Recidivism sentencing premiums and similar criminal justice system policies are based on the theoretical idea that increasing the amount of time a person spends in custody compared to their prior sentence will result in that person experiencing a decrease in their level of offending following community reentry. This idea describes a process of within-individual period-to-period change. Empirically investigating the merit of this idea is crucial given concerns about incarceration and its effectiveness, cost, politicalization, and reported impact on health and social well-being (Berger & Scheidegger, 2022; Zinger, 2022). However, understanding whether incarcerating a person for a longer period of time reduces their future offending has been made difficult in part because research almost exclusively relies on between-group analyses comparing persons who experience incarceration to those who do not. Such analyses are not informative of whether a person who experiences an increase in the length of time they are incarcerated subsequently changes in their level of offending. The current study proposes the use of analyses of period-to-period within-person change as a means of more directly addressing these theoretical questions and policy practices. Threats to reliability that come with selection bias represent another challenge for research on incarceration and reoffending. In this context, selection bias refers to the failure to account for pre-incarceration risk factors (e.g., prior offending, antisocial propensity, substance use, gang involvement) that influence both incarceration and reoffending (Loeffler & Nagin, 2022; Wakefield, 2018; Wermink, Apel, Nieuwbeerta, & Blokland, 2013; Wermink, Nieuwbeerta, Ramakers, de Keijser, & Dirkzwager, 2018). Analyses of period-to-period within-person change address selection bias by accounting for all unobserved time-invariant confounders.

Incarceration is punitive, restricts freedom, and may have negative consequences for life course roles and experiences, including family connectivity, employment, and access to housing (Apel & Sweeten, 2010; Harding, Morenoff, Nguyen, Bushway, & Binswanger, 2019; Kirk & Wakefield, 2018; c.f., Bhuller, Dahl, Løken, & Mogstad, 2020). Arguments in favor of incarceration suggest that these negative consequences are outweighed by incarceration's impact on reducing reoffending. Such justifications have come under scrutiny. Based on their meta-analysis of 116 studies, Petrich, Pratt, Jonson, and Cullen (2021) concluded that the “consensus that custodial sanctions, overall, do not reduce reoffending is universal” (p. 41). It thus would appear that incarceration is expensive, punitive, socially harmful, and fails to achieve its intended purpose of reducing reoffending. However, studies included in Petrich et al.'s (2021) meta-analysis were based on between-group analyses that are not well-suited to addressing selection bias. The meta-analysis also mainly reported on studies that used data from the United States and reflected experiences with mass incarceration. The current study used a first-differenced fixed-effect estimator to examine the relationship between incarceration and reoffending from a different perspective. Specifically, we examine whether year-over-year increases in days spent incarcerated prospectively influenced year-over-year change in convictions among a sample of youth followed into adulthood as part of the Incarcerated Serious and Violent Young Offender Study (ISVYOS) in British Columbia, Canada. Data from the Pathways to Desistance Study were used to examine whether results were consistent in the United States, where incarceration policies and practices differ from Canada (Tonry, 2013a).

1. Differences in the purpose and use of incarceration between Canada and the United States

The use and purpose of incarceration varies across jurisdiction and time (Tonry, 2013a, Tonry, 2013b). Yet, few studies have considered the relationship between incarceration and reoffending outside of the United States (Stam, Wermink, Blokland, & Been, 2023). Compared to the United States, Canada's youth and adult incarceration policies center around the use of shorter sentences that are more individualized, focused on rehabilitation (Tonry, 2013b),¹ and mostly unaffected by political agenda (Neil & Carmichael, 2015). For example, policies created by Canada's Federal government that were designed to expand the use of mandatory minimum sentences minimally impacted judicial decision-making (Doob & Webster, 2016). In the province of British Columbia, which is where the data used in the current study were collected, youth and adult custody facilities are not privatized. Physical and mental health services are delivered by the provincial government. Each custody facility has a mental health professional who coordinates services. All persons receive mental health screening within 24 h of admission (Butler, 2021). Persons in custody have access to programs (e.g., building respectful relationships) that are associated with reductions in reoffending (Gress, 2009). Incarcerated youth have access to high school education that is part of the British Columbia Public School District System (e.g., Fraser Park Secondary School) and have daily programming opportunities for drug and alcohol counseling, mental health support, life skills coaching, and employment readiness (Government of British Columbia, 2024). A survey of 4837 persons involved with British Columbia adult corrections indicated that 81 % were satisfied with service delivery (Gress, 2010). Youth and adult justice system professionals (e.g., probation officers) receive service delivery and supervision training to improve relationships, be responsive to gender-based needs, and reduce criminogenic thinking (Bonta, Bourgon, Rugge, Pedneault, & Lee, 2021; Greiner, 2017).

Historically, Canada's incarceration practices and incarceration rates have differed from those in the United States. These differences magnified during the 1990s and 2010s (Tonry, 2013b). During this period, Canada's correctional philosophy decreased its emphasis on punishment severity in favor of using the risk-need-responsivity (RNR) model to create individualized and evidence-based service delivery plans that aimed to reduce the pains of imprisonment and promote rehabilitation (Bonta & Andrews, 2007; Tonry, 2013a, Tonry, 2022; Webster, Sprott, & Doob, 2019). The RNR model involves a greater level of intervention and supervision for persons viewed as higher risk. The RNR model is used in both youth and adult systems, but legislative changes in 2003 ensured that the RNR model was at the center of Canada's youth justice system, where incarceration is used as a last resort, is designed to emphasize rehabilitation rather than deterrence, and includes detailed community reentry plans (McCuish, Lussier, & Corrado, 2021). In the United States, the 1980s marked the beginning of crime control models of youth justice that focused on responding to perceived out-of-control youth through punishment rather than rehabilitation (Cauffman & Steinberg, 2012). For example, some states mandate youth transfers to adult court. Canada currently prohibits such practices and youth are never incarcerated with adults (McCuish et al., 2021).

Although Canada's youth and adult criminal justice systems differ from those in the United States, because these differences are not fixed over time, it is also important to consider the era in which countries are being compared. Most research on incarceration and reoffending in the United States involved data collected during periods of mass incarceration (Petrich et al., 2021). Situating the era in which these two countries are compared is especially important because the relationship between incarceration and reoffending may vary depending on the era being examined. Specifically, individual-level legal system involvement is more likely in eras with higher incarceration rates (Shen, Bushway, Sorensen, & Smith, 2020). Research should therefore consider whether the relationship between incarceration and reoffending is consistent across different birth cohorts and groups exposed to different justice system policies. Instead, most research on incarceration and reoffending has grappled with how to address selection bias.

2. Generations of research on incarceration and reoffending

Different generations of research on incarceration and reoffending can be organized around their approach to addressing selection bias (Wermink, Apel, Nieuwbeerta and Blokland, 2013, Wermink, Nieuwbeerta, Ramakers, de Keijser and Dirkzwager, 2018). Selection bias threatens the reliability of research on incarceration and reoffending because confounding factors (e.g., age, gender, race, current offense, prior offending) are associated with both the likelihood of incarceration and the likelihood of reoffending (Wakefield, 2018). First-generation research accounted for selection bias by statistically controlling for confounders in regression analyses where incarceration is used to predict recidivism (Gendreau, Cullen, & Goggin, 1999). Second-generation research accounted for selection bias by measuring confounders to generate a propensity score that represents the probability of receiving a custody sentence. Incarcerated persons are then matched with a non-incarcerated person with a similar propensity score (Loughran et al., 2009). Any relationship between incarceration and reoffending is thus assumed to be independent of observed confounders. Summaries of first- and second-generation research are unequivocal that incarceration has a null or criminogenic effect on reoffending (Gendreau et al., 1999; Nagin, Cullen, & Jonson, 2009; Petrich et al., 2021). However, datasets rarely have a sufficient breadth of measures to account for all confounders that create selection bias. Overall, first and second generation methods of accounting for selection bias are considered to produce unreliable estimates of the relationship between incarceration and reoffending (Loeffler & Nagin, 2022).

Third-generation research estimates treatment effects in nonexperimental settings by exploiting exogenous characteristics of justice system policies. Regression discontinuity studies (e.g., Mitchell, Cochran, Mears, & Bales, 2017; Rhodes, Gaes, Kling, & Cutler, 2018; Rose & Shem-Tov, 2021) consider persons who fall just above and just below a sentencing grid's criteria for a custody sentence as similar on relevant confounders. Thus, differences in reoffending between the two groups can be attributed to incarceration. Instrumental variable studies exploit random assignment of cases to judges (e.g., Bhuller et al., 2020; Harding et al., 2019). These studies include individuals who would not be assigned to prison by all judges. Accordingly, those individuals who are assigned to prison differ from those who are not assigned to prison only in terms of the punitiveness of their judge rather than unobserved characteristic of the person that could influence their reoffending. Third-generation research reports mostly null or criminogenic effects of incarceration (Loeffler & Nagin, 2022). However, a few studies, both in the United States but especially outside of the United States, report that incarceration reduces reoffending (Bhuller et al., 2020; Jordan, Krager, & Neal, 2023; Rose & Shem-Tov, 2021; Stam et al., 2023). Differences in findings from first and second generation research may reflect third-generation improvements in accounting for selection bias. However, these improvements came with a cost. To account for confounders, third-generation research involves samples of persons who barely qualified for receiving a custody sentence (i.e., “on the margins” of receiving a custody sentence). By excluding persons involved in more serious offenses and those with longer criminal histories, third-generation studies rely on “implausible comparisons and unrealistic counterfactuals” (Wakefield, 2018, p. 771) that limit generalizability. Analytic strategies that address selection bias without excluding persons who are most likely to experience incarceration can help overcome such concerns.

3. Within-individual analyses as a different methodological approach

Prior generations of research used between-group analytic strategies to compare reoffending rates for those who experienced incarceration and those who did not. Such analyses are incongruent with theories and policies that describe the influence of incarceration on reoffending as a matter of within-individual change. For example, perceptual deterrence describes how a person's perceptions of the costs and benefits of offending change after experiencing an escalation in punishment severity (Stafford & Warr, 1993). Framing describes how the likelihood of reoffending changes when there is a within-person difference between anticipated punishment and actual punishment (Bushway & Owens, 2013). Labeling theory describes a crime-inducing effect of incarceration in which changes in a person's punishment status influences changes in their social ties and identity (Paternoster & Iovanni, 1989). Recidivism sentencing premium policies are based on a relativist perspective in which a person's past punishment influences the determination of punishment severity in the present (Mears & Cochran, 2018). Analyses of person-level period-to-period change align with how these theories and policies describe the relationship between incarceration and reoffending. Such analyses have the added benefit of accounting for selection bias by controlling for all unobserved time-invariant confounders (Allison, 2009). As each person acts as their own control, there is no need to create unrealistic counterfactual groups or restrict samples to persons on the margins of a custody sentence.

Three studies have examined within-person change in incarceration by using a time-demeaned fixed-effect estimator. Two studies using data from the National Longitudinal Survey of Youth 1997 reported that increases in time spent incarcerated were associated with decreases in positive employment outcomes (Apel & Sweeten, 2010) and positive mental health outcomes (Porter & DeMarco, 2019). Using data on incarcerated youth from the Pathways to Desistance Study, Dmitrieva, Monahan, Cauffman, and Steinberg (2012) found that increases in self-reported time spent incarcerated were associated with decreases in psychosocial maturity. We add to this literature by examining the relationship between period-to-period within-person change in incarceration and future period-to-period within-person change in reoffending.

4. Current study

Research on offending trajectories has spanned nearly a century (Glueck & Glueck, 1937). Yet, whether incarceration shapes such trajectories remains poorly understood. This knowledge gap has persisted because of difficulties addressing selection bias (Loeffler & Nagin, 2022; Nagin et al., 2009). We are not proposing that within-individual analyses replace other methods (e.g., regression discontinuity designs). However, within-individual analyses can be useful for (1) addressing selection bias, (2) examining samples beyond those on the margins of receiving a custody sentence, and (3) answering different questions that can help evaluate, refine, and create theories and policies that consider the relationship between incarceration and reoffending from the perspective of within-individual change.

A first-differenced fixed-effect estimator was used to examine the relationship between year-over-year change in the number of days spent incarcerated and future year-over-year change in number of convictions among youth participants from the ISVYOS who were followed through adulthood. The effect of incarceration on reoffending may vary by crime-type (e.g., Loeffler & Nagin, 2022; Stam et al., 2023). For example, persons released from prison may be subject to closer surveillance and monitoring (Liberman, Kirk, & Kim, 2014), which could increase the likelihood of detection for technical offenses (e.g., violating a court order). Accordingly, we examined the influence of changes in incarceration on year-over-year change in violent convictions, technical convictions, and non-technical convictions. To examine the stability of the effect of incarceration, we increased the lag between changes in incarceration and changes in convictions by an additional year. Additionally, the crime-reducing impact of incarceration may be stronger as individuals age and become more adept at considering the consequences of reoffending (Fagan & Piquero, 2007). We therefore examined change from one age-stage to the next (e.g., from adolescence to emerging adulthood) as well as change within specific age-stages (e.g., adolescence as defined by ages 12-17).

Incarceration policies differ across jurisdiction and time period (Tonry, 2013a, Tonry, 2013b). The implications that such differences have for the relationship between incarceration and reoffending are not typically considered (c.f., Shen et al., 2020; Stam et al., 2023). With respect to jurisdiction differences, nearly 80 % of studies on incarceration and reoffending use data from the United States. ISVYOS data reflect Canada's experience using incarceration. Data from the Pathways to Desistance Study (Mulvey, 2013) were used to assess whether ISVYOS findings were similar when applying similar analyses to data on incarcerated youth from the United States. Despite their proximity, Canada and the United States differ regarding how decisions about incarceration are made and how incarceration is experienced (Neil & Carmichael, 2015). Canada did not experience an era of mass incarceration, for-profit prison centers, and routine use of mandatory minimum sentences (Tonry, 2013b). With respect to differences across time period, most research on incarceration and reoffending used samples that were collected during eras of mass incarceration (Petrich et al., 2021). Differences across eras in how incarceration is administered and experienced has implications for reoffending (Shen et al., 2020). Thus, the current study also stratified the ISVYOS sample into cohorts who were born in different eras and who were adjudicated under youth justice systems with different philosophies.

Methodological innovation cannot overcome inadequate data that limit theory-testing. The current study lacked data that could clarify, for example, whether any observed reductions in reoffending were due to specific deterrence or rehabilitation. It is rare for studies to measure individual-level changes that take place during incarceration and after community reentry (c.f., Bhuller et al., 2020). Within-individual analyses do not account for unobserved time-variant factors, but this is true of between-group analyses as well. Although the current study did not measure experiences while incarcerated, given that experiences of incarceration are believed to vary by demographic characteristics (Bucerius & Sandberg, 2022; Cochran, Mears, & Bales, 2014), additional analyses stratified the ISVYOS sample by gender and by ethnicity.

5. Method

5.1. Sample

The Ministry of Children and Family Development is responsible for the care of all incarcerated youth in British Columbia and consented to the ISVYOS' recruitment of youth (n = 1719) from custody centers throughout the province. Data on refusals were not consistently recorded, but for about a 5-year period, only approximately 5 % of youth refused to participate (McCuish et al., 2021 for additional details). The profile of incarcerated youth in Canada likely differs from youth in the United States. Unlike the United States, Canada abolished status offenses, does not incarcerate youth with adults, and does not subject youth to mandatory minimum sentences (Bell, 2014). During the period in which participants were recruited (1998–2011), the youth incarceration rate in British Columbia ranged between approximately 40 to 80 per 100,000. In comparison, the youth incarceration rate in the United States ranged between approximately 196 and 355 per 100,000 (OJJDP, 2019). Compared to incarcerated samples in the United States, the ISVYOS sample is more likely to include a greater proportion of youth who were involved in serious or violent offenses.

5.2. Procedures

ISVYOS Research Assistants typically approached youth to participate in the study within the first week of their incarceration. Some participants were incarcerated prior to their recruitment. These prior periods are accounted for in the measurement of incarceration time in the current study. Youth were informed that their involvement in the study would not affect their stay at the custody center, including the outcome of their sentencing. Interviews occurred in a private room away from other youth and custody staff. To obtain assent, participants were read and given a copy of an information sheet explaining the purpose of the study, how data would be collected, that they could withdraw at any time, and that information would be kept confidential unless they made a direct threat against themselves or someone else. Follow-up data were collected using Corrections Network (CORNET) client management software. CORNET includes youth and adult data for all persons involved in the justice system in the Province of British Columbia. Due to either sealed records or inaccurate record-keeping, CORNET information was available for 1620 of 1719 participants. Our main analyses focused on convictions and days incarcerated between ages 12–25. Over 95 % of the sample were followed until at least age 25.

5.3. Measures

5.3.1. Convictions

During the follow-up period, 142 participants died, and 128 others emigrated outside the province of British Columbia.² Incarceration and convictions after these events were coded as missing. Convictions (see Table 1 for a summary) were measured at each year of age, beginning at age 12 (i.e., the age of criminal responsibility in Canada) and continuing for each year of age during the follow-up period. Conviction type (e.g., violent, non-technical convictions) was coded to examine the influence of incarceration on specific types of reoffending. A challenge in examining the relationship between incarceration and reoffending is that being convicted can result in reincarceration, which in turn limits opportunity to incur additional convictions during the rest of the follow-up period. To account for differences in exposure to the community, as part of our sensitivity analyses, we multiplied the number of convictions during a given year of age by the proportion of that age in which the participant was in the community rather than custody. For example, incurring one conviction at age 25 but spending six months in the community would be equal to two prorated convictions.

5.3.2. Incarceration

Incarceration included both time remanded and time spent in custody while serving a sentence. In British Columbia, the same custody facilities are used for remanded and sentenced youth. However, the adult system has separate facilities for remand and sentenced persons. There are more services available in adult facilities for sentenced persons. The current study was unable to distinguish between amount of time spent remanded versus sentenced. The ISVYOS directly measured each date a participant was admitted to a custody center and each date they were released from a custody center. Days incarcerated were summed for each year of age. This approach accounted for the fact that, in line with the RNR model, the final third of youth custody sentences are typically served in the community with various conditions and supervisory requirements that are deigned to help prevent reoffending (Bell, 2014). This approach avoided limitations with sentencing data that can lead to overestimating the length of time a person spends incarcerated by failing to account for early release. The ISVYOS data did not allow us to measure the number of times a person was incarcerated at each year of age, nor did it allow us to measure the length of time spent in custody for each individual sentence/remand period. The number of days spent incarcerated at each year of age was not necessarily based on a consecutive period. We therefore are not examining the influence of sentencing but rather the impact of time spent incarcerated. This measurement strategy aligns with prior studies (e.g., Loughran et al., 2009) and research on “careers in criminalization” that acknowledges that community reentry is not a one-time event (Western & Harding, 2022, p. 437). CORNET is a provincial database but included information on both federal (two years or more) and provincial (less than two years) custody sentences. For 18 % of the sample who received a federal sentence, if a date of release was not specified, their date of parole eligibility was used as a proxy for date of release.

5.3.3. Demographic characteristics

Gender and ethnicity data were obtained from self-report interviews. The demographic characteristics of the ISVYOS sample resembles the youth custody population in Canada (Malakieh, 2017). The sample includes 1341 male and 372 female participants. Data on gender were missing for six persons. The sample is mainly White (54.7 %). Indigenous persons are overrepresented in the sample (30.1 %) relative to the general population of British Columbia (Statistics Canada, 2013). Remaining participants (15.2 %) from various backgrounds (e.g., Black, East Asian, South-East Asian, Hispanic) were aggregated because base rates were too low to examine each ethnicity individually.

5.3.4. Age-stages

Given the potential age-graded nature of the relationship between incarceration and reoffending, four age-stages were constructed: 12–17; 18–23; 24–29; 30–35. In Canada, persons ages 12–17 are adjudicated under youth justice legislation regardless of the severity of their offense. We created three additional six-year age intervals to facilitate comparisons across age-stages of equal length. We examined the relationship between year-over-year changes in incarceration and convictions within each age-stage. To examine the longer-term impact of incarceration, we summed the total number of days incarcerated and total number of convictions incurred within each age-stage. This facilitated examining whether period-to-period within-person change in days spent incarcerated from one age-stage (e.g., ages 12–17 to the next (e.g., 18–23) was associated with period-to-period within-person change in number of convictions from one age-stage (e.g., 18–23) to the next (e.g., ages 24–29).

5.3.5. Cohort and period effects

Participants were born between 1979 and 1998. Three birth cohorts were created to help account for differences over time in the experience of incarceration (1980–1981, n = 190; 1990–1991, n = 343; 1994–1995, n = 184). These groupings were created to resemble Neil and Sampson's (2021) examination of birth cohort differences in justice system outcomes among participants from the Project Human Development Chicago Neighborhoods. Over time, British Columbia Corrections made concerted efforts to shift both its youth and adult systems to a more rehabilitative focus (Government of British Columbia, 2024; Gress, 2009, Gress, 2010). This shift in focus has the potential to influence experiences and responses to incarceration (e.g., Shen et al., 2020). The declining youth incarceration rate that began in British Columbia in the late 1990s is at least partly related to changes in youth justice legislation. Canada's Young Offenders Act (YOA) was in effect from 1984 until 2003 before being replaced by the Youth Criminal Justice Act (YCJA). We stratified the ISVYOS sample depending on whether they were adjudicated under the YOA (1998–2001; n = 596) or YCJA (2005–2011; n = 1123). In line with RNR principles, the YCJA made a concerted effort to lower incarceration rates by only permitting custody sentences for youth whose index offense met certain criteria of severity. Judges were also required to justify why all other sentencing options were unsuitable for improving public safety. The goal of incarceration under the YCJA was the promotion of public safety through rehabilitation (Bala, Carrington, & Roberts, 2009). Unlike the YOA, the YCJA excluded specific deterrence from its sentencing principles. In sum, youth under the YCJA differ from youth under the YOA, both in terms of being higher risk to reoffend and in terms of incarceration experiences (McCuish et al., 2021), and this may have implications for the relationship between incarceration and reoffending.

5.4. Analytic strategy

We used a fixed-effect estimator to examine the effect of year-over-year within-person change in incarceration on year-over-year within-person change in convictions. By focusing on within-person variation, fixed-effects estimation accounts for between-person differences in unobserved time-invariant factors that create selection bias (Allison, 2009). However, between-person differences remain with respect to the number of days spent incarcerated and number of convictions. Between-person differences in these time-varying factors may be associated with unobserved time-invariant factors. To illustrate, across three waves, Case X spent 14, 28, and 192 days in custody (Mean = 78 days) and Case Y spent 7, 12, and 8 days in custody (Mean = 9 days). Time invariant factors like ethnicity may be related to between-person differences in incarceration length and can confound estimates of the relationship between incarceration and reoffending. Between-person differences in time-varying measures can be removed by detrending the data. Time-demeaning and first-differencing are two approaches to detrending. Both approaches are consistent fixed-effect estimators that produce unbiased estimates and reflect degree and direction of within-person change rather than the magnitude of between-person differences. However, they detrend data using different variations and therefore are prone to give different answers to the same question (Waldfogel, 1997). The detrending method should be chosen based on its appropriateness for the theory or research question being evaluated.

Time-demeaning involves subtracting a person's mean value across all measurement periods from their value at each measurement period. For example, although Case X spent more time incarcerated at Wave 2 compared to Case Y, after time-demeaning, Case X's value at Wave 2 is −50 (28–78), reflecting a decrease in time incarcerated, whereas Case Y's value at Wave 2 is 3 (12–9), reflecting an increase in time incarcerated. Time-demeaning is the more popular approach for removing between-person variation in incarceration (e.g., Apel & Sweeten, 2010) and produces estimates of the relationship between incarceration and reoffending that are not biased by time-invariant confounders. However, these estimates may be theoretically misleading because time-demeaning uses information from the future to calculate whether someone changed in the present. For example, for Case X and Case Y, values at Wave 3 contribute to the calculation of the mean. Thus, the magnitude of change defined at Wave 2 is impacted by values at Wave 3. Time-demeaning may poorly reflect a person's experience of punishment. Case X experienced an increase in time spent incarcerated between Wave 1 and Wave 2, but because of a lengthy period of incarceration at Wave 3, their Wave 2 time-demeaned value reflects a decrease in time incarcerated. From a recidivism premium perspective where repeat offenders are meant to be punished more severely (Roberts, 2008), for Case X, incarceration at Wave 2 represents an escalation in sanction severity. Yet, time-demeaning treats Wave 2 as a de-escalation in sanction severity. From a labeling perspective, incarceration at Wave 2 implies that Case X's “deviant” status was reinforced through an increase in time spent incarcerated when compared to Wave 1. Yet, time-demeaning treats Wave 2 as if a person's label as “deviant” is in remission. People do not have memories of future events that they use to make decisions in the present. Incarceration in the present is not viewed and interpreted through the lens of known future experiences.³ In sum, time-demeaning may influence the misinterpretation of the relationship between change in incarceration and change in offending in ways that impact conclusions about theoretical principles and policy practices.

First-differencing detrends data by subtracting values at time t-1 (e.g., Wave 1) from values at time t (e.g., Wave 2) to examine period-to-period change. For Case X, whereas time-demeaning resulted in a decrease in number of days incarcerated at Wave 2, when using first-differencing, their value at Wave 2 (i.e., 28–14 = 14) reflects an increase in number of days incarcerated. We believe that first-differencing more closely aligns with how people experience change and better captures the variation that motivates policies that increase incarceration to reduce reoffending. First-differencing seems to better capture, for example, the courts' process of applying recidivism sentencing premiums where a person's time incarcerated is increased compared to their prior sentence (Roberts, 2008). First-differencing also captures the concept of labeling in terms of how the courts can reinforce a person's deviant status through increases in the amount of time they are required to spend incarcerated compared to a prior sanction (Paternoster & Iovanni, 1989). By focusing on period-to-period change, first-differencing is useful for avoiding the assumption that long-term patterns over the life-course are inherently predictable (see Paternoster & Bushway, 2009 for a discussion).

Fig. 1 shows that detrending ISVYOS conviction and incarceration data yields different values depending on whether first-differencing or time-demeaning is used. For example, at age 14, using first-differencing, the ISVYOS sample averaged an increase of approximately 20 days in custody compared to age 13. In contrast, at age 14, using time-demeaning, the sample averaged a decrease of approximately 30 days compared to the sample mean. As shown in Table S1 of the Supplemental Materials, the average correlation between a participant's time-demeaned value and first-differenced value at a given year of age is 0.619. Consistent with observations from Waldfogel (1997), the two approaches to detrending reflect different concepts yielding different values.

Our main analysis applied a first-differenced fixed-effect estimator for data spanning ages 12–25. This analysis indicated the contribution of year-over-year change in incarceration (i.e., between time t-2 and time t-1) to year-over-year change in convictions (i.e., between time t-1 and time t). There were no between-participant differences in age because all participants were the same age during each wave. All analyses controlled for linear and quadratic effects of ageing, but for parsimony, these are excluded from Eq. (1).

Con_it-Con_it-1= (β₀+β₁Inc_it-1 + α_i + u_i)-(β₀+β₁Inc_it-2 + α_i + u_i-1)

Con represents the number of convictions for individual i at age t and Inc represents the number of days spent incarcerated for individual i at age t.

The subtraction of each component in the equation differences out all unobserved time-invariant factors (α_i). What remains is the effect of differences in incarceration (β₁Inc_it-1 - β₁Inc_it-2 ) and the error term (u_i- u_i-1). Robust (clustered) standard errors were specified to address serial correlation (Nichols & Schaffer, 2007).⁴ Sensitivity analyses examined whether the relationship between incarceration and reoffending was consistent across types of convictions and age-stages. Our main analyses were replicated after stratifying the sample to account for potential moderators of the relationship between incarceration and reoffending (e.g., gender, ethnicity, birth cohort, exposure to different justice system policies).

6. Results

Between ages 12–25, participants averaged approximately 20 convictions, including 2.5 violent convictions (e.g., assault, robbery, homicide), eight technical convictions (e.g., violating conditions of probation), and 12 non-technical convictions (see Table 1). Participants averaged nearly two years in custody during this period. Although participants were recruited in adolescence, 80 % of the sample spent at least one day incarcerated between ages 18–25. Of the 874 participants with available data between ages 30–35, 35 % spent at least one day incarcerated. Across 21,686 total person-period observations, 629 (2.9 %) represented instances where an ISVYOS participant spent the entire observation period (i.e., year of age) incarcerated. Among participants with complete data between ages 12–25 (n = 1410), 21.2 % (n = 299) spent at least one full year of age incarcerated. Overall, it was common for participants to be incarcerated in adulthood and remain in custody for relatively lengthy periods.

6.1. Within-individual change models

Model 1 of Table 2 shows the results of a first-differenced fixed-effect estimator investigating, between ages 12–25, the effect of year-over-year within-person change in incarceration between time t-2 and time t-1 on year-over-year within-person change in convictions between time t-1 and time t. Each additional day incarcerated from two years ago to one year ago results in a 0.0063 unit decrease in the number of convictions from one year ago to present. Said differently, an increase of one month spent incarcerated from two years ago to one year ago is associated with a 0.19 decrease in number of convictions from one year ago to present. Model 2 increased the lag by an additional year to assess the stability of the impact of changes in incarceration. A year-over-year increase in incarceration between time t-3 and time t-2 was associated with a year-over-year decrease in convictions between time t-1 and time t. The dependent variable in Model 3 reflects the number of convictions incurred per day free in the community during the follow-up period. Consistent with prior models, increases in the amount of time spent incarcerated were associated with decreases in the number of convictions incurred at the next measurement window. Finally, an increase in incarceration between time t-2 and time t-1 was associated with a significant decrease between time t-1 and time t in violent, non-technical, and technical convictions (Models 4–6).

Table 3 presents the effect of changes in incarceration on changes in convictions across different measurement intervals (Model 1), changes between aggregated age-stages (Model 2), and year-over-year changes within specific age-stages (Models 3–6). Consistent with Table 2, for all models, period-to-period within-person increases in the number of days spent incarcerated between time t-2 and time t-1 were associated with significant period-to-period within-person decreases in the number of convictions between time t-1 and time t. Of note, when it came to the examination of the relationship between incarceration and reconvictions within specific age-stages, the lower bound of the 95 % confidence interval of the b coefficient for the analysis between ages 12–17 was higher than the upper bound of the 95 % confidence interval of the b coefficients for the other three age-stages. For ages 12–17, an increase of one month spent incarcerated from two years ago to one year ago is associated with a 0.40 decrease in number of convictions from one year ago to present. For the three age-stages in adulthood, a one month increase in time incarcerated is associated with about a 0.12–0.15 decrease in convictions.

As shown in Table 4, regardless of whether the sample was stratified by gender (Models 1–2) or ethnicity (Models 3–5), between ages 12–25, year-over-year within-person increases in the number of days incarcerated between time t-2 and time t-1 were associated with significant year-over-year within-person decreases in the number of convictions between time t-1 and time t. In Table 5, regardless of when participants were born (Models 1–3) or which youth justice system legislation they were adjudicated under (Models 4–5), increases in time spent incarcerated were associated with, at the next measurement period, significant decreases in convictions. Overlapping confidence intervals implied that the relationship between incarceration and reoffending was not stronger for any particular cohort nor for participants exposed to any particular legislation.

6.2. Supplemental analyses

To evaluate whether findings were similar among incarcerated youth in the Untied States, the same analytic strategy was used to examine self-reported data on incarceration and offending versatility over a seven-year period among boys and girls (n = 1354) from Phoenix and Philadelphia who participated in the Pathways to Desistance Study. In brief (see Supplemental Materials for details), aligning with ISVYOS data, year-over-year within-person increases in days spent incarcerated between time t-2 and time t-1 was significantly (b = −0.001; p = .009) associated with year-over-year within-person decreases in self-reported offending versatility between time t-1 and time t. However, the crime-reducing effect of incarceration was observed for male participants but not female participants and for White participants but not for Black participants nor Hispanic participants.

7. Discussion

Theories and policies typically describe the relationship between incarceration and reoffending in terms of within-individual change. For example, recidivism sentencing premium policies call for a person to spend more time in custody compared to their prior sentence as a means of deterring that person from reoffending (Roberts, 2008). Until this point, research has relied on between-group analyses to interpret whether experiencing incarceration reduces a person's likelihood of reoffending (Petrich et al., 2021). Within-individual methods more directly address theory and policy questions concerning whether a person who experiences an increase in the length of time they are incarcerated subsequently changes in their level of offending. Within-individual analyses also address concerns about selection bias by accounting for all unobserved time-invariant confounders. Although within-individual analyses do not account for unobserved time-variant confounders, this is true of between-group analytic strategies as well. An advantage over between-group strategies like regression discontinuity designs is that within-individual analyses address selection bias without also being forced to focus on samples who barely met a court's criteria for a custody sentence (Loeffler & Nagin, 2022; Wakefield, 2018).

The current study examined the relationship between incarceration and reoffending using data from the ISVYOS on incarcerated youth from British Columbia, Canada (n = 1719). A first-differenced fixed-effect estimator was used to examine whether year-over-year increases in days spent incarcerated prospectively influenced year-over-year change in convictions.⁵ Based on analyses between ages 12–25, year-over-year increases in the number of days spent incarcerated were associated with, at the next measurement window, year-over-year decreases in the number of convictions. An increase of one month spent incarcerated from two years ago to one year ago was associated with a 0.19 decrease in number of convictions from one year ago to present. Changes in age were accounted for, meaning that reductions in convictions were not due to a decrease in risk for reoffending that comes with ageing in adulthood. Results were consistent when looking at different types of convictions, including violent convictions, non-technical convictions, and technical convictions (e.g., violations of court orders). Increases in the number of days incarcerated were associated with decreases in reconvictions even when the lag between incarceration and reconviction was extended by an additional year. This additional lag helped account for delays in court proceedings that could exacerbate the time between offense and conviction. It would be a mistake to interpret these findings as a reason fort Canada's correctional system to maintain the status quo. Issues with Canada's correctional system persist and should not be overlooked (Zinger, 2022). Our findings should not be interpreted as support for expanding the use of incarceration. In fact, our findings may reflect what happens when incarceration is reserved for a small group of youth involved in particularly serious or violent offenses.

Over three quarters of the ISVYOS sample were reincarcerated in adulthood. This facilitated examining the relationship between incarceration and reconvictions at different age-stages. A negative relationship between incarceration and reconvictions was observed when analyses were performed between aggregated age-stages (e.g., the effect changes in incarceration from ages 12–17 to 18–23 on changes in convictions from ages 18–23 to 24–29). This relationship was again observed when analyses were performed within specific age-stages (e.g., ages 12–17 only; ages 30–35 only). However, the lower bound of the 95 % confidence interval of the b coefficient for the analysis between ages 12–17 was higher than the upper bound of the 95 % confidence interval of the b coefficients for ages 18–23, 24–29, and 30–35. Thus, reductions in reconvictions following lengthier stays in custody were especially prominent during adolescence. Compared to its adult system, Canada's youth justice system places more emphasis on rehabilitation than deterrence and youth sentenced to incarceration receive close supervision, monitoring, and services when returning to the community (McCuish et al., 2021). Although speculative, the more purposeful emphasis on rehabilitation in the youth system may explain why incarceration was more strongly related to reductions in reconvictions. Adolescence is also a period in which there is more opportunity for change, including change in response to intervention (Steinberg, 2009). Differences in the magnitude of the incarceration effect may therefore be related to the tendency for youth to have a greater capacity to respond positively to rehabilitative efforts. It should also be pointed out that participants who were not incarcerated during a specific age-stage were dropped from the analysis of that age-stage. Thus, those who remained in the analyses of specific adulthood age-stages were those with the most persistent pattern of offending. Such persons may be less responsive to deterrent and/or rehabilitative efforts. Data limitations precluded direct assessment of this possibility, but our speculation is consistent with Jordan et al.'s (2023) finding that reductions in reoffending following incarceration were weaker for repeat offenders.

7.1. Analysis of potential moderators

The ISVYOS sample was stratified to assess the consistency of findings across different subgroups. For both ISVYOS girls/women and boys/men, year-over-year increases in the number of days spent incarcerated were associated with, at the next measurement window, year-over-year decreases in the number of convictions incurred. Some have reported that incarceration has an especially negative impact on women (e.g., Kruttschnitt, Slotboom, Dirkzwager, & Bijleveld, 2013). However, we found that incarceration reduced reconvictions for girls and women. Our finding aligns with recent research from Alberta, Canada where incarcerated women described their life in custody as an improvement upon their life in the community (Bucerius, Haggerty, & Dunford, 2021; Bucerius & Sandberg, 2022).

Results were also consistent across White, Indigenous, and non-Indigenous racialized persons. Custody facilities in British Columbia provide food, shelter, health care, and programs that are unavailable or difficult to receive in the community (Butler, 2021). We are not suggesting that incarceration is a positive experience for ISVYOS participants. The custody environment in British Columbia may simply be an improvement from severe forms of marginalization experienced in the community. This may be especially true for girls and Indigenous participants who reported experiencing higher rates of crack cocaine and crystal methamphetamine use, family conflict, abuse, and unstable housing (Gushue & McCuish, 2021; McCuish & Corrado, 2018). Our interpretation aligns with Gray and Viljoen's (2023) finding that higher-risk youth from a residential treatment facility experienced a greater number of days free of self-injurious behavior compared to lower-risk youth from the same facility. Gray and Viljoen suggested that higher-risk youth spent a greater amount of time under treatment and supervision at the facility compared to lower-risk youth, which may have resulted in a greater number of days free of self-injury following release. Incarceration may fail to reduce reoffending when marginalization in custody reflects the level of marginalization experienced in the community. When performing similar analyses using data on participants from the Pathways to Desistance Study (see Supplemental Materials), increases in time spent incarcerated were associated with decreases in self-reported offending versatility. However, after stratifying by demographic characteristics, the crime-reducing impact of incarceration was observed for White participants but not Black and Hispanic participants and for male participants but not for female participants. It may be that prison conditions in Arizona and Pennsylvania are less conducive to reducing reoffending. Indeed, the United States has a history of incarcerating youth for minor offenses as a means of controlling behavior (Feld, 2009).

Prior research showed that the relationship between incarceration and reoffending might be influenced by the level of crime during the era in which a person comes of age (Shen et al., 2020). ISVYOS participants born earlier came of age during periods with higher incarceration rates. Accordingly, the ISVYOS sample was stratified into three birth cohorts (1980–81; 1990–91; 1994–95) that resembled birth cohort groupings examined in other studies (Neil & Sampson, 2021). Regardless of when ISVYOS participants were born, year-over-year increases in the number of days spent incarcerated were associated with, at the next measurement window, year-over-year decreases in the number of convictions incurred. ISVYOS participants experienced either Canada's YOA (1984–2003) or YCJA (2003-present) legislation. Moving from the YOA to the YCJA was associated with a decline in incarceration rates and changes to the purpose and experience of incarceration, including an increased emphasis on rehabilitation and closer alignment with RNR principles (Bala et al., 2009). However, year-over-year increases in the amount of time spent incarcerated were associated with year-over-year decreases in the number of convictions incurred regardless of which legislation participants were adjudicated under. The consistency of findings across birth cohort membership and legislative changes may be evidence that Canada is in a “settled period” (Swidler, 1986) where it is difficult to discern an independent influence of cultural factors on individual behavior. Policy changes in Canada occur more slowly, are less dramatic, and are less impactful (Doob & Webster, 2016; Roberts, 2012) compared to policy changes in the United States that allowed for mass incarceration (Tonry, 2013b). The YCJA may not have had a unique impact on the relationship between incarceration and reoffending compared to the YOA because incarceration rates in British Columbia were declining even prior to the enactment of the YCJA (McCuish et al., 2021).

7.2. Study implications, limitations, and future research

In the absence of measures before, during, and after incarceration, studies on the relationship between incarceration and reoffending are more suited to ruling out particular theories than supporting particular theories. For example, our results contradict the notion that prisons serve as schools of crime (Clemmer, 1950). However, like other studies, we lacked data on participant perceptions of punishment and measures of the incarceration experience. Therefore, it would be a mistake to interpret findings as direct support for specific deterrence. ISVYOS participants experiencing a period-to-period increase in number of days spent incarcerated nevertheless may have viewed this situation as less punitive than other options For example, some people prefer to be in custody than to be bound by conditions of a community-based sentence (e.g., probation) for a longer period of time (Cochran et al., 2014; Raaijmakers, de Keijser, Nieuwbeerta, & Dirkzwager, 2017). To examine specific deterrence mechanisms, future research should measure subjective perceptions of incarceration. Instead of reflecting deterrence, it is possible that lengthier periods of incarceration provided greater opportunities for ISVYOS participants to receive services that subsequently influenced decreases in reconvictions. Bhuller et al. (2020) reported that employment training programs influenced lower rates of recidivism among incarcerated persons. Although we did not directly measure incarceration experiences, findings were consistent across groups (e.g., different genders) who typically experience incarceration differently (Bucerius et al., 2021; Bucerius & Sandberg, 2022; Kruttschnitt et al., 2013).

Future research should consider Hickert, Bushway, Nieuwbeerta, and Dirkzwager's (2021) discussion of two stages of incarceration: changes that occur during incarceration and changes in offending post-incarceration. Changes in the amount of time spent incarcerated may have implications for the types of individual-level changes that occur during incarceration. Understanding incarceration as a two-stage turning point requires measures of incarceration experiences such as exposure to programming, visits from supportive contacts, perceptions of incarceration and criminogenic attitudes, and changes in substance use. Measures of post-incarceration experiences are also critical because, especially in jurisdictions where the RNR model is used, those who are viewed as higher risk are allocated greater intervention and supervision resources. Reflecting on findings from the current study, increases in time spent incarcerated from one period to the next may be a proxy for increases in the amount of supervision, monitoring, and rehabilitative support received following community reentry. Reductions in reoffending following increases in time incarcerated may therefore actually reflect a post-incarceration effect in which participants received increases in community supervision, monitoring, and services. Canada's youth system in particular reflects RNR principles, which may explain why coefficients were larger when examining incarceration's influence in adolescence compared to in adulthood. We are not aware of other studies that disentangled the supervision effect from the incarceration effect. As weak evidence against the presence of a supervision effect, our findings were consistent even after increasing the lag between incarceration and reoffending. In other words, reconvictions were lower even when looking beyond initial periods of community reentry where services and supervision and monitoring are more likely.

Based on their review of 116 studies, Petrich et al. (2021) suggested that there is universal consensus that incarceration fails to reduce reoffending. Our findings do not necessarily contradict this literature. By focusing on within-person year-over-year change, we answered a different research question. We also used data on sample of participants who are likely higher-risk when compared to samples in instrumental variable studies that focus on persons on the margins of receiving a custody sentence. Although not measured directly, the experience of incarceration for ISVYOS participants in Canada may be different from experiences in the United States, which is where most of the literature on incarceration and reoffending is based. Canada is more likely to emphasize incarceration as an opportunity for rehabilitation and service delivery rather than deterrence and punishment (Tonry, 2013b). Compared to persons on the margins of receiving a custody sentence in the United States, the ISVYOS sample may be less likely to experience the pains of imprisonment because their experiences in the community are already especially challenging (McCuish et al., 2021; also see Bucerius et al., 2021). Especially since the 1990s, Canadian correctional psychology researchers working in policy positions voiced concerns about punitive components of incarceration (Gendreau et al., 1999), promoted effective correctional practices (Gendreau & Ross, 1983), and developed the RNR model to help identify who should be incarcerated and how to address treatment needs (Bonta & Andrews, 2007). These ideas may hold less weight in the United States, where there is a tendency to hold pessimistic attitudes about whether higher-risk offenders can change (Lussier, McCuish, & Cale, 2021). Such attitudes could subsequently influence the withholding of rehabilitative services. For example, Jordan et al. (2023) found that among persons incarcerated in Illinois, incarceration reduced offending for first-time offenders but not repeat offenders. Future research should examine whether findings generalize to persons with psychopathy traits who may be less likely to engage in and respond positively to treatment programming (Sewall & Olver, 2019; Vasaturo, Krstic, & Knight, 2024).

Like other studies (e.g., Loughran et al., 2009), the ISVYOS measured the number of days spent incarcerated within a particular wave. Two participants may have spent an equal number of days incarcerated at a given year of age but one may have entered and exited custody on five separate occasions whereas the other entered and exited custody only once after receiving a single sentence. ISVYOS data were unable to distinguish between such patterns and so it was not possible to examine the influence of sentence length on reoffending. We also did not examine the functional form of the relationship between incarceration and reoffending. Lengthy stays in custody may have diminishing returns on reducing reoffending (Mears, Cochran, Bales, & Bhati, 2016), especially if such stays result in returning to the community at older ages where the risk of offending is lower.

The use of incarceration reflects a justice system's functioning, and the values, culture, rules, and traditions of this system differ across jurisdictions (Roberts, 2012; Tonry, 2013b). We are not suggesting that incarceration decreases the risk of reoffending irrespective of the jurisdiction, era, and context in which it is used. The ISVYOS sample reflects the types of individuals who engage in repeat offending and who probation and parole officers have regular contact with as part of case management practices. Despite the specificity of the ISVYOS sample, as illustrated by our supplemental analyses using data from the Pathways to Desistance Study, the within-individual analytic strategy used in the current study can be extended to a range of other samples and jurisdictions.

Summary

Policies in the criminal justice system, such as those that increase prison time for repeat offenses, are based on the idea that longer incarceration will reduce future criminal behavior. It is important to study this idea carefully because there are concerns about how effective and costly incarceration is, its political aspects, and its effects on health and well-being. However, most research has compared groups of people who were incarcerated to those who were not. This type of research does not show whether a person’s own criminal behavior changes after they spend more time in prison. This study uses a new approach to examine how changes in a person's incarceration time affect their future offending.

A major challenge in studying incarceration and reoffending is selection bias. This means that factors existing before incarceration, like past crimes, drug use, or gang involvement, can influence both who gets incarcerated and who reoffends. These factors are not always considered in research. The new approach used in this study helps to address selection bias by looking at changes within each person over time, which accounts for all unchanging factors that could affect the results.

Incarceration takes away freedom and can negatively impact a person's life, including family connections, jobs, and housing. Supporters of incarceration believe these negative effects are justified by its ability to reduce reoffending. However, many studies suggest that incarceration generally does not reduce reoffending and can be expensive and harmful. Most of these studies used group comparisons, which may not fully address selection bias. This study uses a specific statistical method to look at whether year-to-year increases in time spent incarcerated affected year-to-year changes in convictions among young people in British Columbia, Canada, who were followed into adulthood. The study also checked if similar results were found using data from the United States, where incarceration policies are different.

Differences in the Purpose and Use of Incarceration Between Canada and the United States

The reasons for and methods of incarceration vary by location and over time. Most studies on incarceration and reoffending have focused on the United States. In contrast, Canada’s policies for youth and adult incarceration emphasize shorter, more personalized sentences focused on rehabilitation, and are less influenced by political agendas. For example, mandatory minimum sentencing laws in Canada have had little impact on judges' decisions. In British Columbia, where the study data were collected, prisons are not privately owned, and the government provides health services. Mental health professionals are available in every facility, and all individuals are screened for mental health within 24 hours of admission. Incarcerated individuals have access to programs and high school education that aim to reduce reoffending. Justice system professionals in British Columbia also receive training to improve relationships and address specific needs.

Historically, Canada's incarceration practices and rates have differed from those in the United States, especially from the 1990s to the 2010s. During this time, Canada focused less on severe punishment and more on rehabilitation, using a model that creates individualized, evidence-based plans to reduce the negative effects of imprisonment and promote recovery. This model involves more support and supervision for individuals considered higher risk. In the United States, starting in the 1980s, youth justice shifted towards punishment rather than rehabilitation, with some states even allowing youth to be tried in adult courts, a practice forbidden in Canada.

It is important to consider the specific time period when comparing incarceration practices between countries, as differences are not always constant. Most research in the United States on incarceration and reoffending uses data from periods of widespread incarceration. The impact of incarceration on reoffending can change depending on the era. Therefore, research should consider whether the relationship between incarceration and reoffending is consistent across different generations and groups affected by different justice policies.

Generations of Research on Incarceration and Reoffending

Research on incarceration and reoffending can be categorized by how it addresses selection bias. Selection bias occurs when factors like age, gender, race, or past offenses influence both the likelihood of incarceration and reoffending. Early research tried to control for these factors using statistical methods, or by matching incarcerated individuals with non-incarcerated individuals who had similar risk factors. These early studies generally concluded that incarceration either had no effect or actually increased reoffending. However, these methods were often limited by incomplete data and are now considered unreliable for measuring the true relationship between incarceration and reoffending.

Later research, known as "third-generation," tried to improve accuracy by using specific characteristics of justice policies to estimate the effects of incarceration. For example, some studies compared individuals who barely met or missed the criteria for a prison sentence, assuming they were otherwise similar. Other studies used random assignments of cases to judges to see how judge leniency affected outcomes. These studies largely reported that incarceration had no effect or increased reoffending, though some studies, especially outside the United States, found it reduced reoffending. While these methods improved the handling of selection bias, they often focused on individuals who were "on the margins" of receiving a prison sentence, excluding those with more serious offenses or longer criminal histories. This limited how broadly the findings could apply. New analytical methods are needed to address selection bias without excluding these important groups.

Within-Individual Analyses as a Different Methodological Approach

Previous research used methods that compared different groups of people. This approach does not align with theories and policies that view incarceration's impact as a change within an individual. For example, theories suggest that a person's view of consequences changes after a harsher punishment, or that their social ties and identity shift after incarceration. Policies that increase sentencing for repeat offenses assume that past punishment influences current decisions. Analyzing changes within a person over time fits better with these ideas. This method also helps to address selection bias by controlling for all unchanging factors that might influence outcomes. Because each person serves as their own comparison, there is no need to create artificial comparison groups or limit the study to people who barely qualified for incarceration.

Three studies have already used within-person change to examine incarceration. Two found that more time incarcerated was linked to fewer positive outcomes in employment and mental health. Another study found that increased self-reported time in prison was linked to lower psychosocial maturity. This current study adds to this research by looking at how changes in incarceration time within a person over a period relate to future changes in their reoffending within the same person over a period.

Current Study

The question of how incarceration affects a person's criminal behavior over time is still not well understood, largely due to challenges in addressing selection bias. This study does not suggest that within-individual analyses should replace other methods. However, these analyses are useful for (1) addressing selection bias, (2) studying a broader range of individuals beyond those who barely qualified for incarceration, and (3) answering different questions that can help improve theories and policies about incarceration and reoffending from the perspective of within-individual change.

The study used a specific statistical method to examine how year-over-year changes in days spent incarcerated affected future year-over-year changes in convictions among young people in British Columbia, Canada, who were followed into adulthood. The researchers considered whether the effect of incarceration varied by crime type (e.g., violent, technical, non-technical offenses). They also looked at whether the impact of incarceration was stable over time by increasing the time gap between incarceration and convictions. Additionally, they considered if the crime-reducing effect of incarceration might be stronger as individuals age. They examined changes from one age group to the next (e.g., adolescence to early adulthood) and changes within specific age groups.

Incarceration policies differ across locations and time. Most research on incarceration and reoffending uses data from the United States. This study uses data from Canada, which has different approaches to incarceration. To see if the findings were consistent, similar analyses were applied to data from incarcerated youth in the United States. Canada and the United States have different ways of deciding who is incarcerated and what the experience of incarceration is like. Canada did not experience a period of widespread incarceration, private prisons, or routine mandatory minimum sentences. Most research on incarceration and reoffending uses data from eras of widespread incarceration. Differences in how incarceration is administered and experienced over time can affect reoffending. Therefore, this study also divided the Canadian sample into groups based on when they were born and which youth justice policies they were under, as these might influence the relationship between incarceration and reoffending.

Even with new methods, inadequate data can limit understanding. This study did not have data to determine if observed reductions in reoffending were due to specific deterrence (punishment discouraging future crime) or rehabilitation. It is uncommon for studies to measure individual changes that occur during and after incarceration. While within-individual analyses do not account for unobserved factors that change over time, this is also true for other types of analyses. Since the experience of incarceration is believed to vary by demographics, additional analyses divided the Canadian sample by gender and ethnicity.

Method

Sample

The Ministry of Children and Family Development in British Columbia allowed the study to recruit 1,719 young people from custody centers across the province. Very few young people refused to participate. The profile of incarcerated youth in Canada likely differs from those in the United States because Canada has different laws (e.g., no status offenses, no incarceration of youth with adults, no mandatory minimum sentences for youth). During the study's recruitment period (1998-2011), British Columbia's youth incarceration rate was significantly lower than that in the United States. The Canadian sample was more likely to include young people involved in serious or violent offenses compared to incarcerated samples in the United States.

Procedures

Research assistants typically approached youth to participate in the study within the first week of their incarceration. Any time spent incarcerated before recruitment was included in the study's measurements. Youth were assured that participating would not affect their custody stay or sentencing outcome. Interviews were conducted in private. Participants received information about the study and were told they could withdraw at any time and that their information would be kept confidential unless they threatened themselves or others. Follow-up data were collected using a provincial client management software system that includes youth and adult justice system data for British Columbia. This information was available for 1,620 of the 1,719 participants. The main analyses focused on convictions and days incarcerated between ages 12 and 25. Over 95% of the sample was followed until at least age 25.

Measures

Convictions

During the follow-up, 142 participants died and 128 moved out of the province. Incarceration and convictions after these events were considered missing. Convictions were recorded for each year of age, starting at age 12 (the age of criminal responsibility in Canada) and continuing throughout the follow-up period. Conviction types (e.g., violent, non-technical) were coded to see how incarceration influenced specific types of reoffending. A challenge is that being convicted can lead to reincarceration, reducing opportunities for more convictions. To adjust for this, the number of convictions in a given year was multiplied by the proportion of that year the person spent in the community. For example, one conviction while spending half a year in the community would be counted as two prorated convictions.

Incarceration

Incarceration included time spent awaiting trial and time served as a sentence. In British Columbia, the same facilities are used for youth awaiting trial and serving sentences. Adult systems have separate facilities, with more services in facilities for sentenced adults. This study could not distinguish between time spent awaiting trial versus serving a sentence. The study directly recorded admission and release dates from custody centers. Days incarcerated were summed for each year of age. This method accounted for early release and community supervision that often occurs in the final third of youth sentences in line with rehabilitation models. This approach avoided overestimating incarceration time based on sentencing data alone. The data did not allow for measuring how many times a person was incarcerated each year or the length of each separate sentence or period of remand. Therefore, the study examined the impact of total time spent incarcerated, not individual sentence lengths. This aligns with prior research and views of community reentry as an ongoing process. The provincial database included information on both federal (sentences of two years or more) and provincial (sentences of less than two years) custody. For participants with federal sentences, if a release date was not specified, the parole eligibility date was used.

Demographic Characteristics

Information on gender and ethnicity was collected through self-report interviews. The study sample's demographic characteristics reflect the youth custody population in Canada. The sample included 1,341 male and 372 female participants; gender data were missing for six individuals. The majority of the sample was White (54.7%). Indigenous persons were overrepresented (30.1%) compared to the general population of British Columbia. Other participants (15.2%) from various backgrounds (e.g., Black, East Asian, Hispanic) were grouped together due to low numbers for individual analysis.

Age-Stages

To account for how the relationship between incarceration and reoffending might change with age, four age-stages were created: 12-17; 18-23; 24-29; 30-35. In Canada, individuals aged 12-17 are subject to youth justice laws. Three additional six-year age intervals were created for consistent comparisons. The study examined the relationship between year-over-year changes in incarceration and convictions within each age-stage. To look at longer-term effects, the total days incarcerated and total convictions were summed for each age-stage. This allowed for examining whether changes in incarceration from one age-stage to the next were linked to changes in convictions from one age-stage to the next.

Cohort and Period Effects

Participants were born between 1979 and 1998. Three birth cohorts were formed (1980-1981, 1990-1991, 1994-1995) to account for changes in the experience of incarceration over time. These groupings were chosen to resemble those used in other studies. Over time, British Columbia Corrections made efforts to shift both its youth and adult systems towards a more rehabilitative approach. This change in focus could influence how people experience and respond to incarceration. The decline in youth incarceration rates in British Columbia starting in the late 1990s is partly due to changes in youth justice laws. Canada's Young Offenders Act (YOA) was in effect from 1984 to 2003, replaced by the Youth Criminal Justice Act (YCJA). The study divided the sample based on whether they were judged under the YOA (1998-2001) or YCJA (2005-2011). The YCJA aimed to lower incarceration rates by allowing custody only for severe offenses and requiring judges to justify why other options were not suitable. Under the YCJA, the goal of incarceration was public safety through rehabilitation, and specific deterrence was not a sentencing principle. Therefore, youth under the YCJA may differ from those under the YOA in terms of risk of reoffending and incarceration experiences, which could impact the relationship between incarceration and reoffending.

Analytic Strategy

A fixed-effect estimator was used to study the effect of year-over-year changes in incarceration within a person on year-over-year changes in convictions within that person. By focusing on changes within individuals, this method accounts for differences between people that are unchanging over time, which helps to address selection bias. However, differences between people in factors that change over time, like the number of days incarcerated or convictions, might still be linked to unobserved unchanging factors. To address this, the data were "detrended" to remove these between-person differences. Two methods for detrending are time-demeaning and first-differencing. Both are valid for fixed-effect estimates, but they can produce different results because they capture different aspects of within-person change. The choice of method should align with the research question.

Time-demeaning involves subtracting a person's average value across all measurement periods from their value at each period. While this is a popular method for removing between-person variation and producing unbiased estimates, it can be theoretically misleading because it uses future information to calculate present change. For example, a person might experience an increase in incarceration time between two periods, but if they have a much longer incarceration period later, the time-demeaned value for the earlier period might appear as a decrease. This may not accurately reflect how a person experiences punishment or how policies like increased sentencing for repeat offenders are meant to work.

First-differencing detrends data by subtracting values from the previous period (time t-1) from the current period's values (time t) to show period-to-period change. This method is believed to better reflect how people experience change and how policies designed to reduce reoffending by increasing incarceration are motivated. For example, it better captures how courts might increase prison time compared to a prior sentence for repeat offenders. First-differencing also avoids assuming that long-term life patterns are always predictable. Comparing the two methods, first-differencing and time-demeaning, shows that they yield different values and represent different concepts when detrending data.

The main analysis used a first-differenced fixed-effect estimator for data from ages 12 to 25. This analysis looked at how year-over-year changes in incarceration (from two years ago to one year ago) contributed to year-over-year changes in convictions (from one year ago to the present). Age differences were not a factor because all participants were the same age at each wave. All analyses controlled for age effects. Robust standard errors were used to address issues with correlated errors over time. Sensitivity analyses examined whether the relationship between incarceration and reoffending was consistent across different types of convictions and age-stages. The main analyses were also repeated after dividing the sample by gender, ethnicity, birth cohort, and exposure to different justice system policies, to account for potential factors that might influence the relationship between incarceration and reoffending.

Results

Between ages 12 and 25, participants had an average of about 20 convictions, including 2.5 violent convictions, eight technical convictions (e.g., probation violations), and 12 non-technical convictions. During this time, they spent nearly two years in custody on average. Although participants were recruited as adolescents, 80% spent at least one day incarcerated between ages 18 and 25. Among the 874 participants with data for ages 30-35, 35% spent at least one day incarcerated. Out of 21,686 total observations over time, 629 (2.9%) involved a participant spending an entire year of age incarcerated. Among the 1,410 participants with complete data between ages 12 and 25, 21.2% (299 individuals) spent at least one full year of age incarcerated. Overall, it was common for participants to be incarcerated in adulthood for relatively long periods.

Within-individual change models

The first model showed that for ages 12-25, each additional day incarcerated from two years ago to one year ago led to a 0.0063 decrease in convictions from one year ago to the present. This means an increase of one month incarcerated was linked to a 0.19 decrease in convictions. The second model, which used a longer time gap, also found that an increase in incarceration from three years ago to two years ago was associated with a decrease in convictions from one year ago to the present. The third model, which accounted for time spent free in the community, also showed that more time incarcerated was linked to fewer convictions in the next period. Finally, an increase in incarceration was associated with a significant decrease in violent, non-technical, and technical convictions (models 4-6).

The study also looked at how changes in incarceration affected changes in convictions across different measurement intervals, between broad age-stages, and within specific age-stages. For all these analyses, increases in days spent incarcerated from two years ago to one year ago were consistently linked to decreases in convictions from one year ago to the present. It was noted that the reduction in convictions following longer stays in custody was particularly strong during adolescence (ages 12-17). For this age group, a one-month increase in incarceration was linked to a 0.40 decrease in convictions. For the adult age groups, a one-month increase in incarceration was associated with a smaller decrease of about 0.12-0.15 convictions.

Regardless of gender or ethnicity, increases in the number of days incarcerated from two years ago to one year ago were consistently linked to decreases in convictions from one year ago to the present for ages 12-25. Similarly, the findings held true regardless of when participants were born or which youth justice system laws applied to them. Overlapping confidence intervals suggested that the relationship between incarceration and reoffending was not stronger for any specific birth group or legal system.

Supplemental analyses

To see if similar patterns were found in the United States, the same analysis was applied to self-reported data from the Pathways to Desistance Study, which followed boys and girls from Phoenix and Philadelphia over seven years. Consistent with the Canadian data, year-over-year increases in self-reported days incarcerated from two years ago to one year ago were significantly linked to year-over-year decreases in self-reported offending for the next period. However, this crime-reducing effect of incarceration was only seen for male participants, not female participants, and for White participants, not Black or Hispanic participants.

Discussion

Theories and policies often describe the relationship between incarceration and reoffending as a change within an individual. For instance, policies that increase sentences for repeat offenders aim to deter future crime. Until now, research has mainly used group comparisons, which do not directly address whether a person's criminal behavior changes after they spend more time in prison. Within-individual methods are better suited to answering these theoretical and policy questions because they directly look at how changes in a person's incarceration time affect their future offending. These methods also address selection bias by accounting for all unchanging factors that might influence the results. While they do not account for unobserved factors that change over time, this is also a limitation of group comparison methods. An advantage of within-individual analyses over some group comparison methods is that they can address selection bias without having to focus only on individuals who barely met the criteria for a prison sentence.

This study examined the relationship between incarceration and reoffending using data from 1,719 incarcerated youth in British Columbia, Canada. The analysis, covering ages 12-25, showed that year-over-year increases in days spent incarcerated were associated with year-over-year decreases in convictions in the following period. Specifically, an increase of one month in incarceration from two years ago to one year ago was linked to a 0.19 decrease in convictions from one year ago to the present. These reductions in convictions were not due to simply aging out of crime, as age effects were controlled. The results were consistent across different types of convictions, including violent, non-technical, and technical offenses. The decreases in re-convictions remained even when a longer time gap was used between incarceration and re-conviction, which helps to account for potential delays in court processes. These findings should not be seen as a reason to maintain Canada's current correctional system without addressing its issues, nor as support for expanding incarceration. Instead, the results may reflect what happens when incarceration is reserved for a smaller group of young people involved in serious offenses.

More than three-quarters of the Canadian sample were re-incarcerated as adults, which allowed for examining the relationship between incarceration and re-convictions at different age-stages. A negative relationship (more incarceration, fewer re-convictions) was observed when comparing aggregated age-stages and within specific age-stages. However, this reduction in re-convictions after longer custody stays was particularly notable during adolescence. For ages 12-17, a one-month increase in incarceration was linked to a 0.40 decrease in convictions, whereas for adult age groups, it was about a 0.12-0.15 decrease. Canada's youth justice system, compared to its adult system, emphasizes rehabilitation more than deterrence and provides close supervision and services upon return to the community. This stronger focus on rehabilitation in the youth system might explain why incarceration had a greater impact on reducing re-convictions during adolescence. Adolescence is also a period with more potential for positive change in response to interventions. The study notes that participants not incarcerated during a specific age-stage were excluded from that stage's analysis. Those remaining in the adult analyses were individuals with more persistent offending patterns, who might be less responsive to deterrent or rehabilitative efforts. This aligns with other research suggesting that reductions in reoffending after incarceration are weaker for repeat offenders.

Analysis of Potential Moderators

The Canadian sample was divided into subgroups to see if the findings were consistent. For both girls/women and boys/men in the study, year-over-year increases in days spent incarcerated were linked to year-over-year decreases in convictions in the next period. While some research suggests incarceration has a particularly negative impact on women, this study found that it reduced re-convictions for girls and women. This finding is consistent with recent research in Canada where incarcerated women described their time in custody as an improvement over their lives in the community.

The results were also consistent across White, Indigenous, and other racialized individuals. Custody facilities in British Columbia provide essential services like food, shelter, healthcare, and programs that might be unavailable or difficult to access in the community. The study does not imply that incarceration is a positive experience for participants but suggests that the custody environment in British Columbia might be an improvement from severe struggles experienced in the community. This could be especially true for girls and Indigenous participants who reported higher rates of substance use, family conflict, abuse, and unstable housing. This interpretation is consistent with findings that higher-risk youth in residential treatment had more days free of self-harm. Incarceration may fail to reduce reoffending if the marginalization experienced in custody mirrors that in the community. In contrast, when similar analyses were performed using data from the United States (Pathways to Desistance Study), increases in time spent incarcerated were linked to decreases in self-reported offending, but this effect was only observed for White and male participants, not for Black, Hispanic, or female participants. This suggests that prison conditions in Arizona and Pennsylvania might be less effective in reducing reoffending, possibly due to a history in the United States of incarcerating youth for minor offenses as a means of control.

Previous research indicated that the relationship between incarceration and reoffending might be influenced by the crime rates during a person's youth. Participants in the Canadian study born earlier grew up during periods with higher incarceration rates. However, regardless of their birth cohort (1980-81; 1990-91; 1994-95), year-over-year increases in days spent incarcerated were associated with year-over-year decreases in convictions in the next period. Participants also experienced either Canada's Young Offenders Act (YOA) or the Youth Criminal Justice Act (YCJA). The shift from YOA to YCJA led to lower incarceration rates and an increased focus on rehabilitation. Despite these legislative changes, increases in time spent incarcerated were linked to decreases in convictions regardless of which law applied to participants. The consistency of these findings across different birth cohorts and legal changes suggests that Canada might be in a stable period where cultural factors have a consistent influence on individual behavior. Policy changes in Canada are typically slower, less dramatic, and less impactful compared to those in the United States that led to widespread incarceration. The YCJA may not have had a unique impact on the relationship between incarceration and reoffending compared to the YOA because incarceration rates in British Columbia were already declining before the YCJA was enacted.

Study Implications, Limitations, and Future Research

Without measures taken before, during, and after incarceration, studies on the relationship between incarceration and reoffending are better at disproving theories than directly supporting them. For example, these results contradict the idea that prisons act as "schools of crime." However, like other studies, this research lacked data on participants' perceptions of punishment and their experiences during incarceration. Therefore, it would be incorrect to interpret these findings as direct support for specific deterrence (punishment discouraging future crime). Participants who experienced an increase in incarceration time might have viewed it as less punitive than other options. For example, some individuals prefer custody over lengthy community-based sentences with strict conditions. To investigate specific deterrence, future research should measure subjective perceptions of incarceration. Instead of deterrence, it is possible that longer periods of incarceration provided more opportunities for participants to receive services that led to fewer re-convictions. Other studies have found that employment training programs reduced reoffending among incarcerated individuals. Although incarceration experiences were not directly measured, the consistency of findings across groups (e.g., different genders) who typically experience incarceration differently suggests a broader impact.

Future research should consider two stages of incarceration: changes during incarceration and changes in offending after release. Changes in time spent incarcerated may influence the types of individual changes that occur while in prison. Understanding incarceration as a turning point requires measuring experiences in prison, such as access to programs, visits from supportive individuals, perceptions of incarceration, attitudes towards crime, and changes in substance use. Measures of post-incarceration experiences are also crucial because, especially in systems using a risk-need-responsivity model, those deemed higher risk receive more intervention and supervision. The reductions in re-offending following increased incarceration in this study might actually reflect an effect of increased community supervision, monitoring, and services after release. Canada's youth system, in particular, aligns with rehabilitation principles, which might explain why the effects were stronger during adolescence. No other studies are known to have separated the effects of supervision from incarceration. As weak evidence against a strong supervision effect, the findings remained consistent even when a longer time lag was used between incarceration and reoffending, meaning reductions in re-convictions were observed even beyond the initial periods of community reentry where services and supervision are most likely.

Based on a review of 116 studies, it has been suggested that there is a general agreement that incarceration does not reduce reoffending. The findings of this study do not necessarily contradict this literature. By focusing on year-over-year changes within individuals, a different research question was addressed. The study also used data from participants who were likely at higher risk compared to those in studies that focus on individuals on the margins of receiving a prison sentence. Although not directly measured, the experience of incarceration for participants in Canada might differ from experiences in the United States, which is where most of the research on incarceration and reoffending originates. Canada tends to emphasize incarceration as an opportunity for rehabilitation and service delivery rather than solely deterrence and punishment. Compared to individuals on the margins of receiving a prison sentence in the United States, the Canadian participants might be less likely to experience the severe hardships of imprisonment because their lives in the community were already particularly challenging. Especially since the 1990s, Canadian correctional researchers in policy roles have expressed concerns about punitive aspects of incarceration, promoted effective correctional practices, and developed models to identify who should be incarcerated and how to address treatment needs. These ideas might be less prominent in the United States, where there can be pessimistic views about whether high-risk offenders can change, potentially leading to a lack of rehabilitative services. For example, one study found that incarceration reduced offending for first-time offenders but not repeat offenders in Illinois. Future research should investigate whether these findings apply to individuals with psychopathic traits, who may be less likely to participate in and respond positively to treatment programs.

Like other studies, this research measured the total number of days spent incarcerated within a specific period. However, it could not distinguish between someone who entered and exited custody multiple times versus someone who had one continuous incarceration period for the same total number of days. Therefore, the study could not examine the influence of sentence length on reoffending. The functional form of the relationship between incarceration and reoffending was also not examined. It is possible that very long stays in custody have diminishing returns on reducing reoffending, especially if they result in individuals returning to the community at older ages when the risk of offending is already lower.

The use of incarceration reflects a justice system's operations, and these systems differ across jurisdictions in their values, culture, rules, and traditions. This study does not suggest that incarceration universally reduces the risk of reoffending regardless of the specific location, time period, and context. The Canadian sample represents individuals who engage in repeat offending and are regularly in contact with probation and parole officers for case management. Despite the specific nature of the Canadian sample, as shown by the supplemental analyses using data from the United States, the within-individual analytical approach used in this study can be applied to various other samples and jurisdictions.

Summary

Policies in the criminal justice system, like longer sentences for repeat offenders, are based on the idea that spending more time in jail or prison will lead to less crime after a person is released. This concept focuses on how a person's behavior changes over time. It is important to study this idea because there are concerns about how well incarceration works, its costs, and its effects on health and well-being.

However, it has been hard to know if longer prison stays reduce future crime. Most research compares groups of people—those who were incarcerated versus those who were not. This type of comparison does not show if an individual person's offending changes after they spend more time in custody. This study suggests using a different approach: analyzing changes within each person over time. This method can help answer these important questions and address selection bias, which happens when pre-incarceration risk factors (like past crimes or substance use) affect both incarceration and future offending.

Incarceration is a punishment that limits freedom and can harm a person's family connections, job prospects, and housing. Supporters of incarceration argue that these negative effects are outweighed by its ability to reduce future crime. Yet, studies have questioned this. A large review of 116 studies found that there is a general agreement that jail and prison sentences do not reduce reoffending. This suggests incarceration is costly, harmful, and fails its main goal. However, many of these studies did not properly account for selection bias. Also, most of these studies used data from the United States, which has a unique history of mass incarceration.

This study used a specific statistical method to look at the relationship between incarceration and reoffending differently. Researchers examined if yearly increases in time spent incarcerated affected yearly changes in convictions among young people in British Columbia, Canada, who were followed into adulthood. The study also looked at similar data from the United States to see if the findings were consistent, since incarceration policies and practices differ between the two countries.

1. Differences in the purpose and use of incarceration between Canada and the United States

The way incarceration is used and its goals vary by location and over time. However, few studies have looked at how incarceration affects reoffending outside of the United States. Compared to the U.S., Canada's policies for both youth and adults involve shorter, more personalized sentences that focus on rehabilitation and are less influenced by politics. For example, laws in Canada that aimed to increase mandatory minimum sentences had little effect on judges' decisions.

In British Columbia, where this study's data came from, youth and adult correctional facilities are not privately owned. The provincial government provides health services, and each facility has a mental health professional. All people entering custody are screened for mental health within 24 hours. Those in custody can access programs, such as those that help build respectful relationships, which are linked to reduced reoffending. Incarcerated youth have access to high school education and daily programs for drug and alcohol counseling, mental health support, life skills, and job readiness. Surveys have shown high satisfaction with services in British Columbia's adult correctional system. Justice professionals also receive training to improve relationships and address individual needs.

Historically, Canada's incarceration practices and rates have differed from those in the States. These differences became more pronounced from the 1990s to the 2010s. During this time, Canada shifted its approach from focusing on severe punishment to using the risk-need-responsivity (RNR) model. This model creates personalized, evidence-based plans to reduce the negative effects of imprisonment and promote rehabilitation. The RNR model involves more intervention and supervision for individuals considered higher risk. While used in both youth and adult systems, legislative changes in 2003 made the RNR model central to Canada's youth justice system. Under this system, incarceration is a last resort, emphasizes rehabilitation over punishment, and includes detailed plans for returning to the community. In contrast, the U.S. in the 1980s began adopting crime control models for youth justice, focusing on punishment rather than rehabilitation. For instance, some U.S. states allow youth to be tried as adults, a practice Canada currently prohibits.

Although Canada's justice systems for youth and adults differ from those in the U.S., these differences are not constant. Therefore, it is important to consider the specific time period being compared. Most U.S. research on incarceration and reoffending collected data during periods of mass incarceration. The relationship between incarceration and reoffending might vary depending on the era. Individual involvement with the legal system is more likely during times of higher incarceration rates. Therefore, research should consider whether the relationship between incarceration and reoffending is consistent across different birth groups and groups exposed to different justice system policies. Instead, most research has focused on addressing selection bias.

2. Generations of research on incarceration and reoffending

Research on incarceration and reoffending can be grouped by how it tries to handle selection bias. Selection bias makes research on incarceration and reoffending unreliable because other factors (like age, gender, race, current crime, or past crimes) affect both the chance of being incarcerated and the chance of reoffending.

Early research (first generation) tried to control for selection bias by using statistics to account for these other factors in analyses where incarceration was used to predict future crimes. Later research (second generation) measured these influencing factors to create a "propensity score," which shows the likelihood of receiving a jail or prison sentence. People who were incarcerated were then matched with non-incarcerated people who had similar propensity scores. This assumed that any link between incarceration and reoffending was separate from the measured influencing factors. Reviews of both first and second-generation research generally concluded that incarceration either had no effect or actually increased reoffending. However, data sets often lack enough information to account for all factors that cause selection bias. Overall, these early methods are thought to produce unreliable results about the relationship between incarceration and reoffending.

Third-generation research estimates the effects of incarceration in real-world situations by using specific characteristics of justice system policies. For example, some studies compare people who just barely met or missed the criteria for a prison sentence, assuming they are similar in other ways. Any differences in reoffending between these groups are then linked to incarceration. Other studies use the random assignment of cases to judges. These studies include people who would not be sent to prison by all judges. This means those sent to prison differ from those not sent to prison only because of the judge's strictness, not due to other unmeasured personal traits that could influence reoffending. Third-generation research largely reports that incarceration has no effect or a negative effect (increases crime). However, a few studies, especially outside the U.S., suggest that incarceration reduces reoffending. The different findings from earlier research might be due to these improvements in accounting for selection bias. But these improvements have a drawback: to account for other factors, third-generation research often focuses on people who were "on the margins" of receiving a prison sentence, excluding those with more serious crimes or longer criminal histories. This can limit how broadly the findings apply, making comparisons unrealistic. New methods are needed to address selection bias without excluding those most likely to be incarcerated.

3. Within-individual analyses as a different methodological approach

Earlier research on incarceration and reoffending compared groups of people. This approach does not match theories and policies that describe how incarceration influences a person's behavior over time. For example, the idea of "perceptual deterrence" suggests that a person's understanding of the risks and rewards of crime changes after experiencing a stricter punishment. Labeling theory suggests that changes in a person's punishment status can affect their social connections and identity, possibly leading to more crime. Policies that give repeat offenders longer sentences are based on the idea that past punishment influences future punishment decisions.

Analyzing how individual people change over time aligns with these theories and policies. This approach also helps address selection bias by accounting for all unmeasured factors that do not change over time within a person. Because each person acts as their own control, there is no need to create artificial comparison groups or limit the study to people who barely qualified for a jail or prison sentence.

Three studies have already looked at changes within individuals over time regarding incarceration. Two studies found that increases in time spent incarcerated were linked to worse job outcomes and poorer mental health. Another study found that more time in jail was linked to less mature behavior among youth. This current study adds to this research by examining the connection between year-to-year changes in incarceration within a person and their future year-to-year changes in reoffending.

4. Current study

Researchers have studied crime patterns for almost a century, but it is still not fully understood how incarceration affects these patterns. This gap in knowledge exists mainly because it is difficult to address selection bias. This study does not suggest that individual-level analyses should replace other research methods. However, these analyses can be useful for: (1) dealing with selection bias, (2) examining groups of people beyond just those who barely qualified for a prison sentence, and (3) answering different questions that can help develop and refine theories and policies about how incarceration affects individuals over time.

This study used a specific statistical method to look at the relationship between year-to-year changes in time spent incarcerated and future year-to-year changes in convictions among young people in British Columbia, Canada, who were followed into adulthood. The effect of incarceration on reoffending might differ depending on the type of crime. For example, people released from prison might be watched more closely, which could increase the chance of being caught for minor violations. Therefore, the study examined how changes in incarceration affected yearly changes in violent, technical (e.g., probation violations), and non-technical convictions. To see how stable this effect was, the researchers also looked at the impact of incarceration over an even longer period. Additionally, the crime-reducing effect of incarceration might be stronger as people get older and better at thinking about the consequences of their actions. So, the study looked at changes between different age groups (e.g., adolescence to early adulthood) and within specific age groups.

Incarceration policies vary by location and time. The implications of these differences for how incarceration affects reoffending are often overlooked. Nearly 80% of studies on this topic use data from the United States. The data for this study reflect Canada's experiences with incarceration. The study also used data from a U.S. study to see if the findings were similar, considering that Canada and the U.S. have different ways of making incarceration decisions and different experiences with incarceration. Canada did not experience an era of mass incarceration, for-profit prisons, or routine use of mandatory minimum sentences. The period of comparison is important because the relationship between incarceration and reoffending might vary depending on the era. Higher incarceration rates are linked to more individual involvement with the legal system. Therefore, the current study also divided the Canadian sample into groups based on their birth year and the youth justice policies they experienced.

Advanced research methods cannot overcome incomplete data that limit how well theories can be tested. This study lacked data that could, for example, clarify whether any observed reductions in reoffending were due to specific deterrence (punishment discouraging future crime) or rehabilitation. It is uncommon for studies to measure individual changes that happen during and after incarceration. While individual-level analyses do not account for unmeasured factors that change over time, this is also true for group-based analyses. Although this study did not measure experiences during incarceration, it is known that these experiences can vary by demographics. Therefore, additional analyses were done by gender and ethnicity.

5. Method

5.1. Sample

The Ministry of Children and Family Development, responsible for all incarcerated youth in British Columbia, approved the recruitment of 1,719 youth from custody centers across the province for the ISVYOS study. Data on refusals were not consistently recorded, but for about a five-year period, approximately 5% of youth declined to participate. The profile of incarcerated youth in Canada likely differs from that in the United States. Unlike the U.S., Canada has eliminated status offenses (crimes only for minors, like truancy), does not incarcerate youth with adults, and does not impose mandatory minimum sentences on youth. During the recruitment period (1998–2011), the youth incarceration rate in British Columbia ranged from about 40 to 80 per 100,000, while in the U.S., it ranged from about 196 to 355 per 100,000. Compared to incarcerated samples in the U.S., the ISVYOS sample is more likely to include a higher proportion of youth involved in serious or violent offenses.

5.2. Procedures

ISVYOS research assistants typically approached youth to participate within the first week of their incarceration. Some participants had been incarcerated before recruitment, and these prior periods were included in the measurement of incarceration time for this study. Youth were informed that their participation would not affect their time in custody or their sentencing outcomes. Interviews took place in a private room, away from other youth and staff. To get consent, participants were given and read an information sheet explaining the study's purpose, data collection methods, their right to withdraw at any time, and that information would be kept confidential unless they posed a direct threat to themselves or others. Follow-up data were collected using the Corrections Network (CORNET) client management software, which contains youth and adult justice system data for British Columbia. Due to sealed or inaccurate records, CORNET information was available for 1,620 of the 1,719 participants. The main analyses focused on convictions and days incarcerated between ages 12 and 25. Over 95% of the sample was followed until at least age 25.

5.3. Measures

5.3.1. Convictions

During the follow-up period, 142 participants died and 128 moved outside British Columbia. Incarceration and convictions after these events were recorded as missing. Convictions were measured for each year of age, starting at age 12 (the age of criminal responsibility in Canada) and continuing throughout the follow-up period. Conviction types (e.g., violent, non-technical) were coded to examine how incarceration influenced specific types of reoffending. A challenge in studying the link between incarceration and reoffending is that a conviction can lead to reincarceration, which then limits chances for more convictions during the remaining follow-up period. To account for different amounts of time spent in the community, for sensitivity analyses, the number of convictions in a given year was multiplied by the proportion of that year a participant was in the community rather than in custody. For example, one conviction at age 25, with six months spent in the community, would count as two prorated convictions.

5.3.2. Incarceration

Incarceration included time spent on remand (awaiting trial or sentencing) and time spent in custody serving a sentence. In British Columbia, the same facilities are used for remanded and sentenced youth. However, the adult system has separate facilities, with more services available in adult facilities for sentenced individuals. This study could not distinguish between time spent on remand versus time sentenced. The ISVYOS directly recorded each admission and release date from custody centers. Days incarcerated were totaled for each year of age. This approach considered that, following the RNR model, the final third of youth custody sentences are typically served in the community with conditions and supervision to prevent reoffending. This method avoided issues with sentencing data that might overestimate time incarcerated by not accounting for early release. The ISVYOS data did not allow for measuring the number of times a person was incarcerated each year or the length of each individual sentence/remand period. The number of days incarcerated each year was not necessarily a continuous period. Therefore, the study examined the impact of time spent incarcerated, rather than sentencing itself. This measurement strategy aligns with previous studies and research on "careers in criminalization," which recognizes that returning to the community is not a one-time event. CORNET is a provincial database but included information on both federal (two years or more) and provincial (less than two years) custody sentences. For 18% of the sample who received a federal sentence, if a release date was not specified, their parole eligibility date was used as a proxy.

5.3.3. Demographic characteristics

Gender and ethnicity information was obtained from self-report interviews. The demographic profile of the ISVYOS sample is similar to the youth custody population in Canada. The sample included 1341 male and 372 female participants, with gender data missing for six individuals. The sample was primarily White (54.7%). Indigenous persons were overrepresented (30.1%) compared to their proportion in the general population of British Columbia. The remaining participants (15.2%) from various backgrounds (e.g., Black, East Asian, South-East Asian, Hispanic) were grouped together due to small numbers for individual examination.

5.3.4. Age-stages

Given that the relationship between incarceration and reoffending may change with age, four age-stages were created: 12–17; 18–23; 24–29; 30–35. In Canada, individuals aged 12–17 are subject to youth justice laws regardless of their offense's severity. Three additional six-year age intervals were created to allow for comparisons between stages of equal length. The study examined the relationship between year-over-year changes in incarceration and convictions within each age-stage. To look at the longer-term impact of incarceration, the total number of days incarcerated and total convictions within each age-stage were summed. This allowed for examining whether changes in time incarcerated from one age-stage (e.g., 12–17) to the next (e.g., 18–23) were linked to changes in convictions from one age-stage (e.g., 18–23) to the next (e.g., 24–29).

5.3.5. Cohort and period effects

Participants were born between 1979 and 1998. Three birth cohorts were created to account for differences over time in the experience of incarceration (1980–1981, n = 190; 1990–1991, n = 343; 1994–1995, n = 184). These groups were designed to resemble groupings used in other studies examining birth cohort differences in justice system outcomes. Over time, British Columbia Corrections made efforts to shift both its youth and adult systems towards a more rehabilitative focus. This shift could influence experiences with and responses to incarceration. The declining youth incarceration rate in British Columbia, starting in the late 1990s, is partly linked to changes in youth justice laws. Canada's Young Offenders Act (YOA) was in effect from 1984 to 2003, then replaced by the Youth Criminal Justice Act (YCJA). The ISVYOS sample was divided based on whether participants were processed under the YOA (1998–2001; n = 596) or YCJA (2005–2011; n = 1123). In line with RNR principles, the YCJA aimed to lower incarceration rates by only allowing custody sentences for youth whose offenses met certain severity criteria. Judges were also required to explain why other sentencing options were unsuitable for public safety. The YCJA's goal for incarceration was to promote public safety through rehabilitation. Unlike the YOA, the YCJA did not include specific deterrence as a sentencing principle. In short, youth under the YCJA differ from those under the YOA, both in terms of being at higher risk of reoffending and in their incarceration experiences, which may affect the relationship between incarceration and reoffending.

5.4. Analytic strategy

A fixed-effect estimator was used to examine how year-over-year changes in incarceration affected year-over-year changes in convictions within each person. By focusing on individual variations, this method accounts for unmeasured factors that do not change over time, which often cause selection bias. However, differences between people still exist regarding the number of days incarcerated and convictions, and these differences might be linked to unmeasured factors that do not change over time. These between-person differences in changing factors can be removed by "detrending" the data.

Two common ways to detrend data are "time-demeaning" and "first-differencing." Both are valid statistical methods that produce accurate estimates and show the extent and direction of change within a person, not just the size of differences between people. However, they process data differently and can lead to different answers to the same question. The choice of method should depend on how well it fits the theory or research question.

Time-demeaning involves subtracting a person's average value across all measurement periods from their value at each specific period. For example, if someone spent 14, 28, and 192 days in custody, and their average was 78 days, their value at wave 2 (28 days) would become -50 (28-78), suggesting a decrease in time incarcerated. This method is common for removing between-person variation in incarceration and produces unbiased estimates of the relationship between incarceration and reoffending that are not affected by unmeasured time-invariant factors. However, these estimates can be misleading because time-demeaning uses future information to calculate past changes. For instance, the long incarceration period at wave 3 influences the average, making the wave 2 value (28 days) appear as a decrease, even if it was an actual increase from wave 1 (14 days). This approach might not accurately reflect a person's experience of punishment or how courts apply policies like longer sentences for repeat offenders. It also doesn't align with how people make decisions in the present based on current and past experiences, not future ones. So, time-demeaning might lead to misinterpretations of how incarceration changes offending, impacting conclusions about theories and policies.

First-differencing detrends data by subtracting the value at the previous time point (e.g., Wave 1) from the current time point (e.g., Wave 2) to show change from one period to the next. For example, with first-differencing, if someone spent 14 days in custody at Wave 1 and 28 days at Wave 2, their Wave 2 value would be 14 (28-14), indicating an increase in time incarcerated. Researchers believe first-differencing better reflects how people experience change and better captures the variation that drives policies aimed at reducing reoffending through increased incarceration. This method also aligns with how courts apply "recidivism sentencing premiums," where a person's incarceration time is increased compared to their previous sentence. It also captures the idea of "labeling theory," where courts can reinforce a person's "deviant" status by increasing their incarceration time compared to a previous punishment. By focusing on change from one period to the next, first-differencing avoids assuming that long-term life patterns are always predictable.

The study showed that processing the Canadian data for convictions and incarceration with first-differencing or time-demeaning led to different results. For example, at age 14, using first-differencing, the sample showed an average increase of about 20 days in custody compared to age 13. In contrast, using time-demeaning, the sample showed an average decrease of about 30 days compared to the overall sample mean. The average correlation between these two methods was 0.619, indicating they reflect different concepts and produce different values.

The main analysis used a first-differenced fixed-effect estimator for data from ages 12–25. This analysis determined how year-over-year changes in incarceration (from two years ago to one year ago) contributed to year-over-year changes in convictions (from one year ago to the present). All participants were the same age at each wave, so there were no between-participant age differences. All analyses accounted for the linear and quadratic effects of aging.

The formula for the analysis was: Convictions at time t - Convictions at time t-1 = (β0 + β1 * Incarceration at time t-1 + αi + ui) - (β0 + β1 * Incarceration at time t-2 + αi + ui-1)

Here, Convictions represents the number of convictions for an individual i at age t, and Incarceration represents the number of days spent incarcerated for individual i at age t. Subtracting each part of the equation removes all unmeasured factors that do not change over time (αi). What remains is the effect of differences in incarceration and the error term. Robust standard errors were used to address potential serial correlation. Sensitivity analyses checked if the relationship between incarceration and reoffending was consistent across different types of convictions and age groups. The main analyses were also repeated after dividing the sample to account for factors that might influence this relationship, such as gender, ethnicity, birth cohort, and exposure to different justice system policies.

6. Results

Between ages 12 and 25, participants averaged approximately 20 convictions. This included 2.5 violent convictions (e.g., assault, robbery, homicide), eight technical convictions (e.g., violating probation conditions), and 12 non-technical convictions. During this period, participants spent almost two years in custody. Although participants were recruited as adolescents, 80% of them spent at least one day incarcerated between ages 18 and 25. Among the 874 participants with data available between ages 30 and 35, 35% spent at least one day incarcerated. Across 21,686 total person-period observations, 629 (2.9%) instances showed an ISVYOS participant spent the entire observation period (i.e., a full year of age) incarcerated. Among participants with complete data between ages 12 and 25 (n = 1410), 21.2% (n = 299) spent at least one full year of age incarcerated. Overall, it was common for participants to be incarcerated in adulthood and remain in custody for relatively long periods.

6.1. Within-individual change models

The first analysis model (Table 2, Model 1) looked at the effect of year-over-year changes in incarceration (from two years ago to one year ago) on year-over-year changes in convictions (from one year ago to the present) for individuals between ages 12 and 25. Results showed that for each additional day spent incarcerated from two years ago to one year ago, there was a 0.0063 unit decrease in convictions from one year ago to the present. This means an increase of one month (about 30 days) spent incarcerated from two years ago to one year ago was linked to a 0.19 decrease in convictions from one year ago to the present.

Model 2 extended the time lag by another year to see if the impact of incarceration changes remained stable. It found that a year-over-year increase in incarceration from three years ago to two years ago was also linked to a year-over-year decrease in convictions from one year ago to the present. Model 3 adjusted the outcome variable to reflect the number of convictions per day spent free in the community. Consistent with the other models, increased time incarcerated was associated with fewer convictions in the next measurement period. Finally, an increase in incarceration from two years ago to one year ago was linked to a significant decrease in violent, non-technical, and technical convictions (Models 4–6) from one year ago to the present.

Table 3 shows the effect of incarceration changes on conviction changes across different measurement intervals and within specific age groups. All models consistently showed that increases in time spent incarcerated from two years ago to one year ago were associated with significant decreases in convictions from one year ago to the present. Notably, the impact of incarceration on reducing re-convictions was stronger during adolescence (ages 12–17) compared to the three adult age groups. For adolescents, an increase of one month incarcerated from two years ago to one year ago was linked to a 0.40 decrease in convictions from one year ago to the present. For the adult age groups, a one-month increase in incarceration was linked to about a 0.12–0.15 decrease in convictions.

Table 4 indicates that the findings were consistent regardless of gender (Models 1–2) or ethnicity (Models 3–5). Year-over-year increases in time incarcerated were associated with significant year-over-year decreases in convictions for all groups. Table 5 shows similar consistency across different birth cohorts (Models 1–3) and youth justice system laws (Models 4–5). Increased time incarcerated was linked to significant decreases in convictions at the next measurement period. The overlapping confidence intervals suggested that the relationship between incarceration and reoffending was not stronger for any specific cohort or group exposed to particular laws.

6.2. Supplemental analyses

To see if similar findings occurred among incarcerated youth in the United States, the same analysis method was used with self-reported data on incarceration and diverse offending behaviors over seven years from boys and girls (n = 1354) in Phoenix and Philadelphia, who were part of the Pathways to Desistance Study. Like the Canadian data, yearly increases in time incarcerated from two years ago to one year ago were significantly linked to yearly decreases in self-reported offending from one year ago to the present. However, this crime-reducing effect of incarceration was observed only for male participants, not female participants, and for White participants, but not for Black or Hispanic participants.

7. Discussion

Theories and policies generally describe the link between incarceration and reoffending as a change within an individual. For example, policies that give longer sentences for repeat offenders aim to deter that person from committing more crimes. Until now, research mostly used comparisons between groups to see if experiencing incarceration reduces a person's likelihood of reoffending. Within-individual methods directly address these theoretical and policy questions by looking at whether a person's offending changes after they spend more time incarcerated. These analyses also help address selection bias by accounting for all unmeasured factors that do not change over time within an individual. While within-individual analyses do not account for unmeasured factors that change over time, group-based analyses also face this limitation. An advantage of within-individual strategies over group comparisons, like regression discontinuity designs, is that they address selection bias without having to focus only on individuals who barely met the criteria for a jail or prison sentence.

This study examined the relationship between incarceration and reoffending using data from youth in British Columbia, Canada. A specific statistical method was used to see if yearly increases in time spent incarcerated influenced yearly changes in convictions. Based on analyses between ages 12 and 25, yearly increases in incarceration were linked to yearly decreases in convictions in the next measurement period. Specifically, an increase of one month spent incarcerated from two years ago to one year ago was associated with a 0.19 decrease in convictions from one year ago to the present. Changes due to age were considered, meaning the reduction in convictions was not simply due to the natural decrease in crime risk that often comes with adulthood. These findings were consistent across different types of convictions, including violent, non-technical, and technical offenses. Decreases in reconvictions linked to increased incarceration were still observed even when looking at an extended time lag between incarceration and reconviction. This longer lag helped account for delays in court proceedings between an offense and a conviction.

It would be wrong to interpret these findings as a reason for Canada's correctional system to stay the same. Problems with Canada's correctional system still exist and should not be ignored. The findings should also not be seen as support for increasing the use of incarceration. In fact, these results might show what happens when incarceration is reserved for a small group of youth involved in particularly serious or violent offenses.

Over three-quarters of the study participants were re-incarcerated as adults. This allowed for examining the relationship between incarceration and reconvictions at different age stages. A negative relationship (more incarceration, fewer reconvictions) was observed when comparing changes between broad age stages (e.g., the effect of changes in incarceration from ages 12–17 to 18–23 on changes in convictions from ages 18–23 to 24–29). This relationship was also found when looking at changes within specific age stages (e.g., ages 12–17 only; ages 30–35 only). However, the reduction in reconvictions after longer stays in custody was particularly strong during adolescence (ages 12–17) compared to the adult age stages.

Compared to its adult system, Canada's youth justice system emphasizes rehabilitation more than punishment. Youth sentenced to incarceration receive close supervision, monitoring, and services when they return to the community. While speculative, this greater focus on rehabilitation in the youth system might explain why incarceration was more strongly linked to reductions in reconvictions during adolescence. Adolescence is also a period when there is more potential for change, including positive responses to interventions. Therefore, differences in the strength of the incarceration effect might be related to youth having a greater capacity to respond well to rehabilitation efforts. It should also be noted that participants not incarcerated during a specific age stage were removed from the analysis for that stage. Thus, those who remained in the analyses for specific adult age stages were individuals with the most persistent patterns of offending. Such individuals may be less responsive to punishment or rehabilitation efforts. Data limitations prevented a direct test of this idea, but this speculation aligns with findings that reductions in reoffending after incarceration were weaker for repeat offenders.

7.1. Analysis of potential moderators

The study divided the sample to see if findings were consistent across different groups. For both girls/women and boys/men, year-over-year increases in time spent incarcerated were linked to year-over-year decreases in convictions in the next measurement period. Some research suggests that incarceration has an especially negative impact on women. However, this study found that incarceration reduced reconvictions for girls and women. This finding aligns with recent Canadian research where incarcerated women described their time in custody as an improvement over their lives in the community.

Results were also consistent across White, Indigenous, and other racialized individuals. Custody facilities in British Columbia provide basic needs, health care, and programs that are often unavailable or hard to get in the community. This does not mean incarceration is a positive experience for participants, but the custody environment in British Columbia might be an improvement from severe challenges faced in the community. This could be particularly true for girls and Indigenous participants, who reported higher rates of substance use, family conflict, abuse, and unstable housing. This interpretation is consistent with findings that high-risk youth in a treatment facility spent more days free of self-harm compared to lower-risk youth, possibly due to more treatment and supervision. Incarceration might not reduce reoffending if the difficulties experienced in custody are similar to those in the community. When similar analyses were done using data from a U.S. study, increases in time incarcerated were linked to decreases in self-reported offending. However, when divided by demographics, this crime-reducing effect was seen for White and male participants, but not for Black, Hispanic, or female participants. This suggests that prison conditions in Arizona and Pennsylvania might be less effective at reducing reoffending. The U.S. has a history of incarcerating youth for minor offenses to control behavior.

Previous research suggested that the relationship between incarceration and reoffending might be influenced by the crime rates during a person's formative years. Participants in this study who were born earlier grew up during times with higher incarceration rates. The sample was divided into three birth cohorts. Regardless of when participants were born, year-over-year increases in time incarcerated were associated with year-over-year decreases in convictions in the next measurement period. Participants experienced either Canada's Young Offenders Act (YOA) or the Youth Criminal Justice Act (YCJA). The shift from YOA to YCJA led to lower incarceration rates and changes in the purpose and experience of incarceration, including a greater focus on rehabilitation and alignment with RNR principles. However, year-over-year increases in time incarcerated were associated with year-over-year decreases in convictions regardless of which law participants were processed under. The consistency of these findings across birth cohorts and legal changes might suggest that Canada is in a "settled period" where it is difficult to see an independent influence of cultural factors on individual behavior. Policy changes in Canada are slower, less dramatic, and less impactful compared to U.S. policy changes that led to mass incarceration. The YCJA might not have had a unique impact on the relationship between incarceration and reoffending compared to the YOA because incarceration rates in British Columbia were already declining before the YCJA was enacted.

7.2. Study implications, limitations, and future research

Without data from before, during, and after incarceration, studies on the relationship between incarceration and reoffending are better at disproving theories than directly supporting them. For instance, these results contradict the idea that prisons act as "schools of crime." However, like other studies, this research lacked data on participants' perceptions of punishment and their experiences during incarceration. Therefore, it would be incorrect to interpret the findings as direct support for "specific deterrence" (punishment discouraging future crime). Participants who experienced a year-over-year increase in time incarcerated might still have viewed this as less harsh than other options. For example, some individuals prefer custody over strict community-based sentences like probation. To examine specific deterrence, future research should measure subjective perceptions of incarceration.

Instead of deterrence, it is possible that longer periods of incarceration offered participants more opportunities to receive services that later led to fewer reconvictions. Research has shown that employment training programs can lead to lower rates of reoffending among incarcerated individuals. Although incarceration experiences were not directly measured, the consistent findings across groups (e.g., different genders) who typically experience incarceration differently suggest that the positive effects may not be due to specific factors related to gender or ethnicity.

Future research should consider that incarceration involves two stages: changes that happen during incarceration and changes in offending after release. Changes in time spent incarcerated might affect the types of individual changes that occur during incarceration. Understanding incarceration as a two-stage turning point requires measuring incarceration experiences, such as exposure to programs, visits from supportive contacts, perceptions of incarceration, attitudes toward crime, and changes in substance use. Measures of post-incarceration experiences are also important because, especially where the RNR model is used, individuals considered higher risk receive more intervention and supervision resources.

Reflecting on the current study's findings, increases in time spent incarcerated from one period to the next might be a stand-in for increases in supervision, monitoring, and rehabilitation support received after returning to the community. Reductions in reoffending after increased incarceration might therefore actually reflect an effect that occurs after incarceration, where participants receive more community supervision, monitoring, and services. Canada's youth system, in particular, reflects RNR principles, which might explain why the positive effects were stronger during adolescence than in adulthood. No other studies are known to have separated the effect of supervision from the effect of incarceration itself. As weak evidence against a supervision effect, the findings remained consistent even when the time lag between incarceration and reoffending was extended. This means reconvictions were lower even beyond the initial periods of community reentry where services, supervision, and monitoring are most likely.

Based on their review of 116 studies, other researchers suggested a general agreement that incarceration fails to reduce reoffending. The findings of this study do not necessarily contradict that literature. By focusing on year-over-year changes within individuals, a different research question was answered. Also, the study used data from a sample of participants who are likely higher-risk compared to samples in other studies that focus on individuals who barely met the criteria for a jail or prison sentence. While not directly measured, the experience of incarceration for Canadian participants might differ from experiences in the United States, where most of the literature on incarceration and reoffending is based. Canada is more likely to emphasize incarceration as an opportunity for rehabilitation and service delivery rather than punishment. Compared to individuals on the margins of receiving a jail or prison sentence in the U.S., the Canadian participants might be less likely to experience the severe hardships of imprisonment because their community experiences are already especially challenging. Since the 1990s, Canadian correctional psychology researchers in policy positions raised concerns about harsh aspects of incarceration, promoted effective correctional practices, and developed the RNR model to help identify who should be incarcerated and how to address treatment needs. These ideas might hold less weight in the United States, where there is a tendency to be pessimistic about whether high-risk offenders can change. Such attitudes could lead to withholding rehabilitation services. For example, some research found that among people incarcerated in Illinois, incarceration reduced offending for first-time offenders but not for repeat offenders. Future research should examine whether these findings apply to people with psychopathic traits, who may be less likely to engage in and respond positively to treatment programs.

Like other studies, this study measured the number of days spent incarcerated within a specific period. Two participants might have spent the same number of days incarcerated in a given year, but one might have entered and exited custody five times, while the other entered and exited only once after a single sentence. The study's data could not distinguish between such patterns, so it was not possible to examine how sentence length affects reoffending. The study also did not examine the specific pattern of the relationship between incarceration and reoffending. Long stays in custody might have diminishing returns on reducing reoffending, especially if they lead to returning to the community at older ages when the risk of offending is naturally lower.

The use of incarceration reflects how a justice system functions, and the values, culture, rules, and traditions of this system differ across locations. This study does not suggest that incarceration decreases the risk of reoffending regardless of the location, era, and context. The study sample represents individuals who frequently reoffend and with whom probation and parole officers regularly interact for case management. Despite the specific nature of the Canadian sample, as shown by supplemental analyses using U.S. data, the within-individual analytic strategy used in this study can be applied to various other samples and locations.

Summary

Policies in the justice system often assume that holding someone in jail or prison for a longer time will make them less likely to break the law after they are released. This idea suggests that a person's behavior changes over time based on their experience with the justice system. It is important to study this idea because there are concerns about how well incarceration works, its cost, political influences, and its effects on health and well-being.

Most research on this topic compares groups of people—those who were incarcerated and those who were not. However, these studies do not show if a person's behavior changes when their time in custody increases. This study looks at changes within the same person over time. This approach can help answer questions about policies and practices, and it also helps address issues like selection bias. Selection bias occurs when pre-existing risk factors (like past criminal behavior or substance use) are not considered, as these factors can influence both incarceration and future offending. Analyzing changes within a person over time helps account for these hidden, unchanging factors.

Incarceration takes away freedom and can negatively affect a person's life, including family connections, jobs, and housing. Supporters of incarceration believe these negative effects are balanced by a reduction in future law-breaking. However, some research suggests that incarceration generally does not reduce reoffending. These past studies often used methods that did not fully account for selection bias and mostly focused on experiences in the United States. This study used a different method to examine if yearly increases in time spent incarcerated affected yearly changes in convictions among young people in British Columbia, Canada, who were followed into adulthood. The study also checked if similar results were found in the United States, where incarceration policies are different.

1. How Incarceration is Used in Canada and the United States

The reasons for and ways that incarceration is used differ depending on the place and time. Few studies have looked at the connection between incarceration and reoffending outside of the United States. Compared to the United States, Canada's policies for young people and adults in custody focus on shorter sentences that are more personalized, aimed at rehabilitation, and less influenced by politics. For example, Canadian government policies meant to increase mandatory minimum sentences had little effect on judges' decisions.

In British Columbia, where the study data were collected, youth and adult detention centers are not privately owned. The provincial government provides physical and mental health services, with a mental health professional at each facility to coordinate care. Everyone entering custody is screened for mental health within 24 hours. People in custody can access programs that help reduce reoffending. Young people in custody receive high school education and daily programs for counseling, mental health support, life skills, and job readiness. Surveys show high satisfaction with services among adults in British Columbia's correctional system. Justice professionals also receive training to improve relationships, address gender-based needs, and reduce criminal thinking.

Historically, Canada's incarceration practices and rates have differed from those in the United States, especially from the 1990s to the 2010s. During this time, Canada's correctional approach shifted from severe punishment to individualized, evidence-based plans based on the "risk-need-responsivity" (RNR) model. This model aims to reduce the harshness of imprisonment and promote rehabilitation. The RNR model involves more intense intervention and supervision for those considered higher risk. While used for both youth and adults, legislative changes in 2003 made the RNR model central to Canada's youth justice system, where incarceration is a last resort, emphasizes rehabilitation over punishment, and includes detailed plans for returning to the community. In contrast, the United States in the 1980s began focusing on punishment for young people rather than rehabilitation. For instance, some U.S. states allow young people to be tried in adult court, a practice Canada currently prohibits.

Even though Canada's justice system differs from the U.S., it is important to consider the specific time periods being compared. Most U.S. research on incarceration and reoffending comes from eras of widespread incarceration. The connection between incarceration and reoffending may change depending on the era. Research should consider if this relationship is consistent across different birth groups and groups affected by different justice policies.

2. Research on Incarceration and Reoffending

Different types of research on incarceration and reoffending have used various ways to address selection bias. Selection bias occurs when factors like age, gender, race, current offense, or past criminal behavior influence both the likelihood of incarceration and future law-breaking, making research results unreliable.

Early research tried to control for selection bias by statistically accounting for these factors in analyses where incarceration was used to predict reoffending. Later research measured these factors to create a "propensity score," which represents the chance of receiving a custody sentence. People who were incarcerated were then matched with those who were not, but had similar scores. This approach assumed any link between incarceration and reoffending was separate from the observed factors. Studies from these early approaches generally found that incarceration either had no effect or actually increased reoffending. However, data sets often lack enough information to account for all factors that create selection bias, making these methods unreliable.

More recent research tries to estimate the effects of interventions in real-world settings by looking at specific aspects of justice system policies. For example, some studies compare people who just barely met the criteria for a custody sentence with those who just missed it, assuming they are similar in other ways. Any differences in reoffending are then linked to incarceration. Other studies use random assignments of cases to judges. These studies include individuals who would not be sent to prison by all judges. This means that those sent to prison differ from those not sent to prison only because of their judge's approach, not because of other hidden traits that could affect reoffending. This type of research usually reports that incarceration has no effect or increases reoffending, but a few studies, especially outside the U.S., suggest that incarceration reduces reoffending. These improved methods for addressing selection bias sometimes come with a cost. To account for confounding factors, these studies often focus on people who were on the edge of receiving a custody sentence. By excluding people involved in more serious offenses or those with longer criminal histories, these studies may make unrealistic comparisons and limit how broadly their findings apply. Better analytical strategies are needed that address selection bias without excluding those most likely to experience incarceration.

3. A Different Approach: Analyzing Within-Individual Change

Past research used methods that compared groups to look at reoffending rates between those who were incarcerated and those who were not. These methods do not align with theories and policies that describe how incarceration influences reoffending through changes within a person. For example, some theories suggest that a person's thoughts about the costs and benefits of crime change after experiencing a harsher punishment. Other ideas suggest that the likelihood of reoffending changes when there is a personal difference between expected and actual punishment. Labeling theory suggests that changes in a person's punishment status affect their social connections and identity, possibly leading to more crime. Policies that increase sentencing for repeat offenders are based on the idea that a person's past punishment influences how severe their current punishment will be.

Analyzing how a person changes over time aligns with how these theories and policies describe the relationship between incarceration and reoffending. This approach also helps address selection bias by controlling for all unchanging factors that are not directly observed. Because each person serves as their own comparison, there is no need to create unrealistic control groups or limit the study to people who barely met the criteria for a custody sentence.

Three studies have looked at changes within a person over time using a specific statistical method. Two studies in the U.S. found that more time spent incarcerated was linked to fewer positive job outcomes and fewer positive mental health outcomes. Another study found that increases in self-reported time spent incarcerated were linked to lower psychological maturity in young people. This study adds to this research by examining the relationship between changes in incarceration within a person over time and future changes in reoffending within the same person.

4. Current Study

Research on criminal behavior patterns has existed for nearly a century. However, it remains unclear whether incarceration shapes these patterns. This gap in knowledge exists because it is difficult to address selection bias. This study does not suggest that within-individual analyses should replace other methods. However, this type of analysis can be useful for (1) addressing selection bias, (2) studying groups beyond those who barely qualified for a custody sentence, and (3) answering different questions that can help evaluate, improve, and create theories and policies that consider the relationship between incarceration and reoffending from the perspective of within-individual change.

The study used a specific statistical method to examine the connection between yearly changes in days incarcerated and future yearly changes in convictions among young people in British Columbia who were followed into adulthood. The effect of incarceration on reoffending might vary by crime type. For example, people released from prison might be watched more closely, which could increase the chance of being caught for minor violations. Therefore, the study looked at how changes in incarceration affected yearly changes in violent, technical (e.g., probation violations), and non-technical convictions. To see how stable this effect was, the study also increased the time gap between changes in incarceration and changes in convictions by an extra year. Additionally, incarceration might reduce crime more effectively as people get older and better understand the consequences of reoffending. So, the study looked at changes between different age groups (like from adolescence to early adulthood) and changes within specific age groups (like ages 12–17).

Incarceration policies differ across regions and over time. The impact of these differences on the relationship between incarceration and reoffending is often not considered. Almost 80% of studies on this topic use data from the United States. The data for this study reflect Canada's experiences with incarceration. Data from a U.S. study were also used to see if the findings were similar when applying similar analyses to young people who had been incarcerated in the U.S. Despite being close neighbors, Canada and the U.S. have different ways of deciding about incarceration and different experiences with it. Canada did not experience a period of widespread incarceration, private prisons, or routine use of mandatory minimum sentences. When comparing countries, it is also important to consider the specific time period, as most U.S. research on incarceration and reoffending used data collected during eras of widespread incarceration. Differences in how incarceration is managed and experienced over time can affect reoffending. Therefore, this study also divided the Canadian sample into groups based on when they were born and under which youth justice system policies they were judged.

New research methods cannot overcome insufficient data that limit the testing of theories. This study lacked data that could clarify, for instance, whether any observed reductions in reoffending were due to specific deterrence (punishment discouraging future crime) or rehabilitation. It is rare for studies to measure individual changes that happen during incarceration and after returning to the community. While within-individual analyses do not account for unobserved factors that change over time, this is also true for group comparison analyses. Although this study did not measure experiences while incarcerated, given that incarceration experiences are thought to vary by demographics, additional analyses divided the Canadian sample by gender and ethnicity.

5. Method

5.1. Sample

The Ministry responsible for youth in British Columbia agreed to recruit young people (1,719 participants) from custody centers across the province for the study. While refusal rates were not consistently recorded, for about a five-year period, only about 5% of young people declined to participate. The profile of incarcerated youth in Canada likely differs from that in the United States. Unlike the U.S., Canada has eliminated "status offenses" (behaviors that are only illegal for minors), does not incarcerate youth with adults, and does not impose mandatory minimum sentences on young people. During the period participants were recruited (1998–2011), the youth incarceration rate in British Columbia was roughly 40 to 80 per 100,000. In comparison, the U.S. youth incarceration rate ranged from approximately 196 to 355 per 100,000. Compared to incarcerated groups in the U.S., the Canadian sample is more likely to include a higher proportion of young people involved in serious or violent offenses.

5.2. Procedures

Study assistants typically invited young people to participate in the study within the first week of their incarceration. Some participants had been incarcerated before their recruitment; these earlier periods were included in the measurement of incarceration time for this study. Young people were told that their involvement in the study would not affect their time in the custody center, including the outcome of their sentencing. Interviews took place in a private room away from other young people and staff. To get their agreement to participate, participants were given and read an information sheet explaining the study's purpose, how data would be collected, their right to withdraw at any time, and that information would be kept confidential unless they threatened themselves or others. Follow-up data were collected using the Corrections Network (CORNET) client management software, which includes data for all young people and adults involved in the justice system in British Columbia. Due to sealed records or inaccurate record-keeping, CORNET information was available for 1,620 of the 1,719 participants. The main analyses focused on convictions and days incarcerated between ages 12–25. Over 95% of the participants were followed until at least age 25.

5.3. Measures

5.3.1. Convictions

During the follow-up period, 142 participants died and 128 others moved outside British Columbia. Incarceration and convictions after these events were considered missing. Convictions were recorded for each year of age, starting at age 12 (the age of criminal responsibility in Canada) and continuing throughout the follow-up period. The type of conviction (e.g., violent, non-technical) was recorded to see how incarceration affected different kinds of reoffending. One challenge in studying the link between incarceration and reoffending is that a conviction can lead to being re-incarcerated, which then reduces the chance of getting more convictions during the remaining follow-up period. To account for differences in how much time people spent in the community, as part of sensitivity analyses, the number of convictions in a given year was multiplied by the proportion of that year the participant was in the community rather than in custody. For example, one conviction at age 25 while spending six months in the community would be counted as two prorated convictions.

5.3.2. Incarceration

Incarceration included time spent in detention while awaiting trial (remand) and time spent in custody serving a sentence. In British Columbia, the same facilities are used for youth on remand and those serving sentences. However, the adult system has separate facilities, with more services available for sentenced adults. This study could not distinguish between time spent on remand versus time serving a sentence. The study directly recorded each date a participant was admitted to and released from a custody center. Days incarcerated were added up for each year of age. This method accounted for the fact that, in line with the RNR model, the final third of youth custody sentences are typically served in the community with conditions and supervision to help prevent reoffending. This approach avoided issues with sentencing data that might overestimate time spent incarcerated by not accounting for early release. The data did not allow for measuring how many times a person was incarcerated each year, nor the length of time spent in custody for each individual sentence or remand period. The total number of days incarcerated each year was not necessarily a continuous period. Therefore, the study examined the impact of total time spent incarcerated, rather than the influence of specific sentence lengths. This measurement strategy is consistent with previous research and the idea that returning to the community is not a one-time event. The CORNET database, while provincial, included information on both federal (two years or more) and provincial (less than two years) custody sentences. For 18% of participants who received a federal sentence, if a release date was not specified, their parole eligibility date was used as a proxy for their release date.

5.3.3. Demographic characteristics

Information on gender and ethnicity came from self-report interviews. The demographics of the study sample are similar to the general population of young people in custody in Canada. The sample included 1,341 male and 372 female participants, with gender data missing for six individuals. The sample was primarily White (54.7%). Indigenous individuals were overrepresented in the sample (30.1%) compared to their proportion in the general population of British Columbia. The remaining participants (15.2%) from various backgrounds (e.g., Black, East Asian, South-East Asian, Hispanic) were grouped together because there were too few in each individual category to study separately.

5.3.4. Age-stages

Given that the relationship between incarceration and reoffending might change with age, four age groups were created: 12–17; 18–23; 24–29; 30–35. In Canada, people aged 12–17 are judged under youth justice laws, regardless of how serious their offense is. Three additional six-year age intervals were created to allow for comparisons between age groups of equal length. The study looked at the relationship between yearly changes in incarceration and convictions within each age group. To examine the longer-term impact of incarceration, the total number of days incarcerated and total convictions within each age group were summed. This allowed for studying whether changes in days incarcerated from one age group (e.g., 12–17) to the next (e.g., 18–23) were linked to changes in convictions from one age group (e.g., 18–23) to the next (e.g., 24–29).

5.3.5. Cohort and period effects

Participants were born between 1979 and 1998. Three birth cohorts were created to help account for differences in the experience of incarceration over time (1980–1981, 1990–1991, and 1994–1995). These groups were designed to be similar to those used in other studies examining birth cohort differences in justice system outcomes. Over time, British Columbia Corrections made conscious efforts to shift both its youth and adult systems towards a greater focus on rehabilitation. This shift could influence how people experience and react to incarceration. The declining youth incarceration rate in British Columbia that began in the late 1990s is partly linked to changes in youth justice laws. Canada's Young Offenders Act (YOA) was in effect from 1984 until 2003, when it was replaced by the Youth Criminal Justice Act (YCJA). The study divided the sample based on whether participants were judged under the YOA (1998–2001) or YCJA (2005–2011). In line with RNR principles, the YCJA aimed to lower incarceration rates by only allowing custody sentences for young people whose offenses met certain severity criteria. Judges also had to explain why all other sentencing options were not suitable for improving public safety. The goal of incarceration under the YCJA was to promote public safety through rehabilitation. Unlike the YOA, the YCJA did not include specific deterrence (punishment discouraging future crime) among its sentencing principles. In summary, young people under the YCJA differ from those under the YOA, both in being at higher risk for reoffending and in their incarceration experiences, which may affect the relationship between incarceration and reoffending.

5.4. Analytic strategy

The study used a statistical method called a fixed-effect estimator to examine the impact of yearly, within-person changes in incarceration on yearly, within-person changes in convictions. By focusing on changes within each individual, fixed-effects estimation accounts for differences between people in unobserved factors that do not change over time and can cause selection bias. However, differences between individuals still exist regarding the number of days spent incarcerated and the number of convictions. These varying factors might be linked to unobserved, unchanging factors between people. For example, differences in ethnicity could be related to differences in incarceration length between individuals and could skew estimates of the relationship between incarceration and reoffending. These differences between people in varying factors can be removed by adjusting the data. Two ways to do this are "time-demeaning" and "first-differencing." Both are reliable methods that provide accurate estimates and reflect the degree and direction of within-person change rather than the size of differences between people. However, they adjust data using different calculations and can, therefore, give different answers to the same question. The choice of method should depend on how well it fits the theory or research question being studied.

Time-demeaning involves subtracting a person's average value across all measurement periods from their value at each period. While this approach is popular for removing differences between individuals in incarceration and produces unbiased estimates, it may be misleading from a theoretical standpoint. This is because time-demeaning uses information from the future to calculate whether someone changed in the present. For example, a person's value at a certain point in time is influenced by their values at later points, which does not reflect how people make decisions in the present based on past or current experiences, not future ones. Therefore, time-demeaning might lead to misinterpretations of the relationship between changes in incarceration and changes in offending, affecting conclusions about theories and policies.

First-differencing adjusts data by subtracting values from the previous period from values in the current period to examine changes over time. This approach better reflects how people experience change and how policies like increased incarceration to reduce reoffending are motivated. First-differencing also aligns with how courts apply increased sentencing for repeat offenders, where a person's incarceration time is increased compared to their previous sentence. It also captures the concept of labeling, where courts reinforce a person's "deviant" status by increasing their incarceration time compared to a previous punishment. By focusing on period-to-period change, first-differencing avoids assuming that long-term patterns in a person's life are always predictable.

The study found that adjusting the conviction and incarceration data using first-differencing or time-demeaning produced different results. For example, at age 14, using first-differencing, the sample showed an average increase of about 20 days in custody compared to age 13. In contrast, using time-demeaning, the sample showed an average decrease of about 30 days compared to the sample mean. These two approaches reflect different concepts and yield different values.

The main analysis used a first-differenced fixed-effect estimator for data covering ages 12–25. This analysis showed how yearly changes in incarceration (from two years ago to one year ago) contributed to yearly changes in convictions (from one year ago to the present). All analyses accounted for the effects of aging. The subtraction of each part of the equation removes all unobserved factors that do not change over time. What remains is the effect of differences in incarceration and the error term. Robust standard errors were used to address related measurements over time. Further analyses examined whether the relationship between incarceration and reoffending was consistent across different types of convictions and age groups. The main analyses were also repeated after dividing the sample to account for factors that might influence this relationship (e.g., gender, ethnicity, birth cohort, exposure to different justice system policies).

6. Results

Between ages 12 and 25, participants averaged about 20 convictions, including 2.5 violent convictions (e.g., assault, robbery, homicide), eight technical convictions (e.g., violating probation conditions), and 12 non-technical convictions. Participants spent nearly two years in custody during this period. Although recruited in adolescence, 80% of the sample spent at least one day incarcerated between ages 18–25. Among the 874 participants with data available between ages 30–35, 35% spent at least one day incarcerated. Out of 21,686 total observations over time for individuals, 629 (2.9%) were instances where a participant spent the entire observation period (i.e., a year of age) incarcerated. Among participants with complete data between ages 12–25 (1,410 participants), 21.2% (299 participants) spent at least one full year incarcerated. Overall, it was common for participants to be incarcerated in adulthood and remain in custody for relatively long periods.

6.1. Within-individual change models

The study found that for every additional day spent incarcerated from two years ago to one year ago, there was a 0.0063 unit decrease in the number of convictions from one year ago to the present. This means that an increase of one month in incarceration from two years ago to one year ago was associated with a 0.19 decrease in convictions from one year ago to the present. When the time lag was extended by an additional year, an increase in incarceration from three years ago to two years ago was also associated with a decrease in convictions from one year ago to the present. The study also considered the number of convictions per day spent free in the community. Consistent with other findings, increased time incarcerated was linked to fewer convictions in the next measurement period. Finally, an increase in incarceration from two years ago to one year ago was linked to a significant decrease in violent, non-technical, and technical convictions from one year ago to the present.

The study also looked at the effect of changes in incarceration on changes in convictions across different time intervals, between aggregated age groups, and within specific age groups. In all cases, increases in days incarcerated from two years ago to one year ago were associated with significant decreases in convictions from one year ago to the present. Notably, the reduction in convictions was particularly strong during adolescence (ages 12–17). For this age group, an increase of one month incarcerated from two years ago to one year ago was associated with a 0.40 decrease in convictions from one year ago to the present. For the three adult age groups, a one-month increase in incarceration was associated with about a 0.12–0.15 decrease in convictions.

Regardless of whether the sample was divided by gender or ethnicity, increases in the number of days incarcerated from two years ago to one year ago were consistently linked to significant decreases in convictions from one year ago to the present, for ages 12–25. Similarly, regardless of when participants were born or under which youth justice laws they were judged, increased time incarcerated was associated with significant decreases in convictions in the next measurement period. The overlapping confidence intervals suggest that the relationship between incarceration and reoffending was not stronger for any specific birth group or group exposed to particular laws.

6.2. Supplemental analyses

To see if similar findings were present among incarcerated young people in the United States, the same analytical approach was used with self-reported data on incarceration and varied offending over a seven-year period from boys and girls in Phoenix and Philadelphia. Similar to the Canadian data, yearly increases in time spent incarcerated from two years ago to one year ago were significantly associated with yearly decreases in self-reported varied offending from one year ago to the present. However, this crime-reducing effect of incarceration was observed for male participants but not female participants, and for White participants but not for Black or Hispanic participants.

7. Discussion

Theories and policies often describe the relationship between incarceration and reoffending as a matter of change within an individual. For instance, policies that increase sentences for repeat offenders aim to deter a person from future crime by increasing their time in custody compared to a previous sentence. Until now, research has mainly used group comparisons to determine if experiencing incarceration reduces a person's likelihood of reoffending. Within-individual methods more directly address these theoretical and policy questions by asking if a person's level of offending changes when their time in custody increases. Within-individual analyses also address concerns about selection bias by accounting for all unobserved factors that do not change over time. While these analyses do not account for unobserved factors that do change over time, this is also true for group comparison methods. An advantage over group strategies, like regression discontinuity designs, is that within-individual analyses address selection bias without having to focus only on individuals who barely met the criteria for a custody sentence.

This study examined the relationship between incarceration and reoffending using data from young people in British Columbia, Canada, who had been incarcerated. A specific statistical method was used to see if yearly increases in days spent incarcerated influenced yearly changes in convictions. Based on analyses between ages 12–25, yearly increases in time incarcerated were associated with yearly decreases in convictions in the next measurement period. An increase of one month in incarceration from two years ago to one year ago was linked to a 0.19 decrease in convictions from one year ago to the present. Changes in age were accounted for, meaning that the reduction in convictions was not just due to the natural decrease in reoffending risk that comes with aging into adulthood. The results were consistent across different types of convictions, including violent, non-technical, and technical convictions (e.g., violations of court orders). Increases in days incarcerated were associated with decreases in re-convictions even when the time lag between incarceration and re-conviction was extended by an additional year. This additional lag helped account for delays in court proceedings that could increase the time between an offense and a conviction. It is important not to interpret these findings as a reason for Canada's correctional system to maintain the current situation. Problems within Canada's correctional system persist and should not be ignored. The findings should not be seen as support for increasing the use of incarceration. In fact, the findings might reflect what happens when incarceration is reserved for a small group of young people involved in particularly serious or violent offenses.

Over three-quarters of the study participants were re-incarcerated in adulthood. This allowed for examining the relationship between incarceration and re-convictions at different age stages. A negative relationship between incarceration and re-convictions was observed when analyses were performed between combined age stages (e.g., the effect of changes in incarceration from ages 12–17 to 18–23 on changes in convictions from ages 18–23 to 24–29). This relationship was also observed when analyses were performed within specific age stages (e.g., ages 12–17 only; ages 30–35 only). However, the reduction in re-convictions following longer stays in custody was especially noticeable during adolescence (ages 12–17). Compared to its adult system, Canada's youth justice system places more emphasis on rehabilitation than deterrence, and young people sentenced to incarceration receive close supervision, monitoring, and services when they return to the community. While this is speculative, the stronger focus on rehabilitation in the youth system might explain why incarceration was more strongly linked to reductions in re-convictions. Adolescence is also a period with more opportunities for change, including responding to interventions. Differences in the strength of the incarceration effect might therefore be related to young people's greater capacity to respond positively to rehabilitation efforts. It should also be noted that participants who were not incarcerated during a specific age stage were excluded from the analysis for that stage. Thus, those who remained in the analyses of specific adulthood age stages were those with the most persistent patterns of offending. Such individuals might be less responsive to deterrent and/or rehabilitative efforts. While data limitations prevented a direct assessment of this possibility, this idea is consistent with findings that reductions in reoffending after incarceration were weaker for repeat offenders.

7.1. Analysis of potential moderators

The study's sample was divided into subgroups to assess whether findings were consistent across different groups. For both girls/women and boys/men in the study, yearly increases in the number of days spent incarcerated were associated with yearly decreases in the number of convictions in the next measurement period. Some research has suggested that incarceration has a particularly negative impact on women. However, this study found that incarceration reduced re-convictions for girls and women. This finding aligns with recent research where incarcerated women in Alberta, Canada, described their lives in custody as an improvement compared to their lives in the community.

The results were also consistent across White, Indigenous, and other racialized individuals. Custody facilities in British Columbia provide food, shelter, healthcare, and programs that are unavailable or difficult to access in the community. This does not suggest that incarceration is a positive experience for participants. The custody environment in British Columbia might simply be an improvement from severe forms of hardship experienced in the community. This could be particularly true for girls and Indigenous participants, who reported higher rates of crack cocaine and crystal methamphetamine use, family conflict, abuse, and unstable housing. This interpretation aligns with findings that higher-risk young people from a residential treatment facility experienced more days without self-injurious behavior compared to lower-risk youth from the same facility. Researchers suggested that higher-risk youth spent more time under treatment and supervision at the facility, which might have led to more days without self-injury after release. Incarceration might fail to reduce reoffending when the hardships in custody mirror the hardships experienced in the community. When similar analyses were performed using data from a U.S. study, increases in time spent incarcerated were associated with decreases in self-reported varied offending. However, after dividing the sample by demographics, the crime-reducing effect of incarceration was observed for White participants but not Black and Hispanic participants, and for male participants but not female participants. This suggests that prison conditions in Arizona and Pennsylvania might be less effective in reducing reoffending. Indeed, the United States has a history of incarcerating young people for minor offenses as a way to control behavior.

Previous research showed that the relationship between incarceration and reoffending might be influenced by the crime levels during the era in which a person grows up. Participants in this study who were born earlier came of age during periods with higher incarceration rates. Accordingly, the sample was divided into three birth cohorts. Regardless of when participants were born, yearly increases in the number of days spent incarcerated were associated with yearly decreases in the number of convictions in the next measurement period. Participants in this study experienced either Canada's Young Offenders Act (YOA) or Youth Criminal Justice Act (YCJA) legislation. The shift from the YOA to the YCJA was associated with a decline in incarceration rates and changes to the purpose and experience of incarceration, including an increased emphasis on rehabilitation and closer alignment with RNR principles. However, yearly increases in the amount of time spent incarcerated were associated with yearly decreases in the number of convictions regardless of which legislation participants were judged under. The consistent findings across birth cohort membership and legislative changes might indicate that Canada is in a "settled period" where it is difficult to see an independent influence of cultural factors on individual behavior. Policy changes in Canada happen more slowly, are less dramatic, and have less impact compared to policy changes in the United States that led to widespread incarceration. The YCJA may not have had a unique impact on the relationship between incarceration and reoffending compared to the YOA because incarceration rates in British Columbia were declining even before the YCJA was enacted.

7.2. Study implications, limitations, and future research

Without measures from before, during, and after incarceration, studies on the relationship between incarceration and reoffending are better at disproving certain theories than supporting them. For example, the study's results contradict the idea that prisons act as "schools of crime." However, like other studies, this one lacked data on participants' perceptions of punishment and measures of their incarceration experience. Therefore, it would be wrong to interpret the findings as direct support for specific deterrence (punishment discouraging future crime). Participants who experienced a period-to-period increase in days incarcerated might have still seen this situation as less punitive than other options. For instance, some people prefer custody to being bound by conditions of a community sentence (e.g., probation) for a longer time. To examine specific deterrence, future research should measure subjective perceptions of incarceration. Instead of reflecting deterrence, it is possible that longer periods of incarceration provided greater opportunities for participants to receive services that then led to fewer re-convictions. Research has shown that employment training programs influenced lower rates of reoffending among incarcerated individuals. Although incarceration experiences were not directly measured, findings were consistent across groups (e.g., different genders) who typically experience incarceration differently.

Future research should consider the two stages of incarceration: changes that occur during incarceration and changes in offending after release. Changes in the amount of time spent incarcerated might affect the types of individual changes that happen during incarceration. Understanding incarceration as a two-stage turning point requires measuring incarceration experiences, such as exposure to programs, visits from supportive contacts, perceptions of incarceration, criminal attitudes, and changes in substance use. Measures of post-incarceration experiences are also critical because, especially in jurisdictions using the RNR model, those considered higher risk are given more intervention and supervision resources. Reflecting on the current study's findings, increases in time spent incarcerated from one period to the next might indirectly represent increases in supervision, monitoring, and rehabilitation support received after returning to the community. Reductions in reoffending after increases in incarceration time might therefore actually reflect an effect that happens after release, where participants received more community supervision, monitoring, and services. Canada's youth system, in particular, reflects RNR principles, which might explain why the effects were stronger when examining incarceration's influence in adolescence compared to adulthood. This study is unaware of other research that separated the effect of supervision from the effect of incarceration. As weak evidence against the presence of a supervision effect, the study's findings were consistent even after increasing the time lag between incarceration and reoffending. In other words, re-convictions were lower even when looking beyond the initial periods of community reentry, when services, supervision, and monitoring are more likely to occur.

Based on a review of 116 studies, researchers suggested that there is a widespread agreement that incarceration does not reduce reoffending. The findings of this study do not necessarily contradict that literature. By focusing on yearly changes within a person, this study answered a different research question. The study also used data from a sample of participants who are likely higher-risk compared to samples in studies that focus on individuals on the edge of receiving a custody sentence. While not directly measured, the experience of incarceration for participants in Canada might differ from experiences in the United States, which is where most of the research on incarceration and reoffending is based. Canada is more likely to emphasize incarceration as an opportunity for rehabilitation and service delivery rather than deterrence and punishment. Compared to individuals on the margins of receiving a custody sentence in the United States, the study's participants might be less likely to experience the harshness of imprisonment because their experiences in the community are already especially challenging. Especially since the 1990s, Canadian correctional psychology researchers in policy positions have raised concerns about punitive aspects of incarceration, promoted effective correctional practices, and developed the RNR model to help identify who should be incarcerated and how to address treatment needs. These ideas might hold less weight in the United States, where there is a tendency to have negative views about whether higher-risk offenders can change. Such attitudes could lead to withholding rehabilitation services. For example, some researchers found that among people incarcerated in Illinois, incarceration reduced offending for first-time offenders but not for repeat offenders. Future research should examine whether these findings apply to individuals with psychopathic traits, who may be less likely to engage in and respond positively to treatment programs.

Like other studies, this study measured the number of days spent incarcerated within a specific period. Two participants might have spent the same number of days incarcerated in a given year, but one might have entered and left custody five separate times, while the other entered and left only once after a single sentence. The study data could not distinguish between such patterns, so it was not possible to examine the influence of sentence length on reoffending. The study also did not examine the specific way the relationship between incarceration and reoffending works. Long stays in custody might have decreasing returns in reducing reoffending, especially if such stays result in returning to the community at older ages when the risk of offending is naturally lower.

The use of incarceration reflects how a justice system functions, and the values, culture, rules, and traditions of this system differ across regions. This study does not suggest that incarceration decreases the risk of reoffending regardless of the region, era, and context in which it is used. The study's sample represents the types of individuals who repeatedly break the law and who probation and parole officers regularly interact with as part of case management. Despite the specific nature of this sample, as shown by the additional analyses using data from a U.S. study, the within-individual analytical strategy used in this study can be applied to a range of other samples and regions.

Summary

Policies in the justice system often suggest that if someone spends more time in jail or prison, they will commit fewer crimes later. This idea is about how one person's behavior changes over time. It's important to study this because many people worry about how well jail works, how much it costs, and how it affects people's health and well-being.

However, it has been hard to know if spending more time in jail truly lowers future crime. Most research looks at groups of people, comparing those who go to jail with those who don't. This type of study does not show if a single person commits fewer crimes after spending more time in jail. This study uses a new way to look at changes in a person's behavior over time to better answer these questions.

Another problem in research is not considering risk factors a person had before going to jail. These factors, like past crimes or drug use, can affect both going to jail and committing new crimes. Looking at how one person changes over time helps to deal with these risk factors that don't change.

1. Jail in Canada and the United States

Jail is used differently in various places and at different times. Most studies on jail and crime have been done in the United States. Canada's approach to young people and adults in trouble with the law is different. Sentences are usually shorter, more focused on helping the person, and less about politics.

For example, in British Columbia, where this study's information came from, jails are not run by private companies. The government provides health services. All people in jail get a mental health check within a day of arriving. They also have access to programs that help reduce future crime, like learning to build good relationships. Young people in jail can go to high school and get help with drugs, alcohol, mental health, and job skills. Many people in adult jails in British Columbia said they were happy with the services. Workers in the justice system get special training to help people better.

Canada's approach to jail has been different from the United States for a long time, especially from the 1990s to the 2010s. Canada focused less on strict punishment and more on plans to help each person, aiming to make jail less harsh and help people change. This approach is used for both young people and adults. For young people, jail is a last resort, meant to help them, not just scare them, and includes plans for when they return to the community.

In the United States, starting in the 1980s, the focus for young people shifted to punishment, not help. Some states even send young people to adult court. Canada does not do this; young people are never put in jail with adults.

Even though Canada and the United States have different ways of using jail, it's important to compare them at similar times. Much of the research in the United States used information from times when many people were in jail. How jail affects someone can depend on when they were born and what justice system rules were in place at that time. Most research has focused on solving the problem of comparing different groups of people.

2. Research on Jail and Crime

Research on jail and crime has tried different ways to compare groups of people fairly. This is important because many things, like age or past crimes, can affect both whether someone goes to jail and whether they commit new crimes.

Early research tried to control for these factors by using statistics. Later research tried to match people who went to jail with those who didn't, based on similar risk factors. These studies often found that jail either did not lower crime or even made it worse. However, these studies often did not have enough information to account for all risk factors. Because of this, these methods are thought to be less reliable.

Newer research tries to get around these problems by looking at specific rules in the justice system. For example, some studies compare people who just barely got a jail sentence with those who just barely did not. The idea is that these groups are very similar, so any difference in crime can be linked to jail. Other studies look at cases assigned by chance to different judges. These studies mostly found that jail either did not lower crime or made it worse.

However, a few studies, especially outside the United States, found that jail does reduce crime. While these newer methods are better at comparing groups fairly, they often only look at people who committed less serious crimes or had shorter criminal histories. This means their findings might not apply to everyone. We need ways to study how jail affects crime without leaving out people who are most likely to go to jail.

3. Looking at Changes within One Person

Old research methods compared groups of people who went to jail with those who did not. But the idea behind jail policies is usually about how one person changes. For example, some theories say that if a person's punishment gets harder, they will be less likely to commit crimes. Other theories say that if a person goes to jail, it can change how they see themselves and their connections to others, which might lead to more crime. Policies about repeat offenders also say that past punishment affects future punishment.

Looking at how one person changes over time fits with these theories and policies. This method also helps to deal with the problem of comparing groups unfairly, by looking at things that don't change about a person. Since each person is compared to their past self, there's no need to create groups that aren't realistic or only look at people with less serious crimes.

Three studies have looked at changes within one person over time. Two of these found that spending more time in jail was linked to fewer job chances and worse mental health. Another study found that spending more time in jail was linked to young people being less grown-up in their thinking. This new study adds to this research by looking at how changes in jail time for one person affect changes in their crime over time.

4. This Study

For nearly 100 years, people have studied how crime changes over time. But we still don't fully understand if going to jail changes these patterns. This is because it's hard to compare groups fairly. This study is not saying that looking at changes within one person is better than other methods. But it can help in three ways:

1. It helps to compare groups fairly.

2. It can look at all types of people, not just those with less serious crimes.

3. It can answer different questions to help us understand and improve theories and policies about jail and crime.

This study looked at how changes in the number of days spent in jail each year affected changes in the number of crimes committed each year. The information came from young people in British Columbia, Canada, who were followed into adulthood. The study looked at different types of crimes, like violent crimes or breaking court rules. It also looked at how stable these effects were over time. Since people might learn more about consequences as they get older, the study also looked at changes in different age groups.

Jail policies are different in different places and at different times. Most studies on jail and crime use information from the United States. This study used information from Canada, which has different ways of dealing with people in jail. It also used information from the United States to see if the results were similar there. Canada did not have a time of "mass incarceration," private jails, or common mandatory minimum sentences. Most past research used information from times when many people were in jail. How jail is run and experienced can affect crime. So, this study also looked at groups of people born in different times and who were under different youth justice system rules.

This study did not have all the information needed to fully explain why crime might go down. For example, it could not tell if crime dropped because people were scared to re-offend or because they got help in jail. It's rare for studies to measure the changes that happen to a person while in jail and after they return to the community. While this study could not look at all changing factors, it did look at how results differed for men and women and for different ethnic groups, since jail experiences can vary for these groups.

5. How the Study Was Done

5.1. People in the Study

The study included 1719 young people from jails in British Columbia, Canada. Very few young people refused to take part. Young people in jail in Canada are likely different from those in the United States. Canada doesn't jail young people for minor offenses, doesn't put them with adults, and doesn't have mandatory minimum sentences for young people. During the time this study gathered information (1998-2011), the number of young people in jail in British Columbia was much lower than in the United States. The young people in this study were more likely to have committed serious or violent crimes compared to young people in jail in the United States.

5.2. Study Steps

Researchers usually asked young people to join the study within the first week of being in jail. Some had been in jail before joining, and that time was counted. Young people were told that taking part would not affect their time in jail or their sentence. Interviews happened in a private room. They were given information about the study and told they could stop at any time and that their information would be private unless they threatened themselves or someone else.

Later information was collected using a computer system that holds records for all people involved with the justice system in British Columbia. This information was available for 1620 of the 1719 young people. The main part of the study looked at crimes and days in jail between ages 12 and 25. Over 95% of the people in the study were followed until at least age 25.

5.3. What Was Measured

5.3.1. Crimes

During the study, 142 people died, and 128 moved out of British Columbia. Jail time and crimes after these events were not counted. Crimes were counted for each year from age 12 (the age when someone can be charged in Canada) onward. The study looked at different types of crimes, like violent crimes or breaking court rules. If someone was convicted of a crime, they might go back to jail, which would limit their chance to commit more crimes. To account for this, the study adjusted the number of crimes based on how much time a person spent outside of jail each year. For example, if someone committed one crime at age 25 but spent half the year in the community, it was counted as two crimes to reflect more time being free.

5.3.2. Jail Time

Jail time included both time spent waiting for a court date and time spent serving a sentence. In British Columbia, the same jails are used for young people waiting for court and those serving sentences. For adults, there are separate jails, with more services for those serving sentences. This study could not tell the difference between these types of adult jail time. The study recorded when each person entered and left jail. The total days in jail were added up for each year of age. This way of counting accounted for the fact that young people often spend the last part of their sentence outside jail with rules to help them. This method avoided overcounting jail time by including early release. The study could not measure how many times a person went to jail each year or the length of each single jail stay. So, the study looked at the total time spent in jail, not the effect of single sentences. This way of measuring is similar to other studies.

The computer system included information about both federal (two years or more) and provincial (less than two years) jail sentences. For 18% of the people who had a federal sentence, if a release date was not clear, the date they could first be let out on parole was used as a guess for their release date.

5.3.3. Personal Details

Information about gender and ethnic background came from what people said in interviews. The group in this study was similar to the general group of young people in jail in Canada. It included 1341 males and 372 females, with 6 people missing gender information. Most people were White (54.7%). Indigenous people were a larger group in the study (30.1%) than in the general population of British Columbia. The rest (15.2%) were from different backgrounds and were grouped together because there were not enough people in each smaller group to study them separately.

5.3.4. Age Groups

Since the link between jail and crime might change with age, the study made four age groups: 12–17, 18–23, 24–29, and 30–35. In Canada, people aged 12–17 are handled by youth justice laws. Three more six-year age groups were made to compare evenly. The study looked at how changes in jail time and crimes happened each year within each age group. To see longer-term effects, the study added up total days in jail and total crimes for each age group. This helped to see if changes in jail time from one age group to the next were linked to changes in crimes from one age group to the next.

5.3.5. Birth Groups and Time Effects

People in the study were born between 1979 and 1998. They were put into three birth groups (1980–1981, 1990–1991, 1994–1995) to see how experiences in jail might have changed over time. Over the years, jails in British Columbia tried to focus more on helping people change. This shift might affect how people respond to jail. The number of young people in jail in British Columbia started to drop in the late 1990s, partly because of new youth justice laws. Canada's Young Offenders Act (YOA) was in place from 1984 to 2003, then replaced by the Youth Criminal Justice Act (YCJA). The study looked at people who were judged under the YOA (1998–2001) or the YCJA (2005–2011). The YCJA aimed to lower jail rates for young people and focused on helping them. Unlike the YOA, the YCJA did not aim to scare people from re-offending. Young people under the YCJA were different from those under the YOA, both in terms of being more likely to re-offend and how they experienced jail, which might affect the link between jail and crime.

5.4. How Information Was Studied

The study used a special method to look at how changes in jail time each year affected changes in crimes each year for the same person. This method helps to remove factors that don't change about a person and could unfairly affect results. However, there are still differences between people in how much jail time and how many crimes they have. These differences might be linked to other factors that don't change.

To deal with these differences, the study used a method called "first-differencing." This method looks at the change from one year to the next for each person. For example, if someone spent 14 days in jail one year and 28 days the next, first-differencing would show an increase of 14 days. The study believes this method better shows how people experience changes and how policies try to use more jail time to lower crime. It matches how courts might give a longer sentence based on a past sentence or how a person's "offender" status might be made stronger by more jail time.

Another common method, "time-demeaning," subtracts a person's average jail time from their jail time each year. But this method uses information from the future to calculate changes in the present, which might not match how people actually experience punishment. For example, if someone had a short jail stay followed by a very long one, time-demeaning might make the short stay look like a decrease in jail time, even though it was an increase from the previous year. This could lead to wrong ideas about why crime changes.

The study found that using first-differencing or time-demeaning gave different results when looking at changes in crimes and jail time. The study used first-differencing for its main analysis, looking at ages 12–25. This analysis looked at how changes in jail time from two years ago to one year ago affected changes in crimes from one year ago to now. The analysis also controlled for the effects of aging. Errors in the data were handled to ensure the results were reliable. The study also looked at whether the link between jail and crime was different for various types of crimes and age groups. Further analyses looked at differences based on gender, ethnic background, birth group, and different justice system laws.

6. Results

Between ages 12 and 25, people in the study committed about 20 crimes on average. This included about 2.5 violent crimes, 8 crimes for breaking court rules, and 12 other types of crimes. On average, they spent nearly two years in jail during this time. Even though they were young when they joined the study, 80% of them spent at least one day in jail between ages 18 and 25. For those with information available up to ages 30–35, 35% spent at least one day in jail. In total, 2.9% of the time, a person spent an entire year in jail. About 21.2% of people spent at least one full year in jail between ages 12 and 25. Overall, it was common for people in the study to go to jail as adults and stay there for fairly long periods.

6.1. Changes within One Person

The study found that when a person spent more days in jail from two years ago to one year ago, they committed fewer crimes from one year ago to now. Specifically, for every extra day spent in jail, there was a small decrease in crimes. This means an extra month in jail led to about 0.19 fewer crimes. This effect was still seen even when looking further back in time. Even when counting crimes based on how much time a person was free in the community, more jail time was linked to fewer future crimes. Also, an increase in jail time was linked to fewer violent crimes, fewer crimes for breaking court rules, and fewer other types of crimes.

The study also looked at how changes in jail time affected crimes at different age periods. In all cases, more jail time from two years ago to one year ago was linked to fewer crimes from one year ago to now. This effect was strongest during adolescence (ages 12–17). For this age group, an extra month in jail was linked to about 0.40 fewer crimes. For adults in the study, an extra month in jail was linked to about 0.12–0.15 fewer crimes.

No matter if the study looked at males or females, or different ethnic groups (White, Indigenous, or other racialized groups), more jail time was linked to fewer future crimes. Similarly, it did not matter when people were born or which youth justice laws they were under; more jail time was linked to fewer future crimes. The findings suggested that the effect of jail on reducing crime was similar across these different groups.

6.2. Extra Analyses

To see if similar findings occurred in the United States, the same study method was used with information from young people in Phoenix and Philadelphia. This study also found that more time in jail was linked to fewer different types of crimes reported by the people themselves. However, in the United States, this crime-reducing effect of jail was seen for males but not females, and for White people but not Black or Hispanic people.

7. Discussion

Theories and policies often talk about how jail affects one person's behavior over time. For example, repeat offender policies suggest that longer jail time should make a person less likely to commit crimes again. Until now, most research has used group comparisons to see if jail reduces crime. Looking at changes within one person helps to answer these questions more directly: Does a person commit fewer crimes after spending more time in jail? This method also helps to deal with the problem of comparing groups unfairly, by taking into account things about a person that don't change over time. While it doesn't account for all changes, it has an advantage over some other methods because it doesn't have to focus only on people with less serious crimes.

This study looked at how changes in jail time affected changes in crimes, using information from young people in British Columbia, Canada. The study found that when people spent more time in jail from one year to the next, they committed fewer crimes in the following year. An extra month in jail was linked to about 0.19 fewer crimes in the next year. This drop in crimes was not just because people got older. The results were the same for different types of crimes, even violent ones. The drop in crimes was also seen even when looking at a longer time between jail and new crimes. These findings should not be taken to mean that Canada's jail system is perfect or that more people should be jailed. The findings might show what happens when jail is used for a small group of young people who have committed serious crimes.

Most people in the study went back to jail as adults. This allowed the study to look at the link between jail and new crimes at different ages. A negative link was seen at all age stages, meaning more jail time was linked to fewer new crimes. However, this link was strongest for young people (ages 12–17). An extra month in jail for young people was linked to a larger drop in crimes than for adults. Canada's youth justice system focuses more on helping young people change and provides more support when they return to the community. This focus on help might explain why jail was more strongly linked to fewer new crimes for young people. Also, young people might be more likely to change after getting help. It's also possible that the people who remained in the study at older ages were those who committed crimes more often and might be harder to help.

7.1. Looking at Other Factors

The study also looked at whether the findings were consistent for different groups of people. For both girls/women and boys/men in the study, more jail time was linked to fewer new crimes in the following year. Some studies have said that jail has a very bad effect on women. But this study found that jail reduced new crimes for girls and women. This matches other research in Canada where women in jail said their life in jail was better than their life outside.

The results were also the same for White, Indigenous, and other racialized people. Jails in British Columbia provide food, shelter, health care, and programs that are hard to get outside. This does not mean jail is a good experience. It might just be that jail is better than the very hard lives some people, especially girls and Indigenous people, had outside of jail, where they faced more drug use, family problems, abuse, and unstable homes. This idea is supported by a study that found young people at higher risk of self-harm showed fewer self-harm behaviors when in a special treatment center.

Past research suggested that the link between jail and crime might depend on how much crime there was when a person grew up. People in this study who were born earlier grew up when more people were in jail. However, no matter when people in this study were born, more jail time was linked to fewer new crimes. Also, it didn't matter if people were under the older (YOA) or newer (YCJA) youth justice laws. Both laws led to a drop in jail rates and changes in how jail was used, with more focus on helping people. The consistent findings suggest that changes in laws in Canada might be slower and less dramatic than in the United States. The YCJA might not have had a unique effect because jail rates were already falling before it became law.

7.2. Study Meaning, Limits, and Future Research

Without knowing what happens before, during, and after jail, studies cannot fully prove why certain theories are correct. But they can show which theories are likely wrong. For example, these results go against the idea that jails are "schools of crime." However, like other studies, this one did not have information on what people thought about their punishment or what their jail experience was like. So, it cannot directly prove that jail scared people from committing more crimes. It's possible that people who spent more time in jail saw it as less harsh than other options, or that longer jail stays gave them more chances to get help. Studies should measure what people actually think about jail.

Future research should look at two stages of jail: changes that happen while in jail and changes in crime after leaving jail. More time in jail might lead to different changes during jail. To understand jail as a turning point, studies need to measure things like access to programs, visits from family, thoughts about jail, and changes in drug use. What happens after jail is also important, especially where programs and supervision are given more to people who are seen as higher risk. The study's finding that more jail time led to fewer crimes might actually mean that people got more supervision, monitoring, and help after leaving jail. This might explain why the effect was stronger for young people, as Canada's youth system focuses more on help. We don't know of other studies that separate the effect of supervision from the effect of jail itself. But the study found that crime rates were lower even when looking beyond the first periods after leaving jail, when services are most likely given.

A review of many studies suggested that there is a general agreement that jail does not reduce crime. This study's findings do not necessarily disagree with that, but it answered a different question by looking at changes within one person year by year. It also looked at people who are likely at higher risk compared to other studies that focus on people with less serious crimes. The experience of jail for people in this study in Canada might be different from experiences in the United States, where most of the research is from. Canada tends to see jail as a chance for help and services, not just punishment. People in this study might also feel the harshness of jail less because their lives outside jail were already very hard. Researchers in Canada have also promoted effective ways to help people in jail and developed methods to identify who needs help. These ideas might be less strong in the United States, where there is often a belief that high-risk offenders cannot change, which might lead to less help being given. Future research should look at whether these findings apply to people who might not respond well to treatment.

Like other studies, this study measured the total number of days spent in jail within a year, but not how many times a person went to jail or the length of each single sentence. So, it could not look at how the length of a sentence affects crime. It also did not look at whether very long stays in jail have less and less effect on reducing crime, especially if people return to the community at older ages when they are less likely to commit crimes anyway.

How jail is used shows how a justice system works, and systems differ across places. This study does not suggest that jail reduces crime everywhere, at all times, or in every situation. The people in this study were those who often commit crimes and have regular contact with probation and parole officers. Despite the specific nature of this study's group, the method used here can be applied to many other groups and places.