Abstract
Homelessness prevention programs intervene with households apparently in imminent danger of becoming homeless, and try to keep them housed. If they are at least partially successful, how do they change the average shelter spell of households actually becoming homeless? We use data from 2003 to 2008 for Homebase, a New York City homelessness prevention program that studies have found to be effective in reducing shelter entries. Homebase made no difference in average shelter spells at the community level. This result, like many results about shelter spell length, is not easy to reconcile with the idea that shelter spell length is a reflection of the seriousness of underlying problems.