Patterns of exits from housing in a homelessness system of care: the case of Calgary, Alberta

Abstract

Public officials around the world seek to target subsidized housing as purposely and efficiently as possible. With limited availability of subsidized housing, it is helpful to know which household types require specific types of housing support and for how long. With this in mind, we undertake survival analysis and hazard models on clients placed into housing funded by the Calgary Homeless Foundation (CHF) to characterize patterns of exit from Calgary’s homeless system of care. To do this, we use data from Calgary’s Homelessness Management Information System from 1 April 2012 until 31 March 2015. We find singles without dependents to require housing support for the longest period of time, while families require the support for the least amount of time. One important finding is that women require housing support for longer periods of time than men (even though we control for employment and income).

Keywords:

• Survival analysis
• hazard models
• housing support
• homelessness

Acknowledgements

The authors wish to acknowledge very helpful comments from the Editor, four anonymous referees, as well as Ron Kneebone and Herb Emery for comments and suggestions that greatly improved the paper. They also wish to thank Meaghan Bell for encouraging the use of Calgary’s HMIS data for a research project. Finally, they wish to thank Carol Adair, Tim Aubry, Carla Babiuk, Rachel Campbell, Janice Chan, Jim Dunn, Herb Emery, William Ghali, Candice Giammarino, Leslie Hill, Stephen Hwang, Eric Latimer, Riley Lendrum, Aaron Li, Kevin McNichol, Steve Metraux, Ann Montgomery, Geoffrey Nelson, Natalie Noble, Pat O’Campo, Steven Richardson, Cheryl Selinger, Richard Shillington and Ken Swift for assistance in writing the paper. Any errors lie with the authors.

Notes

1. Programs that receive funding from the Calgary Homeless Foundation are monitored against Key Performance Indicators. Falvo provides a general overview of them (Falvo, 2017b).

2. Graduation, which will be elaborated on below, is one form of exit from CHF-funded housing.

3. Today, more than one dozen Canadian cities have such a plan.

4. It is not entirely clear to the authors, or senior CHF officials, why the Government of Alberta has used this particular concept.

5. Upon graduation, some clients are eligible to receive the Graduate Rent Subsidy programme. To quality, they must have received Housing First case management for at least 12 months and been stably housed (i.e. no eviction) for at least six months. If their housing is stable, they must no longer require case management. This gets reassessed every six months.

6. Just one housing programme in this study uses the ACT model; it is geared toward clients with higher acuity levels.

7. Interestingly, it found that just drug or alcohol use on its own was not positively correlated with longer spells.

8. Parametric methods assume that the underlying distribution of the survival times follows certain known probability distributions such as exponential and Weibull, while in nonparametric methods, we estimate the model without parametric assumptions about the form of survival distribution.

9. The assessment forms for the housing programs can be found online: http://calgaryhomeless.com/agencies/hmis/user-information-tools/.

10. Calgary’s SPDAT assessment form can be found here: http://calgaryhomeless.com/content/uploads/SPDAT-Demographics-UDEs-Assessment-Form.doc.

11. At any one time, approximately 2000 households are housed in CHF-funded housing programs (throughout Calgary, there are an estimated 3000 units of subsidized housing with supports, managed by 26 different non-profit organizations). The present study contains fewer than 2000 observations due to the following three factors: (1) left-censoring; (2) right-censoring; and (3) the exclusion of multiple-spell households.

12. We started with 2651 observations, but after controlling for left and right censoring – as well as re-entries – we were left with 1542 observations.

13. The event is equal to zero when the client either is still in the programme by the end of our study or s/he exited the programme for non-graduation or death reasons.

14. Lower graduation rates in housing without time limits should not necessarily be viewed as a policy failure, especially when one considers the cost savings associated with housing a person as opposed to paying for them to stay in an emergency shelter (Goering et al., 2014; Pomeroy, 2005).

15. The analysis treats singles who are 24 years of age and younger as youth, even if they are not in youth-specific housing. Results for such individuals are therefore included in Table 7.