High math and science achievement, and expressed and measured interest in STEM-related fields, are positively related to STEM major choice and degree completion in college (Radunzel, Mattern, & Westrick, 2016). Students’ perceived relative strengths, particularly their math and verbal strengths, influence students’ decisions on how they invest their time and effort, and gender differences in relative academic strengths and interests may contribute to gender gaps in math-intensive STEM fields (Coyle, Purcell, Snyder, & Richmond, 2014; Davison, Jew, & Davenport, 2014; Lubinski & Benbow, 2007). This study examined the relationships between ACT test scores, high school GPA (HSGPA) and courses, relative academic strengths or academic tilt, vocational interests tilt, intended academic major, sureness of intended major, and gender to predict STEM degree completion in math-intensive fields within six years of enrollment.

Data included 62,516 ACT-tested college students enrolled as first-time students at 47 four-year institutions. A minimum of 30 declared STEM majors at an institution was required for inclusion. Precollege predictors included standardized ACT composite score, HSGPA, ACT and HSGPA tilt measures (math &science – English & reading/social studies), high school coursework, interests (Data-Ideas tilt, People-Things tilt), intended major, and sureness of intended major. Hierarchical logistic regression models with random slopes and random intercepts were used to estimate students’ chances of earning a math-intensive STEM degree. The accuracy rate (AR) – the estimated proportion of students correctly identified completing or not completing a STEM degree – and the logistic R – defined as the standard deviation of the estimated logit function (Allen & Le, 2008) – were calculated to assess model fit. Models were run to predict degree completion in 4, 5, and 6 years.

Results varied across years 4, 5, and 6, but ACT composite score and HSGPA were the top two predictors of completing a STEM degree in each year. In addition, ACT tilt (higher math & science), HSGPA tilt (higher math & science), tilt toward Ideas (vs. Data), tilt toward Things (vs. People), completion of high school calculus, intention to declare a math-intensive STEM major, and the interaction between ACT-composite score and gender were statistically significant (p<.05) in each year. In years 5 and 6, gender, completion of high school physics, sureness of intended major, and the interaction between Data-Ideas tilt and gender were statistically significant as well. The median AR was highest in year 4 (87%) and lowest in year 6 (77%), as was the median logistic R (1.28, year 4; 1.06, year 6).

The final model in this study provides a high degree of accuracy in predicting completion of a math-intensive STEM degree. Separate analyses found that intended major, sureness of intended major, and the interaction between these variables were the strongest predictors of which students declare a STEM major. However, the results of the current study indicate that cognitive measures – ACT scores and HSGPA – are the strongest predictors of completing a math-intensive STEM degree. Moreover, academic tilt toward math and science (vs. English and reading/social studies) contribute to the prediction of degree completion, as does an interest tilt toward things instead of people and tilt toward ideas instead of data. The effect of academic tilt and interest tilt provide support for investment theories and may help explain gender gaps in math-intensive STEM fields.

Education , Measurement and Psychometrics , Social and Life impacts