Abstract
Spurred by concerns about an inadequately sized science, technology, engineering, and mathematics (STEM) workforce, there has been a growing interest in out-of-school time (OST) science activities as a means to foster STEM career interest. This study examines the association between OST science activities and STEM career interest in university through a logistic regression model and the calculation of prototypical odds ratios. The analysis addresses two main research questions: What is the correlation among different forms of OST activities? And, controlling for student demographic and background variables, what specific forms of OST activities are associated with STEM career interest in university? The study uses data from the ‘Persistence Research in Science and Engineering’ survey (n = 6882), which employs a nationally representative sample of university students enrolled in introductory English courses. Results indicate that students’ participation in OST activities, as well as their middle school interest in science and mathematics and their gender, plays a significant role in university career interest in STEM. Conclusions suggest that making OST clubs and competitions and the inclusion of non-fiction and science fiction within English Language Arts programmes may be beneficial to the development of students in STEM careers. Limitations include the paucity of research examining which students participate in these activities and what specific features or characteristics benefit them.
Notes
These two forms of interest have been further subdivided, and interest is said to develop through four phases including: preliminary situational interest, sustained situational interest, developing individual interest, and finally a maintained individual interest (Hidi & Renninger, Citation2006).
Middle and high school career aspirations were not specifically included as variables in the final model because of their overlap in variance with the Middle School Interest variable already included in the model. The additional explained variance would not offset the loss of parsimony had these two additional variables been included in the final model.