To Match or not to Match?: Improving Student-Program Fit in Dutch Higher Education

Abstract

In the Netherlands, every year a substantial proportion of first-year students in higher education drop out due to a wrong program choice. Matching has been introduced to decrease wrong choices. In this dissertation we studied various types of matching procedures using questionnaire and study progress data from four Dutch universities as well as interviews with (prospective) students. Based on a perspective of person-environment fit, we used three aspects that are important in determining whether students fit with the program of their choice: 1) interests, 2) believe in their abilities, and 3) feeling at home in the program. In this dissertation, we examined the effectiveness of different types of matching procedures as experienced by students, and in relation to final enrollment and first-year student success. The interviews revealed that students believe that matching procedures can contribute to their understanding of the program. In general, the more aspects of fit with the program students could test in matching activities, the more useful the activity was found to be. Moreover, quantitative analyses showed that the types of matching that were found to be more useful by students involved programs where fewer students followed through on their initial enrollment; an indication that more students changed their minds after participating in a matching procedure. Enrollment rates have decreased since the introduction of matching for programs with intensive procedures, but not for less intensive programs, i.e. programs that only offer follow-up activities of questionnaire completion for students identified as at-risk based. Therefore, the introduction of matching procedures in general, and specifically the intensity of these procedures is a likely explanation for lower enrollment rates. Furthermore, we analyzed information from the matching questionnaires for the prediction of study success. It was found that indicators of fit, measured prior to the start of the program, were predictive of grade point average and number of credits earned in the first year of study. Moreover, most fit indicators in our study differed in strength and/or direction between STEM and non-STEM programs. For both STEM and non-STEM students, high school grade point average is the strongest predictor of academic success in the first year of study. However, our findings provide evidence that non-cognitive indicators, such as conscientiousness and interest in the program, are stronger predictors of academic success in the first year for non-STEM students than for STEM students. Lastly, we found that text-mining analyses of motivational texts in the matching questionnaires predicted first-year dropout as well as a set of student characteristics. However, when the motivational texts and student characteristics were combined, the prediction of dropout did not improve. Thus, on the one hand, the use of text mining techniques for dropout prediction seems promising. On the other hand, the fact that combining the text and numerical data does not improve the prediction of dropout in this study could indicate that they measure the same underlying concepts. Combining the findings of all studies provides interesting leads for improving matching procedures, for example in terms of necessary elements in these procedures.