In this bi-weekly blog series, Abe Oudshoorn explores recent research on homelessness, and what it means for the provision of services to prevent or end homelessness. Follow the whole series!
I’ve had the privilege of following Dr. Alina Turner’s work for the past decade including many presentations at a variety of conferences on homelessness. I always appreciate her perspective as she is a perennial ‘systems thinker’, looking to the bigger picture with a drive for change. In a sector where we are often just swapping from one program for the next, it is encouraging that bright minds are looking to the system level.
A tweet from Alina the other week caused me to raise an eyebrow. It spoke to the promise of big data in decision making. However, I also wondered about small data, being participant perspectives… here is the exchange:
With my reply bordering on cynicism, I decided to do the responsible thing and research the program being referenced to see if, in this case, big data was or was not superior to small data.
While there is no published research that I could find on the work, two articles outline the project, one from Fast Company and one from Next City. In a nutshell, while 1/3 of families experiencing homelessness have been evicted, only 5% of evictions lead to homelessness. In the context of New York City, where 200,000 eviction notices are filed, it’s hard for the Department of Homeless Services to reach out to the right families with eviction prevention services. The usual process is to mail everyone who receives an eviction notice, but this interestingly leads to very few people reaching out. Multiple directed contacts have a lot higher responses from the potentially evicted families, but where do you send these?
Well, this is where the big data fits in. If you can use predictive software to identify which of the 200,000 families are most at risk of homelessness, then you can send them multiple personalized letters suggesting they reach out to the eviction prevention team. And this is what they are doing. The results so far? Those neighbourhoods using the predictive software are at least seeing a 50% higher rate of families connecting with the services. Not a clear relationship yet with eviction prevention, but certainly a good first step.
So, in a case where the system relies on families to reach out for help, and most families don’t, big data can be more effective than small data in engaging in targeted awareness raising. Does this ultimately lead to better eviction prevention? Time will tell. It is ultimately a downstream approach versus looking at systemic causes of housing loss, but seems like a promising practice.