STEM integration has become a popular concept not only in the context of education practices but also as a way of learning. The integration of the STEM domains is evident in students' learning experiences when engaging in STEM activities. However, there is a lack of instruments for evaluating students' levels of self-efficacy in these activities. Therefore, the aims of this study were to develop a survey for evaluating upper primary students' self-efficacy in STEM activities and to explore whether a student's gender, grade, and participation in STEM activities predict his or her self-efficacy in STEM activities. A total of 844 fourth- to sixth-grade primary students participated in this study. After pilot testing, exploratory factor analysis and confirmatory factor analysis were conducted, the survey was found to have a single-factor structure with high reliability (Cronbach's alpha = 0.90). Linear regression analysis showed that school and out-of-school participation in STEM activities significantly predicted the students' self-efficacy in STEM activities, while grade and gender did not. The survey developed in this study provides a reliable and valid way to measure students' self-efficacy in STEM activities. These findings also highlight the importance of encouraging students' participation in both formal and informal STEM activities.