High Dosage Math Tutoring in Dutch Primary Education: Evidence from a Disadvantaged Neighborhood in Amsterdam

Randomly selected 5th graders take part in a high dosage math tutoring program. The program was implemented by The Bridge Learning Interventions (www.TBLI.nl). Randomly selected students receive 45 minutes of additional math tutoring classes per day, 5 days a week, for the entire school year. The tutoring took place during regular school hours and replaced some regular instruction. In principle, the tutoring was not meant to replace some of the regular mathematics instruction.

In the beginning of the school year teams of 1 tutor and 2 students were formed. Students and teachers work in these teams for the entire duration of the project. The tutoring classes follow a
mathematics curriculum that is based on the (Dutch) 1F-level, a minimum benchmark level of
achievement that most student should be able to reach by the end of 6th grade. The tutor delivers personalized instruction. Each student's skill level determines the pace at which they work through the curriculum. No steps or subjects are skipped. During the year, tutors build relationships with students and help them develop academic and socioemotional skills (such as confidence). In addition, tutors have weekly phone calls with the parents of their students.

The intention of all parties involved is to repeat the RCT for two more years, for two new cohorts of 5th graders. This however will depend on willingness of (the same or other) schools to participate as well as on funding for implementation.

2019-09-02

2020-06-30

5th grade and 6th grade math test scores and secondary school track placement. If technically feasible, we also intend to follow these students into secondary school and monitor track enrollment, regardless of whether we find effects on test scores.

For test scores we use tests from the Cito student monitoring system ("leerlingvolgsystemen") as well as TOA tests (math, nonlanguage math) that are collected and administered by The Bridge Learning Interventions. Secondary school track placement is determined in 6th grade, towards the end of primary educuation. Teachers and schools use test results and other information and decide on the secondary school track that best matches the student's level of development. This track recommendation is important and can restrict the options for track enrollment in secondary school.

Reading test scores (Cito and TOA), Strengths and Difficulties Questionnaire

The intervention takes place in 5 primary schools in the Amsterdam South-East area. Relatively low achieving students (math scores in the bottom 60% nationally) and who are willing to participate (through active consent) were eligible for the tutoring program. A total of 56 (out of 105) students were selected through stratified/block random sampling to take part in the tutoring program.

The intervention was implemented in cooperation with the 5 primary schools as well as the local govenment who financed the project. Students scoring in the bottom 60% nationally based on the 4th grade test, and for which parents signed a consent form, were eligible and were randomized into a treatment and a control group. We rely on stratified or block randomization:
1. Students were ranked on the basis of their mid-year 4thg grade test score, within school, class and test type (there are two versions of the math test used for testing students).
2. Based on this ranking we made pairs (groups of two) or triplets (groups of three).
3. Within pairs and triplets we select 1 (for pairs) or 2 (for triplets) students into the treatment group at random. (This implies that we have more treated units than control units. This imbalance is due to operational restrictions.)

Randomization was done in the office, by computer (Stata)

student

No

105 students

105 students

56 in treatment, 49 in control

Baseline (4th grade) tests can explain about 75% of the variation in the outcome scores (5th grade, and 6th grade tests). The standard deviation of the residual scores (the unexplained variation in the outcome) is about 6.5 score points (i.e. about 0.6 population standard deviations). Based on this (and the sample sizes in treatment and control group) we estimate that the standard error for the estimated treatment effect will be 1.25 score points. The MDE (power: 80%, alpha: 5%) is then 2.8*1.25=3.5 score points, or 0.30 population level standard deviations. Our previous research on a comparable intervention in the Netherlands (but with a bigger sample size) yielded a point estimate of 3.3 score points (or about 0.28 population standard deviations). The present intervention however is more intensive (5 instead of 4 days a week, and in addition to regular math classes (rather than partly replacing regular math classes), for example).