Evidence shows that early identification of risk factors for homelessness—followed by targeted, data-driven interventions—can help households maintain stable housing and significantly reduce both individual hardship and broader societal costs. As studies continue to validate the impact of these strategies, leaders across sectors are increasingly implementing evidence-based models that intervene before housing instability escalates into homelessness.
Janey Rountree brings a wealth of experience from the fields of education, law, and public policy, with a career shaped by work in public agencies across major cities like New York and Chicago. Throughout her roles, Rountree has focused on improving publicly funded programs by strengthening the relationship between research and effective service delivery. This commitment led to her co-founding of the California Policy Lab (CPL), an institute dedicated to evidence-based partnerships with public agencies to inform and transform programs that impact vulnerable populations.
The California Policy Lab’s mission—to translate rigorous research into meaningful policy change—has positioned it at the forefront of innovations in preventing homelessness. In recognition of CPL’s leadership and data-driven approach, the Hilton Foundation’s Homelessness Initiative partnered with CPL in 2019 and has since expanded its support, with a landmark $5.95 million grant approved in 2025.
To shed light on the evolving landscape of homelessness prevention, the Hilton Foundation communications team recently spoke with Janey Rountree about CPL’s pioneering initiatives and the growing momentum behind proactive, research-driven strategies to keep people stably housed.
Janey Rountree
Executive Director of the California Policy Lab at UCLA
Hilton Foundation: The lab recently put out a report about an innovative homelessness prevention program in LA County, which helps people before they lose stable housing. This pilot program uses a predictive model developed by CPL, which you described as instrumental in preventing individuals and families from experiencing homelessness. Can you explain how this works?
When we talk about homelessness prevention, one of the biggest challenges is knowing who should get assistance. How do we know who is going to become homeless if they don’t get help? And the reason why that’s challenging is that we have millions of people who rely on public assistance in our region, but only a very small percentage of them will actually lose their housing and end up on the street or in shelters.
What the Homelessness Prevention Unit does is test a strategy for figuring out how to target assistance to the people who need it the most. Many programs will screen people who come in based on risk factors, but this one is using a proactive predictive model – which we developed – that runs on county health system data and other data. It creates a list of individuals at high risk of homelessness and then connects them to a program that’s unique.
The county developed the program to be highly flexible and client-centered to meet that household or those clients where they are in that moment. That might be rental assistance or utility debt assistance; it could be transportation to go back to work. It could be workforce development support or a range of other types of services.
What are the benefits of using this model to predict who will fall into homelessness?
Predictive models are more accurate than human judgment at finding people who are very vulnerable. These models analyze large datasets to detect patterns and then make predictions about the future, such as identifying people who are at high risk of homelessness. The people enrolled in the Homelessness Prevention Unit, on average, are more vulnerable, and have significantly higher prior rates of incarceration, serious mental illness, substance use disorder, and other issues. The good news is we know how to prevent homelessness for people who have a variety of needs.
Another aspect of this I’ve already touched on is that it’s proactive. That’s really important because a lot of the individuals who the Homelessness Prevention Unit contacts and enrolls are disconnected from services. There’s almost no overlap between them and people enrolled in other prevention programs.
When HPU case managers connect with them, they find a lot of people have had bad experiences with public programs and may not trust them. So that proactive element, we think, is unique, in addition to the HPU working with people who are much more vulnerable and whose risk of homelessness is so much higher.
We’re running a randomized control trial and while we don’t have the final results yet, we recently released some preliminary findings from a pilot period that show a lot of promise. We looked at a group of about 1,600 people who the model identified as being at high risk of homelessness. HPU case managers attempted to contact all of these individuals, and about one in five people (21%) successfully enrolled.
When we look forward 18 months, people who enrolled in the HPU had a 71% lower rate of subsequently enrolling in interim housing or street outreach program, compared to people who were not enrolled. In a report we released last year, we shared that 92% of HPU participants reported that they had been able to stay housed. We would have predicted a much higher rate would be homeless, given the risk level, so these are promising results. We expect to share the results from the more formal evaluation in 2027.
As you’d expect, it’s challenging to enroll people in this program, both in terms of reaching people (phones may not be working or disconnected), and also convincing people it’s not a scam. We found that enrollment in the HPU increased by 67% after the program made improvements to its outreach and enrollment processes. With more people enrolling, that’s more people being served and hopefully avoiding homelessness altogether.
What are the strengths and weaknesses of the predictive model that you developed to identify people at highest risk of homelessness?
Anytime you’re talking about predictive modeling, equity and data privacy are huge concerns. We write a lot about those issues in our 2024 paper. We also test the model constantly for equity to make sure that if the county is offering this unique new resource, it’s being made available to people in an equitable fashion. We found it performs consistently across race, ethnicity, and gender, with a particular strength in identifying Black individuals at risk of homelessness. This is important because historic and systemic inequities mean that Black people are over-represented among people experiencing homelessness in Los Angeles, compared to their presence in our overall population.
A weakness of this approach is that you can only be predicted if you’re in the data. So, there are going to be people who are at risk of homelessness who are not connected to county services, so they’re not in the data model. However, this is not a weakness if you view the HPU as one part of a larger homelessness prevention strategy. A really comprehensive, healthy prevention system will have something proactive like the HPU in addition to a program that can serve people who raise their hand and say, ‘I’ve never had any contact with county services, but now I really need help.’ If you only have one or the other, you’re going to miss the other group.
What homelessness prevention programs are showing the most promise, according to the latest research?
I love this question. We recently had an event to highlight all the evidence-based models for prevention and to have a conversation not only about what their common elements are, but what we don’t know and what we need to study more.
These are programs in Chicago, Illinois, New York City, Santa Clara, the San Francisco Bay Area, and L.A. They tend to have some common elements. They start with some approach to targeting. Usually, eligibility criteria is very low income: people with 30 to 50% AMI [Area Median Income*] or lower and who have other risk factors. That’s designed to get at people who are not just housing unstable, but who are the ones at risk of becoming homeless.
The programs have other important common elements, such as flexibility around the type of cash assistance so that it could go to a landlord, but if the person’s not a leaseholder, it could instead go to a family member who’s a leaseholder. The programs pay for things that are really important to housing stability. For example, the Homelessness Prevention Unit [in L.A. County] will purchase an e-bike for someone to go back to work.
So, promising prevention programs tend to be targeted and have flexible cash assistance?
Yes, and case management. Another key element of these programs tends to be legal services. Eviction defense is critically important and should be supplemented by other types of legal services. Many people who are housing unstable need legal representation around domestic violence conflicts, employment disputes, and debt forgiveness.
It’s so important to invest in prevention. It’s more humane, frankly, and it’s a moral imperative, but it is also financially responsible [because] we do not have the resources to fund a permanent housing solution for everyone.
I want to emphasize that we know enough to do it now. We have lots of questions about how to do it well or how we can improve it, but it is something that we could be investing in now.
In January this year, Los Angeles suffered through historic wildfires, which delayed the annual Greater Los Angeles homeless count. Early on, many speculated that the numbers of people experiencing homelessness would be exacerbated by the fires. What was your prediction?
I think it’s important to take a step back and acknowledge the root causes of homelessness. It’s fundamentally about the cost of housing and rent burden. Any community where people pay on average more than 30 percent of their monthly income on housing, you will see an increase in people experiencing homelessness. It doesn’t matter what region or state you’re in.
Making a prediction in this case is difficult because there are so many complex factors. There are macroeconomic factors driven by policy and economic policies that are put in place, and of course there will be some effort by the system to meet demand.
What that speculation acknowledges is that we haven’t solved the root cause of homelessness, which is the cost of housing in our region.
* Area Median Income represents the midpoint income for families in a specific geographic area. Each year, the U.S. Department of Housing and Urban Development (HUD) calculates the AMI for each county or metro area. AMI is used to calculate income limits on program eligibility and, in some cases, affordable rents for a host of federal and local programs, including the largest affordable housing programs in the United States.