Pay Prisons to Empty Beds

CoreCivic's revenue per compensated man-day for the first half of 2025, per its latest 10-Q filing with the SEC, was $107.90, which multiplied across 365 days works out to roughly $39,400 per inmate per year flowing into a company that operates 44 correctional and detention facilities with a design capacity of 67,785 beds, posted an operating margin of 26% last quarter, and watched its stock price climb 142% over the past twelve months while achieving none of those gains by keeping a single released prisoner from committing another crime.

CoreCivic and GEO Group, the duopoly controlling approximately 75% of all private prison beds in the United States, earn revenue one way: per diem payments for each body in each bed on each day, with many of their contracts including minimum occupancy guarantees of 90% or even 100%, meaning taxpayers pay for empty beds when the population drops below the contractual threshold regardless of whether the decline reflects a reduction in crime, a change in sentencing policy, or the successful rehabilitation of people who would otherwise be cycling back through the system that profits from their return. An inmate who comes back is a customer retained. An inmate who stays out is revenue evaporated.

Private prisons hold about 8% of the U.S. prison population, but the incentive problem reaches far beyond the private sector into the architecture of public corrections itself. Public prison systems set their budgets based on projected inmate populations, which means wardens who reduce their headcount risk losing staff positions, programming funds, and political relevance, while corrections officer unions lobby against decarceration because fewer inmates translates directly to fewer jobs, and legislators approve construction bonds for new facilities because breaking ground is visible in a way that preventing recidivism never will be. The entire system has arranged its financial plumbing so that money flows toward incarceration and drains away from the thing incarceration is supposed to produce: less crime.

The Recidivism Tax

The Bureau of Justice Statistics has tracked released prisoners for decades and the numbers are remarkably stable in their bleakness: of state prisoners released in 2012, 66% were rearrested within three years, and extending the follow-up to ten years pushed the figure to 82% with at least one new arrest, meaning the system's $80-billion-a-year investment in warehousing human beings produces, as its most consistent output, the next arrest of the person it just released.

Norway's two-year reconviction rate is 20%, according to a systematic review of global recidivism data published in PLOS ONE, and Finland's is 36%, and Denmark's is 29%, and Iceland's is 27%, and the Nordic countries accomplish these numbers while incarcerating roughly 60 to 70 people per 100,000 population compared to the United States' rate of 698 per 100,000, which is more than ten times the Finnish rate applied to a system that produces worse outcomes by every measure that has ever been studied.

The difference is not cultural mystery or Scandinavian exceptionalism. Before Norway overhauled its prison system in the 1990s to prioritize rehabilitation, Norwegian recidivism ran as high as 70%, a number indistinguishable from the current U.S. rate, and the shift followed three principles: normality, where life inside the prison resembles life outside as closely as possible; progression, where inmates gradually earn more freedoms by meeting goals and taking on responsibilities; and reintegration, where the entire purpose of the sentence is preparing the person for a productive return to society rather than punishing them in ways that make them less capable of functioning once they leave. The 70% rate dropped to 20% within a decade. The principles are neither complicated nor mysterious, but they require a system whose financial incentives reward the transformation rather than punishing it.

The Model That Already Exists

Hospitals used to work this way, collecting fee-for-service payments that rewarded volume over outcomes, paying doctors for each procedure performed regardless of whether the patient recovered or deteriorated, which created a financial structure where sicker patients meant more billable encounters and more procedures meant more revenue and nobody in the system had a financial reason to keep people healthy because healthy people do not generate claims. Then came value-based care, and Medicare started bundling payments around outcomes, and hospitals that reduced readmissions got bonuses while hospitals with high readmission rates got penalized, and the Hospital Readmissions Reduction Program launched in 2012 cut 30-day readmission rates across targeted conditions by over 8% in its first five years through a mechanism that was disarmingly simple: make the institution's revenue depend on whether the patient stays well after leaving.

Corrections has the same structure and the same opportunity for transformation. Replace per diem payments with outcome-based contracts that pay a base rate for housing and care, then layer on performance bonuses for every year an ex-inmate remains crime-free post-release, while penalizing facilities whose recidivism rates exceed a risk-adjusted benchmark. When the prison's financial health depends on the released prisoner's civic health, every internal budget decision shifts: the education program stops being a cost center and becomes a revenue driver, the vocational training workshop stops competing with bunk space for square footage and starts competing for funding priority, and the reentry coordinator's salary moves from "nice to have" on the annual budget to "revenue-generating headcount" that directly protects the performance bonus.

The Evidence on What Prisons Could Be Doing

The RAND Corporation conducted the largest meta-analysis ever performed on correctional education, examining 58 studies that controlled for selection effects between participants and non-participants, and their central finding was that inmates who participate in education programs have 43% lower odds of returning to prison than those who do not, with employment after release running 13% higher overall and vocational training specifically increasing post-release employment by 28%, all achieved at a cost of between $1,400 and $1,744 per inmate that the study estimated saves $8,700 to $9,700 per inmate in avoided reincarceration costs within the first three years alone, delivering a four-to-one return on investment that sits on the table, unclaimed, because nobody in the current system captures the upside of keeping people out.

Under per diem contracts, CoreCivic has no financial reason to spend $1,744 educating an inmate who, upon recidivating, will generate another $39,400 in annual revenue. Under outcome-based contracts, that $1,744 becomes the most profitable investment the company can make because the 43% reduction in recidivism odds directly protects the performance bonus and shields the facility from the recidivism penalty, transforming the same arithmetic from a disincentive into the single highest-return line item in the operating budget.

Massachusetts tested this principle through a Pay for Success contract with Roca, a nonprofit providing intensive outreach, life skills training, and employment services to high-risk young men in the justice system, sometimes making contact 40 to 50 times before a young man engaged with the program. The state's own analysis found that addressing both substance use and educational needs simultaneously cut recidivism to 7.8%, compared to 19.7% for those whose needs went unaddressed, with the contract structure ensuring the state paid only for verified outcomes while investors bore the risk if the program failed to deliver.

What We Learned from the Failures

The Rikers Island Social Impact Bond was the first of its kind in the United States, and it failed completely. Goldman Sachs invested $9.6 million in the Adolescent Behavioral Learning Experience program, a cognitive behavioral therapy intervention for 16- to 18-year-olds detained at Rikers, with Bloomberg Philanthropies guaranteeing $7.2 million of the investment and a threshold for success set at a 10% reduction in recidivism. The Vera Institute's independent evaluation found the program produced no reduction in recidivism whatsoever, Goldman Sachs absorbed a $1.2 million loss, and the program was discontinued in August 2015.

But here is what matters about that failure, the part that most critics of social impact bonds overlook entirely: the financial structure worked exactly as designed, killing a bad program fast and protecting public funds while doing it, because in a traditional government contract that $9.6 million would have been spent regardless of outcomes, and the failed program might have continued for years before anyone audited the results or mustered the political will to shut it down, which is precisely how most government-funded corrections programs operate today, spending money year after year with no mechanism to distinguish the programs that reduce crime from the programs that accomplish nothing at all.

The Peterborough Social Impact Bond in the United Kingdom told the opposite story and validated the same structure from the other direction. Launched in 2010 as the world's first SIB, it funded "The One Service," which provided tailored rehabilitation support to short-sentence prisoners after release from HMP Peterborough, and the final evaluation found a 9% reduction in reconviction events across both cohorts, exceeding the 7.5% threshold required to trigger investor payment, with the 17 investors receiving their capital back plus a 3% annual return while reoffending declined and taxpayer money was saved rather than spent.

Two social impact bonds in two countries. One program failed and investors absorbed the loss with no burden on the public. One program succeeded and investors earned a modest return funded entirely by savings the program generated. In both cases the financial structure performed precisely the function that corrections contracts should perform but currently do not: channeling money toward what works and cutting it off from what does not, with private capital bearing the risk of experimentation rather than taxpayers.

The Strongest Counterargument

The most serious objection is cream-skimming, the possibility that if prisons are paid based on recidivism outcomes they will cherry-pick low-risk inmates and reject high-risk ones, gaming the system to post spectacular numbers while dumping the hardest cases on public facilities that receive no performance bonuses at all, and the objection is entirely legitimate because it describes exactly what happened in early healthcare value-based payment models before risk adjustment was introduced. The fix is the same one Medicare now uses: adjust the benchmarks to account for population risk, so that a facility housing violent repeat offenders gets evaluated against the expected recidivism rate for violent repeat offenders rather than against the system average, and a facility that takes easy cases receives a correspondingly lower benchmark and a smaller bonus opportunity. The methodology exists through validated assessment tools like the Level of Service Inventory and the Ohio Risk Assessment System, though it requires transparent auditing to catch and correct the documented racial biases that plague existing instruments like COMPAS, which Dressel and Farid demonstrated in 2018 performs no better than untrained Mechanical Turk workers given only a defendant's age and criminal history.

Measurement lag presents a second challenge because recidivism unfolds over years and quarterly bonus payments cannot depend on three-year outcomes, but the hospital analogy holds once again: a prison outcome-based contract could deploy a tiered payment structure with process bonuses for education enrollment and vocational certification in year one, intermediate outcome bonuses for employment verification and no-arrest milestones at six and twelve months post-release, and full outcome bonuses at the three-year mark, mirroring the way Medicare supplements its long-term outcome measures with interim process and quality indicators that keep the financial signal flowing before the final results arrive.

The Original Calculation Nobody Runs

CoreCivic's 2025 annual report shows total revenue of $2.2 billion across a design capacity of 67,785 beds at an average occupancy of 77.6%, which means roughly 52,600 beds generating revenue on any given day at an operating margin of approximately 25%, implying around $550 million in annual operating profit extracted from a system whose paying customers are taxpayers and whose product, if the word means anything at all, is supposed to be a safer society. If every CoreCivic inmate received the $1,744 education program that the RAND meta-analysis found reduces recidivism by 43%, the total cost would run roughly $92 million per year against the BJS baseline showing approximately 45% of released inmates returning within three years, producing an expected reduction in returning inmates of approximately 19 percentage points, from 45% down to about 26%, which represents thousands of human beings who do not cycle back through a revolving door that currently has no financial reason to stop spinning.

Under the current per diem model, those non-returning inmates are revenue lost from the income statement. Under an outcome-based model, each one triggers a performance payment funded from the state's avoided incarceration costs of roughly $39,000 per year per inmate, and even a modest outcome bonus of $5,000 per non-recidivating individual would generate tens of millions in performance revenue for the operator while saving taxpayers hundreds of millions in prison costs, probation costs, court costs, policing costs, and the economic devastation inflicted on communities where released prisoners cycle through arrest and reincarceration rather than employment and stability and the slow, unglamorous work of rebuilding a life that the system they just left did nothing to prepare them for.

Limitations

This analysis relies on the RAND meta-analysis's 43% figure, which aggregates studies of varying methodological rigor and whose strongest effects emerged from smaller sample sizes that may not generalize to population-level implementation. The BJS recidivism data tracks rearrest rather than reconviction, inflating the apparent failure rate because many arrests never produce charges or convictions. Norwegian comparisons involve a country of 5.5 million people with fundamentally different demographics, drug policy, firearms prevalence, and social safety net infrastructure, and translating Nordic outcomes to the United States requires accounting for vast differences in baseline poverty, racial inequality in policing, and the sheer scale of the American carceral system, which incarcerates more people than any nation in the history of the world. Risk adjustment for outcome-based corrections contracts remains theoretically sound but operationally untested at national scale, and the documented racial biases in existing risk assessment tools could produce contracts that systematically underpay facilities serving predominantly Black and Latino populations unless robust auditing infrastructure is constructed before the first contract is signed.

What You Can Do

If you vote: Ask your state legislator one question at the next town hall or constituent meeting: "What percentage of inmates released from state prisons last year were rearrested within three years, and what is the state's concrete plan to bring that number down?" Most will not know the answer because the system does not force them to learn it, and demanding that corrections departments publish annual recidivism scorecards by facility, the same way hospitals now publish readmission rates under federal mandate, is the single most effective lever a constituent can pull to make invisible failures visible.

If you run a business: The Second Chance Business Coalition, backed by JPMorgan Chase and Walmart among others, connects employers with formerly incarcerated workers, and post-release employment is the single strongest predictor of desistance from crime, which means hiring one returning citizen is not charity but rather the cheapest crime prevention investment available to any employer in any industry in any city in the country.

If you work in corrections: The Bureau of Justice Assistance runs a Pay for Success program providing funding and technical assistance for converting traditional contracts to outcomes-based structures, with pilot results from Massachusetts, New York, and the United Kingdom demonstrating that the model functions when implemented with adequate risk adjustment and transparent third-party evaluation.

If you invest: Social impact bonds have mobilized over $300 million globally across 89 projects in 19 countries since the Peterborough pilot launched in 2010, delivering returns of 3% annually in Peterborough's case while reducing reoffending by 9%, and the Rikers failure demonstrated that investor capital rather than taxpayer money absorbs the downside when a program underperforms, which makes the risk profile meaningfully different from a government grant and the social impact meaningfully larger than a Treasury bill.

The Bottom Line

The United States has built an $80-billion-a-year corrections system whose primary financial incentive is to keep its own customers cycling back through the door, where private prison companies earn more revenue when recidivism runs high, public prison budgets expand when inmate populations grow, corrections officer unions gain membership when cells stay occupied, and legislators win reelection by looking tough rather than by producing results that would make toughness unnecessary. Every stakeholder inside the system benefits materially from the system's failure to achieve its stated purpose, which is why the fix is not ideological or partisan but structural: pay for empty beds that stay empty because the people who left them are employed and housed and sober and not coming back, and stop paying $107.90 per day for the privilege of ensuring that they will.