The present study ascertains the relationship between socioeconomic status (SES) and students’ science self-efficacy using data involving 509,182 15-year-old students and 17,678 school principals in 69 countries/regions who participated in the Programme for International Student Assessment (PISA) 2015. Hierarchical linear modelling results show that, after controlling for science teachers’ instructional practices (science class disciplinary climate, inquiry-based instruction, teachers’ support, direct instruction, provision of feedback, instructional adaptation), school science resources and various student variables (gender, grade levels, type of school programme), SES was related to students’ science self-efficacy in the majority of countries/regions (62–68 countries/regions, depending on the SES indicators used). Specifically, SES was related to students’ science self-efficacy in a larger number of countries/regions when it was measured using home cultural resources, home educational resources or a composite indicator (economic, social and cultural status) than when it was measured using parental education levels or occupational status. In contrast, students’ science self-efficacy was unrelated to the science teachers’ instructional practices examined (except inquiry-based instruction) in most of the countries/regions. These results expand our understanding of students’ science self-efficacy, as a type of learning motivation, from being a largely psychological attribute to one that is also influenced by social origins such as family SES. They imply that SES may have a larger influence on student achievement than we may have assumed if we include the indirect influence of SES on student achievement via students’ self-efficacy.