Download presentation
Presentation is loading. Please wait.
Published byMaximillian Powers Modified over 5 years ago
1
Cost Savings of Housing First in a Non-Experimental Setting
By Ali Jadidzadeh, Nick Falvo and Dan Dutton Annual Conference of the Canadian Economics Association June 1, 2019
2
Disclaimer This body of research is the property of Calgary Homeless Foundation, and is used with permission.
3
Overview Housing First Data Method Results Discussion
4
Housing First (HF) No requirement of housing readiness
Various supports provided Strong support from RCTs. Existing research suggestions important reduction in use of health and justice systems.
5
Housing First (HF) Note: Does not include capital costs.
6
This study In this paper, we investigate the impact of HF programs on public service utilization for people experiencing homelessness in Calgary. Specifically, we look at single adults without dependants.
7
This study (cont’d) Our objective is to study the impacts of HF programs where a large central homeless-serving organization (i.e., a ‘system planner’) provides program oversight. We use longitudinal individual-level data on health and justice system utilization to observe costs to public systems before and after joining HF and cost data from the HF program to estimate cost offsets or savings associated with HF.
8
Research questions Does the delivery of HF reduce utilization of the health and legal systems? If so, by how much and over what period?
9
Data Individuals were triaged to HF based largely on their level of homelessness acuity (as determined by the Service Prioritization Decision Assessment Tool). Data for the present study was shared with us by the Calgary Homeless Foundation. Move-in data + follow-up assessments every 3 months. Includes data on health and justice use.
10
Data (cont’d) There are 2,222 individuals in our dataset.
Years: Administrative (i.e., observational) data. To our knowledge, this is the first such Canadian study (using observational data).
11
Variables Variables used to represent health and justice system utilization (our outcome variables) are: # of hospital visits, # of ER visits, and # of interactions w police. All variables are self-reported by the participant, but recorded by a case manager who typically has some familiarity with the participant.
12
Method We estimate two different models of utilization over time in HF programs. Fixed effect OLS models. These estimate average changes in utilization but are not specific to count data. Fixed effect negative binomial models. These estimate changes in the incidence rate of utilization and are specifically designed for count data. Fixed effects are necessary for panel data. Dummies are used for each individual. We’re controlling the variation between individuals. There are other factors no doubt affecting each individual that we don’t understand (they’re not observable). OLS is good for continuous variables. In out dataset, there lots of 0’s. In order to control for those biases in the data, negative binomial models are necessary.
13
Estimating health costs
Source: Housing First Quarterly Follow-Up Assessment: Calgary HMIS. Retrieved from
14
Estimating health costs (cont’d)
We imputed the actual cost of hospital and ER use w costing data provided by the provincial govt (specifically, Alberta Health). Alberta Health publishes reports on the cost of providing hospital and ER services for people experiencing homelessness (which are higher for each visit than for the general pop.).
15
Estimating justice costs (cont’d)
Source: Housing First Quarterly Follow-Up Assessment: Calgary HMIS. Retrieved from
16
Estimating justice costs (cont’d)
We use the estimated cost of the warrant cycle as the average cost of an interaction with police. Issuing a ticket in Calgary costs $139. Arresting an individual who has not paid their tickets costs an additional $135. Resultant court appearance ranges between $222 and $253.
17
Estimating justice costs (cont’d)
If an individual is convicted in Calgary, then one day in jail costs $220. Each warrant cycle in Calgary is estimated to cost a total of $1,376, which we use as the unit cost for a single interaction with police. Annual costs for a client in HF can range from $14K to $31K depending on specific program type.
18
Results $1 spent on HF is associated with more than $2 of savings to the public system. Ergo: the $42M budgeted by Calgary Homeless Foundation for HF for fiscal year could result in savings of more than $84M in terms of hospital visits, ER visits, and justice services.
19
Results (cont’d) We don’t include homeless shelter utilization as a cost saving because we couldn’t reliably estimate how many shelter stays participants would have used had they not been in HF. However, each year of shelter use avoided represents an additional $12K in savings.
20
Discussion Results shows that the triage process developed in Calgary result in large returns to investment. Future research would benefit from tracking the costs and types of service utilization through linked data.
21
Discussion (cont’d) Cuts to social spending programs like HF will result in greater health spending to fund the consequent increase in hospital and ER utilization.
22
Who saves? In the Calgary context, the systems-level savings estimated here will accrue mostly to the province, which funds the health system and courts, and the municipal government that funds the police services.
23
Caveats HF is not one program, nor are estimates of the benefits of HF generalizable to all cities, so it follows that cost savings estimates at a systems-level are also city-specific. This particular study is focused exclusively on single adults without dependants.
24
Caveats (cont’d) We make no effort in this particular study to estimate the benefits to individuals enrolled in HF, for whom the benefits to successful addiction and mental health management are enormous, and who might experience quality of life improvements in the future such as sustained employment.
25
Thank You This manuscript is currently under consideration by a journal.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.