The evaluation of the short-term and long-term of disruptive peer effects in the primary education stage has become a hot topic in labor economics. Based on the latest Chinese Education Panel Survey（CEPS）data in 2014, this paper empirically examines the impact of disruptive peers on the student test scores.　　When studying the peer effects, there are two key problems to solve, reflection and self-selection. The so-called " reflection” refers to the problem of mutual influences between peers. It is essentially a reverse causality problem that is often encountered in the setting of econometric models, because it is difficult to tell whether the disruptive students have an impact on other students or some students have a bad influence on the surrounding students, which leads to a biased and inconsistent estimator if researchers use the surrounding students’ performance to explain other students’. Self-selection mainly refers to the fact that families don’t choose schools and classes completely randomly. In order to overcome the estimation bias caused by the problem of reflection and self-selection, we define the class environment as the proportion of disruptive students in the class rather than the peer achievement used in the traditional literature, and we also control the cohort features. 　　After controlling the individual and family factors, we find that students exposed to a disruptive environment have a significant lower education performance. This effect is more significant for the third grade students in the junior high school. One percent increase of the disruptive peers will result in about 0.2 decreasing in the average standardized test score. The result also shows that the acadamic performance of male students, students with poor family conditions and boarding students is more sensitive to class peers. The heterogeneity analysis of school types shows that peer effects are more obvious in schools with poor education and in township/rural schools. Furthermore, quantile regression results show that the negative effect of disruptive peers has a greater impact on poorly performing students. 　　This paper is helpful to enrich the academic cognition of educational production function and provide a useful reference for future education policy. The underlying implication of our study is that if there are reasonable and feasible policies that can change students’ class environment or a more reasonable class division plan, it will effectively improve the educational achievements of students. In addition, the attention to students’ family environment should be strengthened, especially those students who perform poorly in school, and strengthening communication and psychological counseling will bring about significant spillovers.