Discrimination in Grading? Experimental Evidence from Primary School

Migrant children are lagging behind in terms of educational performance in many industrialised countries. The relatively low educational level of their parents as well as language difficulties are well-known reasons behind the test score gap between migrant and non-migrant children. However, little is known about the role of teacher expectations in explaining this gap. According to the psychological literature, teacher expectations may affect pupil performance in several ways. For instance, there is evidence that teacher expectations may lead to (unconscious) changes in teacher behaviour that affect the actual performance of the pupils. But besides inducing changes in teacher behavior and pupil performance, expectations may also influence the way in which teachers grade pupils' work. Grades are the main signal of ability and performance that teachers give to their pupils and they can have long term consequences for student achievement. However, grading is a subjective process. Even if teachers generally use point schemes, many subjective impressions may play a role in estimating the quality of student work. In this paper, we randomly assign typical German or Turkish names to identical sets of essays to test the effect of teacher expectations regarding pupil background on grades. We find that essays bearing Turkish names receive significantly worse grades. Moreover, teachers recommend the highest track of secondary school with a 10% lower probability when an essay bears a Turkish name. The effects we measure are relatively small and cannot explain the full gap in the attendance of the highest secondary school track between migrant and German pupils. However, considering the numerous disadvantages faced by migrant pupils, removing the additional penalty resulting from lower teacher expectations would be a welcome step forward. It is therefore good news that the observed grading and recommendation bias originates from a small group of teachers only. Hence most teachers do not grade or recommend different tracks based on ostensible pupil origin. This implies that lower expectations do not necessarily affect teachers' judgment of a pupil's potential and there is scope for getting rid of these biases for the remaining group. For instance, increased awareness about the importance of teacher expectations through teacher training, could contribute to reducing the grading and expectation bias