Leaving welfare for work : evaluation of a French training program for social welfare recipients

We implement our experiment in the region Île-de-France. Participants in our study are unemployed welfare recipients (RSA recipients) who are interested in participating in the training program “Parcours d’Entrée en Emploi” (PEE). Among those interested in the program, we randomly pick participants for the program. The treatment corresponds to being offered to participate in the PEE. Those who are randomly chosen to be not treated are part of the control group and get informed that they cannot be offered participation in PEE.

2020-09-01

2023-07-01

1. Entering (stable) employment

2. Having a professional project

3. Social Inclusion

1. Entering (stable) employment: based on administrative data, we will be able to observe the labor market outcomes for all participants in our experiment (periods of employment, earnings, duration of the employment, type of contract, etc.).

2. Having a professional project: based on a survey which we will conduct around 2 months after randomization, we will observe whether the individuals have an idea about the type of occupation they want to enter. Moreover, we plan to ask them about the importance of different skills for this type of occupation. Based on these questions, we aim to evaluate their knowledge about the occupation they aim to enter.

3. Social Inclusion: in the endline survey (around 10 months after randomization), we will ask the participants in the experiment about their social activities (like meeting friends, watching TV, going out) and their level of mastering the French language. We will also measure to what extent they have access to additional rights (health, housing, etc.). These questions will be used to evaluate the impact of program participation on social inclusion.

In the baseline survey, we try to measure the locus of control and the self-control of individuals using standard items employed in other surveys. We plan to investigate effect heterogeneity along these dimensions and to analyze whether the effectiveness of the program differs depending on the locus of control and the self-control of the participants. The feasibility of this analysis will depend on answers’ reliability (those questions rely on items which might not be understood by by everyone in the whole sample because the share of non-native speakers might be high). We plan to investigate effect heterogeneity along the language dimension if some outcomes are not meaningfully measured for non-native speakers.

We also measure in the baseline survey whether individuals face situations that could potentially deter them from entering a training program (e.g. having no access to childcare and living far away from the closest training center). We aim at analyzing whether the program impacts individuals differently according to these potential barriers. We also measure the maturity of the professional plan and aim at comparing the program effectiveness according to this characteristic.

1. Completing a training program

2. Expectations about their labor market status in one year.

3. Interpersonal trust

4. Self-esteem

5. Ability to deal with everyday life problems.

6. Life Satisfaction

1. Completing a training program: based on administrative and survey data, we will collect information about whether or not individuals completed a training program. We will distinguish between short (up to 4 weeks) and long (more than 4 weeks) training programs.

2. Expectations about their employment in one year : in the midline survey, we will ask the individuals what they think they will do in one year. They have to choose one of the following categories: being employed, being unemployed and being in a training program, being unemployed without being in a training program, being self-employed.

3. Interpersonal trust: in the midline and endline survey, we will observe the level of interpersonal trust. This will be based on a declarative measure (using a Lickert scale) proposed by Falk et al (2016). This measure correlates better with experimental measure of trust comparing to standard survey items.

4. Self-esteem: in the endline survey, we will observe the level of self-esteem based on the Rosenberg scale. The feasibility of this analysis will depend on answers’ reliability (for measuring self-esteem we rely on items which might not be understood by everyone in the whole sample because the share of non-native speakers might be high).

5. Ability to deal with everyday life problems: in the midline and endline surveys we will ask questions about everyday life such as ability to write a letter for administrative purposes or to use a map to go to an appointment.. This will be based on standard survey items used in other surveys.

6. Life Satisfaction : based on a Likert Scale ranging from 0 to 10, we will ask individuals how happy they are in their life. This question will be asked in the endline survey.

In the baseline survey, we try to measure the locus of control and the self-control of individuals using standard items employed in other surveys. We plan to investigate effect heterogeneity along these dimensions and to analyze whether the effectiveness of the program differs depending on the locus of control and the self-control of the participants. The feasibility of this analysis will depend on answers’ reliability (those questions rely on items which might not be understood by everyone in the whole sample because the share of non-native speakers might be high). We plan to investigate effect heterogeneity along the language dimension if some outcomes are not meaningfully measured for non-native speakers.

We also measure in the baseline survey whether individuals face situations that could potentially deter them from entering a training program (e.g. having no access to childcare and living far away from the closest training center). We aim at analyzing whether the program impacts individuals differently according to these potential barriers. We also measure the maturity of the professional plan and aim at comparing the program effectiveness according to this characteristic.

Among the RSA recipients that are interested in the program, we realize a random draw. As soon as two individuals have filled the required surveys to confirm their registration in the same training center, we randomly draw their status: within such a pair, each individual has a 0.5 probability to be assigned to the treatment and comparison group. If during the 5 days period there is not any additional application such that one cannot form a pair, we draw the individual with a probability of 0.5. As soon as another individual applies in the overall Region during the next 5 days, she is assigned the opposite status. Those who are assigned to the treatment group can participate to PEE.

We realize a random draw among the welfare recipients who showed interest in the program. In our case, showing interest means most of the time that the welfare recipient and her counselor filled an online registration form to apply to the PEE program in a particular training center. To confirm the registration the welfare recipient needs to fill a second online form.

As soon as two individuals have filled the required survey to confirm their registration in the same training center, we randomly draw their status: within such a pair, each individual has a 0.5 probability to be assigned to the treatment and comparison group. If during the 5 days period there is not any additional application in the same training center such that one cannot form a pair, we draw the individual with a probability of 0.5. As soon as another individual applies in the region Île-de-France during the next 5 days, the latter is assigned the opposite status.

When an individual is assigned to the treatment group, we send her a notification by sms and email to inform her that she can participate to the next meeting organized by the training center which aims at presenting the program in more details. When the individual is in the control group, she cannot attend the program for a 10 months period ; she is also notified of this information by email and sms. The counselor who referred the welfare recipient is notified of the result of the random draw, and the training center receives a notification when an individual is drawn in the treatment group.

We first collect data through online and phone surveys. All individuals fill two surveys when applying to the PEE program: the first one provides administrative details while in the second one individuals fill additional information regarding their family status or professional project. We call each pair around 45 days after the random draw. We finally interview each individual around 10 months after the random draw.

Randomization done in office by a computer.

The randomization unit is the individual.

No

4000 individuals.

2000 (recipients) * 2 (years) = 4000 (recipients)

2000 recipients control and 2000 recipients treated. The treatment group size is fixed but the sample size can increase if there are more individuals in the control group.

We are able to define a minimum detectable effect (MDE) for the probability of welfare recipients to find a job. The MDE that we can measure depends on the number of prescriptions which are made by social workers and on the compliance rate of individuals once assigned to their treatment status. Relying on a 100% compliance hypothesis, the minimum detectable effect that we can measure on the probability of welfare recipients to find a job is 4 percentage points, with a power of 80% and significance level of 0.05. It can be compared to the average exit rate of RSA of 27% (the most recent French data allow us to compute the exit rate of RSA between the 2016 and 2017 years, see Minima sociaux et prestations sociales – Ménages aux revenus modestes et redistribution, Edition 2019, DREES). It corresponds to 4000 prescriptions, such that 2000 individuals enter the training program (it is the maximum amount of individuals who can enter the program during the experiment) and that 2000 individuals are included in the control group. If the compliance rate decreases to 80%, keeping the power and significance level constant, 5000 prescriptions will be required to ensure that 2000 individuals enter the training program: in this case the MDE that we can measure is 0.046 points of percentage on the probability to find a job. By way of comparison, Blundell et al. (2004) measure the effect of the New Deal for Young People (targeting young unemployed individuals only) and show that the probability of beneficiaries to find a job increases by 5 percentage points, though this effect seems to decrease with time.