Technical appendices to The Cost of Inaction: How Unmitigated Heat in Prisons Extends Incarceration and Costs Taxpayers in California, forthcoming in the Berkeley Public Policy Journal (2026).
Appendix A: Heat Risk Index Methodology
The Prison Heat Risk Index uses an additive risk framework where risk is the weighted sum of three normalized component scores for Hazard, Exposure, and Vulnerability (0.25H + 0.25E + 0.50V), following recommendations from Ovienmhada et al. (2024). Vulnerability is double-weighted to prioritize latent medical risk. The additive form prevents a facility with full mechanical AC from scoring zero risk; the index can then answer “where are people most at risk if cooling fails?”, taking into account that cooling in prisons may be withheld or neglected due to power imbalances between staff and incarcerated people (Brunn et al., 2025). The Hazard, Exposure, and Vulnerability components are an equal-weight composite of sub-indicators normalized 0–1 across 31 facilities; the final score is normalized 0–100 jointly across both time periods (current: 1991–2020; mid-century: 2041–2070, modeled with a SSP3-7.0 scenario for heat projection data).
Hazard combines annual days over 90°F, frequency of hot nights, and air quality. Exposure combines the ratio of indoor to outdoor 78°F days, urban heat island effects, and refrigerated air conditioning rates. Vulnerability combines CCHCS health risk tiers, share of the population 50+, the share enrolled in the Enhanced Outpatient Program for mental health (EOP), the share with a Disability Placement Program designation (DPP), and the share of people of color at each facility. Full documentation as well as all facility-level hazard, exposure, and vulnerability data compiled for the risk index has been published as a dataset on the author’s GitHub, intended as a resource for researchers, advocates, and policymakers working on carceral facility conditions and climate change.
Adaptive capacity is not included as a fourth component, following Ovienmhada et al. (2024)’s treatment of incarcerated populations as having effectively zero adaptive capacity. Incarcerated people cannot relocate, purchase cooling, choose their housing unit, or leave during a heat event; the structure of incarceration removes the individual and collective agency that capacity metrics are designed to measure.
Appendix B: Cost Model
Methods. To estimate heat-related illness and death impact, the marginal acute mortality (M_acute) is calculated by applying Skarha’s (2023) all-cause mortality slope to projected threshold exceedance person-days:
where P_exp is hazard (facility population × average annual days with tmax ≥ mean summer tmax + 10°F, 1991–2020 WMO baseline), T_excess accounts for the continuous dose-response (mean excess of 3.11°F above threshold → 1.31×), μ is the baseline annual all-cause mortality rate, and Δ is the 5.2% mortality increase per threshold exceedance day (Skarha, 2023).
To estimate the marginal increase in time served (Y_total) due to extreme heat exposure, time lost from non-violent behaviors (infractions) is summed with time lost from violent incidents (credit forfeiture and recommitment from victimization):
where λ_nv and λ_v are vulnerability to the rate of non-violent and violent infractions, Δ_v is the hazard exposure effect on violence, ρ_recidivism is the risk rate of recidivism from victimization, and t is time lost in years (for non-violent infractions, violent infractions, and new sentence length for recommitment).
Detailed model findings. Given an exposure rate of three months per year to 90°F days by mid-century (2040–2070), 75% of the incarcerated population is modeled as exposed based on facility location. When estimates from known findings about California’s incarcerated population are applied, 25% are not impacted by the temperatures, 24.6% are estimated to be of highest heat vulnerability to illness and death, and between 8.6–28.6% are estimated to be involved in heat-related violations.
The marginal increase to mortality was calculated using Skarha’s (2023) finding that each day exceeding a facility’s mean summer maximum temperature by 10°F is associated with a 5.2% increase in all-cause mortality. Applied to 840,457 projected Skarha-threshold person-days across 31 CDCR facilities (2016–2025 average), using a baseline mortality rate of 3.07/1,000/year (CCHCS 2016–2019) and a slope adjustment of 1.31×, this yields approximately 0.48 heat-attributable deaths per year currently — one heat-attributable death every ~2 years system-wide. By mid-century there will be approximately 1,256,069 person-days above Skarha’s threshold, increasing heat-attributable deaths ~50% to 0.72 per year, or one every 1.4 years.
The marginal increase to incarcerated person-years was summed from the effects of non-violent and violent violations. A 20% marginal increase applied to the baseline of 9,944 violent incidents per year among the exposed population (CDCR Performance Measures Incidents Reports, 2021–2024) yields 1,066 additional violent incidents annually (a 10.7% increase). Using a midpoint credit forfeiture of 210 days gives 613 total credit-years lost. To estimate additional recidivism from victimization, Listwan et al. (2013) report an odds ratio of 1.325 for recommitment among prison violence victims; applied to the system-wide three-year return-to-prison rate of 23.7% (CDCR Recidivism Reports, 2013–2018), this yields a marginal increase of ~5.5 percentage points among victimized individuals, or ~59 additional recidivists per year. Using the average sentence length for parole violators returning with a new term (4.4 years), steady-state incarceration attributable to recidivism increases by 260 person-years annually. Non-violent infraction rates are unknown, so they are provided in a range of 0–20% with a midpoint of 15 days credit lost, giving a range of 0–382 credit-years lost.
Appendix C: Engineering Alternatives
Engineering intervention mitigations. The provision of air conditioning (mechanical cooling) was shown to effectively eliminate heat-related mortality in Texas prisons, and much of CDCR’s proposed solutions have included mechanical cooling (Skarha et al., 2022; CDCR, 2025). 24% of CDCR housing units currently have mechanical cooling, 52% have evaporative cooling, and the remaining use air handlers or fans (CDCR, 2025; Raychaudhuri et al., 2025). However, as described by CDCR, the materials and age of the buildings often render cooling systems ineffective by urban heat island effects.
Several decades of civil and environmental engineering research has established interventions to effectively reduce the effects of urban heat islands. Cool roof materials and colors reduce the indoor temperature under a roof 50°F compared to traditional roofs (USDOE, 2024; Akbari, 2016). Replacing asphalt with cool pavement materials and colors can reduce the nearby surface temperature 30°F (LBL, 2025; Akbari, 2016). While shade structures can help reduce air temperatures, one of the most effective tools for reducing the physiologically equivalent temperature (e.g. the temperature your body experiences) in an urban heat island is tree shade; trees can reduce temperatures up to 35°F (Wu, 2025).
Recommendation: address heat island effects with lower-cost, short-term interventions. Waiting for complex HVAC overhauls is insufficient. CDCR should immediately incorporate lower-cost, research-backed interventions into existing maintenance cycles. This includes ensuring that all roof upgrades be built with “cool roof” materials as per California Energy Code, Title 24 (Akbari, 2016; CEC, 2022), replacing asphalt with permeable pavements, and increasing tree canopy wherever possible. These measures can reduce urban heat island effects and improve the efficiency of future mechanical cooling systems.
Appendix D: Legal Costs
Between 2022 and 2026, CDCR-related class action and settlement costs ran about $181.2 million. The bulk of that, $155 million, came from Coleman v. Newsom fines over inadequate mental health care (LAO, 2024). A shared $21.1 million attorney fee pool in FY 2022-23 covered Plata v. Newsom, Armstrong v. Newsom, Clark v. California, and Ashker v. Newsom (CDCR, 2024), and a $5.1 million pregnancy discrimination settlement in Carreon v. CDCR (Civil Rights Litigation Clearinghouse, 2024). Factor in $56.9 million in Ashker monitoring-phase fees through 2027 (CDCR, 2024) and the six-year total reaches roughly $238.1 million, averaging close to $39.7 million a year. In Stoetzl v. State, the 2024 Budget Act gave CDCR blanket augmentation authority to draw litigation defense costs from the General Fund with no ceiling (California State Assembly Committee on Budget, 2024); that case involves an ongoing wage/hour dispute covering around 40,000 employees. The 2026-27 budget also includes $23 million for ADA-compliant infrastructure tied to Armstrong v. Newsom (California DOF, 2025).