This working paper compares teacher/student links that have undergone a roster confirmation process—whereby teachers verify the subjects and students they taught—to teacher/student links from unconfirmed administrative data. Due to the departmentalization of instruction in math and reading/English language arts in grades 4 and 5, about one in six teachers in these grades and subjects is linked in the unconfirmed data to a subject that he or she does not teach. The authors discuss the circumstances in which using unconfirmed teacher/student links in value-added models most affect research.