The Longitudinal Interplay of Mathematics Self-concept, Interest, Cost, and Performance – A Random Intercept Cross–Lagged Panel Model

The Situated expectancy-value theory suggests that students' expectancies for success, positive task values, and costs are significant predictors of subsequent performance. Prior research has extensively studied these relations; however, most work is grounded on analyses combining within- and between-student variances. In our study, we utilized a random intercept cross-lagged panel model (RI-CLPM), which separates these variances, thus providing a more nuanced view of the relations. Specifically, our study explores the longitudinal relations between math motivational beliefs (self-concept, interest, emotional cost) and math performance across Grades 4 to 6 (10-12-year-olds, N = 584). The dataset consists of six timepoints, measuring motivation and performance twice per school year. Between-person results showed that the positive motivational beliefs (self-concept and interest) were positively related to each other and math performance, while emotional cost was negatively related to the positive motivational beliefs and math performance. Within-person results revealed higher-than-usual interest to predict subsequent higher-than-usual self-concept and lower-than-usual emotional cost. Higher-than-usual self-concept predicted lower-than-usual emotional cost at the next time point, and vice versa. No cross-lagged relations were found between motivational beliefs and math performance. However, within-student correlations indicated a concurrent relation between these constructs, rather than a longitudinal one. In conclusion, math interest seems as a promising candidate to target in interventions among 10-12-year-olds, as it was linked to subsequent positive (self-concept) and negative (emotional cost) motivation. Emotional cost seems to harm self-concept already in 4th grade, highlighting the importance of also preventing negative emotions related to math.