Computational thinking as a latent predictor of academic success in teacher education

Computational Thinking (CT) has emerged as a fundamental 21st-century cognitive skill, yet its psychometric structure and relationship with academic success within teacher education remain underexplored. Accordingly, this study examined the predictive power of CT on preservice teachers' grade point average (GPA) using three analytical approaches: an individual-path model, a simultaneous multiple-predictor model, and an overarching latent construct model. The sample comprised 368 Indonesian preservice science teachers who completed a validated 30-item computer-based CT assessment measuring decomposition, pattern abstraction, and algorithmic thinking. Data were analyzed using structural equation modeling. Results showed that in individual-path models, each CT component was significantly associated with GPA (all p s < .001), explaining a modest proportion of variance (6.8%–9.1%). In the simultaneous-path model, the three components jointly accounted for 13.1% of GPA variance ( p < .001), though individual path coefficients were attenuated, with algorithmic thinking losing statistical significance ( β = 0.114, p = .056). Subsequently, the latent construct model accounted for the highest proportion of variance in GPA ( R 2 = 15.1%) and demonstrated a statistically significant association ( β = 0.388, p < .001), showing higher explanatory power than individual-component models and a slight enhancement over the simultaneous model. These findings suggest conceptualizing CT as an integrated latent construct that captures the shared variance among subskills in relation to academic performance. Overall, the results support integrating CT into teacher education curricula, not only for pedagogical preparation in teaching CT but also to provide a rationale for further investigating how CT relates to academic performance.