Understanding Public Preferences for Policing Homeless Individuals in the United States: Results from a National Survey

ABSTRACT

The United States has a large and growing homeless population. In the name of public order, municipalities across the country have criminalized behaviors associated with homeless people in public spaces (e.g. panhandling) and tasked police with responding to violations. What should police do in these encounters? This study reports on a nationwide survey experiment that asked US residents whether police should arrest, help, or ignore a homeless individual in several hypothetical scenarios. We estimate (1) aggregate preferences for police response, (2) the association between respondent demographics and individual preferences, and (3) the effect of experimentally manipulated identity – gender and background – of a homeless person on preferences. Results reveal that a helping response from police is generally preferred to arresting or ignoring. An arrest response received greater support from people who perceived homelessness to be a problem locally, as well as men and Republicans. The identity of the homeless individual had little effect on preferred police responses. With respect to public and policy debates about homelessness, these results suggest that there is relatively little public appetite for a heavy-handed police response, though this may not hold in areas where many people perceive homelessness to be a source of problems.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ This study was approved by the Human Research Protection Program and Institutional Review Board of Oregon State University (study #8581).

² YouGov interviewed 1613 respondents who were then matched down to a sample of 1500 to produce the final dataset. The respondents were matched to a sampling frame on gender, age, race, and education. The frame was constructed by stratified sampling from the full 2016 American Community Survey 1-year sample with selection within strata by weighted sampling with replacements. The matched cases were weighted to the sampling frame using propensity scores based on age, gender, race/ethnicity, years of education, and region. The propensity scores were grouped into deciles of the estimated propensity score in the frame and post-stratified according to these deciles. The weights were then post-stratified on 2016 Presidential vote choice, and a four-way stratification of gender, age (4-categories), race (4- categories), and education (4-categories), to produce the final weight. We use these weights provided by YouGov.

³ YouGov invited 2,869 panelists to take the survey. Of these, 2,434 started it, and 435 did not respond. Of those who started, 1,711 completed it, 292 partially completed it, and 431 were subsequently screened out. The corresponding response rate is 72.5%, where response rate is calculated as: RR = complete/(complete + partial + (e*no response)) and e = (complete + partial)/starts.

⁴ The attention check asked: “How much do you agree with the following statement? Select Agree below to show you are paying attention.” Possible responses ranged from 1 (Strongly Agree) to 5 (Strongly Disagree), with 4 (Agree) being the one correct response.

⁵ There is no single marginal effect for an interaction term, as a change in that term may be due to a change in either (or both) of the component terms, which would induce different changes in the outcome. Therefore, the vignette gender*background interaction is omitted.

⁶ Winter’s (2021) `combomarginsplot` for Stata was useful for generating these plots.

⁷ We calculated standard errors for the marginal effects using two methods in Stata: the default delta method (presented here) and the unconditional method, which accounts for sampling variability in the covariates. The methods produce virtually identical results. Changing calculation method causes none of the estimates to move above or below the conventional threshold for statistical significance (p < 0.05).

Additional information

Funding

Data collection was funded by the Russell Sage Foundation, grant #G-1807-07074

Notes on contributors

Brett C. Burkhardt

Brett C. Burkhardt is an Associate Professor of Sociology in the School of Public Policy at Oregon State University. His research examines the exercise of coercive power by legal and quasi-legal authorities, with a focus on prisons, police, and firearms. His work has been funded by the National Science Foundation and has been published in a variety of journals, including Criminology & Public Policy, Police Quarterly, Punishment & Society, and Criminal Justice Policy Review. He holds a Ph.D. in sociology from the University of Wisconsin-Madison.

Mark Edwards

Mark Edwards, Professor of Sociology, conducts research and teaches courses on poverty-alleviation programs and causes of various forms of material hardship, especially food insecurity. He provides extensive applied research assistance to nonprofit organizations and state agencies in Oregon. Edwards directs the OSU Policy Analysis Lab, assisting rural and urban nonprofit organizations with projects such as point-in-time-counts of homelessness, community engagement on laws related to homelessness, and distribution of resources to people struggling with housing insecurity.

Scott Akins

Scott Akins is a Professor of Sociology in the School of Public Policy at Oregon State University. His research interests include drug use and policy, structural criminology, and policing of at-risk and vulnerable populations. With Clay Mosher he is co-author of Drugs and Drug Policy: The Control of Consciousness Alteration (2021) and In the Weeds: Demonization, Legalization, and the Evolution of US Marijuana Policy (2019).

Christopher T. Stout

Christopher Stout is an associate professor in the Oregon State University School of Public Policy, and author of The Case for Identity Politics: Polarization, Demographic Change, and Racial Appeals (University of Virginia Press, 2020).