Longitudinal research in higher education faces several challenges. Appropriate methods of analyzing competence growth of students are needed to deal with those challenges and thereby obtain valid results. In this article, a pretest-posttest-posttest multivariate multilevel IRT model for repeated measures is introduced which is designed to address educational research questions according to a German research project. In this model, dependencies between repeated observations of the same students are considered not, as usual, by clustering observations within participants but rather by clustering observations within semesters. Estimation of the model is conducted within a Bayesian framework. Results indicate that competences grew over time. Gender, intelligence, motivation, and prior education could explain differences in the level of competence among business and economics students.