Program Profile: Denver (Colorado) Supportive Housing Social Impact Bond Initiative (Denver SIB)

This is a program for homeless individuals who are often in jail, to increase their housing stability and reduce their criminal justice involvement. The program is rated Effective. Individuals in the treatment group had statistically significant reductions in arrest, jail stays, days in jail, and shelter stays and visits, and statistically significant increases in mental health services, compared with the control group, but there was no statistically significant effect on substance use services.

An Effective rating implies that implementing the program is likely to result in the intended outcome(s).

This program's rating is based on evidence that includes at least one high-quality randomized controlled trial.

Program Goals/Target Population

In recent years, the number of people experiencing homelessness, particularly chronic homelessness, in Denver County, Colorado, has increased. Chronic homelessness is when people have experienced homelessness for at least 1 year and have a disability, either physical or mental, or a substance use disorder, or some combination of these (Cunningham et al. 2021). If these individuals’ needs are not met, it can result in frequent interactions with police or visits to emergency rooms, jails, and other facilities that offer crisis and institutional care.

To address this need, in 2016 the city and county of Denver launched the Denver Supportive Housing Social Impact Bond Initiative (“Denver SIB”), a supportive housing program to serve the chronically homeless population who frequently cycle in and out of jail. The goal was to increase housing stability and decrease jail stays for individuals who experienced long-term homelessness and had frequent interactions with the criminal justice and emergency health systems.

Program Components/Key Personnel

Denver’s Supportive Housing Social Impact Bond Initiative used an innovative mechanism called a social impact bond, or pay-for-success financing, to pay for part of the program. The city signed a performance-based contract to leverage $8.6 million in up-front capital from eight lenders to fund supportive services and a limited amount of flexible rental assistance. In addition to that funding, the initiative leveraged state and federal housing resources and reimbursements for a share of Medicaid-covered supportive services. The city agreed that if the program worked it would repay the private lenders, potentially with a positive return. If the program did not achieve its performance measures, the city would not repay the investors.

From its launch in 2016 to 2020, Denver’s Supportive Housing Social Impact Bond Initiative combined a permanent housing subsidy with intensive wraparound services, including subsidized housing; a modified assertive community treatment team; behavioral health services, including psychiatric services, individual and group therapy, and substance use treatment; links to community resources (for example, food resources, legal referrals, and advocacy) and to integrated health services (e.g., medical, dental, vision, and pharmacy services); and transportation assistance and referrals (Cunningham et al. 2018). The initiative used a Housing First approach, which placed no requirements or conditions on housing assistance, meaning that it did not require that participants meet preconditions to entry, such as entering treatment, achieving sobriety, or committing to ongoing service participation requirements (Cunningham et al. 2019).

The city developed an agreement with Denver PFS LLC, an entity established by the Corporation for Supportive Housing and Enterprise Community Partners, to execute the Denver Supportive Housing Social Impact Bond Initiative. This entity managed funding to the program’s service providers: the Colorado Coalition for the Homeless, which began providing supportive housing services in the program’s first year, and the Mental Health Center of Denver, which started doing the same in the second year. Denver’s Office of Behavioral Health Strategies provided staff for the program referral process.

After receiving a referral with data available from an individual’s police report (name, date of birth, and location of police contact or arrest), the two supportive housing providers were responsible for finding the individuals in the community and engaging them in the program. Immediately after engaging with service providers, initiative participants often were offered bridge housing while they worked on the documentation necessary for a permanent voucher and lease agreement. Bridge housing was provided in motels and in congregate or single-room occupancy housing in buildings owned and operated by a Denver Supportive Housing Social Impact Bond Initiative service provider. The initiative used a combination of permanent housing types, including scattered-site units rented with a housing subsidy in the private market and single-site buildings with designated supportive housing units. The single-site buildings were designed with trauma-informed spaces, and apartments were furnished according to residents’ needs. Case managers, counselors, peer specialists, psychiatrists, and nurses often were on site, and some type of support staff was available 24 hours a day. The buildings also served as a hub for community gatherings, events, groups, classes, and other supportive services. Some buildings engaged residents to develop community norms and held tenant conferences to address behaviors that went against those norms, which aligns with a Housing First perspective.

Program Theory

Research has shown a link between the criminal justice system and homelessness. People currently or previously involved in the justice system often are not connected to supports and face housing and job discrimination, meaning they are more likely to experience homelessness. Further, by virtue of being forced to live outside, homeless individuals are more likely to interact with the justice system through citations or arrests for low-level offenses such as loitering or sleeping in parks (Cunningham et al. 2021).

The Denver Supportive Housing Social Impact Bond Initiative theory of change identified a lack of stable housing among the target population as the root cause of their frequent arrests, jails stays, emergency room visits, and other negative outcomes that affect their well-being (Cunningham et al. 2021) and that, once these individuals were housed, they would not be drinking openly in public spaces, panhandling, trespassing, or engaging in similar sorts of nuisances or crimes (Cunningham et al. 2016). Thus, the Denver Supportive Housing Social Impact Bond Initiative focused on housing and supportive case management services to reduce criminal justice involvement and improve the lives of the city’s chronically homeless population.

Study

Hanson and Gillespie (2021) used a subset of the full sample from Study 1 (Cunningham et al. 2021) to assess the effects of Denver’s Supportive Housing Social Impact Bond Initiative on healthcare outcomes after 2 years. As described in Study 1 above, the target population included individuals with eight or more arrests, including three or more transient arrests over 3 years. An eligibility list was created in 2016 based on arrests from 2013 to 2015 and then updated every year to cover the 3 previous years. The process of referring people from the eligibility list to the supportive housing program was also the same as described in Study 1. When program slots were available, the list of these individuals was used to conduct a lottery to randomly assign people to the treatment group or to the control group. The control group received services as usual in the community, which primarily included access to emergency shelter and some short-term housing assistance. The individuals assigned to the treatment group were referred to one of two supportive housing providers, the Colorado Coalition for the Homeless or the Mental Health Center of Denver.

Of the initial 361 individuals allocated to the treatment group, 275 were Colorado Access or Denver Health and Hospital Authority members at some point in the year before randomization. Similarly, 274 people from the control group were Colorado Access or Denver Health and Hospital Authority members in the year before randomization. The analysis was conducted with these individuals who had any Medicaid enrollment in the year before randomization.

The treatment group for this study was an average of 43.6 years old, 88 percent men, 44 percent white, 35 percent Black, 13 percent Latino, and 7 percent Native American. Thirty-six percent of the treatment group had a mental health diagnosis at baseline, and 68 percent had a substance use diagnosis. The control group was an average of 44.6 years old, 84 percent men, 51 percent white, 34 percent Black, 10 percent Latino, and 5 percent Native American. Thirty-eight percent of the control group had any mental health diagnosis at baseline, and 66 percent had any substance use diagnosis. There were no statistically significant differences between groups on demographic characteristics or outcomes of interest at baseline.

The primary outcome of interest was use of office-based health services, defined as Medicaid use in the 2 years after randomization. Medicaid claims and encounter data were used to measure health care received in the community. Encounter data came from Colorado Access, a managed-care organization for both behavioral and medical care, and Denver Health and Hospital Authority, another comprehensive medical managed-care program. Demographic and arrest data came from the Denver Police Department, and jail stay data came from the Denver Sheriff Department; both were used as controls in the regression.

To assess the effect of the program on Medicaid services by mental health and substance use diagnosis, a treatment-on-the-treated effect was estimated using an instrumental variables approach for which assignment to the treatment group was an instrument for being housed in the supportive housing program. Results were estimated using ordinary least squares. Regressions controlled for age, race/ethnicity, gender, the number of days in jail in the 3 years before randomization, the number of arrests in the 3 years before randomization, and the value of the outcome in the year before randomization. No subgroup analysis was conducted.

Study

Cunningham and colleagues (2021) conducted a randomized controlled trial to assess the effect of the Denver Supportive Housing Social Impact Bond Initiative on criminal justice system involvement and housing stability after 3 years.

To create a list of eligible individuals, the target population was defined as all persons who had eight or more arrests with the Denver Police Department over 3 consecutive years. Three of the arrests had to be marked as transient, meaning that the person had no address or gave a shelter address. The Denver Police Department identified eligible persons through a data pull and created a master eligibility list that was updated once or twice a year to include newly eligible people. To refer people from the eligibility list to the supportive housing program, the police department established an automated report that matched daily police data with the eligibility list to identify persons from the list who had a police contact or arrest in the previous 24 hours. This process ensured that those who were referred were still in the community and interacting with police. The city also helped screen out people with open felonies from the previous 2 years because they were awaiting sentencing and would likely receive lengthy prison terms, precluding them from accepting the offer of supportive housing. Referrals to the supportive housing program were made on a rolling basis starting in January 2016, and participants were tracked through December 31, 2020. A lottery system was used to randomly assign the 724 eligible individuals to the treatment group (to participate in the supportive housing program, n = 363) or to the control group (whose members did not participate in the program and received usual care services in the community, n = 361). The individuals assigned to the supportive housing program were referred to one of the two service providers, the Colorado Coalition for the Homeless or the Mental Health Center of Denver, which were responsible for finding them in the community and engaging them in the program. Of those in the treatment group, 79 percent (285 individuals) were located, engaged, and housed.

As of 1 year before randomization, the treatment group was on average 44.1 years old, 87 percent men, 45 percent white, 32 percent Black, 15 percent Latino, and 8 percent Native American. The control group was on average 44.3 years old, 83 percent men, 49 percent white, 35 percent Black, 11 percent Latino, and 4 percent Native American. At baseline, the treatment group had an average of 4.2 arrests, 2.5 jail stays, and 67 days in jail, while the control group had an average of 4.7 arrests, 2.7 jail stays, and 69.5 days in jail. At baseline, 67 percent of the treatment group had any shelter stay, and 70 percent of the control group had any shelter stay. There were statistically significant differences between groups on demographic characteristics (percent Native American) and outcomes of interest (average number of arrests), and these were controlled for in the analysis.

Housing stability was measured by participants’ use of shelters (which included emergency shelter overnight and use of day shelter for food, shower, etc.) and access to permanent housing. Data were provided by the Denver Police Department, the Denver Sheriff Department, the Metro Denver Homeless Initiative, the Denver Housing Authority, the Colorado Division of Housing, the Colorado Coalition for the Homeless, and the Mental Health Center of Denver. Criminal justice involvement was measured by participants’ interactions with the system. Arrest data were provided by the Denver Police Department, and jail data were provided by the Denver Sheriff Department. De-identified individual-level data were linked by a unique research ID to facilitate analysis while maintaining confidentiality. A treatment-on-the-treated estimate was used to assess the impact of the Denver Supportive Housing Social Impact Bond Initiative on individuals who became housed, compared with individuals who did not become housed, regardless of to which group they were assigned. This estimate was calculated using an instrumental variables estimate per person served, among those who complied with their referral assignment (to enroll in the housing program or to remain in the control group). Results were estimated using ordinary least squares. The regression-adjusted models included the following control measures: age, gender, and race/ethnicity. In addition, the regressions control for days in jail, number of jail stays, number of arrests, and number of custodial arrests, all measured in the 3 years before randomization. No subgroup analysis was conducted.

The annual per unit cost (spread across multiple levels of government) of the Denver Supportive Housing Social Impact Bond Initiative was $22,265 for Colorado Coalition for the Homeless and $35,770 for the Mental Health Center of Denver. To compare the costs of the program with the costs avoided in other systems because of supportive housing services, Gillespie and colleagues (2021) also calculated per-person costs. These were estimates of the actual costs for the average initiative participant based on the average number of days spent in initiative housing per year, as opposed to the unit cost of a full year of supportive housing.<br><br> On average, participants referred to Colorado Coalition for the Homeless spent 198 days in housing per year, and participants referred to the Mental Health Center of Denver spent 158 days in housing per year. As a result, the average annual per-person cost of the initiative’s supportive housing was $12,078 for Colorado Coalition for the Homeless and $15,484 for the Mental Health Center of Denver. Further, the annual per-person costs of the initiative were compared with the annual per-person cost reductions from avoided outcomes in other services. Among the government funding sources for the various initiative services, the city and county of Denver had the greatest amount of avoided costs ($3,733 per person annually) because they pay for shelter, jail, courts, police, and public safety (Gillespie et al. 2021).

These sources were used in the development of the program profile:

These sources were used in the development of the program profile:

Cunningham, Mary, Sarah Gillespie, Devlin Hanson, Michael Pergamit, Alyse D. Oneto, Prasanna Rajasekaran, Tracey O’Brien, Liz Sweitzer, and Christine Velez. 2019. Maintaining Housing Stability Interim Lessons From the Denver Supportive Housing Social Impact Bond Initiative. Washington, D.C.: Urban Institute.

Cunningham, Mary, Ruth Gourevitch, Michael Pergamit, Sarah Gillespie, Devlin Hanson, Tracey O’Brien, Christine Velez, Daniel Brisson, Gary Sanford, and Abby Magnus. 2018. From Homeless to Housed: Interim Lessons From the Denver Supportive Housing Social Impact Bond Initiative. Washington, D.C.: Urban Institute.

Cunningham, Mary, Mike Pergamit, Sarah Gillespie, Devlin Hanson, and Shiva Kooragayala. 2016. Denver Supportive Housing Social Impact Bond Initiative: Evaluation and Research Design. Washington, D.C.: Urban Institute.

Gillespie, Sarah, Devlin Hanson, Josh Leopold, and Alyse D. Oneto. 2021. Costs and Offsets of Providing Supportive Housing to Break the Homelessness–Jail Cycle: Findings From the Denver Supportive Housing Social Impact Bond Initiative. Washington, D.C.: Urban Institute.