Identifying Gender Peer Effects In Students' Academic And Non-cognitive Outcomes By Higher-order Spatial Autoregressive Models

This article uses the latest China educational survey CEPS,which is a large-scale,national survey to study the gender peer effects in students' academic and non-cognitive outcomes among middle-school students.To avoid selection bias,we choose the random samples.Compared to other literature,th is article adopts higher-order spatial autoregressive models to identify endogenous peer effect and exogenous peer effect.Moreover,we control correlated effect through class and school characteristics.We also investigate the mechanisms behind the endogenous peer effect by using four spatial weighting matrixes,dividing the effect into male to male,female to male,male to female,female to female,respectively.Our estimation results show that there is significantly positive peer effect in students' academic and non-cognitive outcomes among middle-school students and peer effect is stronger intra-gender than across gender.Furthermore,compared with boys,girls are more influenced by intra-gender peer effect in both academic outcomes and non-cognitive outcomes.That is to say,girls are more responsive to peer effect.However,girls are not influential across gender.Then,we explore peer effect further by splitting the whole sample into different grades-seventh and ninth grade,different class sizes-especially the small size classes,and different areas-rural and urban areas,the results are similar.Finally,we separate the whole sample by academic rank and find that girls,especially above average and top girls are influential in academic outcomes but not in non-cognitive outcomes.Our findings show the peer influence in academic and non-cognitive outcomes among middle-school students mainly comes from peers of the same gender,and mixed school may help increase academic outcomes among middle-school students,but not in non-cognitive outcomes.