We are Mobilise: Homelessness risk study

The Program under evaluation is is a direct cash transfer program conditional on attendance at three group workshops over a six-month period.
The main goals of the workshops are to support participants in:
1. Using goal-setting techniques
2. Creating healthy financial habits
3. Aspiring towards attaining a financially sustainable lifestyle

2025-05-31

2026-09-30

1. Housing Security
2. Financial hardship
3. Homelessness service use

*Housing security*
Using survey data, this outcome will be constructed as the proportion of the total nights spent at secure housing over the last 60 days prior the survey. Insecure housing will be defined as sleeping rough, couch surfing, in refuges or temporary accommodation, crisis accommodation or unsafe housing.

*Financial hardship*
Using survey data, this outcome will be constructed by generating an index ranging from 0 to 6, based on the reporting on experiencing the following challenges in the 2 months prior to the survey.
i. Could not pay electricity, gas or telephone bills on time because of lack of money
ii. Could not pay the mortgage or rent on time because of lack of money
iii. Pawned or sold something because of lack of money
iv. Went without meals because of lack of money
v. Was unable to heat home because of lack of money
vi. Asked for help from welfare/community organisations/foodbank because of lack of money
Each item will be counted as one point.
- We will also calculate the financial hardship index following Anderson (2008).

*Homelessness service use* (two outcomes)
Using administrative data from the Specialist Homelessness Services data, we intend to construct 2 binary indicators:
1. Using the service and being identified as homeless (No shelter or improvised/inadequate dwelling, short-term temporary accommodation, couch surfer or with no tenure, other)
2. Using the service and being identified as homeless or at risk of homelessness

The Specialist Homelessness Services data are not well-documented. Depending on its content, we may have to adjust the measures.

1. Housing stability
2. Employment or education status
3. Food insecurity
4. Service use
5. Standard wellbeing scales
6. Domestic violence victimization
7. Health outcomes and service use

1. Housing stability
- Self-reported number of housing relocations in the past two months (survey data). This will be constructed as an indicator of being "very worried" about having an appropriate place to stay in the upcoming months (vs. being "not worried at all" or "a little bit worried") (survey data)
- Social housing application (from HiiP administrative data). This indicator will be constructed as weather a person applied for community or public housing within follow-up time window, conditional on not already being in the social housing waiting list prior to intervention.
The HiiP data are not well-documented. Depending on its content, we may have to adjust the measure.

2. Employment or education status (survey data)
- Currently working or attending an educational institution.
- Total income from job (including zeros) in the last month

3. Food insecurity (survey data)
- Indicator for the number of the following challenges faced in the last two months because of a lack of money (0 to 3):
i. Went without meals
ii. Ate less than you thought you should
iii. Ate less fresh or nutritious food (fruit, vegetables, meat) than you thought you should

4. Service use (survey data)
One variable computed from the number of the following services used in the last month (0 to 5):
i. Medical services (e.g. GP, hospital)
ii. Housing services provided by the government or community (e.g. shelters, emergency housing)
iii. Justice services (e.g. police)
iv. Legal services (e.g. legal aid)
v. Child protection services

5. Standard wellbeing scales (survey data)
- Life satisfaction (1 to 6) – Satisfaction With Life Scale (SWLS)
- Locus of control (1 to 6) – Internal–External Locus of Control Short Scale–4 (IE-4)
- Perceived stress (1 to 5) – Perceived Stress Scale (PSS-4)
- Social connectedness (1 to 8) – Social Connectedness Scale – Revised (SCS-R)
- Self-efficacy (10 to 40) – General Self-Efficacy Scale (GSS)
- Depression and anxiety (0 to 12) – Patient Health Questionnaire (PHQ-4)
- Overall wellebing index. Following Anderson (2008), we will also construct an overall wellbeing scale using all these items.

6. Domestic violence victimization
- Indicator of domestic violence victimization in the follow-up window (from IRIS and The Orange Door administrative data)
These data are not well-documented. Depending on its content, we may have to adjust the measures.

7. Health outcomes and health service use from administrative data
(from Emergency MGT, admitted episodes, mental health, alcohol and drug administrative data):
-Total cost of emergency services in the follow-up window
- Any use of mental health services in the follow-up window
- Any use of alcohol and drug services in the follow-up window
- Any admitted episode in the follow-up window

These data are not well-documented. Depending on its content, we may have to adjust the measures.

1. Recruitment Process
Young adults will be recruited at the charity partner organization and would be referred to the implementing partner who will assess eligibility and conduct the baseline survey.

Eligibility will be assessed using the following rules:
• Aged 18 years or older,
• Has experienced homelessness in the last 2 months,
• Is not currently experiencing an active drug or alcohol addiction.

Approximately 20% are expected to be ineligible, leading to an expected final sample of 200 individuals. The randomization will be carried out from the pool of eligible individuals.

2. Implementation
Implementation, due to budget and logistic constraints, will be carried out over two intakes (waves) over two years (a.k.a one intake per year. This means in each year we will target 100 individuals. If additional funding is available, a third year intake will be included.

3. Randomization
We will stratify the pool of potential participants based on pre-intervention primary outcomes from the baseline survey (housing security and financial hardship). This randomization will be done in each wave, where names will be drawn randomly from the eligible participant pool to be part of the treatment group until we have 25 individuals and 75 individuals will be allocated to the control group. Potentially, we will also stratified based on other covariates that impact pre-intervention outcomes if the sample allows.

4. Follow-ups
We will contact individuals from both groups 6, 12 and 18 months after the start of the intervention to invite them to fill a survey identical to the baseline survey. We will also follow outcomes in the administrative data for 18 months after the start of the intervention.

Not available

The randomization will be done in office by a computer

We have only 2 groups (control and treatment) and the randomization unit is at the individual level

No

Total sample size will be 200 individuals.

Same as above

50 individuals in treatment and 150 in control. If budget allows we will increase the number of individuals in treatment keeping the ratio 1 treatment to 3 controls.

Based on a pilot experiment, we conducted a power calculation for our housing stability primary outcome (the other primary outcomes were not collected in the pilot). This variable is defined as the proportion of days spent in stable housing in the past two months. In the pilot, at baseline, individuals spent on average 44% of the previous 30 days in stable or safe housing. The pilot also revealed significant attrition in the control group (50%). While we expect to reduce this attrition through stronger relationships with participants, we conservatively assume a 50% attrition rate in the control group for the power analysis. This reduces the control group size from 150 to 75. We set power to the conventional level of 0.8 and use significance levels of 0.1 and 0.05. Assuming the intervention has a positive effect (i.e. using a one-sided proportion test), the Minimum Detectable Effect (MDE) for the intervention relative to the control group is 19.2 percentage points (at the 10% significance level) and 22.4 percentage points (at the 5% level), assuming 50 treated participants and 75 controls. This power analysis is conservative for several reasons: - It does not account for the power gains from stratified randomization. - It does not incorporate the expected improvements in precision from regression adjustment using pre-intervention outcomes and other covariates. - It does not apply to outcomes measured through administrative data, which are not subject to attrition and therefore should yield higher power. We did not incorporate these factors into the analysis because doing so would require strong assumptions about the relationships between pre- and post-intervention outcomes and covariates. Moreover, we do not yet have access to the administrative data and cannot assess baseline values for these outcomes.