Investigation of the Role of Pre- and Post-admission Variables in Undergraduate Institutional Persistence, using a Markov Student Flow Model
No Thumbnail Available
Files
Date
2006-06-22
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
This study used selected student record data to investigate the effect of students? characteristics prior to university admission (pre-admission variables), and academic actions and educational achievement indicators (post-admission variables) on retention in higher education. The analysis followed first-year undergraduate students at a large Midwestern university through four academic levels (freshman-senior).
A Markov student-flow model was employed to estimate the probabilities of stopping out, staying at the same academic level, or advancing to a higher academic level up to graduation. Logistic regression was used to calculate fourteen transition probabilities of specific flow-model events given a profile of independent variable scores. Based on the yearly transitions, predicted probabilities of graduating after 4, 5 and 6 years were also computed.
The key results are (a) The Markov student flow model and its use as a predictive tool, which allow calculation of a persistence risk value using institutional data. (b) The finding that many variables vary in predicting persistence depending on the academic level, which corroborates the need to organize the model by academic levels and indicates that it is incorrect to conclude that variables that affect persistence at one academic level do so at all levels.
Relevant to the specific institution studied are the findings that variables such as Age at Entrance, and Pell Grant Indicator consistently predict lower probabilities of progressing towards graduation for all academic levels, holding other variables in the model constant. Cumulative GPA and Not Changing Majors also predict higher transition probabilities, with the strongest effect at the sophomore level. Target Minority, ACT score and High School Percentile predict higher probabilities of persisting at the Freshman level, but the effect becomes negative at the Senior level.
If tested and implemented in an institution, the proposed simulation tool would allow decision-makers to examine potential effects of policies by altering variable profiles and analyzing the predicted changes in the institutional persistence of students. The probabilities obtained can be interpreted as an empirical persistence risk value.
Description
Keywords
Persistence, Logistic Regression, Educational Policy, Markov student-flow model, Higher Education, Doctoral Dissertation
Citation
Degree
PhD
Discipline
Psychology
