Five Reentry Research Corrections Worth Knowing in 2026

Why this matters

Older reentry numbers anchor budget decisions, vendor briefings, and op-eds. When the underlying evidence has moved and the talking points haven’t, programs end up designed for a problem that no longer looks the way the citation implies.

The five corrections below are drawn from our 2026 reentry literature review and represent the places where the 2019–2025 evidence has most materially updated the older narrative.

1. The “83% rearrest rate” is a 2005-cohort number

The widely quoted “83% rearrested within 9 years” figure comes from the BJS 2005 release cohort, published in 2018 (Alper, Durose & Markman). The newer BJS 2012 cohort 5-year follow-up, published in 2021 (Durose & Antenangeli), shows lower rates at the comparable time-window — 62% rearrested at 3 years, 71% at 5 years. Return-to-prison rates have fallen approximately 23% since 2008 (Council of State Governments Justice Center analysis of BJS data).

Why it matters: Citing the 2005 figure as if it describes today’s reentry conditions overstates the problem and understates the policy progress that has produced the decline. Use the 2012 cohort. Better yet, use both, with dates attached.

2. The “129× post-release overdose risk” figure is pre-fentanyl

The much-cited 129× figure comes from a Washington State cohort released 1999–2003 (Binswanger et al. 2007, NEJM), before fentanyl dominated the U.S. drug supply. Current estimates put post-release opioid overdose risk at roughly 10–40× the general population, with the mechanism shifted from heroin-tolerance loss to fentanyl-potency unpredictability (Ranapurwala et al. 2022 AJPH; Hartung et al. 2023, J. Substance Use & Addiction Treatment; Butler et al. 2023, BMJ Open).

Why it matters: Underestimating the fentanyl-era mechanism leads programs to under-prioritize MOUD. Overstating the older multiplier flattens the risk profile in ways that make the first two weeks look the same as the rest of the year — they aren’t.

3. Ban the Box has been shown to amplify racial disparity in some contexts

Ban-the-Box laws were initially treated as an unambiguous reform win. Recent evidence has turned. Agan & Starr (2018, QJE) and Doleac & Hansen (2020, J. Labor Economics) find that in some contexts, removing the record question from job applications amplifies racial disparity through statistical discrimination — when employers cannot ask about records, some appear to use race as a proxy. Rose (2021) reports neutral effects in Seattle.

Why it matters: Reform without enforcement and complementary measures backfires. Cite the 2018–2021 studies, not the 2014 advocacy framing.

4. The race × record hiring penalty is more complex than the 2003 framing

Devah Pager’s foundational 2003 Milwaukee audit established the modern record-penalty literature. The 2025 update — Lageson et al. in Criminology, surveying 1,080 employers — keeps the central finding (the record penalty remains large) but complicates the racial-gap framing in three ways: (a) the penalty for an official background-check “hit” is consistently large, (b) the racial gap interacts with industry, applicant signaling, and ban-the-box context, and (c) sealing or expungement reduces but does not eliminate the penalty.

Why it matters: “Record + race doubles the penalty” is a useful simplification for an op-ed, but for program design and policy, the 2025 framing is more precise — and points to different remedies.

5. Supervision reform is moving — unevenly

The Council of State Governments Justice Center’s 2024 supervision-violations report documents real progress. Parole-violation readmissions decreased by ≥10% in 31 states between 2018 and 2023. Probation-violation admissions decreased by ≥10% in 29 states. But the trend isn’t uniform: admissions in 25 states have started rising again from their 2021 pandemic lows, though not yet returning to 2018 levels (CSG Justice Center 2024). Nationally, nearly 200,000 people were admitted to prison for supervision violations in 2023 — more than 110,000 of them for technical violations alone.

Why it matters: Single-direction narratives (“supervision reform is working” / “supervision reform isn’t working”) both lose. The accurate story is “moving, unevenly, with state-level variation that should drive how each agency frames its own data.”

The through-line

The 2022 National Academies report The Limits of Recidivism said it plainly: recidivism alone is an inadequate measure of reentry success. The five corrections above are specific instances of the broader pattern. The field has updated its evidence base meaningfully in the last seven years. The talking points have not always kept up.

If you cite older statistics, cite them with their cohort and year. And check whether a 2019–2025 update changes the picture before you build a policy or pitch around the older number.