Test anxiety significantly impacts students’ academic performance and mental health, with complex interactions influenced by behavioral and demographic factors. This study examines the relationship between metacognitive self-regulation (MSR) behaviors and test anxiety across demographic groups, explores trade-off between predictive accuracy and fairness in test anxiety prediction models, and investigates how intersecting demographic attributes shape biases. The findings show that specific MSR behaviors, such as classroom distraction and frequent adaptation of study methods, are strongly correlated with test anxiety, highlighting key areas for targeted interventions. Demographic disparities are evident, with females experiencing higher levels of test anxiety and White students reporting more classroom distractions. A trade-off between predictive accuracy and fairness is observed, with highly accurate models not always performing well in terms of fairness, emphasizing the need for balanced model selection. Additionally, the study challenges traditional additive assumptions about fairness, finding that the intersection of demographic attributes produces unexpected compounded effects, such as compounded advantages for Non-White Migrants and mixed outcomes for White Females. We offer insights for designing accurate and equitable predictive models for test anxiety.