Explaining Shelter Resistance: Housing Preferences of Unsheltered Adults

How to cite

Douglass, Trinity; Pierce, Stephanie; McCasland, Jamie; Weiss, Amy. 2025. "Explaining Shelter Resistance: Housing Preferences of Unsheltered Adults." AEA RCT Registry. AEARCTR-0016308. https://doi.org/10.1257/rct.16308-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Trinity Douglass* , Stephanie Pierce , Jamie McCasland , Amy Weiss

* Primary investigator

Registry ID
AEARCTR-0016308
DOI
https://doi.org/10.1257/rct.16308-2.0
First registered
Last updated
Status
In development
Trial dates
2025-06-30 – 2025-12-31
Countries
Private
JEL codes
I15, R21, C93
Registry record
http://www.socialscienceregistry.org/trials/16308

Abstract

In California, an estimated 120,000 individuals experience unsheltered homelessness (de Sousa et al., 2023). While homeless service providers often cite common shelter attributes as critical factors influencing shelter use, there is limited quantitative evidence supporting these claims. This study employs a conjoint choice experiment to identify the shelter preferences of unhoused individuals. Additionally, we collect survey data from individuals experiencing homelessness in the San Francisco area to examine how these preferences vary by baseline characteristics. Our findings aim to provide a clearer understanding of the decision-making processes surrounding shelter use and to identify key barriers to shelter uptake. These insights have the potential to inform more effective and responsive homeless services and housing policies.

De Sousa, Tanya, et al. The 2023 Annual Homelessness Assessment Report (AHAR) to Congress. DC: United States Department of Housing and Urban Development. https://www.huduser.gov/portal/sites/default/files/pdf/2023-ahar-part-1.Pdf

Intervention

This study uses a conjoint choice experiment to determine how individuals who are unhoused in the greater San Francisco Bay Area value attributes of hypothetical shelter options. We present individuals a repeated choice of two hypothetical shelter options where the levels of the attributes are randomly presented to individuals. For each of these questions individuals are then asked if they would prefer this option to where they stayed the previous night.

Experimental design

The experiment contains a series of survey questions and a conjoint choice experiment. The survey questions gather descriptive information about individuals’ history of homelessness, as well as gathers information about drug usage and health usage. We also collect contact information for future possible research.
In the conjoint analysis individuals are presented with repeated questions that have two steps. The first is choosing between two hypothetical housing options, the second step is answering whether they prefer the hypothetical option they selected to where they stayed last night. Lastly, we validate these preferences by having individuals rank the attributes that they are presented in the conjoint analysis. This helps to ensure consistency in the preferences that are stated through the conjoint experiment.

Design details

Randomization method
In the conjoint analysis there are two sources of randomization. The first is the order in which the attributes are presented, this is randomized at the individual level. The order will remain the same amongst all questions within a block containing a particular set of attributes. The second set of randomizations is to determine which level of attribute is presented for each option within a question. This randomization occurs at the attribute level, where a randomly drawn “level” of that attribute is selected with equal probability across all levels. We employ a pure randomization method for both sources of randomization, with all levels of attributes being presented with equal probability. This randomization method is computerized and generated through the survey software Qualtrics.
Randomization unit
Randomization occurs both at the individual level as well as the attribute level, this is further described in the Randomization Method.
Clusters
500-800 people
Observations
500-800 people
Arms
500-800 people there are no identifiable treated and control groups, all individuals receive the same sources of randomized variation.
Primary outcomes
The main outcomes of interest are the relative values individuals hold for each of the shelter attributes.

Cited by

No citation data available for this registration yet.

Keywords