Tracking the expenses of homelessness and determining reduction in costs is inherently problematic. The numerous studies across the nation have used many different methods to create their numbers – each based on contextually valid logic and available knowledge and examples. Some studies have had access to comprehensive HMIS databases and medical records and have tracked individual people. Other studies have found citywide costs and divided them by the homeless count. Others have taken service provider data and made estimates based on average client costs. Ideally, these studies would track individuals and the greater population over time to form a longitudinal study measuring the actual and direct costs for a community. Given the transitory nature of the homeless population and resulting problems in tracking and contacting specific homeless individuals, it has thus far been extremely difficult to construct the type of longitudinal study of the homeless population that would shed light on patterns of individual service usage over time and key services or factors that lead to an exit from the state of homelessness (Allgood & Warren 2003). Also, taking into account the higher costs of such longitudinal studies, most studies to date have been based on cross-sectional data, which are limited in their ability to account for temporal factors in the problem of homelessness and tend to capture those who are homeless for an extended period of time (Early 2002). This study uses a hybrid of several other designs to conform to the needs and realities of Metro Nashville. It is impractical and expensive to create a longitudinal study so a cross-sectional design has been used. As the Nashville HMIS lacks comprehensive tracking of individuals through services, this study used data from service providers, people experiencing homelessness, and archival data to create average costs per individual or total community costs that are divided by the homeless count estimate.
The main weakness of using averages that are not directly attached to specific individuals is determining whether a few people are using many services and therefore incurring much of the costs or if the service use is distributed more evenly. The survey provides data supporting our methodology – in that services are spread out relatively evenly, even though there are some “frequent flyers.” Still, considering the relative cost and effectiveness of the methods used, the hybrid design for this study has captured as many good features from other studies as possible. The main goal is to create numbers that can be used with confidence that are not inflated due to poor methodology or selection bias for people tracked consistently in an HMIS.