It is common to question the equal-interval assumptions of most academic scale scores. Even if interval assumptions hold, it is problematic to assume equal-interval distances with respect to benefits. For example, equivalent gains at the top and bottom of the distribution are unlikely to yield equivalent welfare returns. I develop a method to estimate the welfare returns to academic achievement directly, by making use of established methodologies in health economics. Using performance level descriptors and achievement data from the National Assessment of Educational Progress Long Term Trends, I estimate a random utility model to construct a welfare-adjusted equal interval scale. I then show that welfare returns to achievement are non-linear, convex and lead to different inferences regarding achievement gap trends.