Online tutoring for disadvantaged students in France

This randomized controlled trial (RCT) evaluates the impact of “HomeClasse”, an online tutoring program focusing on academic performance, soft skills, and schooling orientation after middle school. HomeClass is implemented by a French NGO called ZUPdeCO (ZdC hereafter).

HomeClasse is an online tutoring program targeting students in the last year of middle school in France, particularly those experiencing academic difficulties. The last year of middle school is pivotal due to key milestones like the end-of-middle-school national examination (diplôme national du brevet), which offers a comprehensive measure of academic performance, and critical educational orientation decisions (e.g., vocational vs. general tracks).

The program provides weekly one-hour online tutoring sessions tailored to the needs of each student, focusing on French and mathematics, with the goal of improving grades and developing transversal skills such as confidence, grit, and self-efficacy. Students also obtain access to additional educational resources (e.g., quizzes, exercises, or assistance from teachers). Tutors are volunteers and are mainly i) university students who can earn ECTS credits as part of a volunteering assignment and ii) working people who do volunteering (in their spare time or through their company).

Once registered, tutors follow online training modules focusing mainly on how to deal with pupils (e.g. being a good listener, helping pupils to progress and not doing things for them). Tutors indicate their desired school level, subjects and timetable, and ZdC arranges a match with a student who meets these criteria. As far as possible, ZdC also matches on gender, with a focus to ensure that female students have female tutors.

The evaluation is planned to be conducted for two consecutive years in order to reach a sufficient sample size. A first cohort will be recruited in the school year 2024-2025 and a second cohort in the school year 2025-2026.

2025-01-01

2026-06-30

Academic performance (grades in continuous assessment and end-of-middle-school exam)
Orientation (track choice) after middle school

Primary outcomes are key outcomes we expect the intervention to affect.

Reported educational aspirations
Self-efficacy
Grit
Motivation
Willingness to learn
Anxiety
Time spent on homework
Parental involvement

Secondary outcomes are intermediate outcomes that we expect the intervention to affect and that are useful to understand mechanisms.

Recruitment
We identified schools that were interested to participate in the study in the first half of 2024. Each school designated a referent teacher in charge of communicating with the research team and supporting the recruitment of students.

In practice, the referent teacher identified students that might need additional support and presented the program to them and to their family, thanks to a flyer presenting the program and the study. Interested students could sign up to the program together with their parents by answering a baseline questionnaire.


Randomization

Among all students who filled out the baseline survey, and for whom we obtained consent from both students and parents, we randomize students into either treatment or control groups within schools. We randomize students in batches to accommodate varying recruitment timelines in the different middle schools. For each batch, we perform a randomization stratified by school, sex, a dummy indicating whether the student has already retaken a class, and a dummy indicating whether a student is above or below the within-school median grit value of all the students who registered to the study by answering the baseline questionnaire. If strata contain only one student, we assign them to neighboring strata, randomly in the case of two or more strata with the same overlap. Finally, we split strata with more than 3 students based on our continuous grit measure, so that each strata has 2 or 3 pupils, respectively.


Data collection

The baseline questionnaire measures key socio-demographic variables used for stratification, motivation to participate in the program and study environment, as well as baseline values of socio-emotional skills (see primary outcomes). Endline questionnaire measures socio-emotional skills and aspirations (see primary outcomes), time spent on homework, parental involvement as well as program satisfaction. In addition, we have access to administrative data on students’ academic results and track choices after middle school. Finally, we obtain monitoring data from the implementing partner, with information on the number of sessions attended by the pupils, the themes of the sessions, the rate of absenteeism, as well as a grade (and a short qualitative assessment) given by the tutor for each session.

Not available

Randomization is conducted in-office using Stata, following pre-defined strata.

The unit of randomization is the individual participant (student).

No

the intervention is individualized and not clustered - we originally aimed to recruit students from 60-70 schools. However, in the first wave, we were only able to recruit 35 schools. We expect to not reach 60 schools, and instead decided to extend the evaluation by another year.

Planned Number of Observations (originally): 1,000 students. However, following the first wave of recruitment, schools were not able to recruit as many students as planned to the point we revised our expectations in terms of sample size to around 500.

Planned Treatment Allocation: 50% treatment (250 students, originally 500), 50% control (250 students, originally 500).

We conducted power calculations with a significance level of 5% and a power level set at 80%. We expect an attrition rate of 10% and a take-up rate of 80% (with no crossovers). We expect the R² to vary between 20 to 40% depending on the outcomes of interest. We originally aimed for a sample of 1,000 students. With 1,000 students, these parameters lead to a MDES of 18 to 21% of a standard deviation (SD). However, following the first wave of recruitment, schools were not able to recruit as many students as planned to the point we revised our expectations in terms of sample size. For a sample of 500 students -that is more likely to be reached- the same parameters lead to a MDES of 26 to 30% of a SD.