Relational Contracts and Equity-Based Reward in Teamwork : Theory and Evidence from Higher Education

This paper develops a theory of equity-based reward in team production and tests it in higher education. We study n agents who collaborate over the repeated stages of a single project. Under conventional fixed-share rules - whether each member receives a slice of a divided output or, as in education, the whole of a non-rival output - effort falls short of the efficient level, because no member internalises the value that their effort creates for the rest of the team. We characterise an equity sharing rule that ties each member's reward to a logarithmic function of their agreed relative contribution. The rule raises effort above the fixed-share level and, in the limit where the team technology is close to linear, restores the efficient level exactly; under more sharply diminishing returns it over-rewards relative contribution rather than under-rewarding it. Because the rule is enforced not by an outside monitor but by the members themselves through repeated peer evaluation, we show that it is self-enforcing as a relational contract whenever participants are sufficiently patient, and that the patience required is most easily met precisely in the near-linear case relevant to education. We operationalise the mechanism as a design for MSc group dissertations, in which students negotiate contribution shares across three project phases monitored by anonymous weekly peer evaluation, and individual marks are derived from the group mark through an equity-weight formula that is the exact empirical counterpart of the theoretical rule. Using panel data on 926 students at a Russell Group university and combining propensity-score matching with difference-in-differences and the doubly-robust Callaway and Sant'Anna (2021) estimator, we find that adopting the equity-share mechanism raises overall dissertation marks by 6.8 - 8.2 percentage points (7.7 under the preferred doubly-robust estimator) and the individually assessed component by 9.8 - 11.2 points (10.6 preferred) relative to group dissertations without it; the estimates are consistent across estimators and replicate across cohorts, though, because the policy is adopted in a single programme, we read these magnitudes as evidence consistent with the theory rather than as precisely identified causal effects. A companion comparison of group against individual dissertations points in the same direction but does not satisfy the parallel-trends requirement, and we treat it as descriptive. The results suggest that incentive compatible, peer-enforced grading can raise cooperative effort and attainment in team-based assessment.