Theoretical framing of the imprisonment-extremism nexus

The origins of “inmate society,” including the structure of relationships and cultural values of prisoners, and its associated consequences, has been the subject of theoretical and empirical debate for decades [29]. One perspective, deprivation, is that prison institutions bring about functional adaptations to the loss of autonomy, liberty, safety, and relationships experienced while imprisoned [2, 30]. Another perspective, importation, instead emphasizes that behaviors and cultures within prisons are a reflection of the types of people who become imprisoned [31, 32]. More recently, scholars have turned their attention to how the prison experiences of formerly incarcerated persons influence their transition to the community upon their reentry to society. This is known as the exportation hypothesis [14]. We examine this critical stage (transition from prison to society) in the nexus between imprisonment and extremism because intentions to do harm in the name of a cause are somewhat (though not totally) constrained by the parameters of total institutions. Incarcerated people remain under near-constant surveillance and regulation, and the expression of activist and radical intentions may result in the transfer to a different prison unit or placement in solitary confinement [25, 33]. Yet, it is in the release to the community that the greatest potential for harm can be realized through acts of radicalized behavior. After all, over 90% of people incarcerated eventually leave [34]. We highlight three possible relationships between release from prison and activism/radicalization: (a) the extremism-continuity model, (b) the extremism-contraction model, and (c) the extremism-enhancement model. There are sound reasons to anticipate that these models capture the imprisonment-extremism nexus.

The extremism-continuity model holds that the transition from prison to the community is unrelated to activist and radical beliefs. It has long been argued that the beliefs of prisoners, often referred to as the convict or inmate code, are so strongly engrained that they accompany them upon release into the community. There is considerable research on the “convict code” documenting its pervasive character within prison as well as its overlap with the “code of the street” [7, 35–37]. The persistence of these beliefs as a cultural code that guides behavior even after release from prison is supported by arguments that prison culture is exported to the street [14]. Prison norms, behavior, and associations are translated to street settings, underscoring the role of imprisonment for life after release, including activism and radicalism that originated in prison. From this perspective, the empirical prediction is that leaving prison will not correspond with changes in activism or radicalism, and the imprisonment-extremism nexus may exist, but will not worsen or improve with reentry. Extremist orientations after prison are a mirror image of the extremist tendencies embraced while in prison.

The extremism-contraction model, in turn, shifts the focus from radicalization experiences inside of prison by emphasizing the realities of reentry. This model anticipates that activist and radical beliefs will decline upon leaving prison and returning to the community. Reentry is rife with challenges and demands that may require the expression or investment in extremist beliefs to take a backseat to more pressing needs. Examples of these needs include finding housing, reestablishing formal identity (social security cards, driver’s licenses, other licensing), rebuilding personal relationships with family and friends, and finding employment [13, 38]. These challenges are time-consuming for individuals who have been in prison for even short periods of time, much less sustained periods. The strength of values adopted in prison (e.g. activism and radicalism) is likely to wane without continued reinforcement by peers. Unlike prison, where individuals have idle time to socialize with fellow prisoners (e.g., recreational time, partake in religious activities, and interacting with others in prison work programs), the routine activities of reentry may limit access to anti-social peers, religious zealots, and ideas generative of political and social cynicism. The empirical prediction from the contraction model is that levels of activism and radicalism will decline upon release from prison.

The extremism-enhancement model takes the position that support for activism and radicalism increases upon release from prison. This process occurs because reentry experiences fail to meet the expectations on the part of those released from prison [39, 40]. It is well understood that prisoners are hopeful for a successful reentry [41]. Yet, those expectations often meet the harsh reality of reentry, even during periods of economic expansion and a tight labor market. The pace of societal change, particularly in the areas of technology, mobility, and employment, makes old skills and relationships less useful in finding and succeeding in work [42]. The roadblocks posed by a lack of technical skills may produce disillusionment with life on the outside. Former prisoners often must mend relationships with family and friends, who have progressed without the help of spouses, parents, or partners. The anomic state experienced by many such people is a consequence of the inability to master the changing realities of work, family, housing, community, and emerging technologies simultaneously. A response to this disillusionment may be the return to and/or extension of values and relationships, particularly activism and radicalism, learned inside prison, which presents a familiar and known pattern of values and relationships in an uncertain world. The empirical prediction from the enhancement model is that levels of activism and radicalism will exceed what is observed on the inside.

Prior research on imprisonment and extremism

Evidence of a link between serving time in prison and the development of radicalized beliefs finds its strongest support among criminal justice officials and the media. For example, the U.S. Bureau of Prisons identifies “Security Threat Group” members as individuals who hold radicalized beliefs, particularly beliefs against the United States and state governments. Hamm found evidence of a link between gangs and extremist activities in U.S. prisons [43]. These groups can be traditional prison gangs, or groups that espouse race-supremacy, religious extremism, or some combination of ideologies. In the British context, Liebling and Straub interviewed 150 prisoners and 194 staff members at HMP Whitemore [44]. Staff members expressed fears that prisoners were becoming radicalized and were members of well-organized formal groups that had considerable power inside and outside the prison. In contrast, Jones conducted several case studies examining the role of imprisonment in the social process of becoming a terrorist in the United States, United Kingdom, Australia, the Philippines, Indonesia, and Pakistan [45]. He concluded that the evidence did not exist to support the view that radicalization was a consequence of time spent in prison. The case of Canada is interesting in this regard as Haggerty and Bucerius note that few researchers have been given access to interview prisoners in Canadian prisons [27], yet there are claims of a “threat of Islamist prison radicalization in Canada” [24].

De Vito underscores the importance of the relationship between the “inside” and the “outside” of prison in understanding both radicalization and de-radicalization [46]. This reflects the point made by Atran and colleagues regarding the need for a broader understanding of the social contexts of the development of activism and radicalization in relationship to involvement in terrorism [47]. Clearly, the harsh conditions of imprisonment (deprivation of liberty, regimentation of daily life, and isolation from pro-social activities and peers) play a role in fostering conditions where activism and radicalization may flourish. Despite its significance, there are relatively few studies of activism and radicalization in prison, and many such studies depend on the views of correctional employees rather than prisoners themselves. In large part, this is due to the difficulty in gaining access to prisoners and tracking them upon release [48, 49]. In a comprehensive review of the prevalence of Islam in U.S. prisons, Hamm notes that while there has been a dramatic increase in conversions to Islam among U.S. prisoners, it is hard to determine whether such conversions have led to increased or decreased terror activity, given the interventions to reduce such activity [50].

Perhaps the best evidence to date supporting the imprisonment-extremism nexus comes from the work of LaFree, Jiang, and Porter [21]. They used data from the Profiles of Individual Radicalization in the United States (PIRUS) to examine violent political extremism among 675 adults over the age of 21 who espoused far left, far right, Islamist, or single-issue ideologies. Their data were based on “open-source” records, such as newspaper articles, memoirs, and government documents. They made two sets of comparisons: (a) domestic extremists who spent time prison versus those who did not, and (b) domestic extremists who were radicalized while incarcerated versus those who did not. On the whole, their results provide support for the imprisonment-extremism nexus as they conclude that individuals who spent time in prison and were radicalized in prison were more likely to commit politically-motivated violent acts of terror relative to those who did not after release from prison.

One of the challenges in research on activism and radicalization is the lack of validated measurement tools [51]. An exception is the work of McCauley and Moskalenko [26, 52, 53] who developed and tested separate scales, known collectively as the Activism and Radicalism Intentions Scale (ARIS). The scale properties of these measures have been documented with samples comprised of different ages, nationalities, and social statuses, as well as with prisoners [54]. Moskalenko and McCauley argue that radicalism is both qualitatively and quantitatively different from activism and reflects a deeper commitment to action. Activism and radicalism also have different correlates and measure different concepts, leading the researchers to conclude that radicalism is not “simply” a higher score on the scale than activism and that the transition from activism to radicalization required more than a progression in the intensity of feelings toward a political cause. Certainly, the availability of validated scales provides a foundation for a more robust examination of the imprisonment-extremist nexus. Another major shortcoming of prior work in this area is its dependence on staff views of the issues, rather than measuring attitudes and behavior of individuals who have been released from prison. In addition, the design of earlier studies has been cross-sectional, making it difficult to identify the impact of prison while controlling for other factors, much less continuity and change.

The current study

The purpose of this study is to examine continuity and change in activism and radicalism intentions among a sample of 802 prisoners in Texas interviewed several days before their release and reinterviewed around ten months after returning to the community. Texas is an ideal location to conduct this research, as the state has among the highest rates of incarceration in the United States with roughly 550 prisoners per 100,000 citizens. The Activism and Radicalism Intentions Scale was administered to all members of the sample at the baseline and post-release interviews. Using these data, we address two research questions:

1. Do activism and radicalism intentions change upon release from prison into the community?
2. What are the correlates of change in activism and radicalism intentions after release from prison?

A host of demographic, group membership, attitudinal, and behavioral measures were also collected as part of the interviews and are used to examine the correlates of continuity and change. To the best of our knowledge, this represents the first longitudinal assessment of activism and radicalism intentions upon release from prison.

Data

The data used to examine activism and radicalism during and after prison are from the LoneStar Project, or the Texas Study of Trajectories, Associations, and Reentry [55]. The research design involved interviewing prisoners just prior to their release and re-interviewing them upon their return to the community. Prisoners scheduled for release from the largest release unit in Texas constituted the sampling frame. A weekly list of scheduled releases was provided by the prison executive services office that included indicators relevant to the study purpose. Disproportionate stratified random sampling was used to generate a sample of 802 male respondents, drawn from the population of 15,644 prisoners released during the period of data collection. Sampling fractions differed by official gang classification, where people with non-zero levels of gang affiliation were oversampled by a factor of five. Sampling weights are used and denoted. The findings reported herein are representative of prisoners scheduled for release from prison in Texas during the study period.

Study protocols were approved by the Institutional Review Board at Arizona State University (STUDY00001971). Participation in the research was voluntary. Informed consent was obtained from those who chose to participate in the study. The rate of response, or number of respondents who completed an interview divided by the number of people eligible for enrollment, was 61%. The rate of cooperation, or the number of respondents who completed an interview divided by the number of people contacted, was 94%. No benefits were provided to respondents while incarcerated; however, monetary incentives (Walmart gift cards) were offered to respondents for post-release interviews.

Baseline data collection occurred between April and December 2016 in two prison units. The first was the release unit where our respondents were transferred prior to leaving custody. Prisoners who were designated as general population custody level were interviewed at this unit and constituted 95% of the sample. The remaining 5% of the sample was interviewed at the administrative segregation facility of the second unit, where high-custody level prisoners were housed. All these respondents were released from the aforementioned release unit. Trained interviewers affiliated with universities associated with the LoneStar Project employed computer assisted personal interviewing (CAPI) in both settings. Respondents were interviewed around two days prior to their release from prison. The mean duration of the interview was 110 minutes and item refusal/do not know rates were very low (<1%). A complete description of baseline data collection procedures is provided by Mitchell and colleagues [56].

Follow up data collection began May 2016 and was completed February 2018. Respondents were targeted for interviews one and ten months after release from prison, which constitute the wave 2 and 3 interviews. The geographic distribution of former prisoners across Texas required computer assisted telephone interviewing to conduct follow-up interviews, although we conducted face-to-face interviews across the state if respondents were incarcerated at their interview date. Follow-up interviews were facilitated using an extensive contact card generated from the in-prison interview and a client relationship management database to organize reminders, appointments, calls, and notes. The retention rates for these interviews were 66.3% (N = 532) and 64.1% (N = 514), consistent with large-scale studies of reentry [48, 49]. Transient lifestyles, legal cynicism, and limited support systems make former prisoners a hard-to-reach population, and a complete description of post-release collection procedures is provided by Fahmy and colleagues [48].

We examine change in activism and radicalism intentions between the baseline and wave 3 interviews. Ten months affords enough time between imprisonment and reentry to observe change in extremist orientations if it should exist. It is not possible to distinguish between immediate and distal temporal variation in extremism intentions because the scale was not administered during the wave 2 interview owing to study priorities. Missing data due to item nonresponse and a programming error applied to a small (N = 16) group of respondents led us to employ multiple imputation with chained equation for all analyses involving null hypothesis significance testing. However, all descriptive statistics are reported without imputation.

Our analytic sample consists of 514 respondents who completed baseline and wave 3 interviews. While our attrition rate is on par with prisoner reentry studies, it raises concerns of biased results, as respondents who were retained may differ from those who attrited. We have demonstrated in prior work that age, incarceration length, baseline honesty, and study commitment were positively associated with wave 3 retention, while gang membership, rearrest, and extrinsic motivations for study participation were negatively associated with retention [57]. We supplement those results by comparing retained (N = 514) and attrited (N = 288) respondents across the baseline measures derived from the activism and radicalism intentions scale (including reference group and item measures, described below). Only 2 of the 21 measures differed statistically, roughly what would be expected with a Type 1 error rate of 5%. These differences were observed for the importance of “other” groups (attriters score higher) and selecting religious as the most important group (attriters score lower). None of the 10 ARIS items differed.

Dependent variables

The Activism and Radicalism Intentions Scale (ARIS) was administered to our respondents at the baseline and wave 3 interviews. The first step in the administration of the scale involves asking respondents to rank the importance of salient groups, including country, racial/ethnic, religious, or other group. We asked our respondents if there were “any other groups that you feel close to, such as a political group or organization,” and, if so, the name of the group. None of the respondents identified “family” (beyond Latino gang families) or “political” (beyond occasional mentions of Libertarianism or Presidential candidate/elect Trump) in the open-ended responses. The importance of these groups was rated on a scale ranging from 1 (“not at all important) to 7 (“extremely important”). The language of the questionnaire also noted that responding with a 4 was equivalent to a neutral position.

Respondents were then instructed to identify the group that was most important. The groups identified as the most important served as the referent for 10 questions about activism and radicalism intentions. At the baseline and wave 3 interviews, 15% and 25% of respondents indicated that there were no groups that were “most important” to them, respectively, which prevented us from analyzing continuity/change in the ARIS items (although we could still examine group importance). Nonetheless, about two-thirds of the post-release sample (N = 342) identified groups that were important to them at baseline and wave 3 interviews, which are subject to analysis.

Moskalenko and McCauley differentiate between legal and illegal forms of activism. Four of the items pertain to the former and focus on legal activities such as belonging, donating, volunteering, and protesting with groups. These items combine as the activism intention scale (AIS). Six of the items pertain to the latter and focus on illegal activities such as violence and outgroup retaliation. These items combine as the radicalism intention scale (RIS). The likelihood of engaging in these activities was rated on a scale ranging from 1 (“completely disagree”) to 7 (“completely agree”), and included 4 as a neutral response. A complete listing of the items can be found in the results section. Using a latent variable framework, a pooled cross-sectional confirmatory factor analysis led us to a modified two-factor model, consistent with prior research [26, 54]. The model exhibited acceptable psychometric properties (factor loadings>0.53; RMSEA 90% CI 0.04–0.07; CFI = 0.98; TLI = 0.97; AIS Cronbach’s α = 0.85; RIS α = 0.84). AIS and RIS factor scores were standardized across waves and constitute the dependent variable for the respondents who indicated a group was most important to them at the respective interviews.

Independent variables

We examine the correlates of AIS and RIS after release from prison across a series of demographic, in-prison, and post-release measures, as well as ARIS reference groups (described above). Age is measured in years at the time of the baseline interview derived from dates of birth in prison records. Race and ethnicity included the most prevalent groups in the prison system. Four non-mutually exclusive and dummy coded (0 = no, 1 = yes) measures of race/ethnicity derived from self-reports include Black, Latino, white, and mixed/other. Foreign born is a self-reported indicator of whether the study subject was born in the United States (1 = yes, 0 = no). Father is an indicator of whether the respondent reported he had a child (1 = yes, 0 = no). Education was measured by the self-reported last grade of school completed based on seven response categories: less than sixth grade, junior high, incomplete high school, high school graduate/equivalency, incomplete college or specialized training, 4-year college graduate, and graduate professional training. These categories were rescaled to reflect years of educational attainment (e.g., high school graduate or equivalent = 12 years).

Several baseline (pre-release) measures of in-prison attitudes and experiences are included. Gang affiliation was determined by the prison system (1 = yes, 0 = no) [58]. Low self-control is derived from agreement with 13 items from the Brief Self-Control Scale [59], such as “You do certain things that are bad for you if they are fun.” Responses ranged from 0 “not at all like you” to 4 “very much like you” and averaged across the items, where higher scores equal poorer self-control (α = 0.80). Legal cynicism refers to views toward the law [60], and the construct consists of nine items, such as “People in power use the law to try and control people like you.” Response categories ranged from 0 “strongly disagree” to 3 “strongly agree,” and were averaged (α = 0.72). Social support is based on six items from the Multidimensional Scale of Perceived Social Support [61]. Based on a 4-point scale of agreement, the scale contains three measures of familial support (decision-making, general help, and emotional support) and peer support (e.g., sharing feelings, trust, talking), and the average was taken (α = 0.82). Spirituality is a scale derived from the survey instrument used to evaluate the Serious and Violent Offender Reentry Initiative (SVORI). Six items consisted of a 4-point scale of agreement and were averaged, and capture an internalized set of beliefs in a higher power, such as “You feel guided by your God in the midst of daily activities,” and manifestations of the traditions and formal practice of a faith, such as “You pray or meditate regularly” (α = 0.92). Prison misconduct is a variety score composed of 14 types of self-reported misconduct incidents during the current incarceration spell, ranging from less serious acts, such as “Have you refused to obey an order given by a member of the prison staff?,” to more serious acts, such as “Have you attacked a correctional officer with a weapon?” The item was standardized by years of incarceration. Measures from the Texas Departments of Public Safety (DPS) and Criminal Justice (TDCJ) include an indicator of cognitive ability, IQ, derived from the administration of the Wechsler Adult Intelligence Scale-Revised during prison classification, a count of prior arrests which was top-coded to the 99th percentile, the number of years incarcerated, and a count of the number of prison terms served in state facilities.

We examine whether several important characteristics of successful reentry were related to post-release AIS and RIS, consistent with our theoretical models where positive arrangements should be linked to reductions in extremism, while negative arrangements should be linked to increases in extremism. In a relationship measures self-reports of whether respondents were currently married or in a romantic partnership (1 = yes, 0 = no). Employed is a self-reported indicator of whether respondents supported themselves financially via a “legal job where income is taxed” (1 = yes, 0 = no). Stable housing indicates whether respondents lived in the same residence since release from prison (1 = yes, 0 = no). Lives in a good neighborhood is a measure of whether respondents agree with the statement that “You think your neighborhood is a good place to live” (1 = yes, 0 = no). Two measures of recidivism were included. Rearrested is based on DPS records of whether respondents were arrested for a new offense prior to their wave 3 interview (1 = yes, 0 = no). Reincarcerated is based on TDCJ records of whether respondents reentered state jail or prison facilities prior to the wave 3 interview (1 = yes, 0 = no). The final measure, post-release time, is the number of months between the baseline and wave 3 interviews.

Descriptive statistics of the sample

Table 1 displays the descriptive statistics for the non-ARIS study variables. The mean age of respondents was 41 years; the youngest respondent was 19 years and the oldest was 73 years.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics for the study variables.

https://doi.org/10.1371/journal.pone.0242910.t001

Race and ethnicity was rather evenly distributed across Black (33%), Latino (33%), and white (38%) respondents, while 3% of respondents were born outside of the United States. Two-thirds of respondents were fathers at the time of the survey. Consistent with prior research on incarcerated individuals, the mean number of years of education fell short of completing high school (M = 11.7; SD = 2.0). Ten percent of respondents were designated as maintaining a gang affiliation. Prior to their release, respondents spent an average of five years incarcerated in state prison (SD = 5.9) and the average number of prison terms exceeded two (SD = 1.7).

Shifting to the post-release characteristics of the sample, under half of the respondents (41%) were in a romantic relationship when we interviewed them upon returning to the community. Two-thirds indicated they had a legal job with taxable income and maintain stable housing since release from prison. Under half (46%) claimed that they lived in a good neighborhood. In terms of recidivism, 17% and 8% of respondents had been rearrested or reincarcerated, respectively, prior to the date of the wave 3 interview, according to official records. Our wave 3 interview occurred under 11 months after release prison, although there was a fair amount of variation (SD = 2.6) across respondents.

Intra-individual changes in activism and radicalism intentions

Table 2 presents the findings of a comparison of how respondents rated groups during and after imprisonment. The groups identified in prison became less salient 10 months into the reentry process. Respondents reported that country, racial/ethnic, and religious groups were about one-tenth of a standard deviation less important at the follow-up interview than in prison. This is consistent with the hypothesis that reentry is highly consuming of the lives of formerly incarcerated persons. If respondents are less likely to view these groups as important, they are less likely to endorse activist and radical views in support of them.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Differences in ARIS reference groups pre- and post-release from prison.

https://doi.org/10.1371/journal.pone.0242910.t002

There were also changes in the groups that respondents deemed most important. The most notable finding was the overall increase in former prisoners indicating that “none” of these groups were important to them. Whereas 14% of respondents told us this at the baseline in-prison interview, 25% of them revealed this at the wave 3 interview. There were slight reductions in the endorsement of all groups, although respondents were statistically just as likely to select country, race/ethnicity, and other groups. Therefore, the gain in “no groups” category appeared to come primarily from those who endorsed religious groups while incarcerated, a decline of around 6 percentage points, despite religious group remaining the modal category at the post-release interview. To the extent that religious groups promote activism and radicalization, the declines observed in these measures support the view that religious values fall upon release from prison. These declines may cut “both ways” in that religion may play a role in enhancing such activism and radicalization but may also play a role in mitigating the deleterious effects of such beliefs [17]. These groups constitute the reference group that respondents use when responding to statement on activism and radicalism intentions.

Table 3 presents the results of intra-individual changes in the respective AIS and RIS scales and the items used to generate the scales. The overarching finding is that activism drops upon release from prison whereas radicalism essentially remains the same. All the items in the AIS fell in the post-release interview, by anywhere from 0.21 to 0.30 standard deviations. There was an overall 0.30 standard deviation reduction in the AIS; this supports the hypothesis that reentry weakens activism intentions. In contrast, the RIS reduced by 0.08 standard deviations but the effect size was statistically indistinguishable from zero (p = 0.28). The remaining items could not be distinguished statistically from zero. It is important to note that levels of activism far exceed radicalism; the mean response for the former was a “neutral” position, yet the latter was a “moderately disagree” position. Both constructs, and their associated items, maintained a fair amount of variation around the mean.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Differences in ARIS items and scales pre- and post-release from prison.

https://doi.org/10.1371/journal.pone.0242910.t003

The models predicting intra-individual change in AIS and RIS also included a random coefficient for wave, which was coded 0 at the baseline interview and 1 at the follow-up interview. The random coefficients for wave in non-imputed models (N = 327, N*T = 654) contained standard deviations of 0.97 and 0.78 for the AIS and RIS models, respectively, which suggests that there is substantial variability across respondents in changes in activism and radicalism intention. We now move to the question of what explains those changes.

Bivariable correlates of activism and radicalism intentions

The bivariable analyses are reported in Table 4. We find little evidence in support of demographics as correlates of ARIS. Only age was negatively associated with radicalism levels, a standardized drop of 0.20 units. No other measures in the demographic domain were associated with post-release AIS and RIS scores. In terms of the ARIS reference groups selected by respondents, none were unrelated to post-release AIS scores, while three were related to RIS scores. Those who selected race/ethnicity and other groups as the reference category scored 0.69 and 1.52 standard deviations higher in RIS than their comparison groups, respectively. In contrast, respondents who selected religious groups as their reference category scored 0.27 standard deviations lower in RIS. Reference groups appear to matter more for levels of radicalism than activism.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Bivariable OLS regressions predicting post-release standardized levels of Activism Intention Scale (AIS) and Radicalism Intention Scale (RIS).

https://doi.org/10.1371/journal.pone.0242910.t004

The in-prison domain contained several statistically significant correlates of AIS and RIS. The baseline AIS and RIS scores represented the strongest predictors of the respective post-release outcomes. Although large, the standardized scores of 0.36 (p < .001) and 0.47 (p < .001) offer evidence of regression to the mean. Lower levels of self-control was associated with higher radicalism, but not activism. Legal cynicism was positively related to both AIS (0.16 standard deviations) and RIS (0.25 standard deviations). Respondents who spent longer spells in prison maintained higher levels of activism, where an additional six years in prison (relative to the mean of five years) corresponded with 0.14 standard deviations higher in AIS. Respondents who violated prison rules nearly five times per year maintained RIS scores that were 0.09 standard deviations higher than respondents who violated only one rule per year.

Contrary to our expectations, post-release scores on AIS and RIS levels were unrelated to measures capturing relationships, employment, neighborhood, recidivism, or temporal proximity to prison release.

Multivariable correlates of activism and radicalism intentions

The results presented in Table 5 extend our findings by subjecting the imprisonment-extremist nexus to a multivariable analysis. By controlling for baseline levels of AIS and RIS, that is, extremist orientations in prison, our results are capturing the sources of post-release variation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Multivariable OLS regressions predicting post-release standardized levels in Activism Intention Scale (AIS) and Radicalism Intention Scale (RIS).

https://doi.org/10.1371/journal.pone.0242910.t005

In terms of activism, the results indicate that neither the demographic nor the ARIS reference groups were associated with changes in the AIS upon release from prison. We observed a 0.22 standardized increase associated with higher scores on legal cynicism. Similar to our bivariable results, our post-release measures did a poor job of accounting for variation in changes in activism intentions. Only reincarceration was linked to changes in the outcome, where we observed a 0.53 standard deviation reduction in activism intentions among the minority of our sample who recidivated to prison.

In terms of radicalism, our results identify very few correlates of post-release change in RIS scores. Respondents who indicate race/ethnicity as the most important group scored higher than those indicating country or religion (and statistically indistinguishable from “other category”). Among the in-prison measures, only legal cynicism was positively associated with post-release changes in radicalism intentions—a standardized coefficient of 0.16. None of the post-release measures could explain variation in the changes in radicalism intentions.