As coordinated entry systems try to match growing numbers of unhoused people with limited amounts of housing, it’s more like The Hunger Games than Match.com.
This article was originally published in Shelterforce
Mary Kate Bacalao is the director of external affairs and policy at Compass Family Services and the co-chair of the Homeless Emergency Service Providers Association (HESPA) of San Francisco.
In hundreds of communities across the country, coordinated entry systems are attempting to match growing numbers of unhoused people with limited amounts of housing and services. As Virginia Eubanks notes in her book, Automating Inequality, proponents of coordinated entry like to call it “the Match.com of homeless services.” In theory, coordinated entry uses algorithms and other digital tools to streamline the local response to homelessness, putting unhoused people in a database and pairing them up with housing and services calibrated to their needs.
The U.S. Department of Housing and Urban Development (HUD) conceptualized coordinated entry in the early 2010s during a swell in homelessness after the foreclosure crisis and the last recession. With a typical carrot-and-stick approach to policymaking, HUD used a competitive funding program—the Continuum of Care program, which awards about $2.5 billion annually in highly regulated homeless assistance dollars—to push more than 400 communities (called “continuums of care”) to develop and operate their own coordinated entry systems.
HUD’s goal was a paradigm shift from a first-come, first-served model of homeless services—where the concern was that service providers distributed resources willy-nilly—to an efficiency approach, where data systems would distribute resources objectively, based on need. Proponents of coordinated entry used stereotypes to argue that the old model was inequitable: it privileged homeless people who “gamed the system” and service providers who “cherry-picked” the easy clients, over the supposed neutrality of algorithms.
This thinking makes it seem as if homeless response systems are simply disorganized, rather than deeply and dysfunctionally under-resourced. The logic goes: if we could simply line people up outside of a half-empty pantry according to whether they are starving or only very hungry, then we can better stretch the limits of the food we have. This logic may solve incidental problems, but it distracts us from grappling with the essential problem. As Gary Blasi, professor of law emeritus at the UCLA School of Law, points out, “Homelessness is not a systems engineering problem. It’s a carpentry problem.”
Joe Wilson, executive director of Hospitality House in San Francisco, puts it bluntly: “Coordinated entry is a classic case of shrinking the problem to fit the solution.” Coordinated entry systems deliberately work backward from an inadequate supply of housing—using eligibility criteria, assessment tools, and prioritization standards—to justify rationing it out to a small minority. It is a system built to rationalize an unconscionable mismatch between housing options and unhoused people. As Eubanks writes, “Coordinated entry is a machine for producing rationalization.”
Here’s how it works in San Francisco: Unhoused people presenting for services get entered into a centralized database, and trained staff apply several layers of assessments that weed them out of the running for housing. The first layer is an eligibility assessment—only people who meet the definition of homeless can be enrolled. The second layer is a service called “problem-solving”—an effort to divert people from the system they’ve just entered by solving some problem related to their homelessness (e.g., an unpaid utility bill). The third layer is a primary assessment—a standardized set of deeply personal questions (about medical and mental health problems, experiences of physical or sexual violence, and other sensitive topics) designed to probe how vulnerable each person is compared to the others.
The answers get fed into a ranking algorithm, which reduces each household’s vulnerabilities to a single numerical score. Each score gets assessed against a “threshold score”: at or above the threshold, and the household is deemed “housing-referral status,” meaning they scored high enough to get a housing referral. Below the threshold, and the household is deemed “problem-solving status,” meaning they scored too low to get housing. Instead, they get cycled back for another round of problem-solving services, which didn’t work the first time—mainly because people are homeless, and problem-solving is designed to solve problems other than homelessness.
It’s important to note that the threshold score is not a stable number: it goes up or down depending on how much housing is available at a given time. If there’s a lot of housing available, the threshold number goes down, and more people get housing referrals. If there’s not a lot of housing available, the threshold number goes up, and only the most vulnerable people get referrals. And they get referred to whatever is available, not necessarily something suited for their needs (for high-need families, this is almost always a time-limited rental subsidy that may return the family to homelessness when the subsidy ends).
This is a far cry from the efficiency approach touted by proponents of coordinated entry, and it creates an infuriating sense that homelessness is a relative concept: everyone enrolled in the system is homeless, but if they aren’t “homeless enough,” they cannot get meaningful help.
San Francisco’s coordinated entry system assessed 7,406 people in the 2020 fiscal year and weeded that down to 1,332 housing placements. In Los Angeles’s longer-running system, they have assessed 32,728 people (older adults) and narrowed that down to 7,568 permanent housing exits. It’s easy to see in both systems how the population shrinks from about five eligible people to one person ultimately placed in housing. This is the logic of lining up 10 hungry people outside an empty pantry and telling seven or eight of them that they’re not hungry enough to qualify for food.
This is how coordinated entry shrinks the problem—not in the sense of reducing it, but in the sense of putting tens of thousands of unhoused people through a digital process of elimination until the number of people prioritized for housing more or less matches the amount of housing that happens to be available. Ultimately, coordinated entry is not “the Match.com of homeless services.” It is more like the Hunger Games of housing access.
In any human services system, definitions and eligibility criteria play a role in shrinking the problem: they regulate who—and by extension, how many—can access the system’s limited resources. In coordinated entry systems, prioritization goes much further: it provides the rationale for using digital tools to shrink the pool of people who are eligible for housing down to the number of people actually prioritized for and placed in housing.
As Eubanks describes in Automating Inequality, prioritization evolved from research by Dennis Culhane at the University of Pennsylvania, which differentiates between “crisis” and “chronic” homelessness. The idea—based on principles of medical triage—is that the crisis homeless may need the service equivalent of a Band-aid to get back on their feet, whereas the chronic homeless may need the service equivalent of surgery. Under the old first-come, first-served model of homeless services, the crisis homeless were sometimes getting services that should have been prioritized for the chronically homeless.
Coordinated entry endeavored to fix that with a prioritization tool called the VI-SPDAT, or Vulnerability Index—Service Prioritization Decision Assistance Tool. Co-authored in 2013 by OrgCode and Community Solutions, the VI-SPDAT was designed as a pre-assessment triage tool, a precursor to a holistic assessment by a trained case manager. But with the sustained push from HUD and the widespread adoption of coordinated entry, many communities took up the VI-SPDAT as the assessment tool itself, with the result that people’s answers to deeply personal questions get reduced to a single numerical score that is often decisive about who will be prioritized for housing.
In a recent blog post, Iain De Jong, the head of OrgCode, clarified that the VI-SDPAT was not designed to make these decisions: “right in the name of the tool are the words ‘Decision Assistance Tool,’ not ‘Decision Making Tool.’” But in making the VI-SPDAT (or variants of it) the primary assessment tool, coordinated entry systems both automate and over-rely on prioritization to manage a zero-sum level of resources. And ultimately, prioritization only helps us reorganize an empty pantry. It does not push us to confront the fact that it’s empty, and it does not hold us accountable for the people who have not been prioritized.
Courtney Cronley, associate professor at the University of Tennessee, describes the VI-SPDAT as a “single, unvalidated measure of vulnerability” that is used broadly across the U.S. and Canada to determine whose needs are highest and who is most deserving. “The tool’s origins are murky,” she writes in a blog post: its co-authors developed it with demographic samples skewing older and male from a single geographic area. “Community-level studies,” she adds, “show consistent evidence of racial bias and unreliability in its use.” As De Jong readily concedes, “the tool was never designed using a racial or gender equity lens.”
Cronley’s research bears this out: She finds that women are twice as likely as men to report being homeless as a result of trauma, and that white women and Black women have similar odds of experiencing traumas that result in homelessness. But the white women she researched scored consistently higher than Black women on the VI-SPDAT—because the tool measures vulnerability based on behaviors more typical of white women, such as visiting emergency rooms and reporting activities like survival sex to their case managers.
C4 Innovations published a similar racial equity analysis of assessment data from four coordinated entry systems. They found that white people scored statistically significantly higher on the VI-SPDAT than Black and Indigenous people of color. They also found that white people were prioritized for supportive housing at higher rates than BIPOC individuals. (This finding did not apply to families, but many communities do not prioritize families for supportive housing.) Like Cronley, the C4 researchers found that the VI-SPDAT was more likely to identify vulnerabilities based on behaviors more typical of white people.
The result is that coordinated entry systems—by virtue of who they are not prioritizing—may be perpetuating structural racism in ways that communities have called out for years, but that researchers are only just discovering. This is particularly egregious in homeless response systems, given the role of racism in causing homelessness and the stark racial disparities in who experiences homelessness. To name just one example: 50 percent of homeless families in America are Black, yet racial (and other) biases may be intersecting every day to deprioritize women of color, many of them single moms, for housing.
This is a predictable, maddening result of the way coordinated entry was designed to streamline dysfunctionally under-resourced homeless response systems. And it deserves not just research but immediate attention from public officials, system designers, practitioners, and others. We have designed coordinated entry systems to be fundamentally inequitable: every day they’re slicing off shavings from a pie that is too small (resource scarcity) instead of assessing how the pie needs to grow to eliminate disparities—for people of color, for LGBTQ people—and meaningfully improve life and health outcomes for all unhoused people (resource equity).
Where do we go from here? We must get rid of coordinated entry—or redesign it. An equitable redesign would highlight problems and gaps rather than rationalize the mismatch between housing options and unhoused people. It would show the full picture of people and families needing support, rather than using artificial categories—like “problem-solving status” in San Francisco—to minimize the appearance of need and de-prioritize people who should be eligible for more. An equitable redesign would center racial and gender equity, and it would use digital tools transparently, to promote inclusive decision making and help us hold coordinated entry accountable to the goal of ending homelessness.
We must stop reorganizing the empty pantry and focus on putting more food in it. We must bring people in instead of weeding them out, with an emphasis on equity for people of color and LGBTQ people. We must insist on human decision making in the field of human services, and we must stop relying on digital tools to shrink our problems instead of solving them.