Higher Education: Restructuring Student Aid Could Reduce Low-Income
Student Dropout Rate (Letter Report, 03/23/95, GAO/HEHS-95-48).

Pursuant to a congressional request, GAO reviewed how student financial
aid affects low-income students' dropout rates, focusing on whether: (1)
the timing of loan and grant aid influences students dropout rates; and
(2) restructuring federal grant programs could improve low-income
students' dropout rates.

GAO found that: (1) grants and loans do not have the same effects on
reducing low-income college student dropout rates; (2) although grant
aid generally lowers low-income students' dropout rates, loans have no
significant impact on these students' dropout rates; (3) the timing of
grant aid greatly influences students' dropout rates; (4) grant aid to
low-income students is more effective during the first school year than
in subsequent years; (5) although financial aid program participants
have substantially lower dropout rates than other comparable students,
financial aid directors and students have mixed views on the potential
efficacy of frontloading aid packages; (6) a pilot program could be
valuable in evaluating the cost effects of frontloading student aid for
low-income college students; and (7) Department of Education officials
need to further review their legislative authority to determine whether
they are authorized to conduct such a pilot project.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-95-48
     TITLE:  Higher Education: Restructuring Student Aid Could Reduce 
             Low-Income Student Dropout Rate
      DATE:  03/23/95
   SUBJECT:  Grant administration
             Colleges/universities
             Disadvantaged persons
             Student financial aid
             Educational grants
             Student loans
             Higher education
             Aid for education
             Attrition rates
             Minorities
IDENTIFIER:  Pell Grant
             Federal Family Education Loan Program
             Dept. of Education Stafford Student Loan Program
             Dept. of Education Supplemental Loans for Students Program
             Dept. of Education Parent Loans for Undergraduate Students 
             Program
             
**************************************************************************
* This file contains an ASCII representation of the text of a GAO        *
* report.  Delineations within the text indicating chapter titles,       *
* headings, and bullets are preserved.  Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are           *
* identified by double and single lines.  The numbers on the right end   *
* of these lines indicate the position of each of the subsections in the *
* document outline.  These numbers do NOT correspond with the page       *
* numbers of the printed product.                                        *
*                                                                        *
* No attempt has been made to display graphic images, although figure    *
* captions are reproduced. Tables are included, but may not resemble     *
* those in the printed version.                                          *
*                                                                        *
* A printed copy of this report may be obtained from the GAO Document    *
* Distribution Facility by calling (202) 512-6000, by faxing your        *
* request to (301) 258-4066, or by writing to P.O. Box 6015,             *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time.                                    *
**************************************************************************


Cover
================================================================ COVER


Report to Congressional Requesters

March 1995

HIGHER EDUCATION - RESTRUCTURING
STUDENT AID COULD REDUCE
LOW-INCOME STUDENT DROPOUT RATE

GAO/HEHS-95-48

Reducing College Dropouts


Abbreviations
=============================================================== ABBREV

  FFEL - Federal Family Education Loan
  SAT - Scholastic Aptitude Test
  SEOG - Supplemental Education Opportunity Grant

Letter
=============================================================== LETTER


B-253597

Letter Date Goes Here

The Honorable Claiborne Pell
Ranking Minority Member, Subcommittee
 on Education, Arts, and Humanities
Committee on Labor and Human Resources
United States Senate

The Honorable Paul Simon
United States Senate

Postsecondary education is a strong determinant of relative wage
earnings.  College graduates earn much more than those with only a
high-school education or less, and the differential has been
increasing.  Low-income and minority students have traditionally been
underrepresented among college students.  For them, federal student
financial assistance programs have become increasingly critical as
college costs have increased faster than the rate of inflation since
the 1970s and as low-income families lost ground relative to
high-income families.  Federal assistance to students pursuing higher
education has exceeded $300 billion (in 1994 dollars) over the past
15 years. 

While the federal government's investment to improve college access
for low-income students has been substantial, recent changes in
federal financial aid may inhibit broader college access.  A growing
proportion of federal aid has taken the form of loans rather than
grants since the 1970s.  With federal grant aid declining in relative
terms, students and their families have had to shoulder a greater
share of college expenses.  Many policymakers have expressed concern
that this trend in financial aid patterns, which increases students'
net costs for higher education, has diminished college access--both
initial entry and attendance through graduation--for low-income
students. 

As agreed, this report is our second and final product responding to
your request that we compare the relative effectiveness of grants and
loans in helping students stay in college until graduation.  In
previous testimony,\1 we focused on minority students and reported
that while grant aid enhanced the chances that African American and
Hispanic students would stay in college from year to year, loan aid
did not.  In this study, we focused on low-income students.  We
assessed the relative effects of grant and loan aid throughout a
student's academic career and then separately by year.  Specifically,
we addressed two questions regarding low-income students: 

1.  Do grants and loans have equivalent effects on helping students
stay in college? 

2.  Does the timing of grant aid influence the length of time
students stay in college? 

To answer these questions, we analyzed two student-level databases to
examine the statistical relationship between grants, loans, and
staying in college.  One database comprised a national sample of high
school seniors who began full-time study at 4-year colleges; we
traced these students through college.  The other database consisted
of a group of relatively low-income freshmen from a large public
4-year university that frontloaded\2 some of its institutional grant
dollars to improve these students' dropout rates. 

To complement our statistical analysis and provide additional
perspectives on our findings and observations, we talked to financial
aid directors and students from 12 colleges and universities.  We
judgmentally selected four schools from each of three
areas--Washington, D.C.; Philadelphia, Pennsylvania; and Seattle,
Washington--for our study.  We convened a discussion panel with
financial aid directors from the schools in each area and asked them
for their observations on trends in federal student financial aid,
factors affecting students' staying in school, and the potential
benefit of frontloading grants.  We also interviewed 51 students from
these 12 colleges and universities, asking them about their financial
aid packages, the effects working has on their studies, and their
thoughts about debt accumulation.  Although the financial aid
director and student responses illustrated ways financial aid affects
students, neither group's responses were intended to be
representative of the respective population as a whole.  (For further
details on our scope and methodology, see app.  I.)


--------------------
\1 Higher Education:  Grants Effective at Increasing Minorities'
Chances of Graduating (GAO/T-HEHS-94-168, May 17, 1994). 

\2 For our purposes, frontloading grants entails giving students
mostly grant aid in the first year and increasingly substituting loan
aid in subsequent years, which culminates in an aid package
consisting mostly of loans in the final school year. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

Grants and loans do not have equivalent effects on low-income
students' staying in college, according to our statistical results. 
Rather, on average, grant aid lowers the probability of low-income
students' dropping out, while loans have no statistically significant
impact on their dropping out.  Furthermore, the timing of grant aid
influences students' probability of dropping out.  For example, our
results indicated that, on average, for low-income students, grant
aid is relatively more effective during the first school year than in
subsequent years.  Sample-specific information from a university
program that frontloaded grants for some students, and provided them
with academic and administrative support, reinforced these findings. 
Program participants had substantially lower dropout probabilities
than other comparable students.  Financial aid directors and students
with whom we spoke had mixed views on the potential efficacy of
frontloading aid packages.  Some thought the approach would be
beneficial, but others raised such concerns as the bait-and-switch
aspects of replacing grant awards with loans in later years. 

In view of our statistical results, but noting the limited experience
with frontloading grants, we believe the Department of Education
should conduct a pilot program to evaluate the impact of frontloading
on reducing dropouts among low-income college students.  Department
officials told us they would need to further review their legislative
authority to determine whether they could conduct such a pilot or
would need to seek additional authority. 


   BACKGROUND
------------------------------------------------------------ Letter :2

Most federal student financial aid programs are authorized under the
Higher Education Act of 1965.  Federal student financial assistance
exceeded $30 billion in academic year 1993-94, and most assistance
came from two programs--the Pell grant and Federal Family Education
Loan (FFEL) programs.  The Pell grant program, which primarily
targets low-income students, accounted for about $5.7 billion, while
the FFEL program comprised over $21 billion of the total federal
aid.\3 Maximum annual awards to students in each program are capped: 
In 1993-94, the maximum Pell grant was $2,300, and the maximum
subsidized Stafford loan--the largest of the FFEL loan
programs--ranged from $2,625 for freshmen to $5,500 for seniors. 

The Department of Education administers the Pell grant program in
accordance with eligibility criteria and authorized maximum award
amounts set by the Congress.  In addition, the Congress effectively
limits actual maximum Pell award amounts each year through the
appropriations process.  Actual maximum awards have been less than
the authorized levels each year since 1980.  For example, in 1993-94,
the authorized maximum Pell grant was $3,700, but the appropriation
for the program limited the actual maximum award to $2,300. 


--------------------
\3 Subsidized Stafford loans comprised $14 billion of the FFEL
program funds.  Students receiving subsidized loans do not pay
interest while attending school or during a grace period after
leaving school.  When repayment begins, it is with a below-market
interest rate.  The remaining $7 billion of the FFEL program consists
of (1) Supplemental Loans for Students, (2) Parent Loans to
Undergraduate Students, and (3) unsubsidized Stafford loans. 


      CHANGING FEDERAL AID SYSTEM
      COINCIDED WITH RISING
      COLLEGE COSTS
---------------------------------------------------------- Letter :2.1

The composition of student financial assistance has changed
dramatically over the past two decades.  Although total federal aid
has increased since the late 1970s, loan aid has increased far faster
than grant aid.  From 1977 to 1980, grant aid exceeded loan aid, but
since 1985 loan aid has exceeded grant aid by about twice as much
(see fig.  1). 

   Figure 1:  Loan Aid Surpassed
   Grant Aid in Early 1980s

   (See figure in printed
   edition.)

Budgetary concerns and program changes have limited grant aid for
low-income students.  As the deficit rose during the 1980s,
policymakers' awareness of budgetary trade-offs and the need to
leverage resources grew.  Loans are a less expensive form of aid for
federal budgetary purposes than grants because the budget accounts
only for the cost of interest subsidies and default payments.  Thus,
for a given federal expenditure, the government can offer more aid if
it is provided as loans.\4 In addition, the 1978 Middle Income
Student Assistance Act extended eligibility for Pell grants to higher
income students; however, appropriations did not allow for
commensurate increases in program dollars.  Consequently, more
students now receive Pell grants, but the actual maximum award has
remained approximately constant in nominal dollars since 1986. 

Cost pressures for low-income families have increased since the late
1970s, as the average cost of 4-year colleges and universities has
increased faster than the inflation rate.  Between 1978 and 1992, the
average tuition, room, and board charge at 4-year public colleges and
universities rose by 26 percent in real terms.\5 This had two
distinct effects.  First, college expenses at the average public
university absorbed 11 percent of median family income in 1978 and 14
percent in 1992.  For families at the 20th-income percentile, this
charge increased even more, from 22 to 31 percent of income.  Second,
the actual maximum Pell grant, which covered over half the costs at
the average public 4-year school in 1985, now covers less than 40
percent. 


--------------------
\4 For example, for the same appropriation as a $1,000 grant, the
government can lend $3,000 to $4,000. 

\5 These expenses rose by 52 percent at private colleges and
universities. 


      LOW-INCOME STUDENTS LESS
      LIKELY TO ENROLL AND STAY IN
      COLLEGE
---------------------------------------------------------- Letter :2.2

Low-income students are underrepresented among college students. 
Low-income students enroll in college at lower rates than high-income
students, although enrollment rates have been rising for all income
groups (see fig.  2). 

   Figure 2:  College Enrollment
   Rates Highest for High-Income
   Students

   (See figure in printed
   edition.)

Notes:  Low income is the bottom 20 percent of the family income
distribution, middle income is the next 60 percent, and high income
is the top 20 percent.

The enrollment rate is the percentage of high school graduates in a
given year who were enrolled in college the October following
graduation. 

We found no data on students' degree completion by income group. 
However, minorities are overrepresented among low-income families, so
their rates serve as a reasonable proxy for low-income students'
graduation rates.  Sample data show that minority students are less
likely to stay in school and graduate than white students.  For
example, in one sample of students entering 4-year colleges in 1983,
1984, or 1985, 56 percent of white students completed degrees within
6 years, but only 41 percent of Hispanic students and 32 percent of
African American students did so.\6 Policymakers have raised the
possibility that reduced grant aid, relative to the soaring costs of
a college education, have adversely affected graduation rates for
students at the low end of the income scale. 


--------------------
\6 These data pertain to 298 National Collegiate Athletic Association
Division I schools. 


   PRINCIPAL FINDINGS
------------------------------------------------------------ Letter :3


      GRANTS MORE EFFECTIVE THAN
      LOANS IN REDUCING DROPOUTS
---------------------------------------------------------- Letter :3.1

The composition of financial aid packages and the timing of
particular aid components influence education outcomes.  Our results
indicated that, for low-income students, grant aid was effective in
reducing dropouts, but loan aid was not.\7

In addition, grant aid for low-income students was most effective in
the first year, with efficacy decreasing in the second and third
years.  The results of the university frontloading program
strengthened our confidence in this finding.  Students who received
frontloaded grants had a lower dropout probability than other
comparable students.  Results of our statistical work showed the
following. 

Grants versus loans:  Grants significantly reduced dropout
probabilities for low-income students.  In the High School and Beyond
database sample, an additional $1,000 in grant aid for a low-income
student reduced the dropout probability by 14 percent for the award
year.\8

Loans did not have a statistically significant effect for this
group--a commensurate increase in loans did not significantly affect
the student's probability of dropping out. 

First-year students:  Grants were most effective in reducing
low-income students' dropout probabilities in the first year.  For
these students, an additional $1,000 grant in the first year reduced
the dropout probability by 23 percent.  In the second year, the
additional grant reduced the dropout probability by 8 percent\9
while, in the third year, it had no statistically discernable effect. 

Frontloading grants:  The university's program for high-need
freshmen, which included frontloading grants, had a significant
effect on reducing dropouts.\10 Program participants were 39 percent
less likely to drop out in a year than nonparticipants.  For the
lowest income students, those below the poverty line, the program
reduced the dropout probability by 64 percent. 

These results, with certain qualifications, indicate that
frontloading grants for college students, especially low-income
students, could improve dropout rates.  The results pertain only to
4-year college students and thus have no implication for students at
2-year schools.  Also, the frontloading experiment took place at a
university that combined it with other programs to reduce dropouts,
and the results are not generalizable beyond that school.  However,
we believe the sum total of our results shows that frontloading
grants holds promise.  (For detailed information on the analyses that
led to these results, see app.  II.)


--------------------
\7 All results in this section are significant at the 5-percent level
unless noted otherwise. 

\8 A low-income student in this sample was one whose family income
was below $21,000. 

\9 This result was significant only at the 10-percent level. 

\10 Because the program also involved additional student support
services, we cannot unequivocally attribute the entire dropout
reduction to frontloading grant money. 


      RESULTS FROM FINANCIAL AID
      DIRECTOR DISCUSSION PANELS
      AND STUDENT INTERVIEWS
---------------------------------------------------------- Letter :3.2

Comments from financial aid directors and students we interviewed
helped us interpret the statistical results.  One opinion arising in
the directors' panels, for example, was that some low-income students
are reluctant to borrow, especially during their first year or two in
college.  This observation is consistent with our statistical
findings about grants being more effective than loans in increasing
the likelihood that first-year, low-income students will stay in
school.  The directors we spoke with were generally positive about
the potential benefits of frontloading grants, several saying it
could help low-income students stay in college by giving them time to
become acclimated to college and reducing financial pressures when
students are most vulnerable to dropping out.  One concern about
frontloading was that students might perceive it as a bait-and-switch
policy because it would involve reducing grant awards in later years. 
In the student interviews, we sometimes heard that borrowing was
initially difficult for students and that grant aid made the
difference in their being able to start college.  Another theme among
the students was that year-to-year consistency was important in their
aid packaging, so that they could plan ahead without disruptions, and
that frontloading seemed contrary to the principle of consistency. 
(For a discussion of the full range of comments from financial aid
directors and students we spoke with, see app.  III.)


      PILOT FRONTLOADING PROGRAM: 
      DEPARTMENTAL AUTHORITY AND
      IMPLEMENTATION ISSUES
---------------------------------------------------------- Letter :3.3

We discussed with Department officials the value and feasibility of
their conducting a pilot frontloading program.  They thought
frontloading held promise and expressed an interest in such a pilot
program.  They told us that they might have authority under current
law (20 U.S.C.  ï¿½ 1094a(d) (1988 and Supp.  IV 1992)) to conduct a
pilot.  This law authorizes the Department to designate institutions
that volunteer to participate as "experimental sites." These
institutions help evaluate the impact and effectiveness of proposed
regulations or new management initiatives.  The Secretary of
Education may exempt participating institutions from legal
requirements as necessary to conduct the experiments.  The officials
said that they had not yet determined whether this authority would
permit a pilot frontloading program and that they might need specific
authority from the Congress to conduct a pilot.  We defer to the
Department on whether it currently has the authority to do so or
needs additional authority. 

Such a pilot program, moreover, would need to address several
implementation issues.  The potential benefits of frontloading could
be lost if institutional aid policies were changed to offset the
federal change.  Schools would need to be encouraged to ensure that
overall grant aid, meaning federal and institutional aid combined,
were frontloaded.  Also, because eligibility for federal financial
aid is based in part on annual income and other family resources that
change over time, the amount of aid a student qualifies for changes
each year.  Frontloading would entail estimating a 4-year package,
requiring methods not currently employed in aid determination.  It
would also involve adjusting loan limits for third- and fourth-year
students at pilot schools and developing aid award rules for students
who transfer between pilot and nonpilot schools. 

In evaluating a pilot program, changes in dropout rates would have to
be interpreted carefully.  A policy of frontloading grants might
attract students to college who would not have attended otherwise. 
Although some of these students would graduate, on the whole their
dropout rate could be higher than that of the current student
population.  Frontloading might reduce the number of dropouts among
students who now attend college, but high dropout rates among this
new college population could leave the overall dropout rate unchanged
or higher. 


   CONCLUSIONS
------------------------------------------------------------ Letter :4

Our statistical analysis indicates that loans and grants are not
equal substitutes in terms of affecting education outcomes for
low-income students.  Aid packages with relatively high grant levels
may improve low-income students' access to higher education more than
packages that rely more on loans.  In addition, our analysis
indicates that the earlier low-income students receive grant
assistance, the more likely they are to stay in college. 

Departure from the conventional approach to dispersing student
financial aid--relatively proportionate amounts each year--could
further improve low-income students' dropout rates.  Given that the
dropout rate is highest in students' first 2 years, frontloading
grants would appear to provide low-income students with the most
effective means of financial support when they are most likely to
benefit from it.  Restructuring federal grant programs to feature
frontloading could improve low-income students' dropout rates without
changing any student's overall 4-year allocation of grants and loans. 

Given our statistical results, the mixed views of aid directors and
students we spoke with, and the limited experience with frontloading,
we believe the Department needs to shed light on the matter by
undertaking a pilot program. 


   RECOMMENDATION TO THE
   DEPARTMENT OF EDUCATION
------------------------------------------------------------ Letter :5

We recommend that the Department of Education conduct a pilot program
of frontloading federal grants at a limited number of 4-year schools
to evaluate the impact of frontloading on reducing dropouts among
low-income college students.  If, upon review, the Department
concludes that it lacks authority to conduct this pilot, we recommend
that the Department seek legislation from the Congress to authorize
the pilot. 


   AGENCY COMMENTS
------------------------------------------------------------ Letter :6

As agreed, we did not obtain written comments on the report from the
Department of Education, but we discussed our findings with program
officials.  The officials generally agreed with our results and made
suggestions, which we

We conducted our review between March 1993 and December 1994 in
accordance with generally accepted government auditing standards.  We
are sending copies of this report to the Secretary of Education,
congressional committees, and other interested parties.  Please call
Cornelia M.  Blanchette or me on (202) 512-7014 if you or your staff
have any questions about this report.  Other GAO contacts and
contributors are listed in appendix IV. 

Sincerely yours,

Linda G.  Morra
Director, Education
 and Employment Issues


SCOPE AND METHODOLOGY
=========================================================== Appendix I

To examine the effects of grants and loans on the probability of
students' staying in college or dropping out, we analyzed two
databases:  (1) High School and Beyond, a national survey of students
begun in 1980, and (2) financial aid data from a large public
university.\11 The two databases included different information, but
they both contained year-by-year totals for grants and loans each
student received, tuition the student paid, and background
information on the student.  In addition, we could determine the
number of years a student remained in school and if and when that
student dropped out.  We used duration analysis to determine the
factors affecting the probability of staying in college or dropping
out.  To help understand the reasoning and decision-making behind our
statistical results, we conducted discussion panels with financial
aid directors and interviews with students at selected schools.  Our
analysis covered only students in 4-year undergraduate programs; we
did not include community colleges, proprietary schools, or graduate
or professional programs. 


--------------------
\11 In consideration of their help and cooperation in providing data
to us, we have agreed not to identify the university. 


   REGRESSION ANALYSES
--------------------------------------------------------- Appendix I:1


      HIGH SCHOOL AND BEYOND DATA
------------------------------------------------------- Appendix I:1.1

The High School and Beyond survey was first conducted in 1980. 
Graduating high school seniors were asked questions about family
background, educational attainment, and future plans.  To obtain
information on activities since high school, these same students were
then reinterviewed in 1982, 1984, and 1986.  This provided
longitudinal information on students in the initial sample.\12 We
selected for analysis those students who began college full time at a
4-year school immediately after high school.  We followed these
students through their college years and noted whether they continued
from year to year or dropped out.  Our sample consisted of 3,652
students. 

The High School and Beyond survey oversampled African American and
Hispanic students.  This oversampling resulted in sufficient
observations on these populations for meaningful results to be
obtained for them.  We weighted our sample data so that the
proportions of African Americans, Hispanics, whites, and others would
match population proportions.\13 Except when we analyzed data
separately by race, we reported weighted means and regression results
in all cases. 


--------------------
\12 The High School and Beyond survey was conducted by the National
Center for Education Statistics of the Department of Education.  It
is the most recent longitudinal survey the Department has conducted
that has information on students' completed college careers. 

\13 Specifically, we weighted data according to the proportion of
students in the national population entering 4-year colleges for the
first time in 1980. 


      UNIVERSITY FINANCIAL AID
      DATA
------------------------------------------------------- Appendix I:1.2

We analyzed financial aid records from a large public university that
recently implemented a new financial aid packaging strategy, which
included frontloading grant money for certain first-year students.\14
The university designated a group of "high-need" first-year students,
who required additional support because they came from economically
or academically disadvantaged backgrounds or both.  After these
students had received Pell grants, Supplemental Education Opportunity
Grants (SEOG), and a small Perkins loan,\15 they received university
grants to cover remaining need.  In the second and later years, their
financial aid packages were weighted with more loans. 

Some of the high-need freshmen were less academically prepared than
the university's average enrollee, officials at the university said,
but we could not identify these students separately in the data. 
Therefore, to measure differences in student academic readiness for
college, our analysis included controls for a student's score on the
Scholastic Aptitude Test.  In addition, program participants received
additional academic and administrative support, such as precollege
course work in their first year\16 and structured advice on course
schedules and financial aid options.  We thus do not atttribute
program outcomes solely to frontloading. 

The university gave us 5 years of data on a cohort of students that
began as full-time, first-year students under the new system in the
1988-89 academic year.  We constructed records on the students for
the 5 years, noting the type and amount of aid received each year and
how long they remained in school.\17 We also had student background
data that remained constant over time. 

The data provided by the university did not indicate whether students
who left before graduation had transferred to another school.  To
identify transfer students, we matched student records with Pell
grant and Stafford loan data supplied by the Department of Education. 
For students who received Pell grants or Stafford loans within three
semesters of leaving the university, we recoded the dependent
variable so that we would not count them as dropouts. 

For our analysis, we selected students whose family incomes in their
senior year of high school were below 300 percent of the poverty
line.  We did this to ensure that students in and out of the
high-need program were somewhat comparable, although those in the
high-need group were still, on average, from poorer families. 


--------------------
\14 Frontloading grant money entails giving students a proportionally
higher amount of grant money in the first year and less grant money,
with proportionally more loans, in later years. 

\15 The Pell grant and SEOG programs are the federal government's two
programs targeting grant funds to low-income students; the Pell grant
program is by far the larger program.  Perkins loans are also
targeted to low-income students, and their interest rate is lower
than that for the much larger Federal Family Education Loan (formerly
Stafford Loan) program. 

\16 Students in the program could take up to two precollege courses
out of four courses in each semester of the first year.  A student
taking the maximum number of these courses would complete one
semester of college credit in the first full year, after which the
student would be required to make standard academic progress. 

\17 We defined a dropout as a student who left school for more than
two consecutive semesters, whether they ultimately returned or not. 
If a student took two semesters off and then returned, that student
was not considered to have dropped out, but if the time off were
three or more semesters, the student was considered a dropout as of
the last semester in school.  Most students who left for more than
two consecutive semesters never appeared again in the data. 


      DURATION ANALYSIS OF THE TWO
      DATABASES
------------------------------------------------------- Appendix I:1.3

Our duration analysis examined the probability of a student's
dropping out in a particular year, given that he or she attended
school up to the beginning of that year.  Duration analysis, also
known as hazard analysis, is typically used to estimate factors that
result in someone's remaining in a particular state (for example,
"unemployed" or "in college") for a short or long period of time.  As
some students leave the database by dropping out, the sample becomes
smaller each period.  For example, in our data, the first-year
dropout probability was computed for all students in the sample, but
the third-year dropout probability was computed only for those who
completed the first 2 years in college. 

Our analysis was a modified hazard model.  A hazard model treats the
length of time as the dependent variable.  In our analysis, we would
have regressed the number of years in school on the explanatory
variables we chose.  However, because we included some independent
variables whose values changed over time, specifically financial aid
levels received each year, this type of hazard model would have been
complicated to construct.  Instead, we set the data up so that each
person-year was an observation.  A student in college for 1 year, who
then dropped out, appeared in the database only once; someone in
school for 4 years appeared as four separate observations.  The
dependent variable in our regressions was whether or not the student
dropped out in a given year.  We used a logit model to analyze the
resulting database.\18

The independent variables of interest were grants and loans.  To see
whether the impact of grants and loans varied by certain factors, we
analyzed subsamples of the database based on income group, race, and
year. 


--------------------
\18 See Paul D.  Allison, Event History Analysis:  Regression for
Longitudinal Event Data, SAGE University Paper No.  46 (Newbury Park,
Cal.:  1984), pp.  14-19, for further details on this methodology. 


   DISCUSSION PANELS WITH
   FINANCIAL AID DIRECTORS AND
   INTERVIEWS WITH STUDENTS
--------------------------------------------------------- Appendix I:2

We judgmentally selected 12 colleges and universities in three areas. 
We chose six public and six private schools, and we selected schools
that varied by such factors as size, tuition, and urbanicity.  The
schools we selected are shown in table I.1. 



                         Table I.1
          
                     Schools GAO Visted

                                                    Tuitio
                                                        n,
                                                     room,
                                        Undergradu     and
                                               ate  board\
School            Location      Sector  population       a
----------------  ------------  ------  ----------  ------
Washington, D.C., area
----------------------------------------------------------
George Mason      Fairfax, Va.  Public      13,351  $8,728
University

George            Washington,   Privat       5,900  $23,76
Washington        D.C.          e                        8
University

Howard            Washington,   Privat       7,668  $11,67
University        D.C.          e                        6

University of     College       Public      23,331  $8,182
Maryland          Park, Md.


Philadelphia, Penn., area
----------------------------------------------------------
Rutgers           New           Public    22,706\b  $8,841
University        Brunswick,                            \c
                  N.J.

Swarthmore        Swarthmore,   Privat       1,387  $24,78
College           Penn.         e                        2

Temple            Philadelphia  Public      18,239  $10,35
University        , Penn.                                6

University of     Philadelphia  Privat       9,969  $24,63
Pennsylvania      , Penn.       e                        8


Seattle, Wash., area
----------------------------------------------------------
Pacific Lutheran  Tacoma,       Privat       2,882  $16,94
University        Wash.         e                        4

Seattle Pacific   Seattle,      Privat       2,272  $16,50
University        Wash.         e                        3

University of     Seattle,      Public      24,938  $5,760
Washington        Wash.

Western           Bellingham,   Public       9,274  $6,227
Washington        Wash.
University
----------------------------------------------------------
\a In-state rate for public institutions. 

\b Total undergraduate enrollment for seven colleges of Rutgers
University in New Brunswick, New Jersey. 

\c Costs for Rutgers College, largest of the seven colleges. 

Source:  Peterson's Guide to Four-Year Colleges--1995 (Princeton,
N.J.:  Peterson's Guides, Inc., 1994). 

To allow interaction between the financial aid directors, we used
discussion panels.  For the students, we thought a discussion or
focus group might inhibit some from telling us about their financial
situations, so we interviewed them individually.  We did not project
from financial aid director or student responses because we knew our
sample was not representative.  Instead, we used the comments to
illustrate some of the thinking that might have led to our
quantitative results. 


      DISCUSSION PANELS WITH
      FINANCIAL AID DIRECTORS
------------------------------------------------------- Appendix I:2.1

We held three discussion panels with financial aid directors,
bringing together the four directors in each region.\19 We asked them
to describe changes in federal aid policy that they had observed over
time and how these changes had affected institutional or other
patterns of financial aid.  We also asked them whether the changes
had affected student decisions to remain in college until graduation. 
Finally, we asked their opinions on whether grants are more or less
effective than loans and their thoughts on how students made the
trade-off between a small grant and a larger loan and on frontloading
grants. 


--------------------
\19 In one city, one director was sick on the day of the discussion
so that panel had only three participants. 


      STUDENT INTERVIEWS
------------------------------------------------------- Appendix I:2.2

The colleges and universities identified students for us to
interview.  We asked them to select both current students and those
who had dropped out, but none of the schools could give us names and
addresses of dropouts.  We did, however, interview some students who
had taken time off and returned to school as well as some transfer
students. 

In the interviews, we asked students to describe the role of grants
and loans in financing their education and in year-to-year decisions
to stay in school.  We asked how they would describe the trade-off
between grants and loans:  that is, in general, did they prefer a
small grant or a larger loan?  We tried to determine whether debt
burden is a major concern and whether it had caused them to
reconsider staying in school, which major to choose, or whether to go
on to graduate or professional school.  We also asked them about
work, either as work-study or an outside job, including the effect on
their studies of spending time at work. 


DETAILED RESULTS OF REGRESSION
ANALYSES
========================================================== Appendix II


   HIGH SCHOOL AND BEYOND ANALYSIS
-------------------------------------------------------- Appendix II:1

Grants significantly reduced dropout probabilities for most groups of
students, including low-income, first-year, and minority students,
according to our analysis of the High School and Beyond database.  On
the other hand, loans reduced dropout probabilities overall and for
middle-income students but not for others. 

The database contained information on the tuition students paid and
their grant and loan awards.  It also contained a wide variety of
student background information, including family characteristics and
academic achievement, which we included as controls in our
regressions.  The variable definitions, as well as means and standard
errors for first-year observations--that is, for the initial sample
before anyone dropped out--appear in table II.1. 



                         Table II.1
          
          Variables Used in Regression Analysis of
                High School and Beyond Data

                                                  Standard
                                                  deviatio
Variable          Description               Mean         n
----------------  ----------------------  ------  --------
Continuous variables
----------------------------------------------------------
Base year test    Score on a test           56.3       7.1
                   administered to
                   sampled students in
                   1980
High school to    Percentage of previous    54.6      22.5
 college           class at student's
                   high school that went
                   to college
Family size       Number in student's       3.52      1.58
                   family
Tuition           Tuition paid in year\a   5.064     4.548
Grants            Grants received in       1.875     3.461
                   year\a
Loans             Loans received in        1.489     1.949
                   year\a
Cumulative loans  Loans received from         \b        \b
                   start of college
                   through previous
                   year\a

Categorical variables (equal 1 if condition is true)
----------------------------------------------------------
Family income     Lowest income (below     0.050     0.219
 grouping (1993    $12,300)
 dollars)
                  Second lowest income     0.078     0.269
                   ($12,300-$21,000)
                  Third lowest income      0.111     0.314
                   ($21,000-$28,100)
                  Middle income            0.132     0.338
                   ($28,100-$35,100)
                  Third highest income     0.162     0.368
                   ($35,100-43,800)
                  Second highest income    0.213     0.410
                   ($43,800-$66,600)
                  Highest income (over     0.253     0.435
                   $66,600)
Urban             Student went to high     0.205     0.404
                   school in an urban
                   area
Sex               Male                     0.459     0.498
                  Female                   0.541     0.498
Parents college   At least one parent      0.257     0.437
                   graduated from
                   college
Good grades       Student received As      0.806     0.396
                   and Bs in high school
Region of United  Northeast                0.281     0.449
 States in which
 student
 attended high
 school
                  South                    0.266     0.442
                  Midwest                  0.324     0.468
                  West                     0.130     0.336
Race              African American         0.090     0.286
                  Hispanic                 0.030     0.171
                  White                    0.860     0.347
                  Other race               0.020     0.140
----------------------------------------------------------
\a Thousands of 1993 dollars. 

\b Equals zero for all first-year observations. 

Before conducting our regression analysis, we examined
crosstabulations of certain variables with dropouts, the outcome
variable.  Low-income students were more likely to drop out of
college than middle- and high-income students.  In addition, in our
sample, second-year students were more likely than first- or
third-year students to drop out.  We also examined those who dropped
out in the first year to determine their income group; low-income
students were again the most likely to drop out.  The sample dropout
probabilities are shown in table II.2. 



               Table II.2

   Sample Probability of Dropping Out

