SOLAR Resource Guide

Understanding Recidivism Evidence

A plain-language guide to what recidivism statistics do and do not show — and how to read them without turning group data into fear, certainty, or myth.

Jump to Sources

Start Here

There is no single “sex-offender recidivism rate.” The number changes depending on who was studied, what counted as recidivism, how long people were followed, when the clock started, and what comparison group was used.

This guide is a companion to SOLAR’s risk-assessment guide. The risk guide asks, “What does this score or tool mean?” This guide asks, “What do observed reoffending data actually show?”

The strongest takeaway is simple: people convicted of sexual offenses are not uniquely or uniformly high-recidivism compared with other major offense groups. The evidence is more specific, more useful, and much less compatible with slogans.

Read a recidivism statistic in this order

Before accepting a claim, slow the number down and identify what it is really measuring.

Do first

1
Check the population: prison releases, supervision starts, CSEM-only cases, contact offenses, adults, youth, federal cases, state cases, or another group.
2
Check the outcome: any rearrest, sexual rearrest, charge, reconviction, reincarceration, self-report, or another measure.

Then do next

1
Check the follow-up: three years, five years, nine years, fifteen years, or another period.
2
Check the starting point: prison release, supervision start, assessment date, treatment completion, or another milestone.
3
Check the comparison: compared with whom, in the same study, using the same clock and outcome?

Remember

A dramatic relative-risk statement can describe a small absolute rate. Always ask, “What were the actual percentages?”

First question

What counted?

Rearrest, reconviction, reincarceration, official detection, and self-report are different measurements.

Second question

Who was studied?

Offense subtype, age, prior record, supervision context, and jurisdiction can materially change the picture.

Third question

Compared with what?

Same-study comparisons are safer than pulling percentages from unrelated studies and treating them as equivalent.

The public misconception this guide addresses

Public discussion often treats people convicted of sexual offenses as uniquely, uniformly, or “shockingly” high-recidivism. Large official datasets do not support that simplified claim. They show a more careful pattern: overall recidivism is often lower than many other offense groups, sexual-specific rearrest is a different outcome, same-type specialization occurs across many offense categories, and absolute detected sexual-recidivism rates are far below popular assumptions of inevitable repeat offending.

There Is No Single “Sex-Offender Recidivism Rate”

A recidivism number is only meaningful after you know the population, outcome, clock, and comparison.

Recidivism is not one natural fact waiting to be quoted. It is a measurement choice. A three-year rearrest rate for people released from state prison is not the same thing as a five-year reconviction rate for people starting federal supervision. A CSEM-only cohort is not the same thing as a broader contact-offense cohort. A study of any rearrest is not answering the same question as a study of another detected sexual offense.

This is why SOLAR treats precise recidivism claims as more useful than broad labels. A good public statement should preserve the population, outcome, measurement basis, follow-up length, and starting point.

Weak claim

“Sex offenders have a high recidivism rate.”

Stronger claim

“In this specific cohort, using this specific outcome, over this specific follow-up period, the observed rate was ___.”

The label is never the whole measurement

Offense subtype, age, prior record, follow-up period, jurisdiction, treatment history, supervision context, and outcome definition can all change the meaning of a recidivism statistic.

What Does “Recidivism” Mean?

Different measures answer different questions. They should not be collapsed into one generic rate.

“Recidivism” can mean several different things. Some measures are easier to count but less precise. Others are narrower but miss conduct that was never detected or never prosecuted.

Rearrest

A new arrest was recorded. This is commonly used in large official datasets, including BJS and USSC reports.

Watch for: Rearrest is not the same as proof, conviction, or all offending.

Charge

A formal charge was filed after an accusation or arrest.

Watch for: Charging practices vary by jurisdiction and case type.

Reconviction

A new conviction occurred after plea or trial.

Watch for: This is narrower than rearrest and depends on prosecution and court outcomes.

Reincarceration

A person returned to custody after a new sentence or a violation.

Watch for: This can mix new crimes with supervision or release-condition violations.

Official detected offending

Behavior captured through official systems such as arrests, charges, convictions, or corrections records.

Watch for: It undercounts undetected conduct and should not be treated as a full measure of all behavior.

Self-report

A person reports past conduct, often in treatment, research, or clinical settings.

Watch for: It may reveal behavior official records missed, but it is not the same as prospective future recidivism.

Do not swap measures mid-argument

A person can quote a rearrest study, call it “reoffending,” and then talk as if it proved all future conduct. That move is misleading. Keep the measurement basis attached to the claim.

Five Things to Check Before Believing a Recidivism Statistic

Most misuse becomes visible once you ask five simple questions.

The five checks

Population: Who was studied — state prison releases, federal supervision cases, CSEM-only cases, contact offenses, adults, youth, or another group?
Outcome: What counted — any rearrest, sexual rearrest, same-type rearrest, reconviction, reincarceration, self-report, or something else?
Follow-up length: How long were people followed — three years, five years, nine years, fifteen years, or another period?
Starting point: When did the clock begin — prison release, supervision start, assessment date, treatment completion, or another milestone?
Comparison group: Compared with whom — and was it the same study, same clock, same outcome, and same jurisdictional setting?

A simple way to ask for clarification

Use this when someone quotes a recidivism number in a meeting, article, hearing, family conversation, or policy discussion.

When you say that recidivism rate, what population was studied, what counted as recidivism, how long were people followed, when did the follow-up clock start, and what comparison group are you using?

Overall Recidivism and Sexual Recidivism Are Different Questions

Broad claims about being uniquely high-recidivism should not be built from a narrower sexual-specific outcome.

The phrase “recidivism rate” often hides two different questions. Overall recidivism asks whether a person had any new detected justice-system event, such as any rearrest. Sexual-specific recidivism asks whether the new detected event was another sexual offense.

This distinction matters because the public myth usually makes a broad claim: that people convicted of sexual offenses are uniquely high-recidivism in general. A narrow sexual-specific outcome does not answer that broad question.

The BJS 2005 rape/sexual-assault 9-year follow-up illustrates why precision matters. It reported lower overall arrest for people released after rape or sexual assault than for other released prisoners, while also separately reporting rape/sexual-assault arrest as a narrower outcome.

Overall recidivism

Asks: “Was there any new detected justice-system event?”

This is the measure to check before accepting broad claims about whether an offense group is “high recidivism” in general.

Sexual-specific recidivism

Asks: “Was there another detected sexual offense?”

This is a narrower outcome. It should be named clearly and not used as shorthand for overall recidivism.

Keep the outcome attached to the claim

A precise statement can acknowledge sexual-specific rearrest as a distinct outcome without turning it into a broad claim that sexual-offense populations are uniquely high-recidivism overall.

Relative Risk and Absolute Risk Can Sound Very Different

A statement can be mathematically true and still rhetorically misleading if the absolute rates are hidden.

Relative-risk language compares one group to another. Absolute risk tells you the actual percentage. Both can be useful, but they answer different questions.

In the BJS 2005 rape/sexual-assault 9-year follow-up, people released after rape or sexual assault were more likely than other released prisoners to be arrested for rape or sexual assault. The absolute rates were 7.7% versus 2.3% over nine years.

Relative framing

“About 3.3 times as likely to be arrested for rape or sexual assault.”

This sounds dramatic because it compares one rate to another.

Absolute framing

“7.7% versus 2.3% over nine years.”

This shows the actual observed detected rates.

The practical question

When a statistic sounds shocking, ask: “What are the actual absolute rates?” That question does not deny relative elevation. It keeps the number from being turned into a myth of inevitable repeat offending.

Are Sexual-Offense Populations Uniquely High-Recidivism?

Large official datasets do not support that broad claim when the outcome is overall rearrest.

The strongest way to compare offense groups is to use the same dataset, same follow-up clock, and same outcome. Same-study comparisons avoid mixing unrelated percentages from different populations.

Two Bureau of Justice Statistics reports are especially useful for this guide: the 1994 sex-offender prison-release report and the 2005 rape/sexual-assault 9-year follow-up. Both undermine the simple claim that people released after sexual offenses are the highest-recidivating offense group overall.

Same-study comparator examples

Study

BJS 1994 state-prison release cohort

Outcome and clock

Any rearrest within three years of prison release.

What it showed

43% of released sex-offense prisoners were rearrested for any offense, compared with 68% of released non-sex-offense prisoners.

Study

BJS 2005 rape/sexual-assault release cohort

Outcome and clock

Any arrest within nine years of prison release.

What it showed

67% of rape/sexual-assault releases were arrested for any crime, compared with 84% of other released prisoners.

Strong public takeaway

People convicted of sexual offenses are not uniquely or uniformly high-recidivism compared with other major offense groups. The answer changes when the outcome changes, which is exactly why “recidivism” must be defined before it is cited.

Same-Type Recidivism and Offense Specialization

Elevated same-type rearrest is a broader criminal-recidivism pattern, not something unique to sexual offending.

People released after many kinds of offenses are disproportionately likely to be rearrested for the same type of offense. Researchers often call this offense specialization. It matters because sexual-specific rearrest is sometimes treated as if it proves sexual offending is uniquely persistent. The data show a broader pattern.

The BJS 1994 all-prisoner recidivism report included a same-offense relative-likelihood table across many release-offense categories.

Homicide

1.4×