Skip to Content
MDPIMDPI

Advanced

SocietiesSocieties
PDF
  • Article
  • Open Access

4 August 2026

16 Pages

Comparative Analysis of Social Reintegration and Criminal Recidivism in the Spanish Civilian and Military Prison Systems: A Quantitative Study with Sensitivity Analysis

,
and
1
International Doctoral School (EINDOC), Technical University of Cartagena, 30202 Cartagena, Spain
2
Department of Psychology and Sociology, Faculty of Education, University of Zaragoza, 50009 Zaragoza, Spain
3
Department of Aerospace Science and Technology, University Centre of Defence at Spanish Air and Space Force Academy, 30720 San Javier, Spain
*
Author to whom correspondence should be addressed.
Societies2026, 16(8), 247;https://doi.org/10.3390/soc16080247
Version Notes
Order Reprints

Abstract

This study comparatively analyzes the effectiveness of social reintegration in the Spanish civilian and military prison systems, using re-imprisonment recidivism as an inverse proxy indicator. Official administrative data for the 2009–2019 period were analyzed for the civilian system (n = 19,909) and the military system (n = 644), the latter comprising 539 released individuals with complete outcome data and 105 transferred to the civilian system whose post-release outcomes were unavailable. To address this missing information, a sensitivity analysis was implemented under three transparent, a priori scenarios (conservative, intermediate, and optimal). The civilian reintegration rate was 80.02%, whereas the military rate ranged from 95.19% to 98.14% across scenarios. The military advantage was substantial and consistent (absolute risk difference +15.17 to +18.13 percentage points; all p < 0.001; large effect sizes). However, the two populations differ markedly in prior criminal history, offense type, and risk/protective factor profiles; the observed gap therefore reflects the combined effect of population selection and institutional factors and should be interpreted as descriptive rather than causal. These findings establish a robust empirical phenomenon and generate testable hypotheses for future comparative and quasi-experimental research on penitentiary effectiveness.

1. Introduction

Social reintegration after imprisonment is one of the central challenges facing contemporary criminal justice systems. In Spain, Article 25.2 of the 1978 Constitution establishes that custodial sentences must be oriented toward the re-education and social reintegration of convicted persons [1]. Yet the empirical evidence on whether existing prison systems achieve this objective is heterogeneous, context-dependent, and methodologically contested [2,3,4]. Understanding which institutional characteristics promote successful reintegration and for which populations is therefore a problem of direct policy relevance.
The core problem addressed by this study is the following: Spain operates two legally distinct prison systems, the civilian system managed by the General Secretariat of Penitentiary Institutions (Secretaría General de Instituciones Penitenciarias, SGIP), housing approximately 58,000 inmates across 68 facilities, and the military system embodied in the Military Penitentiary Establishment (Establecimiento Penitenciario Militar, ESPEMI) in Alcalá de Henares [5], yet no rigorous empirical comparison of their reintegration outcomes has ever been conducted. This gap is problematic for several reasons. First, it leaves policymakers without evidence on whether the substantially higher per-capita resources invested in the military system (ratio 1:3 vs. 1:8–10 in the civilian system) translate into better outcomes. Second, it precludes informed decisions about transferring effective programmatic components between systems. Third, it leaves the question of whether the two populations—which differ markedly in sociodemographic composition, type of offense, and risk/protective factor profiles—are served appropriately by their respective institutional environments unaddressed.
Two clarifications about the scope and limitations of this comparison are necessary from the outset. First, the two populations are not equivalent: military personnel who enter ESPEMI have passed entry filters requiring minimum educational attainment, absence of a prior criminal record, and demonstrated occupational competence. This creates a population with structural advantages relative to the average civilian prison population. Second, the types of offenses are substantially different: military infractions frequently involve disciplinary or institutional violations (insubordination, abandonment of post, abuse of authority), whereas civilian offenses span a broader and often more severe criminological spectrum. Both of these differences affect recidivism prognosis independently of any institutional treatment effect. This study is therefore not a controlled comparison; it is a descriptive comparative analysis whose findings must be interpreted with explicit attention to these population-level differences.
Given these caveats, the study pursues three complementary objectives: (1) to quantify and compare reintegration rates across both systems using rigorous inferential statistics appropriate for unequal samples; (2) to assess the robustness of the findings through a sensitivity analysis addressing the uncertainty arising from 105 cases transferred between systems; and (3) to situate the findings within the theoretical literature on desistance, risk/protective factors, and penitentiary effectiveness, identifying both the potential contribution of programmatic factors and the methodological limitations that prevent causal attribution.
The scientific contribution of this study is threefold. Empirically, it provides the first systematic quantitative comparison of the Spanish military and civilian prison systems, filling a documented gap in the national and international literature on specialized custodial institutions [6]. Methodologically, it demonstrates the appropriate use of sensitivity analysis under missing-data conditions and the interpretation of effect sizes in contexts of structural population non-equivalence. Theoretically, it situates the findings within a multi-theoretical framework—drawing on life-course theory, desistance theory, the Risk–Need–Responsivity (RNR) model, and social capital theory—to offer a nuanced account of what the observed differences can and cannot tell us about effective reintegration practice.

2. Theoretical Framework

2.1. Defining Recidivism: Conceptual and Typological Clarification

Recidivism is conceptually complex: it encompasses a heterogeneous set of behaviors ranging from technical violations of parole conditions to the commission of serious new criminal offenses, and its legal definition varies substantially across jurisdictions [2,7]. For the purposes of this study, recidivism is operationally defined in its strictest legal and behavioral sense: the commission of a new offense resulting in a final judicial conviction and re-entry into a custodial institution (civilian or military) within the established follow-up period (2009–2019). This definition—sometimes called “re-imprisonment recidivism”—is the most conservative and legally stringent criterion and is preferred in comparative studies because it relies on official records rather than self-reported behavior [2].
This operational definition excludes other forms of recidivism discussed in the literature: re-arrest without conviction, re-conviction without re-imprisonment, and technical violations such as failure to report to a probation officer. It also excludes self-reported offending not captured in official records. These exclusions have implications for the interpretation of rates: the 19.98% civilian recidivism rate reported here captures only the most serious end of the recidivism spectrum, and the true rate of any reoffending (including undetected or unconvicted behavior) would necessarily be higher. The same logic applies to the military system. These limitations are acknowledged explicitly, and their implications for interpretation are discussed in Section 5.
A further conceptual distinction is necessary between recidivism and desistance. Desistance refers to the process by which individuals gradually reduce and eventually cease criminal behavior over the life course [3,8]. Unlike recidivism (a binary outcome at a fixed point in time), desistance is a dynamic process influenced by turning points, identity change, and accumulating social bonds. The reintegration rates reported in this study capture a cross-sectional snapshot—non-recidivism over a defined period—which may underestimate long-term desistance for individuals who reoffend during the study window but subsequently desist. Future research using longitudinal trajectories would provide a more complete picture of post-release outcomes.

2.2. Risk and Protective Factors in Reintegration

A substantial body of evidence documents that reintegration outcomes are shaped by two broad categories of factors: static risk factors (which cannot be modified by intervention) and dynamic risk/need factors (which can be targeted by treatment). The most robustly supported static factors include age at first offense and age at release (younger = higher risk), prior criminal history, and gender (male sex is associated with higher recidivism) [2,9]. Dynamic criminogenic needs with the strongest evidence base include antisocial cognitions and attitudes, antisocial peer associations, substance abuse, and deficits in family/marital bonds (the “Central Eight” risk factors [9,10]).
Protective factors operating at the individual, relational, and structural levels further moderate recidivism risk. At the individual level, these include educational attainment, vocational skills, self-efficacy, and the construction of a prosocial identity [3,11]. At the relational level, they include quality of family ties, social support networks, and mentoring relationships [8,12]. At the structural level, these factors encompass access to stable employment, adequate housing, and community integration resources [13,14].
The populations entering the civilian and military systems differ systematically on virtually all of these dimensions. Military personnel who enter ESPEMI typically possess higher educational attainment (compulsory secondary education is an entry requirement), have no prior criminal record at the time of first military offense, demonstrate lower rates of substance abuse (screened at entry), and retain stronger institutional affiliations. Upon release, the 83% who return to active duty automatically access stable employment, salary, and social benefits—structural protective factors that the life-course literature consistently identifies as turning points promoting desistance [8]. Conversely, civilian prison populations in Spain are disproportionately characterized by low educational attainment, histories of substance abuse, precarious housing, prior criminal records, and limited post-release employment prospects [15]. These structural differences are fundamental to the interpretation of any observed outcome differences.
This study cannot statistically control for these population-level differences because individual-level data are not available—only system-level aggregates. This limitation is acknowledged as the central constraint on causal inference and is addressed in the discussion through a disaggregated theoretical analysis of plausible effect contributions.

2.3. The Risk–Need–Responsivity Model and Its Application to Both Systems

The Risk–Need–Responsivity (RNR) model [9] is the most empirically supported framework for designing effective correctional interventions. Its three core principles are as follows: (1) Risk principle: match intervention intensity to the individual’s risk level; high-risk individuals benefit most from intensive programs, while low-risk individuals may be harmed by intensive intervention. (2) Need principle: target dynamic criminogenic needs, not static background characteristics. (3) Responsivity principle: use cognitive–behavioral approaches matched to the individual’s learning style, motivation, and strengths. Recent survey evidence nonetheless documents a persistent gap between these principles and their implementation in routine correctional assessment practice [16].
Meta-analytic evidence consistently shows that programs adhering to RNR principles reduce recidivism by 25–35% relative to control conditions [9,17,18,19,20]. Specific components with the strongest evidence include cognitive–behavioral programs (25–35% reduction), educational interventions (13–43% reduction), and vocational training linked to real employment opportunities (approximately 13% reduction) [21,22,23].
The military system presents several characteristics aligned with RNR principles: a relatively homogeneous low-to-moderate-risk population (enabling matched intervention intensity); educational programs structured through agreements with the National University of Distance Education (UNED); the INTEGRA-T labor intermediation program for post-release employment; favorable staff-to-inmate ratios (approximately 1:3) enabling individualized attention; and a disciplinary-institutional culture that may reinforce the cognitive-behavioral component of prosocial norm adherence. Whether these elements causally drive the observed outcomes—or whether the population’s pre-existing protective factor profile is the dominant driver—cannot be determined from available data. Both mechanisms likely operate simultaneously, and future research designs would need to disentangle their relative contributions.

2.4. Life-Course Theory, Social Bonds, and Desistance

Life-course criminology [8,14] provides a structural framework for understanding why certain institutional exits promote desistance. The theory posits that informal social bonds—attachment to family, commitment to employment, and involvement in community institutions—progressively accumulate across the life course and reduce the probability of offending through mechanisms of social control, investment in conventional activities, and reputational stake [24]. Critical turning points that disrupt criminal trajectories include marriage, stable employment, military service, and educational attainment [25].
Maruna’s [3] narrative identity theory complements this structural account by identifying the psychological mechanism through which turning points operate: the construction of a “redemption script” through which desisting individuals reinterpret their past and project a prosocial future self. For military personnel returning to active duty, institutional re-entry may facilitate such narrative reconstruction—the offense is reframed as a temporary deviation from an otherwise disciplined professional identity, now restored. This mechanism is not available to civilian releases who face stigma, unemployment, and social exclusion.
Cullen’s [12] social support theory further emphasizes that communities of support—families, peers, institutions—buffer against recidivism by providing material resources, emotional sustenance, and supervisory oversight. The military institution, by reintegrating released personnel into a structured organizational environment with peer accountability, may function as an unusually potent source of post-release social support unavailable to civilian releases [26].

2.5. International Evidence on Military Prison Systems

Empirical research specifically on military prison systems is sparse. The available comparative evidence, concentrated mainly in the United States and United Kingdom, suggests that military populations present a distinctive criminological profile: higher initial human capital, lower rates of prior criminality, greater institutional socialization to discipline and hierarchy, and lower substance abuse prevalence [27,28,29,30,31]. Simultaneously, military populations exhibit specific risk factors absent in civilian populations: combat-related trauma, post-traumatic stress disorder, alcohol misuse associated with organizational culture, and adjustment difficulties upon transition to civilian life [28].
The limited comparative evidence suggests that military systems emphasizing rehabilitation through structured programs achieve better outcomes than those prioritizing punitive approaches; however, the absence of rigorous longitudinal designs with adequate comparison groups and extended follow-up limits causal conclusions [6]. The present study contributes to this sparse literature by providing systematic data on the Spanish case within the same national, legislative, and cultural context—a within-country design that reduces cross-national confounders.

3. Materials and Methods

3.1. Study Design

This study employs a retrospective comparative observational design using system-level aggregate data for the period 2009–2019. The choice of a cross-sectional comparative design with natural (non-randomized) groups reflects the impossibility of experimental assignment—individuals enter either system according to the nature of their offense and their legal status, not by random allocation. This design choice enables the documentation of outcome differences but precludes definitive causal attribution of those differences to specific programmatic or population factors [32].
The period 2009–2019 was selected because it corresponds to the only systematic follow-up cohort study available for the civilian system (see Table 1) with sufficient methodological rigor to permit valid comparison [15] and because it coincides with the period for which military jurisdiction yearbooks provide complete data [33]. The end date of 2019 is partly attributable to data availability constraints; it also predates the institutional disruptions caused by the COVID-19 pandemic (2020–2022), which introduced atypical conditions in both systems that would complicate comparison.
Table 1. Civilian system sample: reintegration outcomes (2009–2019).
Regarding developments between 2019 and the present date (2026), several changes are relevant to contextualizing these findings. At the legislative level, Organic Law 1/2026 introduced reforms regarding persistent recidivism (multirreincidencia), potentially affecting civilian sentencing trajectories for habitual offenders [34,35]. At the institutional level, the civilian prison population has undergone demographic shifts, with increasing proportions of foreign nationals and changing offense profiles. At the military level, the ESPEMI (Establecimiento Penitenciario Militar) population has continued to decline in size (as shown in Table 2), reflecting changes in Armed Forces structure and in military justice application. These post-2019 developments are discussed in Section 5 in terms of their implications for the transferability of current findings but are not incorporated into the quantitative analysis, which strictly covers the 2009–2019 cohort.
Table 2. Military system sample: releases and transfers (2009–2019).

3.2. Data Sources

Civilian system data derive from the Study on Penitentiary Recidivism in Spain published by the SGIP (Secretaría General de Instituciones Penitenciarias) [15], which reports cohort follow-up of individuals released from civilian prisons during the 2009–2019 period. This is the only nationally representative longitudinal study on civilian recidivism in Spain with publicly available methodology, enabling replicability. Military system data are derived from the administrative records of ESPEMI and from the Yearbooks of Military Jurisdiction published by the Ministry of Defence for 2009–2019 [33], which document the number of inmates released, inter-system transfers, and documented re-entries following new convictions within the military justice system. Re-entries into the civilian system for individuals originally processed through ESPEMI were verified through cross-reference with SGIP records provided by MINISDEF (Ministerio de Defensa).

3.3. Samples and Population Characteristics

3.3.1. Civilian System Sample

The civilian sample comprises n = 19,909 individuals released from civilian prisons over the 2009–2019 period. According to the SGIP [15] report, the civilian prison population during this period was predominantly male (approximately 92%), with a mean age at release of approximately 38 years. The most prevalent offense categories were property crimes (approximately 35%), drug trafficking (approximately 25%), crimes against persons (approximately 18%), and sexual offenses (approximately 8%). Approximately 40% had prior imprisonment, and approximately 60% had documented substance use problems. Educational attainment was low: approximately 55% had not completed compulsory secondary education. These characteristics profile a population with multiple dynamic criminogenic needs and limited protective factor endowment—conditions associated in the literature with elevated recidivism risk [9].

3.3.2. Military System Sample

The military sample comprises n = 644 individuals released from ESPEMI over the same period (Table 2). The military population is exclusively male (female military personnel convicted of offenses punishable by sentences exceeding brief periods are exceptionally rare in Spain), with a mean age at release estimated at 32–36 years based on typical military service trajectories. All individuals in this sample had passed the entry filters of the Armed Forces or the Guardia Civil at the time of recruitment, implying completion of compulsory secondary education, no prior criminal record, successful completion of physical and psychometric screening, and absence of documented drug dependency. The predominant offense categories were military disciplinary infractions (insubordination, abandonment of post, and abuse of authority, approximately 60%); crimes against persons within a military context (approximately 20%); gender-based violence (approximately 10%); and property and other offenses (approximately 10%). The offense profile is thus qualitatively different from the civilian system and carries a different recidivism prognosis.
The military sample divides into two analytically distinct subgroups. The first subgroup (n = 539) comprises individuals who completed their sentence at ESPEMI and were released, for whom complete outcome data are available (Table 3). The second subgroup (n = 105) comprises individuals transferred to civilian facilities because their sentence exceeded three years (which under Spanish military law entails loss of military status; Laws 39/2007, 8/2006, and Organic Law 2/1986), for whom post-release outcomes in the civilian system are not available in military records.
Table 3. Military subgroup with complete outcome data (ESPEMI completers, n = 539).

3.3.3. Comparative Population Profile and Analytical Implications

Table 4 summarizes the key sociodemographic and criminological differences between the two populations. These differences are fundamental to the interpretation of outcome comparisons and are discussed systematically in Section 5.
Table 4. Comparative population profiles: civilian vs. military system.

3.4. Operationalization of Variables

The dependent variable is successful social reintegration, operationally defined as the absence of criminal recidivism during the follow-up period. Recidivism is coded dichotomously: 1 = successfully reintegrated (no re-entry into any custodial institution following a new final conviction); 0 = recidivated (re-entry into the prison system following a new final conviction within the 2009–2019 window). This strict operational definition excludes re-arrests without conviction, supervision violations without new offenses, and self-reported offending [38,39].
It is important to note that the follow-up window is not uniform across individuals: those released earlier in the 2009–2019 period have longer observation periods than those released later. This heterogeneity in follow-up time is an acknowledged limitation; ideally, survival analysis with time-to-event outcomes would be employed, but the available aggregate data do not permit individual-level time-to-event analysis. The binary outcome captures whether any recidivism occurred within the study window, not its timing.

3.5. Handling Missing Data: Sensitivity Analysis Design

The 105 individuals transferred from ESPEMI to civilian facilities represent 16.3% of the total military sample. Their absence from military records creates a missing-data problem that, if handled naively (e.g., by simply excluding them), could bias estimates of the military system’s overall effectiveness in either direction [40]. To address this problem transparently, three sensitivity scenarios were specified a priori, each reflecting a different assumption about the likely outcomes of transferred individuals [41]:
Scenario 1 (Conservative): Transferred individuals achieve the civilian system reintegration rate (80.02%), reflecting complete convergence of outcomes after transfer. This represents the most pessimistic estimate for the military system (lower bound).
Scenario 2 (Intermediate): Transferred individuals achieve a rate midway between the ESPEMI completer rate (98.14%) and the civilian rate (80.02%), yielding 89.08%. This reflects partial carry-over of protective factors acquired during ESPEMI detention.
Scenario 3 (Optimal): Transferred individuals achieve the same rate as ESPEMI completers (98.14%), reflecting full persistence of ESPEMI-acquired protective factors. This represents the most optimistic estimate (upper bound).
The convergence of findings across all three scenarios provides the basis for the study’s main conclusions about the robustness of observed differences.

3.6. Statistical Methods

For each scenario, the following statistics were computed: (1) point estimates of reintegration proportions for each system and scenario; (2) 95% confidence intervals using the Wilson method (preferred over the normal approximation for proportions near 0 or 1); (3) a z-test for the difference of proportions for independent samples (H0: p1 = p2; α = 0.05, two-tailed); (4) the chi-square test of independence (2 × 2 contingency table) as a cross-validation of the z-test; (5) absolute risk difference (RD) with 95% CI as the primary effect size measure; (6) relative risk (RR) as a supplementary effect size. All analyses were conducted in R (version 4.3.2) (R Foundation for Statistical Computing, Vienna, Austria) using the stats and PropCIs packages.

3.7. Ethical Considerations

The study uses exclusively aggregated and anonymized data from official administrative records. No personal identifiers were used or accessed. Ethics committee approval was not required. The study complies with Organic Law 3/2018 on the Protection of Personal Data and the Guarantee of Digital Rights. Recidivism data for the military system were verified by the Ministry of Defence (MINISDEF) through consultation of ESPEMI databases and cross-referencing with records of re-entries within the study period.

4. Results

4.1. Civilian System: Descriptive Statistics

Of the 19,909 individuals followed from the civilian system, 15,931 did not reoffend (reintegration rate: 80.02%; 95% CI: 79.46–80.57%), and 3978 recidivated (recidivism rate: 19.98%). These rates are consistent with findings from comparable European jurisdictions using the same operational definition of recidivism [42,43].

4.2. Military System: ESPEMI Completers (n = 539)

Of the 539 individuals who completed their sentence at ESPEMI, 529 did not reoffend (reintegration rate: 98.14%; 95% CI: 96.62–98.99%) and 10 recidivated (recidivism rate: 1.86%). Table 5 presents detailed information on the 10 recidivist cases, illustrating the nature and timing of re-offenses. As is evident from inspection, the re-offenses are predominantly military disciplinary infractions and procedural violations rather than serious criminal offenses—a qualitative difference from civilian recidivism patterns that further complicates direct comparison.
Table 5. Recidivist cases in the ESPEMI completer subgroup (n = 10 individuals, 20 offenses).

4.3. Sensitivity Analysis Results

Table 6 presents reintegration rate estimates under each sensitivity scenario for the total military sample (n = 644), alongside inferential test results.
Table 6. Sensitivity analysis results: military system overall reintegration rates and comparative tests.
Across all three scenarios, the military system demonstrates a statistically significant reintegration advantage over the civilian system (p < 0.001). The absolute difference ranges from +15.17 (conservative) to +18.13 (optimal) percentage points—a magnitude classifiable as a large effect by conventional standards [44], implying 15 to 18 additional successful reintegrations per 100 individuals processed in the military system relative to the civilian system.

5. Discussion

5.1. Analysis of the Civilian System

The civilian system’s 80.02% reintegration rate situates Spain within the European mid-range—above England/Wales (52%) [45] and France (60%) [46], and comparable to Germany (65%) and Norway (80%)—for studies using the same strict operational definition of re-imprisonment recidivism [42,43]. This performance occurs against a background of significant challenges: a prison population with a high prevalence of substance use disorders, low educational attainment, and precarious post-release prospects [47]. That 80% of civilian releases do not return to prison within the study window is, in absolute terms, a considerable achievement, though it must be contextualized alongside the known limitations of re-imprisonment rates as a measure of full social reintegration [48].
The civilian system’s limitations are structural: resource constraints produce staff-to-inmate ratios of approximately 1:8–10, limiting individualized programming. Overcrowding affects program delivery quality. Post-release support services are unevenly distributed geographically [49]. The RNR model predicts that improving the match between risk level and intervention intensity—particularly for the highest-risk 20% of releases—would generate the largest marginal gains in reintegration outcomes [9,19].

5.2. Analysis of the Military System

The military system’s reintegration rate of 98.14% in the ESPEMI completer subgroup represents an exceptional outcome by any international standard. Understanding this result requires disaggregating population and institutional contributions.
Population-level contributions are substantial. The absence of prior criminal records, higher educational attainment, lower substance abuse prevalence, and predominantly institutional/disciplinary offense types all predict lower recidivism independently of any treatment effect. Furthermore, the 83% of completers who return to active military or Guardia Civil duty automatically access the three structural protective factors most robustly associated with desistance: stable employment, prosocial institutional affiliation, and daily routine structure [8,14]. These factors would likely produce low recidivism regardless of the quality of ESPEMI’s treatment programs.
Institutional contributions are plausible but difficult to quantify from available data. The favorable staff-to-inmate ratio enables individualized assessment and programming. The UNED educational agreement provides continuity of academic credentials. The INTEGRA-T program specifically targets the labor reintegration needs of personnel who lose military status. These elements align with RNR principles and with the components shown to be most effective in meta-analyses [18,22]. However, because the population entering ESPEMI is already at low recidivism risk, the RNR model actually predicts that intensive programming for this group would yield only modest incremental gains beyond the baseline provided by their protective factor profile.
The 10 recidivist cases (Table 5) are themselves instructive. The re-offenses are predominantly military infractions, procedural violations, and relational offenses rather than the property crimes, drug offenses, and violent crimes that predominate in civilian recidivism. This qualitative difference suggests that even among recidivists, the offense severity and social disruption are substantially lower in the military context—a dimension not captured by the binary recidivism measure employed here.

5.3. Population Non-Equivalence: Implications for Causal Interpretation

The central interpretive challenge of this study is that the two populations cannot be considered equivalent in either observable or unobservable dimensions. Observable differences (Table 4) include prior criminal history, educational attainment, offense type, substance abuse prevalence, and post-release employment prospects. Unobservable differences may include intrinsic motivation, psychological resilience, prosocial value orientation, and quality of family support—variables that predict both lower recidivism and a higher likelihood of military career entry.
In the absence of individual-level data permitting propensity score matching, regression discontinuity designs, or instrumental variable analysis, it is impossible to partition the observed outcome difference into the proportion attributable to population selection versus institutional treatment effects [50,51]. The most defensible interpretation of the fifteen to eighteen percentage point difference is therefore the following: it represents the combined effect of (a) the military population’s substantially more favorable risk/protective factor profile, and (b) any genuine institutional effectiveness advantage of the ESPEMI environment. The relative contribution of each component cannot be determined from these data.

5.4. Contextualizing Findings in the International Literature

The Spanish military system’s performance compares favorably with all international reference points in the recidivism literature, including Scandinavian systems widely regarded as global benchmarks [43,52]. This finding is consistent with evidence that population selection effects, when sufficiently powerful, can produce outcome differences that exceed those generated by even the most intensive treatment programs [9,53]. It also raises the question—empirically unanswered by this study—of whether civilian programs of comparable quality and intensity could achieve similar outcomes for population segments with equivalent risk profiles.

5.5. Limitations

Five categories of limitations merit explicit acknowledgment. First, the observational design without random assignment prevents causal inference about the sources of the observed differences [54]. Second, the absence of individual-level data prevents multivariate adjustment [55] for observable confounders—an especially significant constraint given the substantial population non-equivalence documented in Table 4. Third, the 105 transferred cases introduce irreducible uncertainty about the military system’s overall effectiveness, partially addressed but not eliminated by the sensitivity analysis [56]. Fourth, the binary recidivism measure captures only re-imprisonment, omitting broader dimensions of social reintegration (employment, housing, family, community participation) and less serious forms of reoffending. Fifth, the 2019 data endpoint means that the study does not capture the post-pandemic period or the effects of recent legislative reforms [35].
The seven-year gap between the end of the study period (2019) and the present (2026) introduces potential limitations on the contemporary relevance of these findings. Key developments include the introduction of Organic Law 1/2026 on persistent recidivism, which creates new sentencing modalities that may alter the civilian recidivism trajectory for habitual offenders; the restructuring of the Spanish Armed Forces, which has further reduced the ESPEMI population (from approximately 80–86 annual releases in 2009–2013 to 22–36 in 2016–2019, as shown in Table 2), raising questions about the continued sustainability and representativeness of the military system; and post-pandemic disruptions to both systems’ educational and vocational programming (2020–2022) that may affect cohorts not captured in this study. Updated comparative data for 2020–2025 would be needed to assess whether the patterns documented here persist under current conditions.

5.6. Practical Applications

A legitimate question—what is the value of this study if its findings cannot be used as causal evidence?—deserves a direct answer. Descriptive comparative evidence of the kind produced here has several distinct and legitimate uses even in the absence of causal identification:
First, it establishes the existence and magnitude of a robust empirical phenomenon. The consistency of the military system’s advantage across three sensitivity scenarios (all p < 0.001, all large effect sizes) confirms that this is not a measurement artifact or statistical noise—it is a real and substantial difference that demands explanation. Establishing the phenomenon is the necessary precondition for designing the studies that will explain it.
Second, it provides the quantitative foundation for resource allocation and policy prioritization decisions. Even without knowing the causal mechanism, policymakers can use the finding that the military system produces approximately 15–18 additional successful reintegrations per 100 individuals to justify investment in studying and potentially replicating its conditions. Cost–benefit analysis of current SGIP data [37] suggests that a conservative 5 percentage-point reduction in civilian recidivism would save approximately €80 million annually in incarceration costs alone.
Third, it generates specific hypotheses for future causal research. The theoretical analysis in Section 5.3 identifies the candidate mechanisms (population selection, post-release employment, and institutional treatment) and specifies the data and designs needed to test each. A study that could neither document the phenomenon nor generate testable hypotheses would have less scientific value, not more. Observational comparative studies are a standard and methodologically legitimate first step in a cumulative research program [32,57].
The findings generate a specific, actionable research agenda. Priority studies include (1) individual-level administrative data linkage studies that merge ESPEMI and SGIP records with Social Security employment data and National Statistics Institute outcome data, enabling propensity score matching and multivariate survival analysis [58,59]; (2) quasi-experimental analyses exploiting the three-year sentence threshold as a regression discontinuity: individuals with sentences just above vs. just below the transfer threshold are plausibly equivalent on unobservable characteristics, enabling a cleaner estimate of the institutional treatment effect [60]; (3) qualitative comparative studies using semi-structured interviews with military and civilian releases to map the subjective mechanisms of reintegration and identify the “active ingredients” of successful post-release trajectories in each context [61]; (4) longitudinal analysis of the 10 recidivist military cases to understand the specific failure mechanisms that break down the protective factor advantages of military service backgrounds [62]; and (5) updated comparative cohort studies using 2020–2025 data to assess the durability of the documented patterns.

6. Conclusions

This study provides the first systematic quantitative comparison of social reintegration outcomes in the Spanish civilian and military prison systems, documenting a robust and large-magnitude advantage for the military system (+15.17 to +18.13 percentage points, p < 0.001) that persists across three sensitivity scenarios addressing the uncertainty from 105 cases with incomplete follow-up data.
These findings are interpreted with explicit caution regarding causal attribution [54]. The two populations compared differ fundamentally on multiple dimensions—prior criminal history, educational attainment, offense type, substance abuse prevalence, and post-release structural opportunities—that independently predict reintegration outcomes. The military system’s superiority likely reflects the combined operation of population selection advantages and institutional program effects [63], but the relative contribution of each component cannot be determined from currently available aggregate data.
The contribution of this study is threefold: it establishes and quantifies a robust empirical phenomenon that demands theoretical explanation; it situates this phenomenon within a multi-theoretical framework drawing on life-course theory [8,14], RNR principles [9], desistance theory [3,11], and social capital theory to generate specific, testable causal hypotheses; and it specifies the research designs needed to test those hypotheses and move from description to causal understanding. Observational comparative evidence of this kind is a legitimate and necessary stage in a cumulative evidence-based program aimed at improving reintegration outcomes across both systems.
Future research using individual-level data [58], quasi-experimental designs [60], and longitudinal outcome tracking will be essential to determine the extent to which the military system’s institutional practices are transferable to the civilian context and to which populations. The stakes of this question are substantial: if even a fraction of the observed advantage reflects replicable programmatic components, the potential public safety and economic benefits of reform are considerable.

Author Contributions

Conceptualization, L.M.O.S. and J.L.R.G.; methodology, L.M.O.S. and R.L.B.; formal analysis, R.L.B.; investigation, L.M.O.S.; resources, J.L.R.G.; data curation, L.M.O.S. and J.L.R.G.; writing—original draft preparation, L.M.O.S.; writing—review and editing, R.L.B. and J.L.R.G.; visualization, R.L.B.; supervision, R.L.B. and J.L.R.G.; project administration, R.L.B. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Ethical review and approval were waived for this study, as it uses exclusively aggregated and anonymized data from official administrative records with no personal identifiers of any kind, in accordance with current Spanish regulations on personal data protection (Organic Law 3/2018).

Data Availability Statement

Data on the civilian prison system are available from the General Secretariat of Penitentiary Institutions (SGIP) approved 2022 [15]. Data on the military prison system were provided by the Ministry of Defence (MINISDEF) [33] and are available in the Yearbooks of Military Jurisdiction (Ministerio de Defensa, 2009–2019).

Acknowledgments

The authors thank the Ministry of Defence (MINISDEF) for providing access to ESPEMI administrative records. During the preparation of this manuscript, the authors used (GPT-5.5 Instant, OpenAI, San Francisco, CA, USA) for language editing, proofreading, and translation from Spanish and Polish into academic English. The authors have reviewed and edited the output and take full responsibility for the content of this publication.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Constitución Española. Boletín Oficial del Estado, 311, 29 December 1978. Available online: https://www.boe.es/buscar/act.php?id=BOE-A-1978-31229 (accessed on 15 June 2026).
  2. Fazel, S.; Wolf, A. A systematic review of criminal recidivism rates worldwide: Current difficulties and recommendations for best practice. PLoS ONE 2015, 10, e0130390. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  3. Maruna, S. Making Good: How Ex-Convicts Reform and Rebuild Their Lives; American Psychological Association: Washington, DC, USA, 2001. [Google Scholar]
  4. Ward, T.; Maruna, S. Rehabilitation: Beyond the Risk Paradigm; Routledge: Abingdon, UK, 2007. [Google Scholar]
  5. Real Decreto 112/2017; De 17 de Febrero, Por el Que se Aprueba el Reglamento Penitenciario Militar; Boletín Oficial del Estado: Madrid, Spain, 2017.
  6. Buckley, S.; Hamilton, C. Within-case comparison of the adult and youth justice systems: New directions for a ‘fractured’ penal state? Criminol. Crim. Justice 2026, 26, 716–733. [Google Scholar] [CrossRef] [Scilit]
  7. Yukhnenko, D.; Blackwood, N.; Fazel, S. Risk factors for recidivism in individuals receiving community sentences: A systematic review and meta-analysis. CNS Spectr. 2020, 25, 252–263. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  8. Laub, J.H.; Sampson, R.J. Shared Beginnings, Divergent Lives: Delinquent Boys to Age 70; Harvard University Press: Cambridge, MA, USA, 2003. [Google Scholar]
  9. Andrews, D.A.; Bonta, J. The Psychology of Criminal Conduct, 5th ed.; Anderson Publishing: Lexington, KY, USA, 2010. [Google Scholar]
  10. Garritsen, K.; Janković, M.; Masthoff, E.; De Caluwé, E.; Bogaerts, S. The role of dynamic risk and protective factors in predicting violent recidivism: Intellectual ability as a possible moderator? Int. J. Offender Ther. Comp. Criminol. 2024, 68, 207–234. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  11. Paternoster, R.; Bushway, S. Desistance and the feared self: Toward an identity theory of criminal desistance. J. Crim. Law Criminol. 2009, 99, 1103–1156. [Google Scholar]
  12. Cullen, F.T. Social support as an organizing concept for criminology: Presidential address to the Academy of Criminal Justice Sciences. Justice Q. 1994, 11, 527–559. [Google Scholar] [CrossRef] [Scilit]
  13. Giordano, P.C.; Cernkovich, S.A.; Rudolph, J.L. Gender, crime, and desistance: Toward a theory of cognitive transformation. Am. J. Sociol. 2002, 107, 990–1064. [Google Scholar] [CrossRef] [Scilit]
  14. Sampson, R.J.; Laub, J.H. Crime in the Making: Pathways and Turning Points Through Life; Harvard University Press: Cambridge, MA, USA, 1993. [Google Scholar]
  15. Secretaría General de Instituciones Penitenciarias. Estudio Sobre la Reincidencia Penitenciaria en España; Ministerio del Interior: Madrid, Spain, 2022.
  16. Duwe, G.; Clark, V. The Gap Between the Ideal and Reality in Risk-Needs-Responsivity Assessments: A Survey of U.S. Prisons. Correct. Policy Pract. Res. 2026, 11, 121–138. [Google Scholar] [CrossRef] [Scilit]
  17. Duan, W.; Wang, Z.; Yang, C.; Ke, S. Are risk-need-responsivity principles golden? A meta-analysis of randomized controlled trials of community correction programs. J. Exp. Criminol. 2024, 20, 593–616. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  18. Lipsey, M.W.; Cullen, F.T. The effectiveness of correctional rehabilitation: A review of systematic reviews. Annu. Rev. Law Soc. Sci. 2007, 3, 297–320. [Google Scholar] [CrossRef] [Scilit]
  19. Lowenkamp, C.T.; Latessa, E.J.; Holsinger, A.M. The risk principle in action: What have we learned from 13,676 offenders and 97 correctional programs? Crime Delinq. 2006, 52, 77–93. [Google Scholar] [CrossRef] [Scilit]
  20. Aos, S.; Miller, M.; Drake, E. Evidence-Based Adult Corrections Programs: What Works and What Does Not, Document No. 06-01-1201; Washington State Institute for Public Policy: Olympia, WA, USA, 2006. Available online: https://www.wsipp.wa.gov/ReportFile/968/ (accessed on 15 June 2026).
  21. Landenberger, N.A.; Lipsey, M.W. The positive effects of cognitive-behavioral programs for offenders: A meta-analysis of factors associated with effective treatment. J. Exp. Criminol. 2005, 1, 451–476. [Google Scholar] [CrossRef] [Scilit]
  22. Roxell, L. What contributes to fewer cases of recidivism? Treatment, education, and work in prison. J. Offender Rehabil. 2024, 63, 480–507. [Google Scholar] [CrossRef] [Scilit]
  23. Duwe, G.; Clark, V. The effect of prison-based college education programs on recidivism: A propensity score matching approach. J. Crim. Justice 2013, 41, 196–204. [Google Scholar] [CrossRef] [Scilit]
  24. Coleman, J.S. Social capital in the creation of human capital. Am. J. Sociol. 1988, 94, S95–S120. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  25. Sharma, K. Turning points: An evaluation of formal and informal interventions in the life of ex-offenders. Themis Res. J. Justice Stud. Forensic Sci. 2024, 12, 13. [Google Scholar] [CrossRef] [Scilit]
  26. Cataldi, L.; Cataldi, S. Prison and love: The role of affection and rehabilitative actions in reducing recidivism and beyond. Soc. Sci. 2024, 13, 323. [Google Scholar] [CrossRef] [Scilit]
  27. Lopoo, M.E. Military matters? The impact of veteran status on recidivism. Socius Sociol. Res. A Dyn. World 2025, 11, 23780231251395474. [Google Scholar] [CrossRef] [Scilit]
  28. Mitchell, O.; Wilson, D.B.; MacKenzie, D.L. The effectiveness of incarceration-based drug treatment on criminal behavior: A systematic review. Campbell Syst. Rev. 2012, 8, 1–76. [Google Scholar] [CrossRef] [Scilit]
  29. Singh, K.; Timko, C.; Yu, M.; Taylor, E.; Blue-Howells, J.; Finlay, A.K. Scoping review of military veterans involved in the criminal legal system and their health and healthcare: 5-year update and map to the Veterans-Sequential Intercept Model. Health Justice 2024, 12, 18. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  30. Tsai, J.; Kelton, K.; Blonigen, D.M.; McInnes, D.K.; Clark, S.; Blue-Howells, J.; Hooshyar, D.; Pugh, M.J. A research agenda for criminal justice involvement among U.S. veterans. Mil. Med. 2024, 189, e481–e485. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  31. United States Disciplinary Barracks. Institution Profile and Statistics 2020; Department of the Army: Washington, DC, USA, 2021. [Google Scholar]
  32. Van Ginneken, E.F.J.C.; Crewe, B. Explaining experiences of punishment and degradation in prisons: A comparative analysis of England & Wales and Norway. Eur. J. Criminol. 2025, 22, 983–1003. [Google Scholar] [CrossRef] [Scilit]
  33. Ministerio de Defensa. Anuarios de la Jurisdicción Militar (Annual Publication Series); Secretaría General Técnica: Madrid, Spain, 2019. [Google Scholar]
  34. Muñoz Cuesta, J. Modificación del Código Penal por LO 1/2026 en Materia de Multirreincidencia y Diversas Figuras Agravadas. 2026. Available online: https://elderecho.com/modificacion-del-codigo-penal-por-lo-1-2026 (accessed on 15 June 2026).
  35. Ley Orgánica 1/2026, De 8 de Abril, en Materia de Multirreincidencia; Boletín Oficial del Estado: Madrid, Spain, 2026; Available online: https://www.boe.es/buscar/act.php?id=BOE-A-2026-7966 (accessed on 15 June 2026).
  36. Karimi-Haghighi, M.; Castillo, C.; Tolan, S.; Lum, K. Effect of conditional release on violent and general recidivism: A causal inference study. J. Exp. Criminol. 2025, 21, 407–449. [Google Scholar] [CrossRef] [Scilit]
  37. Secretaría General de Instituciones Penitenciarias. Estadística Penitenciaria Mensual: Diciembre 2023; Ministerio del Interior: Madrid, Spain, 2024.
  38. Euser, A.M.; Zoccali, C.; Jager, K.J.; Dekker, F.W. Cohort studies: Prospective versus retrospective. Nephron Clin. Pract. 2009, 113, c214–c217. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  39. Wong, J.S.; Neilsen, K.; Gushue, K.; Lee, C. The effects of halfway houses on criminal recidivism: An updated systematic review and meta-analysis. Int. J. Offender Ther. Comp. Criminol. 2025, 69, 1781–1804. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  40. Perley-Robertson, B.; Babchishin, K.M.; Helmus, L.M. The effect of missing item data on the relative predictive accuracy of correctional risk assessment tools. Assessment 2024, 31, 1673–1687. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  41. Cole, S.R.; Zivich, P.N.; Edwards, J.K.; Shook-Sa, B.E.; Hudgens, M.G. Sensitivity analyses for means or proportions with missing outcome data. Epidemiology 2023, 34, 645–651. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  42. Dünkel, F.; Pruin, I.; Storgaard, A.; Weber, J. (Eds.) Prisoner Resettlement in Europe; Routledge: Abingdon, UK, 2018. [Google Scholar]
  43. Kristoffersen, R. Some Results on Recidivism Among Released Prisoners in Norway in 2015–2018 [Research Note]; University College of Norwegian Correctional Service (KRUS): Lillestrøm, Norway, 2020. [Google Scholar]
  44. Cohen, J. Statistical Power Analysis for the Behavioral Sciences, 2nd ed.; Lawrence Erlbaum Associates: Mahwah, NJ, USA, 1988. [Google Scholar]
  45. Ministry of Justice. Proven Reoffending Statistics Quarterly Bulletin, England and Wales; UK Government: London, UK, 2019. Available online: https://www.gov.uk/government/collections/proven-reoffending-statistics (accessed on 15 June 2026).
  46. Kensey, A.; Tournier, P.-V. Prisonniers du Passé? Cohorte des Personnes Condamnées, Libérées en 1996–1997: Examen de Leur Casier Judiciaire 5 Ans Après la Levée D’écrou (Travaux & Documents No. 68); Ministère de la Justice: Paris, France, 2005. [Google Scholar]
  47. Álvarez-Crespo, R.; Arroyo-Cobo, J.M.; Negredo-López, L. The bridging program for social reinsertion: Early detection of people with serious mental illness who are prosecuted in Spain. Int. J. Prison. Health 2022, 14, 252–262. [Google Scholar] [CrossRef] [Scilit]
  48. Mourão, A.; Sousa, M.; Ferreira, M.; Gonçalves, L.; Caridade, S.; Cunha, O. Beyond recidivism: A systematic review exploring comprehensive criteria for successful reintegration after prison release. Crim. Justice Behav. 2025, 52, 1173–1198. [Google Scholar] [CrossRef] [Scilit]
  49. Carrasco Martínez, M.T. La Reinserción Social de Presos en España; Universidad de Valladolid: Valladolid, Spain, 2025; Available online: https://uvadoc.uva.es/handle/10324/79314 (accessed on 15 June 2026).
  50. Calonico, S.; Jawadekar, N.; Kezios, K.; Zeki Al Hazzouri, A. Regression discontinuity design studies: A guide for health researchers. BMJ 2024, 384, e072254. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  51. Koehler, J.A.; Lösel, F.; Akoensi, T.D.; Humphreys, D.K. A systematic review and meta-analysis on the effects of young offender treatment programs in Europe. J. Exp. Criminol. 2013, 9, 19–43. [Google Scholar] [CrossRef] [Scilit]
  52. Pratt, J.; Eriksson, A. Contrasts in Punishment: An Explanation of Anglophone Excess and Nordic Exceptionalism; Routledge: Abingdon, UK, 2013. [Google Scholar]
  53. Gottfredson, M.R.; Hirschi, T. A General Theory of Crime; Stanford University Press: Redwood City, CA, USA, 1990. [Google Scholar]
  54. Shadish, W.R.; Cook, T.D.; Campbell, D.T. Experimental and Quasi-Experimental Designs for Generalized Causal Inference; Houghton Mifflin: Boston, MA, USA, 2002. [Google Scholar]
  55. Diez-Roux, A.V. Multilevel analysis in public health research. Annu. Rev. Public Health 2000, 21, 171–192. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  56. Thabane, L.; Mbuagbaw, L.; Zhang, S.; Samaan, Z.; Marcucci, M.; Ye, C.; Thabane, M.; Giangregorio, L.; Dennis, B.; Kosa, D.; et al. A tutorial on sensitivity analyses in clinical trials: The what, why, when and how. BMC Med. Res. Methodol. 2013, 13, 92. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  57. Negi, S.; Chatterjee, D.; Tripathy, S.; Udhaya, E.; Kumar, D.; Mehta, V. Rehabilitation and re-entry programmes for justice-involved individuals and their effectiveness in social reintegration: A systematic review. Int. J. Offender Ther. Comp. Criminol. 2026, 70, 1037–1057. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  58. Harron, K.; Dibben, C.; Boyd, J.; Hjern, A.; Azimaee, M.; Barreto, M.L.; Goldstein, H. Challenges in administrative data linkage for research. Big Data Soc. 2017, 4, 2053951717745678. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  59. Apel, R.J.; Sweeten, G. Propensity score matching in criminology and criminal justice. In Handbook of Quantitative Criminology; Piquero, A.R., Weisburd, D., Eds.; Springer: Berlin/Heidelberg, Germany, 2010; pp. 543–562. [Google Scholar]
  60. Calonico, S.; Cattaneo, M.D.; Titiunik, R. Robust nonparametric confidence intervals for regression-discontinuity designs. Econometrica 2014, 82, 2295–2326. [Google Scholar] [CrossRef] [Scilit]
  61. Moles-López, E.; Burgos-Jiménez, R.J.; Añaños, F.T. La prisión y su acción re-insertiva. Estudio socioeducativo y de género en España. Converg. Rev. Cienc. Soc. 2023, 30, 14. [Google Scholar] [CrossRef] [Scilit]
  62. Farrall, S.; Calverley, A. Understanding Desistance from Crime: Theoretical Directions in Resettlement and Rehabilitation; Open University Press: London, UK, 2006. [Google Scholar]
  63. Nagin, D.S.; Cullen, F.T.; Jonson, C.L. Imprisonment and reoffending. Crime Justice 2009, 38, 115–200. [Google Scholar] [CrossRef] [Scilit]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Journal Statistics
Multiple requests from the same IP address are counted as one view.
Inclusion and Accessibility in Institutional Communication: Study of the Official Facebook Accounts of Ecuadorian Public Institutions
Previous
Algorithmic Sacredness and Algorithmic Pluralism: Content Moderation and the Symbolic Dispossession of Contemporary European Paganism
Next