In collaboration with the Longhui Educational Bureau, we selected four primary schools to participate in our experiment in the first semester of 2015 (September 2015–February 2016). The experiment lasted for 20 weeks. The four schools were not selected randomly. We required each selected school to have at least three classes in each grade from Grades 3–5 because two types of treatment were introduced, and randomization was implemented at the school-by-grade level. All selected schools are commuter public schools. At the beginning of the first semester, we randomly selected three classes from each grade in each school as our target classes. A total of 36 classes were selected. We randomly chose two classes in each grade in each school as the treated classes (a total of 24 treated classes). These classes were assigned as the complete mixed-seating classrooms, in which low- and high-ability students were randomly paired as deskmates. From the 24 complete mixed-seating classes, half were assigned the MSR strategy and the other half were the pure MS classes. The remaining 12 classes (one class each from Grades 3–5 of the four schools) were the control classes. Our initial sample includes 695 students in the MS classes, 672 students in the MSR classes, and 667 students in the control classes.
At the beginning of the semester, the seats of students in each of the selected 24 treated classes were reassigned to form the complete mixed-seating environment. The basic rules for assigning seats were as follows. First, students in a class were divided into two tracks according to the median score of their average performance on the Chinese and mathematics in the previous end-of-term exam. If a student’s average score on the two courses was higher than the median score of the class, then he/she will be assigned to the upper track. Otherwise, he/she will be assigned to the lower track. Second, students in a class were sorted from the shortest to the tallest within each track and then categorized into tertiles. Specifically, the top 30%, the middle 40%, and the bottom 30% height groups were generated for the upper- and lower-track students. We considered student height in our setting to prevent tall students seated near the blackboard from blocking the view of the short students seated behind them. Finally, we implemented a random sequence of pairing students. We started by assigning lower- and upper-track students in the short-height groups. We randomly selected one student from the upper track and one student from the lower track as deskmates, and the pair was assigned a desk in the front left area of the classroom. Then, we randomly selected the next pair of students, one from each of the two tracks, and assigned them with a desk next to the prior pair of students. Figure 1 illustrates the desk arrangement in a typical classroom. The order of the seating arrangement in the physical space of the classroom for each pair was from left to right and from front to back. After the seats of students in the short-height group were assigned, we assigned seats for students in the middle-height group. Finally, we assigned seats for students in the tall-height group. In case a class had an odd number of students, we picked the tallest student in the lower track and assigned this single student to a desk in the last row of the classroom. In our design, taller students mostly sat behind shorter students.
Next, one of the two treated classes in each grade in each school was chosen to be the MSR treated class, leaving the 12 classes as pure MS classes. Thus, we had 12 MSR and 12 pure MS classes. In the 12 MSR classes, we offered monetary awards and praise certificates as incentives for the 30 upper-track students in the middle and the end of the semester. Award targets were set on the basis of the learning progress of the grade quantiles of the average scores that their low-ability deskmates achieved. Specifically, an upper-track student was offered 100 CNY if his/her deskmate’s improved score ranked in the top 10% in the midterm or final exams among lower-track students in the class. Given that the average class size is 50, a 10% award rate implies that 2.5 students in an MSR class can receive rewards for each exam (in total, 12×2.5=30 upper-track students are offered rewards in each exam). Midterm and final exams were conducted during the experiment. Therefore, an upper-track student may receive a packaged benefit of 200 CNY in total (or approximately 30 USD when using the nominal exchange rate to transfer). This amount can be considered a financial incentive for upper-track students to help their low-ability deskmates on study. The reward plan was announced in the MSR classes after the completion of the seating arrangement, but no students in the MS and control classes were informed about the reward plan. The total expense of our program is 6,000 RMB (or 900USD)