Recidivism

Abstract

Recidivism is a type of criminal behavior: the tendency of a person already sanctioned for an offense to offend again. It is among the most consequential outcome measures in correctional psychology and among the most slippery, because a recidivism rate means little without the definition and the follow-up window that produced it. The same released cohort yields sharply different figures depending on whether reoffending is counted as rearrest, reconviction, or return to custody, and on whether follow-up runs one year or ten (Fazel & Wolf, 2015). Against that measurement problem sits a mature applied science: actuarial instruments predict reoffending better than unstructured clinical judgment, rehabilitation matched to risk and criminogenic need measurably lowers it (Andrews et al., 1990), and desistance research asks how people stop rather than only why they persist.

Keywords: reoffending, criminal behavior, desistance, actuarial risk assessment

Few numbers in criminal justice are quoted as often, or as carelessly, as the recidivism rate. It is treated as a verdict on prisons, programs, and policies, yet the same population can be described as two-thirds reoffenders or one-quarter reoffenders without anyone lying, because the figure depends entirely on what is counted and for how long. Cognitive and correctional psychology approach recidivism on two fronts at once: as a measurement problem that must be disciplined before any comparison is meaningful, and as a behavioral outcome that can be predicted, reduced, and eventually abandoned. This article treats both, because the science of stopping reoffending rests on first measuring it honestly.

What Recidivism Is

Recidivism is the reoffending of a person who has already been sanctioned for a crime. The word carries a built-in before-and-after structure: there must be a first offense and a response to it, and then a subsequent offense that the measure counts. MeSH classifies the descriptor as a form of criminal behavior, and psychology treats it as an outcome to be explained and changed rather than as a trait people simply possess. What makes recidivism distinctive among behavioral outcomes is that it is defined by an administrative event, not by conduct as such: a person who reoffends but is never caught does not recidivate in any measured sense, so every recidivism statistic is also a statistic about surveillance and the reach of the justice system (Bonta & Andrews, 2017).

That dependence on administrative events is the root of the measurement problem. Reoffending can be registered at three successive stages, each rarer and more consequential than the last, and the choice among them changes the headline figure dramatically. A study that counts any rearrest will report a far higher rate than one counting only returns to prison, even when the underlying behavior is identical. Comparing a rearrest-based figure from one jurisdiction with a reconviction-based figure from another is a category error, yet it is done constantly.

Measuring Reoffending

The follow-up window matters as much as the definition. Reoffending accumulates over time, steeply at first and then more slowly, so a rate quoted without a window is uninterpretable. The largest national follow-up studies make the point concretely. Tracking prisoners released across thirty states, Durose, Cooper, and Snyder (2014) found that roughly two-thirds were rearrested within three years and about three-quarters within five, with most of the first reoffenses clustered in the earliest months after release. The updated thirty-four-state cohort showed the same shape, with about 62% rearrested within three years and 71% within five (Durose & Antenangeli, 2021). A one-year figure and a five-year figure for the same people are not in disagreement; they are answers to different questions.

Reoffending accumulates fastest in the first months and then slows, so the same cohort yields a different headline rate at every window. The curve is pinned to the two reference points marked in grey — about 44% rearrested by one year and 71% by five — so a one-year evaluation and a five-year evaluation of one program can differ by roughly 27 percentage points without any change in behavior.

These two sources of variation, definition and window, combine to make naive comparison almost meaningless. Reviewing recidivism rates reported worldwide, Fazel and Wolf (2015) found that studies differ so widely in what they count and how long they wait that cross-national rates cannot be compared without first harmonizing the measures, and they recommended a minimal common standard for reporting. A three-year update confirmed the picture: reconviction rates across countries ranged from roughly 20% to 60% at two years, a spread driven as much by methodology as by any real difference in offending (Yukhnenko, Sridhar, & Fazel, 2020). The practical lesson is that a recidivism statistic should never be read without its two companions, the operational definition and the follow-up period, printed beside it.

The three bars describe the same released cohort scored by the three definitions of Table 1. At the default 1,000 people the funnel reads 710 rearrested, 460 reconvicted, and 250 reincarcerated — so the rearrest rate is 2.84× the reincarceration rate, and a report free to choose its definition can nearly triple or third the apparent failure rate without any change in the underlying behavior.

Predicting Reoffending

If reoffending has identifiable causes, it should be partly predictable, and the applied science of recidivism is largely about prediction and its uses. Meta-analysis established early that some predictors are far stronger than others: synthesizing studies of adult offenders, Gendreau, Little, and Goggin (1996) found that dynamic, changeable factors, antisocial attitudes, antisocial associates, and problems in the domains of family, school, and work, predicted reoffending at least as well as static markers such as age and criminal history, and better than vague clinical impressions. Structured actuarial instruments that tally such validated factors outperform unstructured judgment, though with a characteristic asymmetry: reviewing seventy-three samples covering nearly 25,000 people, Fazel and colleagues (2012) found that these tools identify low-risk individuals well but are considerably weaker at confirming that a high-risk flag will be followed by an actual offense.

Toggle the validated factors present for a hypothetical case. The count maps to a risk tier, and the risk-need-responsivity model matches the intensity of service to that tier — not more, not less.

The tally is a group-level risk estimate, not an individual verdict: two people with the same count share only an average probability of reoffending. The model’s point is allocation — concentrating scarce treatment on higher-risk cases and targeting the specific needs that are checked, since low-risk cases gain little and can be made worse by intensive programming.

The risk-need-responsivity (RNR) model turns this predictive knowledge into a framework for intervention. Andrews, Bonta, and Wormith (2006) distinguished static risk markers from dynamic criminogenic needs, the changeable factors that both predict reoffending and can be targeted by treatment, and set out three principles: match the intensity of service to the offender's risk level, target criminogenic needs specifically, and deliver the program in a style the offender can respond to. The same logic drives specialized actuarial tools for particular populations. For sexual offending, Hanson and Morton-Bourgon (2005) meta-analyzed recidivism studies and isolated deviant sexual interests and an antisocial orientation as the main predictors, the empirical basis for widely used instruments such as the Static-99. In every case the output is a group-level probability, not an individual verdict, and the textbook synthesis of this tradition remains The Psychology of Criminal Conduct (Bonta & Andrews, 2017).

Reducing Reoffending

Prediction matters because it points to what can be changed. The question of whether correctional treatment reduces reoffending at all was once genuinely open, after Robert Martinson's much-cited 1974 review of rehabilitation studies was widely read as showing that nothing works, a conclusion that hardened into policy orthodoxy for a decade. The turning point was a clinically informed meta-analysis: Andrews and colleagues (1990) showed that treatment effects depend on whether a program honors the risk, need, and responsivity principles, that appropriate correctional service reduced recidivism substantially while inappropriate or purely punitive service did little or even increased it. The finding reframed the field from whether treatment works to which treatment works for whom.

A review of systematic reviews consolidated that reframing. Lipsey and Cullen (2007) surveyed the accumulated meta-analytic evidence and concluded that rehabilitation programs reduce reoffending by meaningful margins, whereas sanctions oriented toward deterrence and control show no such benefit and occasionally the reverse. The strongest contemporary evidence comes from randomized trials: Beaudry, Yu, Perry, and Fazel (2021) meta-analyzed randomized controlled trials of psychological interventions delivered in prison and found a modest but genuine reduction in recidivism, while candidly noting how few well-controlled trials exist. Taken together, the treatment literature supports a specific claim rather than a slogan: programs reduce reoffending when they are matched to risk and target criminogenic needs, and fail when they are not.

From Persistence to Desistance

Reducing reoffending through programs is one route out of crime; simply growing out of it is another, and the study of desistance asks how and why people stop. The developmental backdrop is Moffitt's (1993) taxonomy, which distinguishes a small group of life-course-persistent offenders from a far larger group of adolescence-limited offenders whose law-breaking is confined to the teenage years, so that most offending ends without any intervention at all. But desistance is not only maturation. Following a cohort of delinquent boys to age seventy, Sampson and Laub (2003) found that turning points in adult life, a stable marriage, steady employment, military service, restructured the informal social controls around a person and channeled even persistent offenders toward desistance, an age-graded theory in which social bonds, not fixed traits, do much of the work.

Desistance also has a cognitive, narrative dimension. In the Liverpool Desistance Study, Maruna (2001) compared the life stories of persisters and desisters and found that those who stopped had reconstructed their identities around a redemption script, a narrative in which the real self was always good and the criminal past becomes the raw material of a reformed life, whereas persisters told a condemnation script in which change felt impossible. That strengths-based view informs rehabilitation theory directly: Ward and Brown (2004) proposed the good lives model, which complements risk management by building an offender's capacity to secure legitimate goods, giving desistance something to move toward rather than only something to avoid.

Worked Example

Consider an illustrative cohort of 1,000 people released from prison and followed for five years, and suppose the data are recorded at all three stages of Table 1. Let 710 be rearrested at least once, 460 be reconvicted, and 250 be returned to custody for a new offense. Three defensible recidivism rates follow from the identical cohort: 71.0% by rearrest (710 / 1000), 46.0% by reconviction (460 / 1000), and 25.0% by reincarceration (250 / 1000). The rearrest figure is 710 / 250 = 2.84 times the reincarceration figure, so a commentator free to pick the definition can nearly triple or third the apparent failure rate without touching the behavior.

Now hold the definition fixed at rearrest and vary only the window. Reoffending accumulates fastest early, so with cumulative rearrest at roughly 44% by the end of year one and 71% by year five, a one-year evaluation of the same program would report 44% while a five-year evaluation reports 71%, a 27-percentage-point swing that reflects the clock, not the cohort. The two exercises together show why a bare recidivism rate is uninterpretable: the headline number is a joint function of what is counted and how long one waits, and both must be stated for the figure to carry any meaning (Fazel & Wolf, 2015).

Discussion

Recidivism occupies an unusual position between measurement and mechanism. As a measurement, it is fragile: so sensitive to operational choices that much of the methodological literature is devoted simply to making recidivism rates comparable before they are interpreted (Fazel & Wolf, 2015; Yukhnenko et al., 2020). As a behavioral outcome, by contrast, it is one of the better-understood targets in applied psychology, with a replicated structure of predictors, a treatment model whose effects are now estimated from randomized trials, and a desistance literature that explains why most offending ends. The two faces are connected: because the outcome is measured through administrative events, an apparent change in recidivism can reflect a change in policing or recall practice rather than in offending, and sound evaluation must rule that out.

The field also carries a heavy ethical load, because its instruments feed decisions that deprive people of liberty. Actuarial tools predict group probabilities, and their asymmetric accuracy, good at identifying low risk, weaker at confirming high risk (Fazel et al., 2012), means that treating a high-risk flag as an individual certainty will often be wrong. The rehabilitative and desistance traditions push back against a purely predictive use of the science by insisting that criminogenic needs are changeable and that identity and social bonds can be rebuilt (Maruna, 2001; Ward & Brown, 2004), so that the point of measuring reoffending is ultimately to reduce it rather than only to forecast it.

Current Directions

The most consequential recent advance is methodological on both of recidivism's faces. On measurement, the push for harmonized reporting has moved from recommendation toward routine: the three-year international update by Yukhnenko and colleagues (2020) applied a consistent framework across countries and confirmed that the wide spread in published reconviction rates is driven substantially by how studies are built, and the large national follow-up cohorts now report cumulative rates at multiple fixed windows so that the window is explicit rather than hidden (Durose & Antenangeli, 2021). On intervention, the center of gravity has shifted toward randomized evidence: Beaudry and colleagues (2021) show that the effect of prison-based psychological treatment on reoffending can be estimated from controlled trials rather than inferred from observational comparison, while making the thinness of the trial base itself the agenda for future work.

A second front is the growing integration of the desistance and risk traditions, once seen as rivals. The age-graded social-control account (Sampson & Laub, 2003) and the narrative-identity account (Maruna, 2001) are increasingly read alongside the risk-need-responsivity framework rather than against it, with strengths-based models such as the good lives model (Ward & Brown, 2004) offering a bridge between predicting who will reoffend and supporting the person in stopping. The direction of travel, on both the measurement and the intervention side, is from describing reoffending toward specifying the mechanisms that generate it and the conditions that measurably change them.

Common Misconceptions

The recidivism rate is a single, well-defined number.

There is no single rate. Rearrest, reconviction, and reincarceration give different figures for the same cohort, and the follow-up window changes each of them, so a rate is interpretable only with its definition and period attached (Fazel & Wolf, 2015).

Recidivism rates can be compared across countries.

Not without harmonizing the measures first. International rates vary so widely in method that much of the apparent cross-national difference reflects how reoffending was counted rather than how often it occurred (Yukhnenko et al., 2020).

A risk-assessment tool predicts whether a given person will reoffend.

It estimates a group-level probability and is better at ruling out low risk than at confirming that a high-risk individual will actually offend, so a high-risk score is not an individual prediction (Fazel et al., 2012).

Nothing reduces reoffending, so rehabilitation is pointless.

The nothing-works reading was overturned. Programs that match intensity to risk and target criminogenic needs reduce recidivism, while purely punitive sanctions do not (Andrews et al., 1990; Lipsey & Cullen, 2007).

Glossary

Actuarial risk assessment.

Prediction of reoffending by tallying empirically validated risk factors into a score, more accurate on average than unstructured clinical judgment.

Criminogenic need.

A dynamic, changeable risk factor for reoffending, such as antisocial attitudes or associates, that rehabilitation can target.

Desistance.

The process of ceasing to offend, through maturation, changed social bonds, intervention, or a reconstructed sense of self.

Dynamic risk factor.

A predictor of reoffending that can change over time and be influenced by treatment, in contrast to a static factor.

Follow-up window.

The period over which released individuals are tracked for reoffending; a longer window yields a higher cumulative recidivism rate.

Good lives model.

A strengths-based rehabilitation framework that reduces reoffending by building an offender's capacity to achieve legitimate personal goods.

Rearrest.

A new arrest counted as recidivism; the broadest and most sensitive of the three standard measures because it requires no conviction.

Recidivism.

The reoffending of a person who has already been sanctioned for a crime; a primary outcome measure for correctional interventions.

Reconviction.

A new criminal conviction counted as recidivism; a stricter measure than rearrest because it requires proof of guilt.

Redemption script.

A narrative identity, described by Maruna, in which a person's real self is good and the criminal past becomes material for a reformed life; associated with desistance.

Reincarceration.

A return to custody counted as recidivism, which may reflect a new offense or a technical violation of release conditions.

Risk-need-responsivity (RNR).

A model of offender assessment and treatment matching intervention intensity to risk, targeting criminogenic needs, and delivering programs in a responsive style.

Static risk factor.

A predictor of reoffending that does not change, such as age at first offense or number of prior convictions.

Static-99.

A widely used actuarial instrument estimating the risk of sexual recidivism from largely static items.

Key Researchers

Donald A. Andrews

(1941-2010). Professor of psychology at Carleton University; co-developer of the risk-need-responsivity model and co-author of The Psychology of Criminal Conduct. Wikipedia

James Bonta

(living). Former Director of Corrections Research at Public Safety Canada; co-developer of the risk-need-responsivity model and the Level of Service family of risk-need inventories. Google Scholar

Francis T. Cullen

(living). Distinguished Research Professor Emeritus at the University of Cincinnati; co-author of foundational meta-analyses on correctional treatment and the review of reviews establishing that rehabilitation reduces reoffending. Faculty page - ORCID - Google Scholar - Wikipedia

R. Karl Hanson

(living). Adjunct Research Professor at Carleton University and formerly of Public Safety Canada; developer of the Static-99 actuarial tool for sexual-recidivism risk. Faculty page - ORCID - Google Scholar

Shadd Maruna

(living). Professor in the Department of Sociology, Social Policy & Criminology at the University of Liverpool; author of Making Good and a leading figure in desistance research. ORCID - Google Scholar

Robert J. Sampson

(living). Henry Ford II Professor of the Social Sciences at Harvard University; co-author of the age-graded theory of informal social control and the Glueck-cohort study of desistance. Faculty page - ORCID - Google Scholar - Wikipedia

Frequently Asked Questions

What is recidivism?

Recidivism is the reoffending of a person who has already been sanctioned for a crime. It is a primary outcome measure in correctional psychology, used to evaluate prisons, programs, and policies, and it is defined by an administrative event, a rearrest, reconviction, or return to custody, rather than by behavior alone (Bonta & Andrews, 2017).

Why do recidivism rates differ so much between reports?

Because the headline figure depends on two choices. First, what counts as reoffending: rearrest gives the highest rate, reconviction a lower one, reincarceration lower still. Second, the follow-up window: reoffending accumulates over time, so a five-year rate is far higher than a one-year rate for the same people (Fazel & Wolf, 2015).

What is a typical recidivism rate?

There is no single typical rate, but large national follow-ups give a sense of scale. In U.S. data, roughly 62% of released prisoners were rearrested within three years and 71% within five (Durose & Antenangeli, 2021). International reconviction rates vary far more widely, from about 20% to 60% at two years, largely because of methodological differences (Yukhnenko et al., 2020).

Can we predict who will reoffend?

Partly. Structured actuarial instruments that tally validated risk factors predict reoffending better than unstructured clinical judgment, but they estimate group probabilities and are stronger at identifying low risk than at confirming that a high-risk person will actually offend (Fazel et al., 2012).

What is the risk-need-responsivity model?

It is the dominant framework for offender rehabilitation. It holds that treatment reduces reoffending most when its intensity is matched to the person's risk level, when it targets changeable criminogenic needs such as antisocial attitudes, and when it is delivered in a style the person can respond to (Andrews, Bonta, & Wormith, 2006).

Does rehabilitation actually reduce reoffending?

Yes, when it is done well. Programs that match intensity to risk and target criminogenic needs reduce recidivism, whereas purely punitive sanctions do not (Andrews et al., 1990; Lipsey & Cullen, 2007). Randomized trials of prison-based psychological interventions show a modest but genuine reduction (Beaudry, Yu, Perry, & Fazel, 2021).

What is desistance?

Desistance is the process of stopping offending. It is not merely the absence of crime but an active change, driven by maturation, by turning points such as stable work and marriage that reshape social bonds (Sampson & Laub, 2003), and by a reconstructed sense of self in which a person narrates a reformed identity (Maruna, 2001).

How is sexual-offense recidivism assessed?

With specialized actuarial tools built on the predictors that matter for that population. Meta-analysis identified deviant sexual interests and an antisocial orientation as the main predictors of sexual recidivism (Hanson & Morton-Bourgon, 2005), and these form the empirical basis for widely used instruments such as the Static-99.

References

Andrews, D. A., Zinger, I., Hoge, R. D., Bonta, J., Gendreau, P., & Cullen, F. T. (1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28(3), 369-404. https://doi.org/10.1111/j.1745-9125.1990.tb01330.x

Andrews, D. A., Bonta, J., & Wormith, J. S. (2006). The recent past and near future of risk and/or need assessment. Crime & Delinquency, 52(1), 7-27. https://doi.org/10.1177/0011128705281756

Beaudry, G., Yu, R., Perry, A. E., & Fazel, S. (2021). Effectiveness of psychological interventions in prison to reduce recidivism: A systematic review and meta-analysis of randomised controlled trials. The Lancet Psychiatry, 8(9), 759-773. https://doi.org/10.1016/S2215-0366(21)00170-X

Bonta, J., & Andrews, D. A. (2017). The psychology of criminal conduct (6th ed.). Routledge. ISBN 9781138935778.

Durose, M. R., Cooper, A. D., & Snyder, H. N. (2014). Recidivism of prisoners released in 30 states in 2005: Patterns from 2005 to 2010 (NCJ 244205). Bureau of Justice Statistics, U.S. Department of Justice. https://bjs.ojp.gov/content/pub/pdf/rprts05p0510.pdf

Durose, M. R., & Antenangeli, L. (2021). Recidivism of prisoners released in 34 states in 2012: A 5-year follow-up period (2012-2017) (NCJ 255947). Bureau of Justice Statistics, U.S. Department of Justice. https://bjs.ojp.gov/library/publications/recidivism-prisoners-released-34-states-2012-5-year-follow-period-2012-2017

Fazel, S., Singh, J. P., Doll, H., & Grann, M. (2012). Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24827 people: Systematic review and meta-analysis. BMJ, 345, e4692. https://doi.org/10.1136/bmj.e4692

Fazel, S., & Wolf, A. (2015). A systematic review of criminal recidivism rates worldwide: Current difficulties and recommendations for best practice. PLoS ONE, 10(6), e0130390. https://doi.org/10.1371/journal.pone.0130390

Gendreau, P., Little, T., & Goggin, C. (1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34(4), 575-608. https://doi.org/10.1111/j.1745-9125.1996.tb01220.x

Hanson, R. K., & Morton-Bourgon, K. E. (2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73(6), 1154-1163. https://doi.org/10.1037/0022-006X.73.6.1154

Lipsey, M. W., & Cullen, F. T. (2007). The effectiveness of correctional rehabilitation: A review of systematic reviews. Annual Review of Law and Social Science, 3, 297-320. https://doi.org/10.1146/annurev.lawsocsci.3.081806.112833

Maruna, S. (2001). Making good: How ex-convicts reform and rebuild their lives. American Psychological Association. ISBN 1557987319.

Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100(4), 674-701. https://doi.org/10.1037/0033-295X.100.4.674

Sampson, R. J., & Laub, J. H. (2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41(3), 555-592. https://doi.org/10.1111/j.1745-9125.2003.tb00997.x

Ward, T., & Brown, M. (2004). The good lives model and conceptual issues in offender rehabilitation. Psychology, Crime & Law, 10(3), 243-257. https://doi.org/10.1080/10683160410001662744

Yukhnenko, D., Sridhar, S., & Fazel, S. (2020). A systematic review of criminal recidivism rates worldwide: 3-year update. Wellcome Open Research, 4, 28. https://doi.org/10.12688/wellcomeopenres.14970.3