Supporting teachers’ enactment of classroom practices critical to students’ learning engagement in Mathematics

TPD treatment:

The TPD treatment consists of eight digital professional development (TPD) packages provided to teachers over the course of the year, each addressing topics relevant to fostering student engagement in mathematics instruction. Teachers work with the TPD packages in group sessions.

The content is designed to be effective, engaging, and adapted to Norwegian classroom contexts. All materials are research-based, practice-oriented, and aligned with the national curriculum.
Each TPD package consists of three sessions (each 40–60 minutes):
1- A session in which teachers, together with a colleague, reflect on their current instructional practices.
2- A session providing accessible subject-matter content and suggestions for practical implementation in the classroom.
3- A session where teachers reflect together after implementing the practices in their classrooms.

Teachers in the TPD treatment group will also receive access to a task bank, enabling them to adapt the content of each package to the specific topics currently being taught in their classrooms.

Topics covered in the TPD packages include:

- Characteristics of high-quality mathematics tasks that engage students.
- Strategies for introducing mathematics tasks in ways that spark student interest.
- Approaches to responding to student questions and providing feedback that encourages persistence.
- Methods for facilitating whole-class discussions in which students articulate their mathematical thinking.
- Using student errors as instructional resources.
- Representing and analyzing students’ mathematical thinking and representations.
- Guiding students toward the intended mathematical learning goals during instruction.

The TPD packages were developed by the Norwegian Centre for Mathematics Education in collaboration with the Enact project team.

The feedback treatment:
After a short introduction that explains the purpose of the feedback, teachers receive feedback about their teaching practices every three weeks using a web application. All students in a class, independent of whether or not their parents consented to participating in the study, will be asked to provide anonymous feedback to their teachers.

2025-09-08

2027-05-29

Student Outcomes:
Survey:
- Math engagement (Wang et al., 2016): 38 items, 5-point Likert scale
- Math motivation (Gnambs and Handstingl, 2014): 14 items, 5-point Likert scale
Achievement:
- Math test (developed by the Norwegian Math center for this project, based on the items from the national test)

Mechanisms:

Teacher level:
- Mindset (Yeager et al., 2018): 4 questions, 6-point Likert scale
- Knowledge about engaging practices (self-developped)
- Classroom quality (Wang et al., 2020) 13 items, 5-point Likert scale

Student level:
- Mindset (Yeager et al., 2018): 4 questions, 6-point Likert scale
- Teacher Support (Downer et al., 2015) 37 items, 5-point Likert scale
- Perception of teacher practices (self-developed)

Other secondary Outcomes:
- Teacher intention to quit (Michaels and Spector (1982)
- Students’ Math performance (on standardized tests in grades 8 and 9)
- Students' Math grades ( grades in mathematics upon completion in lower-secondary school)
- School completion (completion of lower-secondary school, upper secondary education)
- Choice of further education (field of study/vocational training)

We conduct an RCT in Norwegian schools. At the point of this pre-registration, the
recruitment for the first wave is finished, but we do not know yet how many we will be able to recruit in the second wave. Below, we describe the intervention timeline, randomization, and sample for the first wave. We will update this pre-analysis plan to include the second wave once the recruitment for the second wave is done.

First wave:
In the first wave, we were able to recruit 35 schools.

Before the summer break, we ask the teachers to perform the baseline math-test with the students. Right after the break, we collect baseline survey data on the teachers and the students. In week 37, teachers in the TPD treatment start with the training. The first feedback from students for teachers in the Feedback treatment, is collected in week 39 (and every third week after that). The TPD and the Feedback intervention end in April 2026.

In May 2026, we collect endline data.

We will follow a similar set-up in the second wave.

In addition, we will receive registry data from Statistics Norway on standardized math tests (nasjonale prøver) in grades 8 and 9 as well as grades in mathematics upon completion in of lower-secondary school, upper secondary education and information on education completion and study choice.

The main research question of the project is whether the TPD and/or the Feedback intervention can improve students' math engagement and performance. In addition, we want to investigate interaction effects between the two treatments.

Main hypotheses:
The primary outcome variables will be higher in TPD than in the control schools.
The primary outcome variables will be higher if the students’ teacher received the Feedback than when they didn’t receive feedback.
There is a positive interaction effect between the TPD and Feedback interventions.

Not available

STATA – for school-level randomization we simply generate a random number, for grade-level randomization we use the stratarand command and randomize within school blocks.

We followed a two-step randomization structure. We first generate school pairs using the greedy matching algorithm (King et al, 2008; Bruhn and McKenzie, 2009). We match schools based on test scores in national tests, the number of participating teachers, and dummies for the region the school is in. We then randomly assign one school per pair to the TPD treatment.

Within each school, we randomly assign grades to the Feedback or the control (business as usual) treatment. If a teacher teaches in two classes in different grades, we pool these classes and randomly assign the teacher to treatment or control. We also generate one block with all schools in which only one teacher participated and randomly assign half of these teachers to the treatment.

Yes

In the first wave, we have 35 schools (clusters) for the TPD. One school (school 36) dropped out of the project after randomization but before the data collection or intervention started.

For the Feedback treatment, we have 63 grades (with 94 teachers) across which we randomize in the first wave. In addition, two schools have classes with fewer than 5 students, which forced us to exclude them from the feedback intervention and reduced the number of grades in the first wave to 61.

Assuming an average class size of 21 students per class and at least one class per teacher, we expect a potential sample of around 1974 students. With a conservative estimate of a consent rate of 85%, we expect 1678 students to participate in the first wave of the experiment.

In the first wave, we have 18 control and 17 treatment schools (clusters) in each of the TPD treatment arms. In each of the two Feedback treatment arms, we have 30 grades receiving the feedback and 31 not receiving the treatment (clusters).

We conduct power calculations both for our current sample and with twice our current sample size, hoping that we will be able to double the sample size with the second wave. For simplicity, we use the sample used for randomization with 36 schools and assume that there are 2 grades in each school (following our planned design but not accounting for peculiarities of individual schools), with on average 26 children per grade, so the resulting power is an estimate. This means that for wave 1 (2), we calculate power with 36 (72) schools and 72 (144) grades. We use Optimal Design Software for all power calculations. The main interest of the project is the combined effect of the treatments. To test, we compare 18 (36) grades in TPD schools that receive the FEEDBACK treatment to the 18 (36) grades in NO TPD schools not receiving the FEEDBACK treatment. We account for clustering at the school level and assume a conservative intra-school correlation coefficient of 0.1. We also account for the 18 blocks we get through our pairwise design and assume an effect size variability of 0.01. Assuming that there are 26 children in each grade, so 52 children in each block, we are able to detect an effect of 0.35 (0.24) standard deviations with a power of 80% assuming α = 0.05. If we include student characteristics and pre-treatment assessment scores as controls, then statistical power is likely to be improved. We are able to detect the same effect size for the TPD treatment by comparing the 36 (72) NO FEEDBACK classes between the 18 (36) TPD and the 18(36) NO TPD schools. We are also able to detect a similar effect size for the feedback treatment, where we only consider the 18 (36) NO TPD schools/blocks and compare the 18 (36) Feedback grades to the 18 (36) No Feedback grades within school. As described in more detail in the experimental design, we will also use OLS regressions with interaction terms to measure treatment effects, controlling for baseline measures and block fixed effects, which should improve power.