Citation
A predictive model for the repayment of student loans in community colleges

Material Information

Title:
A predictive model for the repayment of student loans in community colleges
Creator:
Schmidt, James A., 1950- ( Dissertant )
Wattenbarger, James L. ( Thesis advisor )
Nickens, John M. ( Reviewer )
Riker, Harold C. ( Reviewer )
Place of Publication:
Gainesville, Fla.
Publisher:
University of Florida
Publication Date:
Copyright Date:
1983
Language:
English
Physical Description:
vii, 91 leaves; 1983.

Subjects

Subjects / Keywords:
College students ( jstor )
Colleges ( jstor )
Community colleges ( jstor )
Education ( jstor )
Educational research ( jstor )
Higher education ( jstor )
Loan defaults ( jstor )
Modeling ( jstor )
Student financial aid ( jstor )
Student loans ( jstor )
Community colleges -- Administration -- Florida ( lcsh )
Default (Finance) -- Florida ( lcsh )
Dissertations, Academic -- Educational Administration and Supervision -- UF
Educational Administration and Supervision thesis Ph. D
Student loans -- Florida ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Abstract:
The problem of this study was to determine the relationship between selected variables which characterize community college students and student loan defaults and to develop a model using these variables to predict student loan payback. Given the current economic crisis and the increasing reliance on the student loan programs to help students meet educational expenses, a study of the importance of selected student demographic characteristics and their relationships to the student loan default problem is of great importance to the future support of the student loan programs. The literature provided a theoretical basis for this study including appropriate variables for study as predictors of student default. These variables included size of loan total, marital status, sex, grade point average, college standing, and age. The data presented in this study were supplied by the Florida Student Financial Aid Commission, Tallahassee, Florida, and represented a statewide sample of 7 6 community college students who have participated in the Guaranteed Student Loan program. Of the six variables selected, only the size of the loan total and marital status distinguished significantly those who repaid their student loans from those who did not. In addition to these variables, sex, grade point average, college standing, and age were useful in developing a prediction model. Although the model did not provide an infallible formula for predicting those students who are most likely to repay their student loans, the model predicted group membership (defaulter of non-defaulter) for 70% of the sample cases. These findings underscore Pattillo and Wiant's conclusion that items reflecting financial rather than biographical data appear to be better predictors of loan delinquency. Therefore, it appears that the inclusion of additional discriminating variables and a more detailed study design may be necessary in order to improve the identification of students who are likely to repay their student loans.
Thesis:
Thesis (Ph. D.)--University of Florida, 1983.
Bibliography:
Includes bibliographical references (leaves 86-89).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by James A. Schmidt.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
030349339 ( AlephBibNum )
11437236 ( OCLC )
ACL1872 ( NOTIS )

Downloads

This item has the following downloads:


Full Text












A PREDICTIVE MODEL FOR THE
REPAYMENT OF STUDENT LOANS
IN COMMUNITY COLLEGES








BY


JAMES A. SCHMIDT


A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1983














ACKNOWLEDGMENTS


The writer wishes to acknowledge the assistance of the

many persons who took an active interest in the preparation

of this study.

The guidance and assistance provided throughout the last

two years by Dr. James L. Wattenbarger, chairman of the

writer's supervisory committee, are deeply appreciated. In

addition, sincere thanks are also extended to the faculty of

the Department of Educational Administration and Supervision

and in particular to the members of the committee, Dr. John

M. Nickens and Dr. Harold C. Riker.

For their assistance in collecting the data the writer

is indebted to the Florida Student Financial Aid Commission.

The technical assistance provided by Mr. Kenneth Goehle and

Mrs. Betty Davis are also gratefully appreciated.

Finally, the writer wishes to acknowledge the support

provided by his family. For their encouragement, assist-

ance and emphasis on the value of an education, the writer

is deeply indebted to his parents, Dr. and Mrs. Paul H.

Schmidt. To his wife, Linda, and his sons, Daniel and

Timothy, the writer wishes to express his deepest gratitude

for their love, patience and understanding during the prepa-

ration of this study.
















TABLE OF CONTENTS


Page

ACKNOWLEDGMENTS . . . . . .. . . . ii

LIST OF TABLES . . . . . . . . . . v

ABSTRACT . . . . . . . . . vi

CHAPTER

I. INTRODUCTION . . . . . . . . 1

The Role of Financial Aid . . . . . 2
The Growth and Importance of Loans . . .. .10
The Problem of This Study . . . . .. 19
Justification for the Study . . . . .. 20
Delimitations & Limitations . . . . .. 21
Research Design and Procedures . . . .. 22

II. REVIEW OF RELATED LITERATURE . . . .. 26

Problems Inherent in Student Loan
Programs . . . . . . . . . 31
Institutional Responses to
Student Loan Defaults . . . . . .. 38
Professional Organization Action
Regarding Student Loan Defaults . . .. 40
Governmental Action Regarding
Student Loan Defaults . . . . . . 42
Responses to Student Loan Default
Problem by Education Spokesmen . . .. 44
Antecedent Studies Relating to Student
Loans and Demographic Characteristics . 48
Summary . . . . . . . . .. . 53

III. DEVELOPMENT OF THE MODEL . . . . .. 55

Analysis of Variables . . . . . .. 60
Development of the Model . . . . .. 69
Results . . . . . . . . . . 73
Summary . . . . . . . . . . 77

















Page

IV. CONCLUSIONS, IMPLICATIONS AND
RECOMMENDATIONS . . . . . . .. 79

Conclusions . . . . . . . . .. 79
Implications . . . . . . . .. 80
Recommendations . . . . . . . .. 83

BIBLIOGRAPHY . . . . . . . . . .. . 86

BIOGRAPHICAL SKETCH . . . . . . . .. 90
















LIST OF TABLES


Table


1. Philanthropical Support for Higher
Education . . . . . . .

2. Overview of Major Variables Selected
by Previous Researchers . . . .

3. Chi-Square Analysis of Relationship
Between Age and Loan Payback . .

4. Chi-Square Analysis of Relationship
Between Marital Status and Loan
Payback . . . . . . . .

5. Chi-Square Analysis of Relationship
Between Sex and Loan Payback . .

6. Chi-Square Analysis of Relationship
Between College Standing and
Loan Payback . . . . . .

7. Chi-Square Analysis of Relationship
Between Grade Point Average and
Loan Payback . . . . . .

8. Chi-Square Analysis of Relationship
Between Loan Total and Loan
Payback . . . . . . . .

9. Interaction Effects Between the
Variables . . . . . . .


. . 63


. . 63



. . 64



. . 65



. . 66


. . 68


Page










Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfilment of
the Requirements for the Degree of Doctor of Philosophy


A PREDICTIVE MODEL FOR THE
REPAYMENT OF STUDENT LOANS
IN COMMUNITY COLLEGES

By

James A. Schmidt

August, 1983

Chairman: James L. Wattenbarger
Major Department: Educational Administration and Supervision

The problem of this study was to determine the relation-

ship between selected variables which characterize community

college students and student loan defaults and to develop a

model using these variables to predict student loan payback.

Given the current economic crisis and the increasing reliance

on the student loan programs to help students meet educational

expenses, a study of the importance of selected student demo-

graphic characteristics and their relationships to the student

loan default problem is of great importance to the future

support of the student loan programs.

The literature provided a theoretical basis for this

study including appropriate variables for study as predictors

of student default. These variables included size of loan

total, marital status, sex, grade point average, college

standing, and age.

The data presented in this study were supplied by the

Florida Student Financial Aid Commission, Tallahassee,










Florida, and represented a statewide sample of 76 community

college students who have participated in the Guaranteed

Student Loan program.

Of the six variables selected, only the size of the loan

total and marital status distinguished significantly those

who repaid their student loans from those who did not. In

addition to these variables, sex, grade point average,

college standing, and age were useful in developing a pre-

diction model. Although the model did not provide an in-

fallible formula for predicting those students who are most

likely to repay their student loans, the model predicted

group membership (defaulter of non-defaulter) for 70% of the

sample cases. These findings underscore Pattillo and

Wiant's conclusion that items reflecting financial rather

than biographical data appear to be better predictors of

loan delinquency.

Therefore, it appears that the inclusion of additional

discriminating variables and a more detailed study design

may be necessary in order to improve the identification of

students who are likely to repay their student loans.















CHAPTER I

INTRODUCTION



From modest beginnings over three hundred years ago the

role and impact of financial assistance to students in

American colleges have changed dramatically. Like the gift

presented by Lady Anne Mowlsen of London to the deserving

Harvard student, most early student financial aid awards

were granted with money given to colleges by private individ-

uals to aid worthy, needy students. Therefore, the original

purpose of providing student financial aid was to expand

educational opportunities to those students whose incomes

presented a barrier to higher education.

In fact, only during the period of time immediately

following World War II was substantial aid awarded on the

basis of merit alone. From the late 1940s through the early

1960s the Serviceman's Readjustment Act of 1944 (GI Bill)

provided substantial amounts of money to returning service-

men for educational expenses regardless of individual finan-

cial need. Many colleges and universities found that the

large amounts of scholarship funds, previously used to sup-

port needy students, were no longer needed and those insti-

tutions began to use the funds to attract and reward students











with academic or other special talents, with little or no

regard for financial need.

Currently, the essential criterion of the major federal

aid programs (Pell Grant, College Work Study, National

Direct/Guaranteed Student Loan Programs) is the students'

need for funds. Accordingly, the purpose of the principal

financial aid programs of today is remarkably similar to the

original intention of the early student aid programs. The

return to a need-based student aid system has caused a great

deal of consternation to financial aid administrators. Today

it becomes increasingly difficult to reward students with

merit-based aid when the available resources do not ade-

quately take care of the needy students. In order to encour-

age both needy and worthy students, a balance between need-

based and merit-based aid is critical to the development of

an impartial and equitable student financial aid delivery

system. However, most student financial aid administrators

apparently feel that the current, primarily need-based,

delivery system provides the best use of available resources

for it attempts to extend the opportunity for higher education

to as many worthy students as possible.


The Role of Financial Aid


Since 1972 there has been considerable discussion con-

cerning the role of financial aid in American higher educa-

tion. In nearly every deliberation much attention has been










devoted to the related issues of access and tuition charges

because they are so intertwined with the role of financial

aid.

As a nation, the United States has demonstrated its

commitment to citizen access to postsecondary educational

opportunities by instituting programs of assistance to needy

students, establishing support programs for colleges and

universities (i.e., Land Grant College Program) and develop-

ing a strong system of public colleges and universities.

According to Ostar (1978), 80% of all students attend-

ing higher educational institutions today are in programs

at public institutions where the state partially subsidizes

the institutional costs of the student. As a part of the

public higher educational system, the public community

college is an important factor in providing higher educa-

tional opportunities to American students. The impact of

the community college system on the issue of access is high-

lighted in the state of Florida where more than 95% of the

state population live within commuting distance of a commun-

ity college program.

Although the costs of attending a community college

program may be lower because of public subsidies, the indi-

viduals attending the community college seem to have the

greatest need for financial assistance. A 1972 American

College Testing (ACT) Program Study indicated that the

financial needs of community college students sharply










exceeded those of the majority of students enrolled in other

colleges and universities. The ACT study estimated that

financially disadvantaged students constituted 25% of the

total student population at community colleges and cited the

following factors as contributing to this high percentage:

(1) 70% of the students depend on the
automobile to reach campus,

(2) the higher age distribution (average
age of student is 29) indicates the
average student has assumed adult
responsibilities, and

(3) 66% of the students come from families
with (1974) incomes below $10,000.
(Blocker, 1974, p. 126)

Therefore, it would also appear that the retention of

low tuition policies is critical to maintain access to

higher educational opportunities for a large number of stu-

dents in the community college system. In fact, a Univer-

sity of Wisconsin study highlights the impact tuition

policies have on the issue of access (Ostar, 1978). This

study determined that for every 1% reduction in total cost

of attending a University of Wisconsin Center participating

in the study there was a corresponding 1.33% increase in en-

rollment. The study also attributed the increased enroll-

ment to new students who were not enrolled in the University

of Wisconsin system and concluded that many of these new

students would not have been able to attend were it not for

the lower tuition policy (Ostar, 1978).










Despite the establishment of the relationship between

low tuition and access by this and other studies, much de-

bate regarding federal aid centers on the strategy of pro-

viding individual student aid rather than institutional aid.

It is feared that the reliance on the student-based aid

delivery system may cause additional pressure on state govern-

ments and lead both public and private institutions to raise

tuition fees (Ostar, 1978). Although the underlying assump-

tion of raising tuition fees is that the truly needy student

will be taken care of, a University of Tennessee study in-

dicated that a $100 increase in tuition would yield only a

$26 increase in an individual student's Pell Grant Award

(Ostar, 1978). Truthfully, every tuition increase falls

hardest on the $13,000 $25,000 income family and yet this

student is eligible for little federal aid in the form of

grants or College Work Study (Ostar, 1978). In addition,

Ostar maintains that an overemphasis on student-based aid

leads to a growing number of charges involving fraud, waste,

and abuse and leads to ever-tightening regulations and in-

creased numbers of bureaucrats, investigators and bill

collectors.

The previous discussion regarding access would indicate

that the best approach regarding federal assistance may be

the increasing of federal support to higher educational in-

stitutions enabling colleges and universities to retain

lower tuition charges. However, because of economic and











political factors the financial aid delivery system is

oriented toward individual student awards. Given this under-

standing two major questions must be addressed: what aid is

available to students and how effective is the delivery

system?


Student Aid Resources

Essentially student financial aid is available in the

form of grants, work-study opportunities and loans. A grant

is gift money that is offered to those most needy in order

to prevent over-indebtedness through borrowing. Work-study

is a self-help program which provides meaningful, college-

related job experiences and earnings enabling the student to

earn money and meet current expenses from hi-weekly wages.

Loans constitute a program of borrowing whereby students

must repay the loan with interest from future earnings after

leaving college. The federal programs representing these

categories of aid include Pell (BEOG) and Supplemental Educa-

tional Opportunity Grants, College Work-Study, and National

Direct and Guaranteed Student Loan Programs.

Most of the federal programs have not seen substantial

increases in federal money in recent years. However, the

Pell Grant Program (BEOG) which was traditionally limited to

low income students has been expanded to assist families

from middle income levels. In addition, the amount of money

expended and the numbers of students served through the











Guaranteed Student Loan Program (GSL) have risen dramati-

cally.

During the 1982-83 fiscal year modest cuts were made

in the financing of the campus-based programs (SEOG, CWSP

and NDSL). Although support for the Pell Grant Program was

retained at previous levels, eligibility and dollar delivery

regulations were instituted to reduce fraud, abuse and

errors. The combination of reduced support for some pro-

grams and increased regulation of the Pell Grant Program

created a dilemma for student financial aid administrators

who could stretch local and institutional funds only so far.

At the same time, colleges and universities are exper-

iencing higher costs because of increases in major dollar

operating items (i.e., salaries, energy, supplies) which

seem to be severely affected by an inflationary economy.

This increase in costs translates into higher tuition

charges. Since the financial aid application requires the

student's resources to be balanced with the expected ex-

penses, many students are experiencing higher need figures

as the expected expenses rise. Without corresponding in-

creases in student aid to offset the rise in expenses, a

"need gap" crisis occurs for the student. Although existing

support for assistance programs is generally at the highest

level ever recorded, the increasing cost factor is rising

faster than increased federal support and therefore the

number of "need gap" victims continues to grow.











The Student Aid Delivery System

Most of the blame for the "need gap" crisis must be put

on increasing costs and expenses since the majority of the

financial aid researchers conclude that the student aid

delivery system appears to be quite effective (Deitch, 1978).

The financial aid delivery system in the United States

is the mechanism used to provide individual students with the

resources to attend postsecondary schools. Application for

the Federal Pell Grant Program may be made separately on a

free federal application; or, for a nominal charge, an in-

depth assessment of need may be presented along with the

eligibility report for the Pell Grant Program. Although

there has been some recent discussion of the effectiveness

and equity of the need-analysis system, previous studies

indicate the system is in pretty good shape. According to

a study presented at the Annual Meeting of the American

Association for Higher Education in March of 1979 the

following conclusions were presented:

(1) The need based system of financial
aid has gone far in removing a mean-
ingful element of price competition
from the buying and selling of higher
education,

(2) the structure of the aid system (mix
of parental contribution, need-based
grant and self help) is meeting the
public's perception of how post-
secondary education should be financed,

(3) the major goal of providing access is
quite clear, and correspondingly the
programs have been successfully reach-
ing toward the goal (i.e., the number
of minorities has increased signifi-
cantly),











(4) federal, state and institutional
funds seem sufficient to insure
that any qualified student could
afford to attend most of the
nation's institutions, and

(5) the problem is not with the dif-
ferent kinds of aid available but
for more efficient and equitable
distribution of the funds that are
available. ("Premises of Federal,"
1979)

According to Kenneth Dietch (1978), the current system of

student financial aid is a thoroughly developed, pervasive

and well-functioning system of price discrimination. How-

ever, he cites several problems which are likely to arise

with the current system including

(1) competitive pricing (by institutions),

(2) lack of sensitivity to student costs,

(3) loss of financial privacy of families
of applicants,

(4) efforts to depart from the need-based
system to attract (non-needy, out-
standing) students, and

(5) need for fairness in the treatment of
independent students. (Dietch, 1978)

Regardless of the potential problems with the financial

aid delivery system, its current status seems to be that it

is well respected among professionals and researchers.

Financial aid administrators and federal program personnel

have struggled to solve the problems of "need-gap" by work-

ing together to tighten requirements and enforce regula-

tions through extensive validation procedures.











The Growth and Importance of Loans


As other political priorities become apparent in Wash-

ington and free student aid programs (grants) are retained at

previous funding levels, most aid administrators have increas-

ingly used the available loan programs to meet student aid

needs. In fact, loan programs have been increased substan-

tially while most of the other aid programs have been limited

or decreased in size. Therefore, the importance of an

equitable loan system has far greater implications today than

it did previously as more and more students use the device to

solve the "need gap" crisis.


General Definition and Philosophy
of Student Loan Programs

As indicated previously, loans are awards of money

granted with the prior requirement that they subsequently

be repaid, in whole or in part, with or without interest.

According to Davis and Van Dusen (1978), loans are the most

attractive form of student aid to most policy-makers because

they must be repaid and therefore are an investment rather

than a direct expenditure of funds. Recent discussion of

national student aid policies indicates that those in

decision-making roles believe loan programs to be very

effective. Therefore, the great increase in the use of the

current loan programs and the addition of new loan programs

are quite understandable.










The underlying philosophical premise of a loan program

is that the person who derives the benefit from an education

should pay for its costs, without denying low income stu-

dents access to postsecondary education due to lack of

current resources (Davis & Van Dusen, 1978). Not only has

such a philosophy been utilized to develop loan programs

but it has also led to arguments concerning how much of the

cost of an education should be reflected in the tuition

charge. For example, in 1973 the Committee for Economic

Development recommended that tuition costs should be raised

to 50% of the total educational cost within five years at

colleges and universities and within 10 years at community

colleges. Subsequently the issue was studied in detail by

the Carnegie Commission. It suggested that tuition should

be raised to 30% of the educational cost within 10 years at

colleges and universities, but the Carnegie Commission

exempted community colleges due to the historical concept

of two years of free access to higher education. Many

states have cited the Carnegie Commission report in intro-

ducing data supporting the raising of fees at low-cost

public institutions.

However, Howard Bowen (cited in Balderston, 1970) effec-

tively argues that an investment in higher education is

really an investment in the society. It is an investment

in the nation's human capital leading to increased indi-

vidual productivity and societal benefits (i.e., a more

informed electorate, increased national productivity, lower











unemployment rates, a populace more adaptable to technology,

increased social consciousness). Mr. Bowen concludes that

typical students already bear 75 to 80% of the cost of their

education in the form of tuition, books and foregone earn-

ings. Many studies conclude that the heavy reliance on loan

financing has the effect of shifting the eventual burden of

payment for the cost of higher education toward students and

their future income (Balderston, 1970). According to Mr.

Bowen, the increased availability of loans is not an appro-

priate solution to the need gap crisis since students al-

ready bear such a large portion of the cost of their educa-

tion. Mr. Bowen argues that since society receives signifi-

cant benefits from educating students, it should bear an

increasing portion of the cost of educating them.

Those who support the argument that students should

bear the cost of their education also favor the increasing

use of loans to help students finance the increasing costs.

However, the desirability of increasing the burden of cost

on students by raising fees and then assisting the needy

students through loan programs is an issue that will remain

unresolved as both pro and con advocates have many arguments

of merit. However, the current federal administration

favors the use of loans to help needy students bear the in-

creasing costs of higher education and therefore it has

instituted new loan programs to help relieve the "need gap"

crisis.











Types of Loan Programs

Among the original loan programs are the National

Direct Student Loan (NDSL) Program (formerly the National

Defense Student Loan Program) and the Guaranteed Student

Loan (GSL) Program. The NDSL is a low-cost deferred repay-

ment loan administered by the educational institution and

funded 90% with federal monies and 10% with institutional

monies. Under the NDSL Program there are no security or

endorsement requirements and its distinctive feature is that

in addition to a deferment of three years for public service,

the student may arrange for cancellation of the loan based

on military service in an area of hostility, full-time

teaching service in selected schools (for five years) or

full-time staff service in a head start program (for seven

years).

The Guaranteed Student Loan Program is available in

nearly all states where a state level guarantee agency has

been developed. In the remaining states students may apply

for a Federally Insured Student Loan (FISL). Administra-

tively, private parties and non-federal agencies loan their

own capital to students in both programs. In the GSL pro-

gram there is a central monitoring agency. While the

federal government insures the loan and helps pay the in-

terest, the agency acts as intermediary, overseer and

collections processor. However, in the FISL program the










lenders rely directly on the U. S. Department of Education

for an interest supplement, insurance against default and a

collection mechanism. Repayment regulations for both the

GSL and FISL programs are similar to the NDSL; however,

deferments of up to three years are only available for ser-

vice in the armed forces, Action or the Peace Corps, and

there are no provisions for cancellation.


The Increasing Use of Loans

The growth of both of these programs is indicative of the

current emphasis on the issuance of loans to solve the student

aid crisis. In 1977, Ernest L. Boyer (cited in Hauptman, 1978),

then Commissioner of Education, estimated that between 1965 and

1977 six million students borrowed $11.1 billion. Commissioner

Boyer stated that in 1975-76 alone $2.6 billion were loaned

to students and that the national increase rate in the num-

ber of loans processed was averaging 5% each year. Mr.

Boyer concluded that thousands of students have been served

by the loan programs including many students who otherwise

may have been denied a postsecondary education. In addition,

Sanford (1980), in his research of the impact of loans on

graduate education, found that graduates with loans are

slightly more likely to attend graduate or professional

school than those without loans. Therefore, it appears that

there is substantial value in supplying loans beyond merely

providing access to undergraduate degree programs. The most











recent data indicate a total of 2.75 million loans were pro-

cessed through the GSL system alone during the 1981-82

academic year representing an investment of $6.1 billion

(The Chronicle of Higher Education, 1983).

Accordingly, the success of the federal loan programs

seems to be the impetus for the development of new programs.

For example, the Federal Government announced the creation

of the PLUS/ALAS Loan program in 1980. This program is

similar in administrative details to the other federal loan

programs. The major difference between these programs and

the original loan programs appears to be the lack of defer-

ment and cancellation provisions and higher interest rates.

It also should be noted that, in addition to the federal

loan programs, many institutions, states and private founda-

tions provide loan funds to students. Many of these programs

pre-date the federal efforts, but the importance and reliance

on loan programs are a result of the widespread application

of the loan principles effected through the federal programs.

Of the federal loan programs the GSL/FISL program is the

most widespread. The reason is that the NDSL is institution-

ally administered and many institutions choose not to parti-

cipate in the program. Other institutions have been re-

stricted from participating in the NDSL program due to their

high student default rates. Therefore, due to the universal-

ity of the GSL/FISL program, much of the data and research

presented in the remainder of this study will focus on the

GSL/FISL program.











Program Problems and the Default Rate

The federal loan programs have several disadvantages.

According to Tate (1979), abuses in loan programs occur at

each of the three stages through which any student loan

passes: (1) eligibility; (2) disbursement; and (3) collec-

tion. Both institutions and students are responsible for

abuses apparent in the loan programs of today. Institu-

tional abuses range from mismanagement to halfhearted

collection efforts and student abuses include outright fraud

in supplying eligibility data to shrewd bankruptcy declara-

tions in order to discharge outstanding debts. The most

obvious abuse regarding the loan system is the default or

lack of repayment by student borrowers.

By statute a default is defined as an overdue loan pay-

ment of 120 days (20 U.S.C. & 1080 (c), 1976). However, the

Department of Education does not recognize a default until

the lender's obligation is suspended and the subrogation by

the Federal Court takes effect.

Regardless of the procedural issue concerning the

commencement of the default status, the American public is

constantly reminded of the thousands of students who have

participated in the federal loan programs and who have de-

cided not to repay their loan obligations. Headlines in

papers consistently highlight the default rate and recent

efforts to initiate novel collection methods such as im-

pounding automobiles have received national television

attention.











The "default rate" is computed as a ratio of the cumu-

lative amount of delinquent funds to total receivables and

can be measured by the following formula:


cumulative claims paid to lenders

all loans having entered repayment status.

Current researchers estimate the overall default rate to be

around 12%; however, the rate is slightly lower in the GSL

program because the financial institutions exercise more

selective lending practices and the guarantee agencies ex-

ercise more administrative vigilance.

There is much debate concerning the reliability of data

regarding the default rate. Arthur Hauptman (1978) argues

that loan default rates are not consistent due to several

factors:

(1) the measurement standards changed
and the data prior to 1973 are not
consistent with the data collected
after 1974;

(2) the default rate is conceptually
flawed since it (a) uses only the
federal expenditures in the numer-
ator (federal expenditures are only
80% of the costs) and (b) includes
death and disability defaults; and

(3) the default rate is cumulative;
therefore, data for each year's
statistics are not sorted out
(not everyone defaults in the
first year of repayment). (Hauptman,
1978, p. 160)

Hauptman concludes that a single measure of default is in-

adequate to explain levels and trends in the behavior of











students who default because default activity differs by the

type of guarantee (FISL or GSL), the type of educational

institution the borrower attends, the type of lender and

the demographic characteristics of the borrowers.

In addition to the internal problem concerning the re-

liability of the default rate statistics, there is much con-

cern about the use of the default rate as a performance

indicator regarding the success of the student loan programs.

As cited by Hauptman, an Office of Education Report (1979)

illustrates the three major weaknesses of this argument:

(1) too many variables beyond the con-
trol of the Guarantee Agency have a
significant impact on default rates
(i.e., depression/recession, high
unemployment rates, etc.);

(2) loan default rates may only reflect
the availability of loans (higher
rate may be due to significantly
more loans arranged); and

(3) cumulative default rates are his-
torical and may not reflect current
management philosophies. (Hauptman,
1978, p. 162)

In effect, two conclusions can be reached based on the

previous discussion:

(1) the tendency to default may have
increased over the last several
years but it may not be as high
as it has been reported, and

(2) a large number of variables which
are external to the guarantee
agency and the financial lender
affect the default rate.











The Problem of This Study


Regardless of the elusiveness of concrete reliable

statistics, the public perceives the default rate as too

high. The federal government's determination to stem the

default rate must include research on the socio-economic

characteristics of student borrowers and defaulters in

addition to data concerning administrative procedures.

Accordingly this study has attempted to provide data con-

cerning some of the variables that may be related to the

student default rate.

In addition, these variables may be useful in formulat-

ing a prediction model to determine who will repay their

student loans. Such a model would be of valuable assistance

to the student aid administrator in reducing the default

rate because it would distinguish the characteristics of

defaultees and non-defaultees. The aid administrator could

then award and approve loans for only those students with

the greatest likelihood of repayment. Since loan money may

not be as readily available in the future as it is today, a

model that could provide assistance in selecting the "best"

student loan risk may be critical in maintaining the integ-

rity of the loan programs and in replenishing the source of

funds for future loans.











Statement of the Problem

The problem of this study was to determine the relative

influence of selected variables which characterize community

college students who are most likely to repay student loans.


Questions

Specifically, a primary question of this study was to

determine the usefulness of the model derived from the data

presented in this study to predict those students likely to

repay student loans from those students who are not likely

to repay student loans.

Therefore, the following supplemental question was

addressed: will the selected variables be useful in discrim-

inating between students who will repay student loans and

students who will not repay student loans?


Justification for the Study


As indicated previously, throughout the last few years

concern regarding student loan default rates has risen tre-

mendously. The issue has become the source of critical edu-

cational fiscal management decisions and delicate political

debates. During the period from September, 1982, through

February, 1983, The Chronicle of Higher Education reported

on 17 articles relating to recent developments concerning

student loans. Thus, the selection of this issue is











justified on its standing as a practical problem facing

higher education administrators today. Discussion relative

to some of the many student variables contributing to the

default rate is imperative. This study analyzes six vari-

ables (age, marital status, sex, college standing, grade

point average and loan total) identified in the previous

research of Dyl and McGann (1977), Myers and Siera (1980),

Emmert (1978) and other researchers for the impact these

variables have on the student default rate.


Delimitations and Limitations


Since there are several procedural issues impacting the

results of this study, the following delimitations and limi-

tations are applicable.


Delimitations

(1) This study was confined to students
who have received a loan through a
program administered by a Florida
public community college.

(2) Only data made available to the
researcher by the Student Financial
Assistance Commission of the State
of Florida were utilized in this study.

(3) A sample approximating 31% of the
total population was selected by a
systematic random process for both
categories of students identified
in the analysis (i.e., students who
repay student loans and students
who do not repay student loans).

(4) The analysis of data was confined to
the discriminant analysis procedure
utilized by the computer











techniques of the Statistical
Analysis System (SAS), which pro-
duced a set of equations from
which a predictor model could be
evolved.


Limitations

(1) This study was limited to those
variables identified in the re-
search design, specifically:
college standing, age, marital
status, sex, grade point average,
and loan total.


Research Design and Procedures


This study was designed as an ex post facto research

study in which several continuous and discrete independent

variables obtained from data collected on loan application

forms were observed. From these variables, the dependent

variable of loan repayment was predicted.

According to Ary, Jacobs and Razavich (1979), ex post

facto research can supply much information of value in edu-

cational decision making even though its design differs sig-

nificantly from pure experimental research. In an ex post

facto research design an experimenter can study two groups

that are different in some respect in order to discover the

reasons for the difference. Such is the case with this

study. The design format as suggested by Ary et al. is as

follows:













Group Independent Variable(s) Dependent Variable


E (X) (variable) Y1 (Repaid Loan)

C Y 2 (Did Not Repay
Loan)

The supposed effect of treatment occurs in the experimental

group. Therefore, according to Ary et al. (1979), the re-

searcher attempts to relate the dependent variable Y1 to a

previously occurring independent variable, a nonmanipulable

variable indicated by (X) which occurred in the experimental

group but not in the control group.

The purpose of utilizing the ex post facto research

design was to illustrate the significant data to be util-

ized in a model that could be useful in predicting the re-

payment rate probability of students requesting loans for

educational use at community colleges. Such a model would

be helpful to financial aid personnel in determining what

student characteristics are representative of those students

who have repaid their student loans.


Sample

The data for this study were selected randomly from

the total population of students in each of the following

two categories:

(1) students who are repyaing their
student loans and/or are not











identified as defaulting on
their student loan commitment,
and

(2) students who are not repaying
their student loans and have
been identified as defaulting
on their student loan commit-
ment.

These categories were determined based on the data

collection and storage procedures established by the Florida

Student Financial Assistance Commission (FSFAC) in Talla-

hassee, Florida. The FSFAC is the central storage/retrieval

center regarding GSL repayment data for all institutions in

Florida. The sample size (76) represents approximately 31%

of the total population in each category involving loans made

to community college students in Florida.


Data Collection

Data on each student included in this study were

supplied by the FSFAC to the researcher in June, 1982.

During this session files were randomly selected by FSFAC

personnel for review by the researcher.

The appropriate data were then extracted from the files

for those students who were identified as community college

students. Since there are two separate files maintained by

the FSFAC (i.e., those who have not defaulted, those who

have defaulted) the researcher obtained a sample for each

category representing 31% of the total population.











The sample size (38 in each category) meets the re-

quired minimum (30) sample size suggested by Ary et al.

(1979) and the 31% sample size exceeds the 10 to 20% sample

size suggested by Art et al. for descriptive research.

The following variables were extracted from the data

files based on their availability and their identification

in previous research studies by Dyl and McGann (1977), Myers

and Siera (1980), and Emmert (1978):

(1) Age,

(2) Marital Status,

(3) Sex,

(4) College Standing,

(5) Grade Point Average, and

(6) Loan Total.

A full discussion of the selection of these variables and

the results of previous studies utilizing these variables

appears at the beginning of Chapter III.















CHAPTER II

REVIEW OF RELATED LITERATURE


The financial pressures exerted on colleges and univer-

sities have had a direct effect on the expansion of loan pro-

grams. As a result the loan programs are experiencing some

unique problems. An analysis of the responses to the prob-

lems surfacing in the student loan programs and a review of

the research regarding student loan default and student

demographics provide an overview of the issues surrounding

the problem presented in this study.


Fiscal Pressures and the Demand
For Expanded Student Loan Programs


Higher education is a social institution that depends

on tuition and service charges, philanthropy, and govern-

mental support in order to balance its budget. The heavy

reliance on these sources of support has created a dilemma

concerning the funding of higher education during weak eco-

nomic times. The current fiscal crisis can be readily con-

firmed by reviewing the funding sources and relative support

for higher education since 1950.

In terms of tuition and service charges, there has been

a percentage decline in these fees as a source of income











from 1950 to 1974 (Change Magazine, 1976). This decline

has been attributed to the larger percentage of students

attending public colleges in 1974 (76%) compared to 1950

(50%) and the corresponding lower tuition charges at these

public institutions (Change Magazine, 1976). In addition,

colleges and universities have found themselves in compe-

tition with local facilities for the student dollar regarding

auxiliary charges including food, housing and books. There-

fore, there has been a decline in income from this revenue

source as well.

Although all service charges such as room and board

accounted for approximately 40% of the total income in

colleges and universities in 1974, Kelly (1983) estimated

that tuition charges in 1977 accounted for only 11% of the

total income for colleges and universities. As a result of

present trends on college campuses, Kelly (1983) estimates

that this percentage will increase to 20% by 1988. Based

on this estimate it appears that students will be forced

to bear an increased percentage of the cost of their educa-

tion in the form of increased tuition charges. Correspond-

ingly there is likely to be additional pressure on the

student aid delivery system to compensate for this in-

creased cost. However, many colleges and universities are

fearful of an increased dependency on student fees. These

colleges and universities are justifiably concerned that

the resultant increased fees will add a new pressure to











the university budget--declining enrollments and the loss

of dollar income from student charges based on underenroll-

ment.

In terms of philanthropical support, higher education

has been the beneficiary of increased dollar support from

1950 ($200 million) until 1974 ($2.4 billion) (Change Maga-

zine, 1976). As indicated in Table 1, due to the tremen-

dous rise in the cost of higher education, this increased

dollar support represents an actual percentage decline from

1950 (8%) to 1974 (5%) in terms of total income for colleges

and universities (Change Magazine, 1976).


Table 1
Philanthropical Support
for Higher Education



Year Total Dollar Percent of Income for
Support Colleges & Universities


1950 $200 million 8%

1974 $2.4 billion 5%



Since the dollar increase is not consistent with the needed

percentage increase for institutional philanthropical sup-

port, more of the cost of higher education will have to be

absorbed elsewhere.

In the recent past, the government (federal, state,

and local) has provided this increased support for higher











education. In 1974 it was estimated that the government pro-

vided 54% of the total income to colleges and universities

(Change Magazine, 1976). This increase in total income

represents additional support for both private and public

institutions as it includes research monies, building

assistance, and student aid dollars. However today, govern-

ments whose support figures so largely in the financing of

higher education are concentrating on other major concerns

and priorities. Therefore, there has been a decline in the

rate of income growth to higher education from the govern-

ments. For example, the federal government is under con-

trary pressures to hold down expenditures in student aid

(which has risen 600% from 1967 to 1977) on the one hand and

to respond to the continuing demands of middle income par-

ents for selective relief (i.e., additional aid, tuition tax

credits) on the other. In addition, many concerns outside of

higher education are uppermost in the minds of our legisla-

tors. For example, state legislators are highly conscious

of the "Proposition Thirteen" sentiments in their districts.

Therefore, the new depression in higher education is

generally not based upon declining support, but it is due in-

stead to a decrease in the rate of increase of support for

higher education. Nowhere is this more evident than in the

dollar support for student aid during the last two fiscal

years. Colleges and universities have received essentially

the same amount of support in total dollars from the











government, but as the rate of increase has dropped, increas-

ed pressure to raise fees and start fund raising activities

has surfaced. Because the slowdown has not been uniform,

some states and individual institutions have had to contend

with declining income.

As a solution to the pressure added by increased fees,

the Carnegie Council on Policy Studies in Higher Education

(1979) has recommended expanding explicit self-help compon-

ents (work-study and loan opportunities) of the student aid

delivery system and revising the loan system to make loans

more easily attained and widely available with less risk of

default (i.e., longer payback terms, loan consolidation

procedures).

It appears that the Reagan administration has seriously

considered the Carnegie recommendations since they have pro-

posed a 60% increase in College Work-Study funds, a "self-

help" requirement for Pell Grant eligibility and changes in

the Needs Test for the Guaranteed Student Loan Program for

fiscal 1984. In addition, Mr. Reagan has proposed continuing

dollar support for the federal grant programs at the same

level as the past two fiscal years. However, under his pro-

posal, eligibility qualifications for individual students have

been substantially altered. Therefore, it is safe to assume

that as fees increase, endowment expansion decreases and fed-

eral grant money remains the same, an increasing number of

students will seek to participate in the student loan program.











A potential problem arising from the decision to rely

on the student loan programs for the increased dollar support

to higher education is that the private lenders simply may

not have funds they want to lend to students (as other loans

may be more financially advantageous) or there may not be

enough loan money to meet recognized need. Further, the

loan programs in effect today have many problems. Therefore,

before considering an expanding role for student loans in

the financing of higher education, it appears the government

should consider the weaknesses of the current loan system.


Problems Inherent in Student Loan Programs


A closer examination of the student loan default prob-

lem reveals several contributing factors. Therefore, nearly

all researchers agree that the student should shoulder only

a part of the blame for the escalating default ratio.

Foremost among the factors to consider is the history

of incompetent management of loan programs. For years nei-

ther the lender nor the government was making any noticeable

effort to collect many of the outstanding loans (Jenkins,

1978). In addition, Jenkins noted that the Office of Educa-

tion could not afford to install a computerized collection

system until 1978 (Jenkins, 1978). Although governmental

investments in the area of administrative strategies have

resulted in much improvement in collection procedures,

the previous administrative procedures account for a










substantial part of the default problem attributed to the

historically impacted default rate (Jenkins, 1978).

A second factor to consider is the nature of student

loans. Kendis (1978) states that even though student loans

are not the same as consumer loans, the lack of rules con-

cerning asset and income analysis, borrowing limits and

collection procedures are "diametrically opposed" to method-

ologies long since proved successful in consumer credit

(Kendis, 1978). Although Kendis feels that the rules should

be somewhat different from the rules utilized in consumer

credit situations, he states that the result is an "attitude

problem" surrounding student loans. This "attitude problem"

affects not only students but lenders and legislators as well.

Students borrowing as a part of their investment in their

own human capital are forced into a decision process to attend

or not to attend a college or university, wherein the negative

consequences of increasing costs, and debt, are minimized by

the ability to borrow funds. Lenders continue to lend greater

volumes of funds leading to a greater median debt loan (with

increasing financial concern) in the interest of assuring access:

and individual collegiate choice. And finally, the Government

in its effort to assume capital availability guarantees loans

by merely increasing existing programs while disregarding the

long-term costs and the social consequences of the program.

Any one of the conditions highlighted previously would have










significant impact on the student loan problem; however, in

combination there is no doubt that the "attitude problem"

suggested by Klendis is a very real factor affecting student

loan defaults.

A third major factor affecting the student loan problem

is the mechanism for repayment. Balderston (1970) notes in

an analysis of incomes between those students choosing to go

to college and those choosing not to go to college that the

former will not "catch up to" the latter in real dollars

realized until very late in the working life of the individ-

ual primarily due to the foregone earnings (lack of income

during college years) of the college student (Balderston,

1970). In fact, in some instances the college-going student

may never catch up to the non-college student whether the

dollars are discounted for inflation or not. Therefore,

any student researching the wisdom of the investment of a

college education from a purely financial perspective may

decide it is not worthwhile to attend college based on the

market payoff (Balderston, 1970).

Beyond the decision the student must make regarding the

wisdom of attending college, if a loan is utilized, then the

repayment mechanism that currently operates which seeks to

concentrate the repayment in the early years of one's work-

ing life (10 years) may have some deterrent effect upon

college attendance. If loan financing is heavily relied

upon, the effect may be substantial. Balterston's











conclusion is that short repayment periods lead to fewer

students opting to attend college if loans must be utilized

because they impose too heavy a burden of cash outflow on

the student. He also notes the effect is stronger on stu-

dents who do not finish programs or finish programs with

less optimistic forecasts of projected future incomes (i.e.,

humanities, home economics) because these students do not

have the advantage of cashing in on a completed or market-

able degree and the loan payment constitutes a higher per-

centage of their total income than it does for students with

higher incomes.

The solution proposed by Balderston is to make long

term loans and to tie the repayment mechanism into the in-

come tax system or set up a similar system which reduces

administrative costs (Balderston, 1970). Although there is

much debate as to the wisdom of Balderston's proposal,

especially as it relates to the use of the income tax system,

there appears to be little doubt that the repayment system

for current loans is a component of the overall problem

regarding loan defaults.

Many other factors are cited in the literature as con-

tributing to the student loan problem. For example, Jenkins

notes that there is such a "smorgasbord" of student assist-

ance programs that it is possible for one student to be

eligible for eight different loan programs, sponsored by

eight different banks or lending institutions to pay eight











different terms with eight different sets of conditions for

forgiveness and eight different groups to deal with (Jenkins,

1978). Some consolidation appears to be necessary.

Another factor involves the apparent lack of sensi-

tivity to issues of student loans from student financial aid

administrators. For example, the student aid packaging pro-

cess usually relegates the discussion of the terms of a

student loan to the exit interview. Certainly this factor

is within the control of the student aid offices. Hope-

fully, these offices will incorporate the suggestions made

in the National Association of College and University Busi-

ness Officers (NACUBO) manual in order to manage better the

loan programs.

Finally, a legal process has been cited as a major

factor contributing to the student loan problem. The avail-

ability of a bankruptcy action to discharge a student loan

debt has become increasingly apparent to many student borrow-

ers. The use of a bankruptcy action, its effect on the

system of student aid and the governmental response to its

use are important considerations in student loan research

because they demonstrate an area where judicial and legis-

lative cooperation has assisted in reducing the number of

the student loan bankruptcies and encouraging more students

to repay their student loans.











Bankruptcy Actions

Bankruptcy is the mechanism by which insolvent debtors

may be released from the obligation to repay their debts.

The proceedings are handled by a federal bankruptcy court

and conducted under laws enacted by Congress. In the mid-

1970s the problem of student loan bankruptcies received sig-

nificant attention from Congress, colleges and universities,

professional organizations and the courts and governmental

agencies. This attention was not unwarranted as the problem

became quite prominent. During the five fiscal years from

1966 through 1970 only 348 bankruptcies were discharged on

NDSL and GSL loans totaling $400,000. However, in the fis-

cal year 1975 alone, a total of 4,559 bankruptcies were dis-

charged on loans totaling $6,800,000 (Leonard, 1980).

The effect of the tremendous rise in the use of bank-

ruptcy actions to discharge student loan debts resulted in

a great deal of negative publicity for all student aid pro-

grams. Prior to the disclosures regarding the use of bank-

ruptcy actions, support for all federal aid programs was

exceedingly strong and resulted in the passage of the BEOG

(Pell Grant) Program. After considerable press exposure

(i.e., New York Times) public resentment led to a hearing by

the House Subcommittee on Postsecondary Education in October,

1975. The resentment was specifically aimed at a very few of

the 12 to 15% defaulting student borrowers, since very few











students default through bankruptcy. Most defaulting stu-

dents choose not to provide forwarding addresses and thus

are "lost." According to some researchers, the small number

of students abusing the bankruptcy process solely to escape

federal educational debts is not a serious threat to the

loan programs. However, the anger surfacing at the October

committee hearing appeared to be generated by the apparent

misuse of a constitutionally sanctioned remedy to escape a

moral obligation (Tate, 1979).

Consequently, in 1976, Congress passed the Education

Amendments of 1976 which changed the procedures utilized to

discharge student loan debts by setting out specific restric-

tions. The primary restriction was a mandatory five-year

delay from the commencement of the repayment period in the

filing of a bankruptcy action. Although a subprovision

allows for actions based on undue hardship, the purpose of

the amendment is to enforce a moratorium on all but the most

necessary bankruptcy discharges until sufficient time has

passed for a loan holder to either pay off the debt or have

enough at stake to make bankruptcy a genuine last resort

effort (Tate, 1979). This provision has since been incor-

porated into the Bankruptcy Reform Act of 1978.

The Congressional action highlighted above indicates

that student loan problems can be adequately addressed and

resolved. However, some writers feel that the most potent

weapon against bankruptcy abuse and student loan defaults











may rest in the college registrar's office. Therefore,

actions taken by institutions as well as professional organi-

zations, the government and professional educators are worth

researching as possible weapons in combatting the high stu-

dent loan default rate.


Institutional Responses
to Student Loan Defaults


Nearly every college has a statement in its catalog in-

dicating that the college reserves the right to withhold

grades, degrees, statements of honorable dismissal or a

transcript of credits until satisfactory settlement of col-

lege fees and other financial obligations has been made.

This strategy has worked well for many years at several in-

stitutions and is probably partly responsible for keeping

the default rate within its current level.

The appropriateness and legality of utilizing this

strategy have been questioned in several court cases; how-

ever, the courts have not provided a definitive response to

guide colleges and universities especially if the student

has discharged the debt by bankruptcy action. It appears

that in situations where the debt was discharged prior to

October 1, 1979, the court will require state institutions

to release the transcript (Handsome v. Rutgers University,

445 F. Supp. 1362). However, private institutions may have

the authority to withhold transcripts of students whose

debts were discharged prior to October 1, 1979 (Girardier v.












Webster College, 563 F. 2d 1267). Due to the change in the

bankruptcy code, students are not now likely to have their

educational debts discharged, but if they do, it seems that

the college cannot take any action (including withholding

transcripts) to collect the debt.

The Handsome and Girardier cases point out that even if

the debt to the college has been discharged by a bankruptcy

decree there are other legal obstacles to the practice of

withholding transcripts. A future court may decree that the

student has sufficient property interest in the transcript

to require the college to provide adequate notice prior to

withholding the transcript. Or a court may determine that

there exists between the student and the college a contract

and that the college must demonstrate that the contract was

entered into with adequate notice regarding the withholding

of the transcript. In either case, additional responsibility

may be placed upon the colleges utilizing this strategy to

compel students to pay their educational debts.

In addition, courts have indicated that an unpaid debt

must be one that is sufficiently related to the student's

acquiring an education to justify retention of the transcript.

This decision presents several new issues. For example, is

a student's unpaid bookstore bill or a student parking fine

"sufficiently related" to permit the college to withhold

the student's transcript?











A final consideration regarding the withholding of

transcripts rests with the Family Educational Rights and

Privacy Act (Public Law 93-380). This act requires an edu-

cational institution to allow the inspection of a transcript.

Although this act does not require a college to release an

official transcript, a court may rule in the future that

the institution must release an unofficial copy of a tran-

script.

Obviously the strategy of withholding transcripts has

been viable throughout the years, but educational institu-

tions should consider the legal ramifications of their

policies in light of the recent court decisions and the

questions presented in the court dicta.


Professional Organization Action
Regarding Student Loan Defaults


Beyond the strategies utilized by colleges and univer-

sities to encourage student borrowers to repay their educa-

tional loans, a few professions have begun policing their

members regarding loan defaults. The most notable example

is the case of the Florida Board of Bar Examiners re G.W.L.

(364 So 2d 454, Fla. 1978). In this case the Florida

Supreme Court refused admission to the state bar to the

petitioner, who had discharged his student loans in bank-

ruptcy, on the premise that the petitioner's financial










status and apparent motivation at the time of filing re-

flected a sufficient lack of the proper moral character and

requisite values to practice law in the State of Florida

(Martin, 1980). Since the court never questioned the act

of filing for bankruptcy but rather based its decision on

the student's conduct and the apparent breach of the

Florida Character Standards for bar admission (which were

established to protect the public interest), the decision

did not conflict with federal bankruptcy law regarding un-

lawful penalties imposed on bankruptcy petitioners.

This court decision was the first to link discharge in

bankruptcy to moral turpitude and it has been cited in cases

in other states with similar provisions for professional li-

censing. The purpose of the standards as stated by the

Minnesota Supreme Court is not to punish the attorney but to

guard the administration of justice and to protect the gen-

eral public (in re Peterson, 274 N.W. 2d 922, 925, Minn.

1979). Therefore, the legal profession has set a precedent

by refusing bar admission to those who have utilized the

student loan system and irresponsibly neglect their debt

obligation. Since this strategy appears to be useful for

professional student bankruptcy actions and defaults, other

professional organizations, for example the AMA or ADA,

should institute and enforce similar provisions for licen-

sing. This action is particularly relevant since the











pursuit of higher education is valued so greatly by the Amer-

ican public in whose interest the professional serves and

any interference with the financial aid system supporting

the program of higher education in America should not be

treated lightly.


Governmental Action Regarding
Student Loan Defaults


Many governmental organizations and agencies have demon-

strated concern for the student loan default problem and have

initiated reports and actions to help remedy the problem.

In addition, the U. S. Congress has been involved with legis-

lation designed to curtail the default rate among federal

employees and in providing a forum in which to discuss

possible solutions for the loan default problem.

The General Accounting Office (GAO) has published sev-

eral reports concerning the student loan system. Some of the

suggestions listed in its 1977 report included the following:

(1) require financial statements of
all debtors,

(2) refer all defaults immediately
to legal action,

(3) set up guidelines for compromise
settlements,

(4) institute a payroll deduction
system for all federal employees,
and

(5) require any contractor for collec-
tion service to document unsuccess-
ful efforts. (U.S. General Account-
ing Office, 1977)










Although these suggestions seem realistic, they have met re-

sistance from several sources. For example, in 1979 the

U. S. Department of Health, Education and Welfare (now the

U. S. Department of Education) argued successfully that only

it should negotiate settlements of compromise. In a subse-

quent (1981) report the United States Comptroller General in-

dicated that many schools visited by the GAO made inadequate

efforts to collect defaulted loans, that the schools needed

to adopt a tougher attitude toward collecting defaulted

loans and that the Department of Education should take

stronger actions against those institutions that failed to

do so. Consequently, in April of 1982, the Department of

Education notified colleges with a poor record of loan

collection that they would not receive any new federal loan

money appropriated for the 1982-83 year. Although these

actions referred specifically to the NDSL program, they

sufficiently demonstrate the interest in Washington regard-

ing the student loan problem.

During the Oversight Hearing before the Subcommittee

on Postsecondary Education of the Committee on Education and

Labor (House of Representatives) the following methods were

suggested as appropriate to cut down the GSL default rate:

(1) discharging federal employees
who default,

(2) extending deferments to one year
for those who are unemployed,











(3) permitting lower rates of
repayment for those with low
incomes, and

(4) prohibiting defaulters from
obtaining other federal loans.
(U. S. Congress, 1977)

In response to item number one above, Education Secre-

tary Terrence Bell and Senator Charles Percy (R-Ill.) spon-

sored legislation in 1982 enabling the government to garnishee

the wages of federal employees who have defaulted on their

student loans. Recently, the Department of Education sent

letters to federal employees notifying them of the Debt

Collection Act of 1982 and indicating they must make

arrangements for the repayment of their loan by February 1,

1983,or risk the garnishment of their wages. Therefore, it

is apparent that both the Executive and Legislative branches

of the U. S. Government have also taken actions to assist

in the solution of the loan problem.


Responses to the Student Loan
Default Problem by Education Spokesmen


Professional educators have also addressed the problem

of student loans and are often quoted in the literature of

higher education. For example, Robert Hartman identifies

several factors that impinge on the effectiveness of the

current student loan system including the variety of loan

programs and regulations, the use of bankruptcy actions to

discharge debts and the differences among students and











their respective needs. He recommends several restrictions

to the current loan program. Among his suggestions are:

(1) the development of a single
loan instrument,

(2) curtailing the use of bank-
ruptcy actions (at least for
the first 10 years rather
than the five years provided
in the Education Amendments
of 1976), and

(3) a provision allowing different
loan amounts for students of
different needs, and extended
payment and variable repayment
options. (Hartman, 1978, p. 96)

Hartman argues that his suggestions would increase the cost

effectiveness of the loan program (items one and two), in-

crease the manageability of loan repayment procedures (items

one and three), widen student accessibility (items one and

three), and reduce defaults and delinquency (items one and

three). Essentially he is suggesting that the loans be

conformed to fit the student, that the loan programs be

supported identically by each of the fifty states and that

the performance requirements and benefits of the loan pro-

grams be identical regardless of the program utilized.

Hartman's suggested actions seem to be consistent with the

goals of the student aid delivery system and show great in-

sight concerning the many elements constituting the student

loan problem.

Several other plans have been suggested by professional

educators regarding the student loan program. A panel of











advisors during the Kennedy Administration developed the idea

of an Educational Opportunity Bank. The proposal would have

allowed students to borrow money to cover their expenses at

whichever college they decided to attend. The loan would be

paid back over the next 30 to 40 years and an interest rate

of 1% of the students' gross income would be applied. Al-

though this program would have allowed students to pay their

own college expenses, Duke University and Yale University en-

countered several problems when instituting similar programs

including administrative and collection difficulties.

D. Bruce Johnstone, in his proposal of the National

Student Loan Bank, recommends that one new major govern-

mental agency should assume the responsibility for all stu-

dent loans (Johnstone, 1978). In addition, Johnstone sug-

gests that the colleges and universities (rather than the

lending institutions) should originate all loans and that

the repayment of interest should be tax deductible necessi-

tating a tie-in to the Internal Revenue Service. This

radical plan appears to be quite practical; however, it

tends to come into conflict with many existing offices,

agencies and banks as it would drastically change and/or

perhaps eliminate their participation in the student loan

system. Therefore, support for this proposal has never

materialized.

Most attention today regarding changes in the student

loan programs has focused on John Silber's Tuition Advance











Fund (TAF) proposal. The proposal incorporates several key

factors including

(1) loans would only be negotiated
for successful students (after
their freshman year),

(2) individual student TAF accounts
would have to be established
with the respective college
business office,

(3) involvement of the Internal
Revenue Service would be re-
quired in the transaction and
repayment procedures, and

(4) indefinite repayment periods
would be instituted since the
repayment provisions require
students to repay their loan
at the rate of 2% of their
annual gross income each year
until they have paid back 150%
of their loan. (Silber, 1978, p. 7)

To its credit the TAF proposal provides some novel

ideas such as treating loan payments as tax obligations and

pooling the income of married couples with loans for the

purpose of calculating their taxable income. However, the

plan has come under negative scrutiny as many have argued

that the proposal assumes that the burden of paying for an

education should rest with the student. Beyond the previous

arguments concerning who should bear the cost of an educa-

tion, many professionals fear this approach may lead par-

ents, private donors and the taxpayer to abandon their

support for higher education which would force an even

greater burden on the student.










In addition, although research studies show that a long

repayment period provides the best opportunity for a reason-

able return incentive for investing in higher education,

many bankers indicate that the TAF could never be self-

sustaining based on the proposed 2% interest rate over an

estimated 30 year repayment period. Other arguments against

the TAF include the feeling that any plan that provides for

the paying of higher education through long term student

debts will result in limiting equal access and opportunity

for higher education. Further, the tuition and enrollment

controls suggested to distribute TAF funds to institutions

would cause widespread disruptions in the higher education

system. For example, the expanded eligibility requirements

for private college and university students would enable them

to be eligible for a larger proportionate share of the

available money. Regardless of one's point of view, the TAF

represents a proposal which has provided healthy discussions

of the student loan system and the options available to

solve the problems.


Antecedent Studies Relating to
Student Loans and Demographic Characteristics


It is apparent that all the participants in the student

aid delivery system are concerned about the student default

rate on educational loans. The changes implemented thus far

in the student loan system have been instrumental in improv-

ing administrative procedures. For example, new and better











resources for collecting loans and monitoring repayment pro-

gress have been devised. A second approach to solving the

student loan problem should consist of researchers studying

the demographic characteristics of student borrowers in

order to distinguish those students who are likely to repay

student loans from those students who are not likely to repay

student loans. Thus far only a few researchers have explored

the characteristics of student borrowers in the last five

years.

In 1977, Dyl and McGann applied a multivariate discrim-

inant analysis to the problem of identifying defaulting and

non-defaulting student borrowers at the University of Wyoming.

They analyzed 33 different variables including class standing,

college attended, grade point average, amount of the loan,

academic major, marital status, living quarters, total amount

of university loans, and size of the monthly payments. Four

of the significant discriminators displayed direct positive

relationships with actual loan repayment behavior. The stu-

dents who were more likely to repay their student loans in-

cluded those students who

(1) had higher grade point averages,

(2) were married,

(3) were engineering majors, and

(4) chose high monthly payments.

Dyl and McGann also discovered three factors which were nega-

tively associated with repayment of student loans including











the (large) size of the loan, the residence of the student

in an apartment and the total amount of indebtedness to the

university (including previous loans). As cited by Davis and

Van Dusen (1978), this study, in concert with the earlier

studies of Pattillo and Wiant (1972) and Spencer (1974), was

instrumental in identifying for other researchers the factors

most worthy of analysis and providing a format in which to

research the characteristics of student borrowers.

Hauptman, in his 1978 study for the College Scholarship

Service, explored many of the same characteristics for stu-

dents participating in the GSL/FISL programs. One of the

most interesting findings of his study was that the follow-

ing groups of students had a higher percentage of defaults

than would be expected:

(1) students with low family incomes,

(2) female students,

(3) married students,

(4) black students, and

(5) junior college students.

Hauptman discovered that students attending public schools

had a lower default rate than those attending either private

or proprietary schools and that college and university

students had a lower default rate than junior college stu-

dents or vocational students. Although the results of this

study did not duplicate the results presented by Dyl and











McGann (1977), it is equally important because it researched

a variety of institutions and concentrated its data on one

loan program.

Also in 1978, Emmert studied the characteristics of

students defaulting on their NDSL loans and concluded that

the demographic make-up of student populations has an equal

or greater impact upon default rates than do internal admin-

istrative factors. He suggested that schools with higher

default rates are not necessarily administering their pro-

grams any less effectively. Instead he argued that they may

have students from substantially different backgrounds that

account for the different default rate. Emmert also cited

all the previous researchers in synthesizing the crucial

variables selected in his study. Among the variables

selected were:

(1) age,

(2) marital status,

(3) sex,

(4) college standing,

(5) grade point average,

(6) loan total, and

(7) employment status.

Like the Hauptman study (1978), the differing results Emmert

discovered seem to suggest that student demographic character-

istics should be weighted in order to establish variable de-

fault rates for the different postsecondary institutions.











In 1979 the Office of Education completed its first

study of student borrower characteristics. Although the

data collection was restricted to eight proprietary insti-

tutions, the study singled out several variables that may

impact the default rate. Essentially this study underscores

the necessity of establishing statistical validity by re-

searching individual types of institutions independently

since the results vary for the separate types of institutions

studied. Therefore, because of the unique demographic char-

acteristics of the proprietary student, this study concluded

that the higher default rate among proprietary institutions

discovered in the Hauptman study is understandable.

Finally, a study completed by Myers and Siera in 1980

concentrated on data available from the New Mexico Student

Loan Program for the New Mexico State University. Using the

Statistical Package for the Social Sciences (SPSS) Discrim-

inant Program, Myers and Siera attempted to develop a pre-

diction model to discern defaulters from non-defaulters.

The results indicated that the variables selected did not

lead to an accurate prediction of the likelihood of a stu-

dent defaulting on a loan. What prediction was possible

was not substantially different from what one might expect

to accomplish based on chance alone. Although the study did

not present an acceptable prediction model, the work of

Myers and Siera was instrumental in pointing out the need












to explore and refine the application of predictive models

to the student loan problem.


Summary


All of the personnel, institutions, organizations and

governmental agencies involved with student loan programs

have provided input and taken selective action to reduce

the student loan default rate. A majority of the strate-

gies implemented thus far have been concerned with the ad-

ministrative or judicial (bankruptcy) aspects of the pro-

blem. In order to deal effectively with the problem of de-

faults the studies cited demonstrate the necessity of iden-

tifying the characteristics of defaultees and non-default-

ees.

Several of the studies presented emphasized the cru-

cial demographic characteristics worth researching. Other

studies demonstrate the necessity of restricting the popu-

lation researched due to the inherent differences among the

students choosing to attend the different institutions of

higher education. Therefore, this study has restricted its

data base to community college students in the state of

Florida and researched the available data that corresponded

to the variables highlighted in the previous studies.

The student loan problem is not the creation of a

single causality but rather it is the result of the







54



attitudes, procedures, economics and actions of all of the

participants (students, lenders, aid administrators) in the

program. Certainly, additional study needs to be performed

regarding the administrative problems within the student

loan program. However, research regarding the students

who partake in the program has long been relegated to a

subsidiary status. It appears crucial that additional re-

search needs to be conducted to identify those variables

which are effective in predicting the likelihood of a stu-

dent default, because the student is indeed the major

factor contributing to the student loan default problem.















CHAPTER III

DEVELOPMENT OF THE MODEL


The research reviewed in Chapter II provided an over-

view of the demographic variables to consider in the forma-

tion of a prediction model for student loan repayment.

Names of the major contributors to the literature and the

primary variables they studied are shown in Table 2. As

presented in Chapter II many of these studies provided rele-

vant findings concerning relationships between demographic

characteristics of student borrowers and loan default. These

relationships will be reviewed as a basis for developing the

prediction model.

In one of the first studies researching demographics

and student loans, Pattillo and Wiant (1972) as cited by Davis

and Van Dusen (1978), discovered that students who borrowed

late in their academic years, who had previous loans, or who

came from large families were more likely to be delinquent

in their student loan payments (Davis & Van Dusen, 1978). In

addition, Pattillo and Wiant concluded that items reflecting

financial rather than biographical data appeared to be better

predictors of loan delinquency (Davis & Van Dusen, 1978).

Spencer completed a study of the relationship between

demographics and student loan default in 1974. In this study,








Table 2
Overview of Major Variables
Selected by Previous Researchers

(1972) (1974) (1977) (1978) (1978) (1979) (1980)
Pattillo Dyl & Office of Myers
& Wiant Spencer McGann Hauptman Emmert Education & Siera Total


Age X X X X X 5
Marital Status X X X X X X 6
Sex X X X X X X 6
College Standing X X X X X 5
Grade Point Average X X X 3
Loan Total X X X X X 5
Family Income X X X 3
Student Aid Received X X 2
Cost of Attendance X 1
College Major X X 2
Type of College X 1
Ethnic Background X X X 3
Employment Status X X 2
Academic Load X 1
Degree Earned X 1
Type of Residence X X 2
Dependents (Number) X X 2
Age of Automobile X X 2
Family Size X 1
Telephone X X 2


en











according to Emmert (1978), Spencer described the worst

possible student loan risk as an unmarried, unemployed male

student, in his last semester, with an old car, without a

phone, and with a large outstanding loan debt (Emmert, 1978).

The results that Dyl and McGann obtained in 1977 indi-

cated that the following factors related positively to the

repayment of student loans: college major (engineering),

marital status (married), grade point average (high), and

the size of loan payments (low). Dyl and McGann (1977) also

found the following three factors to be negatively associated

with repayment: residence type (apartment), total university

indebtedness, and loan indebtedness (high loan totals). In

this study Dyl and McGann also presented the results of a

discriminant analysis of their data, and they were able to

correctly classify 84% of the cases in their study. However,

as pointed out by Myers and Siera (1980), Dyl and McGann did

not test their model with any new cases. Therefore, the

reliability of the model presented by Dyl and McGann is, as

yet, untested.

Essentially these three studies presented the basis

upon which most of the research in this area has concen-

trated. These researchers have not only provided insight con-

cerning the appropriate variables to research but they also

have suggested the appropriate statistical designs for fu-

ture studies. However, the conclusion originally presented

by Pattillo and Wiant (1972) regarding the inappropriate use










of biographical data for predictive purposes has not been

heeded. Many researchers have continued to search for stu-

dent demographic characteristics which would distinguish

student loan defaulters from non-defaulters.

For example, in 1978 Hauptman studied five demographic

variables and reported that the worst risk student loan par-

ticipant was a married, Black, female student from a low

family income who attended a proprietary vocational school.

In another study presented in 1978, Emmert summarized the

previous studies reported in the literature. In this

article Emmert reported the variables that have been studied

and the conclusions that have been presented concerning the

effect these variables have had on the student loan default

problem. Emmert also noted the different default rates

among postsecondary institutions and suggested that the demo-

graphic make-up of a given student population may place an

institution in a high risk category for student loan defaults.

In conclusion, Emmert suggested that the demographic factors

should be weighted in order to establish variable default

rates for differing types of institutions.

A 1979 study prepared by the Office of Education also

reported student demographic data in relation to the problem

of student loan defaults. Although the study primarily was

designed to research the administrative procedures utilized

by selected proprietary institutions, it did show that many










but not all of the same variables selected by other research-

ers were important variables concerning the student loan

default problem in proprietary institutions. The findings

presented in this study demonstrate the necessity of study-

ing individually the different types of postsecondary insti-

tutions.

The most recent study in this area of financial aid re-

search was carried out in 1980 by Myers and Siera. Through

t-test analysis, Myers and Siera found the following vari-

ables to be highly associated with the repayment or non-

repayment of student loans: college standing, loan total,

amount of loan requested, grade point average, and degree

completion. Myers and Siera also attempted to replicate

the study design presented by Dyl and McGann (1977); and, al-

though they were able to formulate a promising prediction

model, when this model was tested using new cases only 42.5%

of the new cases were correctly predicted. The validity

problem experienced by Myers and Siera in their prediction

model illustrates the difficulty researchers have encountered

in their search for an accurate, useful prediction model for

student loan defaults.

It is apparent from the above discussion that although

the conclusions of the previous studies differ considerably,

there does appear to be some consensus regarding the impor-

tance of researching student demographics and which demo-

graphics should be studied. In selecting which variables











to use in this study, the researcher selected the variables

which had been most frequently used by previous researchers

(see Table 2). Although ethnic status and family income

were studied by several researchers, these variables were

not included in this study because of the difficulty in col-

lecting and verifying these variables. The remaining six

variables were available in the data bank of the Florida

Student Financial Aid Commission in Tallahassee, Florida.

As a result, the following six variables were selected for

the prediction model presented in this study:

(1) age,

(2) marital status,

(3) sex,

(4) college standing,

(5) grade point average, and

(6) loan total.


Analysis of Variables


As indicated in Chapter I, this ex post facto study was

designed to determine the usefulness of a model to distin-

guish those students likely to repay student loans from those

students who are not likely to repay student loans. Prior to

developing any model, there was one major issue to address:

were the selected variables useful in discriminating between

students who will repay student loans and students who will

not repay student loans?











In order to determine the usefulness of the variables

in discriminating those students who will repay student loans

from those students who will not, data representing each of

the six variables were collected for both defaulting and non-

defaulting students in the sample. Since three of the vari-

ables were nominal in nature, the t-test statistic utilized

by previous researchers to determine significant differences

between defaulters and non-defaulters was not applicable to

the data. Therefore, a chi-square analysis was performed on

the data collected for each of the variables in this study

to determine if the differences between the proportions of

subjects that fell into the two different categories (de-

faulters and non-defaulters) were significant.

Since this study represents exploratory research and

any model developed from the data will need to be validated

with new data, the researcher selected a .10 significance

level. Essentially the chi-square analysis determines if

the difference between the expected and observed frequencies

for each of the classifications within each variable is

greater than the difference one would expect to find based

on chance alone. The results of the chi-square analyses

showing the relationship of each of the six variables to

student loan default are presented in Table 3 through

Table 9.











Table 3
Chi-Square Analysis of the
Relationship Between Age and Student Loan Default


Age Defaulters Non-Defaulters


18-20 Number 8 10
% of total frequency 10.53 13.16

21-24 Number 10 9
% of total frequency 13.16 11.84

25-27 Number 8 6
% of total frequency 10.53 7.89

Over 27 Number 12 13
% of total frequency 15.79 17.11



Chi-Square = 0.6010 DF = 3 Probability = .8960



As shown in Table 3, the chi-square analysis of the re-

lationship between age and student loan default indicates

that age was not significant at the .10 level. The expected

frequency for each cell based on chance alone would be 12.5%

and, as shown, the actual frequency percentage for the cells

ranged from 7.9% to 17.1%.

The chi-square analysis of the relationship between

marital status and student loan default is shown in Table 4.

The results of the chi-square analysis for marital status

indicate a statistical significance at the .10 level. The

expected frequency for each cell based on chance alone would

be 25% and, as shown by the table, the actual frequency per-

centage for the cells ranged from 14.5% to 35.5%. Therefore,










Table 4
Chi-Square Analysis of the Relationship
Between Marital Status and Student Loan Default


Status Defaulters Non-Defaulters


Married Number 18 11
% of total frequency 23.68 14.47

Single Number 20 27
% of total frequency 26.32 35.53


Chi-Square = 2.732 DF = 1 Probability = 0.0983



the use of the variable marital status in a subsequent pre-

diction model was appropriate.

As shown in Table 5, the chi-square analysis of the re-

lationship between sex and student loan default indicates

that sex was not significant at the .10 level.


Table 5
Chi-Square Analysis of the Relationship
Between Sex and Student Loan Default


Sex Defaulters Non-Defaulters


Male Number 18 13
% of total frequency 23.68 17.11

Female Number 20 25
% of total frequency 26.32 32.89


Chi-Square = 1.362 DF = 1 Probability = 0.2432










The expected frequency for each cell based on chance alone

would be 25% and, as shown, the actual frequency percentage

ranged from 17.1% to 32.9%.

The chi-square analysis of the relationship between

college standing and student loan default is shown in

Table 6. The data collected did not indicate college stand-

ing to be significantly related to defaulters and non-

defaulters at the .10 level. The expected frequency for each


Table 6
Chi-Square Analysis of the Relationship Between
College Standing and Student Loan Default


College Standing Defaulters Non-Defaulters


Freshman Number 21 17
% of total frequency 27.63 22.37

Sophomore Number 17 21
% of total frequency 22.37 27.63



Chi-Square = 0.8420 DF = 1 Probability = 0.3588



cell based on chance alone would be 25% and, as shown, the

actual frequency percentage ranged from 22.4% to 27.6%.

Table 7 shows the chi-square analysis of the relation-

ship between grade point average and student loan default.

The results of the chi-square analysis indicated that grade

point average was not significant at the .10 level. The











Table 7
Chi-Square Analysis of the Relationship Between
Grade Point Average and Student Loan Default


Grade Point Average Defaulters Non-Defaulters


1.50-1.99 Number 2 0
% of total frequency 2.63 0.00

2.00-2.49 Number 18 13
% of total frequency 23.68 17.11

2.50-2.99 Number 8 12
% of total frequency 10.53 15.79

3.00-3.49 Number 8 8
% of total frequency 10.53 10.53

3.50-4.00 Number 2 5
% of total frequency 2.63 6.58




expected frequency percentage for each cell would be 10%

based on chance alone; however, the actual frequency percent-

age for the cells ranged from 0.00% to 23.68%.

The chi-square results of the relationship between loan

total and student loan default are presented in Table 8. The

loan total variable was significant at the .10 level. The

expected frequency for each cell based on chance alone would

be 8.3% and, as shown, the actual frequency percentage for

the cells ranged from 0.0% to 18.42%. Therefore, the size

of the loan total was a significant variable to include in

any prediction formula designed to ascertain the likelihood

of a student loan default.











Table 8
Chi-Square Analysis of the Relationship Between
Loan Total and Student Loan Default


Size of Loan Defaulters Non-Defaulters



0-$1,000 Number 13 6
% of total frequency 17.11 7.89

$1,001-$2,000 Number 11 6
% of total frequency 14.47 7.89

$2,001-$3,000 Number 13 14
% of total frequency 17.11 18.42

$3,001-$4,000 Number 1 5
% of total frequency 1.32 6.58

$4,001-$5,000 Number 0 5
% of total frequency 0.00 6.58

$5,001-$6,000 Number 0 2
% of total frequency 0.00 2.63



Chi-Square = 15.7530 DF = 5 Probability = 0.0173




The results of the chi-square tests showed that only two

of the variables were significant at the .10 level (marital

status and loan total). The differences found between the

categories (student loan defaulters and non-defaulters) for

both of these variables cannot be attributed to chance alone.

The remaining four variables (sex, grade point average, age

and college standing) were not statistically significant

according to the chi-square results of this study, but they

may demonstrate some differentiation between defaulters and











non-defaulters on student loans based on the results of

previous researchers.

The results of the chi-square analyses indicated that

only two of the six variables studied most frequently by pre-

vious researchers were significant according to the data

collected in this study. The discovery that the only finan-

cial type variable (loan total) was also one of the two vari-

ables found to be significant may underscore the importance

of Pattillo and Wiant's (1972) conclusion that items reflect-

ing financial rather than biographical data appear to be

better predictors of loan delinquency. This finding may

indicate that since students with high loan needs have good

loan payback rates, the students with the highest needs are

the best loan risk for any one of a number of reasons (i.e.,

they may value money more due to its scarcity, they may have

learned to manage money more frugally or they may have

learned how to budget more effectively). The strength of the

relationship of this one variable to student loan payback

indicates that other financial type data should be studied.

Before developing a prediction model based on the re-

sults of the chi-square analyses, the researcher explored

the possibility of interaction effects between the six vari-

ables. These results are presented in Table 9. It can be

observed that no interaction between the variables related

significantly to payback.











Table 9
Interaction Effects Between the Variables


Variables F Pr)F Significant (co =.10)


Loan by Sex 1.45 .2320 No

Loan by Grade Point 0.01 .9185 No

Loan by Marital Status 0.10 .7573 No

Sex by Marital Status 0.05 .8250 No

Sex by Grade Point Aver. 0.47 .4956 No

Grade Point by Marital
Status 2.14 .1483 No




Therefore, in developing the prediction model, the researcher

utilized only the six original variables selected from the

literature review.

Although four of the selected variables were not signi-

ficant according to the chi-square results of this study,

they were included in the formation of the prediction model

because the literature was inconclusive regarding which vari-

ables should be used in a prediction model. It should be

noted that the unique results of the chi-square analyses in

this study may be due to the unique population. Since the

previous studies did not examine the relationship between

selected variables and student loan default for community

college students, the difference in results may be due to the











specific characteristics of the community college population.

Since the results presented in the literature and the results

found for the community college students in this study were

not corroborative regarding the significance of the vari-

ables, all six selected variables were utilized in the pre-

diction model.


Development of the Model


In order to try to predict categorical membership based

on the most discriminating variables, a multiple discriminant

analysis technique was employed. Essentially this statisti-

cal procedure "weights" the predictor variables to yield

maximum discrimination between the two groups (i.e., in this

study it discriminates defaulters from non-defaulters) (Hays,

1981). In addition, the nominal variables of marital status,

college standing and sex were used to define dummy variables

in order to introduce variance into the prediction equation.

Three notable formulae generated by the multiple discrim-

inant analysis function of the Statistical Analysis System

(SAS) program were useful in distinguishing student loan

defaulters. Therefore, three prediction models were devel-

oped.

The first formula included all six variables which were

entered into the data file in the following (arbitrary)

order: age, marital status, sex, college standing, grade











point average, and loan total. The results of this multiple

discriminant analysis indicate that the six selected vari-

ables correctly predicted group membership for approximately

68% of the sample cases. Considering that the prior proba-

bility (based on chance alone) would be a 50% correct pre-

diction, the six selected variables appear to be successful

in improving the accuracy of predicting group membership.

In order to improve the chances of making more accurate

predictions, a second discriminant analysis procedure was

performed. This time the multiple discriminant analysis

enabled the researcher to catalogue the individual effect of

each selected variable on the accuracy of the categorical

prediction. Based on the probability and significance

levels of the chi-square analysis the most differentiating

variables were first included in the prediction model. When

the first variable, loan total, was entered into the pre-

diction formula, the chances of making accurate predictions

improved notably. Instead of the 50% prior probability of

making accurate categorical predictions the accuracy rate

improved to 66%. The only other significant variable

(marital status) was entered into the model next. The in-

clusion of this variable increased the probability of

correctly categorizing the defaulting and non-defaulting

student loan recipient from 66% to 71%. Therefore, this

second formula (generated by the stepwise multiple











discriminant technique) included only the two variables

demonstrating the acceptable degree of significance based

on the chi-square results ( oP = .10) and it provided the

best accuracy for categorical prediction (71%).

When the next two variables, sex and grade point aver-

age, were included in the model, the percentage of correct

categorical classifications essentially remained the same.

Although the variables would appear to be useful in a pre-

diction formula as indicated in the literature review, their

inclusion was not helpful in formulating a more accurate

prediction model. When the variable sex was included in the

model, the accuracy rating remained at 71%; however, when

the variable grade point average was included, the number

of correct classifications decreased by one and reduced the

accuracy rating to 70%. Such a result seems to be incon-

sistent with statistical theory. After studying the values

for each of the sample cases generated by the prediction

formula, the researcher believes that the decreasing

accuracy may be due to the mathematical rounding procedure

within the Statistical Analysis System program. As a result,

one sample case that defaulted on a student loan was classi-

fied by the prediction model as a non-defaulter on the total

probability difference of .0033. Therefore, the apparent

decreasing accuracy of the prediction model may be attributed

to the mathematical precision of the computer program rather

than the reflection of a conflict with statistical theory.











The third formula was generated at the conclusion of

the stepwise multiple discriminant technique and it included

all six selected variables. The addition of the remaining

two variables, college standing and age, did not improve the

accuracy rating for the prediction model. Therefore, the

formula, including all six selected variables, generated by

the stepwise technique was able to predict correct categori-

cal membership for 70% of the sample cases.

Although both the first and third formulae included the

six selected variables, the disparity in the predictive

accuracy of the formulae is accounted for in the statistical

technique employed for the derivation of the formulae.

Since the stepwise multiple discriminant analysis technique

enters the variables in the order of probability, the im-

proved accuracy of the third formula (70%) is a result of

the mathematical process of weighting the variables as they

are entered into the prediction formula. Because the order

of inclusion in the formula is different and the mathemati-

cal weighting is improved, the accuracy of the formula is

improved.

Based on the results of the discriminant analysis, it

would appear that the best predictive formula for determin-

ing correct classification of defaulting and non-defaulting

community college students would include only data regarding

the students' loan totals and marital status. Adding the











variables grade point average, sex, college standing and

age, essentially maintained the effectiveness of the pre-

diction formula. As evidenced, the stepwise technique em-

ployed to produce the second and third formulae improved the

accuracy for categorical classification over the regular

multiple discriminant analysis technique utilized for the

first formula.



Results


Based on chance alone, one would expect to predict

correctly the proper placement of students into the two

categories (defaulting and non-defaulting) 50% of the time.

The prediction formula generated by the inclusion of data

concerning loan total and marital status increases the like-

lihood of making an accurate prediction from 50% to 71%.

Therefore, it would appear that these data are relevant in

predicting whether a student is likely to default on a

student loan. The additional data that were collected for

the other four variables (sex, grade point average, college

standing and age) also appear to be relevant in producing a

prediction model since the prediction formula derived from

all six variables produced a 70% accuracy rating concerning

categorical prediction and this rating is substantially

higher than the 50% probability based on chance alone.










However, all three of the prediction formulae reflect

an error factor of nearly 30% indicating that roughly one-

third of the cases were classified in error. In addition,

the testing of the prediction formula utilizing new cases is

not possible since the formula is derived from a stratified

sample (which was necessary in order to perform the chi-

square tests to determine variable usefulness). Therefore,

the usefulness of the results of this study is limited even

though the prediction formula demonstrates moderate success

in differentiating defaultees from non-defaultees.

Previous researchers also produced prediction models

with limitations. For example, although Dyl and McGann

(1977) reported significant success with their prediction

model, Myers and Siera (1980) indicated that Dyl and McGann

did not validate their results by applying the formula to

cases other than those from which the formula was derived.

In addition, the results of the study performed by Myers

and Siera seem to indicate that although they presented a

useful model, when they attempted to validate their model

by predicting new, non-sample cases, the model was not

reliable or useful.

The results of this study show that the six variables

which were selected based upon the results of the studies

presented by previous researchers were not reliable predic-

tors of payback for community college students. Therefore,

a dependable prediction model cannot be formulated for











community college students based on these variables. Perhaps

the importance of this study is that the results coincide

with findings reported by Hauptman (1978), Emmert (1978) and

Pattillo and Wiant (1972).

Hauptman first suggested that research must be indepen-

dently conducted on the different types of postsecondary in-

stitutions. Underscoring this perception, Emmert concluded

that institutional default rates may not be comparable if

the schools' respective populations are from substantially

different backgrounds. It may be significant that the two

studies previously successful in determining reliable demo-

graphic variables and formulating prediction models (Dyl and

McGann, 1977, and Myers and Siera, 1980) were based on four-

year public univeristy (University of Wyoming and New Mexico

State University respectively) samples.

Previous discussion regarding the importance of finan-

cial aid for community college students demonstrated the

meaningful differences between the typical community college

student and the typical four-year university student. The

disappointing accuracy of this model, specifically the in-

correct classification of 30% of the sample, may be attrib-

uted to the demographic differences between the sample in

this study and the samples used in the previous studies.

For example, the results of the previous research reported

in the literature indicate that the demographic data repre-

sented by the six selected variables discriminate student










loan defaulters and non-defaulters for four-year college or

university students. However, as indicated by the chi-square

results in this study, these same variables do not entirely

discriminate student loan defaulters from non-defaulters for

community college students. Therefore, the variables,

selected based on the success of the previous studies con-

ducted with a university student sample, are apparently not

the best variables to use when attempting to predict

community college student loan defaulters.

The results of this study also appear to support

Pattillo and Wiant's (1972) conclusion that financial rather

than biographical data are better predictors of loan delin-

quency. One of the two variables that met the .10 statistical

significance level in the chi-square analyses of this study

was the loan total. It was also the only variable that was

financially related since all of the other (five) variables in

this study represented purely biographical demographic charac-

teristics. Considering that the only financially related vari-

able was also statistically significant, perhaps more finan-

cially related variables would yield a more accurate prediction

model. Therefore, the results of this study may lead to the

conclusion that, as suggested by Pattillo and Wiant, more

financial data need to be collected and analyzed in order to

make any reliable prediction of loan delinquency. This

conclusion may be also applicable to the results presented

by Dyl and McGann (1977) and Myers and Siera (1980).










Although they have been able to develop successfully a pre-

diction model, neither Dyl and McGann nor Myers and Siera

has been able to validate successfully their models by pre-

dicting accurately new cases, cases not presented in the

sample that derived the prediction model. Perhaps they need

to collect more financial data regarding their students in

order to formulate an accurate prediction model that can be

validated by new cases.


Summary


The prediction formula derived from the data available

through the Florida Student Financial Aid Commission for

Florida community college students indicates that it is

possible to make a useful discrimination regarding who will

and who will not repay their student loans for approximately

70% of the sample cases. This prediction formula was

derived through a (stepwise) multiple discriminant analysis

computer program utilizing the six selected variables high-

lighted in previous research studies as useful discriminators

between defaulting and non-defaulting students. A second

(stepwise) multiple discriminant analysis utilizing only the

two variables demonstrating statistical significance

(c = .10), loan total and marital status, yielded a pre-

diction formula with a categorical prediction accuracy

rating of 71%. A third formula utilizing all six selected

variables entered into the formula in random order










(non-stepwise) generated a model that predicted correct

categorical membership for 68% of the sample. Regardless of

the formula utilized, approximately one-third of the sample

cases were not predicted accurately. Although none of the

prediction models are accurate for one-third of the cases,

each prediction model is capable of predicting nearly 70%

of the cases correctly. Therefore, the models are of some

value in addressing the questions of this study. Future

models may be more useful in predicting those students

likely to repay student loans from those students who are

not likely to repay student loans since the models derived

from the data in this study have limitations concerning

their application as discussed in Chapter IV.

The limited success of these prediction models appears

to be consistent with the findings presented by the previous

researchers in the field. Since the difficulty in validat-

ing even successful prediction models appears to be wide-

spread, as confirmed by the results of previous researchers,

it may be that the variables utilized thus far are not the

most discriminating variables regarding student loan de-

faults. Perhaps future research should focus on financial

type variables in order to provide relevant, statistically

acceptable prediction models to assist student financial aid

administrators in predicting student loan defaulters.















CHAPTER IV

CONCLUSIONS, IMPLICATIONS AND
RECOMMENDATIONS


Previous studies researching the relationship between

student demographics and loan repayment recorded similar

results to this study. However, the results of this study

demonstrate several unique implications for future re-

searchers.


Conclusions


Since the models presented in this study improve the

chances of making an accurate prediction concerning the

likelihood of default on student loans from 50% (based on

chance alone) to nearly 70% (based on the models), the

models are successful in discriminating student loan de-

faulters from non-defaulters. Therefore, the questions of

this study presented in Chapter I must be answered in the

affirmative.

Even though some of the six selected variables were not

statistically significant in this study, the inclusion of

all six variables in the prediction models was warranted

because of the inconclusive results regarding the importance

of the individual variables as demonstrated by previous

university researchers. Inclusion of all six variables was










also appropriate since this study was the first to utilize a

community college sample, and the results of the chi-square

analyses in this study may indicate that the variables

distinguishing community college student loan defaulters may

be different from the variables distinguishing university

student loan defaulters.

Despite the differing results of the chi-square analyses

for the individual variables, the models formulated utilizing

all six selected variables were accurate in predicting

default or non-default for nearly 70% of the sample cases.

Therefore, the models presented in this study are useful in

predicting student loan defaulters. The conclusions of this

study suggest several implications of importance for future

studies.


Implications


The most notable implication of this study is that

there are differences between community college students and

university students regarding the significant variables dis-

tinguishing student loan defaulters from non-defaulters.

This study shows that, although previous researchers identi-

fied the six selected variables and determined that they were

significant in distinguishing defaulting university students,

only two of the six variables were significant in distin-

guishing defaulting community college students. Therefore,

this study underscores the importance of performing separate










study designs yielding specific results for the different

kinds of institutions utilizing the Guaranteed Student Loan

system.

The results of the chi-square analyses also appear to

indicate that the data currently being collected for commun-

ity college students do not adequately distinguish defaulting

from non-defaulting student borrowers. As suggested pre-

viously, it appears that more financial data should be

collected and analyzed if one is to achieve a more accurate

prediction model for community college student loan de-

faults. Although there are inherent differences between the

administrative mechanisms employed for the student loan

system and those utilized by commercial lending institutions,

information found useful in determining eligibility for

commercial loans may be useful in distinguishing student

loan defaulters from non-defaulters. Therefore, a second

implication based on the results of this study is that future

research should provide more focus on the financial data of

community college students in order to improve the chances

of formulating a better prediction model.


Limitations of the Model

The results of this study reveal two major limitations

of importance for future researchers.

First, the models presented in this study do not

accurately categorize nearly one-third of the sample cases.











As indicated, an inaccurate prediction regarding the proba-

bility of default or repayment was made for nearly one-third

of the students in the sample. This imperfection results in

the limited application of the prediction models. Previous

researchers were unable to formulate a prediction model and

validate their models with any great success; similarly, this

study did not find a sufficiently accurate prediction model

for determining who will default on their student loan.

Second, since the models were not effective in making

an accurate prediction for all cases in the sample, it

appears that additional (especially financially related)

variables may be of value in future research studies. This

study was restricted to the use of data highlighted as sig-

nificant in the literature and available through the Florida

Student Financial Aid Commission (FSFAC) in Tallahassee,

Florida. The results of this study and the previous studies

seem to indicate that some variables which were not avail-

able through the FSFAC data files may be of critical impor-

tance in formulating a more accurate prediction model. For

example, the employment status and type of job held at the

time of loan application or the family's financial data in-

cluding loan repayment history may be reliable variables to

consider when predicting student loan defaults. In addition,

data concerning ethnic background, college major, and ad-

justed gross income were found to be of value in prediction

formulae presented by university researchers. Although the










data are not currently available in the data system provided

through FSFAC, perhaps these variables are also important in

the formation of a prediction model concerning the probabil-

ity of default for community college students. Any further

study regarding a prediction model for determining the prob-

ability of default should also research the value of includ-

ing these variables. Based on the difficulty of collecting

data concerning these variables, the inclusion of these

variables would alter considerably the methodology section

of any similar research study. However, the variables do

appear to be of value in examining critically the student

loan default problem and presenting a model that improves

predictive accuracy. Therefore, the resources employed to

collect the data would appear to be prudently utilized.


Recommendations


The use of any of these predictive models for deter-

mining the likelihood of students defaulting on their stu-

dent loans is limited. Therefore, the foremost recommenda-

tion derived from the data and the literature search is the

need to encourage further research. The importance of this

study is that the data indicate it is possible to formulate

a model which will accurately predict 70% of the students

within the sample. It appears realistic to assume that

further research, especially the financially related

research suggested in the previous discussion, may result










in a more inclusive and accurate prediction model which would

be useful in determining the probability of default. Such

research would need to be more inclusive and therefore the

collection of the new data would demand much more time and

financial resources. However, the possibility of finding an

appropriate collection of variables and the usefulness of

the resultant prediction formula would indicate that student

loan default prediction is an appropriate area for further

study.

In the event that a subsequent study does provide a re-

liable prediction model, it should be tested and validated

by the researcher using new cases derived from a different

sample. In addition, before the model is implemented on any

individual college campus, it should be tested and validated

by the financial aid officer for applicability on his/her

campus. Only when these validations are performed can there

be confidence in the predictive capability of the model, for

the results of this study and the research presented in the

literature clearly indicate that the models and appropriate

variables vary from institution to institution. This varia-

tion is dependent upon the type of institution being re-

searched. Another reason for establishing institutional

validation procedures is to determine if significant

differences occur among institutions of the same type. It

is possible that a single, reliable prediction model that is

derived from data representing a state-wide population for a







85



selected type of institution may have no validity or useful-

ness on a single campus.

Because of the increased emphasis on student loan pro-

grams as a major component of the student financial aid

delivery system, further research concerning student demo-

graphics and the probability of default is of utmost impor-

tance. Such importance is underscored by the discussions

presented in the literature search and the successful

results of this study.











BIBLIOGRAPHY


Ary, D., Jacobs, L. C., & Razavich, A. Introduction to
research in education (2nd ed.). New York: Holt,
Reinhart and Winston, 1979.

Balderston, F. E. The repayment period for loan financed
college education. New York: Ford Foundation, 1970.
(ERIC Document Reproduction Service No. ED 081 379)

Balkins, A. J. Withholding transcripts for non-payment
of educational debts: Before and after bankruptcy.
Williamette Law Review, 1979, 15, 563-575.

Bar Admission--Default on student loan warrants denial
of admission to Minnesota bar. William Mitchell Law
Review, 1980, 6, 443-454.

Block, W. R. Other students need money: An approach to
the administration of revolving loan accounts.
Journal of Student Financial Aid, 1977, 7, 26-30.

Blocker, C. E. The reality of student finances. New
Directions for Community Colleges, 1974, 8, 124-128.

The Carnegie Council on Policy Studies in Higher Education.
Next steps for the 1980's in student financial aid.
San Francisco, CA: Jossey Bass Publications, 1979.

Change Magazine and Educational Change. Colleges and money:
A faculty guide to academic economics. Washington, DC:
Author, 1976.

Chronicle of Higher Education, February 24, April 28,
December 1, 1982, February 2, 1983.

Collins, J., Maguire, J. J., & Turner, R. M. Unmet need:
How the gap is filled. Journal of Student Financial
Aid, 1979, 9, 4-15.

Crackdown begins on federal worker's student loan defaults.
Higher Education and National Affairs. December 10,
1982, p. 3.

Davis, J. S., & Van Dusen, W. D. Guide to the literature
of student financial aid. Princeton, NJ: College
Entrance Examination Board, 1978.











Deitch, K. M. Pricing and financial aid in American
higher education: Some interactions. Aspen, CO:
Aspen Institute for Humanistic Studies, 1978.
(ERIC Document Reproduction Service No. ED 184 450)

Dyl, E. A., & McGann, A. F. Discriminant analysis of student
loan application. Journal of Student Financial Aid,
1977, 7, 35-40.

Emmert, M. A. National direct student loan default rates:
A measure of administrative quality or something else?
Journal of Student Financial Aid, 1978, 8, 43-47.

Ferguson, G. A. Statistical analysis in psychology and
education (3rd ed.). New York: McGraw-Hill Book
Company, 1971.

Hartman, R. The National Bank approach to solutions. In
L. Rice (ed.), Student loans: Problems and policy
alternatives. Princeton, NJ: College Entrance
Examination Board, 1978.

Hauptman, A. Student loan defaults: Toward a better
understanding of the problem. In L. Rice (ed.),
Student loans: Problem and policy alternatives.
Princeton, NJ: College Entrance Examination Board,
1978.

Hays, W. L. Statistics (3rd ed.). New York: Holt,
Reinhart and Winston, 1981.

Horch, D. H. A retrospective description of the national
direct student loan program: Administrative practices
and institutional default rates in 1972-/3. Princeton,
NJ: Educational Testing Services, 1978. (ERIC Docu-
ment Reproduction Service No. ED 165 617)

Hyde, W. D., Jr. State guarantee agencies and capital
availability for student loans. Papers in education
finances. Paper no. 22. Denver, CO: Education
Commission of the States, Education Finance Center,
1979. (ERIC Document Reproduction Service No. ED 176 6483

Jenkins, J. A. Whose fault is student loan default? Change,
November, 1978, pp. 44-45.

Johnstone, D. B. Student loans: Some practical radical
alternatives. In Symposium on Federal Student Loans.
Washington, DC: National Association of Student
Financial Aid Administrators and the American Council
on Education, 1978.











Kelly, M., ed. Administrator. Madison, WI: Magna Publica-
tions, January 7, 1983.

Kendis, K. L. Solving the student loan problem by supple-
menting government loan policy--An expanded role for
institutional aid administrators. Student loans:
Alternatives for reauthorization. Washington, DC:
National Association of Student Financial Aid Adminis-
trators, 1978.

Leonard, K. J. Collection of federal student loans.
Columbia Journal of Law and Social Problems, 1980, 15,
317-358.

Martin, P. Student loans: Curing problems and insuring the
future. Capital University Law Review, 1980, 9, 579-
594.

Millsap, W. L., & Wright, P. S. Recent cases on student
transcript rights after bankruptcy. Journal of College
and University Law, 1979, 6, 231-240.

Myers, G., & Siera, S. Development and validation of dis-
criminant analysis models for student loan defaultees
and non-defaultees. Journal of Student Financial Aid,
1980, 10, 9-17.

National Association of Independent Colleges and Universities,
An analysis of the tuition advance fund bill. Washington
DC: Author, 1978. (ERIC Document Reproduction Service
No. ED 171 189)

New York Times, October 12, 1975, Section 3, p. 16.

Ostar, A. W. The effects of inflation and pricing policies
on college enrollments. Washington, DC: American
Association of State Colleges and Universities, 1978.
(ERIC Document Reproduction Service No. ED 177 961)

Premises of federal student assistance. Washington, DC:
American Association for Higher Education, 1979.
(ERIC Document Reproduction Service No. ED 193 999)

Rudolf, F. Myths and realities of student aid. College
Board Review, 1962, Fall, pp. 18-23.

Sanford, T. R. The impact of student loans on graduate
education. In AIR forum 1980 paper. Atlanta, GA:
Association for Institutional Research, 1980.
(ERIC Document Reproduction Service No. ED 189 936)











Silber, J. R. The tuition advance fund: A proposal for
funding higher education. In paper presented at
Assembly of the National Center for Education Manage-
ment Systems. Boston, MA: National Center for
Education Management Systems, 1978. (ERIC Document
Reproduction Service No. ED 163 824)

Study of the requirements for forming state guarantee
agencies, Final report. New York: Research and Fore-
casts, Inc., 1979. (ERIC Document Reproduction Service
No. ED 188 508)

Tate, J. L. Federal aid to postsecondary students. Journal
of Family Law, 1979, 18, 147-178.

U. S. Congress. House. Committee on Education and Labor.
Subcommittee on Postsecondary Education. Student loan
defaults. Hearings before a subcommittee of the Commit-
tee on Education and Labor, House of Representatives,
95th Cong., 1st Sess., 1977. (ERIC Document Reproduc-
tion Service No. ED 153 538)

U. S. Department of Health, Education, and Welfare. Office
of Education. Study of selected educational institu-
tions and students participating in direct school
lending under the federal student loan program, Final
report. Washington, DC: Office of Education, 1979.
(ERIC Document Reproduction Service No. ED 185 953)

U. S. General Accounting Office. Collection efforts not
keeping pace with growing number of defaulted student
loans. Washington, DC: Government Printing Office,
1977.

U. S. General Accounting Office. Stronger action needed to
recover $730 million in defaulted national direct stu-
dent loans. Washington, DC: Government Printing
Office, 1981.











BIOGRAPHICAL SKETCH


James A. Schmidt was born December 11, 1950, at Pitts-

burg, Pennsylvania. He attended local schools in St.

Petersburg, Florida, and graduated from Dixie M. Hollins

High School in 1968. He entered the Florida State Univer-

sity in 1968 and received the Bachelor of Science degree in

criminology in 1972 and the Master of Science degree in

education (student personnel/higher education) in 1973. He

served Southeastern Community College (N.C.) as Director of

Admissions and Student Activities and Researcher from 1973

until mid-1976. In the fall of 1976, he entered Stetson

University College of Law in St. Petersburg, Florida. The

following year, 1977, he accepted a counseling position with

Edison Community College in Fort Myers, Florida. During the

1981-82 academic year, he attended the University of Florida

while on leave from Edison Community College in order to

complete residency and course requirements for the Doctor

of Philosophy degree in educational administration (higher

education). Upon returning to Edison Community College in

July, 1982, he assumed his present position as Director of

Financial Aid.

James A. Schmidt is married to the former Linda Dale

Thompson of Callahan, Florida. They have two sons, Daniel

and Timothy. He is a member of Beta Theta Pi, Kappa Delta






91



Pi, Phi Delta Kappa, several professional and service organ-

izations and has been selected as an Outstanding Young Man

of America for 1983.










I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.




James L. Wattenbarge Chairman
professor of Educational
Administration and Supervision

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.




/John M. Nickens
Professor of Educational
Administration and Supervision

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.




Harold C. Riker
Professor of Counselor Education

This dissertation was submitted to the Graduate Faculty of
the Department of Educational Administration and Supervisi'!
in the College of Education and to the Graduate Council, and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.

August, 1983


Dean for Graduate Studies and
Research




Full Text
A PREDICTIVE MODEL FOR THE
REPAYMENT OF STUDENT LOANS
IN COMMUNITY COLLEGES
BY
JAMES A. SCHMIDT
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA

ACKNOWLEDGMENTS
The writer wishes to acknowledge the assistance of the
many persons who took an active interest in the preparation
of this study.
The guidance and assistance provided throughout the last
two years by Dr. James L. Wattenbarger, chairman of the
writer's supervisory committee, are deeply appreciated. In
addition, sincere thanks are also extended to the faculty of
the Department of Educational Administration and Supervision
and in particular to the members of the committee, Dr. John
M. Nickens and Dr. Harold C. Riker.
For their assistance in collecting the data the writer
is indebted to the Florida Student Financial Aid Commission.
The technical assistance provided by Mr. Kenneth Goehle and
Mrs. Betty Davis are also gratefully appreciated.
Finally, the writer wishes to acknowledge the support
provided by his family. For their encouragement, assist¬
ance and emphasis on the value of an education, the writer
is deeply indebted to his parents. Dr. and Mrs. Paul H.
Schmidt. To his wife, Linda, and his sons, Daniel and
Timothy, the writer wishes to express his deepest gratitude
for their love, patience and understanding during the prepa¬
ration of this study.
11

TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS Ü
LIST OF TABLES V
ABSTRACT vi
CHAPTER
I.INTRODUCTION 1
The Role of Financial Aid 2
The Growth and Importance of Loans 10
The Problem of This Study 19
Justification for the Study 20
Delimitations & Limitations 21
Research Design and Procedures 22
II.REVIEW OF RELATED LITERATURE 26
Problems Inherent in Student Loan
Programs 31
Institutional Responses to
Student Loan Defaults 38
Professional Organization Action
Regarding Student Loan Defaults 40
Governmental Action Regarding
Student Loan Defaults 42
Responses to Student Loan Default
Problem by Education Spokesmen 44
Antecedent Studies Relating to Student
Loans and Demographic Characteristics .... 48
Summary 53
III.DEVELOPMENT OF THE MODEL 55
Analysis of Variables 60
Development of the Model 69
Results 73
Summary 77
iii

Page
IV. CONCLUSIONS, IMPLICATIONS AND
RECOMMENDATIONS 7 9
Conclusions 79
Implications 80
Recommendations 83
BIBLIOGRAPHY 86
BIOGRAPHICAL SKETCH 90
IV

LIST OF TABLES
Table Page
1. Philanthropical Support for Higher
Education 27
2. Overview of Major Variables Selected
by Previous Researchers 56
3. Chi-Square Analysis of Relationship
Between Age and Loan Payback 62
4. Chi-Square Analysis of Relationship
Between Marital Status and Loan
Payback 63
5. Chi-Square Analysis of Relationship
Between Sex and Loan Payback 6 3
6. Chi-Square Analysis of Relationship
Between College Standing and
Loan Payback 64
7. Chi-Square Analysis of Relationship
Between Grade Point Average and
Loan Payback 65
8. Chi-Square Analysis of Relationship
Between Loan Total and Loan
Payback 66
9. Interaction Effects Between the
Variables 68
v

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfilment of
the Requirements for the Degree of Doctor of Philosophy
A PREDICTIVE MODEL FOR THE
REPAYMENT OF STUDENT LOANS
IN COMMUNITY COLLEGES
By
James A. Schmidt
August, 1983
Chairman: James L. Wattenbarger
Major Department: Educational Administration and Supervision
The problem of this study was to determine the relation¬
ship between selected variables which characterize community
college students and student loan defaults and to develop a
model using these variables to predict student loan payback.
Given the current economic crisis and the increasing reliance
on the student loan programs to help students meet educational
expenses, a study of the importance of selected student demo¬
graphic characteristics and their relationships to the student
loan default problem is of great importance to the future
support of the student loan programs.
The literature provided a theoretical basis for this
study including appropriate variables for study as predictors
of student default. These variables included size of loan
total, marital status, sex, grade point average, college
standing, and age.
The data presented in this study were supplied by the
Florida Student Financial Aid Commission, Tallahassee,
vi

Florida, and represented a statewide sample of 76 community
college students who have participated in the Guaranteed
Student Loan program.
Of the six variables selected, only the size of the loan
total and marital status distinguished significantly those
who repaid their student loans from those who did not. In
addition to these variables, sex, grade point average,
college standing, and age were useful in developing a pre¬
diction model. Although the model did not provide an in¬
fallible formula for predicting those students who are most
likely to repay their student loans, the model predicted
group membership (defaulter of non-defaulter) for 70% of the
sample cases. These findings underscore Pattillo and
Wiant's conclusion that items reflecting financial rather
than biographical data appear to be better predictors of
loan delinquency.
Therefore, it appears that the inclusion of additional
discriminating variables and a more detailed study design
may be necessary in order to improve the identification of
students who are likely to repay their student loans.
vil

CHAPTER I
INTRODUCTION
From modest beginnings over three hundred years ago the
role and impact of financial assistance to students in
American colleges have changed dramatically. Like the gift
presented by Lady Anne Mowlsen of London to the deserving
Harvard student, most early student financial aid awards
were granted with money given to colleges by private individ¬
uals to aid worthy, needy students. Therefore, the original
purpose of providing student financial aid was to expand
educational opportunities to those students whose incomes
presented a barrier to higher education.
In fact, only during the period of time immediately
following World War II was substantial aid awarded on the
basis of merit alone. From the late 1940s through the early
1960s the Serviceman's Readjustment Act of 1944 (GI Bill)
provided substantial amounts of money to returning service¬
men for educational expenses regardless of individual finan¬
cial need. Many colleges and universities found that the
large amounts of scholarship funds, previously used to sup¬
port needy students, were no longer needed and those insti¬
tutions began to use the funds to attract and reward students
1

with academic or other special talents, with little or no
regard for financial need.
2
Currently, the essential criterion of the major federal
aid programs (Pell Grant, College Work Study, National
Direct/Guaranteed Student Loan Programs) is the students'
need for funds. Accordingly, the purpose of the principal
financial aid programs of today is remarkably similar to the
original intention of the early student aid programs. The
return to a need-based student aid system has caused a great
deal of consternation to financial aid administrators. Today
it becomes increasingly difficult to reward students with
merit-based aid when the available resources do not ade¬
quately take care of the needy students. In order to encour¬
age both needy and worthy students, a balance between need-
based and merit-based aid is critical to the development of
an impartial and equitable student financial aid delivery
system. However, most student financial aid administrators
apparently feel that the current, primarily need-based,
delivery system provides the best use of available resources
for it attempts to extend the opportunity for higher education
to as many worthy students as possible.
The Role of Financial Aid
Since 1972 there has been considerable discussion con¬
cerning the role of financial aid in American higher educa¬
tion. In nearly every deliberation much attention has been

3
devoted to the related issues of access and tuition charges
because they are so intertwined with the role of financial
aid.
As a nation, the United States has demonstrated its
commitment to citizen access to postsecondary educational
opportunities by instituting programs of assistance to needy
students, establishing support programs for colleges and
universities (i.e., Land Grant College Program) and develop¬
ing a strong system of public colleges and universities.
According to Ostar (1978), 80% of all students attend¬
ing higher educational institutions today are in programs
at public institutions where the state partially subsidizes
the institutional costs of the student. As a part of the
public higher educational system, the public community
college is an important factor in providing higher educa¬
tional opportunities to American students. The impact of
the community college system on the issue of access is high¬
lighted in the state of Florida where more than 95% of the
state population live within commuting distance of a commun¬
ity college program.
Although the costs of attending a community college
program may be lower because of public subsidies, the indi¬
viduals attending the community college seem to have the
greatest need for financial assistance. A 1972 American
College Testing (ACT) Program Study indicated that the
financial needs of community college students sharply

4
exceeded those of the majority of students enrolled in other
colleges and universities. The ACT study estimated that
financially disadvantaged students constituted 25% of the
total student population at community colleges and cited the
following factors as contributing to this high percentage:
(1) 70% of the students depend on the
automobile to reach campus,
(2) the higher age distribution (average
age of student is 29) indicates the
average student has assumed adult
responsibilities, and
(3) 66% of the students come from families
with (1974) incomes below $10,000.
(Blocker, 1974, p. 126)
Therefore, it would also appear that the retention of
low tuition policies is critical to maintain access to
higher educational opportunities for a large number of stu¬
dents in the community college system. In fact, a Univer¬
sity of Wisconsin study highlights the impact tuition
policies have on the issue of access (Ostar, 1978). This
study determined that for every 1% reduction in total cost
of attending a University of Wisconsin Center participating
in the study there was a corresponding 1.33% increase in en¬
rollment. The study also attributed the increased enroll¬
ment to new students who were not enrolled in the University
of Wisconsin system and concluded that many of these new
students would not have been able to attend were it not for
the lower tuition policy (Ostar, 1978) .

Despite the establishment of the relationship between
low tuition and access by this and other studies, much de¬
bate regarding federal aid centers on the strategy of pro¬
5
viding individual student aid rather than institutional aid.
It is feared that the reliance on the student-based aid
delivery system may cause additional pressure on state govern¬
ments and lead both public and private institutions to raise
tuition fees (Ostar, 1978). Although the underlying assump¬
tion of raising tuition fees is that the truly needy student
will be taken care of, a University of Tennessee study in¬
dicated that a $100 increase in tuition would yield only a
$26 increase in an individual student's Pell Grant Award
(Ostar, 1978) . Truthfully, every tuition increase falls
hardest on the $13,000 - $25,000 income family and yet this
student is eligible for little federal aid in the form of
grants or College Work Study (Ostar, 1978) . In addition,
Ostar maintains that an overemphasis on student-based aid
leads to a growing number of charges involving fraud, waste,
and abuse and leads to ever-tightening regulations and in¬
creased numbers of bureaucrats, investigators and bill
collectors.
The previous discussion regarding access would indicate
that the best approach regarding federal assistance may be
the increasing of federal support to higher educational in¬
stitutions enabling colleges and universities to retain
lower tuition charges. However, because of economic and

6
political factors the financial aid delivery system is
oriented toward individual student awards. Given this under¬
standing two major questions must be addressed: what aid is
available to students and how effective is the delivery
system?
Student Aid Resources
Essentially student financial aid is available in the
form of grants, work-study opportunities and loans. A grant
is gift money that is offered to those most needy in order
to prevent over-indebtedness through borrowing. Work-study
is a self-help program which provides meaningful, college-
related job experiences and earnings enabling the student to
earn money and meet current expenses from bi-weekly wages.
Loans constitute a program of borrowing whereby students
must repay the loan with interest from future earnings after
leaving college. The federal programs representing these
categories of aid include Pell (BEOG) and Supplemental Educa¬
tional Opportunity Grants, College Work-Study, and National
Direct and Guaranteed Student Loan Programs.
Most of the federal programs have not seen substantial
increases in federal money in recent years. However, the
Pell Grant Program (BEOG) which was traditionally limited to
low income students has been expanded to assist families
from middle income levels. In addition, the amount of money
expended and the numbers of students served through the

7
Guaranteed Student Loan Program (GSL) have risen dramati¬
cally.
During the 1982-83 fiscal year modest cuts were made
in the financing of the campus-based programs (SEOG, CWSP
and NDSL). Although support for the Pell Grant Program was
retained at previous levels, eligibility and dollar delivery
regulations were instituted to reduce fraud, abuse and
errors. The combination of reduced support for some pro¬
grams and increased regulation of the Pell Grant Program
created a dilemma for student financial aid administrators
who could stretch local and institutional funds only so far.
At the same time, colleges and universities are exper¬
iencing higher costs because of increases in major dollar
operating items (i.e., salaries, energy, supplies) which
seem to be severely affected by an inflationary economy.
This increase in costs translates into higher tuition
charges. Since the financial aid application requires the
student's resources to be balanced with the expected ex¬
penses, many students are experiencing higher need figures
as the expected expenses rise. Without corresponding in¬
creases in student aid to offset the rise in expenses, a
"need gap" crisis occurs for the student. Although existing
support for assistance programs is generally at the highest
level ever recorded, the increasing cost factor is rising
faster than increased federal support and therefore the
number of "need gap" victims continues to grow.

8
The Student Aid Delivery System
Most of the blame for the "need gap" crisis must be put
on increasing costs and expenses since the majority of the
financial aid researchers conclude that the student aid
delivery system appears to be quite effective (Deitch, 1978).
The financial aid delivery system in the United States
is the mechanism used to provide individual students with the
resources to attend postsecondary schools. Application for
the Federal Pell Grant Program may be made separately on a
free federal application; or, for a nominal charge, an in-
depth assessment of need may be presented along with the
eligibility report for the Pell Grant Program. Although
there has been some recent discussion of the effectiveness
and equity of the need-analysis system, previous studies
indicate the system is in pretty good shape. According to
a study presented at the Annual Meeting of the American
Association for Higher Education in March of 1979 the
following conclusions were presented:
(1) The need based system of financial
aid has gone far in removing a mean¬
ingful element of price competition
from the buying and selling of higher
education,
(2) the structure of the aid system (mix
of parental contribution, need-based
grant and self help) is meeting the
public's perception of how post¬
secondary education should be financed,
(3) the major goal of providing access is
quite clear, and correspondingly the
programs have been successfully reach¬
ing toward the goal (i.e., the number
of minorities has increased signifi¬
cantly) ,

9
(4) federal, state and institutional
funds seem sufficient to insure
that any qualified student could
afford to attend most of the
nation's institutions, and
(5) the problem is not with the dif¬
ferent kinds of aid available but
for more efficient and equitable
distribution of the funds that are
available. ("Premises of Federal,"
1979)
According to Kenneth Dietch (1978) , the current system of
student financial aid is a thoroughly developed, pervasive
and well-functioning system of price discrimination. How¬
ever, he cites several problems which are likely to arise
with the current system including
(1) competitive pricing (by institutions),
(2) lack of sensitivity to student costs,
(3) loss of financial privacy of families
of applicants,
(4) efforts to depart from the need-based
system to attract (non-needy, out¬
standing) students, and
(5) need for fairness in the treatment of
independent students. (Dietch, 1978)
Regardless of the potential problems with the financial
aid delivery system, its current status seems to be that it
is well respected among professionals and researchers.
Financial aid administrators and federal program personnel
have struggled to solve the problems of "need-gap" by work¬
ing together to tighten requirements and enforce regula¬
tions through extensive validation procedures.

10
The Growth and Importance of Loans
As other political priorities become apparent in Wash¬
ington and free student aid programs (grants) are retained at
previous funding levels, most aid administrators have increas¬
ingly used the available loan programs to meet student aid
needs. In fact, loan programs have been increased substan¬
tially while most of the other aid programs have been limited
or decreased in size. Therefore, the importance of an
equitable loan system has far greater implications today than
it did previously as more and more students use the device to
solve the "need gap" crisis.
General Definition and Philosophy
of Student Loan Programs
As indicated previously, loans are awards of money
granted with the prior requirement that they subsequently
be repaid, in whole or in part, with or without interest.
According to Davis and Van Dusen (1978), loans are the most
attractive form of student aid to most policy-makers because
they must be repaid and therefore are an investment rather
than a direct expenditure of funds. Recent discussion of
national student aid policies indicates that those in
decision-making roles believe loan programs to be very
effective. Therefore, the great increase in the use of the
current loan programs and the addition of new loan programs
are quite understandable.

11
The underlying philosophical premise of a loan program
is that the person who derives the benefit from an education
should pay for its costs, without denying low income stu¬
dents access to postsecondary education due to lack of
current resources (Davis & Van Dusen, 1978). Not only has
such a philosophy been utilized to develop loan programs
but it has also led to arguments concerning how much of the
cost of an education should be reflected in the tuition
charge. For example, in 1973 the Committee for Economic
Development recommended that tuition costs should be raised
to 50% of the total educational cost within five years at
colleges and universities and within 10 years at community
colleges. Subsequently the issue was studied in detail by
the Carnegie Commission. It suggested that tuition should
be raised to 30% of the educational cost within 10 years at
colleges and universities, but the Carnegie Commission
exempted community colleges due to the historical concept
of two years of free access to higher education. Many
states have cited the Carnegie Commission report in intro¬
ducing data supporting the raising of fees at low-cost
public institutions.
However, Howard Bowen (cited in Balderston, 1970) effec¬
tively argues that an investment in higher education is
really an investment in the society. It is an investment
in the nation's human capital leading to increased indi¬
vidual productivity and societal benefits (i.e., a more
informed electorate, increased national productivity, lower

12
unemployment rates, a populace more adaptable to technology,
increased social consciousness). Mr. Bowen concludes that
typical students already bear 75 to 80% of the cost of their
education in the form of tuition, books and foregone earn¬
ings. Many studies conclude that the heavy reliance on loan
financing has the effect of shifting the eventual burden of
payment for the cost of higher education toward students and
their future income (Balderston, 1970). According to Mr.
Bowen, the increased availability of loans is not an appro¬
priate solution to the need gap crisis since students al¬
ready bear such a large portion of the cost of their educa¬
tion. Mr. Bowen argues that since society receives signifi¬
cant benefits from educating students, it should bear an
increasing portion of the cost of educating them.
Those who support the argument that students should
bear the cost of their education also favor the increasing
use of loans to help students finance the increasing costs.
However, the desirability of increasing the burden of cost
on students by raising fees and then assisting the needy
students through loan programs is an issue that will remain
unresolved as both pro and con advocates have many arguments
of merit. However, the current federal administration
favors the use of loans to help needy students bear the in¬
creasing costs of higher education and therefore it has
instituted new loan programs to help relieve the "need gap"
crisis.

13
Types of Loan Programs
Among the original loan programs are the National
Direct Student Loan (NDSL) Program (formerly the National
Defense Student Loan Program) and the Guaranteed Student
Loan (GSL) Program. The NDSL is a low-cost deferred repay¬
ment loan administered by the educational institution and
funded 90% with federal monies and 10% with institutional
monies. Under the NDSL Program there are no security or
endorsement requirements and its distinctive feature is that
in addition to a deferment of three years for public service,
the student may arrange for cancellation of the loan based
on military service in an area of hostility, full-time
teaching service in selected schools (for five years) or
full-time staff service in a head start program (for seven
years).
The Guaranteed Student Loan Program is available in
nearly all states where a state level guarantee agency has
been developed. In the remaining states students may apply
for a Federally Insured Student Loan (FISL). Administra¬
tively, private parties and non-federal agencies loan their
own capital to students in both programs. In the GSL pro¬
gram there is a central monitoring agency. While the
federal government insures the loan and helps pay the in¬
terest, the agency acts as intermediary, overseer and
collections processor. However, in the FISL program the

14
lenders rely directly on the U. S. Department of Education
for an interest supplement, insurance against default and a
collection mechanism. Repayment regulations for both the
GSL and FISL programs are similar to the NDSL; however,
deferments of up to three years are only available for ser¬
vice in the armed forces, Action or the Peace Corps, and
there are no provisions for cancellation.
The Increasing Use of Loans
The growth of both of these programs is indicative of the
current emphasis on the issuance of loans to solve the student
aid crisis. In 1977, Ernest L. Boyer (cited in Hauptman, 1978),
then Commissioner of Education, estimated that between 1965 and
1977 six million students borrowed $11.1 billion. Commissioner
Boyer stated that in 1975-76 alone $2.6 billion were loaned
to students and that the national increase rate in the num¬
ber of loans processed was averaging 5% each year. Mr.
Boyer concluded that thousands of students have been served
by the loan programs including many students who otherwise
may have been denied a postsecondary education. In addition,
Sanford (1980), in his research of the impact of loans on
graduate education, found that graduates with loans are
slightly more likely to attend graduate or professional
school than those without loans. Therefore, it appears that
there is substantial value in supplying loans beyond merely
providing access to undergraduate degree programs. The most

15
recent data indicate a total of 2.75 million loans were pro¬
cessed through the GSL system alone during the 1981-82
academic year representing an investment of $6.1 billion
(The Chronicle of Higher Education, 1983) .
Accordingly, the success of the federal loan programs
seems to be the impetus for the development of new programs.
For example, the Federal Government announced the creation
of the PLUS/ALAS Loan program in 1980. This program is
similar in administrative details to the other federal loan
programs. The major difference between these programs and
the original loan programs appears to be the lack of defer¬
ment and cancellation provisions and higher interest rates.
It also should be noted that, in addition to the federal
loan programs, many institutions, states and private founda¬
tions provide loan funds to students. Many of these programs
pre-date the federal efforts, but the importance and reliance
on loan programs are a result of the widespread application
of the loan principles effected through the federal programs.
Of the federal loan programs the GSL/FISL program is the
most widespread. The reason is that the NDSL is institution¬
ally administered and many institutions choose not to parti¬
cipate in the program. Other institutions have been re¬
stricted from participating in the NDSL program due to their
high student default rates. Therefore, due to the universal¬
ity of the GSL/FISL program, much of the data and research
presented in the remainder of this study will focus on the
GSL/FISL program.

16
Program Problems and the Default Rate
The federal loan programs have several disadvantages.
According to Tate (1979) , abuses in loan programs occur at
each of the three stages through which any student loan
passes: (1) eligibility; (2) disbursement; and (3) collec¬
tion. Both institutions and students are responsible for
abuses apparent in the loan programs of today. Institu¬
tional abuses range from mismanagement to halfhearted
collection efforts and student abuses include outright fraud
in supplying eligibility data to shrewd bankruptcy declara¬
tions in order to discharge outstanding debts. The most
obvious abuse regarding the loan system is the default or
lack, of repayment by student borrowers.
By statute a default is defined as an overdue loan pay¬
ment of 120 days (20 U.S.C. & 1080 (c), 1976). However, the
Department of Education does not recognize a default until
the lender's obligation is suspended and the subrogation by
the Federal Court takes effect.
Regardless of the procedural issue concerning the
commencement of the default status, the American public is
constantly reminded of the thousands of students who have
participated in the federal loan programs and who have de¬
cided not to repay their loan obligations. Headlines in
papers consistently highlight the default rate and recent
efforts to initiate novel collection methods such as im¬
pounding automobiles have received national television
attention.

17
The "default rate" is computed as a ratio of the cumu¬
lative amount of delinquent funds to total receivables and
can be measured by the following formula:
cumulative claims paid to lenders
all loans having entered repayment status.
Current researchers estimate the overall default rate to be
around 12%; however, the rate is slightly lower in the GSL
program because the financial institutions exercise more
selective lending practices and the guarantee agencies ex¬
ercise more administrative vigilance.
There is much debate concerning the reliability of data
regarding the default rate. Arthur Hauptman (1978) argues
that loan default rates are not consistent due to several
factors:
(1) the measurement standards changed
and the data prior to 1973 are not
consistent with the data collected
after 1974;
(2) the default rate is conceptually
flawed since it (a) uses only the
federal expenditures in the numer¬
ator (federal expenditures are only
80% of the costs) and (b) includes
death and disability defaults; and
(3) the default rate is cumulative;
therefore, data for each year's
statistics are not sorted out
(not everyone defaults in the
first year of repayment). (Hauptman,
1978, p. 160)
Hauptman concludes that a single measure of default is in¬
adequate to explain levels and trends in the behavior of

18
students who default because default activity differs by the
type of guarantee (FISL or GSL), the type of educational
institution the borrower attends, the type of lender and
the demographic characteristics of the borrowers.
In addition to the internal problem concerning the re¬
liability of the default rate statistics, there is much con¬
cern about the use of the default rate as a performance
indicator regarding the success of the student loan programs.
As cited by Hauptman, an Office of Education Report (1979)
illustrates the three major weaknesses of this argument:
(1) too many variables beyond the con¬
trol of the Guarantee Agency have a
significant impact on default rates
(i.e., depression/recession, high
unemployment rates, etc.);
(2) loan default rates may only reflect
the availability of loans (higher
rate may be due to significantly
more loans arranged); and
(3) cumulative default rates are his¬
torical and may not reflect current
management philosophies. (Hauptman,
1978, p. 162)
In effect, two conclusions can be reached based on the
previous discussion:
(1) the tendency to default may have
increased over the last several
years but it may not be as high
as it has been reported, and
(2) a large number of variables which
are external to the guarantee
agency and the financial lender
affect the default rate.

19
The Problem of This Study
Regardless of the elusiveness of concrete reliable
statistics, the public perceives the default rate as too
high. The federal government's determination to stem the
default rate must include research on the socio-economic
characteristics of student borrowers and defaulters in
addition to data concerning administrative procedures.
Accordingly this study has attempted to provide data con¬
cerning some of the variables that may be related to the
student default rate.
In addition, these variables may be useful in formulat¬
ing a prediction model to determine who will repay their
student loans. Such a model would be of valuable assistance
to the student aid administrator in reducing the default
rate because it would distinguish the characteristics of
defaultees and non-defaultees. The aid administrator could
then award and approve loans for only those students with
the greatest likelihood of repayment. Since loan money may
not be as readily available in the future as it is today, a
model that could provide assistance in selecting the "best"
student loan risk may be critical in maintaining the integ¬
rity of the loan programs and in replenishing the source of
funds for future loans.

20
Statement of the Problem
The problem of this study was to determine the relative
influence of selected variables which characterize community
college students who are most likely to repay student loans.
Questions
Specifically, a primary question of this study was to
determine the usefulness of the model derived from the data
presented in this study to predict those students likely to
repay student loans from those students who are not likely
to repay student loans.
Therefore, the following supplemental question was
addressed: will the selected variables be useful in discrim¬
inating between students who will repay student loans and
students who will not repay student loans?
Justification for the Study
As indicated previously, throughout the last few years
concern regarding student loan default rates has risen tre¬
mendously. The issue has become the source of critical edu¬
cational fiscal management decisions and delicate political
debates. During the period from September, 1982, through
February, 1983, The Chronicle of Higher Education reported
on 17 articles relating to recent developments concerning
student loans. Thus, the selection of this issue is

21
justified on its standing as a practical problem facing
higher education administrators today. Discussion relative
to some of the many student variables contributing to the
default rate is imperative. This study analyzes six vari¬
ables (age, marital status, sex, college standing, grade
point average and loan total) identified in the previous
research of Dyl and McGann (1977), Myers and Siera (1980),
Emmert (1978) and other researchers for the impact these
variables have on the student default rate.
Delimitations and Limitations
Since there are several procedural issues impacting the
results of this study, the following delimitations and limi¬
tations are applicable.
Delimitations
(1) This study was confined to students
who have received a loan through a
program administered by a Florida
public community college.
(2) Only data made available to the
researcher by the Student Financial
Assistance Commission of the State
of Florida were utilized in this study.
(3) A sample approximating 31% of the
total population was selected by a
systematic random process for both
categories of students identified
in the analysis (i.e., students who
repay student loans and students
who do not repay student loans).
(4) The analysis of data was confined to
the discriminant analysis procedure
utilized by the computer

22
techniques of the Statistical
Analysis System (SAS), which pro¬
duced a set of equations from
which a predictor model could be
evolved.
Limitations
(1) This study was limited to those
variables identified in the re¬
search design, specifically:
college standing, age, marital
status, sex, grade point average,
and loan total.
Research Design and Procedures
This study was designed as an ex post facto research
study in which several continuous and discrete independent
variables obtained from data collected on loan application
forms were observed. From these variables, the dependent
variable of loan repayment was predicted.
According to Ary, Jacobs and Razavich (1979), ex post
facto research can supply much information of value in edu¬
cational decision making even though its design differs sig¬
nificantly from pure experimental research. In an ex post
facto research design an experimenter can study two groups
that are different in some respect in order to discover the
reasons for the difference. Such is the case with this
study. The design format as suggested by Ary et al. is as
follows:

23
Group
Independent Variable(s) Dependent Variable
E
(X) (variable)
(Repaid Loan)
Y2 (Did Not Repay
C
Loan)
The supposed effect of treatment occurs in the experimental
group. Therefore, according to Ary et al. (1979), the re¬
searcher attempts to relate the dependent variable Y^ to a
previously occurring independent variable, a nonmanipulable
variable indicated by (X) which occurred in the experimental
group but not in the control group.
The purpose of utilizing the ex post facto research
design was to illustrate the significant data to be util¬
ized in a model that could be useful in predicting the re¬
payment rate probability of students requesting loans for
educational use at community colleges. Such a model would
be helpful to financial aid personnel in determining what
student characteristics are representative of those students
who have repaid their student loans.
Sample
The data for this study were selected randomly from
the total population of students in each of the following
two categories:
(1) students who are repyaing their
student loans and/or are not

24
identified as defaulting on
their student loan commitment,
and
(2) students who are not repaying
their student loans and have
been identified as defaulting
on their student loan commit¬
ment.
These categories were determined based on the data
collection and storage procedures established by the Florida
Student Financial Assistance Commission (FSFAC) in Talla¬
hassee, Florida. The FSFAC is the central storage/retrieval
center regarding GSL repayment data for all institutions in
Florida. The sample size (76) represents approximately 31%
of the total population in each category involving loans made
to community college students in Florida.
Data Collection
Data on each student included in this study were
supplied by the FSFAC to the researcher in June, 1982.
During this session files were randomly selected by FSFAC
personnel for review by the researcher.
The appropriate data were then extracted from the files
for those students who were identified as community college
students. Since there are two separate files maintained by
the FSFAC (i.e., those who have not defaulted, those who
have defaulted) the researcher obtained a sample for each
category representing 31% of the total population.

25
The sample size (38 in each category) meets the re¬
quired minimum (30) sample size suggested by Ary et al.
(1979) and the 31% sample size exceeds the 10 to 20% sample
size suggested by Art et al. for descriptive research.
The following variables were extracted from the data
files based on their availability and their identification
in previous research studies by Dyl and McGann (1977), Myers
and Siera (1980), and Emmert (1978):
(1) Age,
(2) Marital Status,
(3) Sex,
(4) College Standing,
(5) Grade Point Average, and
(6) Loan Total.
A full discussion of the selection of these variables and
the results of previous studies utilizing these variables
appears at the beginning of Chapter III.

CHAPTER II
REVIEW OF RELATED LITERATURE
The financial pressures exerted on colleges and univer¬
sities have had a direct effect on the expansion of loan pro¬
grams. As a result the loan programs are experiencing some
unique problems. An analysis of the responses to the prob¬
lems surfacing in the student loan programs and a review of
the research regarding student loan default and student
demographics provide an overview of the issues surrounding
the problem presented in this study.
Fiscal Pressures and the Demand
For Expanded Student Loan Programs
Higher education is a social institution that depends
on tuition and service charges, philanthropy, and govern¬
mental support in order to balance its budget. The heavy
reliance on these sources of support has created a dilemma
concerning the funding of higher education during weak eco¬
nomic times. The current fiscal crisis can be readily con¬
firmed by reviewing the funding sources and relative support
for higher education since 1950.
In terms of tuition and service charges, there has been
a percentage decline in these fees as a source of income
26

27
from 1950 to 1974 (Change Magazine, 1976). This decline
has been attributed to the larger percentage of students
attending public colleges in 1974 (76%) compared to 1950
(50%) and the corresponding lower tuition charges at these
public institutions (Change Magazine, 1976). In addition,
colleges and universities have found themselves in compe¬
tition with local facilities for the student dollar regarding
auxiliary charges including food, housing and books. There¬
fore, there has been a decline in income from this revenue
source as well.
Although all service charges such as room and board
accounted for approximately 40% of the total income in
colleges and universities in 1974, Kelly (1983) estimated
that tuition charges in 1977 accounted for only 11% of the
total income for colleges and universities. As a result of
present trends on college campuses, Kelly (1983) estimates
that this percentage will increase to 20% by 1988. Based
on this estimate it appears that students will be forced
to bear an increased percentage of the cost of their educa¬
tion in the form of increased tuition charges. Correspond¬
ingly there is likely to be additional pressure on the
student aid delivery system to compensate for this in¬
creased cost. However, many colleges and universities are
fearful of an increased dependency on student fees. These
colleges and universities are justifiably concerned that
the resultant increased fees will add a new pressure to

28
the university budget—declining enrollments and the loss
of dollar income from student charges based on underenroll¬
ment.
In terms of philanthropical support, higher education
has been the beneficiary of increased dollar support from
1950 ($200 million) until 1974 ($2.4 billion) (Change Maga¬
zine, 1976). As indicated in Table 1, due to the tremen¬
dous rise in the cost of higher education, this increased
dollar support represents an actual percentage decline from
1950 (8%) to 1974 (5%) in terms of total income for colleges
and universities (Change Magazine, 1976).
Table 1
Philanthropical Support
for Higher Education
Year
Total Dollar
Support
Percent
Colleges
of Income for
& Universities
1950
$200 million
8%
1974
$2.4 billion
5%
Since the dollar increase is not consistent with the needed
percentage increase for institutional philanthropical sup¬
port, more of the cost of higher education will have to be
absorbed elsewhere.
In the recent past, the government (federal, state,
and local) has provided this increased support for higher

29
education. In 1974 it was estimated that the government pro¬
vided 54% of the total income to colleges and universities
(Change Magazine, 1976). This increase in total income
represents additional support for both private and public
institutions as it includes research monies, building
assistance, and student aid dollars. However today, govern¬
ments whose support figures so largely in the financing of
higher education are concentrating on other major concerns
and priorities. Therefore, there has been a decline in the
rate of income growth to higher education from the govern¬
ments. For example, the federal government is under con¬
trary pressures to hold down expenditures in student aid
(which has risen 600% from 1967 to 1977) on the one hand and
to respond to the continuing demands of middle income par¬
ents for selective relief (i.e., additional aid, tuition tax
credits) on the other. In addition, many concerns outside of
higher education are uppermost in the minds of our legisla¬
tors. For example, state legislators are highly conscious
of the "Proposition Thirteen" sentiments in their districts.
Therefore, the new depression in higher education is
generally not based upon declining support, but it is due in¬
stead to a decrease in the rate of increase of support for
higher education. Nowhere is this more evident than in the
dollar support for student aid during the last two fiscal
years. Colleges and universities have received essentially
the same amount of support in total dollars from the

30
government, but as the rate of increase has dropped, increas¬
ed pressure to raise fees and start fund raising activities
has surfaced. Because the slowdown has not been uniform,
some states and individual institutions have had to contend
with declining income.
As a solution to the pressure added by increased fees,
the Carnegie Council on Policy Studies in Higher Education
(1979) has recommended expanding explicit self-help compon¬
ents (work-study and loan opportunities) of the student aid
delivery system and revising the loan system to make loans
more easily attained and widely available with less risk of
default (i.e., longer payback terms, loan consolidation
procedures).
It appears that the Reagan administration has seriously
considered the Carnegie recommendations since they have pro¬
posed a 60% increase in College Work-Study funds, a "self-
help" requirement for Pell Grant eligibility and changes in
the Needs Test for the Guaranteed Student Loan Program for
fiscal 1984. In addition, Mr. Reagan has proposed continuing
dollar support for the federal grant programs at the same
level as the past two fiscal years. However, under his pro¬
posal, eligibility qualifications for individual students have
been substantially altered. Therefore, it is safe to assume
that as fees increase, endowment expansion decreases and fed¬
eral grant money remains the same, an increasing number of
students will seek to participate in the student loan program.

31
A potential problem arising from the decision to rely
on the student loan programs for the increased dollar support
to higher education is that the private lenders simply may
not have funds they want to lend to students (as other loans
may be more financially advantageous) or there may not be
enough loan money to meet recognized need. Further, the
loan programs in effect today have many problems. Therefore,
before considering an expanding role for student loans in
the financing of higher education, it appears the government
should consider the weaknesses of the current loan system.
Problems Inherent in Student Loan Programs
A closer examination of the student loan default prob¬
lem reveals several contributing factors. Therefore, nearly
all researchers agree that the student should shoulder only
a part of the blame for the escalating default ratio.
Foremost among the factors to consider is the history
of incompetent management of loan programs. For years nei¬
ther the lender nor the government was making any noticeable
effort to collect many of the outstanding loans (Jenkins,
1978). In addition, Jenkins noted that the Office of Educa¬
tion could not afford to install a computerized collection
system until 1978 (Jenkins, 1978). Although governmental
investments in the area of administrative strategies have
resulted in much improvement in collection procedures,
the previous administrative procedures account for a

32
substantial part of the default problem attributed to the
historically impacted default rate (Jenkins, 1978) .
A second factor to consider is the nature of student
loans. Kendis (1978) states that even though student loans
are not the same as consumer loans, the lack of rules con¬
cerning asset and income analysis, borrowing limits and
collection procedures are "diametrically opposed" to method¬
ologies long since proved successful in consumer credit
(Kendis, 1978). Although Kendis feels that the rules should
be somewhat different from the rules utilized in consumer
credit situations, he states that the result is an "attitude
problem" surrounding student loans. This "attitude problem"
affects not only students but lenders and legislators as well.
Students borrowing as a part of their investment in their
own human capital are forced into a decision process to attend
or not to attend a college or university, wherein the negative
consequences of increasing costs, and debt, are minimized by
the ability to borrow funds. Lenders continue to lend greater
volumes of funds leading to a greater median debt loan (with
increasing financial concern) in the interest of assuring acces:
and individual collegiate choice. And finally, the Government
in its effort to assume capital availability guarantees loans
by merely increasing existing programs while disregarding the
long-term costs and the social consequences of the program.
Any one of the conditions highlighted previously would have

33
significant impact on the student loan problem; however, in
combination there is no doubt that the "attitude problem"
suggested by Klendis is a very real factor affecting student
loan defaults.
A third major factor affecting the student loan problem
is the mechanism for repayment. Balderston (1970) notes in
an analysis of incomes between those students choosing to go
to college and those choosing not to go to college that the
former will not "catch up to" the latter in real dollars
realized until very late in the working life of the individ¬
ual primarily due to the foregone earnings (lack of income
during college years) of the college student (Balderston,
1970) . In fact, in some instances the college-going student
may never catch up to the non-college student whether the
dollars are discounted for inflation or not. Therefore,
any student researching the wisdom of the investment of a
college education from a purely financial perspective may
decide it is not worthwhile to attend college based on the
market payoff (Balderston, 1970).
Beyond the decision the student must make regarding the
wisdom of attending college, if a loan is utilized, then the
repayment mechanism that currently operates which seeks to
concentrate the repayment in the early years of one's work¬
ing life (10 years) may have some deterrent effect upon
college attendance. If loan financing is heavily relied
upon, the effect may be substantial. Balterston's

34
conclusion is that short repayment periods lead to fewer
students opting to attend college if loans must be utilized
because they impose too heavy a burden of cash outflow on
the student. He also notes the effect is stronger on stu¬
dents who do not finish programs or finish programs with
less optimistic forecasts of projected future incomes (i.e.,
humanities, home economics) because these students do not
have the advantage of cashing in on a completed or market¬
able degree and the loan payment constitutes a higher per¬
centage of their total income than it does for students with
higher incomes.
The solution proposed by Balderston is to make long
term loans and to tie the repayment mechanism into the in¬
come tax system or set up a similar system which reduces
administrative costs (Balderston, 1970). Although there is
much debate as to the wisdom of Balderston's proposal,
especially as it relates to the use of the income tax system,
there appears to be little doubt that the repayment system
for current loans is a component of the overall problem
regarding loan defaults.
Many other factors are cited in the literature as con¬
tributing to the student loan problem. For example, Jenkins
notes that there is such a "smorgasbord" of student assist¬
ance programs that it is possible for one student to be
eligible for eight different loan programs, sponsored by
eight different banks or lending institutions to pay eight

35
different terms with eight different sets of conditions for
forgiveness and eight different groups to deal with (Jenkins,
1978). Some consolidation appears to be necessary.
Another factor involves the apparent lack of sensi¬
tivity to issues of student loans from student financial aid
administrators. For example, the student aid packaging pro¬
cess usually relegates the discussion of the terms of a
student loan to the exit interview. Certainly this factor
is within the control of the student aid offices. Hope¬
fully, these offices will incorporate the suggestions made
in the National Association of College and University Busi¬
ness Officers (NACUBO) manual in order to manage better the
loan programs.
Finally, a legal process has been cited as a major
factor contributing to the student loan problem. The avail¬
ability of a bankruptcy action to discharge a student loan
debt has become increasingly apparent to many student borrow¬
ers. The use of a bankruptcy action, its effect on the
system of student aid and the governmental response to its
use are important considerations in student loan research
because they demonstrate an area where judicial and legis¬
lative cooperation has assisted in reducing the number of
the student loan bankruptcies and encouraging more students
to repay their student loans.

36
Bankruptcy Actions
Bankruptcy is the mechanism by which insolvent debtors
may be released from the obligation to repay their debts.
The proceedings are handled by a federal bankruptcy court
and conducted under laws enacted by Congress. In the mid-
1970s the problem of student loan bankruptcies received sig¬
nificant attention from Congress, colleges and universities,
professional organizations and the courts and governmental
agencies. This attention was not unwarranted as the problem
became quite prominent. During the five fiscal years from
1966 through 1970 only 348 bankruptcies were discharged on
NDSL and GSL loans totaling $400,000. However, in the fis¬
cal year 1975 alone, a total of 4,559 bankruptcies were dis¬
charged on loans totaling $6,800,000 (Leonard, 1980).
The effect of the tremendous rise in the use of bank¬
ruptcy actions to discharge student loan debts resulted in
a great deal of negative publicity for all student aid pro¬
grams. Prior to the disclosures regarding the use of bank¬
ruptcy actions, support for all federal aid programs was
exceedingly strong and resulted in the passage of the BEOG
(Pell Grant) Program. After considerable press exposure
(i.e., Mew York Times) public resentment led to a hearing by
the House Subcommittee on Postsecondary Education in October,
1975. The resentment was specifically aimed at a very few of
the 12 to 15% defaulting student borrowers, since very few

37
students default through bankruptcy. Host defaulting stu¬
dents choose not to provide forwarding addresses and thus
are "lost." According to some researchers, the small number
of students abusing the bankruptcy process solely to escape
federal educational debts is not a serious threat to the
loan programs. However, the anger surfacing at the October
committee hearing appeared to be generated by the apparent
misuse of a constitutionally sanctioned remedy to escape a
moral obligation (Tate, 1979).
Consequently, in 1976, Congress passed the Education
Amendments of 1976 which changed the procedures utilized to
discharge student loan debts by setting out specific restric¬
tions. The primary restriction was a mandatory five-year
delay from the commencement of the repayment period in the
filing of a bankruptcy action. Although a subprovision
allows for actions based on undue hardship, the purpose of
the amendment is to enforce a moratorium on all but the most
necessary bankruptcy discharges until sufficient time has
passed for a loan holder to either pay off the debt or have
enough at stake to make bankruptcy a genuine last resort
effort (Tate, 1979). This provision has since been incor¬
porated into the Bankruptcy Reform Act of 1978.
The Congressional action highlighted above indicates
that student loan problems can be adequately addressed and
resolved. However, some writers feel that the most potent
weapon against bankruptcy abuse and student loan defaults

38
may rest in the college registrar's office. Therefore,
actions taken by institutions as well as professional organi¬
zations, the government and professional educators are worth
researching as possible weapons in combatting the high stu¬
dent loan default rate.
Institutional Responses
to Student Loan Defaults
Nearly every college has a statement in its catalog in¬
dicating that the college reserves the right to withhold
grades, degrees, statements of honorable dismissal or a
transcript of credits until satisfactory settlement of col¬
lege fees and other financial obligations has been made.
This strategy has worked well for many years at several in¬
stitutions and is probably partly responsible for keeping
the default rate within its current level.
The appropriateness and legality of utilizing this
strategy have been questioned in several court cases; how¬
ever, the courts have not provided a definitive response to
guide colleges and universities especially if the student
has discharged the debt by bankruptcy action. It appears
that in situations where the debt was discharged prior to
October 1, 1979, the court will require state institutions
to release the transcript (Handsome v. Rutgers University,
445 F. Supp. 1362). However, private institutions may have
the authority to withhold transcripts of students whose
debts were discharged prior to October 1, 1979 (Girardier v.

39
Webster College, 563 F. 2d 1267). Due to the change in the
bankruptcy code, students are not now likely to have their
educational debts discharged, but if they do, it seems that
the college cannot take any action (including withholding
transcripts) to collect the debt.
The Handsome and Girardier cases point out that even if
the debt to the college has been discharged by a bankruptcy
decree there are other legal obstacles to the practice of
withholding transcripts. A future court may decree that the
student has sufficient property interest in the transcript
to require the college to provide adequate notice prior to
withholding the transcript. Or a court may determine that
there exists between the student and the college a contract
and that the college must demonstrate that the contract was
entered into with adequate notice regarding the withholding
of the transcript. In either case, additional responsibility
may be placed upon the colleges utilizing this strategy to
compel students to pay their educational debts.
In addition, courts have indicated that an unpaid debt
must be one that is sufficiently related to the student's
acquiring an education to justify retention of the transcript.
This decision presents several new issues. For example, is
a student's unpaid bookstore bill or a student parking fine
"sufficiently related" to permit the college to withhold
the student's transcript?

40
A final consideration regarding the withholding of
transcripts rests with the Family Educational Rights and
Privacy Act (Public Law 93-380). This act requires an edu¬
cational institution to allow the inspection of a transcript.
Although this act does not require a college to release an
official transcript, a court may rule in the future that
the institution must release an unofficial copy of a tran¬
script.
Obviously the strategy of withholding transcripts has
been viable throughout the years, but educational institu¬
tions should consider the legal ramifications of their
policies in light of the recent court decisions and the
questions presented in the court dicta.
Professional Organization Action
Regarding Student Loan Defaults
Beyond the strategies utilized by colleges and univer¬
sities to encourage student borrowers to repay their educa¬
tional loans, a few professions have begun policing their
members regarding loan defaults. The most notable example
is the case of the Florida Board of Bar Examiners re G.W.L.
(364 So 2d 454, Fla. 1978). In this case the Florida
Supreme Court refused admission to the state bar to the
petitioner, who had discharged his student loans in bank¬
ruptcy, on the premise that the petitioner's financial

41
status and apparent motivation at the time of filing re¬
flected a sufficient lack of the proper moral character and
requisite values to practice law in the State of Florida
(Martin, 1980). Since the court never questioned the act
of filing for bankruptcy but rather based its decision on
the student's conduct and the apparent breach of the
Florida Character Standards for bar admission (which were
established to protect the public interest), the decision
did not conflict with federal bankruptcy law regarding un¬
lawful penalties imposed on bankruptcy petitioners.
This court decision was the first to link discharge in
bankruptcy to moral turpitude and it has been cited in cases
in other states with similar provisions for professional li¬
censing. The purpose of the standards as stated by the
Minnesota Supreme Court is not to punish the attorney but to
guard the administration of justice and to protect the gen¬
eral public (in re Peterson, 274 N.W. 2d 922, 925, Minn.
1979). Therefore, the legal profession has set a precedent
by refusing bar admission to those who have utilized the
student loan system and irresponsibly neglect their debt
obligation. Since this strategy appears to be useful for
professional student bankruptcy actions and defaults, other
professional organizations, for example the AMA or ADA,
should institute and enforce similar provisions for licen¬
sing. This action is particularly relevant since the

42
pursuit of higher education is valued so greatly by the Amer¬
ican public in whose interest the professional serves and
any interference with the financial aid system supporting
the program of higher education in America should not be
treated lightly.
Governmental Action Regarding
Student Loan Defaults
Many governmental organizations and agencies have demon¬
strated concern for the student loan default problem and have
initiated reports and actions to help remedy the problem.
In addition, the U. S. Congress has been involved with legis¬
lation designed to curtail the default rate among federal
employees and in providing a forum in which to discuss
possible solutions for the loan default problem.
The General Accounting Office (GAO) has published sev¬
eral reports concerning the student loan system. Some of the
suggestions listed in its 1977 report included the following:
(1) require financial statements of
all debtors,
(2) refer all defaults immediately
to legal action,
(3) set up guidelines for compromise
settlements,
(4) institute a payroll deduction
system for all federal employees,
and
(5) require any contractor for collec¬
tion service to document unsuccess¬
ful efforts. (U.S. General Account¬
ing Office, 1977)

43
Although these suggestions seem realistic, they have met re¬
sistance from several sources. For example, in 1979 the
U. S. Department of Health, Education and Welfare (now the
U. S. Department of Education) argued successfully that only
it should negotiate settlements of compromise. In a subse¬
quent (1981) report the United States Comptroller General in¬
dicated that many schools visited by the GAO made inadequate
efforts to collect defaulted loans, that the schools needed
to adopt a tougher attitude toward collecting defaulted
loans and that the Department of Education should take
stronger actions against those institutions that failed to
do so. Consequently, in April of 1982, the Department of
Education notified colleges with a poor record of loan
collection that they would not receive any new federal loan
money appropriated for the 1982-83 year. Although these
actions referred specifically to the NDSL program, they
sufficiently demonstrate the interest in Washington regard¬
ing the student loan problem.
During the Oversight Hearing before the Subcommittee
on Postsecondary Education of the Committee on Education and
Labor (House of Representatives) the following methods were
suggested as appropriate to cut down the GSL default rate:
(1) discharging federal employees
who default,
(2) extending deferments to one year
for those who are unemployed,

44
(3) permitting lower rates of
repayment for those with low
incomes, and
(4) prohibiting defaulters from
obtaining other federal loans.
(U. S. Congress, 1977)
In response to item number one above, Education Secre¬
tary Terrence Bell and Senator Charles Percy (R-Ill.) spon¬
sored legislation in 1982 enabling the government to garnishee
the wages of federal employees who have defaulted on their
student loans. Recently, the Department of Education sent
letters to federal employees notifying them of the Debt
Collection Act of 1982 and indicating they must make
arrangements for the repayment of their loan by February 1,
1983, or risk the garnishment of their wages. Therefore, it
is apparent that both the Executive and Legislative branches
of the U. S. Government have also taken actions to assist
in the solution of the loan problem.
Responses to the Student Loan
Default Problem by Education Spokesmen
Professional educators have also addressed the problem
of student loans and are often quoted in the literature of
higher education. For example, Robert Hartman identifies
several factors that impinge on the effectiveness of the
current student loan system including the variety of loan
programs and regulations, the use of bankruptcy actions to
discharge debts and the differences among students and

45
their respective needs. He recommends several restrictions
to the current loan program. Among his suggestions are:
(1) the development of a single
loan instrument,
(2) curtailing the use of bank¬
ruptcy actions (at least for
the first 10 years rather
than the five years provided
in the Education Amendments
of 1976), and
(3) a provision allowing different
loan amounts for students of
different needs, and extended
payment and variable repayment
options. (Hartman, 1978, p. 96)
Hartman argues that his suggestions would increase the cost
effectiveness of the loan program (items one and two), in¬
crease the manageability of loan repayment procedures (items
one and three), widen student accessibility (items one and
three), and reduce defaults and delinquency (items one and
three). Essentially he is suggesting that the loans be
conformed to fit the student, that the loan programs be
supported identically by each of the fifty states and that
the performance requirements and benefits of the loan pro¬
grams be identical regardless of the program utilized.
Hartman's suggested actions seem to be consistent with the
goals of the student aid delivery system and show great in¬
sight concerning the many elements constituting the student
loan problem.
Several other plans have been suggested by professional
educators regarding the student loan program. A panel of

46
advisors during the Kennedy Administration developed the idea
of an Educational Opportunity Bank. The proposal would have
allowed students to borrow money to cover their expenses at
whichever college they decided to attend. The loan would be
paid back over the next 30 to 40 years and an interest rate
of 1% of the students' gross income would be applied. Al¬
though this program would have allowed students to pay their
own college expenses, Duke University and Yale University en¬
countered several problems when instituting similar programs
including administrative and collection difficulties.
D. Bruce Johnstone, in his proposal of the National
Student Loan Bank, recommends that one new major govern¬
mental agency should assume the responsibility for all stu¬
dent loans (Johnstone, 1978). In addition, Johnstone sug¬
gests that the colleges and universities (rather than the
lending institutions) should originate all loans and that
the repayment of interest should be tax deductible necessi¬
tating a tie-in to the Internal Revenue Service. This
radical plan appears to be quite practical; however, it
tends to come into conflict with many existing offices,
agencies and banks as it would drastically change and/or
perhaps eliminate their participation in the student loan
system. Therefore, support for this proposal has never
materialized.
Most attention today regarding changes in the student
loan programs has focused on John Silber's Tuition Advance

47
Fund (TAF) proposal. The proposal incorporates several key
factors including
(1) loans would only be negotiated
for successful students (after
their freshman year),
(2) individual student TAF accounts
would have to be established
with the respective college
business office,
(3) involvement of the Internal
Revenue Service would be re¬
quired in the transaction and
repayment procedures, and
(4) indefinite repayment periods
would be instituted since the
repayment provisions require
students to repay their loan
at the rate of 2% of their
annual gross income each year
until they have paid back 150%
of their loan. (Silber, 1978, p. 7)
To its credit the TAF proposal provides some novel
ideas such as treating loan payments as tax obligations and
pooling the income of married couples with loans for the
purpose of calculating their taxable income. However, the
plan has come under negative scrutiny as many have argued
that the proposal assumes that the burden of paying for an
education should rest with the student. Beyond the previous
arguments concerning who should bear the cost of an educa¬
tion, many professionals fear this approach may lead par¬
ents, private donors and the taxpayer to abandon their
support for higher education which would force an even
greater burden on the student.

48
In addition, although research studies show that a long
repayment period provides the best opportunity for a reason¬
able return incentive for investing in higher education,
many bankers indicate that the TAF could never be self-
sustaining based on the proposed 2% interest rate over an
estimated 30 year repayment period. Other arguments against
the TAF include the feeling that any plan that provides for
the paying of higher education through long term student
debts will result in limiting equal access and opportunity
for higher education. Further, the tuition and enrollment
controls suggested to distribute TAF funds to institutions
would cause widespread disruptions in the higher education
system. For example, the expanded eligibility requirements
for private college and university students would enable them
to be eligible for a larger proportionate share of the
available money. Regardless of one's point of view, the TAF
represents a proposal which has provided healthy discussions
of the student loan system and the options available to
solve the problems.
Antecedent Studies Relating to
Student Loans and Demographic Characteristics
It is apparent that all the participants in the student
aid delivery system are concerned about the student default
rate on educational loans. The changes implemented thus far
in the student loan system have been instrumental in improv¬
ing administrative procedures. For example, new and better

49
resources for collecting loans and monitoring repayment pro¬
gress have been devised. A second approach to solving the
student loan problem should consist of researchers studying
the demographic characteristics of student borrowers in
order to distinguish those students who are likely to repay
student loans from those students who are not likely to repay
student loans. Thus far only a few researchers have explored
the characteristics of student borrowers in the last five
years.
In 1977, Dyl and McGann applied a multivariate discrim¬
inant analysis to the problem of identifying defaulting and
non-defaulting student borrowers at the University of Wyoming.
They analyzed 33 different variables including class standing,
college attended, grade point average, amount of the loan,
academic major, marital status, living quarters, total amount
of university loans, and size of the monthly payments. Four
of the significant discriminators displayed direct positive
relationships with actual loan repayment behavior. The stu¬
dents who were more likely to repay their student loans in¬
cluded those students who
(1) had higher grade point averages,
(2) were married,
(3) were engineering majors, and
(4) chose high monthly payments.
Dyl and McGann also discovered three factors which were nega¬
tively associated with repayment of student loans including

50
the (large) size of the loan, the residence of the student
in an apartment and the total amount of indebtedness to the
university (including previous loans). As cited by Davis and
Van Dusen (1978), this study, in concert with the earlier
studies of Pattillo and Wiant (1972) and Spencer (1974), was
instrumental in identifying for other researchers the factors
most worthy of analysis and providing a format in which to
research the characteristics of student borrowers.
Hauptman, in his 1978 study for the College Scholarship
Service, explored many of the same characteristics for stu¬
dents participating in the GSL/FISL programs. One of the
most interesting findings of his study was that the follow¬
ing groups of students had a higher percentage of defaults
than would be expected:
(1) students with low family incomes,
(2) female students,
(3) married students,
(4) black students, and
(5) junior college students.
Hauptman discovered that students attending public schools
had a lower default rate than those attending either private
or proprietary schools and that college and university
students had a lower default rate than junior college stu¬
dents or vocational students. Although the results of this
study did not duplicate the results presented by Dyl and

51
McGann (1977), it is equally important because it researched
a variety of institutions and concentrated its data on one
loan program.
Also in 1978, Emmert studied the characteristics of
students defaulting on their NDSL loans and concluded that
the demographic make-up of student populations has an equal
or greater impact upon default rates than do internal admin¬
istrative factors. He suggested that schools with higher
default rates are not necessarily administering their pro¬
grams any less effectively. Instead he argued that they may
have students from substantially different backgrounds that
account for the different default rate. Emmert also cited
all the previous researchers in synthesizing the crucial
variables selected in his study. Among the variables
selected were:
(1) age,
(2) marital status,
(3) sex,
(4) college standing,
(5) grade point average,
(6) loan total, and
(7) employment status.
Like the Hauptman study (1978) , the differing results Emmert
discovered seem to suggest that student demographic character¬
istics should be weighted in order to establish variable de¬
fault rates for the different postsecondary institutions.

52
In 1979 the Office of Education completed its first
study of student borrower characteristics. Although the
data collection was restricted to eight proprietary insti¬
tutions, the study singled out several variables that may
impact the default rate. Essentially this study underscores
the necessity of establishing statistical validity by re¬
searching individual types of institutions independently
since the results vary for the separate types of institutions
studied. Therefore, because of the unique demographic char¬
acteristics of the proprietary student, this study concluded
that the higher default rate among proprietary institutions
discovered in the Hauptman study is understandable.
Finally, a study completed by Myers and Siera in 1980
concentrated on data available from the New Mexico Student
Loan Program for the New Mexico State University. Using the
Statistical Package for the Social Sciences (SPSS) Discrim¬
inant Program, Myers and Siera attempted to develop a pre¬
diction model to discern defaulters from non-defaulters.
The results indicated that the variables selected did not
lead to an accurate prediction of the likelihood of a stu¬
dent defaulting on a loan. What prediction was possible
was not substantially different from what one might expect
to accomplish based on chance alone. Although the study did
not present an acceptable prediction model, the work of
Myers and Siera was instrumental in pointing out the need

53
to explore and refine the application of predictive models
to the student loan problem.
Summary
All of the personnel, institutions, organizations and
governmental agencies involved with student loan programs
have provided input and taken selective action to reduce
the student loan default rate. A majority of the strate¬
gies implemented thus far have been concerned with the ad¬
ministrative or judicial (bankruptcy) aspects of the pro¬
blem. In order to deal effectively with the problem of de¬
faults the studies cited demonstrate the necessity of iden¬
tifying the characteristics of defaultees and non-default-
ees.
Several of the studies presented emphasized the cru¬
cial demographic characteristics worth researching. Other
studies demonstrate the necessity of restricting the popu¬
lation researched due to the inherent differences among the
students choosing to attend the different institutions of
higher education. Therefore, this study has restricted its
data base to community college students in the state of
Florida and researched the available data that corresponded
to the variables highlighted in the previous studies.
The student loan problem is not the creation of a
single causality but rather it is the result of the

54
attitudes, procedures, economics and actions of all of the
participants (students, lenders, aid administrators) in the
program. Certainly, additional study needs to be performed
regarding the administrative problems within the student
loan program. However, research regarding the students
who partake in the program has long been relegated to a
subsidiary status. It appears crucial that additional re¬
search needs to be conducted to identify those variables
which are effective in predicting the likelihood of a stu¬
dent default, because the student is indeed the major
factor contributing to the student loan default problem.

CHAPTER III
DEVELOPMENT OF THE MODEL
The research reviewed in Chapter II provided an over¬
view of the demographic variables to consider in the forma¬
tion of a prediction model for student loan repayment.
Names of the major contributors to the literature and the
primary variables they studied are shown in Table 2. As
presented in Chapter II many of these studies provided rele¬
vant findings concerning relationships between demographic
characteristics of student borrowers and loan default. These
relationships will be reviewed as a basis for developing the
prediction model.
In one of the first studies researching demographics
and student loans, Pattillo and Wiant (1972) as cited by Davis
and Van Dusen (1978), discovered that students who borrowed
late in their academic years, who had previous loans, or who
came from large families were more likely to be delinquent
in their student loan payments (Davis & Van Dusen, 1978). In
addition, Pattillo and Wiant concluded that items reflecting
financial rather than biographical data appeared to be better
predictors of loan delinquency (Davis & Van Dusen, 1978).
Spencer completed a study of the relationship between
demographics and student loan default in 1974. In this study,
55

Table 2
Overview of Major Variables
Selected by Previous Researchers
(1972) (1974) (1977) (1978) (1978) (1979) (1980)
Pattillo Dyl & Office of Myers
& VJiant Spencer McGann Hauptman Emmert Education & Siera Total
Age
Marital Status
Sex
College Standing
Grade Point Average
Loan Total
Family Income
Student Aid Received
Cost of Attendance
College Major
Type of College
Ethnic Background
Employment Status
Academic Load
Degree Earned
Type of Residence
Dependents (Number)
Age of Automobile
Family Size
Telephone
X X
X X
X X
X XX
X
X XX
X
X
X
X
X
X
X
X
X
X
X
X X
X X
X X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
5
6
6
5
3
5
3
2
1
2
1
3
2
1
1
2
2
2
1
2 cn

57
according to Emmert (1978), Spencer described the worst
possible student loan risk as an unmarried, unemployed male
student, in his last semester, with an old car, without a
phone, and with a large outstanding loan debt (Emmert, 1978).
The results that Dyl and McGann obtained in 1977 indi¬
cated that the following factors related positively to the
repayment of student loans: college major (engineering),
marital status (married), grade point average (high), and
the size of loan payments (low). Dyl and McGann (1977) also
found the following three factors to be negatively associated
with repayment: residence type (apartment), total university
indebtedness, and loan indebtedness (high loan totals). In
this study Dyl and McGann also presented the results of a
discriminant analysis of their data, and they were able to
correctly classify 84% of the cases in their study. However,
as pointed out by Myers and Siera (1980), Dyl and McGann did
not test their model with any new cases. Therefore, the
reliability of the model presented by Dyl and McGann is, as
yet, untested.
Essentially these three studies presented the basis
upon which most of the research in this area has concen¬
trated. These researchers have not only provided insight con¬
cerning the appropriate variables to research but they also
have suggested the appropriate statistical designs for fu¬
ture studies. However, the conclusion originally presented
by Pattillo and Wiant (1972) regarding the inappropriate use

58
of biographical data for predictive purposes has not been
heeded. Many researchers have continued to search for stu¬
dent demographic characteristics which would distinguish
student loan defaulters from non-defaulters.
For example, in 1978 Hauptman studied five demographic
variables and reported that the worst risk student loan par¬
ticipant was a married, Black, female student from a low
family income who attended a proprietary vocational school.
In another study presented in 1978, Emmert summarized the
previous studies reported in the literature. In this
article Emmert reported the variables that have been studied
and the conclusions that have been presented concerning the
effect these variables have had on the student loan default
problem. Emmert also noted the different default rates
among postsecondary institutions and suggested that the demo¬
graphic make-up of a given student population may place an
institution in a high risk category for student loan defaults.
In conclusion, Emmert suggested that the demographic factors
should be weighted in order to establish variable default
rates for differing types of institutions.
A 1979 study prepared by the Office of Education also
reported student demographic data in relation to the problem
of student loan defaults. Although the study primarily was
designed to research the administrative procedures utilized
by selected proprietary institutions, it did show that many

59
but not all of the same variables selected by other research¬
ers were important variables concerning the student loan
default problem in proprietary institutions. The findings
presented in this study demonstrate the necessity of study¬
ing individually the different types of postsecondary insti¬
tutions .
The most recent study in this area of financial aid re¬
search was carried out in 1980 by Myers and Siera. Through
t-test analysis, Myers and Siera found the following vari¬
ables to be highly associated with the repayment or non¬
repayment of student loans: college standing, loan total,
amount of loan requested, grade point average, and degree
completion. Myers and Siera also attempted to replicate
the study design presented by Dyl and McGann (1977); and, al¬
though they were able to formulate a promising prediction
model, when this model was tested using new cases only 42.5%
of the new cases were correctly predicted. The validity
problem experienced by Myers and Siera in their prediction
model illustrates the difficulty researchers have encountered
in their search for an accurate, useful prediction model for
student loan defaults.
It is apparent from the above discussion that although
the conclusions of the previous studies differ considerably,
there does appear to be some consensus regarding the impor¬
tance of researching student demographics and which demo¬
graphics should be studied. In selecting which variables

60
to use in this study, the researcher selected the variables
which had been most frequently used by previous researchers
(see Table 2). Although ethnic status and family income
were studied by several researchers, these variables were
not included in this study because of the difficulty in col¬
lecting and verifying these variables. The remaining six
variables were available in the data bank of the Florida
Student Financial Aid Commission in Tallahassee, Florida.
As a result, the following six variables were selected for
the prediction model presented in this study:
(1) age,
(2) marital status,
(3) sex,
(4) college standing/
(5) grade point average, and
(6) loan total.
Analysis of Variables
As indicated in Chapter I, this ex post facto study was
designed to determine the usefulness of a model to distin¬
guish those students likely to repay student loans from those
students who are not likely to repay student loans. Prior to
developing any model, there was one major issue to address:
were the selected variables useful in discriminating between
students who will repay student loans and students who will
not repay student loans?

61
In order to determine the usefulness of the variables
in discriminating those students who will repay student loans
from those students who will not, data representing each of
the six variables were collected for both defaulting and non¬
defaulting students in the sample. Since three of the vari¬
ables were nominal in nature, the t-test statistic utilized
by previous researchers to determine significant differences
between defaulters and non-defaulters was not applicable to
the data. Therefore, a chi-square analysis was performed on
the data collected for each of the variables in this study
to determine if the differences between the proportions of
subjects that fell into the two different categories (de¬
faulters and non-defaulters) were significant.
Since this study represents exploratory research and
any model developed from the data will need to be validated
with new data, the researcher selected a .10 significance
level. Essentially the chi-square analysis determines if
the difference between the expected and observed frequencies
for each of the classifications within each variable is
greater than the difference one would expect to find based
on chance alone. The results of the chi-square analyses
showing the relationship of each of the six variables to
student loan default are presented in Table 3 through
Table 9.

62
Table 3
Chi-Square Analysis of the
Relationship Between Age and Student Loan Default
Age
Defaulters
Non-Defaulters
18-20
Number
8
10
% of total
frequency
10.53
13.16
21-24
Number
10
9
% of total
frequency
13.16
11.84
25-27
Number
8
6
% of total
frequency
10.53
7.89
Over 27
Number
12
13
% of total
frequency
15.79
17.11
Chi-
Square = 0.6010
DF = 3
Probability = .8960
As shown in Table 3, the chi-square analysis of the re¬
lationship between age and student loan default indicates
that age was not significant at the .10 level. The expected
frequency for each cell based on chance alone would be 12.5%
and, as shown, the actual frequency percentage for the cells
ranged from 7.9% to 17.1%.
The chi-square analysis of the relationship between
marital status and student loan default is shown in Table 4.
The results of the chi-square analysis for marital status
indicate a statistical significance at the .10 level. The
expected frequency for each cell based on chance alone would
be 25% and, as shown by the table, the actual frequency per¬
centage for the cells ranged from 14.5% to 35.5%. Therefore,

63
Table 4
Chi-Square Analysis of the Relationship
Between Marital Status and Student Loan Default
Status
Defaulters
Non-Defaulters
Married
Number
18
11
% of total
frequency
23.68
14.47
Single
Number
20
27
% of total
frequency
26.32
35.53
Chi-Square = 2.732 DF = 1 Probability = 0.0983
the use of the variable marital status in a subsequent pre¬
diction model was appropriate.
As shown in Table 5, the chi-square analysis of the re¬
lationship between sex and student loan default indicates
that sex was not significant at the .10 level.
Table 5
Chi-Square Analysis of the Relationship
Between Sex and Student Loan Default
Sex
Defaulters
Non-Defaulters
Male
Number
18
13
% of total
frequency
23.68
17.11
Female
Number
20
25
% of total
frequency
26.32
32.89
Chi-Square = 1.362 DF = 1 Probability = 0.2432

64
The expected frequency for each cell based on chance alone
would be 25% and, as shown, the actual frequency percentage
ranged from 17.1% to 32.9%.
The chi-square analysis of the relationship between
college standing and student loan default is shown in
Table 6. The data collected did not indicate college stand¬
ing to be significantly related to defaulters and non¬
defaulters at the .10 level. The expected frequency for each
Table 6
Chi-Square Analysis of the Relationship Between
College Standing and Student Loan Default
College Standing Defaulters Non-Defaulters
Freshman
Number
21
17
% of total
frequency
27.63
22.37
Sophomore
Number
17
21
% of total
frequency
22.37
27.63
Chi-Square = 0.8420
DF = 1
Probability =
0.3588
cell based on chance alone would be 25% and, as shown, the
actual frequency percentage ranged from 22.4% to 27.6%.
Table 7 shows the chi-square analysis of the relation¬
ship between grade point average and student loan default.
The results of the chi-square analysis indicated that grade
point average was not significant at the .10 level. The

65
Table 7
Chi-Square Analysis of the Relationship Between
Grade Point Average and Student Loan Default
Grade Point Average
Defaulters
Non-Defaulters
1.50-1.99
Number
2
0
% of total
frequency
2.63
0.00
2.00-2.49
Number
18
13
% of total
frequency
23.68
17.11
2.50-2.99
Number
8
12
% of total
frequency
10.53
15.79
3.00-3.49
Number
8
8
% of total
frequency
10.53
10.53
3.50-4.00
Number
2
5
% of total
frequency
2.63
6.58
expected frequency percentage for each cell would be 10%
based on chance alone; however, the actual frequency percent¬
age for the cells ranged from 0.00% to 23.68%.
The chi-square results of the relationship between loan
total and student loan default are presented in Table 8. The
loan total variable was significant at the .10 level. The
expected frequency for each cell based on chance alone would
be 8.3% and, as shown, the actual frequency percentage for
the cells ranged from 0.0% to 18.42%. Therefore, the size
of the loan total was a significant variable to include in
any prediction formula designed to ascertain the likelihood
of a student loan default.

66
Table 8
Chi-Square Analysis of the Relationship Between
Loan Total and Student Loan Default
Size of Loan
Defaulters
Non-Defaulters
0-$l,000
Number
13
6
% of total
frequency
17.11
7.89
$1,001-$2,000
Number
11
6
% of total
frequency
14.47
7.89
$2,001-$3,000
Number
13
14
% of total
frequency
17.11
18.42
$3,001-$4,000
Number
1
5
% of total
frequency
1.32
6.58
$4,001-$5,000
Number
0
5
% of total
frequency
0.00
6.58
$5,001-$6,0 00
Number
0
2
% of total
frequency
0.00
2.63
Chi-Square
= 15.7530
DF = 5
Probability = 0.0173
The results of the chi-square tests showed that only two
of the variables were significant at the .10 level (marital
status and loan total). The differences found between the
categories (student loan defaulters and non-defaulters) for
both of these variables cannot be attributed to chance alone.
The remaining four variables (sex, grade point average, age
and college standing) were not statistically significant
according to the chi-square results of this study, but they
may demonstrate some differentiation between defaulters and

67
non-defaulters on student loans based on the results of
previous researchers.
The results of the chi-square analyses indicated that
only two of the six variables studied most frequently by pre¬
vious researchers were significant according to the data
collected in this study. The discovery that the only finan¬
cial type variable (loan total) was also one of the two vari¬
ables found to be significant may underscore the importance
of Pattillo and Wiant's (1972) conclusion that items reflect¬
ing financial rather than biographical data appear to be
better predictors of loan delinquency. This finding may
indicate that since students with high loan needs have good
loan payback rates, the students with the highest needs are
the best loan risk for any one of a number of reasons (i.e.,
they may value money more due to its scarcity, they may have
learned to manage money more frugally or they may have
learned how to budget more effectively). The strength of the
relationship of this one variable to student loan payback
indicates that other financial type data should be studied.
Before developing a prediction model based on the re¬
sults of the chi-square analyses, the researcher explored
the possibility of interaction effects between the six vari¬
ables. These results are presented in Table 9. It can be
observed that no interaction between the variables related
significantly to payback.

68
Table 9
Interaction Effects Between the Variables
Variables
F
Pr> F
Significant ( o< =.10)
Loan by Sex
1.45
.2320
No
Loan by Grade Point
0.01
.9185
No
Loan by Marital Status
0.10
.7573
No
Sex by Marital Status
0.05
.8250
No
Sex by Grade Point Aver.
0.47
.4956
No
Grade Point by Marital
Status
2.14
.1483
No
Therefore, in developing the prediction model, the researcher
utilized only the six original variables selected from the
literature review.
Although four of the selected variables were not signi¬
ficant according to the chi-square results of this study,
they were included in the formation of the prediction model
because the literature was inconclusive regarding which vari¬
ables should be used in a prediction model. It should be
noted that the unique results of the chi-square analyses in
this study may be due to the unique population. Since the
previous studies did not examine the relationship between
selected variables and student loan default for community
college students, the difference in results may be due to the

69
specific characteristics of the community college population.
Since the results presented in the literature and the results
found for the community college students in this study were
not corroborative regarding the significance of the vari¬
ables, all six selected variables were utilized in the pre¬
diction model.
Development of the Model
In order to try to predict categorical membership based
on the most discriminating variables, a multiple discriminant
analysis technique was employed. Essentially this statisti¬
cal procedure "weights" the predictor variables to yield
maximum discrimination between the two groups (i.e., in this
study it discriminates defaulters from non-defaulters) (Hays,
1981). In addition, the nominal variables of marital status,
college standing and sex were used to define dummy variables
in order to introduce variance into the prediction equation.
Three notable formulae generated by the multiple discrim¬
inant analysis function of the Statistical Analysis System
(SAS) program were useful in distinguishing student loan
defaulters. Therefore, three prediction models were devel¬
oped.
The first formula included all six variables which were
entered into the data file in the following (arbitrary)
order: age, marital status, sex, college standing, grade

70
point average, and loan total. The results of this multiple
discriminant analysis indicate that the six selected vari¬
ables correctly predicted group membership for approximately
68% of the sample cases. Considering that the prior proba¬
bility (based on chance alone) would be a 50% correct pre¬
diction, the six selected variables appear to be successful
in improving the accuracy of predicting group membership.
In order to improve the chances of making more accurate
predictions, a second discriminant analysis procedure was
performed. This time the multiple discriminant analysis
enabled the researcher to catalogue the individual effect of
each selected variable on the accuracy of the categorical
prediction. Based on the probability and significance
levels of the chi-square analysis the most differentiating
variables were first included in the prediction model. When
the first variable, loan total, was entered into the pre¬
diction formula, the chances of making accurate predictions
improved notably. Instead of the 50% prior probability of
making accurate categorical predictions the accuracy rate
improved to 66%. The only other significant variable
(marital status) was entered into the model next. The in¬
clusion of this variable increased the probability of
correctly categorizing the defaulting and non-defaulting
student loan recipient from 66% to 71%. Therefore, this
second formula (generated by the stepwise multiple

71
discriminant technique) included only the two variables
demonstrating the acceptable degree of significance based
on the chi-square results ( e< = .10) and it provided the
best accuracy for categorical prediction (71%).
When the next two variables, sox and grade point aver¬
age, were included in the model, the percentage of correct
categorical classifications essentially remained the same.
Although the variables would appear to be useful in a pre¬
diction formula as indicated in the literature review, their
inclusion was not helpful in formulating a more accurate
prediction model. When the variable sex was included in the
model, the accuracy rating remained at 71%; however, when
the variable grade point average was included, the number
of correct classifications decreased by one and reduced the
accuracy rating to 70%. Such a result seems to be incon¬
sistent with statistical theory. After studying the values
for each of the sample cases generated by the prediction
formula, the researcher believes that the decreasing
accuracy may be due to the mathematical rounding procedure
within the Statistical Analysis System program. As a result,
one sample case that defaulted on a student loan was classi¬
fied by the prediction model as a non-defaulter on the total
probability difference of .0033. Therefore, the apparent
decreasing accuracy of the prediction model may be attributed
to the mathematical precision of the computer program rather
than the reflection of a conflict with statistical theory.

72
The third formula was generated at the conclusion of
the stepwise multiple discriminant technique and it included
all six selected variables. The addition of the remaining
two variables, college standing and age, did not improve the
accuracy rating for the prediction model. Therefore, the
formula, including all six selected variables, generated by
the stepwise technique was able to predict correct categori¬
cal membership for 70% of the sample cases.
Although both the first and third formulae included the
six selected variables, the disparity in the predictive
accuracy of the formulae is accounted for in the statistical
technique employed for the derivation of the formulae.
Since the stepwise multiple discriminant analysis technique
enters the variables in the order of probability, the im¬
proved accuracy of the third formula (70%) is a result of
the mathematical process of weighting the variables as they
are entered into the prediction formula. Because the order
of inclusion in the formula is different and the mathemati¬
cal weighting is improved, the accuracy of the formula is
improved.
Based on the results of the discriminant analysis, it
would appear that the best predictive formula for determin¬
ing correct classification of defaulting and non-defaulting
community college students would include only data regarding
the students' loan totals and marital status. Adding the

73
variables grade point average, sex, college standing and
age, essentially maintained the effectiveness of the pre¬
diction formula. As evidenced, the stepwise technique em¬
ployed to produce the second and third formulae improved the
accuracy for categorical classification over the regular
multiple discriminant analysis technique utilized for the
first formula.
Results
Based on chance alone, one would expect to predict
correctly the proper placement of students into the two
categories (defaulting and non-defaulting) 50% of the time.
The prediction formula generated by the inclusion of data
concerning loan total and marital status increases the like¬
lihood of making an accurate prediction from 50% to 71%.
Therefore, it would appear that these data are relevant in
predicting whether a student is likely to default on a
student loan. The additional data that were collected for
the other four variables (sex, grade point average, college
standing and age) also appear to be relevant in producing a
prediction model since the prediction formula derived from
all six variables produced a 70% accuracy rating concerning
categorical prediction and this rating is substantially
higher than the 50% probability based on chance alone.

74
However, all three of the prediction formulae reflect
an error factor of nearly 30% indicating that roughly one-
third of the cases were classified in error. In addition,
the testing of the prediction formula utilizing new cases is
not possible since the formula is derived from a stratified
sample (which was necessary in order to perform the chi-
square tests to determine variable usefulness). Therefore,
the usefulness of the results of this study is limited even
though the prediction formula demonstrates moderate success
in differentiating defaultees from non-defaultees.
Previous researchers also produced prediction models
with limitations. For example, although Dyl and McGann
(1977) reported significant success with their prediction
model, Myers and Siera (1980) indicated that Dyl and McGann
did not validate their results by applying the formula to
cases other than those from which the formula was derived.
In addition, the results of the study performed by Myers
and Siera seem to indicate that although they presented a
useful model, when they attempted to validate their model
by predicting new, non-sample cases, the model was not
reliable or useful.
The results of this study show that the six variables
which were selected based upon the results of the studies
presented by previous researchers were not reliable predic¬
tors of payback for community college students. Therefore,
a dependable prediction model cannot be formulated for

75
community college students based on these variables. Perhaps
the importance of this study is that the results coincide
with findings reported by Hauptman (1978), Eramert (1978) and
Pattillo and Wiant (1972) .
Hauptman first suggested that research must be indepen¬
dently conducted on the different types of postsecondary in¬
stitutions. Underscoring this perception, Emmert concluded
that institutional default rates may not be comparable if
the schools' respective populations are from substantially
different backgrounds. It may be significant that the two
studies previously successful in determining reliable demo¬
graphic variables and formulating prediction models (Dyl and
McGann, 1977, and Myers and Siera, 1980) were based on four-
year public univeristy (University of Wyoming and New Mexico
State University respectively) samples.
Previous discussion regarding the importance of finan¬
cial aid for community college students demonstrated the
meaningful differences between the typical community college
student and the typical four-year university student. The
disappointing accuracy of this model, specifically the in¬
correct classification of 30% of the sample, may be attrib¬
uted to the demographic differences between the sample in
this study and the samples used in the previous studies.
For example, the results of the previous research reported
in the literature indicate that the demographic data repre¬
sented by the six selected variables discriminate student

76
loan defaulters and non-defaulters for four-year college or
university students. However, as indicated by the chi-square
results in this study, these same variables do not entirely
discriminate student loan defaulters from non-defaulters for
community college students. Therefore, the variables,
selected based on the success of the previous studies con¬
ducted with a university student sample, are apparently not
the best variables to use when attempting to predict
community college student loan defaulters.
The results of this study also appear to support
Pattillo and Wiant's (1972) conclusion that financial rather
than biographical data are better predictors of loan delin¬
quency. One of the two variables that met the .10 statistical
significance level in the chi-square analyses of this study
was the loan total. It was also the only variable that was
financially related since all of the other (five) variables in
this study represented purely biographical demographic charac¬
teristics. Considering that the only financially related vari¬
able was also statistically significant, perhaps more finan¬
cially related variables would yield a more accurate prediction
model. Therefore, the results of this study may lead to the
conclusion that, as suggested by Pattillo and Wiant, more
financial data need to be collected and analyzed in order to
make any reliable prediction of loan delinquency. This
conclusion may be also applicable to the results presented
by Dyl and McGann (1977) and Myers and Siera (1980).

77
Although they have been able to develop successfully a pre¬
diction model, neither Dyl and McGann nor Myers and Siera
has been able to validate successfully their models by pre¬
dicting accurately new cases, cases not presented in the
sample that derived the prediction model. Perhaps they need
to collect more financial data regarding their students in
order to formulate an accurate prediction model that can be
validated by new cases.
Summary
The prediction formula derived from the data available
through the Florida Student Financial Aid Commission for
Florida community college students indicates that it is
possible to make a useful discrimination regarding who will
and who will not repay their student loans for approximately
70% of the sample cases. This prediction formula was
derived through a (stepwise) multiple discriminant analysis
computer program utilizing the six selected variables high¬
lighted in previous research studies as useful discriminators
between defaulting and non-defaulting students. A second
(stepwise) multiple discriminant analysis utilizing only the
two variables demonstrating statistical significance
(<*■ = .10), loan total and marital status, yielded a pre¬
diction formula with a categorical prediction accuracy
rating of 71%. A third formula utilizing all six selected
variables entered into the formula in random order

78
(non-stepwise) generated a model that predicted correct
categorical membership for 68% of the sample. Regardless of
the formula utilized, approximately one-third of the sample
cases were not predicted accurately. Although none of the
prediction models are accurate for one-third of the cases,
each prediction model is capable of predicting nearly 70%
of the cases correctly. Therefore, the models are of some
value in addressing the questions of this study. Future
models may be more useful in predicting those students
likely to repay student loans from those students who are
not likely to repay student loans since the models derived
from the data in this study have limitations concerning
their application as discussed in Chapter IV.
The limited success of these prediction models appears
to be consistent with the findings presented by the previous
researchers in the field. Since the difficulty in validat¬
ing even successful prediction models appears to be wide¬
spread, as confirmed by the results of previous researchers,
it may be that the variables utilized thus far are not the
most discriminating variables regarding student loan de¬
faults. Perhaps future research should focus on financial
type variables in order to provide relevant, statistically
acceptable prediction models to assist student financial aid
administrators in predicting student loan defaulters.

CHAPTER IV
CONCLUSIONS, IMPLICATIONS AND
RECOMMENDATIONS
Previous studies researching the relationship between
student demographics and loan repayment recorded similar
results to this study. However, the results of this study
demonstrate several unique implications for future re¬
searchers .
Conclusions
Since the models presented in this study improve the
chances of making an accurate prediction concerning the
likelihood of default on student loans from 50% (based on
chance alone) to nearly 70% (based on the models), the
models are successful in discriminating student loan de¬
faulters from non-defaulters. Therefore, the questions of
this study presented in Chapter I must be answered in the
affirmative.
Even though some of the six selected variables were not
statistically significant in this study, the inclusion of
all six variables in the prediction models was warranted
because of the inconclusive results regarding the importance
of the individual variables as demonstrated by previous
university researchers. Inclusion of all six variables was
79

80
also appropriate since this study was the first to utilize a
community college sample, and the results of the chi-square
analyses in this study may indicate that the variables
distinguishing community college student loan defaulters may
be different from the variables distinguishing university
student loan defaulters.
Despite the differing results of the chi-square analyses
for the individual variables, the models formulated utilizing
all six selected variables were accurate in predicting
default or non-default for nearly 70% of the sample cases.
Therefore, the models presented in this study are useful in
predicting student loan defaulters. The conclusions of this
study suggest several implications of importance for future
studies.
Implications
The most notable implication of this study is that
there are differences between community college students and
university students regarding the significant variables dis¬
tinguishing student loan defaulters from non-defaulters.
This study shows that, although previous researchers identi¬
fied the six selected variables and determined that they were
significant in distinguishing defaulting university students,
only two of the six variables were significant in distin¬
guishing defaulting community college students. Therefore,
this study underscores the importance of performing separate

81
study designs yielding specific results for the different
kinds of institutions utilizing the Guaranteed Student Loan
system.
The results of the chi-square analyses also appear to
indicate that the data currently being collected for commun¬
ity college students do not adequately distinguish defaulting
from non-defaulting student borrowers. As suggested pre¬
viously, it appears that more financial data should be
collected and analyzed if one is to achieve a more accurate
prediction model for community college student loan de¬
faults. Although there are inherent differences between the
administrative mechanisms employed for the student loan
system and those utilized by commercial lending institutions,
information found useful in determining eligibility for
commercial loans may be useful in distinguishing student
loan defaulters from non-defaulters. Therefore, a second
implication based on the results of this study is that future
research should provide more focus on the financial data of
community college students in order to improve the chances
of formulating a better prediction model.
Limitations of the Model
The results of this study reveal two major limitations
of importance for future researchers.
First, the models presented in this study do not
accurately categorize nearly one-third of the sample cases.

82
As indicated, an inaccurate prediction regarding the proba¬
bility of default or repayment was made for nearly one-third
of the students in the sample. This imperfection results in
the limited application of the prediction models. Previous
researchers were unable to formulate a prediction model and
validate their models with any great success; similarly, this
study did not find a sufficiently accurate prediction model
for determining who will default on their student loan.
Second, since the models were not effective in making
an accurate prediction for all cases in the sample, it
appears that additional (especially financially related)
variables may be of value in future research studies. This
study was restricted to the use of data highlighted as sig¬
nificant in the literature and available through the Florida
Student Financial Aid Commission (FSFAC) in Tallahassee,
Florida. The results of this study and the previous studies
seem to indicate that some variables which were not avail¬
able through the FSFAC data files may be of critical impor¬
tance in formulating a more accurate prediction model. For
example, the employment status and type of job held at the
time of loan application or the family's financial data in¬
cluding loan repayment history may be reliable variables to
consider when predicting student loan defaults. In addition,
data concerning ethnic background, college major, and ad¬
justed gross income were found to be of value in prediction
formulae presented by university researchers. Although the

83
data are not currently available in the data system provided
through FSFAC, perhaps these variables are also important in
the formation of a prediction model concerning the probabil¬
ity of default for community college students. Any further
study regarding a prediction model for determining the prob¬
ability of default should also research the value of includ¬
ing these variables. Based on the difficulty of collecting
data concerning these variables, the inclusion of these
variables would alter considerably the methodology section
of any similar research study. However, the variables do
appear to be of value in examining critically the student
loan default problem and presenting a model that improves
predictive accuracy. Therefore, the resources employed to
collect the data would appear to be prudently utilized.
Recommendations
The use of any of these predictive models for deter¬
mining the likelihood of students defaulting on their stu¬
dent loans is limited. Therefore, the foremost recommenda¬
tion derived from the data and the literature search is the
need to encourage further research. The importance of this
study is that the data indicate it is possible to formulate
a model which will accurately predict 70% of the students
within the sample. It appears realistic to assume that
further research, especially the financially related
research suggested in the previous discussion, may result

84
in a more inclusive and accurate prediction model which would
be useful in determining the probability of default. Such
research would need to be more inclusive and therefore the
collection of the new data would demand much more time and
financial resources. However, the possibility of finding an
appropriate collection of variables and the usefulness of
the resultant prediction formula would indicate that student
loan default prediction is an appropriate area for further
study.
In the event that a subsequent study does provide a re¬
liable prediction model, it should be tested and validated
by the researcher using new cases derived from a different
sample. In addition, before the model is implemented on any
individual college campus, it should be tested and validated
by the financial aid officer for applicability on his/her
campus. Only when these validations are performed can there
be confidence in the predictive capability of the model, for
the results of this study and the research presented in the
literature clearly indicate that the models and appropriate
variables vary from institution to institution. This varia¬
tion is dependent upon the type of institution being re¬
searched. Another reason for establishing institutional
validation procedures is to determine if significant
differences occur among institutions of the same type. It
is possible that a single, reliable prediction model that is
derived from data representing a state-wide population for a

85
selected type of institution may have no validity or useful¬
ness on a single campus.
Because of the increased emphasis on student loan pro¬
grams as a major component of the student financial aid
delivery system, further research concerning student demo¬
graphics and the probability of default is of utmost impor¬
tance. Such importance is underscored by the discussions
presented in the literature search and the successful
results of this study.

BIBLIOGRAPHY
Ary, D., Jacobs, L. C., & Razavich, A. Introduction to
research in education (2nd ed.). New York: Holt,
Reinhart and Winston, 1979.
Balderston, F. E. The repayment period for loan financed
college education. New York: Ford Foundation, 1970.
(ERIC Document Reproduction Service No. ED 081 379)
Balkins, A. J. Withholding transcripts for non-payment
of educational debts: Before and after bankruptcy.
Williamette Law Review, 1979, 15^, 563-575.
Bar Admission—Default on student loan warrants denial
of admission to Minnesota bar. William Mitchell Law
Review, 1980, 6, 443-454.
Block, W. R. Other students need money: An approach to
the administration of revolving loan accounts.
Journal of Student Financial Aid, 1977, 7, 26-30.
Blocker, C. E. The reality of student finances. New
Directions for Community Colleges, 1974, 8, 124-128.
The Carnegie Council on Policy Studies in Higher Education.
Next steps for the 1980's in student financial aid.
San Francisco, CA: Jossey Bass Publications, 1979.
Change Magazine and Educational Change. Colleges and money
A faculty guide to academic economics’! Washington, DC
Author, 1976.
Chronicle of Higher Education, February 24, April 28,
December 1, 1982, February 2, 1983.
Collins, J., Maguire, J. J., & Turner, R. M. Unmet need:
How the gap is filled. Journal of Student Financial
Aid, 1979, 9, 4-15.
Crackdown begins on federal worker's student loan defaults.
Higher Education and National Affairs. December 10,
1982, p. 3.
Davis, 0. S., & Van Dusen, W. D. Guide to the literature
of student financial aid. Princeton, NJ: College
Entrance Examination Board, 1978.
86

87
Deitch, K. M. Pricing and financial aid in American
higher education: Some interactions. Aspen, CO:
Aspen Institute for Humanistic Studies, 1978.
(ERIC Document Reproduction Service No. ED 184 450)
Dyl, E. A., & McGann, A. F. Discriminant analysis of student
loan application. Journal of Student Financial Aid,
1977, 7, 35-40.
Emmert, M. A. National direct student loan default rates:
A measure of administrative quality or something else?
Journal of Student Financial Aid, 1978, 8, 43-47.
Ferguson, G. A. Statistical analysis in psychology and
education (3rd ed.). New York: McGraw-Hill Book
Company, 1971.
Hartman, R. The National Bank approach to solutions. In
L. Rice (ed.), Student loans: Problems and policy
alternatives. Princeton, NJ: College Entrance
Examination Board, 1978.
Hauptman, A. Student loan defaults: Toward a better
understanding of the problem. In L. Rice (ed.).
Student loans: Problem and policy alternatives.
Princeton, NJ: College Entrance Examination Board,
1978.
Hays, W. L. Statistics (3rd ed.). New York: Holt,
Reinhart and Winston, 1981.
Horch, D. H. A retrospective description of the national
direct student loan program: Administrative practices
and institutional default rates m 19/2-/3. Princeton,
NJ: Educational Testing Services, 1978. (ERIC Docu¬
ment Reproduction Service No. ED 165 617)
Hyde, W. D., Jr. State guarantee agencies and capital
availability for student loans. Papers in education
finances. Paper no. 22. Denver, CO: Education
Commission of the States, Education Finance Center,
1979.(ERIC Document Reproduction Service No. ED 176 648]
Jenkins, J. A. Whose fault is student loan default? Change,
November, 1978, pp. 44-45.
Johnstone, D. B. Student loans: Some practical radical
alternatives. In Symposium on Federal Student Loans.
Washington, DC: National Association of Student
Financial Aid Administrators and the American Council
on Education, 1978.

88
Kelly, M. , ed. Administrator. Madison, WI: Magna Publica¬
tions, January 7, 1983.
Kendis, K. L. Solving the student loan problem by supple¬
menting government loan policy--An expanded role for
institutional aid administrators. Student loans:
Alternatives for reauthorization. Washington, DC:
National Association of Student Financial Aid Adminis¬
trators, 1978 .
Leonard, K. J. Collection of federal student loans.
Columbia Journal of Law and Social Problems, 1980, 15,
317-358.
Martin, P. Student loans: Curing problems and insuring the
future. Capital University Law Review, 1980, 9, 579-
594.
Millsap, W. L., & Wright, P. S. Recent cases on student
transcript rights after bankruptcy. Journal of College
and University Law, 1979, 6, 231-240.
Myers, G., & Siera, S. Development and validation of dis¬
criminant analysis models for student loan defaultees
and non-defaultees. Journal of Student Financial Aid,
1980, 10, 9-17.
National Association of Independent Colleges and Universities,
An analysis of the tuition advance fund bill. Washington
DC: Author, 1978. (ERIC Document Reproduction Service
No. ED 171 189)
New York Times, October 12, 1975, Section 3, p. 16.
Ostar, A. W. The effects of inflation and pricing policies
on college enrollments. Washington, DC: American
Association of State Colleges and Universities, 1978.
(ERIC Document Reproduction Service No. ED 177 961)
Premises of federal student assistance. Washington, DC:
American Association for Higher Education, 1979.
(ERIC Document Reproduction Service No. ED 193 999)
Rudolf, F. Myths and realities of student aid. College
Board Review, 1962, Fall, pp. 18-23.
Sanford, T. R. The impact of student loans on graduate
education. In AIR forum 1980 paper. Atlanta, GA:
Association for Institutional Research, 1980.
(ERIC Document Reproduction Service No. ED 189 936)

89
Silber, J. R. The tuition advance fund: A proposal for
funding higher education. In paper presented at
Assembly of the National Center for Education Manage¬
ment SystemsT Boston, MA: National Center for
Education Management Systems, 1978. (ERIC Document
Reproduction Service No. ED 163 824)
Study of the requirements for forming state guarantee
agencies, Final report. New York: Research and Fore¬
casts, Inc., 1979. (ERIC Document Reproduction Service
No. ED 188 508)
Tate, J. L. Federal aid to postsecondary students. Journal
of Family Law, 1979, 18, 147-178.
U. S. Congress. House. Committee on Education and Labor.
Subcommittee on Postsecondary Education. Student loan
defaults. Hearings before a subcommittee of the Commit¬
tee on Education and Labor, House of Representatives,
95th Cong., 1st Sess., 1977. (ERIC Document Reproduc¬
tion Service No. ED 153 538)
U. S. Department of Health, Education, and Welfare. Office
of Education. Study of selected educational institu¬
tions and students participating in direct school
Tending under the federal student loan program, Final
report( Washington, DC: Office of Education, 1979.
(ERIC Document Reproduction Service No. ED 185 953)
U. S. General Accounting Office. Collection efforts not
keeping pace with growing number of defaulted student
loans. Washington, DC: Government Printing Office,
1977 .
U. S. General Accounting Office. Stronger action needed to
recover $730 million in defaulted national direct stu¬
dent loans. Washington, DC: Government Printing
Office, 1981.

BIOGRAPHICAL SKETCH
James A. Schmidt was born December 11, 1950, at Pitts¬
burg, Pennsylvania. He attended local schools in St.
Petersburg, Florida, and graduated from Dixie M. Hollins
High School in 1968. He entered the Florida State Univer¬
sity in 1968 and received the Bachelor of Science degree in
criminology in 1972 and the Master of Science degree in
education (student personnel/higher education) in 1973. He
served Southeastern Community College (N.C.) as Director of
Admissions and Student Activities and Researcher from 1973
until mid-1976. In the fall of 1976, he entered Stetson
University College of Law in St. Petersburg, Florida. The
following year, 1977, he accepted a counseling position with
Edison Community College in Fort Myers, Florida. During the
1981-82 academic year, he attended the University of Florida
while on leave from Edison Community College in order to
complete residency and course requirements for the Doctor
of Philosophy degree in educational administration (higher
education). Upon returning to Edison Community College in
July, 1982, he assumed his present position as Director of
Financial Aid.
James A. Schmidt is married to the former Linda Dale
Thompson of Callahan, Florida. They have two sons, Daniel
and Timothy. He is a member of Beta Theta Pi, Kappa Delta
90

91
Pi, Phi Delta Kappa, several professional and service organ¬
izations and has been selected as an Outstanding Young Man
of America for 1983.

I certify that I have read this study- and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Wattenbargea Chairman
sor of Educational
Administration and Supervision
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Administration and Supervision
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
I \ n \ <\ ^^ y V.'
Harold C. Riker
Professor of Counselor Education
This dissertation was submitted to the Graduate Faculty of
the Department of Educational Administration and Supervision
in the College of Education and to the Graduate Council, and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
Dean for Graduate Studies and
Research
August, 1983

UNIVERSITY OF FLORIDA
3 1262 08552 9427




PAGE 1

A PREDICTIVE MODEL FOR THE REPAYMENT OF STUDENT LOANS IN COMMUNITY COLLEGES BY JAMES A. SCHMIDT A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1983

PAGE 2

ACKNOWLEDGMENTS The writer wishes to acknowledge the assistance of the many persons who took an active interest in the preparation of this study. The guidance and assistance provided throughout the last two years by Dr. James L. Wattenbarger , chairman of the writer's supervisory committee, are deeply appreciated. In addition, sincere thanks are also extended to the faculty of the Department of Educational Administration and Supervision and in particular to the members of the committee, Dr. John M. Nickens and Dr. Harold C. Riker. For their assistance in collecting the data the writer is indebted to the Florida Student Financial Aid Commission. The technical assistance provided by Mr. Kenneth Goehle and Mrs. Betty Davis are also gratefully appreciated. Finally, the writer wishes to acknowledge the support provided by his family. For their encouragement, assistance and emphasis on the value of an education, the writer is deeply indebted to his parents, Dr. and Mrs. Paul H. Schmidt. To his wife, Linda, and his sons, Daniel and Timothy, the writer wishes to express his deepest gratitude for their love, patience and understanding during the preparation of this study.

PAGE 3

TABLE OF CONTENTS Page ACKNOWLEDGMENTS ii LIST OF TABLES v ABSTRACT v i CHAPTER I. INTRODUCTION 1 The Role of Financial Aid 2 The Growth and Importance of Loans 10 The Problem of This Study 19 Justification for the Study 20 Delimitations & Limitations 21 Research Design and Procedures 22 II. REVIEW OF RELATED LITERATURE 26 Problems Inherent in Student Loan Programs 31 Institutional Responses to Student Loan Defaults 38 Professional Organization Action Regarding Student Loan Defaults 40 Governmental Action Regarding Student Loan Defaults 42 Responses to Student Loan Default Problem by Education Spokesmen 4 4 Antecedent Studies Relating to Student Loans and Demographic Characteristics .... 48 Summary 53 III. DEVELOPMENT OF THE MODEL 55 Analysis of Variables 60 Development of the Model 69 Results 73 Summary 77

PAGE 4

Page IV. CONCLUSIONS, IMPLICATIONS AND RECOMMENDATIONS 7 9 Conclusions 79 Implications 80 Recommendations 8 3 BIBLIOGRAPHY 8 6 BIOGRAPHICAL SKETCH 9 IV

PAGE 5

LIST OF TABLES Table Page 1. Philanthropical Support for Higher Education 27 2. Overview of Major Variables Selected by Previous Researchers 56 3. Chi-Square Analysis of Relationship Between Age and Loan Payback 6 2 4. Chi-Square Analysis of Relationship Between Marital Status and Loan Payback 6 3 5. Chi-Square Analysis of Relationship Between Sex and Loan Payback 6 3 6. Chi-Square Analysis of Relationship Between College Standing and Loan Payback 6 4 7. Chi-Square Analysis of Relationship Between Grade Point Average and Loan Payback 6 5 8. Chi-Square Analysis of Relationship Between Loan Total and Loan Payback 6 6 9. Interaction Effects Between the Variables 68

PAGE 6

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfilment of the Requirements for the Degree of Doctor of Philosophy A PREDICTIVE MODEL FOR THE REPAYMENT OF STUDENT LOANS IN COMMUNITY COLLEGES By James A. Schmidt August, 198 3 Chairman: James L. Wattenbarger Major Department: Educational Administration and Supervision The problem of this study was to determine the relationship between selected variables which characterize community college students and student loan defaults and to develop a model using these variables to predict student loan payback. Given the current economic crisis and the increasing reliance on the student loan programs to help students meet educational expenses, a study of the importance of selected student demographic characteristics and their relationships to the student loan default problem is of great importance to the future support of the student loan programs. The literature provided a theoretical basis for this study including appropriate variables for study as predictors of student default. These variables included size of loan total, marital status, sex, grade point average, college standing, and age. The data presented in this study were supplied by the Florida Student Financial Aid Commission, Tallahassee,

PAGE 7

Florida, and represented a statewide sample of 7 6 community college students who have participated in the Guaranteed Student Loan program. Of the six variables selected, only the size of the loan total and marital status distinguished significantly those who repaid their student loans from those who did not. In addition to these variables, sex, grade point average, college standing, and age were useful in developing a prediction model. Although the model did not provide an infallible formula for predicting those students who are most likely to repay their student loans, the model predicted group membership (defaulter of non-defaulter) for 70% of the sample cases. These findings underscore Pattillo and Wiant's conclusion that items reflecting financial rather than biographical data appear to be better predictors of loan delinquency. Therefore, it appears that the inclusion of additional discriminating variables and a more detailed study design may be necessary in order to improve the identification of students who are likely to repay their student loans.

PAGE 8

CHAPTER I INTRODUCTION From modest beginnings over three hundred years ago the role and impact of financial assistance to students in American colleges have changed dramatically. Like the gift presented by Lady Anne Mowlsen of London to the deserving Harvard student, most early student financial aid awards were granted with money given to colleges by private individuals to aid worthy, needy students. Therefore, the original purpose of providing student financial aid was to expand educational opportunities to those students whose incomes presented a barrier to higher education. In fact, only during the period of time immediately following World War II was substantial aid awarded on the basis of merit alone. From the late 1940s through the early 1960s the Serviceman's Readjustment Act of 1944 (GI Bill) provided substantial amounts of money to returning servicemen for educational expenses regardless of individual financial need. Many colleges and universities found that the large amounts of scholarship funds, previously used to support needy students, were no longer needed and those institutions began to use the funds to attract and reward students

PAGE 9

with academic or other special talents, with little or no regard for financial need. Currently, the essential criterion of the major federal aid programs (Pell Grant, College Work Study, National Direct/Guaranteed Student Loan Programs) is the students' need for funds. Accordingly, the purpose of the principal financial aid programs of today is remarkably similar to the original intention of the early student aid programs. The return to a need-based student aid system has caused a great deal of consternation to financial aid administrators. Today it becomes increasingly difficult to reward students with merit-based aid when the available resources do not adequately take care of the needy students. In order to encourage both needy and worthy students, a balance between needbased and merit-based aid is critical to the development of an impartial and equitable student financial aid delivery system. However, most student financial aid administrators apparently feel that the current, primarily need-based, delivery system provides the best use of available resources for it attempts to extend the opportunity for higher education to as many worthy students as possible. The Role of Financial Aid Since 1972 there has been considerable discussion concerning the role of financial aid in American higher education. In nearly every deliberation much attention has been

PAGE 10

devoted to the related issues of access and tuition charges because they are so intertwined with the role of financial aid. As a nation, the United States has demonstrated its commitment to citizen access to postsecondary educational opportunities by instituting programs of assistance to needy students, establishing support programs for colleges and universities (i.e., Land Grant College Program) and developing a strong system of public colleges and universities. According to Ostar (1978) , 80% of all students attending higher educational institutions today are in programs at public institutions where the state partially subsidizes the institutional costs of the student. As a part of the public higher educational system, the public community college is an important factor in providing higher educational opportunities to American students. The impact of the community college system on the issue of access is highlighted in the state of Florida where more than 95% of the state population live within commuting distance of a community college program. Although the costs of attending a community college program may be lower because of public subsidies, the individuals attending the community college seem to have the greatest need for financial assistance. A 1972 American College Testing (ACT) Program Study indicated that the financial needs of community college students sharply

PAGE 11

exceeded those of the majority of students enrolled in other colleges and universities. The ACT study estimated that financially disadvantaged students constituted 25% of the total student population at community colleges and cited the following factors as contributing to this high percentage: (1) 70% of the students depend on the automobile to reach campus, (2) the higher age distribution (average age of student is 29) indicates the average student has assumed adult responsibilities, and (3) 66% of the students come from families with (1974) incomes below $10,000. (Blocker, 1974, p. 126) Therefore, it would also appear that the retention of low tuition policies is critical to maintain access to higher educational opportunities for a large number of students in the community college system. In fact, a University of Wisconsin study highlights the impact tuition policies have on the issue of access (Ostar, 1978). This study determined that for every 1% reduction in total cost of attending a University of Wisconsin Center participating in the study there was a corresponding 1.33% increase in enrollment. The study also attributed the increased enrollment to new students who were not enrolled in the University of Wisconsin system and concluded that many of these new students would not have been able to attend were it not for the lower tuition policy (Ostar, 1978) .

PAGE 12

Despite the establishment of the relationship between low tuition and access by this and other studies, much debate regarding federal aid centers on the strategy of providing individual student aid rather than institutional aid. It is feared that the reliance on the student-based aid delivery system may cause additional pressure on state governments and lead both public and private institutions to raise tuition fees (Ostar, 1978) . Although the underlying assumption of raising tuition fees is that the truly needy student will be taken care of, a University of Tennessee study indicated that a $100 increase in tuition would yield only a $26 increase in an individual student's Pell Grant Award (Ostar, 1978). Truthfully, every tuition increase falls hardest on the $13,000 $25,000 income family and yet this student is eligible for little federal aid in the form of grants or College Work Study (Ostar, 1978) . In addition, Ostar maintains that an overemphasis on student-based aid leads to a growing number of charges involving fraud, waste, and abuse and leads to ever-tightening regulations and increased numbers of bureaucrats, investigators and bill collectors. The previous discussion regarding access would indicate that the best approach regarding federal assistance may be the increasing of federal support to higher educational institutions enabling colleges and universities to retain lower tuition charges. However, because of economic and

PAGE 13

political factors the financial aid delivery system is oriented toward individual student awards. Given this understanding two major questions must be addressed: what aid is available to students and how effective is the delivery system? Student Aid Resources Essentially student financial aid is available in the form of grants, work-study opportunities and loans. A grant is gift money that is offered to those most needy in order to prevent over-indebtedness through borrowing. Work-study is a self-help program which provides meaningful, collegerelated job experiences and earnings enabling the student to earn money and meet current expenses from bi-weekly wages. Loans constitute a program of borrowing whereby students must repay the loan with interest from future earnings after leaving college. The federal programs representing these categories of aid include Pell (BEOG) and Supplemental Educational Opportunity Grants, College Work-Study, and National Direct and Guaranteed Student Loan Programs. Most of the federal programs have not seen substantial increases in federal money in recent years. However, the Pell Grant Program (BEOG) which was traditionally limited to low income students has been expanded to assist families from middle income levels. In addition, the amount of money expended and the numbers of students served through the

PAGE 14

Guaranteed Student Loan Program (GSL) have risen dramatically. During the 1982-83 fiscal year modest cuts were made in the financing of the campus-based programs (SEOG, CWSP and NDSL) . Although support for the Pell Grant Program was retained at previous levels, eligibility and dollar delivery regulations were instituted to reduce fraud, abuse and errors. The combination of reduced support for some programs and increased regulation of the Pell Grant Program created a dilemma for student financial aid administrators who could stretch local and institutional funds only so far. At the same time, colleges and universities are experiencing higher costs because of increases in major dollar operating items (i.e., salaries, energy, supplies) which seem to be severely affected by an inflationary economy. This increase in costs translates into higher tuition charges. Since the financial aid application requires the student's resources to be balanced with the expected expenses, many students are experiencing higher need figures as the expected expenses rise. Without corresponding increases in student aid to offset the rise in expenses, a "need gap" crisis occurs for the student. Although existing support for assistance programs is generally at the highest level ever recorded, the increasing cost factor is rising faster than increased federal support and therefore the number of "need gap" victims continues to grow.

PAGE 15

The Student Aid Delivery System Most of the blame for the "need gap" crisis must be put on increasing costs and expenses since the majority of the financial aid researchers conclude that the student aid delivery system appears to be quite effective (Deitch, 1978) . The financial aid delivery system in the United States is the mechanism used to provide individual students with the resources to attend postsecondary schools. Application for the Federal Pell Grant Program may be made separately on a free federal application; or, for a nominal charge, an indepth assessment of need may be presented along with the eligibility report for the Pell Grant Program. Although there has been some recent discussion of the effectiveness and equity of the need-analysis system, previous studies indicate the system is in pretty good shape. According to a study presented at the Annual Meeting of the American Association for Higher Education in March of 1979 the following conclusions were presented: (1) The need based system of financial aid has gone far in removing a meaningful element of price competition from the buying and selling of higher education, (2) the structure of the aid system (mix of parental contribution, need-based grant and self help) is meeting the public's perception of how postsecondary education should be financed, (3) the major goal of providing access is quite clear, and correspondingly the programs have been successfully reaching toward the goal (i.e., the number of minorities has increased significantly) ,

PAGE 16

(4) federal, state and institutional funds seem sufficient to insure that any qualified student could afford to attend most of the nation's institutions, and (5) the problem is not with the different kinds of aid available but for more efficient and equitable distribution of the funds that are available. ("Premises of Federal," 1979) According to Kenneth Dietch (1978) , the current system of student financial aid is a thoroughly developed, pervasive and well-functioning system of price discrimination. However, he cites several problems which are likely to arise with the current system including (1) competitive pricing (by institutions) , (2) lack of sensitivity to student costs, (3) loss of financial privacy of families of applicants, (4) efforts to depart from the need-based system to attract (non-needy, outstanding) students, and (5) need for fairness in the treatment of independent students. (Dietch, 1978) Regardless of the potential problems with the financial aid delivery system, its current status seems to be that it is well respected among professionals and researchers. Financial aid administrators and federal program personnel have struggled to solve the problems of "need-gap" by working together to tighten requirements and enforce regulations through extensive validation procedures.

PAGE 17

10 The Growth and Importance of Loans As other political priorities become apparent in Washington and free student aid programs (grants) are retained at previous funding levels, most aid administrators have increasingly used the available loan programs to meet student aid needs. In fact, loan programs have been increased substantially while most of the other aid programs have been limited or decreased in size. Therefore, the importance of an equitable loan system has far greater implications today than it did previously as more and more students use the device to solve the "need gap" crisis. General Definition and Philosophy of Student Loan Programs As indicated previously, loans are awards of money granted with the prior requirement that they subsequently be repaid, in whole or in part, with or without interest. According to Davis and Van Dusen (1978) , loans are the most attractive form of student aid to most policy-makers because they must be repaid and therefore are an investment rather than a direct expenditure of funds. Recent discussion of national student aid policies indicates that those in decision-making roles believe loan programs to be very effective. Therefore, the great increase in the use of the current loan programs and the addition of new loan programs are quite understandable.

PAGE 18

11 The underlying philosophical premise of a loan program is that the person who derives the benefit from an education should pay for its costs, without denying low income students access to postsecondary education due to lack of current resources (Davis & Van Dusen, 1978). Not only has such a philosophy been utilized to develop loan programs but it has also led to arguments concerning how much of the cost of an education should be reflected in the tuition charge. For example, in 197 3 the Committee for Economic Development recommended that tuition costs should be raised to 50% of the total educational cost within five years at colleges and universities and within 10 years at community colleges. Subsequently the issue was studied in detail by the Carnegie Commission. It suggested that tuition should be raised to 30% of the educational cost within 10 years at colleges and universities, but the Carnegie Commission exempted community colleges due to the historical concept of two years of free access to higher education. Many states have cited the Carnegie Commission report in introducing data supporting the raising of fees at low-cost public institutions. However, Howard Bowen (cited in Balderston, 1970) effectively argues that an investment in higher education is really an investment in the society. It is an investment in the nation's human capital leading to increased individual productivity and societal benefits (i.e., a more informed electorate, increased national productivity, lower

PAGE 19

12 unemployment rates, a populace more adaptable to technology, increased social consciousness) • Mr. Bowen concludes that typical students already bear 75 to 80% of the cost of their education in the form of tuition, books and foregone earnings. Many studies conclude that the heavy reliance on loan financing has the effect of shifting the eventual burden of payment for the cost of higher education toward students and their future income (Balderston, 1970). According to Mr. Bowen, the increased availability of loans is not an appropriate solution to the need gap crisis since students already bear such a large portion of the cost of their education. Mr. Bowen argues that since society receives significant benefits from educating students, it should bear an increasing portion of the cost of educating them. Those who support the argument that students should bear the cost of their education also favor the increasing use of loans to help students finance the increasing costs. However, the desirability of increasing the burden of cost on students by raising fees and then assisting the needy students through loan programs is an issue that will remain unresolved as both pro and con advocates have many arguments of merit. However, the current federal administration favors the use of loans to help needy students bear the increasing costs of higher education and therefore it has instituted new loan programs to help relieve the "need gap" crisis.

PAGE 20

13 Types of Loan Programs Among the original loan programs are the National Direct Student Loan (NDSL) Program (formerly the National Defense Student Loan Program) and the Guaranteed Student Loan (GSL) Program. The NDSL is a low-cost deferred repayment loan administered by the educational institution and funded 90% with federal monies and 10% with institutional monies. Under the NDSL Program there are no security or endorsement requirements and its distinctive feature is that in addition to a deferment of three years for public service, the student may arrange for cancellation of the loan based on military service in an area of hostility, full-time teaching service in selected schools (for five years) or full-time staff service in a head start program (for seven years) . The Guaranteed Student Loan Program is available in nearly all states where a state level guarantee agency has been developed. In the remaining states students may apply for a Federally Insured Student Loan (FISL) . Administratively, private parties and non-federal agencies loan their own capital to students in both programs. In the GSL program there is a central monitoring agency. While the federal government insures the loan and helps pay the interest, the agency acts as intermediary, overseer and collections processor. However, in the FISL program the

PAGE 21

14 lenders rely directly on the U. S. Department of Education for an interest supplement, insurance against default and a collection mechanism. Repayment regulations for both the GSL and FISL programs are similar to the NDSL; however, deferments of up to three years are only available for service in the armed forces, Action or the Peace Corps, and there are no provisions for cancellation. The Increasing Use of Loans The growth of both of these programs is indicative of the current emphasis on the issuance of loans to solve the student aid crisis. In 1977, Ernest L. Boyer (cited in Hauptman, 1978), then Commissioner of Education, estimated that between 1965 and 1977 six million students borrowed $11.1 billion. Commissioner Boyer stated that in 1975-76 alone $2.6 billion were loaned to students and that the national increase rate in the number of loans processed was averaging 5% each year. Mr. Boyer concluded that thousands of students have been served by the loan programs including many students who otherwise may have been denied a postsecondary education. In addition, Sanford (1980) , in his research of the impact of loans on graduate education, found that graduates with loans are slightly more likely to attend graduate or professional school than those without loans. Therefore, it appears that there is substantial value in supplying loans beyond merely providing access to undergraduate degree programs. The most

PAGE 22

15 recent data indicate a total of 2.75 million loans were processed through the GSL system alone during the 1981-82 academic year representing an investment of $6.1 billion ( The Chronicle of Higher Education , 1983) . Accordingly, the success of the federal loan programs seems to be the impetus for the development of new programs. For example, the Federal Government announced the creation of the PLUS/ALAS Loan program in 1980. This program is similar in administrative details to the other federal loan programs. The major difference between these programs and the original loan programs appears to be the lack of deferment and cancellation provisions and higher interest rates. It also should be noted that, in addition to the federal loan programs, many institutions, states and private foundations provide loan funds to students. Many of these programs pre-date the federal efforts, but the importance and reliance on loan programs are a result of the widespread application of the loan principles effected through the federal programs. Of the federal loan programs the GSL/FISL program is the most widespread. The reason is that the NDSL is institutionally administered and many institutions choose not to participate in the program. Other institutions have been restricted from participating in the NDSL program due to their high student default rates. Therefore, due to the universality of the GSL/FISL program, much of the data and research presented in the remainder of this study will focus on the GSL/FISL program.

PAGE 23

16 Program Problems and the Default Rate The federal loan programs have several disadvantages. According to Tate (1979) , abuses in loan programs occur at each of the three stages through which any student loan passes: (1) eligibility; (2) disbursement; and (3) collection. Both institutions and students are responsible for abuses apparent in the loan programs of today. Institutional abuses range from mismanagement to halfhearted collection efforts and student abuses include outright fraud in supplying eligibility data to shrewd bankruptcy declarations in order to discharge outstanding debts. The most obvious abuse regarding the loan system is the default or lack of repayment by student borrowers. By statute a default is defined as an overdue loan payment of 120 days (20 U.S.C. & 1080 (c) , 1976). However, the Department of Education does not recognize a default until the lender's obligation is suspended and the subrogation by the Federal Court takes effect. Regardless of the procedural issue concerning the commencement of the default status, the American public is constantly reminded of the thousands of students who have participated in the federal loan programs and who have decided not to repay their loan obligations. Headlines in papers consistently highlight the default rate and recent efforts to initiate novel collection methods such as impounding automobiles have received national television attention.

PAGE 24

17 The "default rate" is computed as a ratio of the cumulative amount of delinquent funds to total receivables and can be measured by the following formula: cumulative claims paid to lenders all loans having entered repayment status. Current researchers estimate the overall default rate to be around 12%; however, the rate is slightly lower in the GSL program because the financial institutions exercise more selective lending practices and the guarantee agencies exercise more administrative vigilance. There is much debate concerning the reliability of data regarding the default rate. Arthur Hauptman (1978) argues that loan default rates are not consistent due to several factors: (1) the measurement standards changed and the data prior to 1973 are not consistent with the data collected after 197 4; (2) the default rate is conceptually flawed since it (a) uses only the federal expenditures in the numerator (federal expenditures are only 80% of the costs) and (b) includes death and disability defaults; and (3) the default rate is cumulative; therefore, data for each year's statistics are not sorted out (not everyone defaults in the first year of repayment) . (Hauptman, 1978, p. 160) Hauptman concludes that a single measure of default is inadequate to explain levels and trends in the behavior of

PAGE 25

18 students who default because default activity differs by the type of guarantee (FISL or GSL) , the type of educational institution the borrower attends, the type of lender and the demographic characteristics of the borrowers. In addition to the internal problem concerning the reliability of the default rate statistics, there is much concern about the use of the default rate as a performance indicator regarding the success of the student loan programs. As cited by Hauptman, an Office of Education Report (1979) illustrates the three major weaknesses of this argument: (1) too many variables beyond the control of the Guarantee Agency have a significant impact on default rates (i.e., depression/recession, high unemployment rates, etc.); (2) loan default rates may only reflect the availability of loans (higher rate may be due to significantly more loans arranged) ; and (3) cumulative default rates are historical and may not reflect current management philosophies. (Hauptman, 1978, p. 162) In effect, two conclusions can be reached based on the previous discussion: (1) the tendency to default may have increased over the last several years but it may not be as high as it has been reported, and (2) a large number of variables which are external to the guarantee agency and the financial lender affect the default rate.

PAGE 26

19 The Problem of This Study Regardless of the elusiveness of concrete reliable statistics, the public perceives the default rate as too high. The federal government's determination to stem the default rate must include research on the socio-economic characteristics of student borrowers and defaulters in addition to data concerning administrative procedures. Accordingly this study has attempted to provide data concerning some of the variables that may be related to the student default rate. In addition, these variables may be useful in formulating a prediction model to determine who will repay their student loans. Such a model would be of valuable assistance to the student aid administrator in reducing the default rate because it would distinguish the characteristics of defaultees and non-def aultees. The aid administrator could then award and approve loans for only those students with the greatest likelihood of repayment. Since loan money may not be as readily available in the future as it is today, a model that could provide assistance in selecting the "best" student loan risk may be critical in maintaining the integrity of the loan programs and in replenishing the source of funds for future loans.

PAGE 27

20 Statement of the Problem The problem of this study was to determine the relative influence of selected variables which characterize community college students who are most likely to repay student loans. Questions Specifically, a primary question of this study was to determine the usefulness of the model derived from the data presented in this study to predict those students likely to repay student loans from those students who are not likely to repay student loans. Therefore, the following supplemental question was addressed: will the selected variables be useful in discriminating between students who will repay student loans and students who will not repay student loans? Justification for the Study As indicated previously, throughout the last few years concern regarding student loan default rates has risen tremendously. The issue has become the source of critical educational fiscal management decisions and delicate political debates. During the period from September, 1982, through February, 198 3, The Chronicle of Higher Education reported on 17 articles relating to recent developments concerning student loans. Thus, the selection of this issue is

PAGE 28

21 justified on its standing as a practical problem facing higher education administrators today. Discussion relative to some of the many student variables contributing to the default rate is imperative. This study analyzes six variables (age, marital status, sex, college standing, grade point average and loan total) identified in the previous research of Dyl and McGann (1977), Myers and Siera (1980), Emmert (1978) and other researchers for the impact these variables have on the student default rate. Delimitations and Limitations Since there are several procedural issues impacting the results of this study, the following delimitations and limitations are applicable. Delimitations (1) This study was confined to students who have received a loan through a program administered by a Florida public community college. (2) Only data made available to the researcher by the Student Financial Assistance Commission of the State of Florida were utilized in this study. (3) A sample approximating 31% of the total population was selected by a systematic random process for both categories of students identified in the analysis (i.e., students who repay student loans and students who do not repay student loans) . (4) The analysis of data was confined to the discriminant analysis procedure utilized by the computer

PAGE 29

22 techniques of the Statistical Analysis System (SAS) , which produced a set of equations from which a predictor model could be evolved. Limitations (1) This study was limited to those variables identified in the research design, specifically: college standing, age, marital status, sex, grade point average, and loan total. Research Design and Procedures This study was designed as an ex post facto research study in which several continuous and discrete independent variables obtained from data collected on loan application forms were observed. From these variables, the dependent variable of loan repayment was predicted. According to Ary, Jacobs and Razavich (1979) , ex post facto research can supply much information of value in educational decision making even though its design differs significantly from pure experimental research. In an ex post facto research design an experimenter can study two groups that are different in some respect in order to discover the reasons for the difference. Such is the case with this study. The design format as suggested by Ary et al. is as follows:

PAGE 30

23 Group Independent Variable (s) Dependent Variable E (X) (variable) y (Repaid Loan) c Y 2 (Did Not Repay Loan) The supposed effect of treatment occurs in the experimental group. Therefore, according to Ary et al. (1979), the researcher attempts to relate the dependent variable Y, to a previously occurring independent variable, a nonmanipulable variable indicated by (X) which occurred in the experimental group but not in the control group. The purpose of utilizing the ex post facto research design was to illustrate the significant data to be utilized in a model that could be useful in predicting the repayment rate probability of students requesting loans for educational use at community colleges. Such a model would be helpful to financial aid personnel in determining what student characteristics are representative of those students who have repaid their student loans. Sample The data for this study were selected randomly from the total population of students in each of the following two categories: (1) students who are repyaing their student loans and/or are not

PAGE 31

24 identified as defaulting on their student loan commitment, and (2) students who are not repaying their student loans and have been identified as defaulting on their student loan commitment. These categories were determined based on the data collection and storage procedures established by the Florida Student Financial Assistance Commission (FSFAC) in Tallahassee, Florida. The FSFAC is the central storage/retrieval center regarding GSL repayment data for all institutions in Florida. The sample size (76) represents approximately 31% of the total population in each category involving loans made to community college students in Florida. Data Collection Data on each student included in this study were supplied by the FSFAC to the researcher in June, 1982. During this session files were randomly selected by FSFAC personnel for review by the researcher. The appropriate data were then extracted from the files for those students who were identified as community college students. Since there are two separate files maintained by the FSFAC (i.e., those who have not defaulted, those who have defaulted) the researcher obtained a sample for each category representing 31% of the total population.

PAGE 33

CHAPTER II REVIEW OF RELATED LITERATURE The financial pressures exerted on colleges and universities have had a direct effect on the expansion of loan programs. As a result the loan programs are experiencing some unique problems. An analysis of the responses to the problems surfacing in the student loan programs and a review of the research regarding student loan default and student demographics provide an overview of the issues surrounding the problem presented in this study. Fiscal Pressures and the Demand For Expanded Student Loan Programs Higher education is a social institution that depends on tuition and service charges, philanthropy, and governmental support in order to balance its budget. The heavy reliance on these sources of support has created a dilemma concerning the funding of higher education during weak economic times. The current fiscal crisis can be readily confirmed by reviewing the funding sources and relative support for higher education since 1950. In terms of tuition and service charges, there has been a percentage decline in these fees as a source of income 26

PAGE 34

27 from 1950 to 1974 (Change Magazine, 1976). This decline has been attributed to the larger percentage of students attending public colleges in 1974 (76%) compared to 1950 (50%) and the corresponding lower tuition charges at these public institutions (Change Magazine, 1976) . In addition, colleges and universities have found themselves in competition with local facilities for the student dollar regarding auxiliary charges including food, housing and books. Therefore, there has been a decline in income from this revenue source as well. Although all service charges such as room and board accounted for approximately 4 0% of the total income in colleges and universities in 1974, Kelly (1983) estimated that tuition charges in 1977 accounted for only 11% of the total income for colleges and universities. As a result of present trends on college campuses, Kelly (1983) estimates that this percentage will increase to 20% by 1988. Based on this estimate it appears that students will be forced to bear an increased percentage of the cost of their education in the form of increased tuition charges. Correspondingly there is likely to be additional pressure on the student aid delivery system to compensate for this increased cost. However, many colleges and universities are fearful of an increased dependency on student fees. These colleges and universities are justifiably concerned that the resultant increased fees will add a new pressure to

PAGE 35

28 the university budget — declining enrollments and the loss of dollar income from student charges based on underenrollment . In terms of philanthropical support, higher education has been the beneficiary of increased dollar support from 1950 ($200 million) until 1974 ($2.4 billion) (Change Magazine, 1976). As indicated in Table 1, due to the tremendous rise in the cost of higher education, this increased dollar support represents an actual percentage decline from 1950 (8%) to 1974 (5%) in terms of total income for colleges and universities (Change Magazine, 1976) . Table 1 Philanthropical Support for Higher Education Year Total Dollar Percent of Income for Support Colleges & Universities 1950 $200 million 8% 1974 $2.4 billion 5% Since the dollar increase is not consistent with the needed percentage increase for institutional philanthropical support, more of the cost of higher education will have to be absorbed elsewhere. In the recent past, the government (federal, state, and local) has provided this increased support for higher

PAGE 36

29 education. In 1974 it was estimated that the government provided 54% of the total income to colleges and universities (Change Magazine, 1976). This increase in total income represents additional support for both private and public institutions as it includes research monies, building assistance, and student aid dollars. However today, governments whose support figures so largely in the financing of higher education are concentrating on other major concerns and priorities. Therefore, there has been a decline in the rate of income growth to higher education from the governments. For example, the federal government is under contrary pressures to hold down expenditures in student aid (which has risen 600% from 1967 to 1977) on the one hand and to respond to the continuing demands of middle income parents for selective relief (i.e., additional aid, tuition tax credits) on the other. In addition, many concerns outside of higher education are uppermost in the minds of our legislators. For example, state legislators are highly conscious of the "Proposition Thirteen" sentiments in their districts. Therefore, the new depression in higher education is generally not based upon declining support, but it is due instead to a decrease in the rate of increase of support for higher education. Nowhere is this more evident than in the dollar support for student aid during the last two fiscal years. Colleges and universities have received essentially the same amount of support in total dollars from the

PAGE 37

30 government, but as the rate of increase has dropped, increased pressure to raise fees and start fund raising activities has surfaced. Because the slowdown has not been uniform, some states and individual institutions have had to contend with declining income. As a solution to the pressure added by increased fees, the Carnegie Council on Policy Studies in Higher Education (1979) has recommended expanding explicit self-help components (work-study and loan opportunities) of the student aid delivery system and revising the loan system to make loans more easily attained and widely available with less risk of default (i.e., longer payback terms, loan consolidation procedures) . It appears that the Reagan administration has seriously considered the Carnegie recommendations since they have proposed a 60% increase in College Work-Study funds, a "selfhelp" requirement for Pell Grant eligibility and changes in the Needs Test for the Guaranteed Student Loan Program for fiscal 1984. In addition, Mr. Reagan has proposed continuing dollar support for the federal grant programs at the same level as the past two fiscal years. However, under his proposal, eligibility qualifications for individual students have been substantially altered. Therefore, it is safe to assume that as fees increase, endowment expansion decreases and federal grant money remains the same, an increasing number of students will seek to participate in the student loan program.

PAGE 38

31 A potential problem arising from the decision to rely on the student loan programs for the increased dollar support to higher education is that the private lenders simply may not have funds they want to lend to students (as other loans may be more financially advantageous) or there may not be enough loan money to meet recognized need. Further, the loan programs in effect today have many problems. Therefore, before considering an expanding role for student loans in the financing of higher education, it appears the government should consider the weaknesses of the current loan system. Problems Inherent in Student Loan Programs A closer examination of the student loan default problem reveals several contributing factors. Therefore, nearly all researchers agree that the student should shoulder only a part of the blame for the escalating default ratio. Foremost among the factors to consider is the history of incompetent management of loan programs. For years neither the lender nor the government was making any noticeable effort to collect many of the outstanding loans (Jenkins, 1978) . In addition, Jenkins noted that the Office of Education could not afford to install a computerized collection system until 1978 (Jenkins, 1978). Although governmental investments in the area of administrative strategies have resulted in much improvement in collection procedures, the previous administrative procedures account for a

PAGE 39

32 substantial part of the default problem attributed to the historically impacted default rate (Jenkins, 1978). A second factor to consider is the nature of student loans. Kendis (1978) states that even though student loans are not the same as consumer loans, the lack of rules concerning asset and income analysis, borrowing limits and collection procedures are "diametrically opposed" to methodologies long since proved successful in consumer credit (Kendis, 1978). Although Kendis feels that the rules should be somewhat different from the rules utilized in consumer credit situations, he states that the result is an "attitude problem" surrounding student loans. This "attitude problem" affects not only students but lenders and legislators as well. Students borrowing as a part of their investment in their own human capital are forced into a decision process to attend or not to attend a college or university, wherein the negative consequences of increasing costs, and debt, are minimized by the ability to borrow funds. Lenders continue to lend greater volumes of funds leading to a greater median debt loan (with increasing financial concern) in the interest of assuring access and individual collegiate choice. And finally, the Government in its effort to assume capital availability guarantees loans by merely increasing existing programs while disregarding the long-term costs and the social consequences of the program. Any one of the conditions highlighted previously would have

PAGE 40

33 significant impact on the student loan problem; however, in combination there is no doubt that the "attitude problem" suggested by Klendis is a very real factor affecting student loan defaults. A third major factor affecting the student loan problem is the mechanism for repayment. Balderston (1970) notes in an analysis of incomes between those students choosing to go to college and those choosing not to go to college that the former will not "catch up to" the latter in real dollars realized until very late in the working life of the individual primarily due to the foregone earnings (lack of income during college years) of the college student (Balderston, 1970) . In fact, in some instances the college-going student may never catch up to the non-college student whether the dollars are discounted for inflation or not. Therefore, any student researching the wisdom of the investment of a college education from a purely financial perspective may decide it is not worthwhile to attend college based on the market payoff (Balderston, 1970) . Beyond the decision the student must make regarding the wisdom of attending college, if a loan is utilized, then the repayment mechanism that currently operates which seeks to concentrate the repayment in the early years of one's working life (10 years) may have some deterrent effect upon college attendance. If loan financing is heavily relied upon, the effect may be substantial. Balterston's

PAGE 41

34 conclusion is that short repayment periods lead to fewer students opting to attend college if loans must be utilized because they impose too heavy a burden of cash outflow on the student. He also notes the effect is stronger on students who do not finish programs or finish programs with less optimistic forecasts of projected future incomes (i.e., humanities, home economics) because these students do not have the advantage of cashing in on a completed or marketable degree and the loan payment constitutes a higher percentage of their total income than it does for students with higher incomes. The solution proposed by Balderston is to make long term loans and to tie the repayment mechanism into the income tax system or set up a similar system which reduces administrative costs (Balderston, 1970) . Although there is much debate as to the wisdom of Balderston' s proposal, especially as it relates to the use of the income tax system, there appears to be little doubt that the repayment system for current loans is a component of the overall problem regarding loan defaults. Many other factors are cited in the literature as contributing to the student loan problem. For example, Jenkins notes that there is such a "smorgasbord" of student assistance programs that it is possible for one student to be eligible for eight different loan programs, sponsored by eight different banks or lending institutions to pay eight

PAGE 42

35 different terms with eight different sets of conditions for forgiveness and eight different groups to deal with (Jenkins, 1978) . Some consolidation appears to be necessary. Another factor involves the apparent lack of sensitivity to issues of student loans from student financial aid administrators. For example, the student aid packaging process usually relegates the discussion of the terms of a student loan to the exit interview. Certainly this factor is within the control of the student aid offices. Hopefully, these offices will incorporate the suggestions made in the National Association of College and University Business Officers (NACUBO) manual in order to manage better the loan programs. Finally, a legal process has been cited as a major factor contributing to the student loan problem. The availbility of a bankruptcy action to discharge a student loan debt has become increasingly apparent to many student borrowers. The use of a bankruptcy action, its effect on the system of student aid and the governmental response to its use are important considerations in student loan research because they demonstrate an area where judicial and legislative cooperation has assisted in reducing the number of the student loan bankruptcies and encouraging more students to repay their student loans. a

PAGE 43

36 Bankruptcy Actions Bankruptcy is the mechanism by which insolvent debtors may be released from the obligation to repay their debts. The proceedings are handled by a federal bankruptcy court and conducted under laws enacted by Congress. In the mid1970s the problem of student loan bankruptcies received significant attention from Congress, colleges and universities, professional organizations and the courts and governmental agencies. This attention was not unwarranted as the problem became quite prominent. During the five fiscal years from 1966 through 1970 only 348 bankruptcies were discharged on NDSL and GSL loans totaling $400,000. However, in the fiscal year 1975 alone, a total of 4,559 bankruptcies were discharged on loans totaling $6,800,000 (Leonard, 1980). The effect of the tremendous rise in the use of bankruptcy actions to discharge student loan debts resulted in a great deal of negative publicity for all student aid programs. Prior to the disclosures regarding the use of bankruptcy actions, support for all federal aid programs was exceedingly strong and resulted in the passage of the BEOG (Pell Grant) Program. After considerable press exposure (i.e., New York Times ) public resentment led to a hearing by the House Subcommittee on Postsecondary Education in October, 1975. The resentment was specifically aimed at a very few of the 12 to 15% defaulting student borrowers, since very few

PAGE 44

37 students default through bankruptcy. Most defaulting students choose not to provide forwarding addresses and thus are "lost." According to some researchers, the small number of students abusing the bankruptcy process solely to escape federal educational debts is not a serious threat to the loan programs. However, the anger surfacing at the October committee hearing appeared to be generated by the apparent misuse of a constitutionally sanctioned remedy to escape a moral obligation (Tate, 1979) . Consequently, in 1976, Congress passed the Education Amendments of 1976 which changed the procedures utilized to discharge student loan debts by setting out specific restrictions. The primary restriction was a mandatory five-year delay from the commencement of the repayment period in the filing of a bankruptcy action. Although a subprovision allows for actions based on undue hardship, the purpose of the amendment is to enforce a moratorium on all but the most necessary bankruptcy discharges until sufficient time has passed for a loan holder to either pay off the debt or have enough at stake to make bankruptcy a genuine last resort effort (Tate, 1979) . This provision has since been incorporated into the Bankruptcy Reform Act of 1978. The Congressional action highlighted above indicates that student loan problems can be adequately addressed and resolved. However, some writers feel that the most potent weapon against bankruptcy abuse and student loan defaults

PAGE 45

38 may rest in the college registrar's office. Therefore, actions taken by institutions as well as professional organizations, the government and professional educators are worth researching as possible weapons in combatting the high student loan default rate. Institutional Responses to Student Loan Defaults Nearly every college has a statement in its catalog indicating that the college reserves the right to withhold grades, degrees, statements of honorable dismissal or a transcript of credits until satisfactory settlement of college fees and other financial obligations has been made. This strategy has worked well for many years at several institutions and is probably partly responsible for keeping the default rate within its current level. The appropriateness and legality of utilizing this strategy have been questioned in several court cases; however, the courts have not provided a definitive response to guide colleges and universities especially if the student has discharged the debt by bankruptcy action. It appears that in situations where the debt was discharged prior to October 1, 1979, the court will require state institutions to release the transcript (Handsome v. Rutgers University, 445 F. Supp. 1362). However, private institutions may have the authority to withhold transcripts of students whose debts were discharged prior to October 1, 1979 (Girardier v.

PAGE 46

39 Webster College, 563 F. 2d 1267). Due to the change in the bankruptcy code, students are not now likely to have their educational debts discharged, but if they do, it seems that the college cannot take any action (including withholding transcripts) to collect the debt. The Handsome and Girardier cases point out that even if the debt to the college has been discharged by a bankruptcy decree there are other legal obstacles to the practice of withholding transcripts. A future court may decree that the student has sufficient property interest in the transcript to require the college to provide adequate notice prior to withholding the transcript. Or a court may determine that there exists between the student and the college a contract and that the college must demonstrate that the contract was entered into with adequate notice regarding the withholding of the transcript. In either case, additional responsibility may be placed upon the colleges utilizing this strategy to compel students to pay their educational debts. In addition, courts have indicated that an unpaid debt must be one that is sufficiently related to the student's acquiring an education to justify retention of the transcript. This decision presents several new issues. For example, is a student's unpaid bookstore bill or a student parking fine "sufficiently related" to permit the college to withhold the student's transcript?

PAGE 47

40 A final consideration regarding the withholding of transcripts rests with the Family Educational Rights and Privacy Act (Public Law 93-380) . This act requires an educational institution to allow the inspection of a transcript, Although this act does not require a college to release an official transcript, a court may rule in the future that the institution must release an unofficial copy of a transcript. Obviously the strategy of withholding transcripts has been viable throughout the years, but educational institutions should consider the legal ramifications of their policies in light of the recent court decisions and the questions presented in the court dicta. Professional Organization Action Regarding Student Loan Defaults Beyond the strategies utilized by colleges and universities to encourage student borrowers to repay their educational loans, a few professions have begun policing their members regarding loan defaults. The most notable example is the case of the Florida Board of Bar Examiners re G.W.L. (364 So 2d 454, Fla. 1978). In this case the Florida Supreme Court refused admission to the state bar to the petitioner, who had discharged his student loans in bankruptcy, on the premise that the petitioner's financial

PAGE 48

41 status and apparent motivation at the time of filing reflected a sufficient lack of the proper moral character and requisite values to practice law in the State of Florida (Martin, 1980) . Since the court never questioned the act of filing for bankruptcy but rather based its decision on the student's conduct and the apparent breach of the Florida Character Standards for bar admission (which were established to protect the public interest) , the decision did not conflict with federal bankruptcy law regarding unlawful penalties imposed on bankruptcy petitioners. This court decision was the first to link discharge in bankruptcy to moral turpitude and it has been cited in cases in other states with similar provisions for professional licensing. The purpose of the standards as stated by the Minnesota Supreme Court is not to punish the attorney but to guard the administration of justice and to protect the general public (in re Peterson, 274 N.W. 2d 922, 925, Minn. 1979) . Therefore, the legal profession has set a precedent by refusing bar admission to those who have utilized the student loan system and irresponsibly neglect their debt obligation. Since this strategy appears to be useful for professional student bankruptcy actions and defaults, other professional organizations, for example the AMA or ADA, should institute and enforce similar provisions for licensing. This action is particularly relevant since the

PAGE 49

42 pursuit of higher education is valued so greatly by the American public in whose interest the professional serves and any interference with the financial aid system supporting the program of higher education in America should not be treated lightly. Governmental Action Regarding Student Loan Defaults Many governmental organizations and agencies have demonstrated concern for the student loan default problem and have initiated reports and actions to help remedy the problem. In addition, the U. S. Congress has been involved with legislation designed to curtail the default rate among federal employees and in providing a forum in which to discuss possible solutions for the loan default problem. The General Accounting Office (GAO) has published several reports concerning the student loan system. Some of the suggestions listed in its 1977 report included the following: (1) require financial statements of all debtors, (2) refer all defaults immediately to legal action, (3) set up guidelines for compromise settlements, (4) institute a payroll deduction system for all federal employees, and (5) require any contractor for collection service to document unsuccessful efforts. (U.S. General Accounting Office, 1977)

PAGE 50

43 Although these suggestions seem realistic, they have met resistance from several sources. For example, in 1979 the U. S. Department of Health, Education and Welfare (now the U. S. Department of Education) argued successfully that only it should negotiate settlements of compromise. In a subsequent (1981) report the United States Comptroller General indicated that many schools visited by the GAO made inadequate efforts to collect defaulted loans, that the schools needed to adopt a tougher attitude toward collecting defaulted loans and that the Department of Education should take stronger actions against those institutions that failed to do so. Consequently, in April of 1982, the Department of Education notified colleges with a poor record of loan collection that they would not receive any new federal loan money appropriated for the 1982-83 year. Although these actions referred specifically to the NDSL program, they sufficiently demonstrate the interest in Washington regarding the student loan problem. During the Oversight Hearing before the Subcommittee on Postsecondary Education of the Committee on Education and Labor (House of Representatives) the following methods were suggested as appropriate to cut down the GSL default rate: (1) discharging federal employees who default, (2) extending deferments to one year for those who are unemployed,

PAGE 51

44 (3) permitting lower rates of repayment for those with low incomes, and (4) prohibiting defaulters from obtaining other federal loans. (U, S. Congress, 1977) In response to item number one above, Education Secretary Terrence Bell and Senator Charles Percy (R-Ill.) sponsored legislation in 1982 enabling the government to garnishee the wages of federal employees who have defaulted on their student loans. Recently, the Department of Education sent letters to federal employees notifying them of the Debt Collection Act of 1982 and indicating they must make arrangements for the repayment of their loan by February 1, 1983, or risk the garnishment of their wages. Therefore, it is apparent that both the Executive and Legislative branches of the U. S. Government have also taken actions to assist in the solution of the loan problem. Responses to the Student Loan Default Problem by Education Spokesmen Professional educators have also addressed the problem of student loans and are often quoted in the literature of higher education. For example, Robert Hartman identifies several factors that impinge on the effectiveness of the current student loan system including the variety of loan programs and regulations, the use of bankruptcy actions to discharge debts and the differences among students and

PAGE 52

45 their respective needs. He recommends several restrictions to the current loan program. Among his suggestions are: (1) the development of a single loan instrument, (2) curtailing the use of bankruptcy actions (at least for the first 10 years rather than the five years provided in the Education Amendments of 1976) , and (3) a provision allowing different loan amounts for students of different needs, and extended payment and variable repayment options. (Hartman, 1978, p. 96) Hartman argues that his suggestions would increase the cost effectiveness of the loan program (items one and two) , increase the manageability of loan repayment procedures (items one and three) , widen student accessibility (items one and three) , and reduce defaults and delinquency (items one and three) . Essentially he is suggesting that the loans be conformed to fit the student, that the loan programs be supported identically by each of the fifty states and that the performance requirements and benefits of the loan programs be identical regardless of the program utilized. Hartman' s suggested actions seem to be consistent with the goals of the student aid delivery system and show great insight concerning the many elements constituting the student loan problem. Several other plans have been suggested by professional educators regarding the student loan program. A panel of

PAGE 53

46 advisors during the Kennedy Administration developed the idea of an Educational Opportunity Bank. The proposal would have allowed students to borrow money to cover their expenses at whichever college they decided to attend. The loan would be paid back over the next 30 to 40 years and an interest rate of 1% of the students' gross income would be applied. Although this program would have allowed students to pay their own college expenses, Duke University and Yale University encountered several problems when instituting similar programs including administrative and collection difficulties. D. Bruce Johnstone, in his proposal of the National Student Loan Bank, recommends that one new major governmental agency should assume the responsibility for all student loans (Johnstone, 1978) . In addition, Johnstone suggests that the colleges and universities (rather than the lending institutions) should originate all loans and that the repayment of interest should be tax deductible necessitating a tie-in to the Internal Revenue Service. This radical plan appears to be quite practical; however, it tends to come into conflict with many existing offices, agencies and banks as it would drastically change and/or perhaps eliminate their participation in the student loan system. Therefore, support for this proposal has never materialized. Most attention today regarding changes in the student loan programs has focused on John Silber's Tuition Advance

PAGE 54

47 Fund (TAF) proposal. The proposal incorporates several key factors including (1) loans would only be negotiated for successful students (after their freshman year) , (2) individual student TAF accounts would have to be established with the respective college business office, (3) involvement of the Internal Revenue Service would be required in the transaction and repayment procedures, and (4) indefinite repayment periods would be instituted since the repayment provisions require students to repay their loan at the rate of 2% of their annual gross income each year until they have paid back 150% of their loan. (Silber, 1978, p. 7) To its credit the TAF proposal provides some novel ideas such as treating loan payments as tax obligations and pooling the income of married couples with loans for the purpose of calculating their taxable income. However, the plan has come under negative scrutiny as many have argued that the proposal assumes that the burden of paying for an education should rest with the student. Beyond the previous arguments concerning who should bear the cost of an education, many professionals fear this approach may lead parents, private donors and the taxpayer to abandon their support for higher education which would force an even greater burden on the student.

PAGE 55

48 In addition, although research studies show that a long repayment period provides the best opportunity for a reasonable return incentive for investing in higher education, many bankers indicate that the TAF could never be selfsustaining based on the proposed 2% interest rate over an estimated 30 year repayment period. Other arguments against the TAF include the feeling that any plan that provides for the paying of higher education through long term student debts will result in limiting equal access and opportunity for higher education. Further, the tuition and enrollment controls suggested to distribute TAF funds to institutions would cause widespread disruptions in the higher education system. For example, the expanded eligibility requirements for private college and university students would enable them to be eligible for a larger proportionate share of the available money. Regardless of one's point of view, the TAF represents a proposal which has provided healthy discussions of the student loan system and the options available to solve the problems. Antecedent Studies Relating to Student Loans and Demographic Characteristics It is apparent that all the participants in the student aid delivery system are concerned about the student default rate on educational loans. The changes implemented thus far in the student loan system have been instrumental in improving administrative procedures. For example, new and better

PAGE 56

49 resources for collecting loans and monitoring repayment progress have been devised. A second approach to solving the student loan problem should consist of researchers studying the demographic characteristics of student borrowers in order to distinguish those students who are likely to repay student loans from those students who are not likely to repay student loans. Thus far only a few researchers have explored the characteristics of student borrowers in the last five years . In 1977, Dyl and McGann applied a multivariate discriminant analysis to the problem of identifying defaulting and non-defaulting student borrowers at the University of Wyoming. They analyzed 33 different variables including class standing, college attended, grade point average, amount of the loan, academic major, marital status, living quarters, total amount of university loans, and size of the monthly payments. Four of the significant discriminators displayed direct positive relationships with actual loan repayment behavior. The students who were more likely to repay their student loans included those students who (1) had higher grade point averages, (2) were married, (3) were engineering majors, and (4) chose high monthly payments. Dyl and McGann also discovered three factors which were negatively associated with repayment of student loans including

PAGE 57

50 the (large) size of the loan, the residence of the student in an apartment and the total amount of indebtedness to the university (including previous loans) . As cited by Davis and Van Dusen (1978) , this study, in concert with the earlier studies of Pattillo and Wiant (1972) and Spencer (1974) , was instrumental in identifying for other researchers the factors most worthy of analysis and providing a format in which to research the characteristics of student borrowers. Hauptman, in his 1978 study for the College Scholarship Service, explored many of the same characteristics for students participating in the GSL/FISL programs. One of the most interesting findings of his study was that the following groups of students had a higher percentage of defaults than would be expected: (1) students with low family incomes, (2) female students, (3) married students, (4) black students, and (5) junior college students. Hauptman discovered that students attending public schools had a lower default rate than those attending either private or proprietary schools and that college and university students had a lower default rate than junior college students or vocational students. Although the results of this study did not duplicate the results presented by Dyl and

PAGE 58

51 McGann (1977) , it is equally important because it researched a variety of institutions and concentrated its data on one loan program. Also in 1978, Emmert studied the characteristics of students defaulting on their NDSL loans and concluded that the demographic make-up of student populations has an equal or greater impact upon default rates than do internal administrative factors. He suggested that schools with higher default rates are not necessarily administering their programs any less effectively. Instead he argued that they may have students from substantially different backgrounds that account for the different default rate. Emmert also cited all the previous researchers in synthesizing the crucial variables selected in his study. Among the variables selected were: (1) age, (2) marital status, (3) sex, (4) college standing, (5) grade point average, (6) loan total, and (7) employment status. Like the Hauptman study (1978) , the differing results Emmert discovered seem to suggest that student demographic characteristics should be weighted in order to establish variable default rates for the different postsecondary institutions.

PAGE 59

52 In 1979 the Office of Education completed its first study of student borrower characteristics. Although the data collection was restricted to eight proprietary institutions, the study singled out several variables that may impact the default rate. Essentially this study underscores the necessity of establishing statistical validity by researching individual types of institutions independently since the results vary for the separate types of institutions studied. Therefore, because of the unique demographic characteristics of the proprietary student, this study concluded that the higher default rate among proprietary institutions discovered in the Hauptman study is understandable. Finally, a study completed by Myers and Siera in 1980 concentrated on data available from the New Mexico Student Loan Program for the New Mexico State University. Using the Statistical Package for the Social Sciences (SPSS) Discriminant Program, Myers and Siera attempted to develop a prediction model to discern defaulters from non-defaulters. The results indicated that the variables selected did not lead to an accurate prediction of the likelihood of a student defaulting on a loan. What prediction was possible was not substantially different from what one might expect to accomplish based on chance alone. Although the study did not present an acceptable prediction model, the work of Myers and Siera was instrumental in pointing out the need

PAGE 60

53 to explore and refine the application of predictive models to the student loan problem. Summary All of the personnel, institutions, organizations and governmental agencies involved with student loan programs have provided input and taken selective action to reduce the student loan default rate. A majority of the strategies implemented thus far have been concerned with the administrative or judicial (bankruptcy) aspects of the problem. In order to deal effectively with the problem of defaults the studies cited demonstrate the necessity of identifying the characteristics of defaultees and non-defaultees. Several of the studies presented emphasized the crucial demographic characteristics worth researching. Other studies demonstrate the necessity of restricting the population researched due to the inherent differences among the students choosing to attend the different institutions of higher education. Therefore, this study has restricted its data base to community college students in the state of Florida and researched the available data that corresponded to the variables highlighted in the previous studies. The student loan problem is not the creation of a single causality but rather it is the result of the

PAGE 61

54 attitudes, procedures, economics and actions of all of the participants (students, lenders, aid administrators) in the program. Certainly, additional study needs to be performed regarding the administrative problems within the student loan program. However, research regarding the students who partake in the program has long been relegated to a subsidiary status. It appears crucial that additional research needs to be conducted to identify those variables which are effective in predicting the likelihood of a student default, because the student is indeed the major factor contributing to the student loan default problem.

PAGE 62

CHAPTER III DEVELOPMENT OF THE MODEL The research reviewed in Chapter II provided an overview of the demographic variables to consider in the formation of a prediction model for student loan repayment. Names of the major contributors to the literature and the primary variables they studied are shown in Table 2. As presented in Chapter II many of these studies provided relevant findings concerning relationships between demographic characteristics of student borrowers and loan default. These relationships will be reviewed as a basis for developing the prediction model. In one of the first studies researching demographics and student loans, Pattillo and Wiant (1972) as cited by Davis and Van Dusen (1978), discovered that students who borrowed late in their academic years, who had previous loans, or who came from large families were more likely to be delinquent in their student loan payments (Davis & Van Dusen, 1978) . In addition, Pattillo and Wiant concluded that items reflecting financial rather than biographical data appeared to be better predictors of loan delinquency (Davis & Van Dusen, 1978). Spencer completed a study of the relationship between demographics and student loan default in 1974. In this study, 55

PAGE 63

56 CM M O u CD to xx CD o H Sh X! 03 rd CD H to !h CD ccS Pi > to O CD -r-i-H rH (0 > Xi £ CU rd Sh H H-i Oh o 5 X! CD •H TH. > QJ U 4-> U > CD O .H CD CO -— rd O 10 Jh cxi u CD a\ cd -H ~£ l3 4-1 C O O «-^ -H cn cu -P ru re! m-riu rH m 3 o w c rd g -p rrj IB rrd CTl rH O rH >1 U — Q £ — >H ^r CD rU rH cu — Oj CO o ^r-t 4-1 (NH C r-H rrj en +J -H rH +J JS — rd inVD^LnrOLnr^CNrHCNrHrOCNrHrHCNlrslCNlrHCNl XXXXXXXXXX X XX S/« ^ k>4 rN rN kS S^ »v^ *v* *v* K^ K^* r'N rS i^N kn i^n kn X X *v* *V* 'v* S/* *S*^ S^ N^ S/ 1 r^S rN fN f^N rs cN *-N kS Kj4 K>* VJ Kj* kS i^S rS rS X X X X

PAGE 64

57 according to Emmert (1978) , Spencer described the worst possible student loan risk as an unmarried, unemployed male student, in his last semester, with an old car, without a phone, and with a large outstanding loan debt (Emmert, 1978) . The results that Dyl and McGann obtained in 1977 indicated that the following factors related positively to the repayment of student loans: college major (engineering), marital status (married) , grade point average (high) , and the size of loan payments (low). Dyl and McGann (1977) also found the following three factors to be negatively associated with repayment: residence type (apartment) , total university indebtedness, and loan indebtedness (high loan totals). In this study Dyl and McGann also presented the results of a discriminant analysis of their data, and they were able to correctly classify 84% of the cases in their study. However, as pointed out by Myers and Siera (1980) , Dyl and McGann did not test their model with any new cases. Therefore, the reliability of the model presented by Dyl and McGann is, as yet, untested. Essentially these three studies presented the basis upon which most of the research in this area has concentrated. These researchers have not only provided insight concerning the appropriate variables to research but they also have suggested the appropriate statistical designs for future studies. However, the conclusion originally presented by Pattillo and Wiant (1972) regarding the inappropriate use

PAGE 65

58 of biographical data for predictive purposes has not been heeded. Many researchers have continued to search for student demographic characteristics which would distinguish student loan defaulters from non-defaulters. For example, in 1978 Hauptman studied five demographic variables and reported that the worst risk student loan participant was a married, Black, female student from a low family income who attended a proprietary vocational school. In another study presented in 1978, Emmert summarized the previous studies reported in the literature. In this article Emmert reported the variables that have been studied and the conclusions that have been presented concerning the effect these variables have had on the student loan default problem. Emmert also noted the different default rates among postsecondary institutions and suggested that the demographic make-up of a given student population may place an institution in a high risk category for student loan defaults. In conclusion, Emmert suggested that the demographic factors should be weighted in order to establish variable default rates for differing types of institutions. A 1979 study prepared by the Office of Education also reported student demographic data in relation to the problem of student loan defaults. Although the study primarily was designed to research the administrative procedures utilized by selected proprietary institutions, it did show that many

PAGE 66

59 but not all of the same variables selected by other researchers were important variables concerning the student loan default problem in proprietary institutions. The findings presented in this study demonstrate the necessity of studying individually the different types of postsecondary institutions. The most recent study in this area of financial aid research was carried out in 1980 by Myers and Siera. Through t-test analysis, Myers and Siera found the following variables to be highly associated with the repayment or nonrepayment of student loans: college standing, loan total, amount of loan requested, grade point average, and degree completion. Myers and Siera also attempted to replicate the study design presented by Dyl and McGann (1977) ; and, although they were able to formulate a promising prediction model, when this model was tested using new cases only 42.5% of the new cases were correctly predicted. The validity problem experienced by Myers and Siera in their prediction model illustrates the difficulty researchers have encountered in their search for an accurate, useful prediction model for student loan defaults. It is apparent from the above discussion that although the conclusions of the previous studies differ considerably, there does appear to be some consensus regarding the importance of researching student demographics and which demographics should be studied. In selecting which variables

PAGE 67

60 to use in this study, the researcher selected the variables which had been most frequently used by previous researchers (see Table 2) . Although ethnic status and family income were studied by several researchers, these variables were not included in this study because of the difficulty in collecting and verifying these variables. The remaining six variables were available in the data bank of the Florida Student Financial Aid Commission in Tallahassee, Florida. As a result, the following six variables were selected for the prediction model presented in this study: (1) age, (2) marital status, (3) sex, (4) college standing/ (5) grade point average, and (6) loan total. Analysis of Variables As indicated in Chapter I, this ex post facto study was designed to determine the usefulness of a model to distinguish those students likely to repay student loans from those students who are not likely to repay student loans. Prior to developing any model, there was one major issue to address: were the selected variables useful in discriminating between students who will repay student loans and students who will not repay student loans?

PAGE 68

61 In order to determine the usefulness of the variables in discriminating those students who will repay student loans from those students who will not, data representing each of the six variables were collected for both defaulting and nondefaulting students in the sample. Since three of the variables were nominal in nature, the t-test statistic utilized by previous researchers to determine significant differences between defaulters and non-defaulters was not applicable to the data. Therefore, a chi-square analysis was performed on the data collected for each of the variables in this study to determine if the differences between the proportions of subjects that fell into the two different categories (defaulters and non-defaulters) were significant. Since this study represents exploratory research and any model developed from the data will need to be validated with new data, the researcher selected a .10 significance level. Essentially the chi-square analysis determines if the difference between the expected and observed frequencies for each of the classifications within each variable is greater than the difference one would expect to find based on chance alone. The results of the chi-square analyses showing the relationship of each of the six variables to student loan default are presented in Table 3 through Table 9.

PAGE 69

62 Table 3 Chi-Square Analysis of the Relationship Between Age and Student Loan Default A 9 e Defaulters Non-Defaulters 18-20 Number 8 % of total frequency 10.53 21-24 Number 10 % of total frequency 13.16 25-27 Number 8 % of total frequency 10.53 Over 27 Number 12 % of total frequency 15.79 10

PAGE 70

63 Table 4 Chi-Square Analysis of the Relationship Between Marital Status and Student Loan Default Status Defaulters Non-Defaulters Married Number 18 11 % of total frequency 23.68 14.47 Single Number 20 27 % of total frequency 26.32 35.53 Chi-Square = 2.732 DF = 1 Probability = 0.0983 the use of the variable marital status in a subsequent prediction model was appropriate. As shown in Table 5, the chi-square analysis of the relationship between sex and student loan default indicates that sex was not significant at the .10 level. Table 5 Chi-Square Analysis of the Relationship Between Sex and Student Loan Default Sex Defaulters Non-Defaulters Male Number 18 13 % of total frequency 23.68 17.11 Female Number 2 2 5 % of total frequency 26.32 32.89 Chi-Square = 1.362 DF = 1 Probability = 0.2432

PAGE 71

64 The expected frequency for each cell based on chance alone would be 25% and, as shown, the actual frequency percentage ranged from 17.1% to 32.9%. The chi-square analysis of the relationship between college standing and student loan default is shown in Table 6. The data collected did not indicate college standing to be significantly related to defaulters and nondefaulters at the .10 level. The expected frequency for each Table 6 Chi-Square Analysis of the Relationship Between College Standing and Student Loan Default College Standing Defaulters

PAGE 72

65 Table 7 Chi-Square Analysis of the Relationship Between Grade Point Average and Student Loan Default Grade Point Average Defaulters Non-Defaulters 1.50-1.99 Number 2 % of total frequency 2.63 2.00-2.49 Number 18 % of total frequency 23.68 2.50-2.99 Number 8 % of total frequency 10.53 3.00-3.49 Number 8 8 % of total frequency 10.53 10.53 3.50-4.00 Number 2 5 % of total frequency 2.63 6.58

PAGE 73

66 Table 8 Chi-Square Analysis of the Relationship Between Loan Total and Student Loan Default Size of Loan Defaulters Non-Defaulters 0-$l,000 Number 13 6 % of total frequency 17.11 7.89 $l,001-$2,000 Number 11 6 % of total frequency 14.47 7.89 $2,001-$3,000 Number 13 14 % of total frequency 17.11 18.42 $3,001-$4,000 Number 1 5 % of total frequency 1.32 6.58 $4,001-$5,000 Number 5 % of total frequency 0.00 6.58 $5,001-$6,000 Number 2 % of total frequency 0.00 2.63 Chi-Square = 15.7530 DF = 5 Probability = 0.0173 The results of the chi-square tests showed that only two of the variables were significant at the .10 level (marital status and loan total) . The differences found between the categories (student loan defaulters and non-defaulters) for both of these variables cannot be attributed to chance alone. The remaining four variables (sex, grade point average, age and college standing) were not statistically significant according to the chi-square results of this study, but they may demonstrate some differentiation between defaulters and

PAGE 74

67 non-defaulters on student loans based on the results of previous researchers. The results of the chi-square analyses indicated that only two of the six variables studied most frequently by previous researchers were significant according to the data collected in this study. The discovery that the only financial type variable (loan total) was also one of the two variables found to be significant may underscore the importance of Pattillo and Wiant's (1972) conclusion that items reflecting financial rather than biographical data appear to be better predictors of loan delinquency. This finding may indicate that since students with high loan needs have good loan payback rates, the students with the highest needs are the best loan risk for any one of a number of reasons (i.e., they may value money more due to its scarcity, they may have learned to manage money more frugally or they may have learned how to budget more effectively) . The strength of the relationship of this one variable to student loan payback indicates that other financial type data should be studied. Before developing a prediction model based on the results of the chi-square analyses, the researcher explored the possibility of interaction effects between the six variables. These results are presented in Table 9. It can be observed that no interaction between the variables related significantly to payback.

PAGE 75

68 Table 9 Interaction Effects Between the Variables Variables F Pr>F Significant ( o< =.10) Loan by Sex Loan by Grade Point Loan by Marital Status Sex by Marital Status Sex by Grade Point Aver, Grade Point by Marital Status 2.14 .1483 No 1.45

PAGE 76

69 specific characteristics of the community college population. Since the results presented in the literature and the results found for the community college students in this study were not corroborative regarding the significance of the variables, all six selected variables were utilized in the prediction model. Development of the Model In order to try to predict categorical membership based on the most discriminating variables, a multiple discriminant analysis technique was employed. Essentially this statistical procedure "weights" the predictor variables to yield maximum discrimination between the two groups (i.e., in this study it discriminates defaulters from non-defaulters) (Hays, 1981) . In addition, the nominal variables of marital status, college standing and sex were used to define dummy variables in order to introduce variance into the prediction equation. Three notable formulae generated by the multiple discriminant analysis function of the Statistical Analysis System (SAS) program were useful in distinguishing student loan defaulters. Therefore, three prediction models were developed. The first formula included all six variables which were entered into the data file in the following (arbitrary) order: age, marital status, sex, college standing, grade

PAGE 77

70 point average, and loan total. The results of this multiple discriminant analysis indicate that the six selected variables correctly predicted group membership for approximately 68% of the sample cases. Considering that the prior probability (based on chance alone) would be a 50% correct prediction, the six selected variables appear to be successful in improving the accuracy of predicting group membership. In order to improve the chances of making more accurate predictions, a second discriminant analysis procedure was performed. This time the multiple discriminant analysis enabled the researcher to catalogue the individual effect of each selected variable on the accuracy of the categorical prediction. Based on the probability and significance levels of the chi-square analysis the most differentiating variables were first included in the prediction model. When the first variable, loan total, was entered into the prediction formula, the chances of making accurate predictions improved notably. Instead of the 50% prior probability of making accurate categorical predictions the accuracy rate improved to 66%. The only other significant variable (marital status) was entered into the model next. The inclusion of this variable increased the probability of correctly categorizing the defaulting and non-defaulting student loan recipient from 66% to 71%. Therefore, this second formula (generated by the stepwise multiple

PAGE 78

71 discriminant technique) included only the two variables demonstrating the acceptable degree of significance based on the chi-square results ( u = .10) and it provided the best accuracy for categorical prediction (71%) . When the next two variables, sex and grade point average, were included in the model, the percentage of correct categorical classifications essentially remained the same. Although the variables would appear to be useful in a prediction formula as indicated in the literature review, their inclusion was not helpful in formulating a more accurate prediction model. When the variable sex was included in the model, the accuracy rating remained at 71%; however, when the variable grade point average was included, the number of correct classifications decreased by one and reduced the accuracy rating to 70%. Such a result seems to be inconsistent with statistical theory. After studying the values for each of the sample cases generated by the prediction formula, the researcher believes that the decreasing accuracy may be due to the mathematical rounding procedure within the Statistical Analysis System program. As a result, one sample case that defaulted on a student loan was classified by the prediction model as a non-defaulter on the total probability difference of .0033. Therefore, the apparent decreasing accuracy of the prediction model may be attributed to the mathematical precision of the computer program rather than the reflection of a conflict with statistical theory.

PAGE 79

72 The third formula was generated at the conclusion of the stepwise multiple discriminant technique and it included all six selected variables. The addition of the remaining two variables, college standing and age, did not improve the accuracy rating for the prediction model. Therefore, the formula, including all six selected variables, generated by the stepwise technique was able to predict correct categorical membership for 7 0% of the sample cases. Although both the first and third formulae included the six selected variables, the disparity in the predictive accuracy of the formulae is accounted for in the statistical technique employed for the derivation of the formulae. Since the stepwise multiple discriminant analysis technique enters the variables in the order of probability, the improved accuracy of the third formula (70%) is a result of the mathematical process of weighting the variables as they are entered into the prediction formula. Because the order of inclusion in the formula is different and the mathematical weighting is improved, the accuracy of the formula is improved. Based on the results of the discriminant analysis, it would appear that the best predictive formula for determining correct classification of defaulting and non-defaulting community college students would include only data regarding the students' loan totals and marital status. Adding the

PAGE 80

73 variables grade point average, sex, college standing and age, essentially maintained the effectiveness of the prediction formula. As evidenced, the stepwise technique employed to produce the second and third formulae improved the accuracy for categorical classification over the regular multiple discriminant analysis technique utilized for the first formula. Results Based on chance alone, one would expect to predict correctly the proper placement of students into the two categories (defaulting and non-defaulting) 50% of the time. The prediction formula generated by the inclusion of data concerning loan total and marital status increases the likelihood of making an accurate prediction from 50% to 71%. Therefore, it would appear that these data are relevant in predicting whether a student is likely to default on a student loan. The additional data that were collected for the other four variables (sex, grade point average, college standing and age) also appear to be relevant in producing a prediction model since the prediction formula derived from all six variables produced a 70% accuracy rating concerning categorical prediction and this rating is substantially higher than the 50% probability based on chance alone.

PAGE 81

74 However, all three of the prediction formulae reflect an error factor of nearly 30% indicating that roughly onethird of the cases were classified in error. In addition, the testing of the prediction formula utilizing new cases is not possible since the formula is derived from a stratified sample (which was necessary in order to perform the chisquare tests to determine variable usefulness). Therefore, the usefulness of the results of this study is limited even though the prediction formula demonstrates moderate success in differentiating defaultees from non-def aultees. Previous researchers also produced prediction models with limitations. For example, although Dyl and McGann (1977) reported significant success with their prediction model, Myers and Siera (1980) indicated that Dyl and McGann did not validate their results by applying the formula to cases other than those from which the formula was derived. In addition, the results of the study performed by Myers and Siera seem to indicate that although they presented a useful model, when they attempted to validate their model by predicting new, non-sample cases, the model was not reliable or useful. The results of this study show that the six variables which were selected based upon the results of the studies presented by previous researchers were not reliable predictors of payback for community college students. Therefore, a dependable prediction model cannot be formulated for

PAGE 82

75 community college students based on these variables. Perhaps the importance of this study is that the results coincide with findings reported by Hauptman (1978), Emmert (1978) and Pattillo and Wiant (1972). Hauptman first suggested that research must be independently conducted on the different types of postsecondary institutions. Underscoring this perception, Emmert concluded that institutional default rates may not be comparable if the schools' respective populations are from substantially different backgrounds. It may be significant that the two studies previously successful in determining reliable demographic variables and formulating prediction models (Dyl and McGann, 1977, and Myers and Siera, 1980) were based on fouryear public univeristy (University of Wyoming and New Mexico State University respectively) samples. Previous discussion regarding the importance of financial aid for community college students demonstrated the meaningful differences between the typical community college student and the typical four-year university student. The disappointing accuracy of this model, specifically the incorrect classification of 30% of the sample, may be attributed to the demographic differences between the sample in this study and the samples used in the previous studies. For example, the results of the previous research reported in the literature indicate that the demographic data represented by the six selected variables discriminate student

PAGE 83

76 loan defaulters and non-defaulters for four-year college or university students. However, as indicated by the chi-square results in this study, these same variables do not entirely discriminate student loan defaulters from non-defaulters for community college students. Therefore, the variables, selected based on the success of the previous studies conducted with a university student sample, are apparently not the best variables to use when attempting to predict community college student loan defaulters. The results of this study also appear to support Pattillo and Wiant's (1972) conclusion that financial rather than biographical data are better predictors of loan delinquency. One of the two variables that met the .10 statistical significance level in the chi-square analyses of this study was the loan total. It was also the only variable that was financially related since all of the other (five) variables in this study represented purely biographical demographic characteristics. Considering that the only financially related variable was also statistically significant, perhaps more financially related variables would yield a more accurate prediction model. Therefore, the results of this study may lead to the conclusion that, as suggested by Pattillo and Wiant, more financial data need to be collected and analyzed in order to make any reliable prediction of loan delinquency. This conclusion may be also applicable to the results presented by Dyl and McGann (1977) and Myers and Siera (1980).

PAGE 84

77 Although they have been able to develop successfully a prediction model, neither Dyl and McGann nor Myers and Siera has been able to validate successfully their models by predicting accurately new cases, cases not presented in the sample that derived the prediction model. Perhaps they need to collect more financial data regarding their students in order to formulate an accurate prediction model that can be validated by new cases. Summary The prediction formula derived from the data available through the Florida Student Financial Aid Commission for Florida community college students indicates that it is possible to make a useful discrimination regarding who will and who will not repay their student loans for approximately 70% of the sample cases. This prediction formula was derived through a (stepwise) multiple discriminant analysis computer program utilizing the six selected variables highlighted in previous research studies as useful discriminators between defaulting and non-defaulting students. A second (stepwise) multiple discriminant analysis utilizing only the two variables demonstrating statistical significance ( <* = .10), loan total and marital status, yielded a prediction formula with a categorical prediction accuracy rating of 71%. A third formula utilizing all six selected variables entered into the formula in random order

PAGE 85

78 (non-stepwise) generated a model that predicted correct categorical membership for 68% of the sample. Regardless of the formula utilized, approximately one-third of the sample cases were not predicted accurately. Although none of the prediction models are accurate for one-third of the cases, each prediction model is capable of predicting nearly 70% of the cases correctly. Therefore, the models are of some value in addressing the questions of this study. Future models may be more useful in predicting those students likely to repay student loans from those students who are not likely to repay student loans since the models derived from the data in this study have limitations concerning their application as discussed in Chapter IV. The limited success of these prediction models appears to be consistent with the findings presented by the previous researchers in the field. Since the difficulty in validating even successful prediction models appears to be widespread, as confirmed by the results of previous researchers, it may be that the variables utilized thus far are not the most discriminating variables regarding student loan defaults. Perhaps future research should focus on financial type variables in order to provide relevant, statistically acceptable prediction models to assist student financial aid administrators in predicting student loan defaulters.

PAGE 86

CHAPTER IV CONCLUSIONS, IMPLICATIONS AND RECOMMENDATIONS Previous studies researching the relationship between student demographics and loan repayment recorded similar results to this study. However, the results of this study demonstrate several unique implications for future researchers. Conclusions Since the models presented in this study improve the chances of making an accurate prediction concerning the likelihood of default on student loans from 50% (based on chance alone) to nearly 70% (based on the models) , the models are successful in discriminating student loan defaulters from non-defaulters. Therefore, the questions of this study presented in Chapter I must be answered in the affirmative. Even though some of the six selected variables were not statistically significant in this study, the inclusion of all six variables in the prediction models was warranted because of the inconclusive results regarding the importance of the individual variables as demonstrated by previous university researchers. Inclusion of all six variables was 79

PAGE 87

80 also appropriate since this study was the first to utilize a community college sample, and the results of the chi-square analyses in this study may indicate that the variables distinguishing community college student loan defaulters may be different from the variables distinguishing university student loan defaulters. Despite the differing results of the chi-square analyses for the individual variables, the models formulated utilizing all six selected variables were accurate in predicting default or non-default for nearly 70% of the sample cases. Therefore, the models presented in this study are useful in predicting student loan defaulters. The conclusions of this study suggest several implications of importance for future studies. Implications The most notable implication of this study is that there are differences between community college students and university students regarding the significant variables distinguishing student loan defaulters from non-defaulters. This study shows that, although previous researchers identified the six selected variables and determined that they were significant in distinguishing defaulting university students, only two of the six variables were significant in distinguishing defaulting community college students. Therefore, this study underscores the importance of performing separate

PAGE 88

study designs yielding specific results for the different kinds of institutions utilizing the Guaranteed Student Loan system. The results of the chi-square analyses also appear to indicate that the data currently being collected for community college students do not adequately distinguish defaulting from non-defaulting student borrowers. As suggested previously, it appears that more financial data should be collected and analyzed if one is to achieve a more accurate prediction model for community college student loan defaults. Although there are inherent differences between the administrative mechanisms employed for the student loan system and those utilized by commercial lending institutions, information found useful in determining eligibility for commercial loans may be useful in distinguishing student loan defaulters from non-defaulters. Therefore, a second implication based on the results of this study is that future research should provide more focus on the financial data of community college students in order to improve the chances of formulating a better prediction model. Limitations of the Model The results of this study reveal two major limitations of importance for future researchers. First, the models presented in this study do not accurately categorize nearly one-third of the sample cases.

PAGE 89

82 As indicated, an inaccurate prediction regarding the probability of default or repayment was made for nearly one-third of the students in the sample. This imperfection results in the limited application of the prediction models. Previous researchers were unable to formulate a prediction model and validate their models with any great success; similarly, this study did not find a sufficiently accurate prediction model for determining who will default on their student loan. Second, since the models were not effective in making an accurate prediction for all cases in the sample, it appears that additional (especially financially related) variables may be of value in future research studies. This study was restricted to the use of data highlighted as significant in the literature and available through the Florida Student Financial Aid Commission (FSFAC) in Tallahassee, Florida. The results of this study and the previous studies seem to indicate that some variables which were not available through the FSFAC data files may be of critical importance in formulating a more accurate prediction model. For example, the employment status and type of job held at the time of loan application or the family's financial data including loan repayment history may be reliable variables to consider when predicting student loan defaults. In addition, data concerning ethnic background, college major, and adjusted gross income were found to be of value in prediction formulae presented by university researchers. Although the

PAGE 90

data are not currently available in the data system provided through FSFAC, perhaps these variables are also important in the formation of a prediction model concerning the probability of default for community college students. Any further study regarding a prediction model for determining the probability of default should also research the value of including these variables. Based on the difficulty of collecting data concerning these variables, the inclusion of these variables would alter considerably the methodology section of any similar research study. However, the variables do appear to be of value in examining critically the student loan default problem and presenting a model that improves predictive accuracy. Therefore, the resources employed to collect the data would appear to be prudently utilized. Recommendations The use of any of these predictive models for determining the likelihood of students defaulting on their student loans is limited. Therefore, the foremost recommendation derived from the data and the literature search is the need to encourage further research. The importance of this study is that the data indicate it is possible to formulate a model which will accurately predict 70% of the students within the sample. It appears realistic to assume that further research, especially the financially related research suggested in the previous discussion, may result

PAGE 91

84 in a more inclusive and accurate prediction model which would be useful in determining the probability of default. Such research would need to be more inclusive and therefore the collection of the new data would demand much more time and financial resources. However, the possibility of finding an appropriate collection of variables and the usefulness of the resultant prediction formula would indicate that student loan default prediction is an appropriate area for further study. In the event that a subsequent study does provide a reliable prediction model, it should be tested and validated by the researcher using new cases derived from a different sample. In addition, before the model is implemented on any individual college campus, it should be tested and validated by the financial aid officer for applicability on his/her campus. Only when these validations are performed can there be confidence in the predictive capability of the model, for the results of this study and the research presented in the literature clearly indicate that the models and appropriate variables vary from institution to institution. This variation is dependent upon the type of institution being researched. Another reason for establishing institutional validation procedures is to determine if significant differences occur among institutions of the same type. It is possible that a single, reliable prediction model that is derived from data representing a state-wide population for a

PAGE 92

85 selected type of institution may have no validity or usefulness on a single campus. Because of the increased emphasis on student loan programs as a major component of the student financial aid delivery system, further research concerning student demographics and the probability of default is of utmost importance. Such importance is underscored by the discussions presented in the literature search and the successful results of this study.

PAGE 93

BIBLIOGRAPHY Ary, D., Jacobs, L. C, & Razavich, A. Introduction to research in education (2nd ed.). New York: Holt, Reinhart and Winston, 197 9. Balderston, F. E. The repayment period for loan financed college education . New York: Ford Foundation, 1970. (ERIC Document Reproduction Service No. ED 081 379) Balkins, A. J. Withholding transcripts for non-payment of educational debts: Before and after bankruptcy. Williamette Law Review , 1979, 15, 563-575. Bar Admission — Default on student loan warrants denial of admission to Minnesota bar. Willi am Mitchell Law Review , 1980, 6, 443-454. " Block, W. R. Other students need money: An approach to the administration of revolving loan accounts. Journal of Student Financial Aid , 1977, 7, 26-30. Blocker, C. E. The reality of student finances. New Directions for Community Colleges , 1974, 8, 124^128. The Carnegie Council on Policy Studies in Higher Education. Next steps for the 1980 's in student financial aid . San Francisco, CA: Jossey Bass Publications, 1979. Change Magazine and Educational Change. Colleges and money A faculty guide to academic economics ^ Washington, DC Author, 1976. Chronicle of Higher Education , February 24, April 28, December 1, 1982, February 2, 1983. Collins, J., Maguire, J. J., & Turner, R. M. Unmet need: How the gap is filled. Journ al of Student Financial Aid , 1979, 9, 4-15. " ' Crackdown begins on federal worker's student loan defaults. Higher Education and Nati onal Affairs. December 10, 1982, p. 3. — — Davis, J. S., & Van Dusen, W. D. Guide to the literature of student financial aid . Princeton, NJ: College Entrance Examination Board, 1978. 86

PAGE 94

87 Deitch, K. M. Pricing and financial aid in American higher education: Some interactions . Aspen, CO: Aspen Institute for Humanistic Studies, 1978. (ERIC Document Reproduction Service No. ED 184 450) Dyl, E. A., & McGann, A. F. Discriminant analysis of student loan application. Journal of Student Financial Aid, 1977, 7, 35-40. ~ ' Emmert, M. A. National direct student loan default rates: A measure of administrative quality or something else? Journal of Student Financial Aid , 1978, 8, 43-47. Ferguson, G. A. Statistical analysis in psychology and education (3rd ed. ) . New York: McGraw-Hill Book Company, 1971. Hartman, R. The National Bank approach to solutions. In L. Rice (ed.), Student loans: Problems and policy alternatives . Princeton, NJ: College Entrance Examination Board, 1978. Hauptman, A. Student loan defaults: Toward a better understanding of the problem. In L. Rice (ed.), Student loans: Problem and polic y alternatives. Princeton, NJ: College Entrance Examination Board, 1978. Hays, W. L. Statistics (3rd ed.). New York: Holt, Reinhart and Winston, 1981. Horch, D. H. A retrospective description of the nati onal direct student loan program: Administrative pract ices and institutional default rates in 1972-73 . Princeton, NJ: Educational Testing Services, 1978. (ERIC Document Reproduction Service No. ED 165 617) Hyde, W. D. , Jr. State guarantee agencies and capital availability for student loans. Papers in education finances. Paper no. 22~ T Denver, CO: Education Commission of the States, Education Finance Center, 1979. (ERIC Document Reproduction Service No. ED 176 648) Jenkins, J. A. Whose fault is student loan default? Change, November, 1978, pp. 44-45. Johnstone, D. B. Student loans: Some practical radical alternatives. In Symposium on Federal Student Loans . Washington, DC: National Association of Student Financial Aid Administrators and the American Council on Education, 1978.

PAGE 95

88 Kelly, M. , ed. Administrator . Madison, WI : Magna Publications, January 7, 1983. Kendis, K. L. Solving the student loan problem by supplementing government loan policy — An expanded role for institutional aid administrators. Student loans : Alternatives for reauthorization . Washington, DC: National Association of Student Financial Aid Administrators, 1978. Leonard, K. J. Collection of federal student loans. Columbia Journal of Law and So cial Problems, 1980, 15, 317-358. " — Martin, P. Student loans: Curing problems and insuring the future. Capital University L aw Review, 1980, 9, 579594. Millsap, W. L., & Wright, P. S. Recent cases on student transcript rights after bankruptcy. Journal of College and University Law , 1979, 6, 231-240. Myers, G., & Siera, S. Development and validation of discriminant analysis models for student loan defaultees and non-defaultees. Journal of Student Fi nancial Aid, 1980, 10, 9-17. ~ ~ National Association of Independent Colleges and Universities, An analysis of the tuition advance fund bill . Washington, DC: Author, 1978. (ERIC Document Reproduction Service No. ED 171 189) New York Times , October 12, 1975, Section 3, p. 16. Ostar, A. W. The effects of inflation and pricing policies on college enrollments" ! Washington, DC: American Association of State Colleges and Universities, 1978. (ERIC Document Reproduction Service No. ED 177 961) Premises of federal student assistance . Washington, DC: American Association for Higher Education, 1979. (ERIC Document Reproduction Service No. ED 193 999) Rudolf, F. Myths and realities of student aid. College Board Review , 1962, Fall, pp. 18-23. ~ Sanford, T. R. The impact of student loans on graduate education. In AIR forum 1980 paper . Atlanta, GA: Association for Institutional Research, 1980. (ERIC Document Reproduction Service No. ED 189 936)

PAGE 96

89 Silber, J. R. The tuition advance fund: A proposal for funding higher education. In paper presented at Assembly of the National Center for Education Manag ement Systems . Boston, MA: National Center for Education Management Systems, 1978. (ERIC Document Reproduction Service No. ED 163 824) Study of the requirements for forming state guarante e agencies, Final report ^ New York: Research and Forecasts, Inc., 1979. (ERIC Document Reproduction Service No. ED 188 508) Tate, J. L. Federal aid to postsecondary students. Journal of Family Law , 1979, 18, 147-178. U. S. Congress. House. Committee on Education and Labor. Subcommittee on Postsecondary Education. Student loan defaults . Hearings before a subcommittee of the Committee on Education and Labor, House of Representatives, 95th Cong., 1st Sess., 1977. (ERIC Document Reproduction Service No. ED 153 538) U. S. Department of Health, Education, and Welfare. Office of Education. Study of selected educational institu tions and students participating in direct school lending under the federal student loan program, Final report . Washington, DC: Otfice of Education, 1979. (ERIC Document Reproduction Service No. ED 185 953) U. S. General Accounting Office. Collection efforts not keeping pace with growing number of defaulted student loans . Washington, DC: Government Printing Office, 1977. U. S. General Accounting Office. Stronger action needed to recover $730 million in defaulted national direct stu dent loans . Washington, DC: Government Printing Office, 1981.

PAGE 97

BIOGRAPHICAL SKETCH James A. Schmidt was born December 11, 1950, at Pittsburg, Pennsylvania. He attended local schools in St. Petersburg, Florida, and graduated from Dixie M. Hollins High School in 1968. He entered the Florida State University in 1968 and received the Bachelor of Science degree in criminology in 1972 and the Master of Science degree in education (student personnel/higher education) in 1973. He served Southeastern Community College (N.C.) as Director of Admissions and Student Activities and Researcher from 1973 until mid-1976. In the fall of 1976, he entered Stetson University College of Law in St. Petersburg, Florida. The following year, 1977, he accepted a counseling position with Edison Community College in Fort Myers, Florida. During the 1981-82 academic year, he attended the University of Florida while on leave from Edison Community College in order to complete residency and course requirements for the Doctor of Philosophy degree in educational administration (higher education) . Upon returning to Edison Community College in July, 1982, he assumed his present position as Director of Financial Aid. James A. Schmidt is married to the former Linda Dale Thompson of Callahan, Florida. They have two sons, Daniel and Timothy. He is a member of Beta Theta Pi, Kappa Delta 90

PAGE 98

91 Pi, Phi Delta Kappa, several professional and service organizations and has been selected as an Outstanding Young Man of America for 1983.

PAGE 99

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. James L. Wattenbargery, Chairman 'rofessor of Educational Administration and Supervision I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. /MA VI ' / m //; fuUwe, ,<.A Cr?„. L v Harold C. Riker "' " " Professor of Counselor Education This dissertation was submitted to the Graduate Faculty of the Department of Educational Administration and Supervise •: in the College of Education and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August, 1983 Dean for Graduate Studies and Research

PAGE 100

'iBiilii!, 3 1 262 08552 9427