In order to address the issue of the reliability and possible bias of students' evaluation of the performance of college teachers, an unobtrusive indicator of teacher performance was investigated. The indicator was the degree to which teachers develop a following among students, measured by how often students return to teachers for additional courses. The model uses 5 years of student course and grade data at a large, state university, which entailed the records of 50,000 students, the one-half million courses they took, and 2,275 teachers. The findings suggest that the measure of students taking additional courses from faculty members can be a valid indicator of teaching effectiveness, and can be used along with students' ratings. The ability to use this type of administrative data to augment existing systems of evaluating teacher effectiveness is advocated, since diversity of measurement not only reinforces reliability, but can enhance the effect of valid and unbiased teacher performance indicators. It is concluded that such data, when used in conjunction with other evaluative methods, can be a valuable resource. (SW)