An application of value added models: School and teacher evaluation in Chinese middle schools

Researchers using VAM (Value-Added Models) methods have found that student achievement is substantially impacted by both schools and teachers. However, most of this work has been conducted within the context of the United States. Information is lacking about the educational system in China. ? This first study examines methodological concerns regarding the inclusion of proper controls in the estimation of school value-added models. Using a longitudinal, multi-cohort panel data set from a district in China, I test the effect of controlling for school assignment policies, school external factors, and school-level composite variables that are directly affected by the assignment process. I found that controlling for school assignment policies changed the school effect estimation by between 15-23%. Controlling for external factors changed the school effect estimates by 7%, and the school ranking by 9.6%. Although controlling for school-level composite variables also adjust for school effect estimates by 14%, these controls seem to be problematic due to their correlation with the effect to be estimated. This paper also examines such an issue and found school external factors and assignment policy variables explained over 80% of variance in school level achievement composite, the percentage of top students in schools. Thus, the school level achievements composite variable is not a proper control for school value-added models. ? The second study uses the same longitudinal data and evaluates individual school and teacher performance using VAMs in the context of China. Specifically, I estimate the magnitude of school and teacher effect in math, Chinese Language Arts, and English for middle school grades (seventh to ninth grades). I found the effect size for schools is around 0.12 to 0.17 standard deviation (SD) in math, or 0.15 to 0.23 SD in Chinese Language Arts (CLA), or 0.13 to 0.18 SD in English as a foreign language (English). The effect size for math teachers ranges from 0.17 to 0.21, and the effect size for CLA teachers ranges from 0.19 to 0.25, and 0.15 to 0.19 for ELA teachers.