University of Southern California Dissertations and Theses

(USC Thesis Other)

Factors related to the persistence of students seeking the bachelor's degree at 4-year institutions

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films
the text directly from the original or copy submitted. Thus, some thesis and
dissertation copies are in typewriter face, while others may be from any type of
computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality illustrations
and photographs, print bleedthrough, substandard margins, and improper
alignment can adversely affect reproduction.
in the unlikely event that the author did not send UMI a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and continuing
from left to right in equal sections with small overlaps.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations appearing
in this copy for an additional charge. Contact UMI directly to order.
ProQuest Information and Learning
300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA
800-521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
FA C TO R S R E LA TE D TO TH E P E R S IS T E N C E O F S TU D E N T S
S E E K IN G T H E BACHELOR’S D E G R E E A T 4-Y E A R
IN STITU TIO NS
by
Lee Blecher
A Dissertation Presented to the
F A C U L TY O F THE SC H O O L O F E D U C A TIO N
U N IV E R S IT Y OF SO U TH ER N C A LIFO R N IA
In Partial Fulfillment of the
Requirements for the D egree
D O C TO R OF P H ILO S O P H Y
(Education)
December 2000
Copyright 2000 Lee Blecher
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 3041437
Copyright 2000 by
Blecher, Lee
All rights reserved.
_ _ ®
UMI
UMI Microform 3041437
Copyright 2002 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 90007
This dissertation, written by
under the direction of h.% $...... Dissertation
Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School, in partial fulfillment of re
quirements for the degree of
L e e B le c h e r
D O CTO R OF PHILOSOPHY
Dean of Graduate Studies
D IS S E R TA TIO N C O M M ITTEE
yjhdX a* l > > 1
Chairperson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Lee Blecher William B. Michael, Ph.D.
A BSTRA C T
FACTO RS RELATED TO T H E P E R S IS T E N C E O F STU D EN TS
SEEK IN G TH E BACHELO R’S D E G R E E A T 4-Y E A R
IN STITU TIO N S
Background. This longitudinal study explored the 5-year persistence status of
students seeking the bachelor’s degree, irrespective of whether they had remained at
their initial institution, transferred, temporarily stopped out, or departed from higher ed
ucation altogether (“system persistence"). The sample of 3,278 students derived from
the 1989 Beginning Postsecondary Students Longitudinal Study included only those
students who had begun their postsecondary education at 4-year institutions and had
as their initial goal the completion of a bachelor’s degree.
Purpose. The initial purpose of this study was to determine whether the rela
tionship between selected variables in an hypothesized eclectic model to explain sys
tem persistence would be consistent with the observed data and, if not, whether a
systematically modified model could be developed which would have adequate
goodness-of-fit indices and, therefore, would support the plausibility of the postulated
relationships. The secondary purpose was concerned with determining (a) the degree
of relationship between each of several selected variables and 5-year system persis
tence and (b) the predictive capability of a composite of input variables to classify cor
rectly students as system persisters or nonpersisters.
Conclusions. (1) The background variables of age, socioeconomic status, and
ability, plus educational aspirations, percentage of full-time attendance, hours worked
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
at a job, scholastic achievement, and student involvement could help explain system
persistence. (2) A modified eclectic path analytic model was judged to provide an ade
quate degree of goodness-of-fit to afford a basis for an initial interpretation of both di
rect and indirect effects of key variables in explaining the persistence of bachelor
degree-seeking students. (3) Students who transferred were somewhat less likely to
persist for 5 years toward obtaining the bachelor’s degree. (4) Student satisfaction
levels appeared to have a significant impact on the decision to transfer but little, if any,
direct relationship to persistence in completing the bachelor’s degree. (5) Input vari
ables based on promising results from bivariate analyses led to the formation of a
multiple discriminant analysis that could predict system persistence at a level at least
moderately greater than that afforded by chance alone.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGMENTS
I would like to express my sincere appreciation to those who helped make this
dissertation a reality.
I am deeply grateful to my dissertation committee chairperson, Dr. William B.
Michael, for guiding me through to the completion of the dissertation process. His
help, encouragement, insights, and keen sense of professionalism were all instrumen
tal in helping make this possible. It was a privilege to have had the opportunity to work
under his guidance.
I would also like to extend my appreciation to the other members of my disser
tation committee: Dr. Linda Serra Hagedorn, for her expertise and continued support;
and Dr. Peter J. Robertson, for his guidance in completing the dissertation.
I am grateful to many of the faculty at the University of Southern California for
providing me with the academic preparation in order to complete the doctoral program.
In particular, I would like to thank Dr. Dennis Hocevar for his continued help with statis
tics and Dr. William E. Maxwell for his advice and encouraging words through the
coursework portion of my doctoral program.
I would also like to acknowledge the faculty and administrators at California
State University, Long Beach. Their encouragement and support were significant in
allowing me to complete the doctoral program. In particular, I would like to sincerely
thank Dr. Mary Jacob for her role as a mentor from the CSU system. Her ongoing
ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
support, encouragement, and words of wisdom were instrumental in helping me ac
complish this goal.
Last, but not least, I would like to sincerely thank my family and close friends
for their understanding, help, and encouragement while I was completing the doctoral
program.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CONTENTS
Page
ACKNOWLEDGMENTS ................................................................................................. ii
LIST OF TABLES ............................................................................................................. vii
LIST OF FIGURES .......................................................................................................... viiii
Chapter
1. IN TR O D U C TIO N ............................................................................................ 1
Background of the Problem................................................................... 1
Area of Concern ..................................................................................... 2
Purpose of the S tud y.............................................................................. 4
Importance of the Study ........................................................................ 5
Delimitations............................................................................................ 6
Definition of Terms ................................................................................. 6
Four-Year Institutions..................................................................... 7
Horizontal Transfer.......................................................................... 7
Institutional Persistence................................................................. 7
Reverse T ran sfer............................................................................ 7
Stopout ............................................................................................ 7
System Persistence........................................................................ 7
Transfer............................................................................................ 8
Organization of the Remainder of the Study ..................................... 8
2. REVIEW OF THE RELEVANT LITERATURE............................................ 9
Theoretical Perspectives for Understanding
Persistence ..................................................................................... 9
Tinto’s Model of Student Departure ............................................ 10
Bean’s Model of Student Attrition................................................. 14
Astin’s Theory of Involvement........................................................ 17
Similarities and Differences Among the
Theoretical Models ................................................................. 20
Application of the Theoretical Foundations to
System Persistence........................................................................ 22
Hypothesized Eclectic Model for System Persistence..................... 23
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Page
Review of Factors Selected to Explain System
Persistence ..................................................................................... 24
Socioeconomic Status and Persistence ..................................... 24
Age and Persistence ...................................................................... 24
Gender and Persistence................................................................. 26
Academic Ability and Persistence................................................. 27
Goal Aspirations and Persistence................................................. 27
Enrollment Intensity and Persistence.......................................... 29
Pull Factors and Persistence ........................................................ 29
Student Involvement/Integration and Persistence..................... 30
Scholastic Performance and Persistence................................... 32
Satisfaction and Persistence ........................................................ 33
Transfer Status and Persistence ................................................. 34
Research Questions and Hypotheses................................................. 34
Research Question 1 ..................................................................... 35
Research Question 2 ..................................................................... 35
Research Question 3 ...................................................................... 35
Research Question 4 ...................................................................... 36
Research Question 5 ...................................................................... 37
3. METHODS AND PROCEDURES................................................................. 39
Research S a m p le ................................................................................... 39
Data Collection ....................................................................................... 41
Variables and Their Measures............................................................... 42
Background and Demographic V ariab les................................... 43
Goal Aspiration Variables............................................................... 43
Enrollment Intensity and “Pull-Factors” V ariables..................... 45
Involvement Variable ..................................................................... 46
Scholastic Performance V ariab le................................................. 46
Satisfaction Variables..................................................................... 47
Transfer Variables .......................................................................... 48
Persistence Variables..................................................................... 49
Data Analysis .......................................................................................... 50
Methodological Assumptions................................................................. 51
Limitations................................................................................................. 51
4. PRESENTATION OF RESULTS ................................................................. 52
Analysis of Findings .............................................................................. 52
Descriptive Statistics Related to 5-year System
Persistence.............................................................................. 52
Adequacy of the Goodness-of-Fit Indices of
the Eclectic Model to Explain System
Persistence (Research Question 1 ) ..................................... 53
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Page
Adequacy of the Goodness-of-Fit Indices of a
Modified Model to Explain System Persistence
(Research Question 2) .......................................................... 54
Direct and Indirect Effects of Each of the Predictor
Variables on System Persistence (Research
Question 3 ) .............................................................................. 57
Bivariate Relationship Between Predictor Variables
and 5-year System Persistence (Research
Question 4 ) .............................................................................. 57
Utilizing Multiple Discriminant Analysis to Predict
5-year System Persistence (Research Question
5 ) ................................................................................................. 66
5. DISCUSSION, CONCLUSIONS, AND RECOMMENDA
TIONS ........................................................................................................ 77
Discussion................................................................................................. 77
Adequacy of the Goodness-of-Fit Indices of the
Initial Hypothesized Eclectic Model to Explain
System Persistence (Research Question 1) ..................... 77
Adequacy of the Goodness of Fit Indices of the
Modified Model to Explain System Persistence
(Research Question 2) .......................................................... 78
Direct and Indirect Effects of Each of the Predictor
Variables on System Persistence (Research
Question 3 ) .............................................................................. 79
Bivariate Relationship Between Each of the
Predictor Variables and 5-year System
Persistence (Research Question 4 ) ..................................... 86
Utilizing Multiple Discriminant Analysis to Predict
5-year System Persistence at a Level Greater
Than Chance (Research Question 5) ................................ 92
Conclusions ............................................................................................ 95
Recommendations ................................................................................ 96
R EFER ENC ES................................................................................................................. 98
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Table Page
1. The Direct, Indirect, and Total Effects of Input Variables on
5-year System Persistence................................................................. 58
2. The Relationship Between Gender and 5-year System
Persistence............................................................................................. 60
3. A Comparison of the Means of SAT and ACT Scores in
5-year Persisters Versus Nonpersisters ........................................... 61
4. Relationship Between Transfer Status and 5-year System
Persistence............................................................................................. 66
5. Correlation Matrix for Variables Used in the Discriminant
Analysis .................................................................................................. 68
6. Standardized Coefficients and Structure Coefficients of the
Predictor Variables Entered Independently in the Dis
criminant Analysis ................................................................................. 70
7. Classification Results from the Initial Discriminant Analysis .................. 71
8. Standardized Coefficients and Structure Coefficients of the
Predictor Variables Used in the Reduced Discriminant
Analysis .................................................................................................. 75
9. Classification Results from the Discriminant Analysis
Utilizing Seven Predictor V ariab les................................................... 76
vii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF FIGURES
Figure Page
1. Hypothesized eclectic model of system persistence................................ 25
2. Modified eclectic model of system persistence......................................... 56
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C H A P TE R 1
IN TR O D U C TIO N
Background of the Problem
Understanding the causes of and factors related to the retention and persis
tence of students in higher education has been an area of much discussion and re
search, especially in the latter half of the 20th century (Astin, 1975; Pascarella & Ter-
enzini, 1991; Summerskill, 1962; Tinto, 1993). Despite the large body of knowledge
that has emerged from such interest, student departure from institutions of higher edu
cation continues to be of great concern. Tinto (1993) reported that more than one
quarter (28.5%) of the students entering 4-year institutions of higher education end up
departing those institutions prior to beginning their second year, and only about one
half (50.2%) complete their bachelor’s degree at their initial institution within 5 years.
Attrition rates for those beginning in 2-year institutions have been even higher. The
potential negative impact of departures from higher education is of great concern not
only for the students themselves but also for the institutions, as well as for society as a
whole. When students drop out from college, the benefits associated with degree
completion will presumably go unrealized.
Demographic and longitudinal studies have shown that individuals completing
postsecondary degree programs typically make gains in occupational status and have
benefited from long-term monetary rewards (Leslie & Brinkman, 1986; Mortenson,
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1999). According to the U.S. Department of Education, bachelor degree recipients
ages 25-34 earned more (52% for males and 91% for females) than their counterparts
with only high school diplomas (National Center for Educational Statistics [NCES],
1997). In addition, the perceived value of a postsecondary education seems to be at
an all-time high in American society. According to a 1998 survey reported by the Na
tional Center for Public Policy and Higher Education, Americans feel that higher edu
cation is more important than ever before as a means to a middle-class lifestyle and as
a resource for local economy. Seventy-five percent indicated that its importance had
increased over the past decade (Immerwahr, 1998). Along these same lines, a follow-
up survey indicated that 87% of Americans perceive that “a college education has be
come as important as a high school diploma used to be” (Immerwahr & Foleno, 2000,
p. 1).
The significance of this issue is also reflected by the fact that institutions of
higher education are required to report student retention and completion rates as part
of the Student Right to Know and Campus Security Act of 1991 (Astin, 1997). These
statistics, then, become one of the factors by which potential students can rate an in
stitution when they are determining a suitable location to fulfill their educational aspira
tions. Correspondingly, institutions have developed policies and allocated resources
with the intent of improving the retention rates of students within their institution.
Area of Concern
Although there has been a substantial amount of research in the area of stu
dent retention and persistence, most studies have focused on institutional departure
which, as it turns out, is not the same as departure from the system of higher
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
education as a whole (“system persistence"). Students departing from a given institu
tion could be leaving education altogether (i.e., dropouts), but they could also be
headed to other potential fates such as transferring to a different institution (transfers)
or just temporarily leaving to return to the persistence path at a later time (stopouts).
Although from the institutional point of view these various student paths resemble a
similar outcome, namely institutional departure, from the viewpoint of the student as
well as from that of the society as a whole, these various outcomes are presumably
quite different. The potential benefits of completing a degree in higher education will
most likely be reaped by students irrespective of whether the degree was completed at
the initial institution or after transfer to another institution for final matriculation. Under
standing the characteristics of students who successfully transition through higher ed
ucation by these various paths can further expand current knowledge of persistence
and attrition in postsecondary education.
There is substantially less information regarding retention rates within the sys
tem of higher education as a whole (i.e., “system" persistence). As might be expected,
persistence and completion rates are higher when students are followed whether they
had transferred or stopped out temporarily (Adelman, 1998). For students beginning
at 4-year institutions, bachelor degree completion rates (independent of path) ranged
from 53% to 65%, depending on what the parameters of the study were and whether
the students were assessed after 5, 6, or more years. If one includes students still en
rolled with those who have already completed bachelor’s degrees, then the “system”
persistence rates are somewhat higher again, depending on which study was con
sulted and on how long the students were followed (NCES, 1997; Tinto, 1993).
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The theoretical models posited to explain student persistence and departure
behavior in higher education have focused almost entirely upon institutional, as
opposed to system-wide, departure behavior. The extent to which these models ex
plain the behaviors and causal factors which might predict whether a student persists
(or ceases to persist) through the system of higher education has not been estab
lished. Despite this fact, these models of institutional departure behavior can form the
basis from which to investigate further student persistence in the system of higher ed
ucation.
Purpose of the Study
The purpose of this study was to examine the relationship between a number
of variables and the system-wide persistence and achievement of students enrolled in
bachelor’s degree programs at 4-year postsecondary institutions. Enhancing the cur
rent body of knowledge in this way would help to determine whether the current theo
retical models posited to explain institutional persistence (or portions of the models)
could be applied to explain system persistence. In order to accomplish this goal, this
study utilized the Beginning Postsecondary Students (BPS) longitudinal database from
the NCES (1994a, 1996), which includes a nationally representative sample of stu
dents from various geographical regions and types of institutions. The data for this
study were collected over a 5-year period beginning in the fall of 1989, when the stu
dents in the sample initially began their postsecondary education. The longitudinal
study followed the students irrespective of whether they had remained at their initial 4-
year institution, had transferred to another institution of higher education, temporarily
had left the persistence path, or had departed from higher education altogether.
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
On the basis of incorporating what appeared to be the most promising con
structs and variables from each of three theories dealing with persistence and
achievement of students in higher education as offered by Vincent Tinto (1975,1993),
John P. Bean (1980, 1983, 1990), and Alexander W. Astin (1975, 1984, 1993), the ini
tial purpose of this study was to determine whether these variables and the relation
ships between them as posited in an hypothesized eclectic model to explain system
persistence would be consistent with the observed data (i.e., have adequate
goodness-of-fit indices) and, if not, whether a systematically modified model could be
developed which would be consistent with the data and make a case for the plausibility
of the postulated relationships between the variables. The secondary purpose was
concerned with determining (a) the degree of relationship between each of several se
lected demographic, behavioral, cognitive, and affective variables and the 5-year sys
tem persistence of students seeking bachelor’s degrees in 4-year postsecondary
institutions; and (b) the predictive capability of a composite of input variables that, in
terms of use of multiple discriminant analysis, would predict at a level greater than that
of chance correct classifications of students as system persisters or nonpersisters
within a 5-year period.
Importance of the Study
This study is significant because it attempts to explain further the persistence
behavior of bachelor degree-seeking students in 4-year institutions of higher educa
tion. The study may afford unique opportunities to expand the current theoretical body
of knowledge. Although limiting the number of variables that could be studied to those
available in the database, the Beginning Postsecondary Students longitudinal study
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
represents the first time that an extensive cohort of both traditional- and nontraditional-
aged students beginning their postsecondary education have been followed over mul
tiple years.
Through an analysis of the relationships between predictors and the criterion
variable of persistence, it was determined how well selected tenets and constructs
from the research literature are supported in this study. The increased understanding
of the characteristics and behaviors of students who successfully transition through
higher education can help administrators to develop more effectively policies and in
tervention methods directed towards enhancing student retention.
Delimitations
The following delimitation were applicable to this study:
1. The study sample was restricted to those students who had begun their
postsecondary education during the 1989-1990 academic year.
2. The study involved only those students seeking a bachelor’s degree who
had begun their postsecondary education at 4-year institutions.
3. The variables utilized in this study were limited to those available through
the Beginning Postsecondary Students longitudinal study database.
4. The study sample included only those students for whom information re
garding their 5-year persistence status was available.
Definition of Term s
Definitions of the following terms are given to facilitate understanding of this
study:
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Four-vear Institutions
This terminology represents institutions of higher education that grant bacca
laureate degrees. Institutions that grant graduate degrees in addition to the bachelor’s
degree were also included under the designation “4-year institutions” for the purpose
of this study.
Horizontal Transfer
This term is used to designate the transfer from one 4-year institution to an
other 4-year institution.
Institutional Persistence
Institutional persistence refers to students who have either completed or are
still enrolled (full-time or part-time) in their initial institution within a 5-year period.
Reverse Transfer
This term indicates the transfer from a 4-year institution to a 2-year institution.
Stopout
The term pertains to students who leave the persistence path for 4 or more
months to return at a later time. The term does not imply whether the student returned
to the original institution or to a different institution after “stopping out.”
System Persistence
System persistence refers to students who have either completed or are still
enrolled in the educational system, regardless of whether they transferred from their
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
initial institution. In this case, the educational system referred to is baccalaureate de
gree programs in 4-year institutions. Enrollment on either full-time or part-time basis
qualified as persistence.
Transfer
This term indicates the movement of a student from one institution to another.
Organization of the Remainder of the Study
Chapter 2 contains an overview of theoretical perspectives related to persis
tence in higher education, plus a review of the relevant literature related to variables in
this study. At the end of Chapter 2 the research questions and hypotheses for this
study are presented. Chapter 3 is an explication of the methodology utilized in this
study including a discussion of the sample, research design, methods of data analysis,
and limitations. Chapter 4 is a presentation of the analyses and findings of the study.
Chapter 5 includes the discussion, conclusions, and recommendations.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C H A PTER 2
R E V IE W OF TH E R ELEVA N T LITERATURE
Theoretical Perspectives for Understanding
Persistence
Although most of the early literature was descriptive by nature, William G.
Spady (1970, 1971) made an early attempt to incorporate various psychological and
sociological variables into a complex causative theoretical model to explain student
persistence. In 1975 Vincent Tinto elaborated on the work of Spady and developed a
model of institutional student departure that became the foundation and impetus for
numerous later studies in the area of student attrition (Braxton, Sullivan, & Johnson,
1997; Pascarella & Terenzini, 1980, 1991). His model focused on the need for stu
dents to become adequately integrated into the academic and social communities of
the college in order to promote persistence and to prevent voluntary departure (Tinto,
1975).
Another researcher, John P. Bean, also developed a causative model of stu
dent attrition that was intended to explain why students departed from institutions of
higher education from a slightly different theoretical perspective (Bean, 1980, 1983).
In addition to incorporating some of the psychological, sociological, and environmental
variables associated with student educational outcomes, Bean utilized concepts from
the “organizational” literature to help in framing his theory on student attrition. Later,
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
both Tinto (1993) and Bean (1990) refined their models to be more inclusive, although
these revisions retained their original theoretical frameworks.
Alexander W. Astin is another key figure who has conducted related research
in the field of higher education. Although his theoretical perspective emanates from a
broader view that is intended to explain a variety of educational outcomes, much of his
initial research, which established the foundation for his theoretical perspective, was
focused on student retention issues in higher education. Astin has made significant
contributions to the understanding of the factors associated with student attainment
through extensive longitudinal studies (Astin, 1975, 1984, 1993).
Various researchers (Berger & Braxton, 1998; Cabrera, Castaneda, Nora, &
Hengstler, 1992; Milem & Berger, 1997) have utilized tenets from these basic theoreti
cal models in an effort to further the understanding of student retention and persis
tence. The following is an overview of the essential components and theoretical con
cepts developed by Tinto, Bean, and Astin as they relate to student persistence in
postsecondary education.
Tinto’s Model of Student Departure
Tinto’s model of student departure falls under the general category of what is
referred to as interactional theories of student departure. As a theoretical classifica
tion, interactional theories tend to view the occurrence of student persistence as a
reflection of a positive dynamic interaction between the individual student and the col
lege environment (Braxton et al., 1997). Tinto’s model builds upon the work of Spady
and furnishes a framework for understanding the causes of student attrition at institu
tions of higher education (Spady, 1970, 1971; Tinto, 1975, 1986).
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Like Spady, Tinto derived essential components of his model of student depar
ture from the work of Emile Durkheim, a French sociologist who, among other things,
elaborated on theories of suicide to help to explain its occurrence in society (Durkheim,
1951). Durkheim distinguished among four types of suicide: altruist, anomic, fatalistic,
and egotistical. However, it was the last of these, egotistical suicide, that provided
both Spady and Tinto an analogy for insights into the understanding of student attrition
(Spady, 1970; Tinto, 1975, 1993). According to Durkheim, egotistical suicidal behavior
is preceded by the inability of an individual to integrate and establish membership into
society successfully. Integration into society can occur as “social integration" which
evolves from the day-to-day interactions between the individual and others within so
ciety, as well as “intellectual integration,” which relates to having values congruent with
those of mainstream society. Durkheim contended that both types of integration must
be deficient for egotistical suicide to occur. Although leaving college is not of the same
magnitude as committing suicide, Tinto did draw upon the concepts presented by
Durkheim and applied them to the issue of student attrition in higher education. In
essence, the backbone of Tinto’s theory of student departure is the contention that
successful integration into the academic and social communities of the institution is
paramount for continued persistence in higher education (Tinto, 1975, 1993).
Tinto has also perceived integration as a time-dependent dynamic process that
occurs between the student and the college community. To understand further the
longitudinal nature of this process, Tinto (1986, 1993) drew upon the works of the
Dutch social anthropologist, Arnold Van Gennep (1960), who studied the “rites of pas
sage” of individuals in tribal societies as they pass through various stages in order to
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
obtain social membership as adults in society. The process is described as having
three distinct stages: separation, transition, and incorporation.
Although the situation of individuals experiencing “rites of passage” within a
given “ tribal” society is not the same as that of students from various backgrounds en
tering into and successfully functioning within an academic institution (Tierney, 1992),
Tinto embraced concepts inherent in Van Gennep’s (1960) formulation as a framework
for understanding the time-dependent process of students evolving to the stage of in
corporating and integrating into the academic community (Tinto, 1993).
As viewed from the perspective of Tinto’s model, the first stage, separation, oc
curs as a student disassociates himself or herself from past affiliations such as family,
neighborhood, and/or high school friends. In the second stage, the transition stage,
the student begins to establish new patterns of behavior and affiliations with the new
community and environment. New skills are developed which enable the student to
perform adequately in his or her new role as a student in higher education. The final
stage of Van Gennep’s “rite to passage” is the incorporation stage, which in the con
text of Tinto’s model occurs when a student finally assumes his or her new role as a
member of the campus community. This assimilation can take on varying forms, both
formal and informal, and facilitates the social and academic integration important for
persistence in the college environment (Tinto, 1993).
In addition to integration, Tinto’s theory of student departure considers the
background characteristics (i.e., pre-entry attributes) of each individual, plus his or her
goals, intentions, and commitments. The student’s background characteristics include
socioeconomic status; parents’ education and income; individual attributes such as
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
race, gender, and skills; and the student’s pre-college educational achievements (e.g.,
high school grade point average [GPA]). These pre-entry attributes are predicated to
have both direct and indirect effects upon voluntary departure behavior (Pascarella &
Terenzini, 1991; Summerskill, 1962; Tinto, 1975,1993).
According to Tinto’s theory, a student enters college with goals and intentions
that reflect the level of education or occupation desired. Tinto postulates that the stu
dent’s background characteristics, plus the commitment toward his or her goals, as
well as the initial commitment or affinity to the institution, will all have an impact on the
student’s integration into the academic and social systems of the college community.
As defined in Tinto’s model, academic integration occurs through classroom experi
ences, academic achievements (e.g., grades), and intellectual development. Social
integration, on the other hand, occurs through extracurricular activities and informal
interactions with friends, staff, and faculty.
Tinto theorizes that successful academic integration and social integration
serve to shape further and reinforce a student’s academic goals and institutional com
mitments. These subsequent goals and commitments are then predicted to promote
persistence within the institution (Tinto, 1975, 1993).
Although integration is a critical component of the theory devised by Tinto, he
does not contend that full integration into both the academic and social communities of
the college is required for persistence. Instead, his theory asserts that, as a minimum
condition for continued persistence, there has to be some degree of integration into at
least one social or academic community of the institution. Even though social and aca
demic integration can compensate for each other, Tinto (1993) is quick to point out that
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
a student must maintain at least some level of academic integration to sustain his or
her academic standing.
It should be noted that Tinto’s model focuses on voluntary departure from a
single institution. It is not intended to explain the behavior of students who persist via
transferring to other institutions or of those who “stop out” of college temporarily to re
turn and to persist at a later time.
Bean’s Model of Student Attrition
Bean made an effort to utilize concepts derived both from the prevailing educa
tional retention literature (e.g., Tinto) and from the theoretical models of different dis
ciplines. In particular, he is noted for incorporating organizational and environmental
factors into his model of student attrition (Bean, 1990; Pascarella & Terenzini, 1991).
Bean (1980, 1983) postulated that student attrition is analogous to employee
turnover in the work place. He drew from the works of both Price (1977) and Price and
Mueller (1981) to explain student departure behavior from higher education. The
Price/Mueller model not only incorporates organizational variables as determinants of
employee turnover behavior but also includes the concept of “intent to leave” as a part
of the causal sequence to departure behavior. In essence, the underlying conceptual
linkages that Bean derived from these studies are that the experience of an individual
in the organization, together with his or her beliefs and values, will have an impact on
subsequent attitudes. These attitudes, then, will lead to intentions which, in turn, lead
to behaviors. In the case of student attrition, the final behavior to predict and to explain
is dropout behavior (Bean, 1980, 1983, 1990).
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Employee turnover models also provided Bean with the insight to include “envi
ronmental” or external variables in his theory (Bean, 1982; Price & Mueller, 1981).
Factors external to the college are postulated to have either a positive effect on stu
dent persistence or, on the other hand, a negative effect through the creation of an
environmental “pull." Although institutions of higher education typically have little con
trol over external forces on students, Bean thought that they were important to include
as factors in the model in order to better understand student departure behavior
(Bean, 1990).
Similar to other models of student attrition, Bean’s model is intended to explain
student departure from one institution. Therefore, transfer students and “stopouts” are
also considered cases of attrition. However, it should be noted that, unlike some other
researchers in this area (e.g., Tinto), Bean’s model of student attrition includes both
voluntary departures and academic dismissals as dropouts. Bean contended that aca
demic failure is an example of an outcome of unsuccessful interactions in the aca
demic setting. Therefore, one would expect the determinants of student attrition, and
in particular grades, to predict both voluntary or involuntary (i.e., forced termination
because of academic dismissal) departure. Consequently, Bean thought that eliminat
ing academic dismissals from the model was not necessary and, in fact, would bias the
grade variable by skewing it upward (Bean, 1980, 1983, 1990).
In applying the Price/Mueller model of employee turnover, Bean (1983) found
parallels from the work place to colleges for the determinants of turnover. In some
cases he had to establish surrogate variables found in the educational setting to re
place those formulated by Price and Mueller (1981). For example, Bean utilized
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
“grades” as one of the surrogates to help to establish a “pay” variable as seen in the
employee turnover model (Bean, 1983). This thinking was consistent with that of
Spady, who considered grades to be an extrinsic reward in the “quasi-occupational
role-playing of the career oriented student” (Spady, 1970, p. 77).
In general, Bean considered background characteristics of a student as a nec
essary component in understanding the causes of student departure behavior. These
background variables include educational goals and aspirations, prematriculation
academic performance, college preparatory curriculum, parental income, parental ed
ucation, and parental support. Bean theorized that background variables have an im
pact on a student’s academic and social interactions within the institution as well as on
environmental “pull” factors. His model also reflects the premise that background var
iables have a direct effect on the academic achievement (grade point average [GPA])
of students while in college (Bean, 1990; Pascarella & Terenzini, 1991).
Variables associated with academic interactions include studying skills and
habits, interactions with faculty, major certainty, and absenteeism. Social integration
factors include interactions with friends on campus, informal faculty interactions, and
other interactions provided by social support systems. Environmental “pull” variables
refer to factors that tend to interfere with a student’s interactions with the campus or
ganization. They can include lack of finances, presence of a significant other else
where, opportunity to transfer, work role, and family responsibilities (Bean, 1990). Or
ganizational variables such as rules and regulations, admission and financial aid poli
cies, course schedules, and academic and social services are included in Bean’s
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
model of student attrition, although they are not, in themselves, postulated to be af
fected by students’ background traits (Bean, 1990; Berger & Braxton, 1998).
These organizational, environmental, and academic and social interaction fac
tors are posited by Bean to have effects on various intervening attitudinal variables
and educational outcomes that would ultimately impact departure behavior. The atti
tudinal variables included in Bean's model of student attrition include satisfaction in
being a student, sense of self-development, perception of the practical value or utility
of the education, self-confidence as a student, and stress. Bean theorizes that these
variables directly affect a student’s perception of institutional fit and institutional loyalty,
which is considered critical for continued enrollment (Bean, 1990).
According to Bean’s model, grades (GPA) are an educational outcome that is
directly affected by a student’s background characteristics and their academic interac
tions within the institution. Grades, institutional fit, and institutional loyalty (commit
ment) comprise the three intervening variables theorized to have direct and indirect ef
fects on dropout behavior. The indirect effects are postulated to occur through “intent
to leave,” which can function as a useful intervening variable between attitudes and
behavior to predict dropout behavior (Bean, 1980, 1983, 1990). Bean points out that
identifying departure behavior before it happens can allow appropriate intervention po
tentially to occur.
Astin’s Theory of Involvement
Although the work of Astin has evolved over the course of more than 2 de
cades, the general model underpinning the framework of his studies is an input-
process/output model (Pascarella & Terenzini, 1991). Inasmuch as Astin (1993)
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
utilized the model to focus on interactions as defined within an environment, he later
referred to the model as an Input-Environment-Output (l-E-O) model. However, the
general concept remained the same. According to the model, students enter the pro
cess of postsecondary education with different “input” variables, such as their ability,
socioeconomic status, and gender. Given these background characteristics, the vari
ous interactions and experiences that occur between students and their environment
while attending postsecondary education subsequently shape and affect their educa
tional outcomes. In agreement with this model, Astin developed a theory of involve
ment which maintains that the extent to which students experience various interactions
with the environment determines the level of student involvement (Astin, 1975, 1984,
1993). According to Astin, the environment provides students with multiple opportuni
ties to become actively involved. The level of involvement is reflected by the “amount
of physical and psychological energy that the student devotes to the academic experi
ence” (Astin, 1984, p. 297). This concept is similar to what learning theorists refer to
as “time-on-task” (Astin, 1984; Pascarella & Terenzini, 1991).
Astin postulated that involvement occurs along a continuum. Students who
actively participate in various curricular and extracurricular activities are considered
highly involved. Conversely, the uninvolved student is characterized by his or her lack
of time and energy devoted to participating in these types of activities. The level of
student involvement then becomes a major factor in determining whether the student
will eventually persist in college. According to this view, persistence is seen as contin
ued involvement, and dropping out is viewed as the ultimate form of noninvolvement.
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Astin’s work with longitudinal studies supports the contention that factors asso
ciated with an increase in student involvement are likely to be associated with in
creased persistence, whereas those associated with a decrease in student involve
ment are likely to be negatively associated with persistence (Astin, 1975). For in
stance, student interaction with faculty and other students, studying, and going to class
are examples of involvement behavior that are positively associated with persistence.
On the other hand, activities that “pull" students away from college oriented involve
ment, such as working full-time off campus, are negatively related to persistence (As
tin, 1975, 1984, 1993).
Although Astin recognizes that both psychological factors (perceptions) and
behaviors are important in the understanding of student outcomes, much of his work
has been focused on the behavioral antecedents to student attainment (Astin, 1975,
1984, 1993). The concept of involvement, itself, implies behavioral components. Astin
(1984) contended that the behavioral aspect is critical to emphasize because “it is not
so much what the individual thinks or feels, but what the individual does, how he be
haves, that defines and identifies involvement" (p. 298).
Much of Astin's work is descriptive by nature, as it draws upon a vast amount of
data that he and his colleagues have collected from large longitudinal studies. Al
though it has been questioned whether his model constitutes a theoretical framework
(Pascarella & Terenzini, 1991), Astin’s work has contributed significantly to furthering
the understanding of factors associated with student development, achievement, and
persistence in postsecondary education.
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Similarities and Differences Among
the Theoretical Models
One similarity between Tinto’s model of student departure and Bean’s model of
student attrition is that they are both theoretical longitudinal models specifically in
tended to explain student departure from higher education. Alternatively, Astin’s
model is intended to explain a variety of student outcomes including persistence and
attainment. Although Astin’s model is more simple and parsimonious than the models
established by Tinto and Bean, its key tenets do not seem to contradict and, in fact,
support the underlying orientation found in the more elaborate causal models posited
by Tinto and Bean.
All three models include background variables, ability, and goals as important
determinants of persistence in higher education. In addition, they all consider stu
dents’ interaction (or involvement or integration) with the academic and social aspects
of the college to be important to persistence. All three would contend that formal and
informal interactions with faculty and peers are a significant component in promoting
successful integration into the college community. In addition, all three models would
support the notion that student-institution “ fit” is critical for institutional persistence
(Astin, 1984,1993; Bean, 1990; Tinto, 1993).
Relative to integration, Tinto’s model of student departure seems to reflect a
more encompassing interpretation of the concept. To explain further, when the tenets
of Tinto’s model have been operationalized, a perceptual perspective (i.e., “are you
satisfied with . . . ”) has been employed to measure academic and social integration
(Pascarella & Terenzini, 1980). From Tinto’s standpoint, this perspective comple
ments the contention that successful integration can be determined only through the
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
student’s perception of that integration (Tinto, 1975,1993). However, by way of com
parison, Tinto’s theory has merged what Bean’s model has separated into interaction
versus attitudinal variables. Unlike Tinto, Bean has viewed satisfaction as an educa
tional outcome and as an intermediary attitudinal variable having direct and indirect
effects on student persistence.
Astin also has distinguished between the roles that behaviors and perceptions
play in understanding persistence (Astin, 1996; Berger & Milem, 1999). Accordingly,
Astin has characterized “involvement” using only behavioral measures and, similar to
Bean, has viewed satisfaction as a separate “perceptual” variable.
All three models include grades or GPA in their explanation of departure be
havior. In Tinto’s model, however, grades are perceived as a component or indicator
of academic integration, whereas in Bean’s and Astin’s model, students’ grades are
considered as a separate intermediary variable.
In comparison to Tinto, both Bean and Astin have incorporated environmental
“pull" factors such as financial needs and outside commitments as important determi
nants of student attrition. From Tinto’s theoretical perspective, one might allege that
these factors become important only to the extent that they impact social and aca
demic integration and, therefore, that their presence and impact can be expected to be
subsumed under the integration constructs. However, in more recent writings Tinto
has acknowledged the importance of environmental factors.
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Application of the Theoretical Foundations to
System Persistence
Very little has been done to advance the theoretical understanding of system
persistence. Nevertheless, it seems reasonable to begin that endeavor by using the
theoretical models postulated to explain institutional persistence as a foundation from
which to study and to further the current body of knowledge on system persistence. In
1981 Munro followed this same reasoning and utilized constructs derived from Tinto’s
model of student departure to propose a path analytic model to explain system persis
tence. Although the usefulness of Munro’s research for establishing a theoretical foun
dation for explaining system persistence is questionable because constructs were
somewhat confounded in the study, her work established an approach for investigating
system persistence. The current investigation takes this same approach and utilizes
constructs from the causal models that have been put forth to explain institutional per
sistence (i.e., those by Tinto and Bean) as a framework from which to investigate sys
tem persistence. It might be noted that Munro’s work provided some useful empirical
information related to system persistence. It is referenced as applicable to specific
constructs in the latter portions of this chapter.
One variable that is included in this investigation of system persistence that is
not found in the causal models put forward by both Tinto and Bean is the variable
transfer, itself. The reason that transfer status is not found as a variable in these pre
viously developed causal models is that those models focused on institutional persis
tence, which considers students who transfer as institutional dropouts. However, the
empirical data suggesting a relationship between transferring and persistence
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(discussed later in this chapter) warrants the inclusion of transfer status as a variable
in the theoretical framework for understanding system persistence.
Astin’s work using the l-E-O model as a more general framework from which to
understand a multitude of student outcomes, including educational attainment, also
deserves consideration in a longitudinal study on system persistence because much of
his work utilized students tracked longitudinally irrespective of whether they stayed in
the same institution. Consequently, it seems appropriate to include, in a study of sys
tem persistence, variables from Astin’s work shown to be related to persistence, espe
cially when those variables are similar to ones found in the causal models developed
by either Bean or Tinto.
Hypothesized Eclectic Model for System
Persistence
Key tenets derived from the theoretical models posited by Tinto, Bean, and As
tin were used as a basis to develop a hypothesized model to explain system persis
tence. The selected demographic, behavioral, cognitive, and affective variables uti
lized in the eclectic model and the proposed relationships among them and the final
dependent variable are all, except for the variable transfer status, found in the theoreti
cal models discussed in the previous section. Although most of the key constructs
from the persistence literature are represented in the proposed eclectic model, it is
worthwhile to mention that neither institutional commitment nor intent to persist were
used in the current study as variables intended to explain system persistence, as they
were not included in the database. The essential components of the hypothesized
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
eclectic model of system persistence (variables and specific relationships) are pre
sented in Figure 1.
Review of Factors Selected to Explain System
Persistence
Socioeconomic Status and Persistence
The socioeconomic status (SES) of students has been shown to be positively
associated with bachelor degree persistence and completion in a number of studies
(Astin, 1993; Bean, 1990; Porter, 1990; Summerskill, 1962; Tinto, 1993). The effect
seems to be only modest, and sometimes nonexistent, when other factors such as
ability are controlled (Baker & Velez, 1996). This effect can be understood to reflect
the value that families from various SES groups may put on education (Summerskill,
1962), plus the financial capability to support the student to go to college (Bean, 1990;
Cope & Hannah, 1975).
Age and Persistence
Although most studies related to persistence in postsecondary education
delimit their inquiry to include only traditional-age students, there has been a handful of
studies that have looked into the relationship between age and persistence. Summer
skill (1962) summarized the literature to that point as inconsistent and concluded that
age, as such, does not affect attrition, although older students may encounter more
obstacles as they proceed toward graduation. The fact that older students are more
likely to be married, have children, live off campus, and be working tends to put them
in the position of going to school as an addition to (as opposed to “in place o f) other
obligations and commitments. This observation has been supported by the works of
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
Background and
Demographic
Variables
1. Socioeconomic
Status
2. Age
3. Gender
4. Academic
Ability
Initial Subsequent
Educational ------------------------------- ► Educational
Aspirations Aspirations
Career and
Financial Goals
Enrollment Intensity
(Percentage Enrolled
Full Time)
Pull Factors
1. Hours worked
2. Family
Responsibilities
Transfer
Status
Involvement Satisfaction Persistence
. .. . p
w
Scholastic
Performance
(GPA)
Figure 1 . Hypothesized eclectic model of system persistence
to
tn
both Tinto (1993) and Ozga and Sukhnandan (1998). Astin (1975) found that, in com
parison to students who enter college at a more traditional age (17-19 years), those
entering college at an older age are more likely to drop out. Murtaugh, Burns, and
Schuster (1999) also found that attrition at a 4-year university increased with student
age. Similar findings were noted by the NCES (1997) when it reported that bachelor
degree-seeking students who delayed entry into postsecondary education were less
likely to have completed or still be working towards a bachelor’s degree after 5 years
than those who did not delay.
Gender and Persistence
The majority of the literature has indicated no appreciable difference between
the dropout rates of men and women (Baker & Velez, 1996; NCES, 1997). In a review
of data from another national database, gender did not have an effect on system per
sistence in 4-year institutions (Porter, 1990). However, in another study (Kroc,
Howard, Hull, & Woodard, 1997), females were found to have slightly higher gradua
tion rates than males. Tinto (1993) reported that females had lower attrition rates
when both 2-year and 4-year programs were taken into account.
Astin (1975) found that women were more likely to complete their bachelor’s
degree within 4 years, although he indicated that this might be at least partially attribut
able to the propensity for men to be attracted to longer degree programs (e.g., engi
neering). That men may tend to stop out and return later also may prolong the length
to degree completion (Astin, 1975).
Although there seems to be little difference in overall attrition rates of men and
women, the reasons for their departures may tend to be different. Studies have indi-
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
cated that men are more likely than women to drop out because of dissatisfaction and
poor grades, whereas women are more likely to stop attending college for family re
sponsibilities, marriage, and pregnancy (Astin, 1975; Cope & Hannah, 1975; Summer
skill, 1962).
Academic Ability and Persistence
It seems reasonable that individuals with higher ability would be more likely to
be successful in their pursuit of higher education. This contention seems to be sup
ported fairly well in the literature. Student ability as measured by both high school
GPA and college aptitude test scores (e.g., Scholastic Aptitude Test [SAT], ACT]) has
been shown to be a strong predictor of college completion (Astin, 1975; Summerskill,
1962; Tinto, 1993). Astin (1993) found from longitudinal studies that high school GPA
was “ the single strongest predictor” and that SAT scores were a “relatively strong pre
dictor” of college completion. Cope and Hannah (1975) concurred that dropouts usu
ally have below-average scores on their aptitude tests. Porter (1990) reported from
his analysis of the NCES’s High School and Beyond study that ability test scores for
students attending 4-year institutions were positively correlated with degree comple
tion rates.
Goal Aspirations and Persistence
Educational aspirations at the time of entry have been shown to have a positive
impact on educational attainment in a number of studies (Anderson, 1981; Pascarella
& Terenzini, 1991; Peng & Fetters, 1978; Tinto, 1993). Astin (1975) found that, even in
studies excluding those students who had not aspired to obtain at least a bachelor’s
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
degree, the level of educational aspirations was positively associated with persistence.
Those who aspired to the doctorate or to professional degrees at the time of initial ma
triculation into higher education were the least likely to drop out.
According to Tinto (1975, 1993), student educational and occupational goals
have an impact on persistence. Tinto stated that a student’s goal to graduate college
is the most influential factor after ability is taken into account (Tinto, 1975). In support
of Tinto’s theory, various research studies have provided fairly strong support for the
proposition that initial commitment to the goal of graduation positively affects subse
quent goal commitment, which in turn increases the likelihood of student persistence in
college (Braxton et al., 1997; Pascarella & Terenzini, 1980; Stage, 1989). In some
studies, the impact of goals and aspirations on persistence was determined via its
mediation through the behavioral intention to persist (Bean, 1982; Cabrera et al.,
1992).
Although from a theoretical point of view the construct of career aspirations
would be expected also to have an effect on persistence behavior, its relationship to
persistence behavior has not been studied to the same extent as that of educational
aspirations. Tinto (1993) pointed out that, when attending college is seen as a pre
requisite to a larger goal such as occupational goals, the motivation to complete the
degree program is enhanced. In other words, if a student motivated by career or finan
cial success perceives that degree completion is required to achieve that valued goal,
then the student would be expected to have an increased tendency toward persistence
behavior. However, Tinto (1993) also pointed out that many students experience
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
uncertainty about their long-term educational and/or occupational goals while attend
ing college.
Enrollment Intensity and Persistence
Most of the research on attrition in higher education has been limited to study
ing full-time students. This emphasis is especially interesting given the fact that part-
time enrollments make up approximately 31% of all enrollments in 4-year institutions
(Tinto, 1993). Research on enrollment intensity, which has been primarily focused on
students attending commuter 4-year and 2-year institutions, has generally supported
the contention that students enrolled part-time, in comparison to those enrolled full
time, were more likely to drop out of college (Bean & Metzner, 1985). From a theoreti
cal point of view, part-time attendance is associated with less involvement and fewer
interactions within the academic community, which, in turn, can result in negative
outcomes such as dropout behavior (Astin, 1996; Pascarella, 1980). According to the
U.S. Department of Education report on the Condition of Education 1997, full-time en
rollment is correlated with higher rates of student persistence and attainment in post
secondary education (NCES, 1997).
Pull Factors and Persistence
Factors external to the college can have an impact on persistence behavior.
Conceptually, factors that “pull” students away from participating fully in the academic
and social communities of the institution will potentially have a negative effect on stu
dent outcomes and persistence. Two such “pull” factors that relate to this study
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
include the hours a student may work at a job and the extent to which a student may
have family responsibilities.
There has been fairly strong support in the literature that full-time work, espe
cially when it is off campus, has a negative impact on degree completion (Astin, 1975,
1993; NCES, 1997; Pascarella & Terenzini, 1991; Tinto, 1993). The effect of part-time
work on attrition has not been studied to the same extent as the effect of full-time em
ployment on attrition. Nevertheless, part-time employment, especially when that em
ployment is on campus, has not been generally linked to persistence problems (Pas
carella & Terenzini, 1991). Astin (1975) found that, if part-time employment while at
tending college was continuous from the freshman year, it was associated with de
creased attrition. Astin later reported that working off campus part-time was negatively
related to persistence (Astin, 1993).
Another potential “pull factor,” having family responsibilities, has also been as
sociated with increased withdrawal behavior. As discussed by Tinto (1993), most stu
dents who have significant family responsibilities are likely to be commuters who
spend less time becoming involved and integrated into the social and academic com
munities of campus. A negative relationship between family responsibilities and per
sistence has been supported in the literature (Bean, 1990; Bean & Metzner, 1985;
Nora, Cabrera, Hagedorn, & Pascarella, 1996; Tinto, 1993).
Student Involvement/Integration and
Persistence
Student involvement and integration within the academic and social aspects of
the college experience are central to the theoretical models established by Tinto,
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Bean, and Astin, as previously discussed. Although there are no consistent standards
for using the terminology integration versus involvement, in the broader sense, they
can be considered to represent the same or at least very similar constructs. To elabo
rate, Tinto, and those who have operationalized Tinto’s theory, usually have used the
term integration to refer to a state of being that occurs for the student via his/her inter
action and involvement in the academic and social communities of the campus. It is
typically assessed through measures of student’s perceptions and academic achieve
ment (the latter of which is a separate construct in this study). Astin has used the term
involvement and in doing so has focused on the behavioral measures that quantify a
student’s involvement activities including interaction with faculty and peers. Bean has
appeared to use the terms integration, involvement, and interactions when discussing
his theories on student persistence. However, as noted previously, he has separated
behavioral and attitudinal measures, and has kept academic achievement (GPA) as a
separate construct. Although involvement and integration are discussed here as a
general concept, the term involvement was used as a construct label in the present
study because it was operationalized using behavioral measures (see Chapter 3).
Many studies have lent support to the contention that greater degrees of inte
gration and involvement in the academic and social aspects of the campus community
can increase the likelihood of student persistence (Astin, 1975; Bean, 1990; Berger &
Milem, 1999; Cabrera et al., 1992; Lamport, 1993; Pascarella, 1980; Tinto, 1993). The
literature has also supported the idea that integration into the college community oc
curs to a large extent through formal and informal interactions with fellow students and
faculty (Baker & Velez, 1996; Pascarella & Terenzini, 1991).
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
However, some studies have not shown a relationship between integration and
persistence (Braxton et al., 1997; Ruddock, Hanson, & Moss, 1999). In the meta
analysis of studies testing Tinto’s model, Braxton et al. (1997) found that the overall
support for the relationship between academic and social integration and persistence
is inconsistent and not very strong. They also indicated that the varying results could
be partly attributed to a lack of consistency in how the integration constructs in the var
ious studies were operationalized. Pascarella and Terenzini (1991) pointed out that
social and academic integration might play different roles for different student popula
tions (e.g., residential vs. commuters, traditional vs. nontraditional). They also re
ported that it might have a greater impact on student outcomes during the first year of
college life than during later college years.
As virtually all studies concerning integration and involvement have been done
at the institutional level, they have related more directly to institutional and to system
persistence. However, in one system-wide longitudinal study, Stage and Rushin
(1993) were surprised to find that student involvement factors had a slightly negative
relationship with system persistence. Although the theoretical models are fairly con
sistent in postulating that involvement and integration will have a direct or indirect im
pact on persistence, the findings reported in the empirical literature regarding this con
tention have been mixed.
Scholastic Performance and Persistence
The theoretical longitudinal model posited by Bean (1980, 1983, 1990) posits
scholastic performance to have direct and indirect effects on dropout behavior. A
number of studies can be cited that support the proposition that undergraduate grades
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
are one of the most valid predictors of persistence (Astin, 1993; Bean, 1990; Cabrera,
Nora, & Castaneda, 1993; Murtaugh et al., 1999; Nora et al., 1996; Pascarella & Ter
enzini, 1991; Peng & Fetters, 1978; Summerskill, 1962). Astin (1993) also pointed out
that the direction of causation could also be in the reverse direction, whereby, if a stu
dent decides that he or she will be departing from college (for whatever reason), then
he/she might start to lose interest in studying and consequently earn poorer grades.
Satisfaction and Persistence
Studies have shown a relationship between satisfaction and persistence (Astin,
1993; Hatcher, Kryter, Prus, & Fitzgerald, 1992; Mohr, Eiche, &Sedlacek, 1998; Starr,
Betz, & Menne, 1972). Astin found satisfaction levels to be related to the number of
college years completed, and Starr et al. demonstrated that the level of total satisfac
tion experienced by students, plus their satisfaction levels for compensation (i.e.,
grades), recognition from faculty and peers, and quality of education, were all signifi
cantly higher for those who did not drop out than for those who departed either for aca
demic reasons (e.g., academic dismissal) or nonacademic reasons (e.g., voluntary
departure). Similarly, when Mohr et al. studied attrition in college seniors, they discov
ered that nonretention was significantly related to dissatisfaction with academic guid
ance, access to school-related information, and quality of education. Hatcher et al.
found that student satisfaction levels were negatively correlated with subsequent per
sistence behavior. The assertion that there is an association between student sat
isfaction and retention suggests that student satisfaction may prove to be an important
intermediate outcome on which institutions may focus in their efforts to lower attrition
rates (Astin, 1993; Bean, 1980; Sanders & Burton, 1996).
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The hypothesized eclectic model presented earlier in this chapter (Figure 1)
includes an indirect relationship between satisfaction and persistence that is mediated
by transfer status. Although this exact relationship does not appear to have been
studied previously, Astin (1993) found a positive correlation between “left school or
transferred" and satisfaction measures.
Transfer Status and Persistence
Approximately 15% of all students attending 4-year institutions have trans
ferred at least once during their first 2 years (Pascarella & Terenzini, 1991). Despite
this fact, the persistence of these transfer students has not been studied extensively
(Tinto, 1993). Even though the students at the 4-year institutions who transfer are pre
sumably seeking an institution which will better meet their needs, most of the literature
points to the fact that even horizontal transfer has a negative impact on persistence
(Astin, 1975; Pascarella & Terenzini, 1991). However, Carroll (1989) found that, when
horizontal transfers did not break enrollment continuity (i.e., did not leave the persis
tence path), their chances of leaving the persistence path within 6 years decreased.
Research Questions and Hypotheses
Within the context of the purposes of this study, the following research ques
tions were posed, sometimes in association with hypotheses appropriate for the
inquiry or when rationales have been presented in the previous review of literature.
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Research Question 1
Was the eclectic model posited to explain system persistence consistent with
the observed data as indicated by acceptable levels of goodness-of-fit indices and,
therefore, plausible in accounting for the postulated relationships between variables?
Hypothesis 1. The eclectic model posited to explain system persistence would
be consistent with the observed data (i.e., have adequate goodness-of-fit indices).
Research Question 2
If the hypothesized eclectic model did not have acceptable levels of goodness-
of-fit indices, could a modified model be developed which would be consistent with the
data and would support the plausibility of the postulated relationships between vari
ables?
Hypothesis 2. If Hypothesis 1 was not supported, a modified model could be
developed which would be consistent with the observed data (i.e., have adequate
goodness-of-fit indices).
Research Question 3
If a model with acceptable levels of fit indices could be developed, what would
be the direct and indirect effects of each of the following variables on system persis
tence: SES, age, gender, academic ability (SAT scores), initial educational aspira
tions, subsequent educational aspirations, career/financial success and security,
family responsibilities, involvement at initial institution, scholastic performance (GPA),
satisfaction at initial institution, and transfer status?
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Research Question 4
What was the degree of bivariate relationship between system persistence in
pursuing a bachelor’s degree within a 5-year period in 4-year institutions and each of
the following predictor variables: SES, age, gender, academic ability, goal aspirations,
student enrollment intensity, “pull” factors (i.e., hours worked while enrolled and family
responsibilities), involvement at initial institution, scholastic performance (GPA), satis
faction, and transfer status?
Hypothesis 4a. Students who persisted, in comparison to those who did not
persist, would come from families with higher SES levels.
Hypothesis 4b. Students who persisted, in comparison to those who did not
persist, would be younger in age.
Hypothesis 4c. There would be no relationship between students who per
sisted and those who did not persist with respect to their gender identity.
Hypothesis 4d. Students who persisted, in comparison to those who did not
persist, would have higher academic ability.
Hypothesis 4e. Students who persisted, in comparison to those who did not
persist, would have higher educational aspirations.
Hypothesis 4f. Students who persisted, in comparison to those who did not
persist, would have attributed higher levels of importance to future career and financial
success and security.
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hypothesis 4q. Students who persisted, in comparison to those who did not
persist, would have a greater percentage of full-time enrollment.
Hypothesis 4h. Students who persisted, in comparison to those who did not
persist, would have worked fewer hours per week while attending college.
Hypothesis 4i. Students who persisted, in comparison to those who did not
persist, would have fewer family responsibilities.
Hypothesis 4i. Students who persisted, in comparison to those who did not
persist, would have higher levels of involvement at their initial institution.
Hypothesis 4k. Students who persisted, in comparison to those who did not
persist, would have higher GPAs.
Hypothesis 41. Students who persisted, in comparison to those who did not
persist, would have experienced a higher level of satisfaction at their initial institution
Hypothesis 4m. Students who persisted, in comparison to those who did not
persist, would have experienced a higher level of satisfaction at their last institution.
Hypothesis 4n. Students who persisted, in comparison to those who did not
persist, would be less likely to have transferred.
Research Question 5
In the context of the variables in the previous research question that showed
predictive promise, how effective was a multiple discriminant analysis of scores on a
composite of input variables in predicting those who would or would not persist for a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
period of 5 years in pursuit of a bachelor’s degree at 4-year postsecondary institu
tions?
Hypothesis 5. Multiple discriminant analysis would yield a composite of predic
tor variables that would correctly classify students as system persisters or system
nonpersisters at a level greater than chance.
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 3
METHODS AND PROCEDURES
This chapter provides information regarding the methods and procedures uti
lized for this study. The six major topics discussed in this chapter are the (a) research
sample, (b) procedures used in data collection, (c) variables and their measures,
(d) methods employed in data analysis, (e) methodological assumptions, and (f) limita
tions.
Research Sample
This study utilized the BPS longitudinal database from the NCES (1994a,
1996), which in total follows for a 5-year period a cohort of approximately 7,500 stu
dents who began their postsecondary education sometime between July 1, 1989 and
June 30, 1990. The design involved a multistage probability sampling of 70,000 stu
dents taken from 1,533 institutions in 121 geographical areas across the United States
of America (NCES, 1996). This process yielded a total of 11,700 students from 1,092
institutions as the initial BPS sample. The sample was further reduced to approxi
mately 7,900 students with the first follow-p, and to approximately 7,200 students with
the second follow-up, primarily as a result of ineligibility (e.g., student was not a first
time beginning postsecondary student), inability to locate/contact the student, or a
decision by the student not to participate in the study. Participation in the study was
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
strictly voluntary. Students included in the cohort were first-time postsecondary stu
dents regardless of when they had completed high school.
For the purposes of answering the research questions in this study, only those
students who had begun their postsecondary education at 4-year institutions and who
had had as their initial goal the completion of a bachelor’s degree were utilized. This
constraint provided an initial sample of 3,408 students for the study. The sample size
was further reduced in order to eliminate ambiguities in the data that prevented the
determination of (a) which institution had been the student’s initial institution, and
(b) what had been the student’s final persistence status. These additional constraints
made available a random sample of 3,278 students as the final research sample for
this investigation. Of the 548 four-year institutions represented in the initial sample,
247 were public and 301 were private.
The 3,278 bachelor-seeking students included in this study were all beginning
their postsecondary education for the first time, regardless of when they had com
pleted high school. In the group, 48.7% were males and 51.3% were females. Their
ages when they began were between 16 and 56 years, with a mean of 18.33 years.
The ethnic diversity of the sample included: White, non-Hispanic (82.0%); Black, non-
Hispanic (7.4%); Hispanic (5.1%); Asian/Pacific Islander (4.9%); Native American/
Alaskan (0.3%); and missing (0.2%).
The areas of study for the students in this sample varied. The majors of stu
dents during their longest enrollment spell throughout the 5-year period of the study
were as follows: Humanities (14.0%), Social/Behavioral Sciences (18.0%), Life Sci
ences (7.1%), Physical Sciences (2.3%), Mathematics (1.1%), Computer/Information
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Technology (0.2%), Engineering (8.8%), Education (8.6%), Business and Manage
ment (20.6%), Health (6.3%), Vocational/Technical (2.1%), other (7.3%), and missing
(3.6%).
This study focused on bachelor-seeking students attending 4-year institutions.
Within the sample of 3,278 students used for this study, 41.1 % initially attended a
public 4-year institution and 58.9% initially attended a private 4-year institution.
Data Collection
The data on the 3,278 students were initially obtained both from school records
and from student interviews. The design of the study involved tracing the students to
their current locations and then conducting computer-assisted telephone interviews for
each of two follow-up surveys. The reported overall response rate for sample mem
bers was approximately 87.1 % for the first follow-up and 88.2% for the second follow-
up (NCES, 1996). The average length of time required to complete a telephone inter
view was approximately 35 minutes.
The procedures and computer-assisted survey instruments utilized for the pur
pose of this study were developed by the National Center for Educational Statistics
(NCES, 1996). As stated earlier, the initial base-year data were collected during the
1989-90 academic year. The first follow-up was conducted during the spring of 1992
(Year 3) and the second follow-up took place during the spring and summer of 1994
(Year 5). The initial and follow-up surveys were carried out under the guidelines estab
lished by the NCES, U.S. Department of Education. The procedures and methodolo
gies were all reviewed by a panel of both federal and non-federal academicians and
researchers who were known for their expertise in this field of study and type of
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
research (NCES, 1996). A committee for the Protection of Human Subjects made
sure that the study complied with “applicable regulations regarding informed consent,
confidentiality, and protection of privacy. This committee independently reviewed the
study design, instruments, and data collection/processing procedures to ensure that
sample members’ rights were fully protected” (NCES, 1996, p. 19).
The instrumentation and procedures utilized for data collection were field
tested extensively and subsequently revised as necessary (NCES, 1992, 1994a,
1994b, 1996). In addition, guidelines were established and followed for the selection
and training of the staff selected for the collection of data. Each component of the
study was evaluated and monitored for quality control, including the systems for lo
cating and interviewing students, data quality (including reliability re-interviews), on
line coding, and filing systems. Both formative and summative evaluation methodolo
gies were employed (NCES, 1996).
Field testing of the instrument also examined possible effects resulting from the
order in which the survey questions were asked (i.e., order effects). An analysis of
random starting points revealed that, in most cases, any effects observed because of
differing orders were not significantly different from chance expectations. In some
cases, starting points had to be slightly modified from random selection to adjust for
minor ordering effects.
Variables and Their Measures
This section includes a description of the variables and the way in which they
were measured. Unless otherwise designated, continuous measures were coded so
that higher values reflected higher or greater levels of what was actually being
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
measured. Percentages reported indicate a percentage of the total sample of 3,278
students.
Background and Demographic Variables
Gender. This was a dichotomous variable coded 1 for males (48.7%) and 2 for
females (51.3%).
Age. This was a continuous variable of the subjects’ ages when beginning the
study, which ranged from 16 to 56 years, with a mean of 18.33 years.
SES. This was a continuous composite variable available in the BPS data
base derived by NCES that was derived by combining (a) the highest level of educa
tion of either parent, (b) a set of 1 items (e.g., dishwasher, VCR) in the home that were
used to create a “ things" index, and (c) dependents’ family income.
Ability. The composite scores from ACT and SAT aptitude tests which were
available for a portion of the students as a measure of scholastic ability. From the total
sample of the 3,278 students, 24.7% (n = 811) had ACT scores available and 46.2% (n
= 1,515) had SAT scores available. The two groups were not mutually exclusive (i.e.,
165 students had scores available for both tests). Within the sample, 34.1% (n =
1,117) had missing values.
Goal Aspiration Variables
Initial educational aspiration— Year 1. This was an ordinal-scaled variable
obtained by asking students during their first year (academic year [AY] 1989-90) what
was the highest level of education that they expected to achieve. Of the 3,278
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
students, 30.3% (n = 992) indicated “Bachelor’s degree,” 44.7% (n = 1,464) specified
“Master’s degree," and 23.6% (n = 775) expected that a “PhD/professional degree”
would be their highest level of education. As this study included only those students
who had at least a bachelor’s degree as their initial educational goal, there were no
responses lower than bachelor’s degree on this measure. Responses were missing
for 1.4% (n = 47) of the sample.
Subsequent educational aspiration— Year 3. This was an ordinal-scaled vari
able obtained by asking students at the end of Year 3 (AY 1991-92) what was the
highest level of education that they expected to achieve. Of the 3,278 students, 1.6%
{n = 51) indicated less than a Bachelor’s degree, 21.7% (n = 712) stated “Bachelor’s
degree,” 38.8% (n = 1,272) specified “Master’s degree,” and 24.4% (n = 799) expected
that a “PhD/professional degree” would be their highest level of education. Responses
were missing for 13.5% (n = 444) of the sample. It is interesting to note that although
students who had educational aspirations lower than a bachelor’s degree were initially
filtered out of the sample, the educational goals of some of those students (1.6%)
decreased to less than a bachelor’s degree by the end of Year 3.
Career and financial success/security. This composite variable was derived
from the mean score of responses to three questions asked during the initial year of
the study. Students designated whether it was (a) very important, (b) somewhat
important, or (c) not important with respect to each of the following measures:
1. Be very well off financially,
2. Have good income from the work they would do most of their life, and
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3. Have job security and permanence from the work they would do most of
their life.
The internal consistency of the composite variable utilizing Cronbach’s alpha
was .70. Of the 3,278 students in the sample, 99.1% (n = 3,248) had valid responses,
and 0.9 % (n = 30) had missing data.
Enrollment Intensity and “Pull-Factors"
Variables
Percentage of months enrolled full-time. This was a continuous variable avail
able in the BPS database derived by NCES reflecting the percentage of months that a
student attended full-time, as opposed to part-time, during the academic year while
enrolled in postsecondary education or until first attainment. Information to compute
this variable was taken from monthly enrollment records.
Average hours worked while enrolled. This computed continuous variable re
flected the average hours a student worked per week throughout the first 3 years while
enrolled. Work hours were included only for those months when the student was en
rolled for at least part of the month.
Family responsibilities. This was a composite variable derived from the mean
of the following three measures taken during Year 3 (first follow-up): (a) number of
children, (b) number of individuals besides themselves for whom they had financial
responsibility, and (c) number of individuals besides themselves for whom they had
caretaker or other time-commitment responsibilities. Responses were coded on a 7-
point scale ranging from 0 to 6+. The internal consistency of the scaled variable score
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
was .80 as measured by Cronbach’s alpha. Of the 3,278 students in the sample,
86.8% (n = 2,845) had valid responses and 13.2 % (n = 433) had missing data.
Involvement Variable
Involvement at initial institution. This composite variable was derived from the
mean of six questions measuring the level of the students' involvement in various aca
demic and social activities. These measures of involvement, which were taken at the
end of Year 3 (the first follow-up) ascertained the students’ level of involvement at the
initial institution from the beginning of the study through Year 3. Students’ participation
in these activities is reflective of their involvement in the academic and social aspects
of the college environment.
• The six behavioral measures were how much they (a) talked to faculty outside
of class, (b) socialized with faculty/advisor, (c) participated in study groups, (d) went
places with friends from school, (e) participated in school clubs, and (f) attended con
ventions or academic lectures. Questions were coded as follows: 4 = often per term,
3 = several times per term, 2 = once per term, and 1 = never. The internal consistency
of the composite variable score utilizing Cronbach's alpha was .64. Of the 3,278 stu
dents in the sample, 87.0% (n = 2,853) had valid responses and 13.0 % (n = 425) had
missing data.
Scholastic Performance Variable
Cumulative GPA 1989-1992. Students were asked during the first follow-up
(Year 3) what their overall grades were since they had begun postsecondary educa
tion. Responses were coded as follows: 1 = less than Cs, 2 = mostly Cs, 3 = Bs and
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Cs, 4 = mostly Bs, 5 = As and Bs, and 6 = mostly As. Of the 3,278 students in the
sample, 86.7% (n = 2,842) had valid responses, and 13.3% (n = 436) had missing
data.
Satisfaction Variables
Satisfaction at initial institution (SATSIFYN). This was a continuous variable
computed by the NCES that measured the level of satisfaction expressed by students
at their initial institution. The measure used the number of aspects with which a stu
dent reported having been satisfied at his or her initial institution to determine the level
of satisfaction. The more aspects for which the student reported having been satis
fied, the higher the satisfaction level. These aspects (or dimensions) included the stu
dent’s satisfaction with teaching ability, cost, intellectual growth, prestige, and social
life. This variable was coded as follows: 5 = satisfied with five aspects, 4 = satisfied
with four aspects, 3 = satisfied with three aspects, 2 = satisfied with two aspects, 1 =
satisfied with one aspect, and 0 = satisfied with no aspects. Of the 3,278 students in
the sample, 86.9% (n = 2,847) had valid responses for this measure and 13.1 % (n =
431) had missing data.
Satisfaction at last institution. This composite variable is very similar to the
variable SATISFYN just discussed, except that it measured student satisfaction at the
last institution attended. This variable was computed by NCES to take into account
reported satisfaction levels at Year 5 (or earlier if the student dropped out from the
persistence path) at wherever the last institution might have been. The more aspects
with which the student reported having been satisfied, the higher the satisfaction level.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
These aspects (or dimensions) included the student’s satisfaction with teaching ability,
cost, intellectual growth, prestige, and social life. This variable was coded as follows:
5 = satisfied with five aspects, 4 = satisfied with four aspects, 3 = satisfied with three
aspects, 2 = satisfied with two aspects, 1 = satisfied with one aspect, and 0 = satisfied
with no aspects. Of the 3,278 students in the sample, 89.0% (n = 2,918) had valid re
sponses for this measure and 11.0 % (n = 360) had missing data.
Transfer Variables
Transfer status through first degree. Students were coded as to whether they
had or had not transferred to another institution before obtaining their first degree.
Transfer status was reported whether the student had attended postsecondary educa
tion continuously or noncontinuously (i.e., were stopouts). In addition, both horizontal
and reverse transfers were included as transfers in this variable. The variable was
coded as follows: 1 = transferred and 2 = did not transfer. Of the 3,278 subjects within
the sample, 73.9% (n = 2421) did not transfer and 26.1 % (n = 857) had transfer status.
There were no missing data for this variable.
Type of transfer before first attainment. This categorical variable that defined
the type of transfer experienced by students was utilized in this study for descriptive
statistics. Of the 857 students who had been designated as having transferred (26.1%
of the sample), approximately two thirds (n = 561; 17.1% of the sample) had trans
ferred horizontally and approximately one third (n = 296; 9.0% of the sample) were
students who had transferred in a reverse direction (i.e., from a 4-year institution to a
less-than-4-year institution). There were no missing data for this variable.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Persistence Variables
Overall persistence categories (PERCATGY). This categorical variable was
primarily utilized in this study for purposes of descriptive statistics. It indicated persis
tence status of the student at the time of the second follow-up (Year 5). Categories
were coded as follows: 1 = internal (institutional) persister, 2 = transfer persister, and
3 = nonpersister in a 4-year institution. Although transfer persisters might include all
types of transfers, they were considered persisters only if, at the time of the second
follow-up, they had persisted in a 4-year institution. Nonpersisters could be either
internal and transfer nonpersisters. Enrollment patterns (continuous or noncontinu-
ous) were not taken into consideration in this variable. There were no missing data for
this variable.
Overall system persistence. This was a dichotomous variable derived from
PERCATGY that summarized the status of students at the end of 5 years with respect
to whether they had or had not persisted in the 4-year postsecondary education sys
tem. Students were coded as system persisters if they either had attained a bache
lor’s degree or were still enrolled in a 4-year institution at the time of the second follow-
up (Year 5). This variable coded students as system persisters or system nonper
sisters regardless of their transfer status or enrollment patterns (continuous or noncon-
tinuous). This variable was coded as follows: 1 = persisted in 4-year system and 0 =
did not persist in 4-year system. There were no missing data for this variable.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Data Analysis
The statistical techniques used for analyzing the data include descriptive sta
tistics (frequencies and distributions, plus means and standard deviations for selected
variables). To facilitate answering Research Questions 1-3 and corresponding hypoth
eses, a path analysis utilizing structural equation modeling technique was utilized. To
answer Research Question 4 and corresponding hypotheses, a variety of techniques
was employed, depending on the nature of the variables and the hypothesis being
answered. The statistical techniques were appropriate for a dichotomous dependent
variable, as the criterion variable was stated in such a way as to determine whether a
student successfully “persisted” or “did not persist” toward obtaining a bachelor’s
degree in a 4-year institution. In all cases, the statistical procedures were chosen
based on whether the variables were continuous or categorical. The bivariate sta
tistical techniques used to answer the research questions included t tests and cross
tabulations using a chi-square statistic.
Multiple discriminant analysis was used to answer Research Question 5. This
technique was utilized to determine whether a given set of input variables could suc
cessfully predict group membership, which in this case referred to 5-year system per
sisters versus 5-year system nonpersisters. A step-wise discriminant analysis tech
nique was used to obtain a more parsimonious set of variables which could then be
used to predict system persistence. A final application of the discriminant analysis
technique was conducted on a randomly selected portion (approximately 50%) of the
sample. The classification function derived from the selected portion was applied to
the remaining “unselected” cases in order to cross-validate the results.
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Methodological Assumptions
The following methodological assumptions were evident:
1. The sample selected was representative of the population.
2. Information obtained about the characteristics and behaviors of students
during the period beginning fall 1989 through spring 1994 was accurate.
3. The statistical approaches employed in this investigation were appropriate
and relevant to answering the research questions and hypotheses.
4. The participants in the study understood the intent of the questions that
were asked.
5. The participants answered the questions truthfully.
Limitations
The following limitations were present in this investigation:
1. The study was limited to those students who had agreed to participate.
2. During the 5-year period of this longitudinal study, some of the students
stopped participating and/or no longer could be found for the follow-up surveys. This
circumstance could impact the final recorded percentages of persisters versus nonper
sisters and, accordingly, could bias the results.
3. The extent to which any one of the methodological assumptions was not
met would constitute a limitation.
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4
P R E S E N TA TIO N OF R ESULTS
The results of the statistical analysis are presented in this chapter. The find
ings are reported within the framework of the research questions and corresponding
hypotheses set forth in Chapter 2.
Analysis of Findings
Descriptive Statistics Related to
5-vear System Persistence
The frequencies with which the 3,278 students in the sample achieved or did
not achieve 5-year system persistence are as follows: 76% (n = 2,492) were system
persisters and 24% (n = 786) was system nonpersisters. As mentioned in Chapter 1,
“system persistence” refers to students who have either completed or are still enrolled
in the educational system, regardless of whether they transferred from their initial in
stitution. The educational system referred to in this study is that of baccalaureate de
gree programs in 4-year institutions.
It is interesting to note that, of the 2,492 students who achieved 5-year system
persistence, 80.7% (n = 2021) did so at their initial institution and 19.3% (/? = 480) per
sisted after transferring to another 4-year institution. For comparison purposes, this
information equates to an institutional persistence rate of 61.7%, which is understand
ably lower than the system persistence rate of 76.0%, as was reported previously.
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Adequacy of the Goodness-of-Fit Indices
of the Eclectic Model to Explain System
Persistence (Research Question 1)
The proposed eclectic model was tested using the robust methodology of the
EQS structural equation modeling program (Bentler & Wu, 1995) to determine whether
the model was consistent with the observed data. The results for the corresponding
hypothesis were as follows:
Hypothesis 1. The eclectic model posited to explain system persistence would
be consistent with the observed data (i.e., have adequate goodness-of-fit indices).
Findings. The chi-square statistic was significant, x2(60) = 468.53, p < .001.
Although a promising goodness-of-fit indication occurs when the chi-square value is
not statistically significant, for very large sample sizes even a small chi-square value
can be statistically significant. From a practical standpoint, therefore, a statistically
significant chi-square for a large sample could be misleading. An alternative method
utilizing the ratio of chi-square to the degrees of freedom was employed. If this ratio is
less than 3, a good-fitting model is said to be indicated (Kline, 1998). In this case,
when the proposed eclectic model was tested, the ratio of chi-square to the degrees of
freedom was 7.8, a value which implied that the proposed model did not adequately fit
the observed data.
Other indices of fit used to determine whether the proposed model was consis
tent with the data confirmed that the model did not fit the observed data well. The
Comparative Fit Index (CFI) was .635, which is below the acceptable standard of .90
or greater for indicating an adequate fit of the model to the data (Byrne, 1994). The
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Root-Mean-Square Error of Approximation (RMSEA) was .077, which is above a
conventionally acceptable standard of .05 or less for this index (Hu & Bentler, 1999).
The findings did not support the hypothesis.
Adequacy of the Goodness-of-Fit Indices
of a Modified Model to Explain System
Persistence (Research Question 2)
As Hypothesis 1 was not supported, a systematic method of model modifica
tion was used in order to achieve a model which would be consistent with the observed
data. This task was accomplished, first, by adding parameters identified by the Lag
range Multiplier Test as parameters to add to the model in order to improve the mod
el’s fit and, second, by removing parameters suggested by the Wald Test as paths that
did not significantly contribute to the model’s fit. These parameters were identified
through a methodology available within the computer program that is applied automati
cally for path selections.
Of the seven paths that were added in order to improve significantly the mod
el’s fit, three could be strongly supported by related literature. In particular, a justifica
tion based on previous research could be made for the addition of paths from both
gender and academic ability to scholastic performance (Astin, 1993; Pascarella & Ter-
enzini, 1991) and from enrollment intensity to persistence (Bean & Metzner, 1985).
On the other hand, the paths from both involvement and academic ability to transfer
status were added more for exploratory reasons in order to ascertain possible mediat
ing effects that transfer status might have in explaining system persistence. Because
transfer status has not been previously included in causal models posited to explain
persistence, it seemed prudent to add these significant parameters. Although no
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
literature could be found that could support or refute the addition of paths from either
academic ability or scholastic performance (GPA) to subsequent educational aspira
tions, it seemed very reasonable that both of these relationships might exist and, if so,
their inclusion in the model could potentially help to explain system persistence.
Of the 12 paths that were removed from the model as suggested by the Wald
test, most (9) were from exogenous variables, while only 3 affected endogenous path
ways. These parameters were removed for reasons of parsimony, as none of them
contributed significantly to the model’s fit.
The results for the corresponding hypothesis were as follows:
Hypothesis 2. If Hypothesis 1 was not supported, a modified model could be
developed which would be consistent with the observed data (i.e., have adequate
goodness-of-fit indices).
Findings. Ten iterations of the eclectic model yielded a modified version that
was consistent with the observed data as substantiated by acceptable goodness-of-fit
indices. Although the chi-square was significant, x2(64) = 136.29, p < .001, the ratio of
chi square to degrees of freedom of 2.1 indicated that the modified model was consis
tent with the observed data. This outcome was confirmed by the CFI, which was .935,
and the RMSEA, which was .035. The findings supported the hypothesis. The modi
fied eclectic model with the empirically demonstrated links between variables is illus
trated in Figure 2.
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
.095*
- significant paths
-.062'
-.038 = N.S. paths
= added paths (significant)
Subsequent
Educational
Aspirations
it = .318
Initial
Educational
Aspirations
.164' .451
SES
Transfer
Status
r*=.103
Career and
Financial Goals
System
Persistence
r*= .163
Enrollment Intensity
Percentage Enrolled
Full Time
M L
Involvement
Satisfaction
fi = .005
.221
col
fi = .080
Gender
.098'
- . 110'
Hours
worked
Scholastic
Performance
(GPA)
r* = .170
Academ ic
Ability
(SAT)
Family
Responsibilities
.210’
.382'
-.132'
-.047
cn
O)
Figure 2. Modified eclectic model of system persistence. GPA = grade point average; SAT = Scholastic
Aptitude Test. Asterisk indicates significant paths.
Direct and indirect Effects of Each of the
Predictor Variables on System Persistence
(Research Question 3)
The degree of relationships identified between the variables in the modified
eclectic model is included in Figure 2. The significance of the correlations (exogenous
variables only) and of the standardized path coefficients was determined by a z-test
statistic at the 95% confidence level. Significant paths are indicated by an asterisk (*).
The percentage of variance explained (r2) for some of the intermediary variables and
for the final dependent variable, system persistence, is also indicated on the diagram.
A summary of the direct, indirect, and total effects of each variable on 5-year system
persistence is presented in Table 1.
Bivariate Relationship Between Predictor
Variables and 5-vear System Persistence
(Research Question 4)
Bivariate analyses were conducted to determine the difference between per
sisters and nonpersisters in regard to several selected predictor variables, all (tests
were one-tailed to conform to the directional hypotheses proposed. The results for
each of the hypotheses corresponding to this research question are as follows:
Hypothesis 4a. Students who persisted, in comparison to those who did not
persist, would come from families with higher SES levels.
Findings. An independent-samples (test revealed that individuals who did not
achieve 5-year persistence tended to come from lower SES backgrounds (M = 68.11,
SD = 22.69) than those who did achieve persistence (M = 75.85, SD = 20.05). The
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1
The Direct. Indirect, and Total Effects of Input Variables on 5-vear System Persistence
Degree of relationship to system
persistence3
Direct Indirect Total
Variable effects effects effects
Socioeconomic status .095* .008* .103
Age -.062* -.035* -.097'
Gender -.038 .040* .002
SAT score -.047* .161* -.114'
Initial educational aspirations
—
.102* .102’
Career/financial goals
— — —
Percentage enrolled full-time .122* .033* .155’
Hours worked
—
-.017* -.017’
Family responsibilities
-----
-.008 -.008
Involvement at initial institution .098* .053* .151’
Subsequent educational aspirations .164* .012* .176’
Grade point average .170* .020* .190’
Satisfaction at initial institution
—
.019* .019’
Transfer status -.118* ---- -.118’
a Degree of relationships indicated by standardized path coefficients.
*p < .05.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
difference between the means was statistically significant, f(3,276) = 9.13, p < .001.
The findings supported the hypothesis.
Hypothesis 4b. Students who persisted, in comparison to those who did not
persist, would be younger in age.
Findings. An independent-samples t test was conducted to determine whether
the mean age of students who persisted was significantly different from the mean age
of students who did not persist. Students who persisted were younger (M = 18.15, SD
= 2.05) than those who did not persist (M = 18.90, SD = 3.68), with a mean difference
of 0.75 years. As the results were statistically significant, f(3,276) = 7.16, p < .001, the
hypothesis was supported.
Hypothesis 4c. There would be no relationship between students who per
sisted and those who did not persist with respect to their gender identity.
Findings. A two-way contingency table analysis was utilized to ascertain
whether a relationship existed between gender and whether a student persisted or did
not persist for 5 years towards a bachelor’s degree. As indicated in Table 2, there ap
peared to be no significant relationship between gender and persistence behavior in
the sample tested, Pearson x^ l, N = 3,278) = .61, p = .434, Cramer’s \/= .014). The
Cramer’s V statistic, which measures the strength of the relationship between the vari
ables, is similar to the Pearson product-moment correlation. The findings supported
the hypothesis.
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2
The Relationship Between Gender and 5-vear System Persistence
Five-year persistence
Males
Gender
Females Total
n % n % n %
Persisted 1,203 75.4 1,289 76.6 2,492 76.0
Did not persist 393 24.6 394 23.4 786 24.0
Totals 1,595 100.0 1,683 100.0 3,278 100.0
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hypothesis 4d. Students who persisted, in comparison to those who did not
persist, would have higher academic ability.
Findings. The relationship between system persistence and student ability was
tested for those cases where SAT or ACT scores were available. If scores were ob
tainable for both tests, then the student was included in both analyses. An independ-
ent-samples t test showed that for those in the sample who had scores on the aptitude
tests, the students who persisted 5 years had significantly higher scores on the SAT,
f(1,513) = 6.71, p < .001, and on the ACT, f(809) = 5.34, p < .001, than students who
did not persist. Table 3 contains the means and standard deviations of each measure
according to 5-year persistence status. The findings supported the hypothesis.
Table 3
A Comparison of the Means of SAT and ACT Scores in 5-vear Persisters Versus
Nonpersisters
Variable n M SD
SAT score
Persisters 1,211 1,040.0 206.7
Nonpersisters 304 951.0 209.0
ACT score
Persisters 569 21.9 5.3
Nonpersisters 242 19.6 6.0
Hypothesis 4e. Students who persisted, in comparison to those who did not
persist, would have higher educational aspirations.
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Findings. The relationship between the educational aspirations of students and
whether a student persisted or did not persist for 5 years in 4-year postsecondary in
stitutions was analyzed using an independent-samples t test. The analyses revealed
that the educational goals of students in Year 1 was significantly higher, t{3,229) =
7.64, p < .001, for students who persisted for 5 years (M = 6.99, SD = 0.73) than for
those who did not persist (M = 6.76, SD = 0.73). An even greater difference, t(2,832) =
16.72, p < .001, was found between persisters (M = 7.11, SD = 0.72) and nonpersis-
ters (M = 6.49, SD = 1.14) in their subsequent educational goal levels (measured in
Year 3). The results of the analyses supported the hypothesis.
Hypothesis 4f. Students who persisted, in comparison to those who did not
persist, would have attributed higher levels of importance to future career and financial
success and security.
Findings. An independent-samples t test was conducted to determine whether
the level of importance that students attributed to their future career and financial suc
cess and security was related to 5-year system persistence. On a range from 1 (not
important) to 3 (very important), students who did not achieve 5-year persistence had
a mean on this variable of 2.54 (SD = 0.48), while those who did achieve persistence
had a mean of 2.49 (SD = 0.48). The results were not statistically significant because
the direction of the difference between the means was opposite to that hypothesized,
f(3,246) = -2.50, p > .05. The findings did not support the hypothesis.
Hypothesis 4o. Students who persisted, in comparison to those who did not
persist, would have a greater percentage of full-time enrollment.
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Findings. An independent-samples t test showed a relatively strong relation
ship between the percentage of months that students attended school full-time (as
opposed to part-time) and the incidence of 5-year persistence. Students who per
sisted for 5 years had a significantly greater rate of full-time attendance than those
who did not persist, f(3,276) = 19.53, p < .001. When comparing the group means, the
persisters attended full-time on the average of 91.3% of the time (SD = 14.18), where
as the nonpersisters averaged attending full-time only 75.2% of the time (SD = 32.58).
The results from the analysis supported the hypothesis.
Hypothesis 4h. Students who persisted, in comparison to those who did not
persist, would have worked fewer hours per week while attending college.
Findings. The relationship between the average number of hours that students
worked at a job while enrolled and 5-year persistence status was also shown to be
fairly strong. An independent-samples t test demonstrated that the average number of
hours that persisters worked while enrolled was lower (M = 13.3, SD = 9.3) than the
average number of hours that nonpersisters worked while enrolled (M = 17.5, SD =
12.6). As the results were statistically significantly, t(3,276) =-10.19, p < .001, the hy
pothesis was supported.
Hypothesis 4i. Students who persisted, in comparison to those who did not
persist, would have fewer family responsibilities.
Findings. The results from an independent-samples t test showed that stu
dents who were persisters had fewer family responsibilities than nonpersisters.
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Although the mean value of this variable (scaled 0-6) for persisters (M = 0.07, SD =
.32) was statistically lower than for nonpersisters (M = 0.20, SD = .55), f(783) = -5.76,
p < .001, the difference between the means (0.13) was relatively small.
Hypothesis 4i. Students who persisted, in comparison to those who did not
persist, would have higher levels of involvement at their initial institution.
Findings. An independent-samples t test was conducted to evaluate the rela
tionship between student involvement levels at their initial institution and 5-year persis
tence status. The mean level of student involvement at their initial institution was sig
nificantly higher, f(2,851) = 12.81, p < .001, for those who persisted for 5 years (M =
2.91, SD = 0.56) than for those who did not persist (M = 2.57, SD = 0.64). The findings
supported the hypothesis.
Hypothesis 4k. Students who persisted, in comparison to those who did not
persist, would have higher GPAs.
Findings. The results from an independent-samples t test revealed that mean
GPA (scaled 1-6) of system persisters (M = 4.25, SD = 1.07) was significantly higher
than that for nonpersisters (M = 3.51, SD = 1.27), f(2,840) = 14.85, p < .001. The find
ings confirmed the hypothesis.
Hypothesis 4I. Students who persisted, in comparison to those who did not
persist, would have experienced a higher level of satisfaction at their initial institution.
Findings. An independent-samples t test was utilized to determine whether the
average level satisfaction experienced by students at their initial institution was higher
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
for those who were system persisters than for system nonpersisters. The mean level
of satisfaction at the initial institution (scaled 0-5) experienced by persisters (M = 4.11,
SD = 0.91) was significantly higher, t(2,845) = 3.02, p < .01, than the mean level expe
rienced by nonpersisters (M = 3.98, SO = 1.09). Although the difference in the means
was fairly small, the results supported the hypothesis.
Hypothesis 4m. Students who persisted, in comparison to those who did not
persist, would have experienced a higher level of satisfaction at their last institution.
Findings. An independent-samples t test was utilized to determine whether the
average level of satisfaction at the student’s last institution was related to 5-year per
sistence. The results of the analysis was not significant (p = .83) and showed no sig
nificant difference between the mean level of satisfaction at the last institution (scaled
0-6) experienced by persisters (M = 4.04, SD = 1.13) and the mean level experienced
by nonpersisters (M = 4.03, SD = 1.17). The hypothesis was not supported.
Hypothesis 4n. Students who persist in comparison to those who do not
persist would be less likely to have transferred.
Findings. A two-way contingency table analysis was utilized to determine
whether there was a relationship between students’ transfer status and 5-year system
persistence. Referring to Table 4, the percentage of students who persisted 5 years
was considerably less for those who transferred than for those who did not transfer
during the 5-year period. As the results were statistically significant, Pearson x2(1, N =
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3,278) = 254.95, p < .001), the hypothesis was supported. The strength of the rela
tionship as determined by Cramer's V statistic was .279.
Table 4
Relationship Between Transfer Status and 5-vear System Persistence
Five-year
persistence
Transferred Did not transfer Total
n % n % n %
Persisted 480 56.0 2,012 83.1 2,492 76.0
Did not persist 377 44.0 409 16.9 786 24.0
Totals 857 100.0 2,421 100.0 3,278 100.0
Utilizing Multiple Discriminant Analysis to
Predict 5-year System Persistence
(Research Question 5)
Multiple discriminant analysis was utilized to ascertain whether variables used
in this study would effectively predict 5-year system persistence at a level greater than
chance. The initial set of 13 variables included in the discriminant analysis corre
sponded to those used in the bivariate analyses conducted to answer Research Ques
tion 4, except in the case of the variable satisfaction, where only satisfaction at initial
institution, instead of satisfaction at last institution, was included in the discriminant
analysis. This procedure was followed because the constructs potentially overlap and,
between the two, satisfaction at initial institution showed greater promise as a discrimi
nating variable (see results from Research Question 4). The variables used in the
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
analysis were as follows: (a) SES, (b) age when enrollment began, (c) gender,
(d) scholastic ability (SAT score), (e) initial educational aspiration, (f) career and fi
nancial success/security goals, (g) percentage of months enrolled full-time, (h) aver
age hours worked when enrolled, (i) family responsibilities, 0) involvement at initial in
stitution, (k) cumulative GPA, (I) satisfaction at initial institution, and (m) transfer sta
tus.
A correlation matrix of the variables used in the discriminant analysis is pre
sented in Table 5. The results for the hypothesis corresponding to Research Question
5 were as follows:
Hypothesis 5. Utilizing multiple discriminant analysis would yield a composite
of predictor variables that would correctly classify students as system persisters or
system nonpersisters at a level greater than chance.
Findings. Of the original 3,278 cases, 2,017 were deleted because of missing
data. Most of the missing data was a reflection of the missing values found with the
variable SAT score. However, missing data were encountered among the other pre
dictors, too. Of the 1,261 cases retained, 1,025 (81.3%) were 5-year system persist
ers and 236 (18.7%) were nonpersisters. This percentage of persisters versus non
persisters in the retained sample was somewhat different from the distribution found in
the original (i.e., total) sample used in this study. As noted earlier, in the total sample
76% persisted and 24% did not persist. Evaluations of the assumptions of multivariate
normality, linearity, and homogeneity of covariance matrices did not indicate any threat
to the analysis (Tabachnick & Fidell, 1989).
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 5
Correlation Matrix for Variables Used in the Discriminant Analysis (N= 1.261)
Variable 1 2 3 4 5 6 7 8 9 10 11 12 13
1. SES 1.00
2. Age -.13** 1.00
3. Gender -.07* -.07 1.00
4. SAT score .28** -.14** -.18** 1.00
5. Initial educational
aspiration .12** -.05 -.02 .29** 1.00
6. Career/financial
goals - .09** .03
I
o
*
- .24** -.10** 1.00
7. Percentage en
rolled full-time .04 - .23** -.05 .28** .11** -.05 1.00
8. Hours worked - .09** .06* .02 -.15** -.13** .04 -.15** 1.00
9. Family responsi
bilities -.14**
18*.
.04 - .10** -.05 .03 -.11** .07* 1.00
10. Involvement at ini
tial institution .07* -.11** .05 .16** .18** -.00 .25** - .08** - .08** 1.00
11. Cumulative GPA .12** -.05 .15** .35** .20** -.10** .16**
. 14**
-.03 .14** 1.00
12. Satisfaction at ini
tial institution -.02 .01 .05 -.02 .01 -.03 .04 - .09** .01 .07* .09** 1.00
13. Transfer status -.02 -.01 .00 -.18** -.14** .04 -.21** .09** .03 -.21** -.11** -.18** 1.00
Note. Listwise deletion. SES = socioeconomic status; GPA = grade point average.
*p < .01. **p < .05.
o >
00
The Wilks’s Lambda for the discriminant function was significant, A = .806,
X^l 3, N = 1,025) = 269.99, p < .001. This result indicated that the predictor variables
differentiated significantly between the two groups (i.e., system persisters versus sys
tem nonpersisters). The resulting Eigenvalue was .241, and the canonical correlation
was .440. The canonical correlation, which measures the degree of relationship be
tween the optimally weighted composite of discriminant function scores and the group
membership (persisters versus nonpersisters), is equivalent to the eta from a one-way
analysis of variance. Squaring the canonical correlation corresponds to the eta
squared for the discriminant function (Klecka, 1980; Norusis, 1990). Consequently,
.4402 or 19.4% of the total variance between the two groups could be explained by this
set of predictor variables.
Table 6 presents the standardized discriminant coefficients and the structure
coefficients for each variable used in the discriminant analysis. The standardized co
efficients reflect the “multipliers" that would be utilized in the discriminant function if the
variables were standardized to a z score. These coefficients are considered measures
of the relative contribution of each variable to the discriminant function (Klecka, 1980).
The structure coefficients are also useful in assessing the importance of each variable
to the analysis by providing a measure which is equivalent to the product-moment cor
relation of each predictor variable to the discriminant function. The structure coeffi
cients convey how closely a variable is related to the function (Klecka; Norusis, 1990).
As can be noted from Table 6, the variable cumulative GPA, followed by variables
transfer status and percent enrolled full-time, had the greatest standardized
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 6
Standardized Coefficients and Structure Coefficients of the Predictor Variables En
tered Independently in the Discriminant Analysis
Variable
Standardized
coefficients
Structure
coefficients
Socioeconomic status .233 .309
Age -.135 -.252
Gender -.014 .042
SAT score -.063 .332
Educational aspiration .037 .268
Career/financial goals .114 .002
Percentage enrolled full-time .270 .493
Hours worked -.266 -.387
Family responsibilities -.164 -.256
Involvement .214 .428
Cumulative grade point average .519 .596
Satisfaction -.005 .133
Transfer status -.433 -.524
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
coefficients and the highest structure coefficients (i.e., the highest correlations with the
discriminant function).
The ability of the discriminant function to predict group membership correctly
can be gauged from the hit ratio or overall percentage of known cases correctly classi
fied by the discriminant function. Classification results are presented in Table 7. Of
the total 1,261 cases in the usable sample, 84.3% or 1,063 of the cases were correctly
classified. The significance of this hit ratio is best interpreted by comparing it to the
number of cases one would expect to classify correctly by chance alone, which in this
case would be 875 or 69.6% (Klecka, 1980; Tabachnick & Fidell, 1989).
Table 7
Classification Results from the Initial Discriminant Analysis
Predicted group membership
Actual group
membership
Number of
cases
Ps
n %
NPs
n %
Persisters 1,025 984 96.0 41 4.0
Nonpersisters 236 157 66.5 79 33.5
Note. Percentage of grouped cases correctly classified: 84.3%. Ps = persisters (5-
year system persisters); NPs = nonpersisters (those who did not persist for 5 years.
The method used for determining the number of cases expected to be classi
fied correctly by chance alone was described by Tabachnick and Fidell (1989). To
elaborate on this method, a priori probabilities of group membership were used first to
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
compute the number of cases in each group that would be expected to be correctly
classified by chance alone. These were then summed to determine the overall ex
pected rate of correct classifications. To illustrate how this method was used in this
particular case, the a priori probabilities for group membership as set by actual group
membership were 81.3% and 18.7% for persisters and nonpersisters, respectively.
Accordingly, if 1,025 (the actual number of persisters) were randomly assigned to the
two groups, then one would expect 81.3% or 833 of them to be classified correctly by
chance alone. Similarly, if 236 (the actual number of nonpersisters) were randomly
assigned, then 18.7% or 44 would be expected to be correctly classified by chance
alone. Adding these together (833 + 44), the expected number and corresponding
percentage of cases that one would expect to be correctly classified by chance alone
would be 877 or 69.6% of the total 1,261 cases. Consequently, the percentage of cor
rectly classified cases achieved through the discriminant function (84.3%) was greater
than what could be expected by chance alone (69.6%) by 14.7%.
Looking at the individual groups, the percentage of persisters who were cor
rectly classified (96.0%) was greater than the percentage of nonpersisters who were
correctly classified (33.5%). Although the percentage of nonpersisters correctly clas
sified seems relatively low, it is approximately 1.8 times higher than the percentage
that would have been expected by chance alone (18.7%). These results provided sup
port for the hypothesis that the multiple discriminant analysis would yield a composite
of variables that would correctly classify students as system persisters or nonper
sisters at a level greater than chance.
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A follow-up analysis using step-wise discriminant analysis was conducted to
ascertain whether a subset of the initial 13 variables could be identified which would be
just as effective in predicting whether a student persisted or did not persist toward
achieving a bachelor’s degree in 4-year postsecondary institutions. The results
yielded 7 predictor variables that were, in essence, just as effective as the original 13
variables in discriminating persisters from nonpersisters. The variables that were
eliminated through the step-wise procedure were (a) gender, (b) age, (c) SAT score,
(d) initial educational aspiration, (e) career and financial success/security goals, and
(f) satisfaction with initial institution.
As expected, the results from the step-wise analysis were very similar to those
of the initial analysis. By way of comparison to the initial analysis, the Wilks’s Lambda
was equal to .811 (versus .806), the Eigenvalue was equal to .232 (versus .241), and
the canonical correlation was equal to .434 (versus .440). The percentage of cases
correctly classified was 81.1%, which was slightly lower than the 84.3% found utilizing
the original 13 predictor variables.
In order to determine just how effective the seven variables that had been iden
tified through the step-wise procedure were in predicting students’ 5-year persistence
status, an additional discriminant analysis was performed. This final analysis was con
ducted on a randomly selected portion (approximately 50%) of the sample and then
cross-validated using the remaining “unselected” cases.
Of the 3,278 cases available, 460 were deleted because of missing data, 1,396
cases were used for the analysis, and 1,422 cases were “held back” to use for cross
validating the results. Of the 1,396 cases retained for use in determining the
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
discriminant function, 1,077 (77.1%) were 5-year system persisters and 319 (22.9%)
were nonpersisters. The Wilks’ Lambda for the discriminant function was significant, A
= .772, x*(7, N = 1,396) = 360.11, p < .001. The resulting Eigenvalue was .241 and the
canonical correlation was .478. This corresponds to 22.8% variance explained, which
is better than the 19.4% of the variance that was explained using the original 13 vari
able discriminant function.
The standardized discriminant coefficients and structure coefficients for each of
the seven variables used in the reduced discriminant analysis model are presented in
Table 8. The classification results from the discriminant analysis utilizing the seven
predictor variables including the outcome of the cross-validation technique are sum
marized in Table 9. The percentage of cases correctly classified by the 7 variables
was 81.2% (cross-validated at 80.0%), which is slightly lower than the 84.3% that was
achieved when using 13 variables. Cross-validation was accomplished by using the
classification function derived from the original group above to classify those cases
held out for that purpose (approximately 50% of the sample).
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 8
Standardized Coefficients and Structure Coefficients of the Predictor Variables Used in
the Reduced Discriminant Analysis
Standardized Structure
Variable coefficients coefficients
Socioeconomic status .222 .321
Percentage enrolled full-time .335 .550
Hours worked -.221 -.344
Family responsibilities -.128 -.206
Involvement .206 .463
Cumulative grade point average .425 .518
Transfer status -.527 -.619
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 9
ables
Predicted group membership
Ps NPs
Actual group
membership
Number of
cases n % n %
Original3
Persisters
Nonpersisters
1,077
319
1,013
199
94.1
62.4
64
120
5.9
37.6
Cross-validated6
Persisters
Nonpersisters
1,107
315
1,020
197
92.1
62.5
87
118
7.9
37.5
Note. Ps = persisters (5-year system persisters); NPs = nonpersisters (those who did
not persist for 5 years.
3 Percentage of grouped cases correctly classified: 81.2%. b Percentage of grouped
cases correctly classified: 80.0%. Cross-validation was accomplished by using the
classification function derived from the original group above to classify those cases
held out for that purpose (approximately 50% of the sample).
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C H A P TE R 5
D ISC USSIO N , C O N C LU S IO N S , A N D R E C O M M E N D A TIO N S
Discussion
The following discussion of the findings is presented within the context of the
five research questions that were presented in Chapter 2.
Adequacy of the Goodness-of-Fit Indices
of the Initial Hypothesized Eclectic Model
to Explain System Persistence (Research
Question 1)
The results from analyzing the initial hypothesized eclectic model to explain
system persistence (Figure 1) indicated that the model was not consistent with the
data and that, therefore, the variables and the causal relationships as specifically de
fined by the model were not altogether plausible. However, this circumstance does not
imply that certain components and/or relationships within the model have no merit.
Instead, the interpretation is that, at least for this sample, this model, in its entirety, is
not adequate. An effort to modify the model, while staying within the framework of the
empirical and theoretical literature on attrition, was warranted.
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Adequacy of the Goodness-of-Fit Indices
of the Modified Model to Explain System
Persistence (Research Question 2)
Through systematically adding paths to and subtracting paths from the initial
hypothesized eclectic model, a modified version was produced which provided
adequate goodness-of-fit indices (Figure 2). This outcome indicated that the variables
and the relationships between them, as depicted in the modified model, were plausi
ble.
Of the seven paths that were added to the model, four further defined linkages
from background and demographic variables to intermediary variables, two were be
tween intermediary variables and one established a causal link between an intermedi
ary variable and the final dependent variable, system persistence. Of the paths that
were removed from the model because they did not contribute significantly to the fit of
the model, it is worthy to note that when one of them, the path from career and finan
cial goals to involvement, was removed, it left that variable no longer linked to system
persistence. This result implies that, unless its effect is mediated through some un
known variable, there is little to no relationship between the variable career and finan
cial goals and the variable system persistence as defined by the modified model. An
other path worth noting as being removed was the one that directly linked satisfaction
to persistence. However, satisfaction remained linked to persistence via an indirect
path mediated by transfer status. The percentage of variance that the modified model
explained in persistence was 16.3. This percentage is within the approximate range
(10%-30%) found in other multi-institutional persistence studies (Anderson, 1981;
Munro, 1981; Pascarella, 1985; Peng & Fetters, 1978).
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The fact that the modified eclectic model developed to explain system persis
tence was consistent with the data (and thereby supported the plausibility of the
postulated relationships) indicates that the tenets which were selected from the theo
ries of Tinto, Bean, and Astin to be included in the eclectic model were effective in
establishing at least an initial framework for explaining system persistence. In utilizing
this model as a framework for understanding system persistence, it should be kept in
mind that although the database provided many of the key variables found in the the
oretical models posited by Tinto, Bean, and Astin, it is possible that certain variables
not available in the database could have potentially enhanced the “ fit” of the model
above and beyond what was developed in the current formulation. For example, the
variables “intent to persist” (used by Bean) and “institutional commitment or loyalty”
(used by Tinto and Bean), which are two variables that are often employed in persis
tence research, were not available to incorporate within this study. The extent to
which this omission constitutes a weakness of the eclectic model to explain system
persistence could be ascertained through additional analyses that included these (or
any other potentially useful) variables.
Direct and Indirect Effects of Each of the
Predictor Variables on System Persistence
(Research Question 3)
Table 1 reveals that the input variables not only fluctuated with respect to their
total or complete effects on system persistence but also varied with respect to how
much the direct versus indirect effects contributed to their total effects on 5-year sys
tem persistence. The two variables with the highest standardized path coefficients for
total effects on persistence were GPA (beta = .190), followed by subsequent
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
educational aspirations (beta = .176). The effects of both of these variables were pri
marily realized through direct contributions. The literature that has discussed the re
lationship between academic performance and persistence has supported the conten
tion that undergraduate GPA is a valid predictor of persistence (Astin, 1975; Bean,
1990; Metzner & Bean, 1987; Pascarella & Terenzini, 1991). In addition, the results
indicating a positive relationship between subsequent educational aspirations and
persistence were consistent with a number of studies, especially those that have uti
lized data from multiple institutions (Braxton et al., 1997; Munro, 1981).
In reviewing the path coefficients in the current study, one can note that subse
quent educational aspirations not only had a relatively high direct effect on system per
sistence but also seemed to serve as an important intermediary variable in the modi
fied model. Most noteworthy was its mediating effect in the relationship between
“initial educational aspirations" and persistence. This component of the eclectic model
was derived from Tinto's model of student departure and, although the casual link from
initial educational aspirations to subsequent educational aspirations to persistence is
not delineated in the theories posited by Bean and Astin, it does appear to help explain
system persistence in the current study. Looking at the model further (Figure 2), one
can see that subsequent educational aspirations also provided important mediating
effects between both GPA and persistence, and between “academic ability" (SAT
score) and persistence.
In the current study, the effect of initial educational aspiration on persistence
was realized completely through intermediary variables—an outcome which is
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
consistent with most of the empirical data related to student attrition (Bean, 1990;
Munro, 1981; Tinto, 1975) but not all (Astin, 1975; Bean & Metzner, 1985).
Of the background and demographic variables included in the eclectic model,
SES, age, and academic ability (SAT score) had significant total effects on 5-year sys
tem persistence (beta = .103, -.097, and .114, respectively). For SES, and to a lesser
degree for age, the direct effects on system persistence were much greater than the
indirect effects. Although there is evidence that background variables have a direct
relationship to persistence (Braxton et al., 1997), the general concept in the theoretical
models posited to explain persistence is that background variables have their impact
on persistence primarily through intermediary variables (Bean, 1990; Tinto, 1993).
Consistent with this premise, most of the impact of academic ability (SAT score) on
persistence was through indirect pathways.
An observation worth noting was related to the effects of both academic ability
and gender on system persistence. In both cases, the relationships of the direct and
indirect effects were in opposite directions. This finding is somewhat puzzling. In the
case of academic ability, when the relatively strong positive indirect effect (beta = .161)
that was mediated primarily through GPA and educational aspirations was combined
with the negative direct effect of lower magnitude (beta = -.047), the total effect re
mained positive and significant. However, in the case of gender, the direct effect (beta
= -.038) and the indirect effect (beta = .040) virtually cancelled each other, and thereby
left the total effect of gender on persistence small and insignificant. Most of the rele
vant literature has confirmed that gender has very little to do with persistence behavior
at 4-year institutions (Baker & Velez, 1996; Tinto, 1993). It is unclear why the direct
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
effects of both academic ability and gender on persistence were negative in this anal
ysis.
The variable in the path diagram that showed the third highest total effect on
system persistence was percentage full-time enrollment (beta = .155), and most of the
effect from that variable was attributable to its direct positive effect (beta = .122). A
similar finding was reported by Metzner and Bean (1987), who found a significant
direct effect between the variable hours enrolled and dropout behavior.
Although the contribution of indirect effects of this variable on persistence was
relatively low (beta = .033), one might note that percentage full-time enrollment did
have a fairly high direct positive effect (beta = .221) on the intermediary variable in
volvement, which, from a theoretical point of view, could be interpreted to mean that
full-time attendance increases the likelihood that a student would become involved
with the academic and social aspects of the college community (Astin, 1984; Tinto,
1993). Alternatively, it might also imply that part-time students have other responsibili
ties and/or activities that are competing for their time and interest which, in turn, would
decrease their involvement on campus (Astin, 1996; Pascarella, 1980).
Of the two variables included in the path diagram that represented pull factors,
only hours worked showed any significant relationship to persistence. All of the effect
of hours worked on persistence was realized through a small but significant negative
indirect effect (beta = -.017). This outcome was interpreted to mean that increasing
the hours a student works while going to school could have an impact on factors that,
in turn, could negatively impact persistence. It is noteworthy that the indirect effect of
hours worked was mediated through the intermediary variable percentage full-time
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
enrollment (via a path that was added during model modification), which, in turn, had a
significant impact on persistence, including some effect indirectly through involvement.
The literature related to the relationship of hours worked to attrition has primarily as
sessed the impact of full-time employment. In addition, most of these studies have fo
cused on off-campus work. Given this fact, however, there has been fairly strong
support from the literature that full-time work, especially when that work is off campus,
can have a negative impact on persistence (Astin, 1993; Ehrenberg & Sherman, 1987;
Pascarella & Terenzini, 1991; Tinto, 1993). However, the way in which hours worked
has an effect on persistence (i.e., directly versus indirectly) has not been studied very
much. When Anderson (1981) analyzed a national database using causal modeling
techniques, she found that the effect of hours worked on system persistence was
direct and that the effect was not mediated through intermediary variables, an outcome
which was contrary to the findings of this study. However, the key variable providing
the mediating effect in this study, percentage full-time enrollment, was not included in
Anderson's study.
The variable involvement at initial institution also had a significant total effect
on 5-year system persistence (beta = .151). This effect was both direct (beta = .098)
and indirect (beta = .053), a finding which supports the underlining theoretical perspec
tive that student involvement (and/or integration and/or interaction) helps to promote
persistence (Astin, 1984, 1993; Bean, 1990; Tinto, 1975, 1993). The indirect effects of
involvement on persistence were mediated by subsequent educational aspirations,
scholastic performance (GPA), satisfaction, and transfer status.
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
From a theoretical perspective, the indirect effect through subsequent educa
tional aspirations found in this study was akin to the causal link typically associated
with Tinto’s model of student departure that suggests that academic integration influ
ences subsequent goal commitment, which, in turn, affects persistence. However, a
meta-analysis of studies testing Tinto’s theory indicated only moderate support for that
indirect relationship (Braxton et al., 1997).
The mediating effect of GPA in the relationship between involvement and per
sistence as found in this analysis was consistent with the research and writings of
Spady (1970), Bean (1983, 1990), and Astin (1993). However, the results could not be
adequately compared to Tinto’s model (of student departure) because Tinto’s model
includes GPA as a component (i.e., indicator) of the integration construct, not as a
separate variable (Tinto, 1975, 1993). Despite this operational difference, Tinto’s
theory would support the contention that higher GPAs “indicate” greater academic in
tegration, which, in turn, can impact persistence. No empirical evidence (excluding
that from this investigation) was found to support or to refute the finding of indirect ef
fects of involvement on persistence mediated through GPA.
In a similar fashion, it does not appear that the mediating effects of both satis
faction and transfer status in the causal link between involvement and persistence
have been tested elsewhere to the extent required to permit adequate comparisons to
the literature.
It is interesting to note (in Figure 2) that, even though involvement functioned to
some degree as an intermediary variable for practically every variable feeding into it in
the causal design, the variance explained in involvement itself was only 8%. This
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
result implies that although involvement appears to be significant as an antecedent to
persistence it is likely that t^ere may be factors contributing to and affecting involve
ment that are not currently included in the modified eclectic model to explain system
persistence, as presented iri Figure 2.
Because the direct link fr°m the variable student satisfaction at their initial in
stitution to persistence was dropped during the process of model modification, all of
the effects of satisfaction ori persistence were mediated through the variable transfer
status. Although this positi^e indirect effect was statistically significant, the magnitude
was relatively small (beta = 019). This result is similar to the findings of Bean (1980),
when he applied his modelt0 men, and to those of Metzner and Bean (1987). It is
worthy to note that only 0.5% of the variance in the satisfaction variable was explained
by the model.
It was not possible f° ^nd any research that investigated the relationship be
tween satisfaction levels arid actual transfer behavior that could be referred to for com
parative purposes. Although Astin (1993) reported a correlation between left school or
transferred and satisfactiori measures, he did not distinguish between students who
had left school without trans^ errin9 and those who had transferred and then had per
sisted at another institution- Despite the lack of previous research, it makes sense in
tuitively that, if a student e)£Penences low levels of satisfaction at an institution, then he
or she may be more likely t° transfer to another institution.
In addition to functiPmn9 as an intermediary variable between satisfaction and
persistence, transfer status a* so served to mediate the indirect effects of academic
ability (added path), involvPment (added path), and subsequent educational
85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
aspirations (added path) on system persistence. Transfer status was found to have a
significant direct negative effect on persistence behavior (beta = -.118). No examples
of other studies where transfer status had been integrated into a causal model to ex
plain persistence could be found for comparative purposes.
Bivariate Relationship Between Each of the
Predictor Variables and 5-vear System
Persistence (Research Question 4)
A series of bivariate analyses was conducted to determine whether there was a
statistically significant difference between persisters and nonpersisters relative to each
of the following predictor variables: socioeconomic status, age, gender, academic
ability, educational aspirations, career and financial success/security, percentage full
time enrollment, hours worked, family responsibilities, involvement at initial institution,
scholastic performance (GPA), satisfaction, and transfer status. The following is a
discussion of the findings for this research question.
The results from this study confirmed the hypothesis that students who per
sisted 5 years would come from higher SES backgrounds than those who did not per
sist. The literature had supported a small but positive relationship between SES and
persistence behavior (Astin, 1993; Bean, 1990; Porter, 1990; Summerskill, 1962;
Tinto, 1993).
The hypothesis that students who persisted would be younger than those who
did not persist was supported by the analysis. A relatively small but statistically signifi
cant inverse relationship existed between age and system persistence. Although a
variety of studies has demonstrated a relationship between age and persistence
(Astin, 1975; Murtaugh et al., 1999; Summerskill, 1962), it is generally believed that
86
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
factors other than age are the reason for the apparent relationship. In particular, older
students tend to have responsibilities in addition to those related to school that can
“pull” them away from their studies (Bean & Metzner, 1985; Tinto, 1993). In this par
ticular study, the correlations between age and the two pull factors (a) hours worked
and (b) family responsibilities were .06 (p < .05) and .18 (p < .01), respectively. These
findings provided some support for this contention.
Although females in this sample had slightly higher rates of 5-year persistence
than males, the magnitude of the relationship as determined through the bivariate
analysis was quite small and nonsignificant. This finding, which concurred with the re
sults from the path analysis, provided support for the hypothesis that there would be
no relationship between gender identity and system persistence. Empirical studies
have generally supported the assertion that gender has very little to do with persis
tence behavior at 4-year institutions (Baker & Velez, 1996; Metzner & Bean, 1987;
Peng & Fetters, 1978; Tinto, 1993).
The bivariate analyses using SAT and ACT scores as a measure of academic
ability supported the hypothesis that students who persisted would have higher aca
demic ability than those who did not persist. A positive relationship between ability and
persistence has been well supported by the literature, although high school GPA (not
available for this study) has been shown to be a stronger indicator of college persis
tence and attainment than have scores on aptitude tests (Astin, 1975, 1993; Cope &
Hannah, 1975; Porter, 1990; Summerskill, 1962; Tinto, 1993).
The hypothesis stating that persisters would have greater educational aspira
tions than would nonpersisters was supported by the findings. These results, which
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
were consistent with the “institutional departure” models posited by Tinto and Bean,
have been empirically supported in studies related to system persistence (Anderson,
1981; Peng & Fetters, 1978). The bivariate analysis indicated that the relationship be
tween students’ initial (Year 1) educational goal levels and 5-year persistence was
weaker than the relationship between their subsequent (Year 3) educational goal
levels and persistence. This finding may imply that, as an entry variable, educational
goal levels may not be a very strong predictor of 5-year persistence, and that assess
ing student educational aspirations on an ongoing basis may be more useful in predict
ing persistence behavior.
Although studies have shown that earning a bachelor’s degree provides long
term economical and financial rewards to individuals and their families (Leslie & Brink
man, 1986; Mortenson, 1999; Pascarella & Terenzini, 1991), the results from this
study revealed a small negative relationship between career and financial success
goals of students and their propensity to persist toward degree completion. Even
though the difference between the means was fairly small, it is worth noting that the
direction of the difference was opposite to what had been hypothesized.
The reason that students who persisted had lower career and financial success
goals is not altogether clear. The literature has indicated that a positive relationship
between these factors can be found if the student perceives that attending college is a
necessary prerequisite for obtaining a specific occupational goal. However, the litera
ture also has indicated that most students are uncertain about their career aspirations
when they first enter college (Tinto, 1993). By way of conjecture, it may be possible
that if a student is highly ambitious to receive financial/career gains but does not see a
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
connection between completing a college degree and achieving that financial/career
goal, then the student may eventually leave college in order to pursue that financial/
career goal. If this choice were to be the case, then it may be possible that simple on
going reminders to students about the link between having a bachelor’s degree and
the likelihood of obtaining future career and financial rewards may, in itself, be benefi
cial for promoting persistence.
The percentage of months that a student attended full-time while enrolled had a
relatively strong positive relationship to 5-year system persistence. This finding is of
particular concern because part-time enrollments make up approximately 31% of all
enrollments in 4-year institutions (Tinto, 1993). The results from this study suggest
that institutions may want to develop programs to increase retention that are specifi
cally directed toward helping part-time students.
The hypothesis projecting that persisters would work fewer hours per week
than nonpersisters was supported by the bivariate analysis, an outcome that concurs
with the results from the path analysis. These results could imply that monitoring the
hours a student works may be a useful tool in identifying whether the student is poten
tially at risk of dropping out of college. These findings were in accordance with the ex
isting body of literature which indicates that factors external to college that may “pull” a
student away from fully participating in the academic and social communities of the
institution may negatively impact persistence (Astin, 1993; Bean, 1990; Pascarella &
Terenzini, 1991; Tinto, 1993).
The hypothesis stating that greater family responsibilities would lead to de
creased levels of 5-year persistence was upheld statistically in the bivariate analysis,
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
although the size of the difference between means was relatively small. Consequent
ly, this study only weakly supported the prevailing related literature (Bean, 1990; Nora
etal., 1996; Tinto, 1993).
The finding that students who persisted for 5 years had a significantly higher
mean level of involvement at their initial institution than those who did not persist sup
ports most (Astin, 1975; Bean, 1990; Cabrera et al., 1992; Lamport, 1993; Pascarella,
1980; Tinto, 1993) but not all (Braxton et al., 1997; Ruddock et al., 1999; Stage &
Rushin, 1993) of the literature which has contended that academic and social involve
ment and integration has been paramount for persistence. Consistent with the results
from the path analysis, this study suggests that student involvement has a significant
positive effect on system persistence.
The bivariate analysis also provided fairly strong support for the contention that
system persisters would have higher overall GPAs than nonpersisters. This finding
concurred with the path analysis and, as previously discussed, was weli supported in
the literature (Astin, 1975; Bean, 1990; Pascaralla & Terenzini, 1991). The role that
GPA plays in explaining attrition in Tinto’s model may not be comparable to that sug
gested by this study because (a) Tinto’s model does not include GPA as a separate
variable and (b) Tinto’s theory is intended to explain voluntary departure. It therefore
excludes students with GPAs low enough to warrant academic dismissal (Tinto, 1975).
In the context of the relationship between satisfaction and persistence, the bi
variate analysis weakly supported the hypotheses that persisters would have a higher
level of satisfaction at their initial institution than would nonpersisters. In contrast, the
findings did not support the hypothesis that student satisfaction levels at the last
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
institution attended would be higher for those students who persisted for 5 years. The
reason for the disparity between these two results is unclear. By way of conjecture, it
is possible that, as students approach degree completion, their satisfaction levels
could be influenced by a variety of factors, several of which may have nothing to do
with their experiences on campus but instead reflect their desire to move on to another
stage of life. Nevertheless, the weak connection between satisfaction and persistence
as determined by bivariate analysis plus the lack of any direct effect between satisfac
tion and persistence in the path analysis demonstrated a discrepancy between the
findings of this study and those cited in the literature. This difference might be at least
partially a result of the fact that the studies comparing satisfaction and retention found
in the literature (Astin, 1993; Hatcher et al., 1992; Mohr et al., 1998; Starr et al., 1972)
focus on institutional as opposed to system persistence.
The results offered relatively strong support for the hypothesis that students
who had persisted were less likely to have transferred from their initial 4-year institu
tion than those who had not persisted. Although the relationship between transfer sta
tus and persistence has not been studied extensively (Tinto, 1993), the results have
mostly (Astin, 1975; Pascarella & Terenzini, 1991) but not entirely (Carroll, 1989) con
firmed that transferring sometime after beginning a bachelor’s degree program at a 4-
year institution has a negative impact on persistence. It is understandable that only a
small amount of literature would exist in this area, as most persistence studies have
focused on institutional persistence, which leaves the transfer student to be included
with dropouts or to be eliminated from the study.
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The findings from this study suggest that furthering the understanding of trans
fer behavior may be helpful for improving an understanding of system persistence. If a
student could be identified as someone who might be “at risk” of transferring, an insti
tution might seek to identify and then to rectify whatever problems the student was
experiencing that might be causing him or her to want to transfer. Although this inter
vention could possibly decrease the incidence of institutional departure, it might be a
limited view when discussing system persistence. This view does not take into consid
eration the fact that some students enter an institution with the intention to transfer,
and some choose to transfer in an effort to find an institution with a better “ fit” (Pas
carella & Terenzini, 1991; Tinto, 1993). With this fact in mind, institutions may want to
develop programs that are aimed at facilitating smoother, more successful transfers
which, for some individuals, may be the most appropriate tactic for increasing their
likelihood of persisting toward degree completion.
Utilizing Multiple Discriminant Analysis to
Predict 5-vear System Persistence at a Level
Greater Than Chance (Research Question 5)
The 13 variables utilized in the initial discriminant analysis correctly classified
84.3% of the students as either 5-year system persisters or nonpersisters. Although
this percentage was reasonably high, the result was less remarkable when compared
with the percentage of students that would have been correctly classified by chance
alone (69.6%). Nevertheless, the discriminant function was able to classify students
correctly at a rate greater than chance. In addition, the fact that the discriminant func
tion accounted for 19.4% of the variance between the two groups (persisters versus
nonpersisters) indicated that the model included some variables that were effective in
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
predicting system persistence. The variables contributing most to the predictability of
the discriminant function were GPA, transfer status, percent enrolled full-time, hours
worked, and involvement. This finding further elucidated the strengths of the relation
ships that these factors had to system persistence, as discussed earlier in this section.
If one refers to the percentage of students correctly classified into each group, the rate
of correctly classifying persisters was considerably higher than the rate of correctly
classifying nonpersisters. The tendency for this outcome would be expected when
ever one of the groups is much larger than the other and the prior probabilities are set
to reflect the actual percentages that each group is of the total sample (Norusis, 1990).
However, it might be noted that, by manipulating the prior probabilities, one can in
crease the number of subjects correctly classified in one group at the “ expense" of
decreasing the number correctly classified in the other group. One might follow
through with this course of action if it were determined that misclassification into one
group had greater consequences than misclassification into the other group. In pre
dicting persistence, for example, one could increase the percentage of correctly classi
fying nonpersisters at the expense of decreasing the percentage of correctly classify
ing persisters if the misclassification of nonpersisters was of great concern (Klecka,
1980).
When a step-wise discriminant analysis was conducted as a follow-up, the six
variables— gender, age, SAT scores, initial educational aspirations, career and finan
cial success/security goals, and satisfaction with the initial institution— could be elimi
nated without significantly affecting the predictability of the discriminant function.
When the reduced model was applied to the sample, there was a slightly lower
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
percentage of correctly classified cases (81.2% versus 84.3%), but the variance be
tween the two groups accounted for by the discriminant function that was derived
through the step-wise analysis was slightly higher (22.8% versus 19.3%). By way of
comparison to the path analysis, the variance explained achieved through conducting
the discriminant analysis (22.8% with the reduced model) was slightly higher than the
16.3% variance explained in system persistence that was found with the path analysis.
Although this amount of variance explained was within the range typically found
when persistence studies have been conducted utilizing multiple institutions, it is
worthwhile to put forth some ideas that might provide an explanation for the unex
plained variance (approximately 80%) in the dependent variable, system persistence.
To elaborate, one reason might be that it is feasible that not all of the variables that
could have had an impact on system persistence (and/or on intermediary variables
affecting persistence) were identified for use in the current study. Variables such as
high school GPA, institutional commitment or loyalty, expectations, intention to trans
fer, intention to persist, parental or spouse encouragement, plus organizational factors
such as admission policies, course offerings, and the opportunity to become involved
may all be examples of factors that might help to explain system persistence but were
not available for use in the current study.
It is also possible that part of the unexplained variance might relate to the fact
that the sample in this study was fairly heterogeneous. A more homogeneous group of
students could be obtained by further delimiting the sample by such factors as gender
(only females or males), age (traditional or nontraditional), type of institution (public or
private), institution selectivity, enrollment intensity (part-time or full-time), enrollment
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
continuity (students enrolled continuously or those who were stopouts), or by types of
transfers (horizontal or reverse). Through defining a more homogeneous subset as
the sample, it is possible that the correlations between variables might increase— an
increase which, in turn, would help to enhance the amount of variance explained.
Conclusions
The major conclusions are summarized as follows:
1. Many of the factors demonstrated to be associated with system-wide per
sistence in the current study are similar to those that have been shown to help to ex
plain institutional departure behavior. In particular, the findings suggest that the back
ground variables age, socioeconomic status, and ability, plus educational aspirations,
percentage of full-time attendance, hours worked at a job, scholastic achievement, and
student involvement can all, to various extents, help to explain system persistence.
2. An eclectic path analytic model incorporating in a selective way variables
from an existing database was judged to provide an adequate degree of goodness-of-
fit to afford a basis for an initial interpretation of both direct and indirect (mediating) ef
fects of key variables in explaining the persistence of students seeking the bachelor’s
degree in the 4-year postsecondary educational system.
3. This study suggests that, if a student transfers after beginning a baccalau
reate degree program at a 4-year institution, the student would be somewhat less likely
to persist for 5 years toward obtaining a bachelor’s degree in a 4-year institution.
4. Student satisfaction levels would appear to have a significant impact on the
decision to transfer from one institution to another but to have little, if any, direct rela
tionship to persistence in completing the bachelor’s degree.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5. Input variables based on promising results from bivariate analyses led to
the formation of a multiple discriminant function that could predict system persistence
at a level at least moderately greater than afforded by chance alone.
Recommendations
The following recommendations are offered:
1. Additional studies to strengthen and expand the understanding of system
persistence are warranted. Such studies should include, if possible, factors not in
cluded in the current study, such as high school GPA or ranking level, the types and
locale of jobs at which students worked while enrolled (along with the extent to which
those jobs related to the students’ area of study), institutional commitment, parental or
spouse encouragement, admission policies, the opportunity to become involved, and
student intentions as related to both transferring and persisting.
2. The area of study would benefit from further investigations into why some
students transfer from their initial 4-year institution to continue their education at an
other postsecondary institution. Identifying factors associated with transfer behavior
would aid in the development of effective intervention methods which would be di
rected toward enhancing student persistence toward degree completion.
3. Further investigations into understanding the relationship that student sat
isfaction levels have with transferring and/or persistence behavior would be informa
tive. Such studies might include students’ initial expectations into the satisfaction
measure.
4. Replication of the current investigation might be carried out utilizing a more
homogenous subset of the population of students attending 4-year institutions. Such
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
delimitations could be based on gender, age, institution type or selectivity, enrollment
intensity (part-time or full-time), enrollment continuity (continuous or stopout), and/or
types of transfers (horizontal or reverse transfers).
5. Public and private organizations influential in higher education might en
deavor to implement the findings from studies similar to the current one in evaluating
policy issues related to the persistence of degree-seeking students in 4-year post
secondary institutions.
6. Four-year institutions might utilize information gained from studies similar to
this investigation when developing methods or strategies to promote student persis
tence. Findings from this investigation would suggest that efforts that especially en
courage involvement, enhance academic achievement, promote full-time attendance,
and/or elevate educational aspirations could promote persistence irrespective of
whether a student remained at a given institution or transferred. Resources that may
be available to institutions to support such efforts may be more available if a case
could be made that the support could potentially enhance student success even if the
student transfers.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
REFERENCES CITED
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
REFERENCES
Adelman, C. (1998). What proportion of college students earn a degree? AAHE Bul
letin, 51(2), 7-9.
Anderson, K. L. (1981). Post-high school experiences and college attrition. Sociology
of Education, 54, 1-15.
Astin, A. W. (1975). Preventing students from dropping out. San Francisco: Jossey-
Bass.
Astin, A. W. (1984). Student involvement: A developmental theory for higher educa
tion. Journal of College Student Personnel, 25, 297-308.
Astin, A. W. (1993). What matters in college: Four cntical years. San Francisco:
Jossey-Bass.
Astin, A. W. (1996). Involvement in learning revisited: Lessons we have learned.
Journal of College Student Personnel, 37, 123-133.
Astin, A. W. (1997). How “good” is your institution’s retention rate? Research in
Higher Education, 38, 647-658.
Baker, T. L., & Velez, W. (1996). Access to and opportunity in postsecondary educa
tion in the United States: A review. Sociology of Education [Extra issue], 82-101.
Bean, J. P. (1980). Dropouts and turnover: The synthesis and test of a causal model
of student attrition. Research in Higher Education, 12, 155-187.
Bean, J. P. (1982). Student attrition, intentions, and confidence: Interaction effects in
a path model. Research in Higher Education, 17, 291-320.
Bean, J. P. (1983). The applications of a model of turnover in work organizations to
the student attrition process. Review of Higher Education, 6, 129-148.
Bean, J. P. (1990). Why students leave: Insights from research. In D. Hossler, J. P.
Bean, & associates (Eds.), Strategic management of college enrollments (pp.
147-169). San Francisco: Jossey-Bass.
99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Bean, J. P., & Metzner, B. S. (1985). A conceptual model of nontraditional undergrad
uate student attrition. Review of Educational Research, 55, 485-540.
Bentler, P. M., & Wu, E. J. C. (1995). EQS for Windows user’s guide. Encino, CA:
Multivariate Software.
Berger, J. B., & Braxton, J. M. (1998). Revising Tinto’s interactionalist theory of stu
dent departure through theory elaboration: Examining the role of organizational
attributes in the persistence process. Research in Higher Education, 39, 103-119.
Berger, J. B., & Milem, J. F. (1999). The role of student involvement and perceptions
of integration in a causal model of student persistence. Research in Higher Edu
cation, 40, 641-664.
Braxton, J. M., Sullivan, A. S., & Johnson, R. (1997). Appraising Tinto’s theory of col
lege student departure. In J. Smart (Ed.), Higher education: Handbook of theory
and research (Vol. 12, pp. 107-164). New York: Agathon.
Byrne, B. M. (1994). Structural equation modeling with EQS and EQSAA/indows.
Thousand Oaks, CA: Sage.
Cabrera, A. F., Castaneda, M. B., Nora, A., & Hengstler, D. (1992). The convergence
between two theories of college persistence. Journal of Higher Education, 63,
143-164.
Cabrera, A. F., Nora, A., & Castaneda, M. B. (1993). Structural equation modeling
test of an integrated model of student retention. Journal of Higher Education, 64,
125-139.
Carroll, C. D. (1989). College persistence and attainment for 1980 high school gradu
ates: Hazards for transfers, stopouts, and part-timers (NCES 89-302). Washing
ton DC: U.S. Department of Education, Office of Educational Research and Im
provement.
Cope, R., & Hannah, W. (1975). Revolving college doors: The causes and conse
quences of dropping out, stopping out, and transferring. New York: Wiley.
Durkheim, E. (1951). Suicide (J. Spaulding & C. Simpson, Trans.). Glencoe, NY:
Free Press. (Original work published 1897)
Ehrenberg, R., & Sherman, D. (1987). Employment while in college, academic
achievement, and postcollege outcomes: A summary of results. Journal of Hu
man Resources, 22, 1-23.
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hatcher, L., Kryter, K.t Prus, J. P., & Fitzgerald, V. (1992). Predicting college student
satisfaction, commitment, and attrition from investment model constructs. Journal
of Applied Social Psychology, 22, 1273-1296.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation Mod
eling, 6, 1-55.
Immerwahr, J. (1998). The price of admission: The growing importance of higher ed
ucation (Report No. 98-2). San Jose: CA: National Center for Public Policy and
Higher Education.
Immerwahr, J., & Foleno, T. (2000). Great expectations: How the public and par
ents— White, African American, and Hispanic— view higher education (Report No.
00-2). San Jose, CA: National Center for Public Policy and Higher Education.
Klecka, W. R. (1980). Discriminant analysis. Newbury Park, CA: Sage.
Kline, R. B. (1998). Principles and practice of structural equation modeling. New
York: Guilford.
Kroc, R., Howard, R., Hull, P., & Woodard, D. (1997, May). Graduation rates: Do stu
dents' academic program choices make a difference? Paper presented at the an
nual forum of the Association for Institutional Research, Orlando, FL.
Lamport, M.A. (1993). Student-faculty interaction and the effect on college student
outcomes: A review of the literature. Adolescence, 28, 971-990.
Leslie, L., & Brinkman, P. (1986). Rates of return in higher education. In J. Smart (Ed.),
Higher education: Handbook of theory and research (Vol. 2, pp. 207-234). New
York: Agathon.
Metzner, B. S., & Bean, J. P. (1987). The estimation of a conceptual model of nontra-
ditional undergraduate student attrition. Research in Higher Education, 27, 15-38.
Milem, J. F., & Berger, J. B. (1997). A modified model of college student persistence:
Exploring the relationship between Astin’s theory of involvement and Tinto’s
theory of student departure. Journal of College Student Development, 38, 387-
400.
Mohr, J. J., Eiche, K. D., & Sedlacek, W. E. (1998). So close, yet so far: Predictors of
attrition in college seniors. Journal of College Student Development, 39, 343-354.
Mortenson, T. G. (April, 1999). Family income by educational attainment: 1956 to
1997. Postsecondary Education Opportunity. Iowa City, IA: Author.
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Munro, B. H. (1981). Dropouts from higher education: Path analysis of a national
sample. American Educational Research Journal, 18, 133-141.
Murtaugh, P. A., Bums, L. D., & Schuster, J. (1999). Predicting the retention of uni
versity students. Research in Higher Education, 40, 355-371.
National Center for Educational Statistics. (1992). Beginning postsecondary students
longitudinal study field test methodology report (BPS.90/92) (NCES 92-160).
Washington DC: U.S. Department of Education.
National Center for Educational Statistics. (1994a). Beginning postsecondary stu
dents longitudinal study first follow-up (BPS.90/92): Final public technical report
(NCES 94-369). Washington DC: U.S. Department of Education.
National Center for Educational Statistics. (1994b). Beginning postsecondary students
longitudinal second study: Follow-up field test report (BPS:90/94) (NCES 94-
370). Washington DC: U.S. Department of Education.
National Center for Educational Statistics. (1996). Beginning postsecondary students
longitudinal study second follow-up (BPS.90/94): Final technical report (NCES
96-153). Washington DC: U.S. Department of Education.
National Center for Educational Statistics. (1997). Findings from the condition of edu
cation 1997: Postsecondary persistence and attainment (NCES 97-371). Wash
ington DC: U.S. Department of Education.
Nora, A., Cabrera, A., Hagedorn, L. S., & Pascarella, E. (1996). Differential impacts
of academic and social experiences on college-related behavioral outcomes
across different ethnic and gender groups at four-year institutions. Research in
Higher Education, 37, 427-451.
Norusis, M. J. (1990). SPSS advanced statistics user’s guide. Chicago: SPSS, Inc.
Ozga, J., & Sukhnandan, L. (1998). Undergraduate non-completion: Developing an
explanatory model. Higher Education Quarterly, 52, 316-333.
Pascarella, E. T. (1980). Student-faculty informal contact and college outcomes. Re
view of Educational Research, 50, 545-595.
Pascarella, E. T. (1985). Racial differences in factors associated with bachelor’s de
gree completion: A nine-year follow-up. Research in Higher Education, 23, 351-
373.
Pascarella, E. T., & Terenzini, P. T. (1980). Predicting freshman persistence and vol
untary dropout decisions from a theoretical model. Journal of Higher Education,
51, 60-75.
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Pascarella, E. T., & Terenzini, P. T. (1991). How college affects students. San Fran
cisco: Jossey-Bass.
Peng, S. P., & Fetters, W. B. (1978). Variables involved in withdrawal during the first
two years of college: Preliminary findings from the national longitudinal study of
the high school class 1972. American Educational Research Journal, 15, 361-
372.
Porter, O. F. (1990). Undergraduate completion and persistence at four-year colleges
and universities: Completers, persisters, stopouts, and dropouts. Washington,
DC: National Institute of Independent Colleges and Universities. (ERIC Docu
ment Reproduction Service No. ED 319 343)
Price, J. L. (1977). The study of turnover. Ames: Iowa State University Press.
Price, J. L., & Mueller, C. W. (1981). A causal model of turnover for nurses. Academy
of Management Journal, 24, 543-565.
Ruddock, M. S., Hanson, G., & Moss, M. K. (1999, June). New directions in student
retention research: Looking beyond interactional theories of student departure.
Paper presented at the annual forum of the Association for Institutional Research,
Seattle, WA.
Sanders, L., & Burton, J. D. (1996). From retention to satisfaction: New outcomes for
assessing the freshman experience. Research in Higher Education, 37, 555-567.
Spady, W. G. (1970). Dropouts from higher education: An interdisciplinary review
and synthesis. Interchange, 1, 64-85.
Spady, W. G. (1971). Dropouts from higher education: Toward an empirical model.
Interchange, 2, 38-62.
Stage, F. K. (1989). Motivation, academic and social integration, and early dropout.
American Educational Research Journal, 26, 385-402.
Stage, F. K., & Rushin, P. W. (1993). A combined model of student predisposition to
college and persistence in college. Journal of College Student Development, 34,
276-282.
Starr, A., Betz, E. L., & Menne, J. (1972). Difference in college student satisfaction:
Academic dropouts, nonacademic dropouts, and nondropouts. Journal of Coun
seling Psychology, 19, 318-322.
Summerskill, J. (1962). Dropouts from college. In N. Sanford (Ed.), The American
college (pp. 627-657). New York: Wiley.
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Tabachnick, B.G., & Fidell, L. S. (1989). Using multivariate statistics (2n d e<±). New
York: Harper Collins.
Tierney, W. G. (1992). An anthropological analysis of student participation in college.
Journal of Higher Education, 63, 603-618.
Tinto, V. (1975). Dropout from higher education: A theoretical synthesis of recent
research. Review of Educational Research, 45, 89-125.
Tinto, V. (1986). Theories of student departure revisited. In J. Smart (Ed.), Higher
education: Handbook of theory and research (Vol. 2, pp. 359-384). New York:
Agathon.
Tinto, V. (1993). Leaving college: Rethinking the causes and cures of student attri
tion (2n d ed.). Chicago: University of Chicago Press.
Van Gennep, A. (1960). The rites of passage (M. Viedon & G. Caffee, Trans.). Chi
cago: University of Chicago Press.
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

An examination of the factors that affect education vs non-education career choices of Asian Pacific Americans in higher education

PDF

A typological history of non-heterosexual male college students in the United States, 1945 to the present

PDF

Adult urban community college student success: Identifying the factors that predict course completion and goal attainment for students aged 25 years and older

PDF

Exploring ethnic identity on a university campus: Filipino American students' perspectives

PDF

Community college students: The effect of parenthood and selected variables on degree -seeking aspirations

PDF

"Home boy" to "school boy": Inside the PowerBuilders Workshops at CSULA. A qualitative examination of access, mobility, and leadership qualities of minorities in higher education

PDF

Academic performance and persistence of Asian American students in the Los Angeles Community College District

PDF

Institutional factors affecting academic persistence of underprepared community college freshmen

PDF

Course completion and older adult students: The effects of self-efficacy, self -regulation, and motivation

PDF

Course -taking patterns of Chinese students native and non -native speakers of English at community college

PDF

Factors influencing academic success of Chinese international students in Los Angeles community colleges

PDF

A daily diary approach to compare the accuracy of depressed and nondepressed participants' estimation of positive and negative mood: A test of the depressive realism hypothesis

PDF

Academic success for African -American male community college students

PDF

A cultural study of shared governance at two community colleges

PDF

Gender, learning, and trafficking: Helping vulnerable Thai women through NGO and government non -formal education programs

PDF

International students: Patterns of success

PDF

Border crossing: the experience of international students from Latin America in a California community college

PDF

A qualitative examination of the nature and impact of three California minority engineering programs

PDF

China's critical educational access demand and United States higher education distance learning curriculum: An answer?

PDF

In their own words: Polynesian students' perspectives on persistence in an American university

Asset Metadata

Creator Blecher, Lee (author)

Core Title Factors related to the persistence of students seeking the bachelor's degree at 4-year institutions

School School of Education

Degree Doctor of Philosophy

Degree Program Education

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Education, Guidance and Counseling,education, higher,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Michael, William B. (committee chair), [illegible] (committee member), Hagedorn, Linda Serra (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-118461

Unique identifier UC11338805

Identifier 3041437.pdf (filename),usctheses-c16-118461 (legacy record id)

Legacy Identifier 3041437.pdf

Dmrecord 118461

Document Type Dissertation

Rights Blecher, Lee

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA