Using Machine Learning to Target Assistance: Identifying Tenants at Risk of Landlord Harassment

Rebecca Johnson , Princeton University
Jerica Copeny, Evansville Public Library
Samantha Fu, London School of Economics
Teng Ye, University of Michigan
Bridgit Donnelly, New York City Public Engagement Unit
Alex Freeman, New York City Public Engagement Unit
Joe Walsh, University of Chicago Center for Data Science and Public Policy
Rayid Ghani, University of Chicago Center for Data Science and Public Policy

Rent-stabilization can combat housing insecurity. Yet landlords can use legal loopholes--most notably, the ability to raise the rent each time a tenant moves out until it exceeds a threshold--to deregulate these units. This loophole incentivizes landlords to illegally harass tenants through tactics like neglecting essential repairs or turning off heat to drive tenants out. In this project, we use large-scale administrative and API data to predict tenant harassment in New York City (NYC). We partner with an NYC agency that knocks on the doors of and offers assistance to at-risk tenants. Currently, there is wide variation in the likelihood that a particular knock helps the agency discover harassment. We use machine learning to predict tenant harassment in order to help the agency prioritize door knocks to the highest-risk tenants. We discuss preliminary results that show how model-based targeting can mean the same quantity of resources helps more low-income renters.

See paper

 Presented in Session 100. Using Big Data in Population Research