Enrollment Projection Using Markov Chains: Detecting Leaky Pipes and the Bulge in the Boa

Abstract

While Markov chains are widely used in business and industry, they are used within higher education only sporadically. Furthermore, when used to predict enrollment progression, most of these models use student level as the classification variable. This study uses grouped earned student credit hours to track the movement of students from one academic term to the other to better identify where students enter or leave the institution. Results from this study indicate a high level of predictability from one year to the next. In addition, the use of the credit hour flow matrix can aid administrators in identifying trends and anomalies within the institution's enrollment management process.

Student Trajectories for Enrollment Forecasting, Management, and Planning

“…Although the typical transitions refer to students moving from first year to second year, for example, and eventually to either program completion or dropping out, the state definitions can be extended to help meet needs specific to particular applications. Gandy et al (2019) add cumulative credit hour ranges and Rahim et al (2013) add age group ranges to their state definitions to gain additional insights as well as predictive power. Nicholls (2007) specifically focuses on the use of the Markov chain model for improving the program completion results for master's and PhD students, and also identifies the usefulness of the models for longer-term analyses.…”

Predicting Student Success in College Algebra Classes Using Machine Learning

“…Gandy et al [13] conducted a study using decision tree models and neural networks that utilize the strength of statistical learning to predict student success in the first two years of college. Fifty variables were examined including student retention rates as well as gender, race, academic and economic data.…”