Social investment in practice: The evaluation of RELIANCE, an intensive experimental programme to support the return to work of single parents on long-term welfare in France

The RELIANCE action is led by an association named Arélia in partnership with two others: the Ulis and Ecoval associations. It consists in supporting heads of single-parent families who are RSA beneficiaries and help them reaching sustainable employment in the Nancy Metropolitan area. Each year of the project (2018-2020) sees about 100 participants, with an objective of about 300 beneficiaries at the end of the pilot. The resources of the services and know-how of the 3 associative partners are mobilized in the fields of:

- professional integration: construction, validation of the professional project, training, job tests, evaluations and internships, access to available aid and entitlements, to the resource bank, refresher courses,
- support for parenting and finding answers to custody arrangements,
- exchange networks, mobility.

In addition, the public is supported, backed and encouraged by collective activities and individual coaching, carried by a team of qualified and experienced professionals.

The objective of the support is therefore:

- the development of the person's autonomy in order to carry out the professional integration procedures,
- access to employment through the implementation of career stages adapted to the individual, taken into account as a whole,
- access to training
- social integration (access to social and medical rights, housing, support for parenthood, strengthening of social ties).


The objective for the project is to reach at least 25 \% positive exit to employment, that is: sustainable employment (permanent or fixed-term contracts or temporary assignments of more than six months, business creation or takeover, internship or integration into the civil service), transitional employment (fixed-term contracts or temporary assignments of less than six months and subsidized contracts), positive departures (training, hiring in a SIAE, and other departures recognized as positive).

Arelia's permanent team consists of 4.5 full-time equivalents (FTEs):
- 1 administrative officer (1 FTE)
- a project team leader (0.5 fte)
- 3 social workers (3 fte)

Arelia will be the sole referent for the beneficiaries.

The Reliance scheme has premises in Vandœuvre, close to public transport. In addition to the reception and offices, the premises include two play areas for children, as well as a changing room and a dining area where children can be looked after collectively while waiting to find permanent childcare facilities.

2018-01-01

2021-05-01

- Registration of Labor income within 3 years since randomization
- Exit of RSA
- Entry in PPA
- Long term employment (more than 6 months with labour outcome)
- Amount of all transfers received from the Family benefit allowance
- Subjective well-being based on items from validated research questionnaires.
- A self-assessment of health status
- A self-assessment of living conditions

The primary outcomes for this experiment come from 2 main sources. Administrative data and surveys.
From the administrative database, we have access to extensive and consolidated data on benefit, household incomes from which these benefits are calculated, and many individuals and family characteristics, with a 6 month delay from the actual declaration and payment.
Some family may leave these database due to benefit terminations for other motives than employment. We will assess thoroughly these situations.

The survey are conducted in three waves :
- A first survey will take place in the first quarter of 2021 for all cohorts, which is about 3 years after randomization for the 1st cohort, 2 years for the second and right after the end of the program for the last cohort.
- A second wave will take place one year after and only concerns cohort 2 and 3.
- A third wave will take place in early 2023 only for the 3rd cohort.

These survey aim at assessing long term effect of the program (3 years after randomization) on subjective measures.
Scores will be calculated using IRT models when available, or PCA.

Each year, a sample of 500 beneficiaries is randomly selected from the administrative databases of the departmental council among households that satisfy the eligibility criteria (single parent less than 50 year-old on welfare for more than 2 years). Each file is examined by the territorial integration referents in order to identify those already in training who would therefore not be available for the programme. Each year, approximately 100 families are removed from this initial sample.

The remaining sample (± 400 beneficiaries) is then randomly assigned to "encouragement" and "control" with a block-design according to the variables number of children (1, 2, 3 and +), seniority in RSA (2 to 5 years, 5 to 10, + 10 years) and enrolment in a job centre.

The encouraged group is then contacted to take part in an information meeting by mail and phone follow-up carried out by the CAF of Meurthe-et-Moselle and the RELIANCE team.

The experiment started in early 2018 and concerns 3 experimental cohorts. Each of them participates for a full year to the program although they can leave earlier if they find a job. If the job ends during the period, they can come back to Reliance until the end of the year.

Encouragement design with blocking according to the variables number of children (1, 2, 3 and +), seniority in RSA (2 to 5 years, 5 to 10, + 10 years) and enrolment in a job centre.
The implementation of the randomization was done using Stata by the principal research officer, with fixed seeds for replication.

Unit of randomization is individuals within blocks, within cohorts. There are no clusters

No

1200

1200

600

Initially, the project managers estimated a take-up rate of 50 % and simple power calculation were carried through with this objective. With an initial sample size of 1200, 12 % attrition, mean employment in control group of 10 % (SD=.3), 80 % power and 5 % two sided test, the minimum detectable effect for the probability of employment was 10 pp. However, the take-up rate is closer to 1/3 and the MDE is thus .17pp for a single binary outcome. However, these naïve power calculation were carried through as if we had only one observation per individual whereas for some important outcomes (especially the amount of transfers, probability of being on welfare, etc.) we have monthly administrative data and important socio-demographic characteristics. Thus, we will perform some more precise power calculations using placebo samples with these administrative data.