Skip to main content

Welcome, the Hub connects all projects


Elementary School Data Issues: Implications for Research Using Value-Added Models


"Researchers conducting research using administrative data of ten presume that data from grades 4 and 5 are better than data from grades 6 to 8 for conducting research on teacher effectiveness that uses value-added models because (1) elementary school teachers teach all subjects to their students in individual, self-contained classrooms and (2) elementary school classrooms are more homogenous, with little academic tracking used when assigning students to teachers, unlike in middle school. We examined both assumptions. First, we used data on teacher-student links from DC Public Schools that have undergone a roster confirmation process whereby teachers verify which subjects and students a particular teacher taught. We compared the teacher-student links that resulted from this process to data that approximate the quality of teacher-student links in unconfirmed administrative data. Second, to examine the extent to which tracking of students by achievement segregates students at the middle school level compared to the upper elementary school level, we compared the variation in baseline student achievement at upper elementary and middle school grades within cl asses at the same school, between classes at the same school, and between schools. Results show that departmentalization of teaching instruction across math and reading/ELA is actually quite common in grades 4-5, at least in DCPS; in the unconfirmed administrative data, about one in six teachers of these subjects is linked to a subject that he or she does not teach. In addition, we found more within-school variation in pre-test scores in middle school grades but an offsetting amount of between-school variation in upper elementary grades. As an example of how using unconfirmed administrative data can affect results, we examined how calculations of the year-to-year and cross-subject stability of value-added estimates depended on the quality of the data used."