The Impact of Homeless Encampment Closures: Evidence from Seattle, WA

• In Event: Unsheltered Homelessness, Shelter Attributes, and Housing Stability: Evidence from Policy, Practice, and Lived Experience

Abstract

"Point-in-time" counts of those experiencing unsheltered homelessness in the US are at their highest levels in over a decade. In response, some cities have undertaken "encampment closures" in which individuals residing at an encampment site are made offers for shelter but, regardless of take-up, must evacuate within a short period of time. Prior qualitative research highlights the disruptive and potentially harmful nature of closures for those living unsheltered (e.g., Qi et al., 2022; Kuhn et al., 2022). Nonetheless, closures have continued to grow in prevalence, with justifications including neighborhood benefits such as public safety. Padmanabhan et al. (2024) provides pre-post estimates of crime related to encampment closures in Denver and finds modest, hyperlocal decreases in some crime types but increases in crimes against persons. The present paper builds on this work to (1) estimate the causal impact of encampment closures, and (2) consider a broader range of neighborhood outcomes encompassing not just crime incidence, but also police activity, calls for service (emergency and non-emergency), and foot traffic (as measured by mobile-phone mobility data).

Specifically, this paper studies encampment closures in Seattle, WA from 2017 to 2020. I construct a novel dataset of encampment locations and closures by mapping encampment site boundaries and linking them to administrative records by location. To estimate causal effects, I use a difference-in-differences approach with a matched comparison group designed to avoid bias due to spatial spillovers. For each encampment site, I identify comparison areas ("clusters") within Seattle that (1) are a similar distance to downtown, (2) are located in census tracts with similar socioeconomic characteristics, and (3) are more than 1km away from any encampment closed within the given site's event window (eight weeks prior and sixteen weeks after).

Two broad sets of findings emerge. First, encampment closures do not seem to prevent future encampments at the same location: approximately 46 percent of closures in the sample intersect with the boundaries of at least one previously closed site, and the average time between these repeated closures is 121 days. Second, the effects of closures on surrounding areas are largely null: estimates for crime, emergency calls for service, and foot traffic cannot reject zero and are relatively precise, with confidence intervals often excluding impacts larger than ±10 percent of pre-period means. One notable exception is non-emergency calls for service, where closures appear to produce a modest short-term decline (−11 percent relative to a pre-period mean of roughly 2.4 calls per week).

The high rate of repeated encampment site locations is consistent with encampment closures temporarily displacing unsheltered individuals rather than meaningfully reducing the prevalence of unsheltered homelessness. The largely null effects on crime and emergency calls further suggest that closures do not substantially or durably improve public safety in adjacent areas. At the same time, the decline in non-emergency calls for service offers a potential explanation for why localities continue to pursue closures: they appear to reduce, at least temporarily, the most visible aspects of unsheltered homelessness that originally prompt such records.

Author

• Emily Merola, Princeton University
  Presenting Author