How St. Louis is using data to make its homeless services fairer

Kathy Connors is well acquainted with the personal loss associated with becoming unhoused. As executive director of Gateway180, the largest residential emergency shelter in St. Louis, she has met people who’ve been laid off, suffered debilitating injuries, fallen victim to natural disasters or watched loved ones die.

“By the time they come to us, people have experienced overwhelming devastation,” Connors says. “Homelessness is trauma in itself. But they’re also incredibly vulnerable — to illness, to violent crime.”

Connors’ job is to find beds for those who need them most. Much to her and her colleagues’ dismay, this goal is growing increasingly out of reach.

In 2023, 1,252 people in the city of St. Louis lacked access to stable, long-term housing; around 100 were sleeping on the streets, and some 200 more were couch-surfing or squatting in abandoned buildings. Many have chronic physical or mental health issues, and those who have received assistance are ending up on the streets again due to the rising costs of housing, childcare and transportation.

The St. Louis City Continuum of Care is a system of more than 100 public and private organizations such as Gateway180 that coordinates homeless services. Unfortunately, the people who need housing outnumber the CoC’s available beds. Around 10 households a month get referred to a housing program, while at any time another 1,000 families are on the waiting list. So the CoC uses information gathered from surveys to identify and prioritize those in the most danger.

“Data can help us do what’s best for the community,” says Kaitlyn Poepsel, manager of data analysis and reporting at the Institute for Community Alliances (ICA), a nonprofit that maintains the homeless management information system for the city of St. Louis.

But capturing the right data can be tricky. In the 2010s, for example, St. Louis, like hundreds of cities and counties around the U.S., adopted the Vulnerability Index–Service Prioritization Decision Assistance Tool, a questionnaire about the client’s housing history, health and safety, to help measure need and allocate housing.

By the end of the decade, the tool had come under criticism for being unreliable and racially biased. White people, and white women in particular, tended to score highest — even though Black people are more likely to experience homelessness. In 2022, ICA discovered that veterans in the city of St. Louis also scored lower on average than non-veterans, even though veterans nationwide are disproportionately affected by homelessness.

So the next year, the CoC stopped using the Vulnerability Index and focused instead on a homegrown survey called the Risk/Medical Frailty (R/MF) assessment to see if it sidestepped those biases — or introduced any of its own.

“Homelessness disproportionately affects persons of color,” says Isaac Fox-Poulsen, a data analyst at ICA. “Society marginalizes certain communities, and we needed to make sure that we are not exacerbating that problem.”

Harnessing data for a fairer process

Around this same time, Mastercard data scientists were already helping law enforcement and social services agencies in St. Louis test the effectiveness of the city’s new Crisis Response Unit, which dispatches mental health professionals on some 911 calls to deescalate situations, identify services and follow-up care, or provide a compassionate ear to troubled individuals.

When the company learned that the community wanted to capitalize on data to improve the city’s homeless services as well, it offered its expertise and resources pro bono as well. “By harnessing our expertise and technology, we’re helping the social sector realize the power of data insights to shed light on disparities and accelerate inclusion,” says Eric Schneider, an executive vice president in Mastercard Services who oversees these “data-for-good” initiatives. “Ultimately, this creates a stronger community and a more robust economy for everyone.”

ICA happily partnered with Mastercard. “Given the day-to-day job, we don’t always have time to do long-term studies,” Poepsel says. “For service providers, there’s always someone at their door who needs housing right now.”

What ICA and Mastercard discovered from analyzing the anonymized responses of 4,900 clients was illuminating: Though the new survey avoids past biases against men and veterans, Black people are still underrepresented among high scorers. While more than two-thirds of the clients in the study were Black, they made up just 56% of the highest-scoring group.

The disparity is driven in large part by questions about health conditions, the analysis found. Given the legacy of exploitative medical research and ongoing racial disparities in the quality of health care, only 59% of Black Americans trust doctors, as compared with 78% of white Americans. As a result, Black respondents generally score lower on sections of the R/MF that frame health problems around their interactions with the medical system, such as a series of questions about the applicant’s diagnoses and prescriptions.

The researchers also discovered that the survey favors elderly respondents at the expense of children. A question that awards a five-point bonus to respondents 65 and over, designed to identify clients with the highest COVID risk, seems to grant them a disproportionate advantage.

New information in hand, the St. Louis CoC has assembled a team to revise their assessment. They are discussing refocusing the health section on symptoms rather than medical services; they are also looking at how they assign points to older adults, while continuing to track changing demographics of homelessness — experts say that homelessness among older adults is on the rise.

“We serve so many individuals who have been slammed backwards by unjust systems,” Connors says. “We must make the CoC a system they can trust, and it all starts with the assessment. This work is critical for giving fair entry to the people who need it most.”