Tracking homeless individuals over time has proved to be extremely difficult; thus, only limited longitudinal data on the homeless exist. We analyze longitudinal data originally collected from the New Orleans Homeless Substance Abusers Program in 1991-1993, supplemented with mortality data for the same sample by year 2010. We use social bonding theory to examine the effect of conventional social ties on mortality among a sample of substance abusing homeless people. This is of special concern when researching the older homeless persons. We find that social bonding theory does not help to understand mortality among this population. However, alcohol abuse, as compared to crack cocaine, does increase the likelihood of early mortality.