It Takes Two: Expectancy-Value Constructs and Vocational Interests Jointly Predict STEM Major Choices

The supply of STEM graduates is an issue in many countries—raising the question of why more students do not choose STEM careers. We aimed to address this question by drawing on two of the most prominent theories on career choices: Eccles et al.'s (1983) expectancy-value theory (EVT) and Holland's (1997) theory of vocational interests. In a large longitudinal data set, we used EVT constructs and vocational interests assessed at the end of high school (T1) to explain differences in math achievement (a critical filter for later STEM choices) at the end of high school (school sample: N = 4,984) and to predict actual STEM major choices in college 2 years later (T2; college sample: N = 2,145). To investigate their distinct and relative contributions, we conducted a set of hierarchical regression analyses (linear and logistic) in which we considered EVT constructs and vocational interests separately and simultaneously. Whereas both groups of constructs were significant predictors of both outcomes, vocational interests were better predictors of STEM major choice (Pseudo-R² = .26) than EVT constructs (Pseudo-R² = .19), and EVT constructs were better predictors of math achievement (R² = .47) than vocational interests (R² = .30) when investigated separately. In addition, the combined analyses revealed that the best predictions of math achievement (R² = .48) and STEM major choice (Pseudo-R² = .28) were achieved when EVT constructs and vocational interests were used jointly. Results suggest that EVT constructs and vocational interests contribute differently to STEM major choices, a finding that may have important implications for research and practice, including the choice of which constructs to target in interventions.