Association for Public Policy Analysis & Management

Prior research investigating temporal typologies of homelessness has had a meaningful impact on policy responses intended to prevent and end homelessness.  For example, Kuhn and Culhane (1998) developed a three-category temporal typology of homelessness in which persons identified as experiencing long-term or chronic homelessness were found to comprise roughly 10% of all users of emergency shelter, but consume roughly half of all emergency shelter days.  This finding proved highly influential in the creation of targeted policy initiatives at the federal, state and local levels intended to address chronic homelessness.  However, there has been no attempt to date to develop a temporal typology of homelessness specifically within the population of U.S. military veterans who experience homelessness.  This lack of evidence presents an obstacle to the U.S. Department of Veterans Affairs’ (VA) goal of preventing and ending homelessness among Veterans.  The need for a specific typology of homelessness among Veterans is especially pressing because prior research on the more general homeless population has relied solely on data from emergency shelters, whereas the VA operates a range of specialized homeless assistance programs, including those that provide emergency shelter, transitional housing, permanent supportive housing, and homelessness prevention.  Thus, findings from existing research may be less relevant to the unique service context that exists for veterans experiencing homelessness.  

To address this need, the present study develops a temporal typology of veteran homelessness using administrative data from the U.S. Department of Veterans Affairs (VA) National Homeless Registry and VA electronic medical record data.  The study uses a time-patterned approach to identify distinct groups within a cohort of nearly 90,000 Veterans who used VA homeless services for the first time 2013. I track this cohort’s use of VA homeless programs prospectively over a three-year period, which is divided into 36 discrete one month periods.  I use optimal matching to create a dissimilarity matrix capturing differences among veterans’ sequences of VA homeless program use over the study’s observation period.  This dissimilarity matrix is then subjected to hierarchical cluster analysis using Ward’s method to identify distinct groups of homeless Veterans.  This analysis identifies four groups of homeless veterans:  short-term single program users, long-term transitional housing users, permanent supportive housing users, and multi-program users.  Comparisons of Veterans in each of these groups reveal important differences in terms of their characteristics, and use of VA inpatient and outpatient healthcare services.  I conclude by discussing implications of study results for resource allocation and programmatic decisions.