|Migrant children in a suburb
Regardless of their urban tenure, these migrants are classified as “rural,” and their children are not routinely allowed to enroll in urban public schools. As a consequence, thousands of privately run elementary schools have cropped up to serve these migrant children. Unlike public schools, however, these private “migrant schools” require tuition and fees. They are also very weakly regulated and the educational outcomes they produce are often dismal. A conspicuous gap in educational achievement has therefore appeared between migrant children and both their urban and rural counterparts.
Students with weak scholastic performance in migrant schools are particularly vulnerable. Already members of a marginalized student population, these under-performing students are the least likely to acquire the basic skills necessary to succeed in China’s increasingly modern economy. What can be done in the classroom to help these young, at-risk students get on track with their peers?
Peer-tutoring, in which students work together to increase performance, has been shown to be a promising avenue for improving educational achievement among marginalized student populations. Previous studies have argued that tapping into peer resources may be more important in the developing world, where other resources are in serious shortage and many students are not learning effectively. One pilot study in particular was conducted in migrant schools in Shanghai and measured the effects of putting strong and weak performing students together into pairs and having the pairs compete in a semester long competition. The participants were knowingly judged by the improvement in performance of the weak students who participated. The increases in student performance observed in the two-school pilot study were much larger than other more costly remedial educational interventions that have been evaluated in the literature.
REAP designed the current study to encompass a much larger and randomly selected sample of schools, classes and students than the first pilot study. The goal of the experiment is to definitively assess the effect of two different interventions on student performance: peer-tutoring or individual rewards. The results will be compared to the student outcomes in control groups that receive no treatment at all.
Sample and baseline survey
From a previously established list of 340 private migrant schools in and around Beijing, REAP randomly selected 24 schools to participate in the study. From these schools, REAP randomly selected approximately 24 third, fourth, fifth and sixth grade classes, respectively. Each class has approximately 50 students, creating a total pool of approximately 4,800 selected students [(24 schools) X (4 grades) X (@50 students per class)]. All students were between the ages of 7 and 12, half were male and half were female, and nearly all were ethnically Han Chinese. All students took a standardized math test developed by Beijing test experts. This test serves as the baseline survey of academic performance.
|High and low performing students are paired to determine whether peer-tutoring can improve academic performance.|
Of the 24 participating schools, REAP randomly selected 12 schools to serve as the treatment group. In each of these schools one class each from the 3rd, 4th, 5th and 6th grades were selected to conduct peer-tutoring. For the purposes of this study, the ten students in each selected class with the highest baseline test scores were considered the strongest performers. These students were invited to participate as peer tutors. If any declined to participate, REAP invited the next highest scoring student in the class until 10 from each class agreed to participate. For the purposes of this study, the 20 students with the lowest baseline tests scores in each class were considered the poorest performers. Of these 20 students, ten were randomly invited from each class to participate as peer tutees. If any declined, REAP randomly selected one more of the lowest scoring twenty students until 10 students agreed to participate.
The 10 high performing tutors in each class were then matched with the 10 low performing tutees in the same class to make 10 tutoring pairs. The tutor-tutee pairs were taken aside by the teacher who explained that the three tutor-tutee pairs in which the poorer performing students achieve the greatest improvement (out of the 10 participating), as measured on the evaluation test in six months, would receive small rewards. The team with the most (2nd most, 3rd most) improved score by the tutee would share a 100 yuan reward, to be split evenly between tutor and tutee.
Individual reward schools
|What incentives motivate improvement?|
The remaining 12 schools in the study were designated for individual reward testing. The 10 poorest performing students in each class, similarly selected randomly from the 20 poorest performing students, also had the reward and competition explained to them: whoever among them achieved the greatest increase from their baseline test scores on the evaluation test after six months would receive a reward of 100 yuan. Second and third runners up will receive 50 yuan each.
In both sets of schools, REAP also monitored the performance of students in 3rd through 6th grade classes that were not selected for the intervention. The students in these classes serve as the de facto control group for this study (as will the 10 poor performing students in each peer-tutoring and individual class that were not selected to participate in the reward competition), against which we will compare the participating classes and students.
REAP will return six months after the start of the peer-tutoring and individual reward programs to repeat the baseline surveys and give new standardized tests. In addition to the rewards for students with the highest degree of improvement between tests, token gifts will also be provided to all other participating students.
Our results show that those students who participated in the peer tutoring intervention increased their standardized test scores by a statistically significant 0.14 standard deviations.
What does a shift of 0.14 standard deviations mean in day-to-day terms? A loose representation of such a shift might be turning a "C" student into a "B" student. It is worth noting that far more complex and expensive interventions that have been assessed in the past, such as class size reductions, have been shown to produce less positive effects.
There was no positive effect among those students in the individual rewards intervention. Click here to see our results in REAP's working paper, "Cash incentives, peer tutoring, and parental involvement: A study of three educational inputs in a randomized field experiment in China".
The solid lines represent intervention groups, while the
dotted are control groups.