Predictors of Students' Evaluations of Teaching:A Multilevel Analysis

Abstract

Despite the widespread use of student evaluations of teaching the validity of these ratings is still called into question. In particular it is suspected that student ratings are affected by variables that are not related to the quality of teaching and thus are biased. In a dataset with 2 898 courses the influences of 10 potentially biasing variables are analyzed using multilevel modelling. The variables are located on the student level (student gender, prior interest, surrounding conditions, absenteeism), the course level (class size, dropout, mandatory participation) and the teacher level (faculty status, gender, teaching experience). Nearly all of the variables on the first two levels and a cross-level interaction between student and teacher gender were statistically significant predictors of the student evaluations. However, with the exception of prior interest and the surrounding conditions these relationships are only weak. It is also often difficult to decide whether these variables are biasing factors.