Comparing last stable address parcel value versus neighborhood values for a random sample of homeless: An initial test of the “tenant of last resort” hypothesis

Introduction: To date, research on homelessness has failed to develop causal models distinguishing homeless individuals from other high-risk poverty populations using individual-level characteristics. Other research has found availability and affordability of rental units associated with neighborhood rates of homelessness. Based on these sets of findings, it appears possible that economic factors related to housing stock may represent the link between individual-level characteristics and the likelihood of becoming homeless. Building on this, it is hypothesized that “future homeless” individuals may be identified through examining individual housing units, that these persons represent what we term the “tenant of last resort”. This presentation presents an initial, simplified test of this hypothesis.

Methods: The sample of individuals was randomly recruited for a NIDA-funded homelessness study in a large Midwestern city. As part of the interview, individuals identified their last stable address prior to becoming homeless. Using public data from the City Assessor's Office, each address was geocoded and matched to parcel values assessed in the year 2000. This parcel was matched with “neighborhood” parcels (residential units within 400 feet of the address), and mean value calculated. To control for gentrification, parcel values greater than $150,000 were eliminated. To examine the impact of extremely marginal units, assessed values of less than $1,000 were coded as “marginal,” greater than $1,000 as not. Last housed address and neighborhood values were compared using paired-sample t-tests (n=168); marginal units with chi-square tests.

Results: Last stable housing address values averaged less than half of the surrounding neighborhood parcels ($9,000 vs. $21,000); however, these differences were not statistically significant (p=.14). Number of marginal units approached statistical significance (p=.06), with 10% of the last stable units valued less than $1,000 in 2000, as opposed to 2% of neighborhood parcels.

Discussion: While results failed to disprove the null hypothesis, they suggest the promise of the approach. More sophisticated modeling (e.g., change in values over time, occupancy rates for parcels over time) represent a next step in testing the “tenant of last resort” hypothesis, as well as controlling for the date on which the future homeless person became homeless versus the date of the assessed value. If upheld, the ability to identify addresses from which individuals are likely to become homeless has potential to create a true primary prevention response to homelessness.