The gap in supply (i.e., shortage) and demand of the STEM workforce have prompted extensive research on identifying factors that predict STEM outcomes and retention of students. Few studies, however, have examined the relationships between STEM outcomes and predictors in an integrated model, taking into account measurement errors in the predictors. Drawing upon the Expectancy-Value Model of Achievement Related Performance and Choice, I conducted a structural equation modeling (SEM) analysis to examine the relationships between academic support, academic engagement, mathematics readiness, student hours worked, and first-year STEM students’ academic success and retention. The SEM allowed me to investigate the relationships between predictors and outcomes simultaneously while accounting for the measurement errors. The sample consisted of 798 first-year STEM majors who took the National Survey of Student Engagement during the 2016, 2018, and 2020 academic years in a large urban university. Results indicated that academic support was a statistically significant predictor of first-year STEM students’ academic success and retention. Additionally, mathematics readiness was found to be a statistically significant predictor of first-year retention. Lastly, results suggested that female students on average were more likely than their male counterparts to engage in academic support and academic engagement activities even though females worked more hours than males. The results have implications for policies and practices aimed at improving STEM retention. Areas of further research are also identified. Keywords: expectancy-value model, structural equation modeling, STEM retention, academic success predictors, mathematics readiness