Multilevel modeling of undergraduate student attrition across the University of North Carolina system

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Elizabeth M. Reilley (Creator)
Institution
The University of North Carolina at Greensboro (UNCG )
Web Site: http://library.uncg.edu/
Advisor
John Willse

Abstract: The volume of articles, books, and studies about increasing the retention, persistence, and graduation of undergraduate college students is nothing short of prolific (Seidman, 2005). However, only modest gains in undergraduate graduation rates have been made nationally (Chen, 2012; Seidman, 2005). Six-year graduation rates at all four-year colleges and universities rose minimally from 54.4% for the entering cohort of 1996 to 54.9% for the cohort beginning fourteen years later (U.S. Department of Education, 2019) with 35% of institutions experiencing declines in graduation rates during part of this period (Brainard & Fuller, 2010). Persistently low graduation rates coupled with recent leaps forward in technology, including processing speeds, statistical software, and data warehousing, have led many higher education researchers, practitioners, and companies to apply statistical models to examine what variables have a relationship with graduation. Many multi-university models suffer from a variety of hurdles including large amounts of missing data, missing important variables, questionable data quality and lack of common definitions across colleges or universities, and/ or inappropriate statistical methods that do not account for the nested nature of the data (students within universities). This study sought to avoid many of the limitations of past studies and used a two-level logistic hierarchical generalized linear model to comprehensively model six-year graduation in the UNC System. Included in this study were 406,909 undergraduate students who began undergraduate degree-seeking enrollment in any of the 16 public universities in the state of North Carolina from 2000 until 2010. Each variable included in the model was selected based on evidence in the literature of significant relationships with retention and persistence found in regression-based models. In comparison to past literature, this study included a wider array of financial and financial aid-related variables and examined more closely the relationship between university characteristics and student characteristics. Most level-1, student, variables included in this study were significant. The level-2, university, characteristics residential status and selectivity were found to have a significant relationship with six-year graduation and to have an influence on the relationship between some of the student-level covariates and six-year graduation. The results confirmed many of the relationships in the literature between the variables studied and student attrition with some fascinating deviations explored in the discussion. Limitations and suggestions for future research are provided. The results of this study will equip university practitioners and policy-makers in North Carolina with information to improve graduation and further explore student attrition. This study can act as a model for how other states or higher education systems use their own administrative data for comprehensive, multi-institutional modeling.

Additional Information

Publication
Dissertation
Language: English
Date: 2019
Keywords
Graduation, HGLM, Hierarchical Linear Modeling, Higher Education, HLM, Multilevel model
Subjects
University of North Carolina (System) $v Statistics
College dropouts $z North Carolina $v Statistics
College graduates $z North Carolina $v Statistics
College attendance $z North Carolina $v Statistics

Email this document to