The question of precollege physics access and performance has been a persistent concern when considering the goal of diversifying participation in post-secondary STEM study and careers. This observational study examined school-level academic and demographic predictors of high school physics enrollment and performance in the USA. Due to the sequential and hierarchical nature of high school science and mathematics coursework, related predictors were included in multivariable regression and structural equation models to understand whether selectivity constrains the STEM pipeline. Descriptive and inferential analyses of state-level data were conducted (N = 663 high schools enrolling 559,044 students) to define the scope of physics enrollment and performance, and the predictive value of ethnicity, socioeconomic status, and science and mathematics course enrollments and performance. Physics was taken by fewer students when compared to biology, earth science, and chemistry. Correlations, multiple regression, and structural equation modeling indicated that although socioeconomic status was the main predictor of student "enrollment" in physics, this was mediated by schoolwide enrollment in chemistry and algebra II as well as schoolwide performance in chemistry and geometry. School-level "performance" in physics was negatively predicted by the percentage of students traditionally underrepresented in STEM, yet the predictive value was mediated by chemistry and algebra II performance. Results suggest that the prevalent science and mathematics sequence correlates to diminished physics enrollment, particularly with regard to students traditionally marginalized in STEM. US school leaders and policy makers should implement more proactive interventions earlier in the STEM pipeline to promote diverse physics participation and more equitable performance outcomes.