Computational thinking has become an essential competency for prospective teachers to support technology-integrated learning in the digital era. However, empirical studies integrating project-based learning with Scratch programming that simultaneously examine cognitive and affective outcomes in teacher education contexts remain limited. This study aims to examine the effectiveness of Project-Based Learning integrated with Scratch programming in improving prospective teachers' computational thinking skills and self-efficacy. A mixed-methods approach employing a quasi-experimental design with experimental and control groups was implemented involving 64 students enrolled in a teacher education program. Quantitative data were collected through computational thinking tests and self-efficacy questionnaires. Qualitative data were obtained from semi-structured interviews, which were analyzed using NVivo to explain students' learning experiences. Quantitative data were analyzed using Multivariate Analysis of Variance, normalized gain, and effect size analysis. The results revealed a significant multivariate effect of Scratch-based project learning on computational thinking skills and self-efficacy. The experimental group demonstrated moderate learning gains and large effect sizes, indicating substantial improvement over conventional learning. Qualitative findings supported these results by showing increased engagement, collaboration, and confidence in computational problem-solving during project development. The study contributes theoretically by reinforcing the integration of constructivist learning principles with the development of computational thinking and the formation of self-efficacy. In practice, the findings suggest that Scratch-based project-based learning provides an effective instructional framework for preparing prospective teachers with the computational competencies required for 21st-century education.