San Francisco Rations Housing by Scoring Homeless People’s Trauma. By Design, Most Fail to Qualify.

Tabitha Davis had just lost twins in childbirth and was facing homelessness. The 23-year-old had slept on friends’ floors for the first seven months of her pregnancy, before being accepted to a temporary housing program for pregnant women. But with the loss of the twins, the housing program she’d applied to live in after giving birth — intended for families — was no longer an option.

After several weeks in a hotel, which a prenatal program for homeless people had paid for while she recovered, Davis went to a brick building in San Francisco’s South of Market neighborhood to apply for a permanent, subsidized housing unit. There, a case worker she’d never met asked her more than a dozen questions to determine if she was eligible.

Some of the things he asked: Have you ever been sexually assaulted while experiencing homelessness? Have you ever had to use violence to keep yourself safe while experiencing homelessness? Have you ever exchanged sex for a place to stay? “Those are the questions that really bothered me,” she said. “Whatever my experience is of being sexually assaulted, or what I had to do in order to stay safe on the streets, shouldn’t pertain to whether or not I deserve housing.”

That day, Davis was informed that the score she’d been given based on her answers to the questionnaire wasn’t high enough to qualify for permanent supportive housing. It was a devastating blow after an already traumatizing few months. “I thought, ‘You put me on the streets right now, mentally, I will kill myself,’” she said.

What Davis encountered with those questions is called coordinated entry, a system designed to match people experiencing homelessness with housing. In San Francisco’s system, applicants are asked 16 core questions, and their answers are given a point value which is then tallied. The total number is intended to reflect applicants’ vulnerability; currently, a score of 118 points means they qualify for one of the city’s permanent supportive housing units, which is subsidized by the government and comes with wraparound supportive services. Applicants with lower scores may qualify for rent assistance or a bus ticket out of town, but if they want housing in San Francisco, they have to wait six months before taking the test again.

Though the city’s Department of Homelessness and Supportive Housing has an annual budget of $598 million and the majority of that is spent on housing, there simply aren’t enough permanent supportive housing units available to accommodate the thousands of homeless people in San Francisco. (A 2019 survey estimated the number of homeless people at more than 8,000.) The threshold for approval is directly tied to housing availability, and right now, roughly one-third of people who take the assessment score high enough to qualify.

“It’s really prioritizing scarce resources,” said Cynthia Nagendra, the department’s deputy director of planning and strategy. “There has to be some prioritization, unfortunately, until we have some housing resource for every single person.”

Coordinated entry was meant to be a more objective tool than the previous system, which offered resources on a first-come, first-served basis. In contrast, coordinated entry aims to determine who is most vulnerable and who should therefore get access to the limited supply of available housing.

Through records requests, the San Francisco Public Press and ProPublica obtained the questions and scoring algorithm used in San Francisco’s coordinated entry questionnaire, which has never before been made public. The news organizations solicited feedback on that tool from front-line workers, academics and people experiencing homelessness. Some raised objections to how the questions were phrased. Others pointed out inequities in the scoring. And many more criticized the way it was administered, suggesting that the process itself — in which applicants are asked very personal questions by a stranger — might make it unlikely that already-distressed people would answer accurately.

In our interviews, it became clear that the survey fails to identify many of the vulnerabilities it was intended to catch. And what was supposed to be an objective tool winds up, as a result of how it’s written and administered, making it harder for certain populations — immigrants, young people and transgender people, among others — to get indoors, experts and advocates told us.

For Davis, that meant some of the hardships she was experiencing were overlooked. For instance, there was no question in the survey that would give her points for the losses she had just suffered. Failing to qualify for housing resulted in weeks of stress and instability while she recovered from the trauma of losing her children. Eventually, with the assistance of case workers at several organizations, she found a place in a transitional housing program for youth. But being told, during the lowest moment of her life, that she did not qualify for permanent housing left its mark. “It made me feel invalid in my own experience,” she said.

In response to these critiques, homelessness department spokesperson Denny Machuca-Grebe said in an email, “I want to make it clear that anyone who comes to our department for help should NOT ‘be left out.’” For those deemed ineligible for housing, he said the city offers other services; these may include shelter placements, relocation help and rental assistance. In general, the department had not responded to requests for comments about individual cases in the past, and it didn’t comment on Davis’ experience.

Excluded Populations

Coordinated entry was first implemented in 2018, after the Department of Housing and Urban Development began requiring regions that apply for federal homelessness funds to create a tool “to ensure that people who need assistance the most can receive it in a timely manner.” Much of the rest of the country adopted a tool called the Vulnerability Index, Service Prioritization Decision Assistance Tool. San Francisco developed its own set of questions, intended to determine which unhoused people are in greatest need of a home.

In the four years since the requirement was implemented, some cities and counties have reviewed their coordinated entry systems and uncovered trends such as significant racial or gender biases. A 2019 analysis of data from Oregon, Virginia, and Washington found that even though people of color were overrepresented in the homeless population, they tended to score significantly lower than their white counterparts, making it harder for them to access permanent supportive housing. The study recommended that HUD consider revising its coordinated entry guidelines to ensure that communities “equitably allocate resources and services.” This year, San Francisco started its own analysis of its coordinated entry process, and it expects to present the findings before the end of the year.

Nearly every expert we interviewed suggested that the experiences of people of color may not be fully reflected in their answers to the coordinated entry questions. San Francisco’s own data shows Black, white, Asian and Indigenous people being approved for housing at roughly equal rates. But Nagendra, from the Department of Homelessness and Supportive Housing, is looking into concerns that conditions that often make people of color more vulnerable are not being fully captured and that the numbers may not tell the whole story. “When you look at quantitative data, ours will show we are actually prioritizing people who are Black at an equitable rate. But when we talk to people, they might tell a different story,” she said.

Courtney Cronley, an associate professor of social work at the University of Tennessee who has written about racial bias in coordinated entry systems, pointed to one of San Francisco’s questions as an example of possible bias in action: “How many times have you used crisis services in the past year (for example, mental health crisis services, hospital, detox, suicide prevention hotline)?”

“Black people are less likely to use formal health care systems,” Cronley said. “They’ll reach out to family and friends and social support systems rather than going to the doctor. The doctor is not someone that they necessarily trust. These questions are biased towards persons who are white in our communities and biased against African Americans.”

The Department of Homeslessness and Supportive Housing has also said that very few transgender and gender-nonconforming people have been taking coordinated entry assessments. In a December 2021 meeting, Megan Owens, the department’s coordinated entry manager, presented demographic data on who was being assessed. She said that the number of people reporting those gender identities during assessments is “lower than in the best estimates of the homeless population.” In March, city data showed that transgender and gender-nonconforming people constituted only 2% of those taking assessments to try to get housing.

Critics of San Francisco’s coordinated entry system also say that one of the most basic questions, “How long have you been homeless this time?” leads to the exclusion of immigrants and younger people.

That question might sound simple, but it’s difficult for many people to say how long they’ve been homeless — and answering accurately can be critical to getting housing. That’s because San Francisco’s algorithm grants people more points the longer they have been unhoused: A person who has been homeless for more than 15 years receives 12 more points than someone who’s been homeless for one to two years. Anyone who says they’ve been homeless for less than a year gets zero points on this question. (On average, adults who qualify for housing in San Francisco report being homeless for six years.)

Gayle Roberts, the chief development officer at Larkin Street Youth Services, a nonprofit serving young homeless people in San Francisco, said it is “common knowledge among social service providers that it [the coordinated entry system] is weighted heavily toward serving the needs of those who have experienced homelessness the longest.”

Laura Valdéz, executive director of Dolores Street Community Services, is one of several nonprofit leaders who questioned the efficacy of the system. “For many newly arrived immigrants, the way they literally interpret that question is since they’ve been here in San Francisco,” she explained. “So their scores are really low in comparison to other folks. But a large percentage of our immigrant community were unhoused in their home country.”

Valdéz also said the coordinated entry system can lead people living outdoors to accrue significant trauma before they qualify for permanent supportive housing. The program, she said, “requires people to stay in that system that is creating greater and greater harm to them for them to be able to score higher.”

The duration-of-homelessness question can also be tricky for homeless youth, defined as those between 18 and 24. In a 2019 count, they accounted for 14% of the city’s homeless population. Many young people are intermittently homeless, making it difficult to calculate the full length of that experience, said Dr. Colette Auerswald, a professor of community health sciences at the University of California, Berkeley.

“Maybe they stayed on their friend’s couch for five days and they were on a bus last night,” she said. “So they may be like, ‘Well, one day,’ but actually they’ve been in an unstable situation for a really long time.”

San Francisco’s homelessness department acknowledges this bias against young people seeking housing. In an attempt to address the age gap, the department included two questions that are only scored for people ages 18 to 24: “In the place you are staying, are you experiencing physical or sexual violence?” and, “In the last 12 months have you traded sex for a place to stay?” If they answer yes to either one, it provides a significant bump in their overall score: 12 points for each question. But if anyone older than 24 who has been sexually assaulted or has traded sex for a place to stay gets no points at all. (While the answers to these questions are only scored for 18-to-24-year-olds, they are asked of every person who takes the assessment. When asked why these questions were asked of people who could not receive points for answering, the department said it was for “data gathering.”)

Machuca-Grebe, the department spokesperson, explained that the question was added because “we have found that without the score placed on the questions for youth, they would be seriously under prioritized — leading to a disproportionate exclusion of youth.”

Davis was in the 18-to-24 age range when she first took her coordinated entry assessment, so those questions were scored. But she does not believe they should be asked at all.

“There’s not a single person that I can think of that is female-presenting that hasn’t been sexually assaulted while experiencing any part of their life, not just homelessness,” she said. “So you’re telling me that because someone hasn’t been raped, that she doesn’t get housing, and then she stays on the streets and then does get raped? And now she can? No, that doesn’t make sense.”

Questions From a Stranger

It is not just the wording and scoring of the questions that give experts pause. They also said that the way the assessment is given can fail to accurately assess a person’s vulnerability.

In San Francisco, all questions must be read by a trained staff member from one of the nonprofits that contract with the city to conduct the assessment. The questions are pulled up on an iPad or a computer. A drop-down menu offers a prewritten set of answers to select from, and the score is automatically added up by the software.

Coordinated entry assessments are frequently conducted in semi-public places, like a bustling office or a street corner under a highway. Applicants rarely have a preexisting relationship with the person asking the questions, and, due to understaffing at many nonprofits conducting assessments and the high number of people in need, there may not be time to build one.

“You really need to have interviewers establish rapport and relationship with the client prior to conducting or doing any assessment, because if they don’t trust interviewers, they’re just not going to talk to them,” said Cronley, the University of Tennessee professor.

The stakes are high: When an interviewer chooses the “Client refused” option from the pull-down menu of potential answers, the applicant receives zero points for that question.

Valdéz also sees lack of trust as a problem in the communities she serves. “Many of us would not feel comfortable speaking about our personal traumas, in 45 minutes, to a complete stranger,” she said. “My family experienced homelessness, and I can tell you right now, if I’m sitting in front of someone that I’ve just met, it is very unlikely that I would share that in an assessment.”

This was a concern voiced by Auerswald, the Berkeley professor, about the youth questions on violence and trading sex for a place to stay. She said the phrasing would not secure accurate results.

“My worries here is that a lot of young people are gonna say no,” she said. “And obviously, here, they really need to say yes. It’s one of their only hopes at prioritizing for housing, even though it’s a super traumatizing question.”

People’s personal interpretation of each question can affect their answers, Auerswald said. “A lot of young people who are trafficked would say no to this question,” she said. “They’d say, ‘Well I wasn’t raped, it wasn’t violent. I have someone taking care of me and I am paid or given something in exchange.’ Definitions of violence are different now. Violence is a lot of things. You can have sex under threat of violence, even if you don’t have a mark on you.”

Cronley said racial bias in child welfare and policing plays a similar role in determining how forthcoming people are willing to be when answering these questions.

“Black women are going to be more likely to fear that their children will be taken away from them if they report illicit behaviors, or if they report any sort of mental health challenges,” she said. “If you’ve got kids and you’re homeless and you’ve traded sex for money, you’re not going to tell them that you did that. No way.”

Davis had enough experience with systems for homeless people that she knew not answering the questions was not an option. “I had no choice but to answer them or I couldn’t get into housing,” she said.

For some, though, the experience is so uncomfortable that they drop out of the process entirely. A native of El Salvador, Luis Reyes has lived in San Francisco for 30 years and been homeless for 10 of those. Reyes said he has taken the coordinated entry questionnaire twice — once in 2019 and again in 2020, right before the pandemic hit. Like Davis, he went to the brick building at 123 10th St., the city’s largest drop-in center for these assessments.

“There was a guy who did the assessment in Spanish,” Reyes said, through an interpreter, of his 2020 interview. “‘Are you incapacitated? Are you a senior citizen? Do you have AIDS?’” Reyes remembers him asking. “He even asked me if I was gay,” he recalls — a question that is not included in the coordinated entry assessment. Reyes answered no to all of the above and says he was then told he didn’t qualify for housing.

The experience discouraged Reyes, who was living in a shelter at the time of his second assessment. He decided not to take the questionnaire again. He has spent some months sleeping in his car, and more recently he stayed with his girlfriend at a senior living facility. But she’s not allowed to have guests, and soon he will have to return to the streets.

System Under Review

Across the country, cities and counties are starting to critically examine their coordinated entry systems. Last year, eight communities, including Chicago and Austin, Texas, studied the data on their coordinated entry results and discovered significant racial disparities. Both cities revised their systems using community feedback, redesigned their processes and wound up approving more people of color for services.

In San Francisco, 17,000 coordinated entry assessments were conducted between the launch of the system in 2018 and the middle of 2021. This year, the city announced it would be undertaking its own review to determine if the government is serving people equitably and if the housing options offered are a good fit for those in need. Nagendra, at the Department of Homelessness and Supportive Housing, is overseeing the city’s review.

“If things have gotten away from our overall intention and design, we can look at those things and figure out where we need to redesign, refresh, whatever it might be,” she said in an interview.

The city’s approach to its review is driven by data and leans heavily on interviews, which are being conducted in focus groups and through outreach at encampments. The agency plans to make the research findings public in late May.

Critics would like to see a more radical overhaul of the coordinated entry system and the way it is pegged only to the supply of housing.

Joe Wilson, executive director of Hospitality House, a community center for homeless people in the Tenderloin neighborhood, where the majority of the city’s unhoused population resides, explains the problem with that approach.

“This algorithmic-based decision-making process is designed to keep the problem small enough so we don’t have to truly address it,” he said in an interview. “They’re not filling housing based on need, they’re assigning it based on capacity. It is not logical, it’s not consistent, and it’s not effective.”

For example, families used to be required to hit 40 points to qualify for housing. In February, the Department of Homelessness and Supportive Housing doubled that number to 80 points due to a shortage of family-specific housing. Owens, the coordinated entry manager at the department, estimated that the change would reduce the number of families who qualified for housing to between 50% and 60% of those taking the assessment, down from 75%.

Critics of the coordinated entry program have been proposing solutions as the city begins its review. In a February report, the Coalition on Homelessness, San Francisco’s largest nonprofit advocating for homeless people, recommended that the city “develop an assessment tool that categorizes people according to what type of housing would be the most suitable for their situation, instead of assigning them an eligibility score. This will tell us what type of housing and assistance is needed, versus how much housing we have.”

The organization also proposes letting case workers and housing providers work together to identify the best place to house an applicant. This approach, the Coalition argues, would create “a real-time housing placement system” that would more quickly bring vulnerable people indoors. This could help address the city’s chronic difficulty in filling the vacant units it has available: As the San Francisco Public Press and ProPublica reported in February, 1,633 people who had been approved for housing were still waiting to move in — some for months — even as more than 800 apartments sat vacant. At least 400 people had been on the waitlist for more than a year.

For those working on the front lines of the homelessness crisis, change to the coordinated entry system can’t come fast enough. Last July, in a meeting with the Department of Homelessness and Supportive Housing, Wilson told a story about a client his organization had helped.

“We have an 86-year-old woman who has been homeless for 14 years who has not been prioritized for housing,” he said, noting that she took a coordinated entry assessment but did not hit the 118-point threshold for housing.

A key insight from that experience, he said: Algorithmic decision-making “moves us away from the absolute necessity of human judgment and human interaction in human services.”