Recidivism & Reentry

The Bounded Metric Problem: What Georgia Counts and What It Doesn't

Georgia's official reentry statistics start with a narrow window. The state's three-year felony reconviction rate counts only felony convictions within three years of release and excludes misdemeanor arrests and later reoffending (Recidivism & Reentry Failures in Georgia). Federal data shows what that window hides. Among 404,638 state prisoners released in 30 states in 2005, 67.8% were rearrested within three years and 76.6% within five years (A Sense of Purpose as a Driver of Rehabilitation in Incarcerated People: An Evidence Brief). The 9-year follow-up of the same cohort found 68% rearrested within three years, 79% within six years, and 83% within nine years (A Sense of Purpose as a Driver of Rehabilitation in Incarcerated People: An Evidence Brief). No comparable long-term statewide rearrest rate is published by Georgia, leaving a data vacuum at exactly the point where most reoffending occurs.

The bounded metric also sits inside a system of unusually deep confinement. Georgia's incarceration rate is 881 per 100,000 residents, the 7th highest in the nation and higher than any country in the world except El Salvador (Recidivism & Reentry Failures in Georgia). Across all facility types, 95,000 people are behind bars in Georgia (Georgia Incarceration Trends: Population, Demographics & National Context), with approximately 53,000 in state prisons as of 2025 (Georgia Incarceration Trends: Population, Demographics & National Context). This means the reentry pipeline is structurally large, but the state's primary outcome measure is engineered to look small.

A Shrinking Parole Valve: Releases, Votes, and Completion Rates

The Parole Board's own annual reports trace a sustained contraction in releases. In FY16, the Board released 13,374 offenders from prison through all forms of clemency (BOPP Annual Report FY2016 — Georgia Parole Board). By FY18, that number had dropped to 10,363, then to 6,245 in FY22, 5,863 in FY23, 5,443 in FY24, and 5,588 in FY25 (BOPP Annual Report FY2018 — Georgia Parole Board; BOPP Annual Report FY2022 — Georgia Parole Board; BOPP Annual Report FY2023 — Georgia Parole Board; BOPP Annual Report FY2024 — Georgia Parole Board; BOPP Annual Report FY2025 — Georgia Parole Board). FY24 was 420 fewer releases than the previous year; FY22 was 2,389 fewer than FY21 (BOPP Annual Report FY2024 — Georgia Parole Board; BOPP Annual Report FY2022 — Georgia Parole Board). At the same time, the Board considered fewer cases: 24,738 in FY19, 21,790 in FY20, and 16,255 in FY21, with total Board votes falling from 86,054 in FY19 to 59,354 in FY21 (BOPP Annual Report FY2019 — Georgia Parole Board; BOPP Annual Report FY2020 — Georgia Parole Board; BOPP Annual Report FY2021 — Georgia Parole Board). This contraction is not explained by a shrinking prison population: Georgia still holds roughly 50,000 to 53,000 people in state prisons (Georgia Department of Corrections Budget FY2026-FY2027; Georgia Incarceration Trends: Population, Demographics & National Context).

Against that contraction, parole completion rates look strong. The Board reports 66% completion in FY09, 72% in FY12, 74% in FY13, 72% in FY14, 70% in FY23, 72% in FY24, and 73% in FY25, consistently above national averages estimated between 45% and 60% (BOPP Annual Report FY2009 — Georgia Parole Board; BOPP Annual Report FY2012 — Georgia Parole Board; BOPP Annual Report FY2013 — Georgia Parole Board; BOPP Annual Report FY2014 — Georgia Parole Board; BOPP Annual Report FY2023 — Georgia Parole Board; BOPP Annual Report FY2024 — Georgia Parole Board; BOPP Annual Report FY2025 — Georgia Parole Board). This is the recidivism paradox: if a parole board releases fewer people and selects lower-risk individuals, success rates can rise while overall reentry opportunity shrinks. The state has not published a risk-profile analysis showing that the declining release cohort is comparable to those denied parole, leaving the high completion rate as much a measure of selection as of supervision.

Reentry Infrastructure: Housing, Supervision, and the Employment Gap

Nearly all people in Georgia prisons will eventually return to communities. Nationally, 95% of all incarcerated people will be released, most having received almost no programming or support (National Prison Reform Models & Georgia Comparison — Brennan Center 2026 Report). In Georgia, the release flow is substantial but supervised reentry is shrinking: FY25 saw 5,588 parole releases, while Georgia's total behind-bars population is 95,000 (BOPP Annual Report FY2025 — Georgia Parole Board; Georgia Incarceration Trends: Population, Demographics & National Context). Nationally, nearly 60% of formerly incarcerated people remain unemployed a year after release (National Prison Reform Models & Georgia Comparison — Brennan Center 2026 Report). Georgia does not publish equivalent post-release employment data for the full population leaving its prisons, only supervision-era snapshots.

Supervision data describes a system that invests in oversight but not economic integration. In FY09, Georgia parole officers made 447,234 face-to-face contacts with offenders, enrolled 9,394 parolees in drug treatment programs, and administered 99,941 drug tests (BOPP Annual Report FY2009 — Georgia Parole Board). Parolees under active supervision had a 78% employment rate, and the Board cited studies showing a 1% decrease in the chance of rearrest for every day a parolee is employed (BOPP Annual Report FY2009 — Georgia Parole Board). Housing and financial obligations add another layer of precarity. In FY09, the Board counted 28 approved Reentry Partnership Housing providers and 133 approved Transitional Housing for Offender Reentry providers, with 45 parolees in nursing homes (BOPP Annual Report FY2009 — Georgia Parole Board). Between the program's April 2006 start and FY09, 793 parolees had been placed at a cost of $1,097,600, while the Board estimated $8,609,927 was spent to hold 788 inmates past their parole release month, for $7,512,327 in avoided incarceration costs (BOPP Annual Report FY2009 — Georgia Parole Board). In the same year, offenders on parole paid $4,103,252.63, including $3,293,911.92 in supervision fees, $99,989.35 in restitution, and $709,351.36 in victim compensation (BOPP Annual Report FY2009 — Georgia Parole Board). The most recent dedicated federal reentry grant identified in state reports was $2 million in FY2002, directed to Albany, Atlanta, Augusta, Macon, LaGrange, and Savannah (BOPP Annual Report FY2002 — Georgia Parole Board). This is episodic money for a permanent reentry population.

Fiscal and Human Stakes: The Cost of Failed Reentry

Georgia's reentry failure is expensive. The state's prison system costs approximately $1.8 billion annually, with Amended FY 2026 total funds of $1,799,204,979 (Recidivism & Reentry Failures in Georgia; Georgia Department of Corrections Budget FY2026-FY2027). The Senate's approved FY2027 GDC budget is $1,770,903,120 in total state funds, an increase of $8,833,156 over the Governor's proposal (FY2027 GDC Approved Budget — HB 974 Senate Appropriations Committee Substitute). The State General Funds component is $1,762,261,281 (FY2027 GDC Approved Budget — HB 974 Senate Appropriations Committee Substitute). An earlier Parole Board analysis found a one-percent drop in recidivism translates to $6 million in taxpayer savings (BOPP Annual Report FY2002 — Georgia Parole Board). By that measure, the entire FY2027 Senate add-back of $8.8 million is barely more than the fiscal impact of a single-point improvement in recidivism, while the system still lacks durable reentry appropriations.

Georgia's fiscal imbalance between incarceration and community supervision is longstanding. A 1978 sentencing-alternatives study found that in FY 1977 Georgia prison incarceration cost $3,858 per inmate per year ($10.57 per day), while immediate probation cost $145 per probationer per year ($0.40 per day) — a more than 26-to-1 ratio (Utilization of the Split Sentence and Shock Probation as Sentencing Alternatives and Implications for Their Expanded Use in Georgia (NCJ 52262, 1978)). That early cost gap has since scaled into the current $1.8 billion prison appropriation, while probation and other community supervision remain comparatively low-cost but underused alternatives.

The evidence base for returns on reentry investment is strong. RAND's analysis of correctional education estimates $4 to $5 saved in reincarceration costs for every $1 spent and found participants had 43% lower odds of recidivating, a 13-percentage-point reduction, and 13% higher odds of post-release employment (Evidence-Based Prison Rehabilitation: Verified Outcomes, Recidivism Baselines, and Georgia Program Delivery (Research Pass, Aug 2026)). The deeper historical roots are also fiscal: by 2001, the federal Violent Offender Incarceration and Truth-in-Sentencing grant program had disbursed $2.7 billion to 29 jurisdictions to build longer carceral capacity (Truth in Sentencing & Fiscal Impact: The $40 Billion Story). That architecture brought more people into prison for longer periods but did not fund a parallel post-release infrastructure, a mismatch now visible in Georgia's release and recidivism data.

Aging, Life Sentences, and the Delayed Reentry Horizon

The population Georgia sends back to communities is increasingly old and long-sentenced. Of 47,391 active inmates in the GPS database, 12,777 — 27.0% — are age 50 or older (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). Those age 55 or older number 8,694, or 18.3% (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). In 2024, 10,392 people in Georgia prisons — 20% of the reported population — were serving life sentences: 7,679 life with parole, 1,949 life without parole, and 764 virtual life sentences of 50 years or more (A Matter of Life: Life and Long-Term Imprisonment in the United States — Georgia in National Context (2024 Census)). One in five prisoners will not reach reentry under current policies, and many who do will be elderly.

Age also drives the fiscal and health dimensions of release. CDCR data shows people 60 and older were 14% of the population but accounted for 27% of health system expenses in 2024 (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). Estimated annual health care costs run from $87,552 at ages 60 to 69, to $139,056 at 70 to 79, and $237,325 for those 80 and older (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). Compassionate release can change those trajectories, but access depends heavily on counsel: under Illinois' Joe Coleman Medical Release Act, 60% of applicants with legal counsel were granted release compared with 19% without (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). In 2022, 78 Illinois medical release applications produced 28 approvals and 50 denials (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). Georgia's own data shows 27,471 people — 51% of the custody population — with no hepatitis C test result on file as of June 2026 (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape), a basic health-reentry indicator that remains unknown for half the system.

What Works: Evidence-Based Models and Georgia's Delivery Gap

Evidence-based programming changes outcomes when it is delivered. A 2009 evaluation of Thinking for a Change found 23% of participants recidivated compared with 36% in a control group during a six-month follow-up (Evidence-Based Rehabilitation Curricula: Cognitive-Behavioral, Trauma-Informed, and Mentorship Programs for Correctional Settings). Correctional education has stronger long-term evidence: 43% lower odds of recidivism and higher post-release employment (Evidence-Based Prison Rehabilitation: Verified Outcomes, Recidivism Baselines, and Georgia Program Delivery (Research Pass, Aug 2026)). Other systems provide the delivery infrastructure: the Prison Entrepreneurship Program invites more than 6,000 men across 80 TDCJ units annually, and California's Innovative Programming Grants funded 299 programs since 2014, with a $12 million current cycle for 2025 to 2028 (Prison Program Structure Models: Cohorts, Tiers, Mentorship Pipelines, and Outcomes from Leading U.S. Correctional Programs; California Prison Programs: From Brown v. Plata to National Model — CDCR's Rehabilitative Transformation).

Georgia ran its own randomized field experiment on cognitive skills programming. The Georgia Cognitive Skills Program — the first known statewide effort to employ cognitive skills training with parolees — began in 1997, when Elizabeth Fabiano and Frank Porporino trained approximately 50 parole officers across Georgia as cognitive coaches; since implementation, a total of 30 parole districts and three correctional institutions have administered Reasoning and Rehabilitation (R&R) (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)). Phase I randomly assigned 468 parolees to an experimental group (N=232) or a control group (N=236) across 16 parole districts. Phase II, operating from July 1998 through April 2000, randomized 1,193 offenders at 28 sites — 25 parole districts and three correctional institutions — into a treatment group (N=609) and a control group (N=584) (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)).

The Phase II treatment group was not a low-need sample. Men made up 91.3% of participants; 69.0% were between ages 18 and 35, with an average age of 32; and 68.5% were nonwhite — 409 participants in all, one Native American and the rest African American (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)). At intake, 79.8% were assessed as maintaining at least a minimum standard of living, but 50.9% were not employed full time at the time of their arrest. Educationally, 65.4% had not completed high school, and 26.4% scored below a fifth-grade reading level on the WRAT. The average IQ was 100.8, but 8.2% scored below 80. Most were single (60.8%), and 56.8% had a documented history of substance abuse. Half (49.2%) were classified as medium/high risk by the study's risk score. Prior system involvement was extensive: 58.1% had multiple prior incarcerations, 44.1% had prior parole revocations, only one participant (0.2%) had no prior felony convictions, and 48% had three or more prior felony convictions. Prior property convictions were most common (72.4%), followed by drug possession (62.0%), violent convictions (43.9%), and drug sale convictions (42.0%) (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)). The median number of prior property convictions for both groups was two, though a greater proportion of the experimental group had eight or more prior property convictions than the control group (23.3% vs. 19.0%, p < .05).

Screening departed from the program's own eligibility logic. Despite criteria excluding sex offenders, 18 experimental-group members (3.0%) had sex-crime conviction histories, and one had up to 10 prior sex-crime convictions (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)). By security classification, 4.9% of the experimental group were I-level 2, 55.2% were I-level 3, and 40.0% were I-level 4. On the Jesness Inventory, 32.2% were classified as Aggressives, 27.9% as Dependents, 22.4% as Situationals, and 17.5% as Neurotics. Low-risk offenders constituted 50.8% of the experimental group (299 people), and 48 experimental-group members had IQs below 80. Overall, the evaluators concluded 10.2% of the experimental group were inappropriate for the program because of low cognitive ability or sex-offender history; when the 299 low-risk offenders are added, over half (54%) of the experimental group could be considered inappropriate. Still, the random assignment held up: group distributions were nearly identical on background characteristics, and only one statistically significant difference emerged, the property-conviction imbalance noted above (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)).

Delivery fidelity was mixed. At program start, class size ranged from 6 to 28 participants, with a mean of 12 and a median of 10; 76.9% of classes exceeded the recommended maximum of six participants (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)). By session 30, class size had shrunk to between 2 and 20 participants, but 40.4% of classes still exceeded the recommended size. Start-up losses were significant: 80 offenders (13.1% of the experimental group) did not start the program, and of those who did begin, 24.1% were no longer participating by the 30th class session. Overall, 382 of the 609 experimental-group members completed Phase II, a completion rate of 62.7%; Phase I completed at 59.6%. Among Phase II program starters only, completion rates ranged from 33% to 100%, averaging 72.0% with a median of 73.2%. Observers rated general issues and class structure as less than satisfactory for 41.3% of groups at session 17 and 22.2% at session 30. Coaches met expectations in prosocial interpersonal skills in 78.3% of groups at session 17 and 91.1% at session 30. Most participants (68.9%) said coaches read from the manual a good deal of the time, but 97.2% rated the classes as relevant or very relevant to their daily lives (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)).

The process evaluation's findings are a warning about scale and selection. The evaluators concluded that randomization procedures were implemented successfully, but that offenders were not screened into Phase II as carefully as they might have been, with IQ the only eligibility assessment consistently aiding screening; they also noted nearly all program groups exceeded the recommended class size of four to six members (The Georgia Cognitive Skills Experiment: Process Evaluation, Phase Two (NCJ 248529, 2001)). If the Phase I outcome results hold for Phase II, the inclusion of low-risk offenders may not have been detrimental, but screening was not systematically aided by risk or eligibility assessments. For Georgia's current reentry system, the experiment is an early, concrete example of the same delivery gap that still recurs in its programming: rigorous design and real participation, but weak selection, oversized classes, substantial attrition, and no durable statewide infrastructure.

Historical sentencing-alternative experiments also carried warning signs for Georgia. In FY 1976, split sentences accounted for 5.6% of all convicted defendants sentenced in U.S. District Courts, but only 4.0% of convicted defendants sentenced in Georgia District Courts (Utilization of the Split Sentence and Shock Probation as Sentencing Alternatives and Implications for Their Expanded Use in Georgia (NCJ 52262, 1978)). Ohio's shock probation data from the same era showed a mixed record: among 554 shock probationers released from Ohio prisons in 1969, plus female releases from 1966–1970, 47.7% were arrested at least once after release, 31.3% were subsequently convicted of a felony within five years, and 24% served at least one additional prison sentence (NCJ 52262, 1978). The population receiving shock probation was not low-need: 40% had prior convictions, 12.1% had served prior prison terms, and 20% were classified as potentially violent offenders (NCJ 52262, 1978). The same cohort was young and economically precarious — 49% were ages 18–22, 78% had not finished high school, 76% were unskilled, and 53% of men and 69% of women were unemployed (NCJ 52262, 1978). Release practice also fell short of the statute's design: in one 1970 sample, 17% of timely filed shock-probation motions were not ruled on until after the statutory deadline, and in a 1973 study, 15.6% of shock probationers were released after the 130-day limit (NCJ 52262, 1978). California's period data pointed in the same direction: only 52% of people given probation with jail and 41% of those given straight jail remained violation-free during their first year, compared with 66% for straight probation (NCJ 52262, 1978). California's Section 1203.03 diagnostic study estimated more than $1.6 million in monetary savings for 1970, but those savings came from diversion to treatment, not from jail as a condition of supervision (NCJ 52262, 1978). These older results caution that a short incarceration "shock" does not automatically improve reentry outcomes and may select people with higher needs into supervision.

Georgia's institutional base is uneven. The state operates 85 correctional facilities and highlights 49 ACA-accredited facilities, but accreditation measures operational standards rather than rehabilitation outcomes (2024 Georgia Senate Study Committee on the Department of Corrections — Final Report (SR 570)). Research-pass inventories of Georgia's evidence-based and postsecondary programs describe institutional fragility rather than stable, scaled delivery (Evidence-Based and Postsecondary Rehabilitation Programs in Georgia Prisons: Inventory, Funding Sources, and Institutional Fragility (Research Pass, Aug 2026)). The result is a delivery gap that mirrors the parole numbers: Georgia has access to programmatic knowledge and national models that reduce recidivism, but release volume has collapsed while program capacity remains fragmented. The Cognitive Skills Experiment itself is a case in point — strong randomization and real participation, but uneven screening, inflated class sizes, and enough attrition to undercut scale. Until Georgia ties its reentry spending, release decisions, and outcome metrics to the same evidence base, its official low recidivism rate will continue to measure the system's narrow definitions, not its public safety performance.