The positive impact of Math Learning Support Centers (MLSCs) on student performance is well-documented in existing literature. However, we argue in this study that many of these evaluations could be improved methodologically. Using a more rigorous approach, we investigated the effect of an MLSC on students' performance and affective variables in an economic study program. Our data, collected over six years, was analyzed using Propensity Score Matching and Difference-in-Differences methods, to control for observable and time-invariant unobservable characteristics. Contrary to a majority of the existing literature, we found no direct impact of students' participation in the MLSC on students' math performance. However, we identified positive effects on students' test anxiety and other affective variables. Moreover, we found that students' engagement in the learning process during the semester (e.g. by attending classes and solving exercise sheets) was a key driver of skill growth. These findings underscore the importance of more rigorous methodological approaches in future evaluations of MLSCs and other similar support systems.