A growing shortage of technical and science graduates means that the question of determinants of differences in performance and choices in that field based, among other things, on social background characteristics (gender, socioeconomic status, ethnic origin, family composition) remains a topical one. In this study, a rational choice model is proposed in which background-specific comparative advantages in different subjects are in focus. Secondary analyses of a large-scale cohort study (VOCL) show that in addition to the most important explanatory variable, namely gender, the gap between mathematical and language skills makes a relevant contribution in predicting the choice of science subjects in secondary education. This gap can, in turn, be predicted on the basis of gender and family composition.