Abstract This paper is one of the first systematic assessments of ex-felons’ workplace performance. Using FOIA-requested data from the Department of Defense, we follow 1.3 million ex-offender and non-offender enlistees in the US military from 2002 to 2009. Those with a felony background show no difference in attrition rates due to poor performance compared to those without criminal records. Moreover, ex-felons are promoted more quickly and to higher ranks than other enlistees. At the same time, we find that ex-felons are slightly more likely to commit a legal offense in the military system (5 percent of non-felons compared to 6.6 percent of ex-felons). We also find a higher rate of work-related deaths among the ex-felon sample; closer evaluation of limited data suggests this may be driven by ex-felons being assigned more often to combat positions. Overall, our study shows that the military’s criminal history screening process can result in successful employment outcomes for ex-felons, at least in terms of job mobility and reliability, to the mutual benefit of employer and employee. An important question arising from this analysis is whether the military’s “whole person” review can apply successfully to the civilian sector. The assumption that a criminal record signals an undesirable employee is widespread, and guides the hiring behavior of many US employers. More than 70 percent of businesses conduct background checks (SHRM 2012), and the majority report that they would be unwilling to hire an individual convicted of a serious criminal offense (Holzer, Raphael, and Stoll 2004). Employers report concerns over ex-offenders lacking character, work ethic, ability or some combination thereof (Holzer 1999; Pager 2007). At the same time, we know that the vast majority of individuals who commit crimes desist in the transition to adulthood (Hirshi and Gottfredson 1983), with employment itself playing an important role in this transition (Sampson and Laub 1995; Uggen 2000). How, then, should we think about the prospects of ex-offenders in the workplace? Is the widespread aversion to hiring ex-offenders warranted or a self-fulfilling prophecy? Very little empirical evidence exists with which to assess the workplace risk or potential of individuals with criminal records. Meanwhile, a little-known initiative has been operating in the US military, which regularly recruits and hires individuals with felony-level criminal records to serve. With appropriate adjustments, this program presents a compelling test case to understand ex-offenders’ work performance when they are given a reprieve to enter a workplace, in this case one of America’s largest employers.1 This paper represents one of the first efforts to provide a systematic evaluation of ex-offenders in the workplace. Using original data assembled from military administrative records, we follow 1.3 million ex-felon and non-felon individuals who enlisted during the period 2002–2009. On a number of dimensions, we find that ex-felons perform as well as or better than counterparts with no criminal record. Those admitted with felony waivers are no more likely to attrite due to poor performance compared to those without. Moreover, ex-felons are promoted more quickly and to higher ranks than other enlistees. At the same time, some outcomes appear less favorable for felon enlistees: We see a small increase in the likelihood of being discharged for committing a legal offense in the military system (5 percent of non-felons are discharged for criminal behavior compared to 6.6 percent of ex-felons). We also find that those with serious criminal records have a higher rate of death, a finding that raises a number of questions. Given this mix of risk and rewards, we consider the role of ex-offenders in the workplace, and the potential implications of this research for civilian employers. Background The Criminal Justice System and Its Implications for Employment The expansion of the criminal justice system in the United States is by now familiar to scholars, with dramatic increases in the use of policing, community corrections, and incarceration over the past four decades (Travis, Western, and Redburn 2014). It is estimated that roughly 8 percent of the working-age population now has a felony conviction and 92 million Americans have some form of criminal record (Love 2011). The rapid increase of job seekers with criminal pasts has important repercussions for the American labor market. Former felons are barred from a variety of occupations, with the number of licensure restrictions increasing significantly since 1970 (Love, Roberts, and Klingele 2013). Criminal arrest and conviction records have become increasingly easy for employers to access via online repositories, and it is now commonplace to require applicants to report criminal histories and to undergo background checks (SEARCH Group 2005). Not surprisingly, the rate of employment among ex-offenders is extremely low. Studies of ex-prisoner populations report that roughly half remain jobless up to a year after prison release (Petersilia 2003; Sabol 2007; Visher, Yahner, and Vigne 2010). Likewise, general population panel studies find a strong negative association between the experience of incarceration and subsequent employment rates (Apel and Sweeten 2010; Freeman 1992; Western 2006) and earnings (Pettit and Lyons 2009; Western 2006). Given the size of the ex-offender population, these patterns suggest troubling implications for aggregate patterns of joblessness and economic insecurity (Western and Pettit 2005). Much of the scholarship on the employment of ex-offenders focuses on the barriers that result from a criminal record. Employers demonstrate strong reluctance to consider applicants with criminal records (Holzer, Raphael, and Stoll 2004), and matched-pair audit studies suggest that ex-offenders are roughly half as likely to receive a callback or job offer relative to equally qualified applicants with no criminal record (Pager 2003). These barriers hinder the reentry process and make it more difficult for those with criminal pasts to move on to stable, productive lives. Research points to employment as one of the strongest predictors of desistance from crime (OPRE 2012; Uggen 2000). If employers are unwilling to hire ex-offenders, however, this critical pathway remains largely out of reach. At the same time, employers may have reason to be concerned about hiring ex-offenders. Many of the characteristics leading to criminal justice involvement in the first place—crime, drug and alcohol addiction, mental health problems, disadvantaged background, and lack of human and cultural capital—may also make ex-offenders poor prospects for stable employment (BJS 2001; Grogger 1995). Each of these characteristics is arguably related to worker quality, and may be difficult to observe directly. To what extent are employers’ concerns about ex-offender behavior borne out? Currently, we have little direct evidence with which to answer this question. But we at least know that the number and severity of prior offenses shows a strong relationship with the likelihood of reoffense (Barnes and Hyatt 2012). Beyond these fixed characteristics, many of the relevant risk factors vary with time and context. For example, the risk of recidivism falls steadily with time since arrest, with nearly 60 percent of recidivism occurring within the first year (BJS 2014). Over time, those with a prior arrest start to appear indistinguishable from the general population in the likelihood of subsequent offense (Blumstein and Nakamura 2009; Kurlyche, Brame, and Bushway 2006). Additional evidence suggests that this recidivism curve is substantially reduced through employment. A randomized evaluation of a transitional jobs program found that those who started working within three months of release from prison were 22 percent less likely to be convicted of a new crime within three years of release than those in the control group (OPRE 2012). Likewise, a randomized evaluation of a national jobs program found that, among those over age 26, assignment to low-wage employment post-release reduced the risk of rearrest by 22 percent (Uggen 2000). Thus, the predictive value of a criminal record is heavily influenced by contextual factors such as age, time since offense, and intervening work experience. Reflecting these complexities, the Equal Employment Opportunity Commission issued new guidelines in 2012 proscribing hiring policies that categorically ban applicants with criminal records (EEOC 2012). Instead, the commission directs employers to consider factors such as time elapsed since the offense, evidence of rehabilitation, and the relationship between the crime and job in question, in order to offer a case-by-case assessment of the relevance of a given criminal record. Can such a review process effectively address employers’ concerns about hiring ex-offenders? There is very little direct evidence with which to evaluate how ex-offenders fare in the workplace, with or without extensive screening. Our research offers among the first available evidence with which to assess this question, using the military context as case study. Prison-Military Linkage While representing highly distinct arms of the state, the military and the criminal justice system have enjoyed an important and long-standing association. Part of this is reflected in the popular assumption that barracks discipline will set delinquent youth on the right path, as embodied by the militaristic structures of many programs for juvenile delinquents. During World War II, judges often ordered criminals to military service as an alternative to trial or prison (Mattick 1954; Shattuck 1945). And a number of studies have, in fact, found some desistance effect of military service for criminal delinquents who joined the military during the Vietnam, Korean War, and WWII eras (Bouffard and Laub 2004; Sampson and Laub 1996). Today it is illegal to send people to the military for reasons of punishment or rehabilitation. But the ties between criminal justice and military institutions remain important. According to the US federal legal code, individuals convicted of a felony may not enlist in any branch of the military. However, exceptions may be made by the military through the issuance of “moral character waivers.” A felony waiver approval is required of military recruits if they have a history of misconduct,2 such as a conviction or other adverse adjudication of a major misconduct offense. We refer to this waiver type as a “felony waiver.” Misconduct waivers are especially prevalent when the supply of enlistees is low, such as during wartime and when the national economy is thriving. To illustrate, criminal waivers (especially those for misdemeanors and other less serious offenses) doubled during the height of the wars in Iraq and Afghanistan compared to the late 1990s, a peacetime era of economic expansion (Boucai 2007). In the past five years, as the military has downsized its personnel and upscaled its recruitment criteria, felony waivers have become rare again. Our study examines an era when felony waivers were relatively plentiful and asks, what happens when the typical barriers facing ex-offenders in the labor market are reduced, and a screening process similar to that recommended by the EEOC is used? How does the attrition and performance of those with felony-level criminal records compare to recruits with no reported felony history? It is important to note that conduct waiver enlistees are not a random draw from the ex-offender population. Like most civilian employers, the military screens carefully on educational background and applicant skills. Specifically, the military typically requires a high school degree or equivalent and average scores on the Armed Forces Qualifying Exam (AFQT). In addition, when considering individuals for conduct waivers, the military conducts what is referred to as a “whole person” evaluation, a review that bears striking similarity to the EEOC guidelines for evaluating applicants with criminal histories. This military screening process takes into consideration the “who, what, when, where, and why” of the offense, which includes the age when the crime was committed and the circumstances and severity of the offense.3 In addition, the recruit’s compensating qualifications are taken into consideration and evaluated in a personal interview, and letters of recommendation are requested attesting to the applicant’s character or suitability for enlistment. Scrutiny is greater the more serious the offense, with felony offenses requiring approval by the highest level of authority on the recruiting command. Furthermore, a waived recruit’s offender status is sealed once they enlist, so as to protect them from labeling bias as they move through their military career (Malone 2013). Our analyses thus allow us to explore whether, given a careful screening process, individuals with criminal records perform similarly to their counterparts with no criminal history. As one of the largest employers in the United States, employing between 7 and 10 percent of men ages 18–24 (Teachman and Tedrow 2016), the military represents an important case study for broader labor market dynamics. Recent research looks to the military workforce as a prototype to analyze the performance of GED holders (Heckman, Humphries, and Kautz 2014, ch. 6) and prior studies note the importance of military service in better understanding employment patterns in the civilian labor market (e.g., Mare and Winship 1984). Given its large and diverse workforce, the military offers opportunities to explore employment dynamics that generalize to the broader labor market. Occupations in the military are a microcosm of the many types of labor found in the civilian economy. At the same time, the military does exhibit key features that distinguish it from typical civilian employ. In particular, the military represents a highly structured environment in which enlistees both live and work. On the one hand, ex-offenders may have less opportunity to perform poorly in such an environment. Studies have shown a “knifing off” effect of military enlistment, where individuals are removed from negative pre-existing peer group influences (Laub and Sampson 2001). To the extent that the military ethos emphasizes discipline, responsibility, and self-efficacy, the environment may also mitigate against poor performance. A recent study using the NLSY 1997 cohort finds that veterans and active duty soldiers who enter the military with previous histories of arrest are less likely than nonveterans to reoffend (Teachman and Tedrow 2016). At the same time, the greater surveillance of military service may expose the recruit to more intense behavioral scrutiny,4 while also exacerbating any pre-existing tendency of resistance to authority. Furthermore, the stress of war deployment and separation from one’s family and community may exacerbate negative impulses. It is thus unclear whether the military setting represents a conservative or liberal test of the performance of individuals with criminal records. In either case, it represents an undeniably central component of the American labor market. Although the waiver program is not widely known to the public nor to scholars, the DoD has a keen interest in analyzing the success of its waiver programs. Waiver programs enable the military to be flexible in response to the changing labor market and recruitment pool, allowing leeway in recruiting individuals who would normally be ineligible to serve in poor recruitment climates. It is a costly investment to recruit and train any service personnel, and so the military has a vested interest in recruiting individuals who will serve out their full contracts. As potentially high-risk hires, the trajectory of individuals with conduct waivers of any kind has been subject to examination by the Department of Defense (DoD). The earliest evidence of ex-offender performance in the military stems from both World Wars, when convicts were drafted to meet manpower needs in the battlefield. In both periods, the performance of ex-felons was found to be no different from those without criminal records (Mattick 1954; Shattuck 1945). While data was not systematically collected in WWI, in WWII the army ran an experiment with the Illinois Penitentiary System, whereby 2,942 ex-felons entered the army. The parole violation rate of army-bound ex-felons was four times lower than the violation rate of the average ex-offender during that period (Mattick 1954). As a result, the army revised its regulations to allow men with criminal records to serve in place of parole (Shattuck 1945). In the 1950s, the army conducted a matched sample of ex-felons with non-felon controls and found that promotions and merit awards were the same; however, ex-felons were more likely to be discharged for misconduct (Mattick 1954). A number of military reports and unpublished theses from military service academies have examined the program in the post-draft era to assess how different waiver programs (drug, health, conduct, dependent, etc.) impact manpower effectiveness, readiness, and troop morale in the military. These internal documents are instrumental in informing our study and the data. Evidence from these studies for the success of conduct waiver recruits is mixed. Most analyze first-term attrition, defined as failure to complete the initial contract period of a recruit. This is a widely used metric in the military to measure the investment payoff of an incoming cohort. An analysis by the US General Accounting Office found no difference between first-term attrition rates and reasons between standard recruits and those with conduct waivers (GAO 1999), and another study from the navy similarly found no difference in first-term disciplinary problems (Noble 2005). But a thesis examining the likelihood of attrition among conduct waivers in the marines from 1997 to 2005 found that felony waivers were 17 percent more likely than non-waivers to separate for unsuitability reasons (Jeppe 2008). Another thesis focusing on sailors joining the navy in two different locations found that those with waivers (conduct and drug combined) were more likely to terminate before the end of their first contract (Huth 2007). Another study found conduct waiver status to be a minor predictor of attrition, but only for the navy and the marines, not for the army or air force (Wenger and Hodari 2004). In an analysis of army recruits from 2000 to 2005, a thesis found that those with waivers for criminal records were more likely than other recruits to terminate early, though significantly less so than non–high school diploma holders (Sahin and Ayhan 2009). While these studies are informative, they do not provide a complete picture on how those with and without criminal waivers stack up against one another throughout the length of service. In a 2007 congressional hearing on the National Defense Authorization Act, references are made to a number of unavailable reports that indicate positive effects for conduct waivers in the army (US Congress 2007). While a small sub-analysis of all conduct waivers showed they were more likely to terminate early, when limited to felony waivers, there were no differences from enlistees without waivers. Furthermore, overall performance was found to be higher than for other types of waivers, such as medical and dependency. There is one published paper, which evaluates both first-term attrition and promotion rates for recruits with any kind of waiver separately for each military branch (Malone 2013). The paper finds small but significantly lower attrition rates for felony and misdemeanor conduct waiver recruits within all military branches. Findings on promotion are more branch specific. In the army and marines, conduct waivers with felony backgrounds are promoted faster. Conduct waivers with misdemeanors also get promoted faster in the army and in the navy. The only negative promotion rates found are those for air force misdemeanor waivers. Malone (2013) takes an important step forward in analyzing this important data. We build on these existing studies with civilian employers in mind. First, we pay attention to specific causes of attrition. Second, we consider poor performance as a time-dependent continuing risk rather than a risk that employers face only for fixed terms, such as first six months or a year, which we believe better approximates common concerns held by civilian employers when hiring those with criminal records. Third, in addition to examining attrition and its causes, we examine pace and level of promotion through the ranks, which provides a longer-term and more holistic perspective on career success. Finally, the majority of existing research was conducted prior to the rapid escalation in use of felony waivers associated with recent large-scale deployments in Iraq and Afghanistan. Relying on more recent data offers the opportunity to observe how individuals with criminal records fare when the standards are relaxed to allow higher rates of participation. In the following analysis, we examine the performance of individuals with felony waivers relative to their non-felony peers in the military. Concerns about ex-offenders center on assumptions around continuing criminal activity and poor work performance. But if employers’ concerns are exaggerated or unwarranted, or if effective screening can adequately select those ready for work, individuals given this second chance may perform similarly to their non-offender counterparts. Sample and Methods We received our data through a Freedom of Information Act (FOIA) request for all new active-duty enlistments who lack a college degree in calendar years 2002 through 2009. The Data/Methods Background Appendix provides more detailed information. The sample size is 1,274,926 individual cases. The unit of analysis is the individual soldier, and we restrict our empirical analysis to those with valid termination and pay records. Although 0.5 percent of our valid cases report multiple separations, we only analyze the first termination on records, due to the lack of re-enlistment information in our data. Our analysis proceeds in two stages. The first set of analyses focuses on job performance indicators that are relevant to the civilian employment settings, and the second stage of our analyses dives deeper into attrition by examining the unique separation reasons that are often not observable in the civilian settings—early terminations due to violation of the Uniform Code of Military Justice (UCMJ) and death.5 In the first stage of analyses, we use a Cox proportional hazard model to examine time-dependent risks of termination due to poor job performance. Early exits from the military are usually seen as negative employment outcomes, known as BCDs (Bad Conduct Discharge). However, discharges also occur for other reasons, including combat-related injury and disability, which may be seen as a neutral or even a positive employment outcome within the context of the military. In order to focus on the kind of separations that are most concerning to civilian employers, we analyze the risk for poor performance termination using the four-digit separation codes. Those who separated within the same month of entry were excluded.6 An innovative aspect of our approach compared to prior studies is that we consider poor performance separation as a time-dependent continuing risk rather than a risk employers face for fixed terms. The fixed term approach is problematic if rates of poor performance separation cluster differently between those with and without felony waivers, because estimates are sensitive to the chosen time period. If the number of felony waiver and non-waiver employees differs across the length of service, estimated risks should be proportional across all time points of observation. Next, we separately analyze the rate of promotion and the likelihood of promotion to sergeant. We use ordinary least squares regression to assess the rate of promotion, using the difference between pay grade at entry and the latest pay grade on record as the outcome measure. In order to account for curvilinear effect of tenure on the rate of promotion, we include each individual’s total years served and its squared term. Military pay grades for enlisted personnel range across nine different levels, and promotions follow centralized or semi-centralized guidelines depending on the class of soldier. To differentiate semi-automatic promotions based on time-in-service from more competitive promotions based on rigorous screening processes and occupation-specific supply and demand of personnel, we also analyze promotion to sergeant (enlisted rank 5) and above. Promotion to this title is considered to be more meritorious than progression in the lower levels, which is a more standardized process for junior enlistees. In order to examine non-time-dependent likelihood of promotion to rank of E5 and above, we use a logistic regression model. We perform robustness checks using Coarsened Exact Matching (CEM) OLS and logistic models, to ensure adequate covariance balance among our waiver and non-felony samples. Additionally, we examine time-dependent risks of termination due to violation of the UCMJ and death based on the four-digit separation codes using two Cox proportional hazard models. By looking at these two unique circumstances that are less likely to occur (or be observed) in civilian work settings, we deepen our understanding of complex issues surrounding the enlistment of individuals with past criminal records. Data and Analyses Table 1 presents basic descriptive information about our sample, comparing the distributions of those with felony waivers to those without. Felony waivers comprise 7,840 cases out of 1,267,086, or about 0.6 percent of the valid sample. Felony conduct waivers are required for felony offense histories that include arrest, conviction, and/or incarceration. We do not know the circumstances behind each felony waiver category code, because DoD did not provide our requested offense description codes. Felony refers to offense types for which the maximum period of confinement under local law is one year or more (similar to the definition of a felony-level offense). For comparison, those entering with waivers for misdemeanor-level offenses comprise a larger percentage, at 3.6 percent.7 Table 1. Descriptive Table of Means for Covariates, by Non Felon and Felon Status Non-felony Felony Total N = 1,274,926 1,267,086 99.39% 7,840 0.61% Job performance indicators Mean SD Mean SD Pay increase from accession 1.6973 (1.34) 2.0230*** (1.32) Poor performance separation 0.1422 0.1538** Promoted to sergeant and above 0.1600 0.2180*** Continuous independent variables Max years served 3.1781 (2.14) 3.5329*** (2.00) Age at accession 20.0550 (2.87) 21.8222*** (3.65) AFQTP 60.8515 (18.59) 61.3722* (18.19) Categorical independent variables Period Recession 0.2469 0.0541*** Surge 0.2499 0.1704*** Prewar 0.1372 0.1611*** Misdemeanor 0.0361 0.0000 Branch Army 0.3860 0.7524*** Air force 0.1830 0.1577*** Marine corps 0.1999 0.0390*** Navy 0.2310 0.0509*** Delayed entry 0.3789 0.2876*** Education Below HS 0.0087 0.0041*** GED equivalent 0.1037 0.1974*** HS graduate 0.8478 0.7168*** Some college 0.0398 0.0816*** Female 0.1637 0.0661*** Citizen at entry 0.9634 0.9795*** Race White 0.7479 0.7879*** Black 0.1465 0.1087*** Asian 0.0255 0.0202** AI/AN 0.0221 0.0139*** NAT HI/OT PI 0.0089 0.0069 2+ Races 0.0118 0.0059*** Missing race 0.0374 0.0566*** Ethnicity Non-Hispanic 0.8051 0.7722*** Hispanic 0.1332 0.1055*** Missing ethnicity 0.0618 0.1223*** Non-felony Felony Total N = 1,274,926 1,267,086 99.39% 7,840 0.61% Job performance indicators Mean SD Mean SD Pay increase from accession 1.6973 (1.34) 2.0230*** (1.32) Poor performance separation 0.1422 0.1538** Promoted to sergeant and above 0.1600 0.2180*** Continuous independent variables Max years served 3.1781 (2.14) 3.5329*** (2.00) Age at accession 20.0550 (2.87) 21.8222*** (3.65) AFQTP 60.8515 (18.59) 61.3722* (18.19) Categorical independent variables Period Recession 0.2469 0.0541*** Surge 0.2499 0.1704*** Prewar 0.1372 0.1611*** Misdemeanor 0.0361 0.0000 Branch Army 0.3860 0.7524*** Air force 0.1830 0.1577*** Marine corps 0.1999 0.0390*** Navy 0.2310 0.0509*** Delayed entry 0.3789 0.2876*** Education Below HS 0.0087 0.0041*** GED equivalent 0.1037 0.1974*** HS graduate 0.8478 0.7168*** Some college 0.0398 0.0816*** Female 0.1637 0.0661*** Citizen at entry 0.9634 0.9795*** Race White 0.7479 0.7879*** Black 0.1465 0.1087*** Asian 0.0255 0.0202** AI/AN 0.0221 0.0139*** NAT HI/OT PI 0.0089 0.0069 2+ Races 0.0118 0.0059*** Missing race 0.0374 0.0566*** Ethnicity Non-Hispanic 0.8051 0.7722*** Hispanic 0.1332 0.1055*** Missing ethnicity 0.0618 0.1223*** *p < 0.05 **p < 0.01 ***p < 0.001 The top three variables shown in table 1 are the dependent variables we use in the main analyses. The first is poor performance termination, which is a dichotomous indicator. Its mean suggests that ex-felons may be slightly more likely than standard recruits to break their contract due to poor work performance, absent controls. The next two dependent variables are promotion indicators: the rate of pay grade increase and the percentage who have been promoted to sergeant. Absent controls, it appears that felony waivers significantly outperform non-felony waivers on both measures. The control variables that follow include demographic and educational backgrounds of enlistees and branch of service and contextual variables based on enlistment years. There are some interesting contrasts worth highlighting. First, we see some variation in the distribution of felony waivers across periods. We separately identify three distinct periods during which the context of military recruitment shifted in important ways: the prewar period (2002), the Iraq War troop surge (2007 and 2008), and the Great Recession (2008 and 2009). Note that these are non-exhaustive and partially overlapping periods; and yet each captures a distinct environment shaping the supply and demand for military service. For example, severe contraction of the civilian economy during the Great Recession may have increased the supply of qualified candidates, leading to increased selectivity and lower recruitment of ex-felons. Recruiting cycles vary inversely with employment cycles, with more civilians interested in the military when employment rates are low (Kleykamp 2006). By contrast, the “surge” generated escalating demand for new recruits, with a doubling in non-felony recruits from the prewar period and a slight increase among those with felonies. Second, we see important differences in the distribution of felony waivers across branches of military service. The army comprises about 39 percent of our sample overall, but 75 percent of enlistees with felony-level waivers are army personnel. The largest and least specialized of the military branches, the army has the least selective recruitment standard criteria in terms of education and test scores. Finally, educational distributions at time of enlistment also differ between those with and without a felony wavier. The reference group is high school graduates compared against high school non-completers, GED equivalents, and those with some college education. Roughly 10 percent of non-felony enlistees have a GED; among our felony waiver sample, this fraction is roughly twice as large (20 percent). Some of this difference may be structural, with some ex-offenders completing their degrees while incarcerated (Heckman and LaFontaine 2007). GEDs are associated with poorer performance relative to those with high school diplomas (Heckman, Humphries, and Kautz 2014), a factor that may contribute to performance differences between those with and without felony waivers. However, ex-felons are not uniformly at the bottom of the educational distribution. In fact, enlistees with felony waivers are twice as likely to have some college education as those without. In addition, the Armed Forces Qualifying Test (AFQT) percentile rankings that are used as a predictor of job performance (bottom of table 1) are higher for ex-felons by half a point. The Delayed Entry Program is also included as a proxy measure for readiness to serve because delaying one’s date of military entry following recruitment tends to increase early attrition. In our sample, fewer felony waivers delayed their entry. Consistent with expectations, ex-felons are more likely to be male (reference category); more likely to be older; and more likely to be US citizens (reference category) at time of enlistment. Other demographics of felony waiver holders do not mirror the prison population, however, such as racial demographics. We code racial categories as white (reference), black/African American, Asian American, American Indian/Native Alaskan, Native Hawaiian/other Pacific islanders, mixed races, and missing/other. At the aggregate level, 4 percent more white enlistees have felony waivers, relative to their population size in the military. African Americans are more likely to have a felony-level criminal record than whites in the general population (Uggen, Shannon, and Manza 2012), and so this may indicate racial bias in the whole-person screening process. It is plausible that what the military considers as unwaivable crimes disproportionately affect the black population. Unfortunately, we lack data for rejected applications to assess whether these patterns match the racial distribution of the applicant pool and/or to what degree race shapes the waiver and selection process. Analyses Predicting Performance Of particular concern to employers within and outside the military is the possibility of poor performance among those with criminal pasts. Termination for this reason is due to overall unsatisfactory work and any disciplinary problems, such as positive testing for drugs and alcohol. The military is arguably an upper bound compared to civilian employment for its sensitivity to misconduct. Given that the ultimate goal of enlistee training is manpower readiness for war, the armed forces take dismissible behaviors like disrespect for authority, lateness for work (“shirking”), lack of motivation, inaptitude, irresponsibility, and so on, very seriously. This is made clear in a Supreme Court decision noting that “the special character of the military requires civilian authorities to accord military commanders some flexibility in dealing with matters that affect internal discipline and morale” (Secretary of the Navy v. Huff 1980). We examine whether individuals with felony waivers are more likely than their non-felony counterparts to be terminated for poor work performance. Figure 1 shows the baseline survival rates of each group over the course of service, indicating that both groups are equally likely to attrite, with approximately 20 percent of all enlistees dismissed for poor performance in the first four years of service. This finding contradicts employer concerns about ex-offenders’ ability to do satisfactory work as employees, indicating that they perform similarly to the average employee. Figure 1. View largeDownload slide Estimated risks of poor-performance separation by month of active service Figure 1. View largeDownload slide Estimated risks of poor-performance separation by month of active service Controlling for all background and service characteristics, our model shows no significant difference between ex-felons and standard recruits in their risks of poor performance separation (full model results are in appendix 1). Our focus is on how felony waivers perform, but it is worth noting that entrants with lower-level misdemeanor waivers face 21 percent higher risk of early termination for poor performance.8 We discuss this finding further below. Predicting Promotions Figure 2 plots the model estimates of promotion rates for those with and without felony waivers (gray and black lines, respectively) with fitted probability values (gray and black circles, respectively), controlling for the rank at which the individual entered the service and total years enlisted, among all the other controls from table 2 (full model is in the second panel of appendix 1). Our estimates suggest that on average those with felony waivers are promoted roughly 0.05 ranks more than non-felons with similar starting points and years of experience. This effect is seen most in the early period of enlistment and toward the end. The fitted values indicate that there is substantially less variation among felony waivers, which are more tightly clustered toward higher numbers of promotions with each year of service. Table 2. Cause of Death by Felony Status Non-felony Felony Death, battle casualty 1,910 (39.02%) 28 (45.16%) Death, non-battle, disease 89 (1.82%) 0 (0.00%) Death, non-battle, other 2,595 (53.01%) 33 (53.23%) Death, cause not specified 301 (6.15%) 1 (1.61%) Total 4,895 (100.00%) 62 (100.00%) Non-felony Felony Death, battle casualty 1,910 (39.02%) 28 (45.16%) Death, non-battle, disease 89 (1.82%) 0 (0.00%) Death, non-battle, other 2,595 (53.01%) 33 (53.23%) Death, cause not specified 301 (6.15%) 1 (1.61%) Total 4,895 (100.00%) 62 (100.00%) Figure 2. View largeDownload slide Predicted number of promotions by years of service Figure 2. View largeDownload slide Predicted number of promotions by years of service The overall magnitude of this promotion effect is not large. Promotion in the lower ranks progresses fairly steadily for the average recruit due to strongly enforced EEO policies in operation in the military. However, promotion to sergeant rank and above is considered to be more meritorious than the more standardized process at the junior ranks. Next we test this promotion outcome, and here we see a more pronounced promotion success for enlistees with felony waivers, who are 32 percent more likely to be promoted to the rank of sergeant than similar enlistees with no felony history (see full model in appendix 1, Panel 3). At this more selective stage, then, we see individuals with serious criminal histories performing better than their counterparts with no criminal record. As a way of calibrating the relative promotion rates of those with felony waivers to other status groups of interest, figure 3 shows a series of predicted probabilities for key groups, with all other values set to their means. Vertical whiskers indicate 95 percent confidence intervals for the predicted probabilities. In particular, we compare felony waivers to those who lack a criminal record but have characteristics that generally predict lower success rates, such as lower measured cognitive test scores (AFQT = 25th percentile) and those entering with only a GED, and those with characteristics that generally predict higher success rates, such as high cognitive test scores (AFQT = 75th percentile). We also show a comparison category in which all characteristics have been set to their mean values. Figure 3 shows that enlistees with a felony waiver do better in their promotions than do average enlistees with no waiver, and notably better than non-felons who are in the 25th AFQT percentile or who have a GED. In fact, their promotion performance is akin to non-felons in the top 75th percentile of cognitive ability (AFQT). Figure 3. View largeDownload slide Probability of being promoted to rank of sergeant or more after 6 years of service Figure 3. View largeDownload slide Probability of being promoted to rank of sergeant or more after 6 years of service The results from our promotion models provide consistent evidence: To the extent that those with felony waivers differ from those without, it is in the direction of more successful advancement through the ranks. Despite widespread concerns about the competence of individuals with serious criminal histories, these analyses suggest that ex-felons screened into military service meet and exceed standards of performance.9 Predicting Risk The first part of this paper examines the work performance of those with felony waivers, pointing to the nontrivial benefits associated with workers with serious criminal records. To the extent that these results generalize to civilian employers, they suggest that appropriately screened ex-felons can represent a desirable source of workers. But employers are not concerned only with the work-related skills or performance of their employees. They are also concerned about the risks associated with ex-offenders; namely, the risks of dangerous or illegal behavior. Such behavior may be of concern both for the immediate risks it poses for the workplace, as well as any possible legal liability employers might face as a result of negligent hiring lawsuits. Some criminological theory predicts that those with histories of serious criminal offending will demonstrate risky or dangerous behavior across a wide range of domains. Gottfredson and Hirschi (1990), for example, theorized that offenders often display a poor internal locus of control, which is associated with impulsivity and risk-seeking, among other negative traits. To the extent that ex-offenders engage in more risk-taking, we may see elevated rates of dangerous and illegal activity. In this section, we consider two kinds of risks: the commission of a legal infraction and the risk of a fatal injury. Each of these outcomes teaches us something about the risk profile of ex-offenders and its possible implications for the employers who hire them. We begin with the risk of legal infraction. This category maps onto employers’ concerns about criminal activity, but here it refers to the military’s distinctive system of court martials, court convictions, and serious offenses. Unfortunately we do not have information on the specific reasons for the legal infraction discharge. This is an important omission, as military law operates under the UCMJ, which has a lesser burden of proof and defines some behaviors as more serious criminal acts than those under standard civilian law (for example, adultery). Indeed, it has been said that “military justice is to justice what military music is to music” (Sherrill 1969), meaning that the system tends to be harsh and extreme. We likewise note that, unlike the civilian context, military personnel are under surveillance at all hours, with little separation between home and work. Some criminological theory suggests that the swiftness and certainty of punishment can promote deterrence (e.g., Hawken and Kleiman 2009), in which case the round-the-clock military-style oversight may be associated with less criminal activity. At the same time, there is a long line of research in criminology pointing to the net-widening effects of increased surveillance, with increasing opportunities for detection often leading to a higher incidence of infractions, independent of actual criminality. For example, “intensive probation” programs, popularized in the 1970s and 1980s, attempted to reduce recidivism by providing more active oversight and support. In many cases, however, those in intensive supervision programs demonstrated higher recidivism rates than those under traditional probation supervision, an effect largely driven by technical violations (e.g., Petersilia and Turner 1993). In such cases, added oversight appears to increase the detection, rather than commission, of misconduct. If this is the case here, our analysis would provide an upper bound on this possible manifestation of ex-felon behavior, as military personnel are at far greater risk of detection than their civilian counterparts. Five percent (N = 62,998) of the non-felony sample is discharged early due to legal infraction, compared to 6.6 percent (n = 519) of the ex-felon sample. In models controlling for the full set of covariates (available upon request), we find that the difference in legal infractions between felony and non-felony recruits is statistically significant. Figure 4 provides some context for the magnitude of this effect by generating predicted Cox regression proportional risks of how education mediates this effect. Predicted probabilities indicate that those with felony waivers are more likely to commit a legal infraction than the standard recruit with a high school degree or some college; however, non-felony recruits with a GED or no high school degree have overall higher predicted risks of legal infraction than does the average recruit with a felony waiver. Relative to common educational credentials (e.g., GED), then, those with felony waivers compare favorably. Nevertheless, the models confirm that individuals with felony records are at elevated risk of termination for legal infractions, even if they otherwise have high job performance.10 Figure 4. View largeDownload slide Predicted proportional risks of legal infraction separation Figure 4. View largeDownload slide Predicted proportional risks of legal infraction separation As a second take on the risk profile of ex-offenders, we examine the risk of fatal injury. Death in the course of service is a nontrivial risk of military enlistment, and often is an unavoidable outcome, particularly during times of war. But, contrary to popular imagination, a minority of deaths in the military occur in the course of combat. The majority of deaths, even in periods of wartime, result from accidents or illness (Defense Casualty Analysis System 2017). To the extent that those with felony records engage in more risky behavior, they may be more prone to fatal injuries. Figure 5 shows the survival likelihood for both groups over the service period. We find that enlistees with felony waivers do in fact have a significantly higher risk of death, which increases the longer they serve (at an average hazard that is 80 percent higher than non-felons). As with the legal infractions findings, the overall numbers of deaths are relatively small. Nevertheless, it is a sobering difference. Figure 5. View largeDownload slide Estimated risks of death Figure 5. View largeDownload slide Estimated risks of death Is this “death disparity” indicative of a higher risk profile on the part of ex-felons? If this were the case, we would expect to see elevated death rates within both combat and non-combat settings. Particularly given the high fraction of accidental deaths, unrelated to combat, we might expect thrill-seekers or risk-takers to be particularly overrepresented among this group. To investigate this question, we analyzed combat and noncombat-related deaths separately in table 2. Although the samples are very small, we find a higher fraction of combat-related deaths for ex-felons compared to non-felons (45 percent vs. 39 percent). Felony waivers are not overrepresented for any other “cause of death” category. Contrary to the prediction of a higher overall risk profile among ex-offenders, these individuals appear to be of elevated risk only in the riskiest of environments. As a final take on elevated death rates among those with felony waivers, we look to the occupational assignments of our sample for some insight into the context in which their work takes place. While we do not have occupational data for the full sample, we are able to identify occupation codes for those in the army, the branch representing the largest fraction of our felony waiver sample (75 percent). Given the small data limitations, we consider this analysis suggestive rather than definitive, and encourage future research in this vein.11 In table 3, we display the 10 occupations with the highest fatality rates, by felony waiver status. We also display the distribution of felony waivers across each occupation. Not surprisingly, infantrymen and cavalry account for the highest fraction of deaths, together accounting for nearly 20 percent of deaths overall. The majority of occupations on this list involve significant combat exposure, including combat engineers, M1 armor crewmen, motor transport operators, and cannon crewmembers. In each of these high-fatality occupations, we see further elevated death rates among those with felony waivers. For example, more than 19 percent of felony deaths occur among infantrymen, relative to less than 8 percent among non-felons. Does this suggest that those with felony waivers are riskier in combat and therefore more likely to lose their lives? The evidence does not support such a conclusion. While those with felony waivers are disproportionately likely to die in combat occupations, they are even more disproportionately likely to be assigned to a high-risk occupation in the first place. Returning to the example of infantrymen (11X), we see two and a half times the proportion of felony deaths in this occupation relative to non-felony deaths. But we also see a much higher fraction of those with felony waivers assigned to this occupation: Nearly 9 percent of the felony waiver sample are assigned to infantry occupations, relative to less than 3 percent of the non-felon sample, a threefold surplus. Across each high-risk occupation, the ratio of felon versus non-felon workers exceeds the ratio of felon versus non-felon deaths within that occupation. While we cannot identify the precise source of this occupational overrepresentation, these findings do suggest that, rather than being riskier overall, those with felony waivers are assigned to more risky contexts.12 Table 3. List of Top 10 Military Occupational Specialties by Number of Deaths Non-felons Ex-felons All Dead All Dead 11X Infantryman 36,152 2.85% 377 7.70% 703 8.96% 12 19.35% 11B Infantryman 29,234 2.31% 287 5.86% 517 6.59% 6 9.68% 101 Unknown 12,622 1.00% 188 3.84% 213 2.72% 4 6.45% 19D Cavalry scout 11,690 0.92% 143 2.92% 226 2.88% 4 6.45% 21B Combat engineer 11,660 0.92% 89 1.82% 220 2.81% 0 0.00% 19K M1 armor crewman 7,393 0.58% 73 1.49% 103 1.31% 1 1.61% 88M Motor transport operator 16,022 1.26% 71 1.45% 339 4.32% 4 6.45% 13B Cannon crewmember 10,798 0.85% 70 1.43% 293 3.74% 1 1.61% 31B Military police 14,558 1.15% 70 1.43% 0 0.00% 0 0.00% 91W Health care specialist 9,426 0.74% 67 1.37% 105 1.34% 2 3.23% Other army MOSs 320,562 25.30% 1,309 26.75% 3,079 39.26% 21 33.88% Non-army MOSs 787,259 62.12% 2,151 43.94% 2,045 26.07% 7 11.29% Total 1,267,376 100.00% 4,895 100.00% 7,843 100.00% 62 100.00% Non-felons Ex-felons All Dead All Dead 11X Infantryman 36,152 2.85% 377 7.70% 703 8.96% 12 19.35% 11B Infantryman 29,234 2.31% 287 5.86% 517 6.59% 6 9.68% 101 Unknown 12,622 1.00% 188 3.84% 213 2.72% 4 6.45% 19D Cavalry scout 11,690 0.92% 143 2.92% 226 2.88% 4 6.45% 21B Combat engineer 11,660 0.92% 89 1.82% 220 2.81% 0 0.00% 19K M1 armor crewman 7,393 0.58% 73 1.49% 103 1.31% 1 1.61% 88M Motor transport operator 16,022 1.26% 71 1.45% 339 4.32% 4 6.45% 13B Cannon crewmember 10,798 0.85% 70 1.43% 293 3.74% 1 1.61% 31B Military police 14,558 1.15% 70 1.43% 0 0.00% 0 0.00% 91W Health care specialist 9,426 0.74% 67 1.37% 105 1.34% 2 3.23% Other army MOSs 320,562 25.30% 1,309 26.75% 3,079 39.26% 21 33.88% Non-army MOSs 787,259 62.12% 2,151 43.94% 2,045 26.07% 7 11.29% Total 1,267,376 100.00% 4,895 100.00% 7,843 100.00% 62 100.00% Unfortunately, we have little information about the process of occupational assignment, which is likely determined by some combination of military need, enlistee’s skills and preferences, and perceived fit between the two. Whether or how an enlistee’s waiver status may be taken into account in this process is largely unknown (to us).13 Furthermore, our small sample of deaths do not allow for a robust analysis of how death rates may vary by time period.14 Future research on the determinants of combat exposure and its consequences would be useful in more fully exploring this relationship. Overall, our risk analysis points to a mixed set of conclusions. On the one hand, those with felony waivers do appear to exhibit modest increased risk when it comes to legal infractions, as least as classified according to the UCMJ. On the other hand, while those with felony waivers do exhibit an increased risk of fatality, this appears to be driven more by their disproportionate assignment to combat, rather than more risky behavior overall. As employers consider the possible trade-offs involved in hiring those with criminal records, these risks should be considered against the more positive findings of performance and promotions discussed earlier. Discussion The military waiver program offers a rare glimpse into the work performance of thousands of individuals with criminal records who are offered positions with one of America’s largest employers. Using a process known as the “whole person” review, the military takes into account the nature of the individual’s offense, compensating skills and experience, and a range of input from personal interviews. In theory, this screening process for felony waivers should result in job performance that is indistinguishable from non-felony recruits. And, indeed, we find that felony waivers are no more likely to be terminated for reasons of poor performance. At the same time, tempering the null findings on work performance separation, we do find that those with felony records have a higher likelihood of being terminated for a legal infraction, although the effect is weaker than, for example, the effect of low education levels in predicting separation. While estimates from the armed forces represent an upper bound on the risks of criminal activity in the civilian workplace—because military personnel are under surveillance day and night, and because military law defines as criminal behaviors that are legal (although certainly not model behaviors) in civilian contexts—these patterns merit further investigation. In future work, we hope to gain access to nuanced data that would allow us to differentiate between legal infractions that might matter less to civilian employers, for example adultery or failure to dress properly,15 and those that matter more, for example workplace theft or workplace assault. We find it particularly intriguing that ex-felons are more likely than the standard recruit to experience promotions during their term of service. What might be the reason for this comparative success? One possibility is that the “whole person” selection process does not just weed out potentially poor-performing ex-offenders, but it also selects individuals who will ultimately perform at higher than average levels. If so, rather than categorically turning applicants with criminal records away, civilian employers may have much to gain in considering a similar screening process. One potential criticism, relevant for the generalizability of these results to the civilian sector, is the issue of selectivity. What if the whole-person screening process in the military is so selective and exclusive that the population of ex-felon waivers does not come close to comparing to the pool of ex-felon applicants for a typical civilian job? This may be; however, let us remember that the size of the ex-felon population accepted into the military is not tiny. Furthermore, the entry characteristics of those with felony waivers do not differ dramatically from those of other recruits, at least on observable characteristics such as educational attainment and cognitive ability.16 We speculate that the success of the “whole person” screening process may derive from qualitative factors, such as self-presentation in face-to-face interviews, careful consideration of the nature and timing of the crime, and recommendations from referrals who can provide evidence of character traits or trajectories not otherwise evident in an applicant’s file. These more subjective components of the screening process are important, and they likewise represent a standard feature of the civilian hiring process. We believe that an additional likely explanation for felony waivers’ high job performance relates to their relatively poor employment prospects overall. The scarcity of stable employment for ex-felons is likely to generate greater commitment to an employer who is willing to take a chance on them. In the case of the military’s felony waiver program, ex-felons may work harder to get promoted and look to military service as a more long-term career because they believe they will not find such opportunities in the broader labor market.17 This is an interesting prospect for employers. Those who adopt this sort of screening process before the practice becomes more widespread could benefit from an early innovator effect. At the same time, the higher combat death rates experienced by ex-felons suggests that they may also be in a position of accepting assignments that put them at greater risk. How occupational assignment is determined for those with felony pasts is a research question in need of pursuit. We have focused our attention in this paper on felony waivers because ex-felons experience the most discrimination on the job market. However, it is important to note that our full analyses do not show high levels of performance for all categories of ex-offenders. Those with misdemeanor-level waivers are more likely to attrite due to poor performance. This may again be a reflection of the whole-person screening process, the criteria for which are more stringent for serious offenders (Powers 2014a). If this is the case, it suggests that the whole-person screening should be equally applied to job applicants with more minor infractions. On the other hand, it may be the case that the severity of the offense is less predictive of employment success than a history of repeat low-level offending, particularly in highly structured environments like the military. Given the available data, we cannot identify the specific factor(s) that may differentiate those former felons with more and less serious felony offenses. What we can safely conclude is that, while not all those with criminal records perform well in this particular employment context, the integration of those with felony-level criminal records—the type of criminal record of most concern to employers—have patterns of attrition that are indistinguishable from those without records. Additional research with more detailed information on type of criminal histories would help answer this question. Conclusion Our research demonstrates that, given a holistic screening process, hiring ex-felons can result in adequate, and even advantageous, worker outcomes. However, while they are promoted faster than the average recruit, we also find that they are modestly more likely to commit a military legal offense than those who lack a felony history. Does this mean, then, that hiring an ex-felon is a high-risk, high-reward premise? We remind the reader that the overall percentages of people who are terminated for legal infractions in the military is very small relative to the numbers who remain to complete their service successfully. We believe the next step to being able to answer this question is first to better understand what the re-offense categories actually are, given differences in the military legal code. Second, we plan to further investigate the screening standards to understand if there are further screening processes that might identify the small number of enlistees who separate early for re-offense. To what extent can these findings generalize to the civilian labor market? As discussed previously, the military has several unique characteristics that distinguish it from typical civilian jobs: It is a highly structured environment in which soldiers both live and work. Hierarchy, authority, and discipline are at a premium in this context, with a high degree of surveillance across all aspects of performance. It is difficult to know whether this setting is a best-case or worst-case scenario for the successful integration of ex-offenders. Highly structured environments can be helpful for those who have made poor decisions in the past. There is arguably an ethos associated with military service that emphasizes toughness and masculine heroism. It is possible that ex-felons may thrive more in these types of settings, in which case future analyses might compare other similar types of occupations, such as construction work, mining, bar bouncing, and so on,18 to more conventional workplace settings. On the other hand, the strict emphasis on hierarchy and discipline is likely to be challenging, particularly for those who have had negative experiences with authority figures. We remain agnostic as to whether the typical civilian employment context would have stronger or weaker results, given similar conditions of entry. A recent ACLU report features emerging research from several single-employer case studies that point to reduced turnover among companies hiring individuals with criminal records. For example, the report notes: “At Total Wine & More, human resources managers found that annual turnover was on average 12.2 percent lower for employees with criminal records. Electronic Recyclers International (ERI) saw a similar outcome: by adopting a program to recruit employees with criminal histories it reduced turnover from 25 percent to just 11 percent” (ACLU 2017, 8). These findings are consistent with our own research. We encourage large private employers to make data available to study more systematically the performance of ex-offenders in the civilian context. At present, the military is the only large-scale employer that accommodates the hire of ex-felons in significant numbers. Moreover, it carefully measures and documents their performance over time. While generalizability to the civilian labor force remains an open question, these data allow us to assess the important question of ex-offender work performance across a wide range of occupations and with multiple dimensions of performance. We hope that future research and data collection will extend this analysis to the few civilian contexts that regularly hire ex-felons to test whether our results are replicated in non-military contexts. The existing research on employers’ assessments of applicants with criminal records suggests that the military’s “whole person” review is far from the norm. Rather, employers appear to use criminal records as an easy screen to weed out applicants at the early stages of review (Holzer 1999; Pager 2007). If our findings extend to the civilian labor market, employers may thus be missing out on a large number of potentially high-quality workers who are disqualified by virtue of their criminal record. Efforts to encourage a more holistic review of these candidates, such as Ban the Box laws19 that delay the introduction of discrediting information, may be helpful in this regard. We also wish to emphasize that the work performance of ex-offenders is both an outcome and a treatment. In the present study, we have examined the work performance of individuals with felony waivers as a key outcome for investigating the integration of ex-offenders into the workplace. But work experience has an impact in and of itself. Evidence suggests that ex-offenders who find employment are significantly less likely to recidivate (OPRE 2012; Uggen 2000). Work experience following a spell of incarceration has also been shown to reduce the stigma of a criminal record (Pager 2007). Similar benefits may attach to ex-offenders with intervening military service. In this sense, the enlistment or employment of ex-offenders may have beneficial effects for the workplace as well as benefits for the longer-term integration and well-being of ex-offenders within their communities. Finally, it is important to note that even if our findings are replicated in the civilian sector, and non-military employers began adopting similar ex-felon hiring programs, there will still be a large number of individuals with criminal records who get screened out of that second-chance opportunity. In addition, periods of peace and robust employment conditions will periodically lessen employers’ incentives to hire ex-felons. Thus, criminal justice reform must develop in tandem with labor market reform. Most importantly, efforts to reduce the number of individuals processed through the criminal justice system from the start will have more impact than any amount of post-incarceration reform. Notes 1 Throughout most of the study period of our sample, the US military was the largest single employer in the continental United States, with roughly 1.3 million active duty personnel in the US military (DMDC 2015). For comparison, Walmart, now America’s largest employer, had 800,000 workers in the United States in 2002 at the beginning of our study period (see annual report, http://s2.q4cdn.com/056532643/files/doc_financials/2002/2002-annual-report-for-walmart-stores-inc_130202939459524585.pdf), which grew to 1.4 million by 2009 (see Hess ). 2 We focus primarily on felony-level criminal offenses, but most conduct-related waivers in the military are granted for lower-order forms of misconduct, such as misdemeanors and traffic violations, self-disclosed admission of drug or alcohol addiction, or testing positive for drugs or alcohol at the recruitment physical exam. 3 DoDI 1304.26. 4 Greater surveillance can serve as a deterrent to continued offending, but previous research also suggests that intensive oversight often leads to a higher violation rate, due to increased detection of minor offenses (e.g., Petersilia and Turner 1993). 5 This is an interesting way in which the “total institution” setting of the military allows us to assess ex-felon performance in detail that would be almost impossible to measure in the civilian labor market. When civilian employees recidivate or simply quit their job, employers lack complete knowledge on why their employee has stopped coming to work. The military maintains a full spectrum of separation data, allowing us to assess the reasons behind all military exits. 6 We measure the survival rate in the smallest time increment available to us, months of service. As such, we must drop those who join in the same period in which they drop out. An examination of the 0.7 percent of the sample who dropped within the first month shows that a higher percentage of those separating were from the non-felon population (just 0.4 percent of ex-felons fall into this category), which is in keeping with the longer-term patterns we document. 7 Our data only report the highest-level offense waiver. Therefore, waivers are mutually exclusive in our data. Any given felony-level waiver may have misdemeanor offenses as well. 8 We tested for interactions with the felony variable and found none to be statistically significant. 9 One weakness of the promotion models is that the demand and supply of specific occupations, particularly those in combat and at higher paygrades, may also influence how quickly and to what level one can be promoted. We do not have occupational data with which to address unobserved heterogeneity of this sort across branches. If enlistees with waivers are more or less likely to be placed in higher-mobility occupations, then this could influence the observed pattern of results. 10 In separate models, analyses show that even the small group of ex-felons committing legal infractions during active duty were still more likely to get promoted than non-felons. 11 Even with “big data,” rare events can still be difficult to analyze reliably. Future research on this topic could benefit from a more in-depth ethnographic account, or could pool additional years of data as they become available, to allow for more detailed quantitative analysis. 12 This finding may also point to relatively harsher and more stressful conditions under which felony recruits are operating, which could likewise contribute to higher rates of legal infractions. 13 According to Powers (2014b), recruiters (who are aware of waiver status) tend to play a larger role in occupational assignment in the army relative to other branches, where assignments fall largely to commanders following basic training. 14 While death rates for all soldiers are higher once the war begins, we find that the higher death rates for ex-felons varies by period in the war years. Among those recruited during the recession, when the supply of potential military recruits was higher, felony deaths were 58 to 95 percent higher. But the death rates of felons recruited during the war surge are more than 300 percent higher. We speculate that when recruitment standards go down due to greater recruiting demand and reduced supply, ex-felons who enlisted during those less selective periods are more likely to be assigned to combat occupations. 15 See 10 U.S.C. §886. 16 One way to address this question would be to gain access to data comparing ex-felon applications to those that gain entrance. This would require an FOIA request targeted at branch-specific recruiting command stations. In addition, an audit design study of specific recruitment stations could shed light on the first stage of the process, showing whether fictitious profiles with and without felony histories are encouraged or discouraged by the recruiter. Anecdotally, one of the authors contacted a Massachusetts army recruitment office to ask what estimated fraction of felony waiver applications were approved on average. The recruiter estimated that roughly 80 percent of all misconduct waivers are approved; however, this estimate also includes misdemeanors and traffic violations. His best guess was that the felony rate alone was quite a bit smaller, but since his station does not receive a high enough volume of ex-felon applications, he was unable to estimate what that might be. This is a question we will continue to investigate. 17 Similar views are expressed by some civilian employers who hire ex-offenders. For example, one employer reported that he “liked hiring people who had just come out of prison because they tend to be more motivated, and are more likely to be hard workers” (Pager 2003, 956–57). 18 Policing and emergency services also have cultural associations with manliness and toughness, but their rigorous background checks will generally weed out ex-felons. 19 See http://www.nelp.org/publication/ban-the-box-fair-chance-hiring-state-and-local-guide/. Appendix. Full Model Results 1. Poor performance early separation (Cox model) 2. Rate of promotion (OLS model) 3. Likelihood of E5 and above (logistic model) (N = 1,266,230) (N = 1,274,926) (N = 1,274,926) Exp. coef. Z Coefficient SE Exp. coef. Z Waiver Felony 0.9426* (−2.03) 0.0515*** (0.0079) 1.3233*** (7.67) Misdemeanor 1.2092*** (16.80) 0.0007 (0.0033) 1.0477** (2.78) Branch (Omitted: Army) Air force 1.0000 (0.00) −0.1346*** (0.0019) 0.4598*** (−76.88) Marines 0.7994*** (−30.85) −0.1790*** (0.0019) 0.8232*** (−19.29) Navy 0.8717*** (−20.44) −0.2598*** (0.0017) 0.7479*** (−31.73) Delayed entry 0.9328*** (−13.76) 0.1616*** (0.0013) 1.2844*** (36.38) Female 1.2581*** (36.79) −0.0196*** (0.0017) 0.8562*** (−15.84) Citizen at entry 1.5876*** (27.90) −0.0320*** (0.0034) 0.9704 (−1.77) Race (Omitted: White) Black/AA 0.9937 (−0.93) −0.0447*** (0.0018) 0.7549*** (−27.38) Asian 0.7054*** (−18.96) −0.0002 (0.0040) 0.9014*** (−5.05) AI/AN 1.0510** (3.05) −0.0118** (0.0043) 0.9568 (−1.89) NAT HI/OT PI 0.6708*** (−12.61) −0.0196** (0.0066) 0.9300 (−1.95) 2+ Races 0.8908*** (−4.77) −0.0095 (0.0057) 0.9335* (−2.28) Missing/NA 0.9201*** (−6.03) 0.0101** (0.0036) 1.0282 (1.67) Ethnicity (Omitted: Non-Hispanic) Hispanic 0.7437*** (−36.57) 0.0097*** (0.0019) 1.0205* (1.96) Missing/NA 1.0218* (2.11) −0.0033 (0.0029) 1.0311* (2.34) Education (Omitted: HS graduates) Below HS 1.6027*** (21.24) −0.0683*** (0.0067) 0.7529*** (−6.17) GED equivalent 1.5519*** (61.91) −0.0514*** (0.0021) 0.7875*** (−18.30) Some college 1.1004*** (7.57) −0.0215*** (0.0033) 0.9163*** (−5.13) AFQTP 0.9932*** (−50.52) 0.0031*** (0.0000) 1.0177*** (92.75) Age at accession 0.9842*** (−17.51) 0.0101*** (0.0002) 1.0362*** (28.08) Accession pay rank − − −0.6538*** (0.0009) 1.5422*** (96.29) Years of service − − 0.8176*** (0.0012) 8.7624*** (103.21) Years of service squared − − −0.0508*** (0.0001) 0.8992*** (−56.46) Early separation − − −0.1161*** (0.0019) 0.6909*** (−33.60) Period Recession 1.0817*** (11.02) −0.2033*** (0.0020) 0.2049*** (−18.19) Surge 0.9143*** (−13.58) 0.1602*** (0.0016) 0.5227*** (−26.82) No-war 1.1104*** (15.82) 0.1030*** (0.0020) 0.9932 (−0.69) Constant − − 0.5863*** (0.0070) − − 1. Poor performance early separation (Cox model) 2. Rate of promotion (OLS model) 3. Likelihood of E5 and above (logistic model) (N = 1,266,230) (N = 1,274,926) (N = 1,274,926) Exp. coef. Z Coefficient SE Exp. coef. Z Waiver Felony 0.9426* (−2.03) 0.0515*** (0.0079) 1.3233*** (7.67) Misdemeanor 1.2092*** (16.80) 0.0007 (0.0033) 1.0477** (2.78) Branch (Omitted: Army) Air force 1.0000 (0.00) −0.1346*** (0.0019) 0.4598*** (−76.88) Marines 0.7994*** (−30.85) −0.1790*** (0.0019) 0.8232*** (−19.29) Navy 0.8717*** (−20.44) −0.2598*** (0.0017) 0.7479*** (−31.73) Delayed entry 0.9328*** (−13.76) 0.1616*** (0.0013) 1.2844*** (36.38) Female 1.2581*** (36.79) −0.0196*** (0.0017) 0.8562*** (−15.84) Citizen at entry 1.5876*** (27.90) −0.0320*** (0.0034) 0.9704 (−1.77) Race (Omitted: White) Black/AA 0.9937 (−0.93) −0.0447*** (0.0018) 0.7549*** (−27.38) Asian 0.7054*** (−18.96) −0.0002 (0.0040) 0.9014*** (−5.05) AI/AN 1.0510** (3.05) −0.0118** (0.0043) 0.9568 (−1.89) NAT HI/OT PI 0.6708*** (−12.61) −0.0196** (0.0066) 0.9300 (−1.95) 2+ Races 0.8908*** (−4.77) −0.0095 (0.0057) 0.9335* (−2.28) Missing/NA 0.9201*** (−6.03) 0.0101** (0.0036) 1.0282 (1.67) Ethnicity (Omitted: Non-Hispanic) Hispanic 0.7437*** (−36.57) 0.0097*** (0.0019) 1.0205* (1.96) Missing/NA 1.0218* (2.11) −0.0033 (0.0029) 1.0311* (2.34) Education (Omitted: HS graduates) Below HS 1.6027*** (21.24) −0.0683*** (0.0067) 0.7529*** (−6.17) GED equivalent 1.5519*** (61.91) −0.0514*** (0.0021) 0.7875*** (−18.30) Some college 1.1004*** (7.57) −0.0215*** (0.0033) 0.9163*** (−5.13) AFQTP 0.9932*** (−50.52) 0.0031*** (0.0000) 1.0177*** (92.75) Age at accession 0.9842*** (−17.51) 0.0101*** (0.0002) 1.0362*** (28.08) Accession pay rank − − −0.6538*** (0.0009) 1.5422*** (96.29) Years of service − − 0.8176*** (0.0012) 8.7624*** (103.21) Years of service squared − − −0.0508*** (0.0001) 0.8992*** (−56.46) Early separation − − −0.1161*** (0.0019) 0.6909*** (−33.60) Period Recession 1.0817*** (11.02) −0.2033*** (0.0020) 0.2049*** (−18.19) Surge 0.9143*** (−13.58) 0.1602*** (0.0016) 0.5227*** (−26.82) No-war 1.1104*** (15.82) 0.1030*** (0.0020) 0.9932 (−0.69) Constant − − 0.5863*** (0.0070) − − *p < 0.05 **p < 0.01 ***p < 0.001 About the Authors Jennifer Lundquist is Professor of Sociology and Associate Dean at the University of Massachusetts–Amherst. Her scholarship analyzes racial preferences in dating and marriage behaviors; the US military as a unique test case to better understand civilian racial and gender disparities in health and family formation behaviors; and tracing the impact on the American welfare state of the US prison and military systems. Her recently published work appears in Social Science Research, the American Journal of Sociology, and American Sociological Review. Devah Pager is Professor of Sociology and Public Policy and the Susan and Kenneth Wallach Radcliffe Professor at Harvard University. She has conducted several field experiments studying the employment barriers facing individuals with criminal records. Her current research focuses on the organizational context of discrimination and the long-term consequences of legal debt. Eiko Strader is Assistant Professor of Public Policy and Women’s, Gender, and Sexuality Studies in the Trachtenberg School of Public Policy and Public Administration at the George Washington University. Much of her work tries to understand how and under what conditions gender becomes relevant in predicting life chances across different levels of geographical location. References ACLU. 2017. Back to Business: How Hiring Formerly Incarcerated Job Seekers Benefits Your Company. A Report by the Trone Private Sector and Education Advisory Council. https://www.aclu.org/sites/default/files/field_document/060917-trone-reportweb_0.pdf. Apel, Robert, and Gary Sweeten. 2010. “ The Impact of Incarceration on Employment During the Transition to Adulthood.” Social Problems 57( 3): 448– 79. Google Scholar CrossRef Search ADS Barnes, Geoffrey C., and Jordan M. Hyatt. 2012. “Classifying Adult Probationers by Forecasting Future Offending.” https://www.ncjrs.gov/pdffiles1/nij/grants/238082.pdf. Blumstein, Alfred, and Kiminori Nakamura. 2009. “ Redemption in the Presence of Widespread Criminal Background Checks.” Criminology 47: 327– 59. Google Scholar CrossRef Search ADS Boucai, Michael. 2007. “ Balancing Your Strengths Against Your Felonies: Considerations for Military Recruitment of Ex-Offenders.” University of Miami Law Review 61( 4): 997– 1032. Bouffard, Leana Allen, and John H. Laub. 2004. “Jail or the Army: Does Military Service Facilitate Desistance from Crime?” In After Crime and Punishment: Pathways to Offender Reintegration , edited by S. Maruna and R. Immarigeon, 129– 51. Portland, OR: Willan Publishing. Bureau of Justice Statistics (BJS). 2001. Trends in State Parole, 1990–2000 . Washington, DC: US Department of Justice. Bureau of Justice Statistics (BJS). 2014. 3 in 4 Former Prisoners in 30 States Arrested Within Five Years of Release. Washington, DC: US Department of Justice. http://www.bjs.gov/content/pub/press/rprts05p0510pr.cfm. Defense Casualty Analysis System. 2017. https://www.dmdc.osd.mil/dcas/pages/report_number_serve.xhtml. Defense Manpower Data Center (DMDC). 2015. Active Duty Military Personnel by Service by Region/Country, September 30, 2015. DRS #54601. Alexandria, VA: Defense Manpower Data Center. Equal Employment Opportunity Commission (EEOC). 2012. Consideration of Arrest and Conviction Records in Employment Decisions Under Title VII of the Civil Rights Act of 1964 (No. 915.002). Washington, DC. Freeman, Richard B. 1992. “On the Economic Analysis of Labor Market Institutions and Institutional Change.” Working Paper 1587, Harvard Institute of Economic Research. General Accounting Office (GAO). 1999. Military Recruiting: New Initiatives Could Improve Criminal History Screening (GAO/NSIAD-99-53). Washington, DC. Gottfredson, M. R., and Hirschi, T. 1990. A general theory of crime . Stanford University Press. Grogger, Jeffrey. 1995. “ The Effect of Arrests on the Employment and Earnings of Young Men.” Quarterly Journal of Economics 110: 51– 72. Google Scholar CrossRef Search ADS Hawken, A., and Kleiman, M. 2009. Managing drug involved probationers with swift and certain sanctions: Evaluating Hawaii's HOPE . Washington, DC: National Institute of Justice. Heckman, James J., John Eric Humphries, and Tim Kautz, eds. 2014. The Myth of Achievement Tests: The GED and the Role of Character in American Life . Chicago: University of Chicago Press. Heckman, James J., and Paul A. LaFontaine. 2007. “The American High School Graduate: Trends and Levels.” NBER Working Paper 13670. http://www.nber.org/papers/w13670.pdf. Hess, Alexander E M 2013. “The 10 Largest Employers in America.” USA Today, August 22. http://www.usatoday.com/story/money/business/2013/08/22/ten-largest-employers/2680249/. Hirshi, Travis, and Michael Gottfredson. 1983. “ Age and the Explanation of Crime.” American Journal of Sociology 89( 3): 552– 84. Google Scholar CrossRef Search ADS Holzer, Harry J. 1999. What Employers Want: Job Prospects for Less-Education Workers . New York: Russell Sage. Holzer, Harry, Steven Raphael, and Michael Stoll. 2004. “ How Willing Are Employers to Hire Ex-Offenders?” Focus (San Francisco, Calif.) 23( 2): 40– 43. Huth, Richard A. 2007. “The Effect of Moral Waivers on the Success of Navy Recruits.” Master’s thesis, Naval Postgraduate School. Monterey, CA. Jeppe, Adam L. 2008. “Determining the Relationship Between Moral Waivers and Marine Corps Unsuitability Attrition.” Master’s thesis, Naval Postgraduate School, Monterey, CA. Kleykamp, M. A. 2006. “ College, Jobs, or the Military? Enlistment during a Time of War.” Social Science Quarterly 87( 2): 272– 90. Google Scholar CrossRef Search ADS Kurlychek, Megan C., Robert Brame, and Shawn D. Bushway. 2006. “ Scarlet Letters and Recidivism: Does an Old Criminal Record Predict Future Offending?” Criminology Public Policy 5: 483– 504. Google Scholar CrossRef Search ADS Laub, John H., and Robert J. Sampson. 2001. “ Understanding Desistance from Crime.” Crime and Justice 28: 1– 69. Google Scholar CrossRef Search ADS Love, Margaret Colgate. 2011. “ Paying Their Debt to Society: Forgiveness, Redemption, and the Uniform Collateral Consequences of Conviction Act.” Howard Law Journal 54( 3): 753– 93. Love, Margaret Colgate, Jenny Roberts, and Cecelia M. Klingele. 2013. Collateral Consequences of a Criminal Conviction: Law, Policy and Practice . Eagan, MN: NACDL/West. Malone, Lauren. 2013. “ Hiring from High-Risk Populations: Lessons from the US Military.” Contemporary Economic Policy 32( 1): 133– 43. Google Scholar CrossRef Search ADS Mare, Robert D., and Christopher Winship. 1984. “ The Paradox of Lessening Racial Inequality and Joblessness Among Black Youth: Enrollment, Enlistment, and Employment, 1964–1981.” American Sociological Review 49( 1): 39– 55. Google Scholar CrossRef Search ADS Mattick, Hans W. 1954. “ Parolees in the Army during World War II.” Federal Probation 24: 49– 55. National Resource Council (NRC). 2014. The Growth of Incarceration in the United States: Exploring Causes and Consequences , edited by J. Travis, B. Western, and S. Redburn. Washington, DC: National Academies Press. Noble, John L. 2005. “Navy Recruits with Moral Waivers.” Paper presented at the Accessions Research Conference, Navy Support Training Center, Great Lakes, IL. Office of Planning, Research, and Evaluation (OPRE). 2012. “More Than a Job: Final Results from the Evaluation of the Center for Employment Opportunities (CEO) Transitional Jobs Program.” Administration for Children and Families, US Department of Health and Human Services. http://www.mdrc.org/sites/default/files/full_451.pdf. Pager, Devah. 2003. “ The Mark of a Criminal Record.” American Journal of Sociology 108( 5): 937– 75. Google Scholar CrossRef Search ADS Pager, Devah. 2007. Marked: Race, Crime and Finding Work in an Era of Mass Incarceration . Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS Petersilia, Joan. 2003. When Prisoners Come Home: Parole and Prisoner Reentry . New York: Oxford University Press. Petersilia, Joan, and Susan Turner. 1993. “ Intensive Probation and Parole.” Crime and Justice 17: 281– 335. Google Scholar CrossRef Search ADS Pettit, Becky, and Christopher Lyons. 2009. “ Incarceration and the Legitimate Labor Market: Examining Age-Graded Effects on Employment and Wages.” Law & Society Review 43( 4): 725– 56. Google Scholar CrossRef Search ADS Powers, Rod. 2014a. “Criminal History Waivers.” http://usmilitary.about.com/od/armyjoin/a/criminal.-u59.htm. Powers, Rod. 2014b. “What the Recruiter Never Told You: Enlistment Process and Job Selection.” http://usmilitary.about.com/cs/joiningup/a/recruiter3.htm. Sabol, William J. 2007. “Local Labor Market Conditions and Post-Prison Employment Experiences of Offenders Released from Ohio State Prisons.” In Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America , edited by Shawn Bushway, Michael A. Stoll, and David F. Weiman, 257– 303. New York: Russell Sage Foundation. Sahin, Fatih, and Serhat Ayhan. 2009. The Next Best Alternative to an Ideal Recruit: Attrition Characteristics of Recruits with Waivers and Low Educational Credentials in the US Army. Naval Postgraduate School. Monterey, CA. Sampson, R. J., and Laub, J. H. 1995. Crime in the making: Pathways and turning points through life . Harvard University Press. Sampson, R., and Laub, J. 1996. Socioeconomic Achievement in the Life Course of Disadvantaged Men: Military Service as a Turning Point, Circa 1940-1965. American Sociological Review , 61( 3), 347– 367. Google Scholar CrossRef Search ADS SEARCH Group. 2005. “Report of the National Task Force on the Commercial Sale of Criminal Justice Information.” Sacramento, CA: SEARCH Group. Secretary of the Navy v. Huff, 444 U.S. 453, 1980. Ante at 360. Shattuck, Edward S. 1945. “ Military Service for Men with Criminal Records.” Federal Probation 9( 1): 12– 14. Sherrill, Robert. 1969. Military Justice Is to Justice as Military Music Is to Music . New York: Harper & Row. Society for Human Resource Management (SHRM). 2012. “Background Checking—The Use of Criminal Background Checks in Hiring Decisions.” http://www.shrm.org/research/surveyfindings/articles/pages/criminalbackgroundcheck.aspx. Teachman, J., and L. Tedrow. 2016. “ Altering the Life Course: Military Service and Contact with the Criminal Justice System.” Social Science Research 60: 74– 87. Google Scholar CrossRef Search ADS PubMed Travis, J., Western, B., and Redburn, F. S. 2014. The growth of incarceration in the United States: Exploring causes and consequences Uggen, Christopher. 2000. “ Work as a Turning Point in the Life Course of Criminals: A Duration Model of Age, Employment and Recidivism.” American Sociological Review 65( 4): 529– 49. Google Scholar CrossRef Search ADS Uggen, Christopher, Sarah Shannon, and Jeff Manza. 2012. State-Level Estimates of Felon Disenfranchisement in the United States, 2010 . Washington, DC: Sentencing Project. US Congress. 2007. Hearing on National Defense Act for Fiscal Year 2008 and Oversight of Previously Authorized Programs. Committee on Armed Services, H.A.S.C. # 110-20. US Government Printing Office, Washington, DC. Visher, Christy, Jennifer Yahner, and Nancy La Vigne. 2010. Life after Prison: Tracking the Experiences of Male Prisoners Returning to Chicago, Cleveland, and Houston. Research Brief. Urban Institute Justice Policy Center, Washington, DC. Wenger, Jennie W., and Apriel K. Hodari. 2004. Predictors of Attrition: Attitudes, Behaviors, and Educational Characteristics (No. CRM-D0010146. A2). Center For Naval Analyses, Resource Analysis Division. Alexandria, VA. Western, Bruce. 2006. Punishment and Inequality in America . New York: Russell Sage Foundation. Western, Bruce, and Becky Pettit. 2005. “ Black-White Wage Inequality, Employment Rates, and Incarceration.” American Journal of Sociology 111: 553– 78. Google Scholar CrossRef Search ADS Author notes This paper was presented at the 2016 Population Association of America meetings and Harvard’s Research for Justice Reform 2016 conference. Our thanks to Alec Ewald, Matt Salganik, Isaac Wisniewski, and four anonymous reviewers from Social Forces. We acknowledge the support of the Project on Race, Class, and Cumulative Adversity at Harvard University, funded by the Ford Foundation and the Hutchins Family Foundation. Authorship is alphabetical and reflects equal contributions. Direct correspondence to Jennifer Lundquist, SBS Dean’s Office, 236 Draper Hall, 40 Campus Center Way, University of Massachusetts, Amherst, MA 10003, USA; e-mail: firstname.lastname@example.org. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US.
Social Forces – Oxford University Press
Published: Mar 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera