Academic Leniency
Relaxing academic standards raises grades without raising achievement, and the students the policy is meant to help are the ones who disengage. Bowden, Rodriguez and Weingarten establish the causal chain with a statewide grading reform in North Carolina: GPA rose mechanically, test performance did not move, absences rose among low-ability students, and achievement gaps widened across high school. The work is archived here as EdWorkingPaper 23-836 (September 2023).
The policy and the design
In fall 2014 the North Carolina State Board of Education standardized public high school grading to a ten-point scale — a move the paper treats as a leniency increase, and one that gave students at the margin of passing a class a larger buffer. The research design combines that statewide grading-policy variation with school entry rules: administrative records for the universe of North Carolina public school students, with exact birth dates used to measure each student’s relative distance to the kindergarten entry cutoff, supply comparison groups around the stricter and more lenient grading regimes.
Theory
The paper’s model makes student effort endogenous to the grading policy. Designs that ignore either the spread of student ability or the size of the leniency change can widen achievement gaps even while lifting measured outcomes.
Results
Leniency mechanically increased GPA without increasing student achievement, and it increased school absences. The effects split by prior ability: the GPA gains came entirely from high-ability students, the attendance losses entirely from low-ability students. In the grade distribution, the share of A’s in core academic courses rose almost 20% in 2016 and stayed elevated through 2018 while the share of F’s fell about 20% in 2016 against the prior year; the middle letter grades moved much less. The cohort differences compounded across high school and widened long-run achievement gaps as measured by ACT scores.
The measurement-channel reading
The grade scale is a measurement channel, and loosening it moved the measured quantity (GPA) without moving the one it stands for (achievement) — while effort, which the channel exists to induce, fell at the margin. The pattern is the student-side instance of the Goodhart dynamic tracked in grader-orientation: when the metric is the operative target, changes to the metric’s difficulty get absorbed by behavior instead of passing through to the outcome. It is also direct evidence for the education-policy channel in the post-literacy-debate: standards changes move effort and gaps on their own, so achievement trends cannot be attributed to exposure effects such as screens without accounting for grading policy.
Open questions
- The design identifies a leniency increase. Whether re-tightening standards symmetrically restores effort, and for whom, is not answered here.
- The compounding mechanism — heterogeneous responses accumulating across grades — is observed on ACT outcomes; how much runs through course-taking and tracking rather than within-course effort is unresolved.
- The reform predates the post-2020 wave of leniency; whether the same ability-split response appears there is untested.
Related
- eval-awareness-and-grader-orientation — measurement-channel capture in AI evals; the same Goodhart structure on a different substrate
- post-literacy-debate — where the education-standards channel of achievement change is argued
- ai-mathematical-practice — Goodhart’s law in mathematical practice: community norms against optimizing the measure