Citation
An examination of the relationship between student financial aid and retention of students of differing ethnic and income backgrounds in baccalaureate programs in Colorado

Material Information

Title:
An examination of the relationship between student financial aid and retention of students of differing ethnic and income backgrounds in baccalaureate programs in Colorado
Creator:
Thayer, Paul B
Publication Date:
Language:
English
Physical Description:
635 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Student aid -- Colorado ( lcsh )
Student aid -- Social aspects -- Colorado ( lcsh )
College dropouts -- Prevention -- Colorado ( lcsh )
Minority college students -- Colorado ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 626-635).
General Note:
School of Public Affairs
Statement of Responsibility:
by Paul B. Thayer.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
40336971 ( OCLC )
ocm40336971
Classification:
LD1190.P86 1997d .T43 ( lcc )

Full Text
AN EXAMINATION OF THE RELATIONSHIP BETWEEN STUDENT
FINANCIAL AID AND RETENTION OF STUDENTS OF DIFFERING
ETHNIC AND INCOME BACKGROUNDS
IN BACCALAUREATE PROGRAMS IN COLORADO
by
Paul B. Thayer
B.A. Williams College, 1969
M.P.A. University of Colorado at Denver, 1985
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Public Administration
1997


1997 by Paul B. Thayer
All rights reserved.


This Thesis for the Doctor of Philosophy
degree by
Paul B. Thayer
has been approved


Thayer, Paul B. (Ph.D., Public Administration)
An Examination of the Relationship Between Student Financial Aid and Retention of
Students of Differing Ethnic and Income Backgrounds in Baccalaureate
Programs in Colorado
Thesis directed by Associate Professor Linda deLeon
ABSTRACT
This thesis tested the proposition that human capital development could be
increased for the least advantaged members of society through financial aid strategies
that affected enrolled students chances of persisting in college through graduation. A
sample of Colorado higher education institutions and students enrolled in those
colleges was identified for the study.
The study utilized logistic regression methods to estimate the effect of various
financial aid factors, including receipt of aid, kind of aid, amount of aid, and
proportion of need met by aid, on persistence. The analysis controlled for student
background and academic preparedness factors.
The study found that financial aid does appear to equalize educational
opportunity by allowing students from low income backgrounds to continue
successfully in academic programs, particularly from the first year to the second. For
the aggregate population in the in-state sample, there was strong evidence that receipt
of aid, amount of aid, and high grant and loan amounts were associated with retention.
When individual institutions were considered, four colleges showed similar patterns of
relationship between financial aid variables and retention. Findings for the thirteen
individual institutions, however, were more ambiguous than for the aggregate sample.
In-state students were found to benefit to a greater degree from financial aid than out-
of-state students.
While financial aid appeared to affect retention when controlling for
background and academic preparedness, academic preparedness emerged as a
powerful factor affecting retention across ethnic and income groups, and exhibiting
differential impacts on ethnic minority groups whose chances of being well prepared
were substantially less than those of white students. Academic preparedness, along
with institutional characteristics, appeared to account for considerable variability in
retention rates and behaviors.
IV


Based on the findings, the study recommends that consideration be given to
increasing the amount of financial aid, particularly in the form of high amounts of
grants, available to low income students. In addition, it is recommended that higher
education institutions become more actively involved in increasing the pool of
academically prepared low income and ethnically diverse students available for
enrollment in higher education.
This abstract accurately represents the content of the candidates thesis. I recommend
its publication.
Signed
Linda deLeon
v


DEDICATION
This thesis is dedicated to the idea of educational opportunity
which continues to be our best chance for a more just society.


ACKNOWLEDGMENTS
Many individuals contributed to the development and completion of this project. I am
appreciative of important support from:
My committee chair, Linda deLeon, whose every contribution was
constructive, energetic, encouraging, and helpful;
My committee members, Julie Carnahan, Peter deLeon, Bob Gage, and Cheryl
Presley, who were willing to read and critique the many pages of the
proposal and thesis;
The Colorado State University Center for Applied Studies in American
Ethnicity, whose support for the project was critical to its completion;
James Zumbrunnen of the Colorado State University Center for Applied
Statistics for statistical expertise and the patience to share his
knowledge;
Cheryl Presley, whose encouragement as my supervisor encouraged me to
make the thesis a priority;
The Colorado Commission on Higher Education for access to the data;
My parents and sisters for always believing in me; and
Most of all my family, Marilyn, Daniel, and Maile, who so willingly gave me
the time, support, and love necessary to bring this project to fruition.


CONTENTS
Abstract ......................................................iv
Dedication ....................................................vi
Acknowledgments ............................................. vii
CONTENTS............................................viii
FIGURES..............................................xxi
TABLES .....................................................xxiii
CHAPTER
1. INTRODUCTION............................................... 1
The Problem............................................... 1
Organization of the Thesis ............................... 5
2. REVIEW OF THEORY AND RESEARCH ............................. 7
Basis in Human Capital Theory............................. 7
Historical Roots of Financial Aid Policy ................ 11
Trends in Student Financial Aid ......................... 13
Increases in Prices.................................. 13
Erosion of Student Financial Aid..................... 14
Redistribution in the Balance of Grants and Loans ... 14
vm


Retargeting Aid Away from Low Income Students . 15
Shifting the Cost Burden of Higher Education..... 16
Questioning of the Financial Aid Strategy ....... 16
Research on the Costs and Benefits of Financial Aid ... 18
Absence of Adequate Research..................... 19
The Combination of Attrition/Persistence Theory
with Investigation into Financial Aid Impact .... 20
Limitations of the Research...................... 24
Areas for Further Exploration of Financial Aid
Impact on Retention ................................... 27
Receipt of Aid................................... 27
Type of Aid...................................... 27
Amount of Aid ................................... 28
Proportion of Need Funded by Aid ................ 28
Differential Impact of Aid on Subpopulations .... 28
Patterns of Effects Across Institutions.......... 29
Potential Contributions of the Study .................. 29
3. RESEARCH HYPOTHESES AND
METHODS OF INVESTIGATION .............................. 31
Primary Research Hypothesis......................... 32
Secondary Research Questions........................ 33
ix


Research Design
34
Assumptions ........................................ 36
The Sample.......................................... 37
Data Available for Analysis................... 38
Institutions in the Sample ................... 38
Students in the Sample ....................... 39
Persistence: The Dependent Variable .............. 40
The Independent Variables......................... 41
Background Factors............................ 41
Academic Preparation.......................... 41
Financial Aid Factors ........................ 41
Statistical Procedures............................ 43
Estimating Effects: the Delta-P Statistic .... 53
Internal Validity ................................ 53
External Validity................................. 57
Conclusion............................................ 58
4. RESULTS AND ANALYSIS ..................................... 60
Organization of the Chapter........................... 60
Tables Summarizing Results for All Students
And for Individual Institutions ...................... 61
x


Examination of Student Retention Behavior
Across Institutions................................... 68
Hypothesized Relationship between the
Variables and Retention .......................... 68
Tables Summarizing the Estimated Effects of
Variables in the Equations........................ 70
Aggregated In-State Students ..................... 73
Aggregated Out-of-State Students ................. 75
Retention to Year 3 and Year 4
for In-State Students............................. 76
Retention to Year 3 and Year 4 for
Out-of-State Students............................. 79
Subgroups of In-State Students ................... 80
Subgroups of Out-of-State Students................ 86
Out-of-State Students by Ethnic Group......... 86
Out-of-State Students by Income Group......... 88
Discussion of Aggregated In-State and
Out-of-State Students............................. 91
Observed Relationships Compared to
Expected Relationships............................ 94
Summary ......................................... 102
Examination of Financial Aid Factors and their
Impact on Retention at Individual Institutions....... 105
Criteria for Judgment ........................... 105
xi


Use of the Tables for Analyzing Student
Behavior by Institution.......................... 110
Extent to which Individual Institutions Conform
to the Expected Model............................ Ill
All In-State Students at
Individual Institutions........................ Ill
All Out-of State Students at
Individual Institutions........................ 114
Subgroups of In-State Students at Individual
Institutions................................... 115
Subgroups of Out-of-State Students at
Individual Institutions........................ 119
Conclusions Concerning the Conformance of
Individual Institutions to the Expected Pattern ... 121
Examination of the Impact of Individual
Financial Aid Factors...................................... 126
Observations Concerning All In-State Students....... 130
Observations Concerning All In-State Students
Retained to Years Three and Four ..................... 132
Observations Concerning In-State Students
by Ethnic Background ................................. 133
Observations Concerning In-State Students by
Income Background..................................... 136
Observations Concerning All Out-of-State Students .. 137
Observations Concerning Out-of-State Students
Returning for Years Three and Four.................... 138
xii


Observations Concerning Out-of-State Students
by Ethnic Background .............................. 139
Observations Concerning Out-of-State Students by
Income Background.................................. 140
Overall Observations Concerning Individual Financial
Aid Factors........................................ 142
Predicted and Observed Relationships of Aid Factors
with Retention .................................... 148
Socio-Economic and Institutional Factors and their
Interplay with Student Response to Financial Aid....... 151
Retention by Ethnicity ............................ 152
Retention by Ethnicity and Income.................. 157
Retention by Ethnicity, Income, and Academic
Preparedness ...................................... 158
Institutional Financial Aid Policy Differences .... 161
Comparison of Aid Proportions with Retention Rate . 170
Summary ........................................... 174
Conclusion ............................................ 175
Primary Research Hypothesis........................ 176
Secondary Research Hypotheses...................... 176
xm


5. CONCLUSION
181
Generalizaility Of Findings.............................. 181
Important Findings ...................................... 183
Variation of Institutional and Student Response..... 183
Expected Findings ................................... 184
Contribution of Receipt of Aid to Persistence .... 184
Positive Relationship between Amount of
Aid and Persistence.............................. 186
Unexpected Findings ................................. 186
Total Amount of Aid ............................. 186
Student Response to Grant and Loan .............. 187
Effect of Aid on Out-of-State Students........... 189
Effect of Aid on Return for the Third
and Fourth Year.................................. 189
Factors Obscuring the Critical Role Played
by Financial Aid......................................... 190
Price Responsiveness of Students from Low Income
Backgrounds ......................................... 190
Lack of Attention to Low Income Students in
Higher Education Institutions........................ 194
Retention, Access, and Academic Preparation.......... 198
Implications for Theory and Research .................... 202
xiv


Financial Aid and Out-of-State Students
203
Differences in Institutional Aid Policies............... 203
Utility of High Grants and High Loans................... 204
Detailed Research on Student Subgroups.................. 204
Interaction of Financial Aid with Institutional Factors 205
Creative Combinations of Financial Aid with
Retention Programs ................................. 205
Measurement of the Payoff from Human Capital
Investment.......................................... 205
Anomalous Findings at Particular Institutions....... 206
Comparisons of Private and Public Institutions..... 206
Implications for Policy and Practice.................... 207
Summary ................................................ 210
xv


APPENDICES
A. Source Tables: Delta-p Values
from Logistic Regression Analysis......................... 212
Part 1: In-State Students................................. 212
All Students......................................... 213
Adams State College ................................. 224
Colorado School of Mines ............................ 232
Colorado State University............................ 239
Fort Lewis College................................... 248
Mesa College......................................... 257
Metropolitan State College........................... 263
University of Colorado at Boulder ................... 271
University of Colorado at Denver..................... 287
University of Colorado at Colorado Springs........... 281
University of Northern Colorado ..................... 295
University of Southern Colorado ..................... 303
Western State College ............................... 311
Part 2: Out-of-State Students............................. 317
All Students......................................... 318
Colorado State University............................ 329
Fort Lewis College................................... 337
University of Colorado at Boulder ................... 343
University of Northern Colorado ..................... 352
Western State College ............................... 357
Regis University (Private)........................... 361
B. Summary Tables Showing Association of Ethnic, Income,
and Financial Aid Factors with Retention to Year Two
at Individual Institutions ............................... 365
Part 1: In-State Students................................. 365
Adams State College ................................. 366
Colorado School of Mines ............................ 367
xvi


Colorado State University............................ 368
Fort Lewis College................................... 369
Mesa College......................................... 370
Metropolitan State College.......................... 371
University of Colorado at Boulder ................... 372
University of Colorado at Denver..................... 373
University of Colorado at Colorado Springs........... 374
University of Northern Colorado ..................... 375
University of Southern Colorado ..................... 376
Western State College ............................... 377
Part 2: Out-of-State Students............................. 378
Colorado State University............................ 379
Fort Lewis College................................... 380
University of Colorado at Boulder ................... 381
University of Northern Colorado ..................... 382
Western State College ............................... 383
Regis University (Private)........................... 384
C. Summary Tables Showing Significant Associations
of Variables Across Institutions ......................... 385
Parti: In-state Students.................................. 385
All Students......................................... 386
Hispanic Students.................................... 399
African American Students............................ 410
Native American Students ............................ 421
Asian Students....................................... 430
White Students....................................... 441
Low Income Students.................................. 452
Medium Income Students............................... 463
High Income Students ................................ 474
xvii


Part 2: Out-of-State Students
485
All Students........................................ 486
Students Retained to Year Three .................... 499
Students Retained to Year Four..................... 506
Hispanic Students.................................... 513
African American Students............................ 524
Native American Students ............................ 535
Asian Students....................................... 546
White Students....................................... 557
Low Income Students.................................. 568
Medium Income Students............................... 579
High Income Students ................................ 589
D. Detailed Tables Presenting Information on Institutional
and Financial Aid Factors................................. 600
STUDENT ENROLLMENT AND RETENTION:
D.l: Ethnic and Income Background for In-State Students
in the Sample, by Institution..................... 601
D.2: Ethnic and Income Background for Out-of-State
Students in the Sample, by Institution............ 602
D.3: Attrition Rate by Ethnicity and Institution:
Retention to Year Two, All Students ............. 603
D.4: Attrition Rate by Ethnicity and Institution:
Retention to Year Two, In-State Students ......... 604
D.5: Attrition Rate by Ethnicity and Institution:
Retention to Year Two, Out-of-State Students ... 605
D.6: Attrition Rate by Ethnicity and Institution:
Students Graduation from Same Institution
Within Five Years, All Students .................. 606
XVlll


D.7: Attrition Rate by Ethnicity and Institution:
Students Graduation from Same Institution
Within Five Years, In-State Students .............. 607
D.8: Attrition Rate by Ethnicity and Institution:
Students Graduation from Same Institution
Within Five Years, Out-of-State Students........... 608
D.9: Attrition Rate by Income and Institution:
Retention to Year Two, All Students ................. 609
D. 10: Attrition Rate by Income and Institution:
Retention to Year Two, In-State Students ............ 610
D. 11: Attrition Rate by Income and Institution:
Retention to Year Two, Out-of-State Students ... 611
D. 12: Attrition Rate by Income and Institution:
Retention to Graduation within Five Years,
All Students......................................... 612
D. 13: Attrition Rate by Income and Institution:
Retention to Graduation within Five Years,
In-State Students ................................... 613
D. 14: Attrition Rate by Income and Institution:
Retention to Graduation within Five Years,
Out-of-State Students................................ 614
ACADEMIC PREPARATION AND INDEX SCORES:
D.15: Index Profiles of Institutions in the Sample......... 615
D. 16: Index Ranges of Students by Institution ............ 616
xix


FINANCIAL AID AWARD LEVELS:
D. 17: Aid Awarded at Selected Percentiles for All Students
in the Sample with Calculated Need ................. 617
D. 18: Aid Awarded at Selected Percentiles for In-State Students
in the Sample with Calculated Need ......................... 618
D. 19: Aid Awarded at Selected Percentiles for Out-of-State
Students in the Sample with Calculated Need .... 619
D.20: Average Total Aid Awarded to In-State Students,
by Institution ..................................... 620
D.21: Average Grant Aid Awarded to In-State Students,
by Institution ..................................... 621
D.22: Average Loan Aid Awarded to In-State Students,
by Institution ..................................... 622
D.23: Average Work Aid Awarded to In-State Students,
by Institution ..................................... 623
D.24: Average Merit Aid Awarded to In-State Students,
by Institution ..................................... 624
D.25: Average Other Aid Awarded to In-State Students,
by Institution ..................................... 625
REFERENCES ............................................................ 626
xx


FIGURES
Figure
3.1 Retention Model Showing Relationship of Independent
Variables to Dependent Variable (Persistence/Graduation) .... 35
4.1 Attrition by Return to Second Year, by Institution ........... 152
4.2 Attrition by Graduation in Five Years, by Institution......... 153
4.3 Attrition Rates by Graduation in Five Years by
Student Type and Ethnic Group ............................... 154
4.4 Attrition Rate Differential: Minority Groups
Compared to Whites 155
4.5 Attrition Rate Differential: Low and Medium Income Students
Compared to High Income Students............................. 156
4.6 Proportion of Calculated Need Funded by Aid
for Low Income, In-State Students............................. 163
4.7 Proportions of Need and Cost of Education
Funded by Aid 166
4.8 Proportion of Low-Income Students Cost of Education
Funded by Total Aid, Grant, and Loan ........................ 167
4.9 Other Aid as a Proportion f the Students Cost of Education 169
4.10 Low Income Students Retention Rate, Compared to
Proportion of Student Costs Funded............................. 171
4.11 Low Income Students Retention Rate and Proportion
of Costs Funded, with Institutions by Index Score.............. 172
xxi


4.12 Low Income Students Retention Rate and Proportion
of Costs Funded by Grant Plus Other Funds, with
Institutions in Order of Mean Index Score...................\ .
xxu


TABLES
Table
3.1 Summary of Dependent and Independent Variables.............. 45
3.2 Steps of the Logistic Regression Analysis................... 48
4.1 Sample Source Table Showing Delta-p Values from
Logistic Regression Analysis ............................... 62
4.2 Stages of Analysis for Aggregated In-State Students........ 67
4.3 Expected Relationship of Financial Aid Variables
to Retention 68
4.4 Summary Delta-p Statistics for Aggregated In-State
and Out-of-State Students: Background and Academic Factors
in Relation to Retention to Year Two................. 71
4.5. Summary Delta-p Statistics for Aggregated In-State and
Out-of-State Students: Financial Aid Factors in Relation to
Retention to Year Two 72
4.6 Summary Delta-p Statistics for Aggregated In-State
and Out-of-State Students: Financial Aid Factors
in Relation to Retention to Years Three and Four........... 76
4.7 Background/Academic Factors for Ethnic Groups (In-State) .... 80
4.8 Financial Aid Factors for Ethnic Groups (In-State)......... 81
4.9 Background/Academic Variable for Income Groups (In-State) .. 83
4.10 Background/Academic Variable for Income Groups (In-State) .. 84
4.11 Background/Academic Variables for Out-of-State Ethnic Groups 86
xxm


4.12 Financial Aid Factors for Out-of-State Ethnic Groups........ 88
4.13 Background/Academic Factors for Out-of-State Income Groups 89
4.14 Delta-p statistics by income level for Out-of-State Students:
Financial Aid Factors for Out-of-State Ethnic Groups......... 90
4.15 Relationship of Independent Variables to Research Hypotheses 107
4.16 Institutions with Four or More Positive Associations........ 113
4.17 Significant Associations for In-State Ethnic Groups......... 116
4.18 Significant Associations for Low and Medium Income Groups 118
4.19 Significant Associations for Out-of-State Ethnic Groups..... 120
4.20 Significant Associations for Out-of-State
Low and Medium Income Groups................................. 121
4.21 All In-State Students: Retention to Year Two................ 131
4.22 In-State Students by Ethnic Group: Retention to Years
Three and Four ............................ 133
4.23 In-State Students by Ethnic Group: Retention to Year Two ... 134
4.24 In-State Students by Income: Retention to Year Two........ 136
4.25 All Out-of-State Students: Retention to Year Two ............. 138
4.26 All Out-of-State Students: Retention to Years Three and Four . 138
4.27 Out-of-State Students by Ethnic Group,
Retention to Year Two ............................ 139
4.28 Out-of-State Students by Income: Retention to Year Two .... 141
4.29 Summary of Significant Associations of Variables with
Retention to Year Two for In-State Students ................. 145
xxiv


4.30 Summary of Significant Associations of Grant and Loan
Variables with Retention to Year Two for In-State Students ... 146
4.31 Summary of Significant Associations of Variables with
Retention to Year Two for Out-of-State Students............... 147
4.32 Summary of Significant Associations of Grant and Loan
Variables with Retention to Year Two for
Out-of-State Students 148
4.33 Comparison of Attrition Rates by Income and Ethnicity:
All Students at All Institutions Attrition through Five-Year
Graduation 157
4.34 Attrition Rates to Five-Year Graduation, Controlling
for Income and Academic Preparation........................... 158
4.35 Composition of Students in the Sample by Ethnicity,
Income, and Academic Preparation.............................. 160
4.36 Estimated Undergraduate Costs at Four-Year Colleges
and Universities, 1990-91 165
xxv


CHAPTER 1
INTRODUCTION
The Problem
The modem form of student financial assistance for higher education was bom
with the authorization of the Higher Education Act of 1965. It represented a strong
endorsement of human capital theory and constituted a statement of confidence in the
public role in higher education.
The Higher Education Act must be re-authorized every five years or so to
continue as law. The reauthorization process begins once again in 1997. More than
thirty years since the law was first enacted, reauthorization of the Higher Education
Act begins in a dramatically different atmosphere. Federal and state budgets are far
more constrained, making resource decisions more problematic. The consensus about
the stature and role of higher education institutions has eroded, making policy choices
more political and contentious.
1


Given this context, the reauthorization of the Higher Education Act will take
place not only with political leadership from a different party, but in an environment of
skepticism and disagreement that contrasts sharply with its birth more than three
decades earlier. Most parts of the act will be the subject of considerable scrutiny, but
none will attract greater attention than Title IV, the part authorizing the federal
financial aid programs.
Concurrent with the national debate on the reauthorization of the Higher
Education Act, states will be enacting their annual budgets. Not so many years ago,
education claimed the largest part of those budgets. Today, however, education
interests must compete with pressures to fund Medicaid, welfare reform, and prisons.
Legislative committees are likely to greet requests for higher education expenditures,
including state financial aid, with demands for greater accountability and proof of
efficiency and effectiveness.
Higher educations response to these new demands at federal, state, and local
levels may be problematic, at least in the case of financial aid. Despite annual
expenditures in the billions of dollars, there are many unanswered questions about
results. For many years, a strong national consensus on the desirability of financial aid
dampened motivation to evaluate seriously the outcomes of such programs. Even as
that consensus deteriorated, however, financial aid programs proved difficult to
2


evaluate. The programs themselves are complex, but even more complicated is the
phenomenon of student enrollment and retention behavior that underlies judgments
about financial aid.
Financial aid was conceived as a means for access to educational
opportunities for economically disadvantaged students and for groups who continued
to face the legacy of racial and cultural discrimination. Over time, however, access
came to be viewed as too limited a goal. To fulfill the promise of greater educational
opportunity, students must not just enter the doors of colleges and universities, but
also leave them with degrees in hand. Retention and graduation took their place
alongside access as the intended outcomes of financial aid, even while little conclusive
evidence was available to confirm that financial aid could play such a role.
The study undertaken here is designed to contribute to the understanding of
the role of financial aid in promoting retention and degree completion. It promises to
contribute to this understanding not only because it is timely hearings on
reauthorization of the Higher Education Act are about to be underway but because
its scope is sufficiently large to make such a contribution.
The study utilizes a large and diverse sample of students and institutions. Over
12,000 Colorado students in the entering class of 1990 are included, as are thirteen
different colleges and universities in the state of Colorado. A exceptionally complete
3


database, the Student Unit Record Database (SURDS) of the Colorado Commission
on Higher Education, permitted examination of a many relevant demographic,
academic, and financial aid variables. The data were timely as well. Most of the body
of research on financial aid was developed before the considerable changes in financial
aid policy in the late 1970s and in the 1980s. The subjects of this study, on the other
hand, entered college in the freshman class of 1990, and if they remained in school,
graduated in 1995 or 1996.
The study asks whether financial aid plays a significant role in increasing
retention of college students with financial need. If so, what is the extent of that role?
In addition, the study examines the mechanisms through which financial aid is
delivered in order to assess whether particular forms of aid, amounts of aid, and
proportions of aid are more or less effective. Finally, the study seeks to determine
whether the effects of financial aid vary among different groups of students in the
population, particularly students of color and students from low income backgrounds.
Indeed, these were the students who were intended to benefit by the original
enactment of financial aid legislation three decades ago.
These questions are important in the development of financial aid policy at
national, state, and institutional levels. The size and diversity of the sample, the level
of detail of data analysis, and the comprehensiveness of the research questions suggest
4


that this study can make a meaningful contribution to the research and policy
discussion of financial aid, and to the larger issues of increasing educational
opportunity and investing in human capital.
Organization of the Thesis
Following this introductory chapter, Chapter 2, Problem Statement and
Review of Theory and Research, discusses the research literature associated with
financial aid and student retention in higher education. The review describes existing
literature that illuminates aspects of the research problem and identifies opportunities
for contributing to the research base. The following chapter, Research Hypotheses
and Methods of Investigation, specifies the hypotheses that provide direction to the
study. The chapter also describes the sample of students and institutions that are the
subjects of the investigation. In addition, the chapter discusses the statistical methods
that are employed as a means of testing the research hypotheses.
Chapter 4, Results and Analysis, reviews in detail the findings of the study.
Results are presented in four sections: analysis of the entire student population in the
sample, analysis of student retention behavior by institution, analysis of financial aid
variables in the study, and exploration of student characteristics and financial aid
policies that may influence retention patterns.
5


The final chapter, the Conclusion, discusses the importance of the findings,
their place in the body of research, and their potential relevance to policy formulation
in the area of financial aid at national, state, and institutional levels.
6


CHAPTER 2
REVIEW OF THEORY AND RESEARCH
Basis in Human Capital Theory
Title IV of the Higher Education Act of 1965 passed into law a policy designed
to make higher education affordable and accessible for students from families whose
resources would otherwise put such education out of reach. Embedded in this policy
were two fundamental assumptions. First, there was an assumption that education
could be the means for breaking down cycles of poverty and social stratification.
Second was the assumption that the entire society would be better off if higher
education were accessible to students from low income backgrounds. Both these
assumptions have roots in human capital theory.
Human capital refers to an individuals skills, talents and knowledge. That
capital can be put to use in work, and can be measured as the value of goods and
services produced (Thurow, 1970). Lester Thurow (1969) described education and
training as the proxy for human capital.
7


Investment in education is a choice. Becker (1964) suggested that an
individual will choose to invest in education so long as the benefit, perceived as higher
wages from increased productivity in the future, outweighs the cost, usually the costs
of securing the education and the costs of foregone earnings or opportunity cost of
alternative activities. In other terms, an individual should invest in education and
training as long as discounted benefits exceed discounted costs, or as long as the
internal return of human capital investment is greater than the rate of return on other
possible investments (Thurow, 1969).
The difficulty is that individuals and families from low income backgrounds do
not have ready access to capital, and so their choice with relation to human capital
investment is restricted. According to Thurow (1969), to the extent that individuals
must rely on their own assets for human capital investment, the poor will fall further
and further behind... (p. 84).
Indeed, human capital development does explain a great portion income
disparity. In contemporary terms, the difference in 1995 average annual income
between males with a only a high school degree and males with a bachelors degree is
$20,904 (Mortenson, 1997a). For females, the difference is $11,568 (Mortenson,
1997b). The difference in average family income between a householder with a high
school degree and a householder with a bachelors degree was $32, 287 (Mortenson,
8


1996). There is a steep, consistent, upward slope describing the difference in annual
income based on educational attainment. While productivity (measured by income) is
clearly related to educational attainment, educational attainment is in turn related to
income background. In 1994, the chances of a student with a family income in the top
quartile earning a bachelors degree by age 24 was nine times greater than the chances
of degree attainment by a student from the bottom income quartile (Mortenson,
1995b).
Human capital investment is not only an issue of private income for private
citizens, however, but a societal issue as well. Human capital has some characteristics
of a pure public good, including democracys need for an educated citizenry, the
relationship of education to social stability, and educations capacity to eliminate
extreme differences in standards of living. Market mechanisms alone may not produce
the amounts of human capital that society as a whole may deem appropriate (Thurow,
1970).
Efforts to measure the outcome of investment in human capital have indicated
that there is both private and social return. In a meta-analysis of contemporary
research on the benefits of higher education, Leslie and Brinkman (1988) estimated a
private rate of return of 11.8 to 13.4 percent, and a social return of 11.6 to 12.1
percent. In a study of investment of state funds in higher education, Cunningham
9


(1994) estimated a modest private return for most students, and a small social return
for the state.
Some have also argued that investment in human capital is necessary even if
certain costs are high. Arthur Okun (1975) describes the relationship between equity
and efficiency as The Big Tradeoff, with efficiency as the cost of equity. While
arguing that efficiency should be societys choice in most cases, he also recognizes that
inefficiency caused by inequality of opportunity can grow at compound interest. In
such cases, including the inadequate development of the human resources of the
children of the poor, and the differential entry to college by persons of similar ability
but unequal incomes, the choice should be made to increase human capital, serving
the interests of equity and efficiency at the same time (Okun, 1975, p. 77).
In fact, investment in human capital is often the publics preferred means of
addressing income inequity. Given the choice between direct transfer payments to
individuals and families requiring similar expenses every year, or investment over a
limited time in human capital, the public is inclined toward the latter (Thurow, 1970).
The investment alternative not only has a finite period over which expenses are
incurred, but it carries the potential of later payoffs, including the pride and
independence of the individuals in whom the human capital is developed.
10


Society may choose to increase the investment in human capital by providing
incentives to individuals or institutions. In the case of financial aid, the incentive is a
discounting of the cost of education. Where the discounting is targeted upon those
whose opportunities are most circumscribed by socio-economic circumstance, the
goals of the investment may include not only the private benefit of the individuals
involved, but also the societal benefits of increased equity of income and opportunity,
society stability, and increased productivity in society as a whole.
Historical Roots of Financial Aid Policy
Financial support for higher education is an American tradition. Beginning
with colonial times, the federal and state governments have been deeply involved in the
financing of colleges and universities. The Morrill Act of 1862 was a striking example
of making education affordable to the common man through subsidies to state
educational structures (Fenske & Huff and Associates, 1983). Direct student financial
aid, however, is a more recent phenomenon.
Student financial aid has served multiple purposes over the last five decades
(Jensen, 1981). Hartle (1985) identifies these as rewarding military service and easing
the return of veterans to the labor market as in the G.I. Bill of 1944, and encouraging
students to pursue careers in mathematics, science, and foreign languages as in the
11


National Defense and Education Act of the late 1950s. Finally, and most relevant to
our concerns, financial aid has been used as an instrument to break the cycle of
poverty and increase access to higher education programs.
In terms of the equity purposes of financial aid, J.S. Hansen (1984) identifies
the establishment of the College Scholarship Service in 1954 as the beginning of a
remarkable consensus that lasted thirty years. Not long after, the Higher Education
Act of 1965 signaled a major expansion of the federal financial aid effort, and also the
beginning of a major effort to help individuals from low income families attend college
through educational grants.
In the 1970s, there was considerable debate about the use of subsidies to
increase access by low-income students to higher education. Many voices objected to
such a strategy for reasons of economic inefficiency (Dresch, 1977; Hansen, 1971;
Johnson, 1971; Shultz, 1971; Windham, 1971). Still, political forces worked in favor
of the reauthorization of the subsequent Higher Education Act in 1972 with student
financial aid as a main feature (Gladieux, 1983). The policy debate dwindled markedly
after that time. Successive reauthorizations of the Higher Education Act expanded this
effort, and gradually broadened eligibility to include the middle class. The cost of this
policy grew as well. In 1994-95, over $26 billion was provided to students in federal
loans alone. With increasing concern over government spending and federal deficits,
12


many are now calling into question the appropriations, values, mechanisms, and results
of federal financial aid policy (Gladieux and Hauptman, 1995).
Increased sophistication in the measurement of the equity and efficiency
impacts of financial aid did not accompany growth of financial aid programs. The
absence of measurement may be partly related to the culture of higher education itself,
since goals are vaguely stated (Bowen, 1980), and definitions of quality are more
related to institutional resources than to educational outcomes (Dunn, 1992; Tan,
1986; Gilmore and To, 1992). It is also attributable to the growing complexity of the
aid process as well as to the rapidly changing procedures, funding levels, and targets
of financial aid policy.
Trends in Student Financial Aid
A number of trends have interacted in such a way as to affect the student
subjects in this study far differently than the students in earlier days of the aid
programs.
Increases in Prices
Beginning around 1980, after a long period of relative stability, the rate of
increase in the price of higher education began doubling the rate of increase in the
13


consumer price index (Lewis, G., 1989; Gladieux, 1995). While the rate of increase in
price has moderated somewhat in recent years, it continues to be a topic of intense and
contentious debate in the public policy arena. The precipitous increase in price made
higher education less affordable for low and middle income students at the same time
as the economy made a college education increasingly necessary.
Erosion of Financial Aid
While the cost of attending college in the 1980s and early 1990s out-paced
both inflation and increases in family income, student financial aid failed to keep up
(Odden and Massey, 1992). Private institutions tended to adjust their prices and
institutional aid policies to compensate for these losses, but public institutions did not
(St. John, Oescher, and Andrieu, 1992; McPherson and Schapiro, 1995). Not
surprisingly, the erosion of aid had the greatest impact on those at the lowest incomes
(Mortenson 1995a).
Redistribution in the Balance of Grants and Loan
It has always been an assumption that grants (originally the Basic Education
Opportunity Grant, and later the Pell Grant) were the foundation of the aid program.
Federal aid began with a small loan program and a large grant program. Between
14


1980-81 and 1984-85, however, the value of the Guaranteed Student Loan (GSL) and
PLUS loans increased in constant dollars by 10.8%, while the value of Pell, SEOG and
SSIG grants declined by 5.7%, 17.2% and 18.8%, respectively (Andrew and Russo,
1989). Grants grew from 40% of federal aid in 1963-64 to 76% in 1975-76, but
subsequently shrank to 47% in 1980-81. By 1988-89 grants were only 31% of federal
aid (Lewis, G., 1989). Since the mid-1970s federal student loans have increased from
about one-fifth to two-fifths of all available student aid. Gladieux and Hauptman
(1995) observe that the loan program has grown to five times the size of the very Pell
Grant program that was meant to be the foundation of the student financial aid system.
Re-targeting Aid Away from Low-Income Students
Beginning in 1965, students from low income families were the primary focus
of the federal student aid effort. The Middle Income Assistance Act of 1978 marked
the beginning of broadening eligibility for aid (Jensen, 1983). As a result, more and
more financial aid dollars were spent to benefit students from middle income
backgrounds. At the same time, middle income students were feeling an increasing
financial squeeze from the effects of the rapidly escalating cost of attendance, for
which even new financial aid dollars (primarily in the form of loans) could not
compensate (Morning, 1993).
15


Shifting the Cost Burden of Higher Education
Mortenson (1995a) notes a shift of the costs of financial aid from society to
students and families as aid declines. At the same time, he notes another shift of the
costs of operating institutions amounting to $14.2 billion in higher education costs
from taxpayers to students in the form of rising tuition rates. Because most of the
borrowers in the loan programs are students, the growth in borrowing also represents
a shift of the cost of education from one generation to another. While parents used to
bear primary responsibility for these costs, students now bear an ever-increasing share.
The long-term effects of these changes remain to be seen. It is not clear whether
todays graduating students, many of whom will be making loan payments far into the
future, will be in a position to save for their own childrens future education (Baum,
1996).
Questioning of the Financial Aid Strategy
By the mid-1980s, low- and middle-income students were having great
difficulty affording college. However, as Michael Mumper (1993) notes, the problems
of affording college extended well beyond those students and their families. The
federal government, faced by a mammoth national debt, was finding it impossible to
expand student aid sufficiently to keep up with tuition inflation or even to fund the
16


Federal Pell Grant program at authorized levels (Frances and Harrison, 1993).
Similarly, state governments, also under severe budgetary pressures, were facing
choices among difficult alternatives: cutting back higher education funding, reducing
other essential services, or raising taxes. Finally, budgetary pressures were forcing
many colleges to both raise tuitions and reduce programs and staff in order to pay their
bills (Mumper, 1993, p. 27).
The combination of factors described here have escalated the erosion of
consensus about financial aid policy. Financial pressures on students, parents,
educational institutions, and governments have translated into discontent heard by
legislators and other policy makers. Congressional skepticism is apparent in the
appropriations process. The FY 96 appropriation originally proposed by the majority
party in the 104th Congress would have dramatically reduced many educational
programs, including financial aid. While the eventual appropriations were far less
draconian than first proposed, the initial proposals reflect a fundamental questioning of
the value of investing in students and higher education through financial aid.
Unfortunately, such questioning is occurring without the benefit of policy examination
and debate. The discussion has been driven by budget, rather than by detailed policy
analysis, sound research, or thoughtful public philosophy (Merisotis, 1995).
17


Recent trend studies only intensify skepticism about financial aid investments.
Frances and Morning (1993), for example, after examining enrollment patterns over
time concludes that after more than two decades of need-based aid...the vast
differences in college-going rates of students from higher income families and lower
income families remain (p. 3). Indeed, Mortenson (1997b) reports that Asians and
whites graduate from college at higher rates than do blacks, American Indian and
Hispanics... [and] students from high income families graduate from college at higher
rates than do students from low income families (p. 9). If these observations are
correct, it is reasonable to ask whether financial aid truly works.
Research on the Costs and Benefits of Financial Aid
Over a long period, there was a virtual absence of policy research informing
student aid policy discussion. Larry Gladieux (1983) has chronicled the use, misuse,
and non-use of policy research in student aid policy-making. After years of
complacency, the financial aid community was shaken by publication of a paper
(Hansen, W.L., 1983) questioning the effectiveness of financial aid. This critique of
financial aid coincided with Reagan Administration efforts to further reduce financial
aid. Only at that point did researchers turn serious attention to assessing the impacts
of student aid policy. Even so, as shall be apparent below, important questions remain
18


to be answered about a financial aid policy that has cost billions of dollars over three
decades (S42 billion in 1993-94 alone, according to the College Board, 1994), yet has
not been subjected to adequate analysis or debate.
Absence of Adequate Research
Mow and Nettles (1989), in their comprehensive review of literature on
Minority Student Access to, and Persistence and Performance in, College, take
particular note of the
paucity of empirical research into the relationship between financial aid
and student performance and persistence, particularly among
undergraduate minority students. What is available is not very
illuminating. Much of the research either simply describes results of
surveys, administrative procedures, and problems, or investigates
financial assistance without looking at other economic variables (p. 70).
The problems with existing research, they note, include difficulties associated with
variation in the distribution of aid from institution to institution, the complexity of how
financial aid is packaged, and the absence of accurate classifications of students with
different kinds of aid (p. 72).
Leslie and Brinkman (1988) observe that there are three possible ways to study
the impact of student aid: opinion surveys, trend studies, and econometric studies.
Many studies of financial aid fall in the first two categories. Trend studies, which limit
19


analysis to the relationship between financial aid and enrollment or persistence, have
been used to make generalizations about the impact of financial aid. (See for example:
Boyd, Fenske, and Maxey, 1978; McCreight and LeMay, 1982; Nichols, 1980;
Heamdon, 1984.) Because trend studies do not take into account the many other
variables that contribute to students decisions to persist in or withdraw from higher
education, they have not been able to distinguish the particular role that financial aid
may play.
Studies using the econometric approach utilize multivariate statistical
techniques, and offer the advantage of the opportunity to control for the influence of
intervening variables in a complex model (Leslie and Brinkman, 1988). Examples of
econometric research include Cabrera, Stampen, and Lee (1990), St. John, Kirshtein
and Noell (1991), Somers (1994), and Whitaker and Pascarella (1994).
The Combination of Attrition/Persistence Theory
with Investigation into Financial Aid Impact
There is a great deal of literature on college persistence and attrition, including
both theory and empirical research. This literature provides useful multivariate models
for student enrollment and persistence behavior. Unfortunately, most of that literature
does not include either financial aid or ability to pay as variables. (See for example,
Munro, 1981; Pascarella, Smart, and Ethington, 1986.)
20


Much research on retention has generated from the field of sociology, and the
theories of Vincent Tinto (1975, 1987). Tintos theories of academic and social
integration continue to guide most of retention research. The power of his model lies
in the comprehensive set of variables, including background, socioeconomic status,
academic ability, motivation, and institutional factors, all of which contribute to the
students decision to remain engaged or depart from higher education. Studies of
financial aid need to take into consideration such factors if the part of financial aid is to
be explained. Tinto himself, however, has not considered financial aid a primary factor
in persistence, concluding instead that finances play at best an indirect role in students
withdrawal decisions (Tinto, 1987).
Others, however, began to combine broader retention theory with the concern
for financial aid impacts. As they did so, research improved immeasurably. Astin
(1975) developed a theory of persistence that emphasized student fit with the
environment of the institution. His study found that work was strongly associated and
grant somewhat associated with persistence, while loans had a negative effect. Jensen
(1981) utilized Tintos attrition theory in designing a study of financial aid recipients,
eligible non-recipients, and ineligible non-recipients and found that financial aid made a
small contribution to retention when issues of socio-economic status and academic
21


performance in high school were taken into account. These studies were far more
powerful than those that did not utilize broader theories of attrition or retention.
Additional researchers used persistence theory in their studies of financial aid
and found that financial aid had a demonstrable influence on student success.
Voorhees (1985), using fewer variables, found that all types of financial aid programs
played a major role in persistence. Terkla (1985) did not distinguish among different
types of aid nor aid packages, but found that aid was the third strongest factor directly
impacting student decisions to withdraw, and that controlling for other factors, aided
students were more likely to complete college. Stampen and Cabrera (1986) found
that financial aid has both direct and indirect effects on persistence.
Murdock (1987), in a meta-analysis of forty-six studies of financial aid and
persistence found that for all studies, financial aid demonstrated a significant effect on
persistence. Among the relatively few studies that took academic preparation into
account, however, the effect dropped to inconsequential levels. In fact, evidence is
mixed on the overall effectiveness of aid, as well as the effect of types or combinations
of aid on persistence (Pascarella and Terenzini, 1991).
St. John (1989), St. John and Noell (1989), and St. John, Kirshtein, and Noell
(1991) found that all forms of aid are effective in promoting either enrollment or
persistence. St. John (1989) found that this was true over time, testing student
22


samples at three different points in time and under conditions of different financial aid
policy. St. John and Noell (1989) discerned a growing negative impact of loans in
attracting students to enroll as loans increased in student aid packages. St. John,
Kirshtein, and Noell (1991) found that different packages of aid types were effective at
different points in students college careers.
Edward St. John has become not only a strong advocate for the combination of
persistence theory with financial aid impact research, but a leading researcher in this
field. In 1992, he proposed a model for use by institutions to assess the effectiveness
of student financial aid at the institutional level using available data, so that colleges
could become involved in locally-relevant research.
Cabrera, Stampen, and Hansen (1990) combined persistence theory with ability
to pay theory. Their findings were that ability to pay had both direct and indirect
effects on persistence decisions, and moderated academic and social integration.
Hanson and Swann (1993) found that financial aid had an effect on persistence, but
that the effects varied with different levels of student preparation. Somers (1994)
found that higher amounts of aid are more effective, whether amounts of loan, of
grant, or of total package.
23


Limitations of Research
Nevertheless, research is still limited. Many prominent researchers (Jensen,
1981; and Cabrera, Stampen, and Hansen, 1990, for example) comment on how little
is really understood of the relationship between financial aid and college success. In
particular, there are problems in research that derive from limitation of the samples,
from the currency and time-frame of studies, and from the complexity of financial aid
processes.
Of the studies cited above, six (Nichols, 1980; Jensen, 1981; McCreight and
LeMay, 1982; Vorhees, 1985; Somers, 1983 and 1984; Luan and Fenske, 1996) used
samples drawn from a single institution. The ability to generalize from single-
institutions is extremely limited, since institutions have varying packaging policies and
may have student populations that reflect local situations (Pantages and Creedon,
1978, Cabrera, Stampen, and Hansen, 1990). Studies using samples from a single
state have been limited as well, either by their use of a narrow population sample
(Boyd, 1978) or the use of one particular sample of students in a single state system
(the University of Wisconsin System in Stampen, Cabrera, and Hansen, 1990).
Several researchers have utilized national samples to achieve greater
generalizability of results. The national samples include the High School and Beyond
data sets (HSB-72, HSB-80, HSB-82) (Cabrera, Stampen, and Hansen, 1990; St.
24


John, 1989b; St. John, Noell, and Kirshtein, 1991), the National Longitudinal Study
(NLS-72, 80, 82) (Munro, 1981; St. John and Masten, 1990; St. John and Noell,
1989; Terkla, 1985), or the National Postsecondary Assistance Study (NPSAS-86)
(Flint, 1994; St. John, Oescher, and Andrieu, 1992). These studies benefit from the
national scope of the sample and from the depth of student- and family-reported
background data. However, there are missing data in the sample, and the self-reported
data may be less than completely accurate. In particular, the data concerning financial
aid are entirely self-reported and do not contain details of the financial aid calculation.
The NPSAS samples include a high level of financial aid detail, but lack background
on high school performance.
Many of the studies are limited by the date of the study or the limited time-
span. The data used in the single-institution studies, with the exception of Somers
(1993 and 1994), are now fifteen to twenty-five years old, not sufficiently current to
illuminate todays situation given the rapidly changing nature of financial aid policy
and student response. Except for Somers, none of these studies uses data from the
last ten years, during which the dynamics of student borrowing, family income levels,
minority enrollment trends, labor-market trends, and many other factors have changed
dramatically. The same is true for the state and national samples. Data for these
studies were generated from the 1970s to 1987, limiting considerably the applicability
25


of findings to the present. A new national database has been initiated, but the National
Educational Longitudinal Study of 1988 (NLS-88) will not contain college graduates
for another four years, i.e., 1999 or beyond (U.S. Department of Education, 1988).
The frequency and duration of the sampling is also limited in many cases.
Many of the studies address only within-year persistence, or year-to-year persistence
from the first to second year. With the exception of Munro (1980), Terkla (1985),
Vorhees (1985), or St. John (1989), studies have adopted relatively short time-frames
and are therefore not able to demonstrate the effects of financial aid or other factors
over the period of time required for students to graduate.
Finally, the studies are limited in their ability to account for the complexity of
student financial aid processes and information. As noted previously by Mow and
Nettles (1989), this has been one of the major problems with financial aid research.
Students eligibility changes from year-to-year and even semester-to-semester. It is
difficult to track these changes, and also difficult to categorize students in terms of
their aid packages, except according to gross categories such as aided students
versus non-aided students; or students receiving grants versus students not
receiving grants. Although several researchers (Jensen, 1981; Hanson and Swann,
1993; and Somers, 1994; for example) express the suspicion that many aided students
in public institutions actually had high levels of unmet need that precluded significant
26


effects from their meager financial aid, none of the studies was able to confirm this
suspicion.
Areas for Further Exploration and Research
Many questions remain to be answered with respect to the impact of financial
aid on students decisions to remain in higher education rather than to withdraw.
Several of these are identified here.
Receipt of Aid
Do those who receive aid, presumably students at lower income levels, persist
at lower rates, making financial aid a particularly costly investment? If so, is the
higher attrition attributable to students income background or to differences in
academic preparation? If academic preparation is accounted for, does persistence
increase? Finally, if academic preparation is controlled, is receipt of aid a negative,
neutral, or positive factor in relation to retention?
Type of Aid
There are a variety of instruments for the delivery of aid: grants, loans, work
and work-study programs, merit-based aid to needy students, and other aid,
27


including private and institutional scholarships. Are some of these more associated
with retention behaviors than others? Are others related more to attrition than
retention?
Amount of Aid
Is it possible that aid at low levels is marginally effective, while aid at high
levels is more effective? Given the erosion of financial aid noted earlier, questions
concerning effectiveness of aid in terms of amount are particularly critical.
Proportion of Need Funded by Aid
Questions have been raised by research about the adequacy of aid in relation to
students need (Richardson and Bender, 1987). A different way of addressing the
amount of aid question is to measure the extent to which students calculated need is
met by aid.
Differential Impact of Aid on Subpopulations
Financial aid was conceived as a policy to increase equity. Are the effects of
aid consistent for students who demonstrate financial need but have different levels of
28


income? Are the needs of particular ethnic groups sufficiently different that there are
differential benefits from various aid strategies (Castle, 1993)?
Patterns of Aid Effects Across Institutions
The limitations of single-institution studies have been noted earlier. Will
patterns appear across contrasting institutions, or will response to aid be unique at
each individual institution? If so, are the differences attributable to variations in aid
policy or rather to other characteristics and cultures of those institutions?
Potential Contributions of the Study
The study undertaken here has the potential to contribute to the body of
knowledge and research about financial aid. First, the research is based on an
econometric model, taking into account multiple variables. Second, financial aid
factors are understood within a larger model based on retention theory. Third, the
sample includes a fairly large number of students and a diverse group of public
institutions. Fourth, the data are relatively contemporary, and are likely to reflect
student responses to nearly-current financial aid policy. Fifth, the database includes
detailed and reliable information on many variables, including those relating to student
background, academic preparation level, and financial aid package.
29


Each of the elements just described were identified in the review of research as
important to the study of the relationship of financial aid to retention. In addition, the
research model undertaken is based on information that is generally available to
institutions, making replication of the study for institutional policy-makers practical, as
recommended by Edward St. John (1992) and Vincent Tinto (1990).
The next chapter describes the theoretical model and research methods through
which these elements will be incorporated into the study.
30


CHAPTER 3
RESEARCH HYPOTHESES AND METHODS OF INVESTIGATION
The previous discussion reviewed the history of student financial aid, and
found that the initial purpose of aid was that of access to higher education for
individuals and groups with less access to capital and, therefore, less access to the
benefits of higher education. If financial aid were provided to needy students, it was
thought, the access problem would be solved.
Over time, however, the access objective was found by many to be too
limited, since students who had access to higher education did not necessarily
complete a degree and reap the associated benefits. This critique of higher education
asserted that institutions were luring unqualified students with the promise of student
aid, but that students from economically disadvantaged backgrounds, including a large
proportion of students of color, were not emerging from the educational pipeline at
the other end. This critique called for a new definition of effectiveness, one that
measured success in terms of degree achievement. The persistence or retention of
31


students from one semester to the next constitutes the incremental steps students
complete along the way to their degree.1
In the context of this new criterion for educational opportunity, it was unclear
what role, if any, financial aid actually played in persistence and graduation. As noted
in the earlier review, research on this point was virtually absent for decades, even after
the amount of public funds committed to financial aid swelled to the billions of dollars.
Retention literature, usually sociologically-based, tended to ignore financial aid as a
factor. The few studies which considered financial aid either failed to take retention
research into account, thereby weakening their explanatory power, or produced
contradictory conclusions. Present research continues to be ambiguous on this point.
Primary Research Hypothesis
This study addresses the question of whether, and to what degree, student
financial aid contributes to the persistence of students from economically
disadvantaged backgrounds in their quest for bachelors degrees.
H,: The receipt of student financial aid by students with calculated
financial need will contribute to those students persistence
from one year in college to the next.
while
"Persistence best describes student behavior in continuing toward degree completion,
retention best describes institutional strategy for increasing student persistence. Because
they are often used interchangeably in the literature, however, they are also used interchangeably
here.
32


Secondary Research Hypotheses
If there is a relationship between financial aid and persistence toward degree
completion, the study will proceed to draw finer distinctions with respect to the
delivery of student aid. The following questions are associated with this line of
inquiry:
H2a: As the total amount of financial aid increases, persistence will
increase as well for students of equivalent family incomes.
H2b: Grant funds, that is, those financial aid dollars that do not need
to be repaid by students, will be more effective in increasing
retention than other forms of aid.
H2c: The greater the proportion of the cost of education that is
funded by financial aid, the greater will be the resulting
persistence.
In addition, the study will examine whether all students appear to benefit in the
same way, or whether certain groups of students benefit to a greater or lesser degree
than others. Of greatest concern will be those groups of students who have been most
disadvantaged in society and who have been historically most underrepresented in
higher education: students of color and students from the lowest income backgrounds.
H3a: There will be differences in the benefits of financial aid
according to differences in ethnic background, with white
students demonstrating greater benefits than students of color,
other factors being equal.
H3b: Students from the least advantaged economic backgrounds will
benefit from financial aid less than students from more
advantaged, but still needy, backgrounds.
33


H3c: Students from the lowest income groups, and those from ethnic
backgrounds who traditionally experience the greatest
discrimination, are likely to experience the greatest benefit from
grant aid and the least benefit from loan aid.
Finally, the study will examine the institutional policies reflected in the
packaging patterns of financial aid in an attempt to conclude whether institutional
choices result in greater or lesser effectiveness in encouraging persistence.
H4a: There will be variation in financial aid packaging from
institution to institution, and those institutions that provide
greater levels of aid, particularly higher proportions of need met
by grant aid, will be more successful in retaining students from
socio-economically disadvantaged backgrounds.
Research Design
Financial Aid as an Element in a Retention Model
Financial aid must be studied within the context of the system of factors that
together influence student retention behavior. Drawing from the many sources
identified in the previous chapter, this study views retention according to model
depicted on the following page.
34


Figure 3.1: Model Showing Relationship of Independent Variables to
Dependent Variable (Persistence/Graduation)

Financial Aid Factors

Financial Aid:
Receipt, Type, Amount,
Adequacy________________
Family Income
X,
X
y
Ethnic Background f
Background, SES Factors
A H.S. GPA, Class
Rank, ACT/SAT
r| Scores
w Academic Preparedness
~ Factors
Persistence/
Graduation
(Bachelors
Degree)
Outcome
Background factors, such as gender, ethnicity, and economic background exert
fundamental influences on students likelihood of entering and persisting in college.
In addition, parent education, parent aspirations for their childrens education and
career, and student aspirations all play a part in student preparedness and persistence.
Academic preparedness, in turn, is a significant factor related to student success and
persistence in college.
Of the background and academic preparedness factors identified above, several
will not be examined in this study because they are not available in the data. These
include parent education, parent career, and parent aspirations for their children. Also
unavailable is information on student aspirations for their own education and career.
35


Once in college, a multitude of factors have an impact on student persistence,
including financial resources, peer relationships, relationships with faculty, involvement
in campus activities and support services, and degree of certainty about the choice of
major. The data selected for the study do not contain information about any of these
factors except financial aid. Fortunately, the data do provide a wealth of detail about
the calculation upon which financial aid is based, and about the financial aid package
actually received by the students.
In summary, the study will take into account several of the background and
socio-economic factors that have been established in the research literature as major
influences on student persistence. It will also take into account the academic
preparedness level of students entering college. After taking these into account, the
study will examine the residual effect of student financial aid on student persistence
behavior in comparison to students who receive no financial aid.
Assumptions
A number of assumptions have been made concerning the data and its meaning.
1. In general, the calculation of need through Federal Methodology
(National Association of Student Financial Aid Administrators, 1996)
will be assumed to represent an accurate assessment of the students
and familys economic condition. There is certainly room for some
debate on this issue. For example, some have taken issue with the fact
that calculated need stops artificially at zero for the lowest income
families; that assets are not fully taken into account for some families;
and that independent students are favored in the need analysis process.
For the purpose of this study, however, it is assumed that the need
36


analysis process does generally identify appropriate students as having
financial need, and that those with higher amounts of calculated need
are in fact more economically disadvantaged than those with lesser
amounts.
2. The study also operates on the assumption that those who did not apply
for financial aid, and so were not subject to the need analysis process,
were not economically disadvantaged. Certainly, there may be some
students who did not apply for financial aid who are economically
disadvantaged, and yet found a way to enroll in four-year higher
education programs. Given the high cost of a bachelors degree
program, however, it is a safe assumption that these exceptions will be
extremely rare. It is nearly impossible for a low-income student to
attend a four-year program without significant financial assistance.
The Sample
The Colorado Commission on Higher Education (CCHE) keeps records on all
students applying to, and entering, higher education institutions in Colorado. Records
were obtained from the CCHEs Student Unit Record Database for student cohorts
entering higher education in the years 1987 through 1991. Although it was originally
intended to use all five cohorts in the study, it was discovered that only the class of
1990 had complete records for all semesters for the full five years (fall of 1990 through
spring of 1995). For other cohorts, there were missing data for one or more semesters
over the five years necessary to track persistence through graduation from
baccalaureate programs.
It was also intended to include the states three major private four-year
institutions in the study. It was found, however, that only Regis University, an urban
sectarian school, had complete financial aid and enrollment data for the entire period
37


under investigation. This presented a serious limitation. It was theorized that public
institutions had not adjusted their pricing policies in response to the erosion of
financial aid over the past two decades, while private institutions had done so to a far
greater degree (McPherson and Schapiro, 1995). It turned out to be impossible to test
this theory, since two of the major private institutions, and those with the greatest
financial resources, had to be excluded from the study for lack of complete data. The
one private institution that was included was insufficient to test the theory, because its
limited resources probably cause it to act in a manner more characteristic of public
institutions than wealthier private institutions.
Data Available for Analysis. An enrollment extract file contained individual
student records for demographic data, term-by-term enrollment and performance data,
basic financial aid data, and graduation and degree attainment data. A separate file
provided detail on financial aid packages and family income. The files were
reformatted and combined into one large file for analysis.
Institutions in the Sample. The study was designed to address students who
entered baccalaureate programs. There are twelve public four-year institutions in
Colorado, and three major private institutions. All twelve public institutions were
included in the study. As noted previously, financial aid and other records were
complete for the required five-year period for only one private institution. The final
sample, then, contained records from thirteen institutions.
38


Students in the Sample. For students enrolled in the thirteen four-year
institutions in the sample, the following selections were made:
SELECTION CRITERION RATIONALE
First-time students
Four-year program type
Freshman level
Full-time status at initial enrollment
Ethnicity: Black, non-Hispanic; American
Indian or Alaskan Native; Asian or Pacific
Islander; Hispanic; and White, non-
Hispanic
Age 16 through 49 at initial enrollment
Students with admissions index scores
Standardize for experience of first-time
students without prior collegiate
experience
Eliminate two-year, vocational, or
certificate programs
Eliminate students entering with additional
credits
Eliminate part-time students
Eliminate non-resident aliens, and those
who did not indicate an ethnic group
Eliminate extremes of age
Requirement to be able to standardize for
academic preparation
A follow-up study might focus on students excluded from this study, such as
students in two-year schools, non-traditionally aged students, part-time students,
transfer students, or others. Given the complexity of this study to begin with, the
choice was made to focus on first-time, full-time, freshman in four-year programs.
Although a wide range of ages was included, the need to exclude students without
index scores eliminated the majority of non-traditionally-aged students.
39


When the institutional and student selections were made, the original number
of 44,303 students was reduced to 13,608 students who were on record at four-year
schools and who were first-time, full-time freshmen enrolled in baccalaureate
programs in the fall of the first year. Of these, 12,468, or 91.6 percent, were selected
for the sample on the basis of age and presence of an index score. This number was
sufficient to conduct most analyses for most institutions; however, some analyses of
subgroups by income or ethnicity were not possible.
Persistence: The Dependent Variable
For the purpose of this study, student success was measured in terms of the
students return to college the following year. For example, a student enrolled in the
fall semester freshman year is counted as retained when they enroll in a fall or spring
semester of the following year. (The choice was made to use either fall or spring
semester, since many students stop out for at least a semester.) Retention beyond
the first year was measured similarly. A student who was enrolled in the fall of year
one (1990) was counted as retained through the second year if they met the criterion
for persistence in year one, were enrolled in fall or spring of year two, and re-enrolled
in the fall or spring of year three (1992). The definition of retention through the third
year follows the same pattern.
40


The Independent Variables
Background Factors. Four factors were selected to represent student socio-
economic background: gender, ethnic background, family income, and age. To
represent academic preparedness, the admissions index score was used.
Academic Preparedness. The admissions index score was adopted by the
Colorado Commission on Higher Education in 1986, and was mandated for use by all
Colorado public higher education institutions. It is a composite of the students grade
point average in high school at the time of the admissions decision, and the students
score on the ACT or SAT standardized college entrance examination. The Colorado
Commission on Higher Education has conducted analyses that lead to the conclusion
that the admissions index is a valid predictor of student success in college. (Colorado
Commission on Higher Education, 1986; Chisholm, 1992) Four-year public higher
education institutions are required to admit a proportion of students at or above a
particular index threshold, and these institutions are ranked by selectivity according to
their required index standard.
Financial Aid Factors. Financial aid was represented in a variety of ways in
order to measure different aspects of aid policy or aid delivery. Financial aid was only
measured when awarded to students with a calculated need. Calculated need is the
difference between the student budget or cost of education (estimated by the
institution) and the family resources (including both student and parent resources).
Calculated need is a figure that is only available for students who have applied for
41


financial aid. Students who applied for aid but whose resources exceeded the
estimated budget showed negative need, and were not counted as having received aid.
Financial aid was operationalized by eight different representations. Each
representation was used to the exclusion of the others; that is, the various
representations were substituted for one another systematically in order to be able to
evaluate each one. The representations utilized in the study included:
Any Aid: whether the student received any aid or no aid
Aid Type: whether the student received aid in the form of grant, loan,
work, merit, or other; or whether they received a combinations of these
Aid Amount: the total amount of the aid package received by students
with calculated need
Aid Amount by Kind of Aid: the amount of aid received in each form
(grant, loan, work, merit, or other)
Proportion of Aid: the proportion of calculated need met by the total
aid package
Proportion of Aid by Kind of Aid: the proportion of calculated need
met by individual forms of aid (grant, loan, work, merit, or other)
High Aid Amount: exceptionally high amount of total aid received
(75th or 90th percentile)
High Aid by Kind of Aid: exceptionally high amount of aid received in
the form of grant and in the form of loan (75th and 90th percentile)
42


Statistical Procedures
For each institution, baseline data was collected. This data included:
the number of students at the institution, by ethnicity, gender, and
income level. (Income level is only available for students who have
applied for financial aid; others are assumed to be of higher income.)
index profile of the institution, by ethnicity, gender, and income level
number of students receiving aid, by ethnicity, gender, and income level
retention rates of students by ethnicity, gender, income level, and by
receipt or non-receipt of aid, and receipt of various types and amounts
of financial aid
This data was gathered by generating frequency tables, descriptive statistics, and/or
cross tabulations.
To assess the relationship between the dependent and independent variables,
regression analysis was employed. Logistic regression was the particular method
appropriate for this study because the dependent variable of retention has only two
possible outcomes: was the student retained or not? Logistic regression not only
produces a measure of association with the outcome, but also an estimation of the
probability of the event (retention) occurring. (Norusis/SPSS Inc., 1990)
Twenty-eight data files were created. For in-state students, one file was
created for all students enrolled at any of the thirteen institutions under consideration,
one file for each of the twelve public institutions in the study, and one file for the lone
private institution. Similar files were created for out-of-state students. A series of 116
43


regression equations was written, each testing a different factor or set of factors
theorized to affect retention. The entire series of regression equations was then run
for the data in each of the twenty-eight files.
With one exception, forward stepwise selection method was used to control
the entry of factors into the retention model. Removal of factors from the model was
based on the likelihood-ratio statistic based on the maximum-likelihood estimates.
(Norusis/SPSS Inc., 1990, pp. 15-18) The exception was first the logistic regression
procedure used for each file. In this case, the forced entry method was used to enter
all variables in a single step without any controls. The results of this forced entry
analysis were used for reference in the subsequent forward stepwise analyses.
The logistic regression model followed this format:
e1
Prob(event) =--
1 +e z
where e is the base of the natural logarithms, approximately 2.718, and Z is the
linear combination
^vww---+v;
in which B0 and B, are coefficients estimated from the data, and Xp represents the
independent variables. In this study, X, represents the set of background variables; X2
44


the academic variable, and X3 the various financial aid variables substituted for one
another in sequence.
A description of the variables and their formulation is displayed in Table 3.1.
Table 3.2, which follows, describes the steps of the analysis, and how the variables are
included in each step.
Table 3.1: Summary of Dependent and Independent Variables
VARIABLE CODE HOW CALCULATED BASIS OF COMPARISON

DEPENDENT VARIABLES: RETENTION
RETENTION TO YR. 2 RET2 Student was enrolled in fall of year one, and re-enrolled at same Institution for fall or spring of year two Students who were retained through year one are compared to those who did not re-enroll. (1=retained; 0=not retained)
RETENTION TO YR. 3 RET3 Student met criteria for RET2 and re-enrolled at same institution for fall or spring of year three Students who were retained through year two are compared to those who did not re-enroll. (1 retained; 0=not retained)
RETENTION TO YR. 4 RET4 Student met criteria for RET3 and re-enrolled at same institution for fall or spring of year four Students who were retained through year three are compared to those who did not re-enroll. (1 retained; 0=not retained)
INDEPENDENT V ARIABLES: BACKGROUND
ETHNICITY E. AS IAN E.BLACK E.HISP E.NATAM Students ethnic background as self-reported to institution Compared to white students, and to other students not of that group. (ForE.ASIAN, 1=Asian; O=non-Asian. Same pattern for other groups.)
NON- TRADITION-AL AGE NTAGE Students who were less than 17 or over 22 in the fall of year one Compared to students of traditional age (17 to 22 in fall of year one)
GENDER GENDER Sex as self-reported to institution Compared to males. (Female=1; male=0)
45


Table 3.1: Summary of Dependent and Independent Variables (Cont.)
VARIABLE CODE | HOW CALCULATED BASIS OF COMPARISON

FAMILY INCOME INC1.LO INC1.MD Adjusted gross income taken from financial aid files for students for whom need was calculated Compared to students for whom need was calculated whose family income was $50,000 and over, and to those for whom need was not determined. (INC1.LO=$0- $19,999; INC1.MD=$20,000- $49,999)
INDEPENDENT VARIABLES: ACADEMIC PREPARATION
ADMISSIONS INDEX INDEX.LO INDEX.MD Admissions index taken from enrollment files. Index is composite of high school GPA and SAT or ACT scores. Compared to high index. High index is defined as top third of index scores for the institution; medium index=middle third of scores for institution; low index=lower third of scores for institution.
CUMULATIVE GRADE POINT AVG. CGPA2 CGPA3 Grade point average for each semester, adjusted for total credits taken. Continuous variable.
INDEPENDENT VARIABLES: FINANCIAL AID FACTORS
DEPENDENCY STATUS DEPEND Dependency status for financial aid purposes. Recorded only for those for whom need was determined through financial aid application. Compared to independent students. (1 =dependent; 0=independent)
ANY AID Y1 .ANY Recorded for any student for whom need was determined who exhibited positive need and received any grant, loan, or work aid. Compared to students for whom need was not determined, and students for whom need was determined but showed negative need.
AID TYPE Y1 .GRANT Y1.LOAN Y1 .WORK Y1 .MERIT Y1 .OTHER Recorded for student who received any aid of any type. Compared to students who did not receive aid of the given type. (1 received aid of that type; 0=did not receive aid of that type)
AID COMBIN- ATIONS Y1.LNGT Y1.LNWK. Y1 .GTWK Y1 ALL Recorded for students who received any amount of aid of the combination noted. (Loan/grant; loan/work; grant/work; grant/loan/work) Compared to students who did not receive aid type of the given combination. (1 =received aid combination type; 0=did not receive aid of given combination type)
46


Table 3.1: Summary of Dependent and Independent Variables (Cont.)
VARIABLE CODE HOW CALCULATED BASIS OF COMPARISON

TOTAL AMOUNT OF AID Y1 .TAMT For students for whom aid was determined and who demonstrated positive need, the sum of aid of all types. Total aid was divided by $1,000. Continuous variable. Compared to those whose total amount of aid was lower.
AMOUNT OF AID BY TYPE Y1 .GAMT Y1 .LAMT Y1 .WAMT Y1 MTAMT Y1.0AMT For students for whom aid was determined and who demonstrated positive need, the amount of aid by type (grant, loan, work, merit, other). The aid for each category was divided by $1,000. Continuous variable. Compared to those whose amount of given type was lower.
EXCEPTION- ALLY HIGH TOTAL AID Y1.HIT75 Y1.HIT90 By institution, aid packages were determined for the 75th and 90th percentiles. Each students total aid package was then compared to this standard. Compared to students whose aid was less than the 75th or 90th percentile. (1 =total aid at 75th percentile; 0=aid lower than 75th percentile. Same for 90th percentile)
EXCEPTION- ALLY HIGH LOAN OR GRANT Y1.HIL75 Y1.HIL90 Y1.HIG75 Y1.HIG75 By institution, loan and grant amounts were each determined for the 75th and 90th percentiles. Each students loan or grant was then compared to this standard. Compared to students whose loan or grant was less than the 75th or 90th percentile. For Loan: 1=loan at 75th percentile; 0=loan lower than 75th percentile. Same for 90th percentile. Same system for Grant)
PROPORTION OF NEED FUNDED BY AID Y1 PROP For each student for whom need was calculated, the total aid package was divided by the calculated student budget as determined by the institution. Continuous variable. Students with need and with aid were compared to those without need or aid.
PROPORTION OF NEED FUNDED BY EACH AID TYPE Y1.GPROP Y1.LPROP Y1 .WPROP Y1.0PR0P For each student for whom need was calculated, the amount of aid for each aid type was divided by the calculated student budget as determined by the institution. Continuous variable. Students with need and with aid of the given type were compared to those without need or aid of that type.
47


Table 3.2: Steps of the Logistic Regression Analysis
Step Number Sample Dependent Variable Independent Variables
Background Academic Financial Aid
1 All Students Retention to second fall Age, Ethnic Background, Gender, Family Income
2 semester Index Score (GPA + ACT/SAT)
3 Any aid
4 Kind of aid package: grant, loan, work, merit, other
5 Kind of aid package: combinations of types
6 Total amount of aid
7 Amount of aid by kind
8 Total aid at the 75th percentile
9 Total aid at the 90th percentile
10 Grant or loan at the 75th percentile
11 Grant or loan at the 90th percentile
12 Proportion of aid to calculated need
13 Proportion of aid by kind to calculated need
48


Table 3.2: Steps of the Logistic Regression Analysis (Cont.)
'O
Independent Variables
Step
Number
Sample
Dependent Variable
Background
14
through
Each ethnic
group, in turn
Retention to second fall
semester
Age, Ethnic Background,
Gender, Family Income
Academic
Financial Aid
68
Index Score (GPA +
ACT/SAT)
Any aid
Total amount of aid
Total aid at 75th percentile
Total aid at 90th percentile
Aid by kind (grant, loan, work,
merit, other)
Grant or loan aid at 75th
percentile
Grant or loan aid at 90th
percentile
Proportion of total aid to
calculated need
Proportion of aid by kind to
calculated need
49


Table 3.2: Steps of the Logistic Regression Analysis (Cont.)
o
Independent Variables
Step
Number
Sample
Dependent Variable
Background
69
through
101
Each income
group (low,
medium, high),
by turn
Retention to second fall
semester
Age, Ethnic Background
Gender, Family Income
Academic
Financial Aid
Index Score (GPA +
ACT/SAT)
Any aid
Total amount of aid
Total aid at 75th percentile
Total aid at 90th percentile
Aid by kind (grant, loan, work,
merit, other)
Grant or loan aid at 75th
percentile
Grant or loan aid at 90th
percentile
Proportion of total aid to
calculated need
Proportion of aid by kind to
calculated need
50


Table 3.2: Steps of the Logistic Regression Analysis (Cont.)
Step Number Sample Dependent Variable Independent Variables
Background Academic Financial Aid
102 through All students who were enrolled in year Retention from second year to fall semester of third year Age, Ethnic Background, Gender, Family Income Cumulative college GPA through year two Any aid in year two, or any aid in year two
103 two Total amount of aid in year one, and total amount of aid in year three
104 Cumulative amount of aid for years one and two combined
105 Amount of aid by kind for year two
106 Cumulative amount of aid by kind for years one and two combined
107 Proportion of aid for year one and year two compared to the calculated need for each year
108 Proportion of aid by kind for year two to calculated need for year two
51


Table 3.2: Steps of the Logistic Regression Analysis (Cont.)
Step Number Sample Dependent Variable Independent Variables
Background Academic Financial Aid
109 All students who were enrolled in year three Retention from third year to fall semester of fourth year Age, Ethnic Background, Gender, Family Income Cumulative college GPA through year three Any aid in year two, or any aid in year two
110 Total amount of aid in year one, and total amount of aid in year three
111 Cumulative amount of aid for years one and two combined
112 Amount of aid by kind for year two
113 Cumulative amount of aid by kind for years one and two combined
114 Proportion of aid for year one and year two compared to the calculated need for each year
115 Proportion of aid by kind for year two to calculated need for year two

52


Estimating Effects: the Delta-P Statistic. To show the effect on the dependent
variable of a change in the independent variable, a delta P statistic has been calculated
using the formula2:
AP=p,P(l-P) (])
For dichotomous variables, the delta P shows the change in probability of
retention for a student possessing that characteristic. For continuous variables, delta P
shows the percentage increase or decrease in retention that will result from a one-unit
change in the variable.
Internal Validity
There are a number of considerations which bear on whether the results of this
study should be judged dependable. These considerations relate to the complexity of 1
2Peterson (1984) recommends an alternate formula:
AjP- exP(Zi) exP(Lo)
1 +exp(Z.,) 1 +exp(L0)
(2)
where L0 is the logit score before the change in the /th variable, and L, = L0 + B, is the
logit score after the unit change in X,. Peterson prefers (2) because formula (1) can
produce delta-p values greater than 1.0. However, because formula (2) requires that
changes in all variables in the formula be compared with changes in the particular
independent variable in question, the formula proved impractical for the purpose of this
study. Indeed, occasional delta-p values greater than 1.0 were produced. In such cases,
the symbol X+ or HI was substituted for the delta-p value.
53


retention behavior, the question of price-responsiveness of students (especially those
from low-income backgrounds), and the selection of subjects for study.
The behavior of students in persisting or leaving a higher education program is
exceedingly complex. The most accepted model of retention, as articulated by Vincent
Tinto (1975, 1987, 1990), takes into account an extensive calculation of background
factors, initial experiences in college, and then a complex of interactions between
students and their academic and social environments. Financial assistance can certainly
fit in such a model, yet its importance in relation to the multitude of other factors
affecting student behavior has not been made clear. The present study attempts to
clarify that relationship, but the very complexity of retention phenomena dictates
caution in interpreting study results. It is undeniable that financial aid may be
necessary, but not sufficient, for success and persistence in college. The effect of the
many other factors, both measured and unmeasured, on persistence makes
interpretation of results problematic.
The model adopted by this study assumes that students are responsive to
financial concerns related to prices and resources. While some have concluded that
students are price-responsive (St. John, Oescher, and Andrieu, 1992, and Leslie and
Brinkman, 1988), the rapidly-changing financial environment for higher education and
for financial aid may affect this responsiveness. The price charged for higher
education has escalated rapidly, as noted in the previous chapter, while the structure of
financing has changed as well. A few short years ago, it was assumed that families
54


were responsible for their childrens education and that if the family were unable to
afford the cost, society would assist. More recently, the student is most often bearing
the primary cost through loans and work, with the family and society playing much
reduced roles. Have these changes been recorded fully in students consciousness, or
is there a lag between the changing circumstances and the students awareness of
them? If so, then the relationship between financial aid and student retention behavior,
particularly in todays rapidly-changing environment, may reflect a delayed response
that is not measured in the present study.
Another question concerning internal reliability has to do with students level
of information and awareness of prices and financial aid resources. One might imagine
that older students, who have been largely left out of the current study because they
often lack the admissions index scores needed to standardize for academic preparation,
are the most mature in terms of their understanding of financial issues. Younger
college students, especially those who are just entering college at around age 18 and
who leave higher education at alarming rates within the first four semesters, are
inexperienced with decisions about high-cost purchases of goods or services. In the
case of middle and upper income students, parents have most often made those
decisions for the student up to the point of college. For low-income students, neither
the parent nor the student has had such experience. The student, therefore, is less
likely to make decisions about persistence in college as a result of a complete
55


calculation of the costs and benefits, including the long-term investment value, of
higher education.
Additionally, students often lack full information about the cost of college.
Again, the experience varies by income level. The middle or upper income student is
likely to be insulated from such information by parents who may be paying much of the
cost and making most of the financial decisions about college. For the low-income
student information may be available, but not sufficiently complete. For example, a
low-income student who takes out a student loan may know the amount of the loan,
but not the schedule of interest payments, or even more important, the effect of those
interest payments on future lifestyle possibilities.
Finally, there is information that suggests that students and parents
perceptions about college costs do not square entirely with reality. For example,
Jacqueline King (1996) discovered that both parents and students overestimate the
cost of education to a significant degree. This leads us to be concerned that students
persistence in college may be affected as much by these perceptions of exaggerated
cost as by a realistic assessment of tuition costs, loan costs, and financial aid resources.
Each of these factors poses a challenge to internal validity that must be taken
into consideration in interpreting results. One other consideration deserves attention.
This study looks at students as they choose to return or not return after their first year,
their second year, and their third year. The study considers the same factors at each
point, yet the composition of the group under examination changes with each year. It
56


is possible that the most vulnerable and unsophisticated students drop out in the first
year, leaving a group of enrolled students who respond differently to prices and to
financial aid, and are more sophisticated and mature, than those in the first year of
enrollment.
On the other hand, the study presents a detailed examination of the records of
a large sample of students. These records are gathered through a carefully controlled
process that has been standardized and monitored by the Colorado Commission on
Higher Education. Most of the data is recorded by the institution, rather than the
student, and the likelihood of reporting errors is small.
External Validity
This study includes a substantial proportion of the student population in
Colorado higher education. It also includes a variety of institutions and institutional
types. For these reasons, the results may reasonably be generalized beyond the single-
institution studies of financial aid that have been most prevalent.
On the other hand, the Colorado student population has its own particular
characteristics, such as a high level of adult educational attainment, a healthy
economy, a low proportion of Native Americans and African Americans, and other
such circumstances, all of which may limit generalizability to states and institutions
whose socio-economic conditions are similar to those in Colorado.
57


The post-analysis nature of the study has the advantage that it does not intrude
upon the subjects. Students in the sample would not have been affected or influenced
at all by the study, and it is unlikely as well that they would have been influenced by
the data collection process conducted by institutions at the behest of the Colorado
Commission on Higher Education in the development of its Student Unit Record
Database.
One of the greatest limitations of the study is the lack of data from the already
small number of Colorado private institutions, a lack that dictated their exclusion from
the study. Because only one fairly unusual private institution was included, very few
conclusions can be drawn concerning private higher education, or concerning the
comparison of public and private institutions with relation to retention and financial
aid.
All things considered, it is likely that the results of this study can be generalized
to states with similar demographics and similar levels of support for higher education,
and to institutions with similar populations, missions, levels of selectivity, and
environments.
Conclusion
Even considering the limitations of the study that have been identified, the
breadth of the sample, the detailed analysis of numerous aspects of financial aid, the
separate analysis of subgroups by income and ethnicity, and the appropriate use of
58


statistical procedures outlined in this chapter suggest that a rational, ordered, and
detailed approach has been adopted in order to produce greater understanding of the
complex interaction between financial aid and retention behavior. The following
chapter reviews the specific results of the analyses that were conducted.
59


CHAPTER 4
RESULTS AND ANALYSIS
Organization of the Chapter
The logistic regression analysis generated a large volume of data. To make this
data more understandable, summary tables were created. Separate tables were created
for in-state and out-of-state students. In consideration of the length of these summary
tables, they have been included in the Appendix A: Delta-p Values from Logistic
Regression Analysis: Source Tables. The discussion of results is based upon these
summaries.
This chapter will first discuss the extent to which all students at all institutions
exhibit the predicted behavior with respect to the issues of retention and financial aid.
After exploring the results for the aggregate of in-state and out-of-state students, each
institution is examined separately to discern whether the student population and
subpopulations at each individual institution have behaved according to prediction.
Following the analysis by institution, data is examined for each regression
equation across institutions. This analysis is designed to reveal whether particular
equations or particular variables have greater explanatory power than others.
60


Finally, the chapter concludes with a discussion of institutional and socio-
economic factors that may be associated with patterns of student behavior related to
financial aid and retention.
Tables Summarizing Results for All Students f Aggregated In-State and Out-of-State
Students! and for Individual Institutions
The entire set of regression equations was run for the aggregate file of in-state
students and the aggregate file of out-of-state students. The tables on pages 213-223
(in-state students) and pages 318-328 (out-of-state students) show the results of these
analyses.
Please refer to the sample table on the following page entitled Delta-p Values
from Logistic Regression Analysis: Source Tables All Institutions: Residents Only
(90A). In the table, the independent variables used in various versions of the
regression equation are shown in the left-hand column. (Please refer to the table in
Methods, pages 45-47 for definitions of variables.) In all cases, the dependent
variable is a dummy variable for retention, indicating a students enrollment in the
second year (value=l) or no enrollment the second year (value=0). The column
marked (1) BACKGROUND shows the results of the regression equation for
background variables only. The regression equation could be written as:
61


On
to
Table 4.1: Delta-p Values from Logistic Regression Analysis: Source TablesAll Institutions, Residents Only (from Appendix A)
SUMMARYO r .OGIT RESU rs
1
1
YEAR: 1800

INSTITUTION i LL INSTITU 3NS: RESID E JTS ONLY n* 9480

74.96 74.90 74.96 74.96 74.96 74.96 74 96 74.96 74.96 74.96 74 96 74.96 74.96
in (2) (3) (4) (5) (6) (6a) (6b) (7) (7a) (7b) (8) (9)
FINANCIAL AID FACTO! S:
tVpes AMCH >4TS: PROPC RTION
G,L.W,M,0 | COMBOS
TOTAL-75 TOTAL-90
ntage
eastern 0.0630 0.0741 0.0741 0.0907 0.0741 0.0560 0.0575 0.0656 0.0604 0.0632 00642 0.0714 0.0741
e.black
e hlsp 4).0687 -0.0249 -0.0258 -0.0279
e.natam
qender-f 0.0346 0.0366 0.0366 0.0368 0.0366 0.0364 0.0364 0.0363 0.0368 0.0348 0.0367 0.0348 0.0366
inc1 lo -0.0200 -0 0255
ind.md 0.0361 -0.0416
depend
GT/LN-75 GT/LN-90
index, lo -0.1236 0.1236 -0.1201 -0.1236 -0.1210 0.1193 -0.1224 -0.1174 -0.1218 0.1227 0.1232 -0.1236
index.hl 0.1302 0.1302 0.1231 0.1302 0.1278 0.1262 0.1275 0.1176 0.1305 0.1280 0.1281 0.1302
Iy1 any
yj.grnt
ylloan
yl.work
y1. merit
yj.oth
y1.hlt75/90
y1.htg75/90
y1,hi!75/90
i^pfp
0.1235 0.1374
0.00881 0.1003
0.0850 00818
iUSi
ylflprop
yj.lprop
yl.wprop
y1 mprop
yl.oprop


Z=B0+BlXl+B2X2+...+BpXp
where X, represents non-traditional age, X2 represents Asian ethnic background, Xs
African American ethnic background, X4 represents Hispanic ethnic background, X5
Native American ethnic background, X6 being female, X7 low income background,
and Xs middle income background. (Dependency status is only run when financial aid
variables are present.)
The values shown in column one are the delta-p values calculated from
regression coefficients generated from logistic regression analysis. Only regression
coefficients for independent variables showing relationship with the dependent variable
at a significance level of .01. .05. or .10 were converted to delta-p statistics. To
indicate the significance level of the regression coefficient from which the delta-p
values were derived, the following code has been employed:
values in bold = significance level of .01
values in italics = significance level of .05
values in standard type = significance level of. 10
If the significance level was greater than .10, the variable was eliminated from the
equation, no regression coefficient was calculated, and no delta-p is shown in the
table.
63


For the analysis of the aggregate samples, large sample sizes were involved
(9,490 for in-state students, and 2,689 for out-of-state students). To avoid making the
error of attributing importance to a statistically significant but practically
inconsequential relationship (Blalock, 1972, pp. 162-163), attention was given to the
magnitude of delta-p values. Values less than .01 were not considered. Delta-p values
greater than .01, representing a change in persistence of one percentage point or more,
were considered since changes in persistence rates of this magnitude are important in
the higher education environment. For example, the difference in ethnic minority
retention rates often differ from white student retention rates by something like ten
percentage points. A change of one percentage point would equate to a ten percent
reduction (or increase) in the gap between ethnic groups, and would represent an
important change in the rates of student success. In general, delta-p values of .01 to
.03 are characterized in the discussion as small magnitude effects, while values of
.04 to .07 are characterized as moderate effects and values of .08 or greater as
large effects.
In the first equation, the factor Asian has a delta-p value of .0630. The
italics signify that the regression coefficient from which the value was derived was
significant at the .05 level. The delta-p value can be interpreted as an increase (the
value is positive) in the rate of retention of 6.3 percentage points if a student is Asian.
The delta-p of -0.0597 for black signifies that the original regression coefficient was
significant at the .01 level (represented by bold type), and that the probability or rate
64


of retention is likely to decrease (the value is negative) by 6.0 percentage points if a
student is African American. Likewise, the probability of retention is predicted to
increase by 3.5 percentage points if the student is female, and so on.
The column marked (2) ACADEMIC shows the result of the regression
equation when all the background factors are retained, while new factors representing
academic preparedness (low admissions index score, high admissions index score) are
added to the equation. The regression coefficients resulting from this analysis have
been converted to delta-p values in column two for any variables whose effects are
found to be significant at the .01, .05, or .10 criteria.
All background variables and all academic preparedness variables continue to
be included in the regression equations represented by columns 3 through 9, but in
each of these columns a variable or variables representing financial aid factors have
been substituted for one another. For example, in the column marked (3) AID-
ANY, the variable representing receipt of any aid (yl.any) has been included as the
aid factor. In this case, the empty cell for yl.any signifies that no effect was identified
at the required significance level. In column (4) TYPES: G.L.W.M.O, however, it
was determined that receipt of grants (yl.grnt), loans (yl.loan), and merit (yl. merit)
types of aid did show effects at the required level. Receipt of grants showed a
negative effect on retention (delta-p of -.04), while the effects of loans and merit aid
showed positive effects on retention rates (.02 and .07).
65


In each of the subsequent columns, a different representation of aid has been
substituted in the regression equation, and the delta-p values calculated from
regression coefficients have been displayed. Background factors and academic
preparedness factors are always included (controlled! in every equation. Therefore.
the delta-p values for financial aid factors can be interpreted as the marginal change in
probability of retention when background and academic preparedness have been taken
into account.
The same system has been used in the tables on pages 213-223. In each of
these, background and academic preparedness factors are taken into account first,
followed by insertion of various representations of financial aid. The tables vary from
the first table described in the following ways:
66


Table 4.2: Stages of Analysis for Aggregated In-State Students
STAGE OF ANALYSIS PAGE NUMBER TITLE CONTENT
All Students; Retention to Year 2 ALL INSTITUTIONS: RESIDENTS ONLY Analysis of retention to the second year for all in-state students
All Students; Retention to Year 3 RETENTION TO YEAR 3 Analysis of retention to the third year for all in-state students
All Students; Retention to Year 4 RETENTION TO YEAR 4 Analysis of retention to the third year for all in-state students
Students by Ethnic Background; Retention to Year 2 HISPANIC Analysis of retention to the second year for Hispanic students only
BLACK Analysis of retention to the second year for African American students only
NATIVE AMERICAN Analysis of retention to the second year for Native American students only
ASIAN Analysis of retention to the second year for Asian students only
WHITE Analysis of retention to the second year for White students only
Students by Income Background; Retention to Year 2 LOW INCOME Analysis of retention to the second year for low income students (for whom need has been determined) only
MEDIUM INCOME Analysis of retention to the second year for medium income students (for whom need has been determined) only
HIGH INCOME Analysis of retention to the second year for high income students (for whom need has been determined) only
The same pattern shown above is used in the analysis of aggregated out-of-
state students on pages 318-328.
The entire process described above was then performed for each individual
institution in the sample. The tables summarizing results for individual institutions,
67


including all eleven stages of analysis for each institution, are produced on pages 224-
316 for in-state students and on pages 319-364 for out-of-state students. (Tables are
absent only where the sample size is insufficient for logistic regression analysis [<50])
Examination of Student Retention Behavior Aggregated Across Institutions
Hypothesized Relationship between the Variables and Retention
In the examination of the aggregate files for all students, the following
relationships are expected, consistent with the research hypotheses specified in
Chapter Three.
Table 4.3: Expected Relationship of Financial Aid Variables to Retention
Variable Expected Relationship to Retention
The receipt of student financial aid by students with calculated financial need will contribute to those students persistence from one year in college to the next.
Receipt of any aid [y1 .any] Neutral or positive delta-p value
H2a: As the total amount of financial aid increases, persistence will increase as well for students of equivalent family incomes.
Total amount of aid received [yltamt] in the analysis of students by income group Positive delta-p value
High total amounts of aid received at the 75th and 90th percentile [y1.hit75; y1.hit90] in the analysis of students by income groups Positive delta-p value
H2b: Grant funds, that is, those financial aid dollars that do not need to be repaid by students, will be more effective in increasing retention than other forms of aid.
Grant funds as type of aid received [y1 .grnt] Neutral or positive delta-p value, and higher value than for other aid types
68


Table 4.3: Expected Relationship of Financial Aid Variables to Retention (Cont.)
Variable Expected Relationship to Retention
Other types of aid received: loan, work, merit, other [y1 .loan; y1 .work; y1 .merit; y1 .other] Lower delta-p values than for grant aid
Amount of grant funds received [y1 .gamt] Positive delta-p value, and higher value than for other amounts of aid types
Amount of other types of aid received [y1 .lamt, ylwamt; ylmtamt; yloamt] Lower delta-p values than for grant amounts
The greater the proportion of the cost of education" that is funded by financial aid, the greater will be the resulting persistence.
Proportion of calculated need funded by aid of any type [y1 .prop] Positive delta-p value
H^: There will be differences in the benefits of financial aid according to differences in ethnic background, with white students demonstrating greater benefits than students of color other factors being equal.
All financial aid variables in the analysis of students by ethnic group Lower delta-p values for ethnic minority groups than for white students
H^: Students from the least advantaged economic backgrounds will benefit from financial aid less than students from more advantaged, but still needy, backgrounds.
All financial aid variables in the analysis of students by income group Lower delta-p values for low income group as compared to middle and high income groups
Students from the lowest income groups, and those from ethnic backgrounds who traditionally experience the greatest discrimination, are likely to experience the greatest benefit from grant aid and the least benefit from loan aid.
Receipt of grant as a form of aid [y1 .grnt] in the analysis of ethnic and income groups Higher delta-p values for ethnic minority and low income groups as compared to delta-p values for loans
Amount of grant aid received [y1 .gamt] in the analysis of ethnic and income groups Higher delta-p values for ethnic minority and low income groups as compared to delta-p values for loan amounts
High amounts of grant aid at the 75th and 90th percentiles [y1.hig75; y1.hig90] in the analysis of ethnic and income groups Higher delta-p values for ethnic minority and low income groups as compared to delta-p values for loan amounts
69


Table 4.3: Expected Relationship of Financial Aid Variables to Retention (Cont.)
Variable Expected Relationship to Retention
Proportion of calculated need funded by grant aid [y1 .gprop] in the analysis of ethnic and income groups Higher positive delta-p values for ethnic minority and low income groups as compared to delta-p values for loan proportion
Loan as a form of aid [yl.loan] in the analysis of ethnic and income groups Lower delta-p values for ethnic minority and low income groups as compared to delta-p values for grants
Amount of loan aid received [y1 .lamt] in the analysis of ethnic and income groups Lower delta-p values for ethnic minority and low income groups as compared to delta-p values for grant amounts
High amounts of loan aid at the 75th and 90th percentiles [y1.hil75; y1.hil90] in the analysis of ethnic and income groups Lower delta-p values for ethnic minority and low income groups as compared to delta-p values for grant amounts
Proportion of calculated need funded by loan aid [y1 .Iprop] in the analysis of ethnic and income groups Lower delta-p values for ethnic minority and low income groups as compared to delta-p values for grant proportion
Tables Summarizing the Estimated Effects of Variables in the Equation
In the discussion that follows, information has been extracted from the Source
Tables for aggregated students in the sample (in-state students on pp. 213-223 and
out-of-state students on pages 318-328). The extracted data appears in tables in the
text. For each variable, the most prevalent delta-p value appearing in the source tables
across the nine regression equations containing financial aid variables (equations 3
through 9) has been entered into tables such as the one below. These tables provide a
summary description of the estimated effects of each variable in the analysis.
70


The tables appearing in the text are summaries that are a step removed from
the source tables. Information on the significance level of the regression coefficient
from which delta-p values have been derived, for example, is omitted from the
summary tables in the text. For the most comprehensive and detailed information, the
reader is referred to the original source tables in the Appendix.
Tables 4.4 and 4.5 summarize the effects of background, academic, and
financial aid factors for aggregated in-state and out-of-state students.
Table 4.4: Summary Delta-p Statistics for
Aggregated In-State and Out-of-State Students:
Background and Academic Factors in Relation to
Retention to Year Two
Variable In-State Students Out-of-State Students
Asian .07
African American
Hispanic
Native American -.18
Gender (Female) .04 .04
Low Income
Medium Income -.09
Low Index -.12 -.14
High Index .13 .04
71


Table 4.5: Summary Delta-p Statistics for
Aggregated In-State and Out-of-State Students:
Financial Aid Factors in Relation to Retention to
Year Two
Variable In-State Students Out-of- State Students
Any Aid -.06
Aid by Kind
Grant -.04 -.05
Loan .02
Work
Merit .07 -.10
Other -.06
Total Amount of Aid .02
High Total Aid:
75th percentile .12
90th percentile .14
Aid Amount by Kind:
Grant -.01
Loan .02
Work .03
Merit .11
Other .03 .01
High Grant:
75th percentile .07
90th percentile .10 -.13
High Loan
75th percentile .10 .08
90th percentile .09 .10
72


Table 4.5: Summary Delta-p Statistics for
Aggregated In-State and Out-of-State Students:
Financial Aid Factors in Relation to Retention to
Year Two (Cont.)
Variable In-State Students Out-of- State Students
Proportion of Need met by Aid .02
Proportion of Need met by Kind of Aid: Grant Loan Work Merit Other





Aggregated In-State Students
When background variables were entered in the first step, a positive
relationship with retention to the second year appeared for Asians and for women and
a negative relationship appeared for Hispanics. When academic preparedness variables
were entered in the next step, the positive relationship increased slightly for Asians and
continued for women. (See Source Table in Appendix A, page 213.) The negative
relationship between Hispanics and retention to the second year was no longer
significant, however, leading to the conclusion that lower levels of academic
preparedness accounted for most of the initial difference in retention for Hispanic
73


students. Low and high index scores show a large negative and large positive effect
on retention, respectively.
Next, different representations of financial aid were substituted for one
another. (Please refer to the Source Table, page 213, and Table 4.5.) The first test of
the effects of financial aid was the entry of the any aid variable. There was no
significant association of this variable with retention. When type of aid was
substituted, grants were negatively associated with retention, while loans and merit aid
were positively associated.
The next series of financial aid variable substitutions took dollar-amounts of
aid into account. The total amount of aid was associated with retention. The
change in the probability of retention for each $1,000 of aid was estimated at
approximately 1.7 percentage points. When exceptionally high amounts of total aid
packages were considered, however, the estimated changes were far greater: 12.4 for
grant aid at or above the 75th percentile, and 13.7 percentage points for grant aid at or
above the 90th percentile. The delta-p for aid-amount variables for every aid type was
positive, with the exception of grants. Grants did have a positive association with
retention, however, when awarded at the higher levels. The change in probability of
persistence increased an estimated 6.9 percentage points for those with grant awards
at or above the 75th percentile level, and 10.0 points for grants at or above the 90th
percentile. High levels of loan were also associated with persistence: 9.6 points at the
75th percentile and 9.2 points at the 90th percentile.
74


The proportion of calculated need met by aid was significant, but the estimated
change in probability of retention was small: 1.5 percentage points. No significant
relationships appeared between the proportions of need met by particular aid types.
Out-of State Students
For the remainder of the discussion, delta-p statistics, rounded to the nearest
hundredth, are included in parentheses. They indicate the estimated change in
probability of retention given the characteristic in question.
The initial analysis of background factors identified a negative relationship
between Native Americans and retention and medium income students and retention,
and a positive relationship for women and retention. (Please refer to the source table
on page 318 of Appendix A.) Hispanics were less likely to be retained when
background characteristics were considered. Again, however, the negative
relationship between Hispanics and retention disappeared when academic preparedness
was taken into consideration. The positive relationship between women and retention
continued after accounting for academic preparedness. The estimated change in
probability of Native Americans being retained continued to be negative and high
(about -.18).
The receipt of any financial aid was negatively associated with retention (-.06).
(Please refer to Tables 4.4 and 4.5 in the text.) Several types of aid were also
negatively associated with retention: grants merit, and other aid, as well as the
75