A journey home: What drives how long people are homeless?

This paper uses survival analysis to model exits from two alternative forms of homelessness: sleeping on the streets (‘literal homelessness’) and not having a home of one's own (‘housing insecurity’). We are unique in being able to account for time-invariant, unobserved heterogeneity. Like previous researchers, we find results consistent with negative duration dependence in models which ignore unobserved heterogeneity. However, controlling for unobserved heterogeneity, we find that duration dependence has an inverted U-shape with exit rates initially increasing (indicating positive duration dependence) and then falling. Exit rates out of both literal homelessness and housing insecurity fall with age. Women are more likely than men to exit housing insecurity for a home of their own, but are less likely to exit literal homelessness. Persons with dependent children have higher exit rates. Finally, education seems to protect people from longer periods of housing insecurity.