Registry Effectiveness: What Does the Evidence Show?

The third guide in SOLAR’s evidence sequence

The Understanding Sex-Offense Risk Assessment guide explains why risk is heterogeneous and why individualized assessment matters. The Understanding Recidivism Evidence guide explains why recidivism depends on population, outcome, follow-up, subgroup, and comparison.

This guide takes the next step: given what we know about risk and recidivism, does categorical registration or public notification produce enough measurable safety benefit to justify the system built around it?

Registry laws can serve several claimed functions at once. A law-enforcement database may be intended to help police identify or locate people. Public notification may be intended to help residents change their behavior. Verification rules may be intended to keep records current. Some policies are also defended as deterrence: the prospect of registration or public exposure is supposed to discourage offending.

Those are different mechanisms. A database can have investigative utility without proving that public Internet disclosure reduces sexual offending. A website can be widely available without proving that people use it, act on it, or prevent victimization.

A fair evaluation begins by naming the outcome. Researchers can ask whether a policy changes sexual recidivism, overall recidivism, first-time sexual offending, victimization, deterrence, apprehension, public protective behavior, information accuracy, administrative workload, cost, or collateral effects.

Calling a registry “effective” without specifying which outcome improved can hide the central question. People looking at a website is not the same outcome as fewer victims. Police having access to a database is not the same outcome as lower recidivism.

The SOLAR Evidence Matrix separates registry-related policies because studies do not all evaluate the same intervention. This is essential to reading the evidence accurately.

A useful example comes from Prescott and Rockoff. Their study estimated different effects for registration and public notification. Registration was associated with reductions in some reported sex offenses against local victims, while public notification appeared to operate through a different mechanism and may have increased recidivism among people already registered.

The lesson is not that one study settles the whole debate. The lesson is that “registration” and “public notification” are not interchangeable concepts.

The pooled result is supported by several influential individual evaluations. Agan analyzed three different datasets and found no detectable public-safety benefit across the designs examined. Letourneau and colleagues’ South Carolina analysis found that registration status did not significantly predict sexual recidivism in the modeled analyses.

A New York time-series study by Sandler, Freeman, and Socia found no support for SORN reducing the studied categories of sexual offending, including sexual recidivists and first-time offenders. Because it is a time-series study, it cannot perfectly isolate registration from notification or broader secular trends.

A large Freeman notified-versus-non-notified comparison also sits uneasily with a simple claim that notification lowers recidivism. Notified people were rearrested about twice as quickly for a sexual offense and 47% more quickly for a nonsexual offense. That result should not be read as proof that notification caused faster rearrest: notification assignment, baseline risk, surveillance intensity, and detection can all confound the comparison.

Public notification is sometimes defended as a general deterrent, not only as a recidivism intervention. In Prescott and Rockoff, notification appeared to reduce offending by people who were not already registered. That finding matters because it identifies a possible benefit operating through a mechanism different from reducing recidivism among registered people.

But the public-notification theory still contains a chain of assumptions: information must be available, people must find and understand it, they must change behavior, and those changes must prevent victimization. Each step can fail even if the website itself functions exactly as designed.

The Anderson and Sample Nebraska survey directly tested part of that chain. Among respondents with valid access data, 34.8% had accessed the registry and 65.2% had not. Among registry users with action data, 37.6% reported taking a preventative action and 62.4% did not.

That does not mean nobody uses registries. It means public availability does not automatically produce public use, and public use does not by itself establish a measurable crime-prevention effect.

Collateral consequences matter for two separate reasons. First, they are human and fairness costs borne by people on registries and their families. Second, they can affect the stability that public-safety systems ordinarily try to build: housing, work, relationships, treatment engagement, and successful reintegration.

The matrix does not support saying that notification simply “causes people to reoffend.” It does support saying that recurring reintegration burdens are well documented and that more intrusive notification has been associated with more socially destabilizing consequences.

The Lasher and McGrath quantitative review synthesized eight studies involving 1,503 people and found recurring housing, employment, social, and psychological burdens. The review also found that more intrusive notification strategies were associated with more socially destabilizing consequences.

Earlier work by Levenson and Cotter likewise documented job and housing loss, threats or harassment, psychological effects, and reports of inaccurate Internet-registry information among surveyed respondents.

Even if a registry has some informational value, the safety theory still depends on implementation. Information must be accurate. Systems must communicate. Homelessness and transience must be handled in ways that do not make location data meaningless. Public-facing information must be understandable. Agencies must have enough capacity to maintain the system they are asked to operate.

National law-enforcement research by Harris and colleagues identified concerns involving information reliability and utility, inter-system communication, homelessness and transience, and the public’s ability to interpret registry information.

A separate study by Cubellis, Walfield, and Harris found mixed law-enforcement views. Respondents in states with larger registries expressed greater concern about collateral consequences and less belief in SORN public-safety efficacy, while personnel more engaged in SORN work also sometimes reported more belief in its effectiveness.

These questions also connect back to SOLAR’s earlier evidence guides. If risk is heterogeneous, then a finding about a selected high-risk group should not automatically be generalized to everyone with a registry label. If recidivism varies by population, outcome, subgroup, and time, then policy evaluation should be at least as specific.

The component-level literature is heterogeneous, but the center of the evidence is not neutral. The strongest broad pooled evidence in the SOLAR matrix has not demonstrated an overall recidivism-reduction effect. Several major evaluations found no detectable benefit. Public-notification systems impose measurable burdens. And narrower favorable findings—especially for selected high-risk notification—do not establish that universal public disclosure produces the same result.

That leaves a legitimate public-safety question: what measurable safety benefit are we buying with all of this?

A policy can have some administrative or informational utility and still fail to justify its scale, publicity, duration, collateral consequences, or categorical design. Evidence-based public safety should be willing to distinguish individualized risk, targeted intervention, focused supervision where warranted, effective treatment, and prevention aimed at actual pathways to harm from a broad system whose benefits are too often assumed rather than demonstrated.