Measuring systematic gaps in teacher judgement: A new approach

Subjective performance evaluations play a prevalent role across many parts of society, despite concerns regarding bias. Consistently measuring the extent of bias with such evaluations is inherently difficult. We propose a new approach to tackle this longstanding question in a setting where teachers were required to assign students both grades and rankings within each grade. By exploiting the discrete nature of these evaluations, we compare the concentration of student types of adjacent students either side of grade boundaries. Through an application of a local randomization approach, we establish systematic bias favoring higher income female students. We then establish that these grading decisions carry real consequences: students just above the grade threshold are significantly more likely to attend university and secure their first-choice degree.