Replication and robustness analysis of the randomized peer intervention study by Wu et al. (2023), examining the effects of mixed seating arrangements and incentive-based peer interactions on students’ academic performance in Chinese primary schools.
This project replicates key results from Wu et al. (2023), who study how peer effects in the classroom influence academic outcomes using a randomized controlled trial (RCT).
The original experiment randomly assigned students to:
- Mixed Seating (MS): pairing high- and low-performing students as desk mates
- Mixed Seating with Rewards (MSR): the same setup, but high-performing students received rewards if their low-performing desk mate improved
The replication focuses on whether these interventions affect average test scores, Chinese scores, and mathematics scores, with particular attention to differences between lower-track and upper-track students.
Completed (2025). Academic replication project.
Low-and high-performing deskmatepeers have a positive effect on students’ academic achievements, measured by average test scores, Chinese test scores and Math test scores.
Other hypotheses were tested in the original paper but are beyond the scope of this replication.
- Source: Wu et al. (2023), randomized peer intervention experiment
- Location: Longhui County, Hunan Province, China
- Population: Primary school students (grades 3–5)
- Period: September 2015 – February 2016
- Sample size: 1,802 students across 36 classes
- Treatment variables:
treat1(MS),treat2(MSR)
- Outcome variables:
- Average test score (
ave) - Chinese score (
tchn) - Mathematics score (
tmath)
- Average test score (
- Controls:
Gender, age, height, health, hukou status, nationality, parental education, household assets, siblings - Fixed effects:
Grade and class
-
Descriptive Analysis
- Sample characteristics and balance checks
- Visualization of key demographic variables
-
Replication of Main Results
- Linear regression models replicating Table 3 in Wu et al. (2023)
- Separate analyses for lower-track and upper-track students
- Clustered standard errors at the class level
-
Robustness Check
- Permutation test to assess whether the estimated MSR effect could arise by chance
- Comparison of the real treatment effect with a distribution of randomized placebo effects
- R (≥ 4.2 recommended)
- RStudio (optional)
install.packages(c(
"stargazer",
"dplyr",
"lmtest",
"sandwich",
"car",
"ggplot2"
))- The MSR treatment has a positive and statistically significant effect on math scores of lower-track students (≈ +0.24 standard deviations).
- Mixed seating without rewards (MS) shows no significant effect.
- Upper-track students are largely unaffected by either intervention.
- The permutation test confirms that the MSR effect is unlikely to be driven by random chance.
- Overall, the replication results are consistent with Wu et al. (2023).
- This is a replication, not an original experimental design.
- Results rely on listing-based outcome measures provided by the original dataset.
- External validity is limited to similar educational settings.
The code in this repository is provided for academic and analytical purposes.
The dataset from Wu et al. (2023) cannot be published or redistributed due to licensing restrictions. All data usage follows the terms set by the original authors.
Celina Breuer