Breaking the Negative Spiral of Disruptive School Behavior – Experimental Evidence about how an Exogenous Shock to Primary School Students’ School Behavior Affects their Achievement

Treated and control students both will receive an intervention implemented by the same coaches. The intervention consists of the same elements: information and a 50-day campaign.
Treated students will receive information about troublesome school behavior and in a 50-day campaign will be trained to avoid troublesome school behaviors. Control students will receive information about unhealthy food and in a 50-day campaign will be trained to avoid the consumption of unhealthy food.

2020-02-03

2020-06-30

Our primary outcome variable is students’ endline GPA calculated from teacher-awarded grades from three core subjects (Hungarian grammar, Hungarian literature, and mathematics). Our second primary variable is the students’ reading score, which is a grade-specific, competency-based test from the test-bank of the National Assessment of Basic Competencies (NABC) of the Hungarian Educational Authority.

Students’ students’ grades for behavior (1); students’ self-control (2); students' self-reported reported school behavior (3); students' teacher-reported reported school behavior (4).

For another field experiment, in the spring of 2017, we contacted all primary schools in seven contiguous counties of central Hungary via the heads of the local school districts to elicit information about seating practices and obtained initial participation agreements from 55 schools. Out of the original 55 schools that participated in our prior research, 13 schools agreed to join in this new research. We obtained consent from 16 additional schools by sending out letters in autumn 2019 again to all the schools in the same counties.
We contacted these schools again in autumn 2019 and asked for their consent for participation in a new project. We told schools that our recent research has two arms concerning the consequences of troublesome school behavior / unhealthy food. We communicated to schools that we will tell them about the allocation of classrooms to the research arms. We offered schools 45-minute long coaching sessions accompanied by a 50-day campaign aimed at changing maladaptive behavior. We specified for schools the type of data that we will collect from students.
Following the recommendations of Imai, King, and Nall (2009), we undertook pair-matched cluster-randomization, with classrooms as the clusters.
Treated and control students both will receive an intervention implemented by the same coaches. The intervention consists of the same elements: information and a 50-day campaign.
Treated students will receive information about troublesome school behavior and in a 50-day campaign will be trained to avoid troublesome school behaviors. Control students will receive information about unhealthy food and in a 50-day campaign will be trained to avoid the consumption of unhealthy food.
The 50-day campaign is designed to raise and maintain students' awareness. It will help students to recall the information they have received (troublesome behavior / unhealthy food) and motivate them to change their behavior.
In sum, the elements of the intervention will be the same for treated and control students. Both groups will receive information from the same coaches, and both groups will participate in a 50-day intervention aimed at changing their behavior. The only difference between the treated and the control group is the content of the information.

Following the recommendations of Imai, King, and Nall (2009), we undertook pair-matched cluster-randomization, with classrooms as the clusters. The following steps describe the randomization we carried out.
1. We collected baseline grades referring to the end of the last academic year (2018/19).
2. We created a data set containing the class-average grades for the following six school subjects: mathematics (1), Hungarian literature (2), Hungarian grammar (3), diligence (4), behavior (5), and class size (6).
3. In the dataset of class-averages, we calculated a principal component from the six variables. We will refer to the principal component as the blocking score.
4. In each grade level, classrooms were ranked according to the blocking score.
5. We identified the most similar classroom pairs in the ranked data set.
6. Within each classroom-pair, we allocated one classroom to the treated, and one to the control group based on a randomly generated number. To optimize balance, we have repeated this step 1000 times. We computed the p-values for balance on each baseline variables known at randomization (not only for the blocking variable) and chose the randomization with the largest minimal p-value across baseline variables.

classrooms

Yes

148 classrooms and 29 schools

2,898 students

treated classrooms: 74
control classrooms: 74

We used the program “optimal design” (Spybrook et al. 2011) to calculate the statistical power of the field experiment. The study will be a cluster-randomized trial that treats students at Level 2 (classroom level). We will randomize half of the classroom to treated and half of the schools to the control group. We assume we will have 20 students in each classroom (n=20). We have 148 classrooms, half of which will be randomized to the treated, and half to the control group. From our baseline data, we know that the inter-class correlation (ρ) in the baseline GPA is about 0.2. We assume that our blocking variables such as grades in mathematics, Hungarian literature, Hungarian grammar, diligence, and behavior in January 2020 will explain at least 50% of the variance in students' GPA (average of mathematics, Hungarian literature, Hungarian grammar ) in June 2020. Under these assumptions, we receive an effect size (MDES) of 0.17. Thus our design is powerful enough to detect small effects.