Allocating Interventions Based on Predicted Outcomes: A Case Study on Homelessness Services

Abstract

Modern statistical and machine learning methods are increasingly capable of modeling individual or personalized treatment effects. These predictions could be used to allocate different interventions across populations based on individual characteristics. In many domains, like social services, the availability of different possible interventions can be severely resource limited. This paper considers possible improvements to the allocation of such services in the context of homelessness service provision in a major metropolitan area. Using data from the homeless system, we use a counterfactual approach to show potential for substantial benefits in terms of reducing the number of families who experience repeat episodes of homelessness by choosing optimal allocations (based on predicted outcomes) to a fixed number of beds in different types of homelessness service facilities. Such changes in the allocation mechanism would not be without tradeoffs, however; a significant fraction of households are predicted to have a higher probability of re-entry in the optimal allocation than in the original one. We discuss the efficiency, equity and fairness issues that arise and consider potential implications for policy.

Advancing a Five-Level Typology of Homelessness Prevention

“…Despite an evidence base demonstrating that some groups are at much higher risk of homelessness, efforts to prevent their homelessness have been underwhelming and traditionally dogged by difficulties in accurately predicting who would become homeless absent of intervention (Apicello, 2010). Nonetheless, recent studies suggest that improved data gathering and sharing can aid better predictive decision-making (Gaetz & Dej, 2017;Thomas & Mackie, 2020), and there is a growing practice and literature, mostly from the US, which uses statistical modelling and forms of machine learning to identify individuals at particular risk (Pleace, 2019; see also Greer et al, 2016;Hudson & Vissing, 2010;Kube et al, 2019;Shinn et al, 2013).…”

Subsidy Allocations in the Presence of Income Shocks

“…Our paper can be viewed as belonging to an emerging style of work that uses computational and optimizationbased methods to inform assistance programs aimed at improving access to opportunity for vulnerable populations (Abebe and Goldner 2018a;2018b). A recent study considers allocating interventions for homelessness services using a mixture of counter-factual reasoning and mechanism design (Kube, Das, and Fowler 2018); another studies optimal allocation of financial aid in US colleges based on students' parental income (Findeisen and Sachs 2016). Relative to these papers, our work is the first that we are aware of to take a computational approach to assistance in a setting where the underlying optimization problem exhibits the type of rich stochastic dynamics that characterizes our domain.…”

Event-Driven Continuous Time Bayesian Networks

“…While there has been some related work on helping homeless individuals using planning, e.g. with MDPs (Gehlbach et al 2006;Yi, Finkel, and Goldsmith 2008;Dekhtyar et al 2009) and POMDPs (Yadav et al 2016a;2016b), as well as other studies of intervention allocation (Kube, Das, and Fowler 2018), none of these efforts use the dynamic events setting considered here.…”