Abstract: In recent decades, due to the effectiveness of the “OneChild” policy in the cities, and the social campaign to improve the enrollment and reduce the dropout rate of female students especially in the countryside, the gender gap in educational attainment up to the college level has narrowed. While female students are now as likely as male students to finish high school and enter college, they are also leading the male students in academic performance in schools. Different studies from various countries point to the same result across societies including China.Legewie and DiPrete (2012) found that male students in schools with low average SEI (social economic index) families were more likely to be negatively affected by peers’ antisocial attitudes than female students, and regard goodatstudy as less masculine, which resulted in male students’ poor academic performances. Female students in similarly disadvantaged schools, however, were less prone to such negative peer pressures.Using data collected from seven middle schools in Guangzhou and the same school fixedeffect model employed by Legewie and DiPrete, the paper seeks to understand how school environment affects gender differences in academic performances in different schools in China and how it compares to Legewie and DiPrete’s study of the German case. The pattern of gender differences in academic performances, measured by standardized math tests, is similar to Legewie and DiPrete’s findings that the gaps between average female test scores and average male test scores are larger in socially disadvantaged schools than elite schools. While students from families of different socialeconomic statuses are sorted into different schools of the twotiered (public vs. private schools) and highly differentiated (selective elite schools vs. nonelite schools) school system in Chinese cities, the class assignment within schools is supposed to be random. As a matter of fact, most of the classes within the seven schools studies were based on random assignment, only a few classes were based on preferential selections either by test scores or by higher fees. Classes within schools, therefore, can be regarded as naturally formed experimental groups by random assignment. Since school class is the most immediate and salient micro environment for the students, especially in Chinese schools because students mostly stay with the same class throughout the school years, how variations among classes within the same school affect students’ performances provides a way to assess specifically how micro environment in school affects students’ attitude, behavior, and performance.The results show that the gender gap in math test scores is larger in classes whose average father’s educational level is lower. Average father’s educational level of the class is also related to the differences in male and female students’ gender attitudes and learning behavior which are also shown to have significant effects on academic achievement.

Key words: gender gap, social status, school environment, peer effect