HRD Candidate: Sembukutti Arachchige Bhagya Gunawardena

Milestone: Mid-Candidature Review

Title: Peer Effects in Education - Ability, Incentives, Information, and Gender

When: 2-4pm, Tuesday 4 May 2021

Where: Zoom -


The focus of this thesis is to identify how peers influence the learning outcomes of students and how can we use these findings to improve educational performance. This thesis consists of three chapters that primarily focus on the direct effects of peers on learning outcomes using experimental and quasi-experimental methods.

The first chapter investigates the effects of peers' ability in cooperative learning environments when the incentives for performance depend on group performance compared to individual performance using a lab experiment. The findings support mixed ability grouping where mixing high ability subjects with medium and low ability subjects under group incentives being Pareto optimal for direct learning from peers. That is, high ability subjects are better off by not being with similar ability peers regardless of the incentive mechanism. However, medium and low ability subjects are better off by being with high ability subjects under group incentive treatment, whereas they are worse off by being with high ability subjects under individual incentive treatment. These findings support the idea that high ability subjects learn by teaching but discouraged by being with similar ability peers.

The second chapter focuses on the effects of providing peers' investment information and performance information on students' monetary and effort investments in acquiring knowledge and improving performance while controlling for the pressure generated through being observed by peers. For this purpose, this chapter also utilizes a lab experiment as it allows greater controllability and provides clean measurements.

The third chapter analyses the effects of classroom peers' gender and ability composition on students' performance, major field choices, and university outcomes using a rich Greek set representing approximately 10\% of public schools in Greece between 2003 - 2011. This data, as good as random assignment of students into schools and classrooms, allows us to identify the effects of gender-ability variations across classes within the same school cohorts. Therefore, we can control for any confounding factors at the school-cohort level.