Reform Models & Programs

The Evidence Base: What Works in Correctional Rehabilitation

Structured cognitive-behavioral interventions remain the strongest national evidence for reducing recidivism. A 2009 evaluation of the Thinking for a Change curriculum found that 23% of participants recidivated within six months, compared with 36% of a control group—a statistically significant reduction (Evidence-Based Rehabilitation Curricula: Cognitive-Behavioral, Trauma-Informed, and Mentorship Programs for Correctional Settings). The default release trajectory is far worse: the Bureau of Justice Statistics 2005 release cohort saw 67.8% rearrested within three years, 76.6% within five years, and 55.1% reconvicted within five years (A Sense of Purpose as a Driver of Rehabilitation in Incarcerated People: An Evidence Brief). RAND's cost analysis estimates $4 to $5 saved in reincarceration costs per $1 spent on correctional education (Evidence-Based Prison Rehabilitation: Verified Outcomes, Recidivism Baselines, and Georgia Program Delivery).

California and Texas have turned this evidence into institutional funding structures. California's Innovative Programming Grants program has funded 299 programs since 2014 and operates a 2025–2028 cycle at $12 million—$4 million per year for three years (California Prison Programs: From Brown v. Plata to National Model — CDCR's Rehabilitative Transformation). Texas's Prison Entrepreneurship Program reaches over 6,000 men annually across 80 TDCJ units through a fixed cohort leadership and accountability curriculum (Prison Program Structure Models: Cohorts, Tiers, Mentorship Pipelines, and Outcomes from Leading U.S. Correctional Programs). These models share features GPS research repeatedly identifies as protective: consistent dosage, identity transformation, peer mentorship, and post-release continuity.

The GPS brief A Sense of Purpose as a Driver of Rehabilitation in Incarcerated People argues that meaningful activity—education, mentorship, employment pathways—is not an amenity but a criminogenic protective factor. That reframing shifts the burden from moral argument to correctional outcomes: a system that ignores purpose is not simply harsher; it is measurably less effective at reducing future crime.

Georgia's Fiscal Envelope vs. Its Programming Outputs

Georgia's corrections budget has reached historic highs without producing the infrastructure of a rehabilitative system. GDC's actual budget was $1,526,654,104 in FY2024, peaked at $1,913,888,054 in FY2025, and stands at $1,770,903,120 in total approved state funds for FY2027, including $1,762,261,281 in State General Funds (GDC Budget Baseline FY2025–FY2027; FY2027 GDC Approved Budget — HB 974 Senate Appropriations Committee Substitute). In 2025 alone, the General Assembly added approximately $634 million in new corrections spending across the amended FY2025 and FY2026 budgets, the largest increase in state history (Georgia's $600 Million Prison Spending Infusion: An Accountability Analysis). The total annual appropriation is roughly $1.8 billion per year (GDC Budget Baseline).

The line items that shape daily life remain at subsistence levels. The Food and Farm Operations budget was $30.9 million in FY2024, $31.7 million in FY2025, $31.1 million amended in FY2026, and $31.3 million approved for FY2027—about $0.54 to $0.55 per meal, or 14.8% of the American Correctional Association's recommended $3.66 per meal (GDC Budget Baseline). That figure represents a roughly 60% real-terms decline in per-meal spending since 2015. GDC Mission vs. Reality: The Rehabilitation That Does Not Exist documents the consequences: an agency with a $1.5 billion to $1.9 billion budget whose core rehabilitative function remains largely unbuilt.

This is not solely a resource problem; it is an allocation and accountability problem. The $634 million infusion arrived without an evaluation framework that ties spending to evidence-based program expansion, staffing stabilization, or measurable recidivism reduction (Georgia's $600 Million Prison Spending Infusion). In the absence of California-style innovation grants or Texas-style cohort funding, Georgia's budget growth has been primarily an expansion of custody capacity rather than a reform model.

Georgia has attempted structured incentive programming before, with documented implementation strain. The Earned Time System: First Annual Evaluation (NCJ 67770, 1977) examined the state's earned-time model: 53 institutional privilege lists drawn from a master list of 68 privileges; privilege lists ranged from 1 to 13 items and averaged 8 privileges at state institutions versus 4 at county institutions, with eight county institutions offering only two privileges and four state institutions offering fewer than six. Telephone calls were available at 48 of 53 institutions (91%), extra or extended visits at 41 (77%), supervised fishing trips at 23 (43%), detail changes at 17 (32%), and a day off work or activity at 15 (28%). Across 11 sampled institutions, 1,396 privileges were recorded—566 telephone calls (55%), 272 visiting privileges (26%), and 51 fishing trips (5%)—and phone and visit privileges accounted for 81% of all grants, with a mean of 8.8 days between award and redemption. The average monthly privilege award rate was 0.11 per inmate, with Good Time inmates showing a slightly higher rate (0.11) than Earned Time inmates (0.09). Yet the system operated inside a capacity crisis: at enactment, Georgia's institutions housed more than 11,000 inmates in facilities designed for 4,200; by September 1, 1977, the total accountable population was 11,697, of whom 4,737 (40%) were Earned Time inmates, and state institutions held 3,670 ETS inmates—42% of their total population. Across sampled institutions, 568 Earned Time inmates were permanently assigned longer than 60 days; the evaluation's institutional file sample covered only 156 cases, or 28% of that group. At one Southwest District institution, counselor caseloads reached 190 each. That is not a failed idea; it is an underfunded and underdocumented one.

Unit Design, Violence, and the Cost of Isolation

Some of the most instructive U.S. reform models are small. Pennsylvania's 64-bed Little Scandinavia unit at SCI Chester cost approximately $310,000 to create and recorded almost no violent episodes in 2024 (Scandinavian-Inspired Prison Reform in U.S. States: Pennsylvania, California, and Connecticut Pilot Programs; Comparative Solutions Evidence Base: Prison Reforms That Have Demonstrably Worked in Other States and Nations). The unit translates Scandinavian normalization—single cells, communal space, and staff trained in de-escalation—into a standard American prison setting. Texas's Prison Entrepreneurship Program operates at a different scale, using an 80-unit distance-education screen and fixed cohorts to deliver high-dosage programming to more than 6,000 men annually (Prison Program Structure Models). Both suggest that structure, not scale, determines whether a unit becomes a program or a warehouse.

Georgia's violence data moves in the opposite direction. GPS documented 48 killings in Georgia prisons from 2018 to 2020 and 94 from 2021 to 2023—a 95.8% increase (Who Is Responsible for Violence in Georgia's Prisons? An Evidence-Based Analysis). That 142-death count is confirmed across GPS collections, including Legal Access in Georgia Prisons. For 2024, GDC acknowledged 66 homicides, while the Atlanta Journal-Constitution confirmed at least 100 in the same year (Gang Separation as Violence Reduction Strategy: Georgia vs. Other States). Solitary confinement remains the default response: 50% of prison suicides nationally occur among people in solitary, who constitute only 6–8% of the population, and 78% of prisoners in Georgia's Special Management Unit had been held more than two years as of July 2017 (Solitary Confinement & Restrictive Housing).

The contrast is not merely philosophical. A purpose-built unit like Little Scandinavia costs less than many individual corrections settlements. Georgia has already paid $50,633,556 across at least 261 settled claims, a figure the GPS grievance collection describes as a floor, not a total (The Grievance Machine: How GDC's Internal Complaint System and the PLRA Exhaustion Requirement Suppress Accountability). The cheap option—programmatic housing—is also the fiscally conservative option.

Parole and Community Supervision: A Proven System in Retreat

Georgia's supervision records expose a paradox. The state leads the nation in felony probation population, with 191,000 people on felony probation—more than any other state—and a probation rate more than triple the national average (Georgia Probation & Community Supervision: Reform, Costs & Outcomes). Georgia's incarceration rate of 881 per 100,000 residents is the seventh highest in the nation, higher than any country in the world except El Salvador (Recidivism & Reentry Failures in Georgia), and its prison population ranked fifth nationally in FY2006 (BOPP Annual Report FY2006 — Georgia Parole Board). By those measures, Georgia is a system built on mass supervision and mass incarceration.

Yet the Board of Pardons and Paroles' own completion data shows community supervision can succeed with structure and support. In FY2009, 66% of Georgia parolees completed supervision against a national rate of 45%; FY2011 was 71% versus 51%; FY2014 reached 72% against a national average of 59% (BOPP Annual Report FY2009; FY2011; FY2014). Those are historically significant differences. Agency research from FY2002 found that a one-percent drop in recidivism translates to a $6 million taxpayer savings (BOPP Annual Report FY2002).

The problem is scale and front-end overuse, not capacity. Georgia uses probation as a mass disposal mechanism while steadily shrinking parole releases—from 6,245 in FY2022 to 5,863 in FY2023 and 5,588 in FY2025 (BOPP Annual Report FY2022; FY2023; FY2025). The 1977 Earned Time evaluation's parole file review offers an early warning about decision-making under those pressures: among 46 sampled Earned Time inmates due for parole, a Parole Review Summary or Inmate Evaluation was present in all 46 files, but Performance Plans appeared in only 13%, Quarterly Review Summaries in 24%, and staff recommendations correlated with the Parole Board's actual decision only about 50% of the time. At the same time, the custody population is aging: 27.0% of active inmates are age 50 or older, and 18.3% are 55 or older (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). The U.S. Sentencing Commission found older offenders recidivated at 21.3%, less than half the 53.4% rate for offenders under 50 (Comparative Solutions Evidence Base). Georgia is moving releases down even as its lowest-risk, highest-cost population grows.

Oversight, Accreditation, and the Rule of Law as Reform

Reform models are not only programs; they are accountability structures. New Jersey operates a fully independent corrections ombudsperson for about $2.8 million per year with 26 staff (Comparative Solutions Evidence Base: Prison Reforms That Have Demonstrably Worked in Other States and Nations). That is a modest line item inside Georgia's $1.8 billion corrections budget, but Georgia has no comparable independent oversight body. The GPS collection Nobody Watches the Watchmen: Independent Correctional Oversight Models and the Statutory Path for Georgia treats that absence as a structural failure, while The Grievance Machine documents how GDC's internal complaint system and the Prison Litigation Reform Act's exhaustion requirement suppress accountability before external review can occur.

Georgia leans on American Correctional Association accreditation instead. The Department of Corrections reports 49 ACA-accredited facilities (2024 Georgia Senate Study Committee on the Department of Corrections — Final Report (SR 570)). Accreditation checks practices against minimum standards, but it does not constitute independent investigation of deaths, excessive force, or retaliation. The same agency that holds accreditation for 49 facilities spends $0.54 per meal against the ACA's own $3.66 recommendation—only 14.8%—while the Senate Study Committee report operates without binding enforcement authority (GDC Budget Baseline; 2024 Senate Study Committee). Accreditation is a floor, and on basic conditions the agency is below it.

The settlement record quantifies the accountability gap. Georgia paid at least $50,633,556 across at least 261 settled claims, a total that does not include defense costs or non-monetary harm (The Grievance Machine). That amount is equivalent to about 18 years of a New Jersey-style independent ombudsperson at $2.8 million annually. The state is paying for harm after the fact rather than investing in oversight before it occurs.

Data Gaps, Contradictions, and What Reform Requires

Any reform model depends on trustworthy data, and Georgia's data systems are failing. GDC has no hepatitis C test result on file for 27,471 people—51% of its custody population—as of June 2026 (Aging Prison Population & Compassionate Release: Georgia Data, National Research, Fiscal Analysis, and Legislative Landscape). A system that cannot report hepatitis C prevalence cannot target health programming or compassionate release. The same pattern appears in mortality and violence data: GDC acknowledged 66 homicides in 2024, while the Atlanta Journal-Constitution confirmed at least 100 (Gang Separation as Violence Reduction Strategy: Georgia vs. Other States). If the baseline count is understated by a third, evaluations of violence reduction programs inherit the same error.

Georgia's documentation failures have a forty-year-old baseline. The 1977 Earned Time System evaluation sampled 156 institutional files and found the Supervisor's Evaluation—the primary periodic assessment document—in only 40% of files (63 of 156). Performance Plans were present in 80% of files, but only 34% (53) were signed and dated by the Classification Committee; 35% (55) were signed but not dated; 11% (17) were neither signed nor dated; and 20% (31) had no Performance Plan at all. Quarterly Review Summaries were present in 70% of files. Participation status forms were present in 98% (153), Performance Assessment Sheets in 92% (143), Intake Assessments in 90% (141), and Extended Assessments in 80% (125), indicating that the front end of the system was documented while the review and evaluation stage fell away. The same evaluation sampled 46 of 175 Earned Time inmates considered for parole: a Parole Review Summary or Inmate Evaluation was present in all 46 files, but Performance Plans appeared in only 13% (6), Quarterly Review Summaries in 24% (11), DOR Case Studies in 46%, Complete Performance Assessment Sheets in 44%, Employment Interview Data in 35%, and Computerized Psychological Profiles in 30%. Staff recommendations correlated with the Parole Board's ultimate decision only about 50% of the time; Case Managers averaged 5.1 comments on Parole Review Summaries and 10.4 on Inmate Evaluations, while Wardens averaged 2.1 comments on each. Program-output data were similarly incomplete: Earned Time inmates completed only 1.7% of selected vocational program terminations successfully over a 12-month period, academic completion for ETS inmates ranged from 21% to 27% across four quarters, and transfers accounted for 43% of ETS vocational terminations and 38% of combined program terminations, compared with 27% for the first-year cohort. The evaluation also recorded 161 Use of Force reports at state institutions in FY76 and 129 in FY77—monthly averages of 13 and 10.8—while county institutions recorded 12 and 10 in the same years.

The evaluation also found privilege distribution disparities: lower- and medium-security inmates received 755 of the privileges awarded, or 74%, while comprising only 44% of the sampled population; white males represented 38% of the population but received 48% of privileges; and within lower-security classifications, Black inmates were 47.4% of the population but received 70% of privileges, while white inmates were 42.6% but received 75.7%. These are not 2024–2027 anomalies; they are long-lived data and equity failures that undermine every subsequent claim that a program worked.

Budget data reveal a different kind of gap. Food and Farm Operations sits at $0.54 per meal across FY2024–FY2027, roughly 15% of the ACA recommendation and a 60% real-terms decline since 2015 (GDC Budget Baseline FY2025–FY2027). That is not only a programming failure; it is a basic-conditions failure that shapes whether any curriculum, cohort, or mentorship model can operate. The $634 million in 2025 new spending was the largest corrections increase in state history, but the accountability analysis documents no evaluation mechanism connecting that money to programming outputs or recidivism benchmarks (Georgia's $600 Million Prison Spending Infusion: An Accountability Analysis).

What the evidence demands is not mystery. Structured cognitive-behavioral programs reduce recidivism; small purpose-built units reduce violence; independent oversight reduces impunity; and parole supervision with services succeeds in Georgia at rates above the national average. The missing ingredient is not evidence. It is political will to shift money from reactive custody to proactive rehabilitation, and to build data systems that tell the public whether the spending works.