Every renter’s bad dream is to get evicted, but that happens for 200,000 households a year in New York City. Not all evictions, however, are equal in terms of the devastation they cause. A college student who has blown his rent money on beer, for example, is much less likely to end up in a homeless shelter than a single mother struggling to get by.
The problem for social service agencies is figuring out who’s at risk of imminently becoming homeless amid thousands of eviction notices, and reaching those who need help. The nonprofit data analytics firm SumAll is trying to help with that challenge.
Taking a page from the playbook of marketing firms who use “big data” to target potential customers, SumAll has developed a tool that uses data — court records, shelter history, demographic information — to identify people at risk of becoming homeless. SumAll’s algorithm helps social workers decide where to focus their efforts — down to the individual.
“Think about a highly targeted marketing campaign trying to sell something,” explains Stefan Heeke, SumAll’s CEO. “This is the same thing.”
Last summer, SumAll conducted its first pilot in collaboration with CAMBA, a Brooklyn-based nonprofit social service organization. They tracked four districts and targeted outreach in one, the Bedford-Stuyvesant neighborhood.
Pre-SumAll, CAMBA’s efforts to reach out to families going through the protracted eviction process was arduous. Social workers would look through the entire list of new eviction cases at Brooklyn’s Kings County Housing Court — roughly 5,000 per month — and then manually figure out which addresses fell in the areas they serve, says Melissa Mowery, vice president of CAMBA’s eviction counseling project, HomeBase. They then would mail letters to those in their zones, about 400 a month telling people about their eviction prevention counseling services.
With the help of SumAll, they were able to first geo-code the list and figure out which addresses were in the right neighborhoods — a process that took hours rather than days. “Hello, that saved a lot of manpower,” Mowery says. Then they used SumAll’s tool to figure out the 30 to 50 most at-risk cases. “At-risk” indicators included previous experience with the shelter system (using data from the city’s Department of Homeless Services), education level, employment status, age and even factors going back to childhood such as interaction with the foster care system. (All of this is public information.)
During the pilot, at-risk families received several letters in the mail — more personalized than previous outreach letters — and they were given a hotline number to call or text to set up an appointment. Only targeted families were given the hotline number, so their calls went to the top of the priority list, making the process of connecting them with services as efficient as possible.
The pilot results were promising. With SumAll’s data analysis, CAMBA was able to connect 50 percent more families in the test neighborhood with eviction prevention services compared to demographically similar neighborhoods nearby over the same period of time. That’s about 65 families that avoided ending up in a shelter. “That was huge, really huge,” says Mowery.
The use of big data to target individuals has a shaky reputation. “You don’t want big data to be used in ways that would invade privacy or would be used in ways that could have negative repercussions in other contexts,” says Jay Stanley, senior policy analyst with the ACLU’s Speech, Privacy and Technology Project. But, he says, “Incentives matter. The incentives here are for the social service agency to provide services to those in need the most, which is different from other contexts where the interests of the entity doing the data mining may run counter to the subjects of the data mining.” He warns that the analysis must be sound, or it could potentially exclude some in need of services.
“It’s kind of controversial, targeting in the social service world,” Mowery admits.
She says part of the controversy comes from the idea that social service agencies should cast a wide net, to everyone who might be in need. But this approach of trawling for any and all possible clients, she says, wastes time and resources — both of which are thin at CAMBA, as with most nonprofits.
With SumAll’s tool, her team got the most at-risk families through the door. “Bam, these are the 50 people you need to be thinking about,” she says. “Don’t worry about these other 4,550.”
Several nonprofits in New York are interested in using the tool. The system requires access to a robust database from government and the courts, Heeke says. One challenge is in accessing eviction filings, which lie in the hands of the city and housing courts. Currently SumAll is re-negotiating relationships with officials in the new administration to streamline access to data. But with the right cooperation, Heeke says, “targeting critical services to most vulnerable populations could be done in any neighborhood, in any city.”
The Equity Factor is made possible with the support of the Surdna Foundation.