Signal Study models how test-retake and scoring rules create wealth-based inequities in admissions

Summary

A paper develops a strategic framework analyzing how students allocate effort across multiple standardized-test retake attempts under different institutional scoring policies. It shows that "Superscoring", combining the best section scores across multiple sittings, introduces systematic score inflation through order-statistic selection over random noise, which degrades signal accuracy and disproportionately rewards applicants who can afford to retake tests repeatedly, thereby amplifying wealth-based disparities. By contrast, "Single-Sitting" scoring, which keeps only the best overall score from one sitting rather than combining section scores, preserves signal fidelity but excludes high-ability students who lack the resources to prepare for all subjects simultaneously. The authors find that neither rule uniformly dominates the other, forcing a structural trade-off between statistical precision and fair outcomes. To address this, they propose three algorithmic interventions that either change how scores from multiple attempts are combined or apply a post-hoc correction to observed scores. The proposed interventions are compared against standard scoring rules using simulations calibrated to 2025 College Board data.