Recent research and policy making concerning homelessness has focused on the particular challenges associated with ending chronic homelessness. It is generally assumed that individuals who have been homeless for longer durations have a greater impact on the public system of care and are in need of the most assistance. The aim of this study was to examine key sample characteristics by duration of homelessness (total lifetime duration and longest single episode) using bivariate and multivariable regression analysis. Participants (n = 425) were adults recruited in Vancouver, BC on the basis of absolute homelessness or precarious housing and a current mental disorder. Interviewer-administered questionnaires elicited details concerning demographics, mental illness, substance use, service use, and duration of homelessness. In multivariable models, “persistent” homelessness (lifetime duration of three or more years) was independently predicted by male gender, older age, younger age when first homeless, incomplete high school, past month alcohol use, and daily illicit drug use. “Prolonged” homelessness, a single episode of one year or more, was independently predicted by older age, younger age when first homeless, current substance dependence, daily illicit drug use, and not seeing a psychiatrist in the past month. Substance use is strongly associated with prolonged and persistent homelessness among people with mental disorders, as is the early experience of first becoming homeless. Our findings replicate and extend those of previous studies, and are discussed in terms of their implications for service delivery and the broader construct of social inclusion.