This study focuses on trends in STEM performance and inclusiveness. We examine performance trajectories in STEM subjects, the predictive role of social support and psychological well-being of students, and variations across student gender and school locality (rural vs urban). We used three waves of data from 135 junior high school students in Ghana. Multilevel growth curve modeling was used to assess the trajectories and the socio-environmental predictors of STEM performance, and posthoc power calculation was used to confirm the adequacy of the sample size. Results show that overall, students' STEM performance improves over time. Minimal gender differences exist but depend on the subject area and evolve with time. We observed a nuanced "urban advantage," with rural students starting well but declining over time. Among various indicators of social support and psychological well-being, teacher support was the strongest positive predictor of STEM performance. The study highlights the need to focus on the structural and cultural impediments to STEM education at the lower levels of education in order not to risk excluding marginalized groups early in the education system. Further, STEM interventions may do well to incorporate long-term measures to sustain girls' interest, motivation, and efforts in STEM.