Preventing homelessness: the tough job of predicting who is at real risk

When the unknown number popped up on her phone, Jocelyn Escanuela was in the checkout line at Walmart.She still can’t explain why she picked up and then listened to a cold-caller’s pitch that sounded a lot like a scam.She had been selected to receive a grant of $6,000, the caller told her.

And she would have a personal assistant to help her get her through her “crisis.”How did they even know she was in a crisis?It turned out the caller was legitimate.She was from the Homelessness Prevention Unit, an experimental Los Angeles County program that is testing whether it is feasible to stop homelessness before it starts — one person at a time — by picking them out of mountains of data.

Escanuela’s crisis was detected not by a person but a predictive statistical model that was developed to solve a conundrum that has made homelessness prevention a tantalizing but underused strategy.Despite sound evidence that services such as eviction defense and financial assistance can prevent people from becoming homeless, it’s impossible to know after the fact whether any given person would have become homeless without the help.Research has shown that only a small percentage would.

The elusive goal of prevention is to identify that small percentage.“With limited prevention resources to work with, there are real consequences to not getting them to the people who need them most,” said Steve Berg, chief policy officer for the National Alliance to End Homelessness, which has historically frowned on costly prevention programs.But Berg said “it would be good news if these emerging technologies turned out to be effective at predicting who’s most likely to become homeless if they don’t get help.”Attaining that elusive precision will be increasingly important as both the city’s ULA “mansion tax” and the countywide Measure A sales tax begin to direct millions of dollars into homelessness prevention.

The model that picked Escanuela as high risk is being test...