at least two years before incarceration. This resulted in a sample size of 16,956. Descriptive information about the sample is shown in Table 1.

Table 1. Descriptive Statistics for Dependent and Independent Variables Used in the Regression Analyses, Washington State Inmates

Employment and Wages

An individual was coded as being employed if he had positive reported earnings within a quarter. The first column of Table 1 shows that offenders in Washington are only sporadically employed in UI-covered jobs. In order to be included in our analysis, the men we observed had to be employed at least once prior to incarceration and once after incarceration in UI records. That is, all men included in our analyses worked in employment reported for state unemployment insurance purposes during at least two quarters between 1988 and 2001. On average, however, inmates worked in UI-covered jobs in only 26 percent of our quarterly observations.²

Hourly wages were constructed by taking quarterly earnings divided by reported hours. In approximately 12 percent of cases where there were positive quarterly wages, the hours data were either missing or misreported. We found few systematic differences in the misreporting of hours, and for the wage analysis we excluded all quarters with zero reported wages or hourly wages of more than five times the median.³ Wage data were in constant dollars indexed to 1995. The average hourly wage for the sample was $9.02. In comparison, the minimum wage in Washington was $7.35 in 1995.

Analyses of employment also included information on prior work experience. We measured prior work experience as the proportion of previous quarters that the individual was reported working in UI-covered employment. At the time of admission to prison, the sample had worked, on average, in 29.4 percent of previous quarters. Wage analyses included information on industry of employment. Table 1 shows the distribution of inmates by industry in the last job prior to incapacitation. Like similarly educated noninmates, inmates were concentrated in the service, manufacturing, and retail trades. There were no significant shifts in the distribution of ex-inmates across industries after incapacitation.

Incarceration Variables

We estimated the effects of incarceration on labor market outcomes using several incarceration variables. We were primarily interested in the effects of having been incarcerated on employment and wages. Incarceration was coded to 1 in all quarters after release from prison. The coefficient on incarceration captured the mean difference in employment or log wages before and after incarceration, and can be thought of as the effect of having spent time in prison on post-release employment and wages. We attended to employment and wage effects immediately after release because prior research suggests that labor market experiences in the short term after release are critical determinants of longer-term employment and incarceration outcomes (e.g., Yahner & Visher 2008; Visher et al. 2008). We focused our attention on employment and wage outcomes following the first observed incarceration in the 1990s. We excluded all quarters in which an individual was currently incarcerated, and we also excluded quarters from, and following, subsequent prison stays.

To investigate the possibility that the effects of incarceration depend on age at admission to prison, we divided age at admission into four categories: 20–24, 25–29, 30–34, and 35 or older. The sample was relatively evenly divided into the four groups; 29 percent of the sample was admitted to prison between ages 20 and 24, 22 percent was admitted between ages 25 and 29, 19 percent was admitted between ages 30 and 34, and 29 percent was admitted after age 35. We analyzed the effects of incarceration separately by age at admission groups. Table 1 also shows descriptive information for these subgroups.

Additional information about incarceration included the number of quarters served during the most recent incarceration, involvement in a work release program, and time since incarceration. The average length of stay in prison was 6.3 quarters.⁴ Just over 30 percent of inmates participated in a work-release program while incarcerated. We do not have specific information about the nature of work-release, although there is evidence that work-release is more common among older offenders, long-serving offenders, and those with particularly poor employment histories. That is, initial descriptive information suggests that work-release programs may attract particularly disadvantaged workers. A priori, we expected participation in work-release programs to have positive effects on post-release employment and earnings. However, if inmates who participate in work-release programs are a particularly disadvantaged group, any positive effects of work-release might be offset by existing liabilities.

The mean length of follow-up was 18.6 quarters.⁵ The vast majority of inmates (78.8%) served only one prison term between 1990 and 2000, and we observed their employment and wage outcomes through the first quarter of 2002. For inmates who were readmitted before 2000, we only included information on wage and employment outcomes up to a new prison stay. We restricted the data in this way in an attempt to disentangle the effects of age from the cumulative effects of multiple prison stays, which we expected to correlate strongly with age.

Other Variables

Analyses of employment and wages included both quarterly age and year effects. The mean age of the sample in 1988 was 23.8 years and ranged from 8 to 82. We did not include individuals in the analysis until they reached age 18. In 2002, after 14 years, the mean age of the sample was approximately 37 and ranged from 22 to 96. Although we observed inmates at a range of ages, our sample was concentrated in the early adult years. In effect, we examined the consequences of incarceration on employment and wages through the prime working ages although the variance of age in our sample was much larger than in panel survey data such as the NLSY. As a consequence, we could carefully examine how the effects of incarceration vary by age at admission to prison.

Table 1 also shows additional time-invariant characteristics. Because we employed a modeling strategy that controlled for both measured and unmeasured time-invariant characteristics via individual fixed effects, we did not estimate time-invariant parameters such as offense type, race, ethnicity, and education. Nonetheless, we report time-invariant characteristics in Table 1 to better describe the sample.

Data on type of offense showed that two-fifths of inmates served time for violent offenses, but that almost 60 percent served time for nonviolent drug or property crimes. Much of the growth in the prison population has been fuelled by nonviolent property and drug offenders and, as a consequence, our results likely generalize to a growing fraction of the prison population. We considered offense type to be a proxy for underlying criminal propensities and treated it as fixed.

Just as there is racial and educational disproportionality in imprisonment nationally and in other states, incarceration is concentrated among minority, low-skilled men in Washington. The majority of the sample was white, although blacks and Hispanics were overrepresented in relation to the racial composition of the state. And the sample was relatively poorly educated; fully one-third of the sample had not finished high school, and another 46 percent had only a high school diploma. Many inmates acquire additional education while in prison, but we do not have reliable time-varying measures of educational attainment, so we treated education as fixed.

Method

We used a pooled cross-sectional time-series analysis to analyze the effects of incarceration on employment and log hourly wages. Much of the research studying the effects of incarceration on labor market outcomes is concerned with how selection into prison affects estimates of incarceration effects. That is, spending time in prison is highly correlated with existing social and economic disadvantage. Therefore, are the effects of spending time in prison on employment and wages actually due to incarceration or to preexisting labor market liabilities that are also associated with spending time in prison?

We attempted to deal with this concern in three ways. First, we included a number of previously unmeasured controls that influence both the likelihood of incarceration and employment outcomes. Specifically, we included information on length of stay, whether or not the individual participated in a work-release program while in prison, and information about prior work experience in UI-covered jobs. Second, as previously mentioned, we estimated models using individual fixed effects to account for time-invariant observed and unobserved characteristics that vary across individuals. We used a conditional fixed-effects logit model to estimate the effects of incarceration on post-release employment and a fixed-effects regression to estimate log hourly wages. To explore whether the effects of incarceration vary by age at admission, we estimated separate fixed-effects models for each age-at-admission subgroup.⁶ Third, we tested the sensitivity of our results to different sample specifications. The main results in the article focus on the analysis of the full complement of inmates for whom we had complete data and observed employed in UI-covered jobs at least once prior to and once following incarceration (N=14,485). Although these data can describe the incarceration effects post-release among inmates in Washington State, they tell us less about employment and wage outcomes in the absence of incarceration for similarly situated men. We tested the sensitivity of our results by comparing them to results generated by two additional sample specifications in supplementary analyses, described below, that attempted to isolate a comparison group to our inmate sample. Descriptive information on the supplemental samples is shown in Appendix Table 1.

The first supplemental analysis examined the labor market experiences of a subsample of men admitted early in the observation period in relation to experiences of similar men who were incarcerated later in the observation period. From our larger sample of inmates, we first selected data from 1,188 “early admits” incarcerated in state correctional facilities between 1990 and 1995. We then matched early admits with a comparison group of men we considered “at risk” of incarceration but did not observe in prison until at least 1997. We defined men in the comparison group as “at risk” because they eventually would be incarcerated at least once between 1997 and 2001. We matched early admits with at-risk men on the basis of chronological age, race, education, offense type (for first observed offense), and the risk of recidivism as measured by a risk-assessment inventory (The Level of Service Inventory - Revised [LSI-R]; Andrews & Bonta 1995; Gendreau et al. 1996).

This method has been used in previous research (Grogger 1995) and should provide a very conservative test of incarceration effects since we restricted our analyses to a subset of inmates and other individuals we knew would be admitted to prison for the same offense at a later date. Nonetheless, there was some concern that early admits differed from later admits in unobserved ways that might undermine our results. On the one hand, if the same skills that enabled “at risk” men to elude prison were also associated with better labor market outcomes, we would be likely to overstate the negative consequences of incarceration. On the other hand, if those we defined as “at risk” were disproportionately involved in the illegal economy, we might only observe particularly strong performers in the legitimate labor market and therefore understate the negative effects of incarceration.

The second supplemental analysis examined the labor market experiences of a subsample of inmates with general equivalency degrees (GEDs) in relation to a sample of similarly low-skilled men we did not observe in prison. We began with data from 3,362 inmates who received a GED in the state of Washington. Drawing from a statewide database of GEDs awarded, we matched inmates with GEDs with a comparison group of men with GEDs we did not observe in state correctional records from 1990 to 2000. Following Reville et al. (2002), we matched inmates with noninmates on the basis of age, GED date and test scores, employment, and wages in the last quarter employed before incarceration. We only kept matches where the hourly wage difference between inmates and noninmates was less than 10 percent in the last quarter employed before incarceration.

This strategy allowed us to examine the labor market experiences of ex-inmates in contrast to noninmates with similar cognitive skills, and potentially draw comparisons to analyses conducted with survey data that include inmates and noninmates, such as studies by Freeman (1991) or Western (2002) using the NLSY. The comparison might be biased, however, if there were unobservable differences between inmates and noninmates that affected labor market outcomes in ways that we could not capture with available measures. Indeed, despite a careful matching strategy, consistent pre-incarceration differences in employment and wages among this sample suggested that inmates and noninmates with GEDs likely differ on unobserved factors associated with labor market productivity. Therefore we may have run the risk of attributing unobserved differences in productivity to incarceration in this subsample. It also may be the case that inmates with the initiative to get GEDs may differ from otherwise similar inmates, undermining the comparability of results across samples. Nonetheless, given the challenges in isolating the treatment effect of incarceration, comparing the results of the full complement of inmates to GED and early/late alternative sampling strategies provided an indication of the robustness of our findings.

Results

Incarceration had significant and sizeable effects on both employment and wages after release from prison. Perhaps surprisingly, but consistent with other recent research using similar administrative data (e.g., Tyler & Kling 2007; Sabol 2007; see also Holzer 2008), incarceration had positive short-term effects on employment for all but the youngest offenders. Incarceration had negative effects on wages for all inmates, and there was some evidence that the effects were largest among the oldest offenders.

Table 2 shows the results from regressions of employment on incarceration and a number of other covariates (also see Appendix Table 2). While incarceration had no significant effect on the immediate post-release employment fortunes of men admitted to prison between ages 20–24, it had significant and positive effects on the immediate post-release employment of the other age groups over age 25. Men admitted to prison between ages 25–29 experienced a 15 percent increase in the probability of employment following a stay in prison, compared to their employment prior to entering prison. The effect of spending time in prison on employment outcomes was more dramatic among men incarcerated over age 30.

Table 2. Unstandardized Coefficients From the Logistic Regression of Employment on Incarceration (Standard Errors in Parentheses)

^* p less than 0.05.

Note: Standard errors are in parentheses. All models include year dummies. Complete results for year dummies are shown in Appendix Table 2.

Our analysis included information about conditions of confinement, including time spent in prison, involvement in a work-release program, and time since incarceration. Therefore, the coefficient on incarceration captured the difference in the probability of employment prior to incarceration and the probability of employment in the quarter immediately following release from prison associated with incarceration. It is perhaps somewhat surprising that incarceration was positively related to post-release employment for many inmates, especially in light of evidence from employer surveys and audit studies indicating employer preferences for employees without criminal records (Holzer 1996; Pager 2003).

However, these results were not entirely unexpected. The short-term increase in employment could be due to the deterrent effects of spending time in prison or increased motivation on the part of ex-offenders to avoid reincarceration and find work in the legitimate labor market, at least in the quarters immediately following release. Evidence of short-term boosts in employment after release from prison was consistent with research using administrative data from other states that reports similar post-release increases in employment in formal-sector jobs (Tyler & Kling 2007; Sabol 2007; and possibly Uggen 2000; see also Holzer 2008). Tyler and Kling (2007) show that inmates in Florida experience increases in UI-reported employment immediately upon release from prison. Sabol (2007) documents a similar trend using data from prisoners in Ohio.

Unique features of Washington's correctional system may also partially explain positive effects of incarceration on post-release employment. Washington State has an extensive post-release supervision program, and more than 70 percent of inmates released in 2002 had one year of mandatory post-prison community supervision. Some of the positive post-release employment effect may be due to job placements following release. Supervisory personnel may engage in positive labeling of ex-convicts, and employers may be encouraged by supervisory personnel to employ recently released inmates. Ex-inmates assigned to community supervision also have access to a network of potential employers and employment contacts through the supervisory program.

Important to note, however, the immediate post-release increase in employment was transitory. Our estimates suggested that employment probabilities fell to pre-incarceration levels for all age groups within 6–10 quarters of release from prison. The sizeable, and significant, negative effect for quarters since release (−0.032 to −0.049) indicated that on average the probability of employment declined with time out of prison. For men incarcerated in their twenties, there was substantial evidence that their employment fortunes went from bad to worse after release from prison: The effects of incarceration on immediate post-release employment were not statistically different from zero, and the significant and relatively large negative effect of time out of prison (quarters out) indicated that the employment prospects of men incarcerated at young ages only got worse with time out of prison. Consistent with research from audit studies, this finding highlights the poor employment fortunes of young ex-inmates (e.g., Pager 2003).

In summary, our results indicated that most men experience a short-term employment boost after incarceration, followed by a return to pre-incarceration employment levels. There was some evidence that men incarcerated at young ages have less promising employment prospects after release from prison compared to men incarcerated at older ages. It is possible that men incarcerated in their twenties may be penalized more harshly for spending time in prison, or they may be opting out of the legal labor market in pursuit of illegal gain. We had no direct way to measure the effects of stigma, but we could examine whether young men were opting out of the job market by raising their reservation wage. The term reservation wage refers to the lowest wage at which one is willing to accept work in a particular type of job. If men have alternative sources of income or strong preferences for engaging in other activities, they may require a particularly high wage—reservation wage—to take a UI-covered job. If this is the case, we should have found an increase in the post-release wages of men admitted to prison in their twenties.

While we could not rule out the possibility that older men are shifting careers into occupations where they are more likely to find employment, differences in the effects of incarceration by age at admission led us to suspect that as men finish their “criminal careers,” they are likely to lower their reservation wage in order to accept employment in the legitimate labor market. If this is the case, we should have also found more negative effects on the post-release wages of men incarcerated at older ages.

Turning our attention to the analysis of incarceration on post-release wages in Table 3, we found that spending time in prison was associated with declines in hourly wages for men admitted to prison at any age (also see Appendix Table 3). Although the magnitude of the effect of spending time in prison on post-release wages differed by age at admission to prison, men in every age group experienced declines in wages of between 4.6 and 6.8 percent associated with spending time in prison. Men admitted to prison in their early twenties could expect a wage loss of 4.6 percent after release from prison, whereas those incarcerated after age 35 could expect a 6.8 percent loss (a 47 percent larger penalty). These effects were large and consistent with previous research that finds a significant wage penalty associated with spending time in prison.

Table 3. Unstandardized Coefficients From the Regression of Log Hourly Wages on Incarceration (Standard Errors in Parentheses)

^* p less than 0.05.

Note: Standard errors are in parentheses. All models also include industry and year dummies. Complete results for industry and year dummies are shown in Appendix Table 3.

Again, these results are particularly noteworthy because we controlled for conditions of confinement in addition to prior work experience and other factors associated with wages. Although we did not have extensive information on program participation while in prison, the effect of incarceration persisted even when we controlled for time spent in prison and involvement in work release programs. These results suggest that arguments emphasizing skill deterioration or the loss of human capital are insufficient to explain incarceration effects. In other words, we cannot explain the effects of incarceration on wages solely as a function of the loss of skills and competencies associated with time out of the paid labor market (and in prison). That these results persisted when controlling for other explanatory factors lends some merit to the idea that spending time in prison confers an enduring stigma that undermines men's labor market opportunities.

Because we included a measure of time since incarceration, the incarceration variable captured the difference in mean log hourly wages before incarceration, with the mean log hourly wage in the quarter immediately following release (see Appendix Table 4). The effect of time out of prison was positive, though small, indicating that wages began to recover slowly in the post-incarceration period.

Sensitivity Analyses

Supplementary analyses with two additional samples provide some evaluation of the robustness of the age-graded effects of incarceration reported above. These sensitivity analyses are shown in Table 4 and illustrated in Figures 1 and 2. Table 4 compares the effects of incarceration on employment and wages across the three samples. The results were consistent in sign and significance across all three samples. That is, there was consistent evidence that inmates exhibited a temporary increase in the probability of working in a UI-covered job after release from prison and exhibited significant wage declines after release from prison.

Table 4. Comparison of Effects of Incarceration by Age, 3 Samples (Standard Errors in Parentheses)

^* p less than 0.05.

Note: Standard errors are in parentheses. All models also include additional controls, and complete results are available from the authors.

Figure 1. Relative effect sizes of incarceration on post-release employment by age at admission.

Figure 2. Relative effect sizes of incarceration on post-release log hourly wages by age at admission.

For employment, there were differences in the magnitude of age-graded effects of incarceration across samples. The early/late sample generated larger estimates of incarceration effects across all age groups. Although the pattern of relative effects by age group was somewhat consistent with the results reported above—estimates for men admitted at older ages were greater than those for men admitted in their twenties—the differences in effects were not statistically significant across age groups. The employment patterns from the GED sample did exhibit statistically different effects by age, but in patterns divergent from the other samples. Specifically, among inmates with a GED, young inmates exhibited greater employment gains than older inmates.

Figure 1 graphs the unstandardized coefficients of incarceration on employment by age at admission for the three samples. Although the relative pattern of age-graded effects varied by sampling strategy, Figure 1 shows that all of the estimates of post-release employment effects were generally positive effects. Among the all-inmate sample (hatched bars), the effect of incarceration on post-release employment was indistinguishable from zero for young inmates, whereas it was positive, and significant, for older inmates. The early/late and GED samples confirmed the positive post-release effects of incarceration on employment but did not confirm the same age differences in the effects. The effects of incarceration on employment were larger in magnitude among the early/late sample, yet the differences in the effects by age at admission were not statistically significant. Among the GED sample, incarceration also had a positive effect on post-release employment, but the age-graded pattern of results was inconsistent with the full sample of inmates. Among GED holders, young ex-inmates exhibited the largest employment increases in the short-term after release.

Figure 2 graphs the unstandardized coefficients of incarceration on wages by age at admission. Among all samples and age groups, there was consistent evidence that inmates exhibited significant wage declines after release from prison. Although the magnitude of the declines ranged across age groups (4.6–6.8% for all inmates, 9.1–22.7% for the early/late sample, and 7.5–8.8% for the GED sample), the differences were not statistically significant. The lone exception occurred with the all-inmate sample (hatched bars) between the wage effects for men admitted to prison at the youngest (20–24) and oldest (35+) ages. The general pattern of results across samples, however, confirmed negative and sizeable post-release wage effects regardless of the timing of incarceration.

Discussion

Using particularly rich administrative data about inmates, this article examines how the consequences of incarceration depend on the timing of incarceration in the life course. In summary, the results emphasize enduring disadvantage resulting from imprisonment, regardless of timing in the life course. Even the most motivated ex-offenders, older men and younger men with GEDs, suffer sizeable and significant wage penalties and, over time, decreased likelihood of employment.

To return to our initial hypotheses, we do not find consistent evidence that incarceration effects vary by age at admission. Instead, incarceration appears to have important consequences for employment and wage outcomes regardless of when individuals are admitted to prison. Although results provide some indication that the effects of incarceration may depend on age of admission, the weight of evidence across multiple sampling strategies underscores the consistent consequences of incarceration on employment and wage outcomes despite variation in age at admission to prison.

For example, initial results with the full sample of inmates indicate that men incarcerated in early adulthood have the poorest post-release employment outcomes, whereas those admitted to prison at older ages do significantly better securing jobs in the legitimate labor market in the short term after release from prison. These patterns, however, are not robust to alternative sampling strategies. Supplementary analyses with a GED inmate/noninmate sample suggest that the relationship between age at admission and employment effects may be more complex than initially indicated. For example, when we focus on a sample that includes only men with a GED, we find the most positive employment prospects after release from prison among men incarcerated at the youngest ages. It may be the case that older ex-inmates and ex-inmates with a GED—especially young ex-inmates with a GED—are particularly motivated to work in the legitimate labor market.

For wages, we conclude that spending time in prison exacts significant penalties regardless of age at admission. While initial results with the full sample indicate that men admitted to prison later in life—over age 35—experience the greatest wage penalties associated with spending time in prison, results from supplementary analyses do not find similar age-graded effects of incarceration. The supplementary analyses establish that all ex-inmates face large wage penalties. The magnitudes of the effect of incarceration on post-release wages fall between 4 and 23 percent, in line with those established by previous research (Western et al. 2001).

The sensitivity analyses, therefore, confirm short-lived employment gains and persistent wage penalties associated with incarceration. At the same time, the sensitivity analyses challenge ideas about age-graded effects of incarceration and suggest that the effects of incarceration may be affected not only by a shifting calculus of crime that occurs across age, but may also vary in relation to other social and demographic characteristics such as education.

We caution that there is reason to suspect that our results may be overly optimistic about the employment and earnings of ex-inmates. Our analysis is restricted to inmates with relatively short periods of incarceration and to those incarcerated in the 10-year period between 1990 and 2000. Former inmates were entering a relatively tight labor market, particularly later in the decade, as unemployment reached historic lows. In addition, our results are based solely on administrative records and on employment covered by UI. We do not have information on employment not covered by UI records, and if inmates are disproportionately involved in the informal economy prior to incarceration we may understate relative losses in employment and wages in the post-release period (or overstate gains).

At the same time, our results may be overly pessimistic. Although we are able to capture the incapacitative effect of readmission to state prison facilities, we do not have information about stays in local jails. As a consequence, we may overstate incarceration effects because we cannot determine if men are unemployed because they are reincarcerated in jail facilities. Unfortunately, we cannot know the direction or extent of bias of our results without additional (and unavailable) information.

We also note that our data come from a single state, in which incarceration rates are relatively low and unemployment is high. Washington's incarceration rate is among the lowest in the nation, and as a consequence the state's inmate population may represent particularly severe offenders. The incarceration rate in Washington, which includes inmates in state and federal prisons, was 271 (per 100,000) in 2007. By contrast, the incarceration rate was 506 nationwide in 2007 (West & Sabol 2008:20). On the one hand, we might expect that the post-release effects in Washington would be smaller than those in other states because the system is less punitive. On the other hand, the effects of incarceration may be more severe in Washington because of the increased stigma associated with incarceration. Whereas low-level offenders in other states may be imprisoned, low-level offenders may be less likely to serve time in Washington. In addition, Washington consistently has among the highest unemployment rates in the United States, and deindustrialization across the state has led to particularly grim employment prospects for low-skilled men. We cannot resolve whether or to what extent these conditions influence the generalizability of our results, but they certainly highlight the need for additional research in other contexts.

Despite all these caveats, our results have important theoretical significance. Theoretically our results are consistent with structural explanations for declining economic fortunes among low-skilled men. Post-release employment instability and wage declines could be explained by the movement of offenders into short-term or temporary jobs, rapid dismissals from jobs, prohibitions on employment of ex-inmates in industries with job security and strong earnings growth, and discrimination on the part of employers. The ubiquity of wage penalties among ex-inmates regardless of age at incarceration emphasizes the central relevance of the growth in incarceration for contemporary accounts of economic inequality.

Inmates exhibit increases in employment in the immediate post-release period, yet these employment gains are temporary and employment rates soon return to their pre-incarceration levels. At least some of the post-release employment boosts may be particular to Washington either because of programmatic initiatives associated with supervisory programs or labor demand. However, similar findings in other states, including Ohio and Florida (Sabol 2007; Tyler & Kling 2007), suggest that our findings may generalize beyond Washington and are consistent with theories of orientational shifts among ex-inmates in the immediate post-incarceration period.

While these data do not allow for a careful evaluation of shifts in the calculus of crime in relation to age or additional education, they imply that life course accounts of incarceration effects may be trumped by the stigma of incarceration. Our results suggest that offenders at all ages are highly motivated to seek employment in the legitimate labor market immediately after release from prison. However, high turnover and low pay in the legitimate labor market do not foster the kind of social bonds associated with criminal desistance.

Our results could also be explained by poor health associated with incarceration (e.g., Massoglia 2008a, 2008b; Schnittker & John 2007). Poor health may affect a person's ability to work, thereby impacting wages. Our data do not include measures of health, so we cannot directly test this proposition. However, the persistence of incarceration effects even when controlling for work experience lends support to arguments that emphasize the economic relevance of the “scarlet letter” (Nagin 1998) that adheres to ex-convicts. In any case, additional research on the consequences of incarceration on labor market outcomes should pay careful attention to how the effects of incarceration are influenced by age, education, health, and other characteristics that may not only affect the calculus of crime, but also the stigma associated with spending time in prison.

Conclusion

Throughout the late twentieth century, the prison system expanded not only in size but also in scope. Prison time is no longer reserved for youthful indiscretions or long-term chronic offenders. Spells of incarceration now punctuate the life course of a growing fraction of Americans, with the fastest growing group of new inmates over age 30. As the prison population has grown, so has the number of inmates released to the community. In 2007 alone, more than 750,000 individuals were released from federal or state prisons (West & Sabol 2008:3), and analysts expect that almost half of them will be returned to prison within three years.

States and localities are increasingly turning to prisoner reentry programs to help ease returning offenders' transition to the community to increase the chances of successful reintegration and reduce the likelihood of reincarceration. Even the federal government has lent support to the project through the Second Chance Act of 2007, signed into law by then-President George W. Bush. In contrast to the punitive penal philosophy that guided prison expansion, prisoner reentry programs typically are founded with rehabilitative aims (Visher & Travis 2003; Travis & Visher 2005). Federal, state, and local governments are all invested in successful reentry programs as the prison system expands beyond its capacity and corrections account for an increasing proportion of government budgets.

Yet the rehabilitation of offenders may prove elusive if the economy is the central culprit. As the prison system has expanded, spending time in prison has grown increasingly concentrated among low-skilled men. Nearly 1 in 10 white men who have not finished high school can expect to serve time in a federal or state prison; 6 in 10 black male high school dropouts will spend at least a year behind bars (Pettit & Western 2004:164). Just as declining economic opportunities for low-skilled men are implicated in the prison buildup, increasing economic opportunities is critical for offender reintegration. Our research joins a chorus of previous studies calling for the inclusion of the prison system in accounts of economic inequality. However, by documenting the durability of incarceration effects over the life course our research not only suggests that we would do well to revisit long-held beliefs about the age-crime profile, but also draws attention to the philosophical tensions between the premise of the prison buildup and the aims of prisoner reentry programs.

Inmates may experience programmatic incentives or orientational shifts in prison that connect them to legitimate employment, but the boost is not sustained. Many reentry programs face similar challenges of sustaining motivation on the part of ex-inmates to stay clean, which is in part a function of the resources and institutional support available to ex-inmates. If, as our results suggest, even those who are highly motivated to work in the legal labor market cannot sustain meaningful employment and attain economic self-sufficiency, then attention should be redirected to how to better incorporate offenders into the economy. Short of that, even those with strong motivation to work in the legitimate labor market are at greater risk of returning to prison or jail.

The findings reported here raise questions about whether the life course events that capture the experience reported by Sampson and Laub (1990) and Laub and Sampson (2001) are relevant in the contemporary context, where incarceration has become a normative life event in some social and demographic groups, and the labor market increasingly preferences highly skilled labor. While we cannot alone resolve those issues, future research should investigate how incarceration affects life course transitions in the contemporary context.

Our results imply that reentry initiatives targeted at older offenders or highly motivated young offenders may find a receptive audience, yet their effectiveness is dependent on the availability of high-quality employment, job stability, and prospects for wage growth. The inequality-generating effects of incarceration may be better stemmed through a process of deincarceration or widespread adoption of legal protections for ex-inmates as workers. Employment and wage discrimination is pernicious, and ex-felons may represent an important social class in need of legal protections.

An even more radical solution would involve enhancing the economic skills and capacities among the most disadvantaged Americans; such investments may keep a large fraction of nonviolent drug and property offenders out of prison or jail in the first place. Even before incarceration, offenders have relatively poor labor market fortunes, sporadic employment histories, and extremely low wages. In fact, prison expansion of the past 30 years has disproportionately affected the lives of low-skilled men (Western 2006). Insofar as legitimate work provides a turning point for some offenders (Uggen 2000), the promise of reintegration for crime control hinges on improving labor market opportunities to make work in the legitimate labor market a viable and sustainable path to economic self-sufficiency.