Standard achievement scales aggregate test questions without considering their relationship to economic outcomes. This paper uses question-level data to improve the measurement of achievement in two ways. First, the paper constructs alternative achievement scales by relating individual questions directly to school completion and labor market outcomes. Second, the paper leverages the question data to construct multiple such scales in order to correct for biases stemming from measurement error. These new achievement scales rank students differently than standard scales and typically yield achievement gaps by race, gender, and household income that are larger by 0.1 to 0.5 standard deviations. Differential performance on test questions can fully explain black-white differences in both wages and lifetime earnings and can explain roughly half of the difference in these outcomes between youth from high- versus low-income households. By contrast, test questions do not explain gender differences in labor market outcomes.