Effects of grading leniency and low workload on students' evaluations of teaching: Popular myth, bias, validity, or innocent bystanders?

Herbert W Marsh, Lawrence A Roche
2000-03-01
Abstract:Two studies debunk popular myths that student evaluations of teaching (SETs) are substantially biased by low workload and grading leniency. A workload bias is untenable because the workload–SET relation is positive. The small grade–SET relation (. 20 for overall ratings) has many well-supported explanations that do not involve bias. Some SET factors (eg, Organization, Enthusiasm) are unrelated to grades, and the highest relation is with Learning (. 30), implying valid teaching effects rather than bias. Structural equation models confirmed that perceived learning and prior characteristics (course level, prior subject interest) account for much of the grade–SET relation. The relation is also nonlinear, so that high grades (sometimes misused as a leniency measure) are unrelated to SETs. Contrary to dire predictions based on bias claims, Workload, expected grades, and their relations with SETs were stable over 12 …
What problem does this paper attempt to address?