Why Recidivism Rates Disagree, and What Rearrest, Reconviction and Return to Prison Each Measure
By MHB Admin ·
Few criminal-justice statistics are quoted as often, or as loosely, as recidivism rates. One figure says more than 80% of people released from state prisons are arrested again within a decade. Another says fewer than half of people released from federal custody are rearrested in eight years. A widely cited study concludes that about two-thirds of people who go to prison never return. In England and Wales, official figures show fewer than half of adults released from custody reoffending within a year.
These numbers are not necessarily in conflict. Most are accurate descriptions of different things. They count different events, follow different populations, use different time windows and rely on different records. The word "recidivism" covers all of them, which is why the same justice system can be described as a revolving door or as one that most people leave for good, depending on which figure is chosen.
This article explains how recidivism is measured, why the definitions matter, and how to read the most commonly cited US figures and the international comparisons often drawn from them.
Recidivism is an event, a population and a clock
Every recidivism statistic is built from three choices.
The event. What counts as returning to crime? The most common options are a new arrest, a new conviction, a new prison sentence, or any return to prison, including returns for violating the conditions of parole or probation without a new crime.
The population. Who is being followed? Common groups include people released from state prison in a given year, people released from federal prison or starting federal probation, people convicted in court, or everyone who received any criminal sanction.
The clock. How long are people followed, and from when? Follow-up periods range from one year to ten or more. A longer window always produces a higher rate, because more time allows more events.
Changing any one of these choices changes the result, often dramatically. The table below shows how the main measures differ.
html
<table>
<thead>
<tr>
<th>Measure</th>
<th>What counts</th>
<th>What it captures</th>
<th>Main limitation</th>
</tr>
</thead>
<tbody>
<tr>
<td>Rearrest</td>
<td>Any new arrest after release</td>
<td>The broadest sign of renewed contact with police</td>
<td>An arrest is not proof of an offence; many arrests do not lead to charges or convictions</td>
</tr>
<tr>
<td>Reconviction</td>
<td>A new conviction for an offence committed after release</td>
<td>Offences proven in court</td>
<td>Depends on prosecution and court decisions; can take months or years to record</td>
</tr>
<tr>
<td>Reincarceration for a new sentence</td>
<td>A new prison sentence for a new offence</td>
<td>Serious reoffending that results in imprisonment</td>
<td>Reflects sentencing policy as well as behaviour</td>
</tr>
<tr>
<td>Return to prison (any reason)</td>
<td>Return for a new sentence or for violating parole or probation conditions</td>
<td>Total flow back into prison</td>
<td>Includes technical violations, such as missed appointments or failed drug tests, that are not new crimes</td>
</tr>
<tr>
<td>Proven reoffending (England and Wales)</td>
<td>An offence committed within one year that leads to a conviction or caution, with six further months allowed for it to be proven</td>
<td>Officially proven reoffending across all sanctions</td>
<td>Short follow-up; definitions differ from US measures</td>
</tr>
</tbody>
</table>The most cited US numbers, read carefully
The Bureau of Justice Statistics (BJS), part of the US Department of Justice, produces the most widely cited studies of state prisoners. It follows cohorts of people released from state prisons using criminal history records, including arrests anywhere in the country, not only in the state that released them.
The best-known figures come from people released in 30 states in 2005, followed for nine years in a report published in 2018. About 44% were arrested within the first year after release, 68% within three years and 83% within nine years. Those numbers are often quoted without the definition. They are rearrest rates: any arrest, for any offence, in any state.
Two later BJS studies show similar patterns over different periods. Among people released from state prisons in 24 states in 2008 and followed for ten years, about 66% were arrested within three years and about 82% within ten years. About 61% returned to prison within ten years, either for a new sentence or for violating the conditions of their release. Sixteen per cent were arrested at some point in a state other than the one that released them, which shows why studies relying only on a single state's records undercount.
For people released from state prisons in 34 states in 2012, BJS reported that 62% were arrested within three years and 71% within five. Within five years, 46% had returned to prison for a violation or a new sentence. Rearrest rates fell with age: 81% for people aged 24 or younger at release, 74% for those aged 25 to 39 and 61% for those aged 40 or older.
Federal figures are lower. The US Sentencing Commission, which studies people convicted in federal courts, followed people released from federal prison or placed on federal probation in 2005 over eight years. Of that group, 49.3% were rearrested, 31.7% were reconvicted and 24.6% were reincarcerated. A later study of 32,135 people released or starting probation in 2010 found the same eight-year rearrest rate, 49.3%, with a median of 19 months to rearrest among those rearrested.
The federal and state groups differ in several ways that explain much of the gap. The federal studies include people sentenced to probation as well as prison, and federal offenders differ from state prisoners in offence types, criminal histories and ages. The federal figures are not evidence that federal prisons "work better". They describe a different population.
The same cohort, many answers
The importance of the definition is clearest when one cohort is measured several ways. For the federal 2005 cohort, the eight-year rate was 49.3% for rearrest, 31.7% for reconviction and 24.6% for reincarceration. All three describe the same people over the same period. Which is "the" recidivism rate depends entirely on the question being asked.
Rearrest is the broadest measure and the one most often cited, partly because arrest records are the most complete source available. But an arrest is not proof of an offence. Arrests can reflect policing priorities as well as behaviour, and some people are arrested and later released without charge or acquitted. Reconviction requires proof in court but depends on prosecutors' decisions and court delays, and convictions can take months or years to appear in records. Reincarceration is narrower still and reflects sentencing laws, which vary between states.
Why "return to prison" is not the same as "new crime"
Returns to prison include a category that rarely features in public discussion: technical violations. People released under supervision on parole or probation must meet conditions such as reporting to an officer, staying in a particular area, holding a job, attending treatment or passing drug tests. Breaking those conditions can lead to a return to prison even without a new crime.
These returns are substantial. A 2019 analysis by the Council of State Governments Justice Center, based on data from state corrections agencies, found that 45% of admissions to state prisons were due to violations of probation or parole. These include both new offences and technical violations, with technical violations alone accounting for about a quarter of all admissions. The share varies widely from state to state.
A rise or fall in return-to-prison rates may therefore reflect changes in supervision policy as much as changes in behaviour. A state that responds to missed appointments with short jail stays rather than prison, or that reduces the number of conditions imposed, may see its reincarceration rate fall without any change in reoffending.
The cohort problem, and why most people may never return
The most counterintuitive finding in recidivism research comes from a 2014 study published in Crime & Delinquency by William Rhodes, Gerald Gaes and colleagues. Using national corrections data covering 13 years of admissions and releases in 17 states, they concluded that about two of every three people who enter and leave prison never return.
That appears to contradict the BJS figures. It does not, once the populations are understood.
BJS studies follow a release cohort: everyone released from prison in a particular year. A release cohort over-represents people who cycle in and out of prison frequently, because they are released more often. Someone released five times in a decade is five times as likely to appear in a given year's release cohort as someone released once. A release cohort therefore contains a higher proportion of people with repeated returns than the population of everyone who ever goes to prison.
Rhodes and his colleagues followed individuals over time instead of following a single year's releases. Counted that way, most people who go to prison go once. A smaller group returns repeatedly and accounts for a large share of releases and returns.
Neither approach is wrong. A release cohort answers a practical question for the prison system: of the people leaving this year, how many will come back? An individual-based approach answers a different question: of everyone who goes to prison, what share returns? The headline "most released prisoners are rearrested" and the headline "most people who go to prison never return" can both be accurate, because they describe different groups.
International comparisons usually compare different things
Recidivism rates are frequently compared between countries, often to argue that one country's prison system is more effective than another's. Most such comparisons are unreliable because the measures do not match.
England and Wales illustrate the problem. The Ministry of Justice publishes "proven reoffending" statistics. These count offences committed within a one-year follow-up period that lead to a conviction or caution, with a further six months allowed for the offence to be proven. Among adults who left custody between January and March 2024, the proven reoffending rate was 43.8%. Among adults released from sentences of less than 12 months, it was 66.0%.
Those figures cannot be set beside the US rearrest rates above. The English measure covers one year rather than three, five or ten. It counts proven offences rather than arrests, and it includes cautions, which have no direct US equivalent. Comparisons with other countries often have the same problem, setting a short-window reconviction or reimprisonment rate against a long-window US rearrest rate. The gap between the numbers may reflect the measures as much as the systems.
Meaningful comparisons require the same event, a similar population and the same follow-up period. Researchers who have attempted careful cross-national comparisons emphasise how rarely those conditions are met.
What recidivism rates are used for, and the limits of each use
Recidivism statistics serve several purposes, and each use needs a measure suited to it.
Evaluating programmes. When a prison education or treatment programme reports lower recidivism among participants, the question is compared with whom. People who choose or are selected for programmes may differ from those who are not. Credible evaluations use comparison groups that are as similar as possible, ideally through random assignment. A lower rate among participants on its own does not show that a programme caused the difference.
Comparing jurisdictions. States publish their own recidivism rates, often using return to prison within three years. Differences between states can reflect supervision practices, sentencing laws, policing and record-keeping as much as reoffending. State-to-state rankings using these figures should be treated with caution.
Assessing risk. Risk-assessment tools used in sentencing, parole and pretrial decisions are trained on past outcomes, usually rearrest or reconviction. Their predictions inherit the strengths and weaknesses of those measures, including any differences in how often different groups are arrested.
Tracking trends. Following the same measure in the same jurisdiction over time is one of the more reliable uses, provided the definition and data sources stay constant. Even then, changes in policing, prosecution or supervision can shift the numbers without any change in behaviour.
What the evidence consistently shows
Despite the differences in definitions, several findings recur across the major studies.
Risk is highest soon after release. In the BJS studies, the largest share of first arrests occurred in the first year, and annual arrest rates declined in later years.
Age matters. Younger people at release have higher rates of rearrest than older people, a pattern found consistently across studies and consistent with long-established research showing that offending declines with age.
A minority accounts for a disproportionate share of repeated returns. That is the central insight of the cohort research, and it helps explain why release-cohort statistics look different from individual-based ones.
Measures diverge substantially. For the same group of people over the same period, rearrest, reconviction and reincarceration rates can differ by 20 percentage points or more.
How to read the next recidivism figure
A few questions help make sense of any recidivism statistic.
What event was counted? Arrest, conviction, a new prison sentence, or any return to prison including technical violations?
Who was followed? People released from prison, people on probation, everyone convicted, or a programme's participants? From which jurisdiction, and in which year?
For how long? One year, three, five or ten?
Whose records? Do the data include arrests in other states, or only the releasing state's records?
Compared with what? If a figure is presented as evidence that a programme or policy works, is there a credible comparison group?
Without those details, a recidivism rate is close to meaningless. With them, the apparently contradictory numbers become consistent descriptions of the same system from different angles. Prisons in the United States release many people who are rearrested, a significant number who return under supervision rules, a smaller number who are convicted and imprisoned again, and a large share of individuals who never come back.
