Biases in Student Evaluations of Teaching: An American Case Study

This work contributes to the literature raising concerns with the use of SET (student teaching evaluation) scores to evaluate teaching effectiveness and to motivate or demotivate faculty tenure and promotion decisions. It shows that the non-deterministic and qualitative nature of the SETs controverts their analysis and interpretation. The evidence of the strong selection of the (un)happiest students into survey participation since the recent switch to the online format with voluntary participation further demonstrates the subjective nature and invalidity of the SETs. The paper also provides empirical evidence of the unidimensionality of the SET answers: various SET items convey uniform content (satisfaction with students' in-class experience) regardless of the questions' specificity. Further, it reinforces empirical evidence that the SET usage introduces multiple biases related to professor, course, and class characteristics and facilitates grade inflation. These biases unbalance the SET scores differently at different tiers of the scores distribution. The results suggest that the biases based on gender and penalty to teaching very weak students are particularly strong. Use of recent and large American data** raises the validity and relevance of the findings relating to gender biases induced by the SETs and mainly reported by researchers from European institutions providing large data at the university level.