Point scale framing and student performance: Evidence from a field experiment

Does the numerical scale used to score an exam affect how hard students try? We test this question with a preregistered natural field experiment in which 1235 university students in India took identical tests framed as scored out of either 20 or 100 points. Exploiting both a high-stakes graded quiz and a zero-stakes additional test, we find no statistically or economically meaningful effects of the point-scale manipulation on test scores or on four effort measures, regardless of stakes, gender, or prior ability. These consistently null results, robust across settings, outcomes, and subgroups, identify a boundary condition for the numerosity heuristic: when scoring units are transparently arbitrary and evaluation ultimately rests on percentages, numerical rescaling does not shift behavior. The findings support the behavioral neutrality of a widely used and easily modified feature of assessment design.