General education astronomy courses are an important vehicle for helping students develop “real world” quantitative skills, but performance on quantitative assessments is not a simple measure of skill alone. Affective variables such as numerical self-efficacy and math anxiety level have a strong influence on students’ performance. We will present results of a mediation analysis of Quantitative Reasoning for College Science (QuaRCS) study data; this work was designed to inform the interrelationships between student performance, demographics, and affective measures. Our results indicate that accounting for demographic differences in affect can explain up to 50% of observed gender, race/ethnicity, and socioeconomic status achievement gaps. At the same time, the specific contributing factors and effect sizes vary greatly across demographic groups. We will discuss what we think this analysis tells us about possible approaches for closing QR achievement gaps in astronomy classrooms, which is critically important to our field because quantitative achievement gaps are at least partially responsible for underrepresentation of certain demographic groups in STEM fields.