Institutional Retention Strategies at Historically Black Colleges and Universites and Their Effects on Cohort Default Rates

N O V E M B E R 1 9 9 9

M O N O G R A P H S E R I E S

Institutional RetentionStrategies at Historically

Black Colleges andUniversities

and Their Effects onCohort Default Rates:

1987 - 1995By Fred J. Galloway

Associate DeanSchool of Education

University of San Diego

Watson Scott SwailAssociate Director for Policy Analysis

The College Board

This publication provided courtesy of

epiepi

Educational Policy Institute Washington, DC

Watson Scott Swail, Ed.D. President

www.educationalpolicy.org

Improving educational policy & practice through research

Washington Office • 25 Ludwell Lane • Stafford, VA 22554 • 1 (877) e-POLICY Los Angeles Office • Occidental College • 1600 Campus Road • Los Angeles, CA 90041 • 1 (877) e-POLICY

Institutional Retention Strategies at Historically Black Colleges and Universities and Their Effects on Cohort Default Rates: 1987 - 1995

By Fred J. GallowayAssociate DeanSchool of EducationUniversity of San Diego

Watson Scott SwailAssociate Director for Policy AnalysisThe College Board

Published by the Sallie Mae Education Institute901 E Street NWWashington, DC 20004

November 1999

Copyright © 1999 Sallie Mae Education Institute, all rights reserved.Portions of this Monograph may be reproduced with written permission from the publisher.

The Sallie Mae Education Institute (SMEI) is a non-profit corpo-ration that focuses its attention on issues in higher educationfinance, especially student and family ability to pay for postsec-ondary education. SMEI conducts research and publishes find-ings, works with higher education leaders to offer creative solutionsto problems, and furnishes objective analyses to policy-makers andother interested parties.

A B O U T T H E A U T H O RS

Fred J. Galloway is currently associate dean and tenure-track associateprofessor in the School of Education at the University of San Diego. Prior tojoining the University faculty, he was project director for the national DirectStudent Loan Evaluation project at Macro International, as well as director offederal policy analysis at the American Council on Education in Washington,D.C., where he represented the interests of the higher education communitybefore the Executive and Legislative branches of the federal government. Dr.Galloway received his bachelor’s and master’s degrees in economics from theUniversity of California, San Diego, and his doctoral degree in the econom-ics of education from Harvard University.

Watson Scott Swail is associate director for policy analysis of The CollegeBoard, where he provides data and analysis on issues relating to three prima-ry areas: academic preparation, access to college, and postsecondary success.In addition to his research projects, he directs and produces The CollegeBoard’s annual Trends in Student Aid, Trends in College Pricing, and Trendsin Academic Preparation. Dr. Swail is also adjunct professor in theEducational Leadership Department at George Washington University. Hereceived his bachelor’s degree from the University of Manitoba, his master’sdegree from Old Dominion University, and his doctorate in educational poli-cy from George Washington University.

As college prices have risen, so has student borrowing to pay for them. These growthpatterns have led to concerns about borrower defaults, as not all students successful-ly complete their education programs and, consequently, some default on their loans.Defaults cost the taxpayers, the government, and the lenders. They also cost the stu-dent borrowers a great deal of money in lost credit and related negative effects ontheir lives. Thus, there is great interest in studying the causes of defaults and usingthis research to reduce default costs and probabilities. The literature of student finan-cial aid abounds with research reports on defaulter characteristics and how they dif-fer from student borrowers who do not default, accounts of successful actions takenby loan program administrators to avert defaults, and distinguishing characteristics ofinstitutions whose students have above average default rates.

The literature contains very little discussion of the relationships between what insti-tutions do, particularly how they allocate their resources among manifold activities,and the effects their acts might have on student borrowers' repayment experiences.This study employs sophisticated statistical methods to assess the effects of institu-tional spending (resource allocation) in such areas as instruction, academic supportservices, student services, and student financial aid, on students' subsequent cohortdefault rates. The subjects of this research are colleges whose students have experi-enced above-average default rates in federal loan programs, the Historically BlackColleges and Universities. These colleges were chosen as subjects for two reasons.First, there are great individual differences in their students' default rates, theresources they have to apply to general and education expenditures, and in the waysthey divide funds among various program activities. Second, some of these collegesare in danger of losing institutional eligibility to participate in the federal studentloan programs because of their students' higher default rates. So the researchershoped that their analyses would help the colleges to address defaults and the rootcause of defaults, student attrition, more effectively.

The Sallie Mae Education Institute is pleased to publish this important study. Webelieve it offers institutions guidance in how they might better allocate resources toreduce their students' default rates. And we believe it offers student defaultresearchers ideas and hypotheses to explore when they conduct their next studies.There is still much to learn.

Jerry S. DavisPresident

Sallie Mae Education InstituteNovember 1999

F O R E W O R D

During the latter half of the 1980s, annual costs to the federal government for defaultsin the nation’s largest student loan program, the Federal Family Education LoanProgram (FFELP), began to soar. Between 1984 and 1986, annual default costs near-ly doubled, rising from about $712 million to over $1.3 billion, before reaching a highof $3.2 billion in federal fiscal year (FFY) 1991 (Conner, Saab, and Cicmanec, 1997).Congress, the federal government, postsecondary education institutions, lenders, andall FFELP participants became concerned about the threat default costs represented tocontinued support of student loans. During those years, and in the years hence, theseparties cooperated in many ways to help reduce default costs and suppress defaultprobabilities. One of those strategies was to use borrower cohort default rates torestrict institutional participation in federal student aid programs.

Unfortunately, America’s Historically Black Colleges and Universities (HBCUs) havemuch higher cohort default rates than the national average. At the time of the 1998Higher Education Act amendments, about one-third of the HBCUs had cohort defaultrates above the legislated threshold rate. Therefore, many of those colleges were indanger of losing eligibility to participate in the federal loan programs. A few remain insuch danger today.

This study focuses on an analysis of the factors that significantly affect the cohortdefault rates at HBCUs, and on the potential for reducing these rates. It is hoped thatfindings from the analysis will help pave the way for further analysis, and ultimatelydevelop a better understanding of which institutional practices have greater success inreducing the higher default rates for these colleges and universities.

I N T R O D U C T I O N

M O N O G R A P H S E R I E S 1

BackgroundThree discoveries contributed to the implemen-

tation of institutional cohort default rates as eligibil-ity criteria. First, the data on default paymentsshowed that a disproportionate number of defaultingborrowers had attended proprietary (private, for-prof-it) business, trade, and technical schools (Merisotis,1988; General Accounting Office, 1989). Second, itwas discovered that many defaulting borrowers spentonly a year or two in postsecondary education anddropped out before completing their academic pro-grams (Beanblossom and Rodriguez, 1989). It wasconcluded that colleges and schools were at least par-tially responsible for borrower defaults. The logicwas as follows: If schools enrolled only students whohad the ability to benefit from their curricula, ifschools offered programs and services that helpedretain admitted students until completion of theireducation goals, and if schools helped their studentsfind satisfactory employment after graduation, thenstudent borrowers would be much less likely todefault on their loans. Congress reasoned that if theevidence indicated that the schools and collegescould not do these things, then their students shouldno longer be eligible for federal student loans.

A third discovery, that the vast majority ofdefaults occur in the first year or two after borrowersleave school or college (Davis and Knapp, 1990), ledCongress to consider their students’ cohort defaultrates as primary evidence that certain postsecondaryinstitutions could not successfully serve borrowers.The term “cohort default rate” was first defined inthe Omnibus Budget Reconciliation Act of 1990(P.L. 101-508) as the percentage of students at a par-ticular institution who enter repayment during a fis-cal year who default before the end of the followingfiscal year.

For example, here is the formula for the mostrecent cohort default rate (CDR) available, the onefor FFY 1997:

The Number of Students Who EnteredRepayment in FFY 1997 and Defaulted Beforethe End of FFY 1998 (Numerator) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _x100= CDR

The Number of Students Who Entered Repayment in FFY 1997 (Denominator)

Using such a simple measure as evidence of suc-cess with students as a criterion for continued partici-

pation in the federal student loan programs is notwithout complications. First, many factors that arebeyond the control of a school or college contributeto defaults on student loans. For example, a borrow-er’s unwillingness to seek any job or a job that earnsenough to meet loan payments, or a borrower’s deci-sion to spend earnings on consumer goods ratherthan loan payments, can lead to defaults. Changesin the economy or local job market can inhibit bor-rowers’ employment opportunities and subsequentability to repay their student loans.

Second, many institutions are committed to thenational goal of providing access to postsecondaryeducation and training to all who might benefit fromit. However, it is not always easy for such institutionsto determine which applicants will benefit.Sometimes institutions that are committed to theaccess goal admit and enroll students who are unableto meet their curricular requirements. Such studentsmay drop out and, because their education and train-ing were not successful, may have difficulty findingjobs. If these students were borrowers, they maydefault on their loans. Other “open door” institu-tions, such as community colleges, that are commit-ted to offering all students the opportunity to try tosucceed in postsecondary education will probablyenroll some who will fail, drop out, and subsequentlyhave difficulty repaying a student loan.

Third, the mission of some institutions is to servepopulations of students whose socioeconomic status-es, previous education and training, and lack of pre-paredness make it difficult for all of them to succeed,regardless of what the institutions do for them. Thustheir student borrowers very likely will have higherdefault rates than will the student borrowers at elite,selective colleges. Only one of the HBCUs in thestudy sample is selective in its admissions practices.

Congress was faced with choosing a specificcohort default rate which would not discriminateagainst colleges and schools that were trying toachieve laudable goals and sometimes failed to do so,and yet a rate which would restrict institutions whowere doing disservice to their student borrowers.Congress decided that the threshold for loss of insti-tutional eligibility would be a cohort default rate of35 percent for FFY 1991 and FFY 1992, 30 percentfor FFY 1993, and 25 percent for any succeeding year.

Because of their special historical mission, andthe characteristics of the students they serve, theHBCUs were exempted from these limitations until

M O N O G R A P H S E R I E S2

July 1, 1993. The Higher Education AmendmentsAct of 1992 extended that exemption to July 1,1998. (The loss of eligibility sanctions also does notapply to tribally controlled community colleges andNavajo community colleges.) The current law man-dates that a school with an official cohort default rateof 25 percent or greater for the three most recent fed-eral fiscal years (in this case, FFY 1995, FFY 1996,and FFY 1997) lose eligibility to participate in theFFELP and the Direct Loan Program [HigherEducation Act, Section 435(a)(2)]. Such a schoolbecomes ineligible to participate in the FFELP and/orDirect Loan program for the current and two suc-ceeding fiscal years, unless the institution successful-ly appeals its loss of eligibility.

Were it not for the exemption, when thisresearch was begun 18 months ago, fourteen HBCUswould have lost their eligibility to participate in thefederal student loan programs because their cohortdefault rates for FFY 1993, FFY 1994, and FFY 1995each exceeded 25 percent (General AccountingOffice, 1998). Another nineteen institutions were insome potential danger of losing eligibility becausetheir cohort default rates exceeded 25 percent in oneor two of the three most recent fiscal years. Stillanother 13 institutions had FFY 1995 cohort defaultrates in excess of the Department of Education’s“cautionary” level of 20 percent.

Many HBCUs have achieved success in loweringtheir cohort default rates. For example, nineteenHBCUs whose FFY 1988, FFY 1989, and FFY 1990rates exceeded the statutory threshold of 25 percenthave fallen below the threshold for the most currentcohort years (General Accounting Office, 1998).However, many have not done as well as Congressinitially hoped they would. Their students continueto default at a much higher rate than students at mostother types of postsecondary institutions. The aggre-gate 1995 student loan cohort default rate for HBCUswas 18.5 percent; the rate for non-HBCUs was only7.8 percent (General Accounting Office, 1998).

Congress and the federal government remainconcerned about cutting federal budget costs. Due tothis concern, the recently-enacted Higher EducationAmendments of 1998 (P.L. 105-244) contains a pro-vision which extended the exemption for HBCUs forjust one year (to July 1, 1999). After this date, allHBCUs and tribally-controlled colleges that havehad cohort default rates of 25 percent or higher forthe past three federal fiscal years must submit defaultmanagement plans to the U.S. Secretary of

Education in order to remain eligible for participa-tion. Thirteen HBCUs were required to submit theseplans. There is reason to worry that these HBCUswill no longer be able to provide federal loans to theirstudents if their default rates remain high.

That most HBCUs have successfully loweredtheir students’ cohort default rates indicates that theychanged their policies, programs, and practices toachieve that goal effectively. Research has shownthat only 5 percent of all student borrowers whoreceive their degrees default on their FFELP loans(Davis, 1996). For this reason, it was hypothesizedthat the HBCUs that have been most successful inlowering their students’ cohort default rates werethose that were also successful in increasing their stu-dent retention rates. Put another way, if their cohortdefault rates went down, then the colleges’ retention(and graduation) rates must have gone up.

Swail’s Conceptual Framework for StudentRetention (Swail, 1995) was used to help identifyand assess which factors might have contributed tosuccess in reducing cohort default rates. In doing so,it was assumed that factors positively related toimprovements in retention are positively related toimprovements in default rates. The research goal wasto identify and quantify primary factors that are sta-tistically associated with cohort default rate reduc-tions. The intention is that identifying these factors

at colleges that have successfully cut their cohortdefault rates will lead to potential replication byother relatively unsuccessful institutions.

For example, if it is determined that a certainbase level of per student expenditures on academicsupport services is necessary to achieve reductions incohort default rates, then colleges must be able toachieve that level to expect to meet the federal cri-teria. Or, for another example, if it is determined thatan increase in expenditures on student support ser-vices is more closely related to a reduction in cohort

Figure 1. Five components of Student RetentionFramework (Swail, 1995)


default rates than an increase in expenditures on aca-demic support services, then HBCUs will know betterhow to target limited resources to reach their defaultreduction goals. It is also intended that the researchfindings will help federal policymakers understandwhether it is realistic to expect every single HBCU toreach the target cohort default rates for all schoolsparticipating in the FFELP and Direct Loan program.

The Student Retention ModelSwail’s Conceptual Framework for Student

Retention (1995) focuses on the barriers and issuesfacing minority students at HBCUs and other minor-ity-serving institutions. Originated from a synthesisof previous retention research, specific retentionstrategies were then formed into a five-part frame-work. A validation process by way of delphi tech-nique resulted in the final framework, whosecomponents are presented in Appendix A. The fivebasic components include student services, academicservices, curriculum and instruction, financial aid,and admissions and recruitment.

Swail’s framework pulls together many of thewidely-held tenets of minority student retention, pre-viously illustrated through landmark studies by Tinto(1975), Astin (1977), Beal and Noel (1980), Bean(1986), and others. While some may argue the

specifics of the models developed by these and otherresearchers over the years, all subscribe to a similarphilosophy about the problems facing disadvantagedstudents in higher education and the basic strategiesto increase student retention.

Figure 1 illustrates how the five components ofthe conceptual framework act as a basic foundationfor student retention at a college or university. Whilethe geometric frame simplifies the nature of the stu-dent experience in higher education, it does offer avisual interpretation of how students and institutionsinteract.

Figure 2 shows the relationship between theinstitutional factors or practices of the college andthe academic and social skills students bring withthem to college. If one conceptualizes the triangle asthe student college experience, one can thinkabstractly about the various factors that may impactthat experience. Academic preparation, readiness forcollege, maturity, social awareness, ability to getalong, and the myriad of other developmental issuesall have some impact on a student’s ability to persistin school. While the factors presented in Figure 2 arenot exhaustive, they do provide some idea of whatstudents bring with them from high school and/orthe workforce.

Figure 2. Factors impactingupon the geometricmodel of student per-sistence


The ultimate responsibility of the insti-tution is to provide a quality education forits students. Through each of the five com-ponents of the framework, institutions havean opportunity to build a support system tohelp students stay in school and receive adegree. For example, an institution mayprovide specialized orientations (studentservices) to help freshman students accli-mate to their new surroundings, or providetutoring and study assistance opportunities(academic services) to assist the academicdevelopment of their students. (SeeAppendix A for details of the originalframework).

One challenge facing HBCUs is the setof issues students bring with them to cam-pus. It was mentioned earlier in this paperthat the mission of many HBCUs is to servepopulations of students whose backgroundsmay make it difficult to succeed. Generallyspeaking, the academic preparation of stu-dents attending HBCUs is not commensu-rate with their peers at other institutions.These students, on average, have not takenthe same rigorous course work in highschool, have not been exposed to high-leveleducation, have not always had the supportof family and community, and lack some ofthe general skills required to excel in post-secondary education (e.g., study, time man-agement, etc.). Additionally, researchsuggests that these students also have a dif-ficult time making the social adjustment topostsecondary education.

These factors alone provide a formida-ble barrier to success for many students.However, academics represent only part ofthe challenge. Many of the students attend-ing HBCUs require financial aid to attend.How students and their families pay for col-lege varies widely depending on their acad-emic ability, current income status, andavailable savings or other financialresources. Low-income students may be eli-gible for Pell Grants, but these provide onlymodest support at most colleges. Theremainder is often made up by loans, manyof which are subsidized and offered throughthe federal government. In 1996, about 53percent of undergraduate students at

HBCUs received Stafford Loans of $3,629, on aver-age. While average tuition costs at HBCUs are lowerthan national averages, many students will haveaccumulated loan debt in excess of $10,000.1

For graduates of HBCUs, repaying their loans isnot particularly difficult. Most will find jobs that willallow them to begin repayment soon after gradua-tion. And while no one likes to repay loans, a$10,000 debt is not a formidable barrier to most grad-uates. The bigger problem is with students who leavecollege before completing their degree programs.These students do not have the advantages affordedbachelor’s degree recipients, including increasedskill, knowledge, and a degree to flaunt around thejob market. In a society that rewards credentialism,the degree itself—regardless of skill in many cases—is valued by employers. Unfortunately, students with-out degrees will have a much more difficult timefinding a job (let alone a decent paying job) thanthose with degrees. Still, the loan debt persists, andmany of these students face a difficult situation:choosing between paying rent or paying their studentloan. Understandably, repaying student loans quicklybecomes of secondary importance.

It was hypothesized that the answer to decreasinginstitutional cohort default rates lies in increasingstudent persistence to a degree. While persistencerates are affected by admissions policies, what insti-tutions do with their students once on campus maybe the only direct strategy of retaining students fromyear-to-year.

Research Design and MethodologyA mixed time-series/cross sectional model was

used to examine the effectiveness of various institu-tional strategies designed to reduce cohort defaultrates. The pooled model contained data for 80HBCUs2 over the entire period for which cohortdefault rates have been calculated by the U.S.Department of Education (1987 - 1995).3 The modelitself can be thought of as a series of cross-sectionalmodels over time describing changes in both cohortdefault rates and the institutional strategies designedto reduce them. The financial and enrollment datafor the HBCUs came from the U.S. Department ofEducation’s Integrated Postsecondary EducationData System (IPEDS) annual finance and fall enroll-ment surveys, the undergraduate admissions selectiv-ity data came from the 1998 Peterson’s Guide toFour-Year Colleges, and the cohort default rates wereprovided by the U.S. Department of Education.


For each of the five strategic areas identified inSwail’s retention framework, there were at least atleast two conceptual ways in which each strategycould be measured. For those strategies best repre-sented through institutional expenditures, a “bud-getary shares” and a “spending per FTE” frameworkwas developed. In this manner, four expenditure-based strategies (academic services, student services,instruction, and financial aid in the form of grants)could all be measured as either their respective shareof total educational and general expenditures in agiven year, or in terms of annual expenditures per full-time equivalent student. To test the inferentialrobustness of the theory, estimates were developedunder both frameworks. For the fifth and final type ofstrategy, institutional selectivity, an ordinal measuredescribing the selectivity of the institutions’ admis-sions policies based on the SAT scores of admitted

freshmen was developed. The precise definitions ofthe variables, along with their sources, are presentedin Table 1.

These measures were then used to first estimatethe effects of each one of the institutional strategiesindependent of the other strategies, beginning byusing ordinary least squares to linearly regress eachone of the measures against the institutional cohortdefault rates. However, since it takes time for theseinstitutional changes to effect an institution’s cohortdefault rate, lag structures of between one and fouryears were tested, to represent the length of time itmight take for a change to work its way through thesystem.

After measuring the magnitude of the individuallinear effects, tests were conducted for non-linearitiesin the estimated effects. Since in particular, many of

Table 1Variable Definitions and their Sources

Variable Name Definition Source

CDF Cohort Default Rate U.S. Department of Education

ISPCT Percentage of Total Educational and General IPEDSExpenditures Devoted to Instructional Services

ISFTE Expenditures on Instructional Services Per IPEDSFull-Time Equivalent Student

ASPCT Percentage of Total Educational and General IPEDS Expenditures Devoted to Academic Services

ASFTE Expenditures on Academic Services Per IPEDS Full-Time Equivalent Student

SSPCT Percentage of Total Educational and General IPEDS Expenditures Devoted to Student Services

SSFTE Expenditures on Student Services Per Full-Time Equivalent Student IPEDS

AIDPCT Percentage of Total Educational and General IPEDS Expenditures Devoted to Financial Aid

AIDFTE Expenditures on Financial Aid Per Full-Time Equivalent Student IPEDS

SELECT Categorical Variable Describing the Selectivity of the Institutional Admissions Policy (Noncompetitive, Minimally

Difficult, Moderately Difficult, Very Difficult, Most Difficult ) Peterson’s Guide


the expenditure-based strategies are hypothesized tohave their greatest effects at low levels of spending,several functional forms designed to capture this rec-iprocal effect were tested. Together, the linear andnon-linear effects were then used to estimate boththe relative size of the effects produced by each insti-tutional strategy, as well as to identify the most effi-cient lag structure for each strategy.

In the final part of the analysis, the individualeffects were combined into a more general modeldescribing all five of the institutional strategies. Todetermine the optimal lag structure and specificationfor the final “budgetary shares” and “expenditure perFTE” models, the results from the first two parts ofthe analysis were used to sequentially narrow therange of acceptable lag structures and specifications.

This sequencing was done until it created a set ofmodels that produced the same inferences andexplained the same variation in cohort default rates.From among these few models, ones with con-formable lag structures were selected to allow com-parisons of the results of the two conceptual models.

ResultsThe first part of the analysis involved estimating

the effects of each one of the institutional strategiesindependent of the other strategies. Each one of the“budgetary shares” and “spending per FTE” variableswas linearly regressed against the cohort default rates.The elasticities, significance levels, lag structures,and explained variation for both models are present-ed in Tables 2 and 3.

Examination of these tables shows that three of

Table 3 Elasticities, Lag Structures, and Explained Variation for the

Independent Variables in the Linear Expenditure per FTE Model

Elasticity Significance Optimal Lag Range of Explained Level Structure Variation (R2)

Academic Support -.119 1% 1 Year 3% - 5%

Student Services .040 Insignificant 4 Years 0% - 0%

Instructional -.294 1% 4 Years 14% - 15%

Grants .253 1% 4 Years 8% - 11%

Admissions -.147 1% 4 Years 12% - 14%


Independent Variables in the Linear Budgetary Shares Model

Elasticity4 Significance Optimal Lag Range of Explained Level Structure Variation (R2)


Student Services .543 1% 4 Years 1% - 3%

Instructional -.412 1% 4 Years 12% - 14%

Grants .528 1% 4 Years 19% - 21%


per FTE results in a smaller decrease in the institu-tion’s cohort default rate (.29). To a large extent, thesame holds true for academic expenditures: a 1 per-cent increase in the share of the budget devoted toacademic services results in a decrease of slightly lessthan half a percentage point (.39) in the institution’scohort default rate in one year, while a similarincrease in academic expenditures per FTE results ina smaller decrease in the institution’s cohort defaultrate over the same time period (.12). For both mod-els, tightening up admissions policies by admittingmore “select” students results in a decrease in theinstitution’s cohort default rate by .15 percent overfour years.

Two strategies result in higher cohort default rates.

the institutional strategies led to decreases in theinstitution’s cohort default rates, and the other twoled to increases or had no effect at all. The threestrategies that led to decreases involved tighteningup admissions policies, increasing spending on acad-emic services, and increasing spending on instruc-tional services. For all three of these institutionalstrategies, the effect sizes are modest, and take frombetween one and four years to work their waythrough the system.

For example, a 1 percent increase in the share ofthe budget devoted to instructional services results ina decrease of slightly less than half a percentage point(.41) in the institution’s cohort default rate in fouryears. A 1 percent increase in instructional services



Independent Variables in the Non-Linear Budgetary Shares Model




Instructional -.385 1% 1 Year 13% - 19%

Grants .313 1% 4 Years 15% - 16%


Table 5 Nonlinear Effect Sizes, Lag Structures, and Explained Variation

For the Independent Variables in the Expenditure per FTE Model


Academic Support -.084 1% 3 Years 3% - 5%


Instructional -.236 1% 1 Year 9% - 17%

Grants .234 1% 4 Years 6% - 9%



These are increasing expenditures on student servicesand financial aid, both of which take four years towork through the system and, in the “budgetaryshares” model, result in slightly greater than half apoint increases in an institution’s cohort default rate(.54 for student services and .53 for financial aid).However, in the “expenditure per FTE” model, thereis no effect associated with increases in student ser-vice expenditures, and a more modest increase incohort default rates is associated with a 1 percentincrease in expenditures on grants per FTE (.25).

Although these results are only preliminary, onepossible explanation for the positive, and possiblycounter-intuitive, effect of financial aid may be thatincreasing expenditures on grants encourages moremarginal students, who are heavily dependent onloans as well as grants, to enroll and then subse-quently drop out. As for the positive results associat-

ed with student services, perhaps increases in thisshare of the budget are spent on non-academic ser-vices that actually decrease the student’s chances ofcompleting their degree (e.g., expenditures on frater-nities and sororities, or for purely social events).

In the next stage of the analysis, non-linearitieswere tested for in ways that institutional strategiesworked to change the institution’s cohort defaultrates. This was done because many of the expendi-ture-based strategies were hypothesized to have theirgreatest effect at low levels of spending, and this sortof reciprocal effect can only be captured through theuse of non-linear functional forms. This informationis presented in Tables 4 and 5, and suggests that, forseveral strategies, the non-linear approach explainsmore variation in cohort default rates than does thelinear specification. For example, in the “budgetaryshares” model, the reciprocal specification associated

Table 6 Specifications, Elasticities, Significance Levels, and Lag Structures for

The Independent Variables in the Full Budgetary Shares Model (R2 = .41)

Specification Elasticity Significance Level Lag Structure

Academic Support Reciprocal .032 Insignificant 1 Year

Student Services Linear .426 1% 4 Years

Instructional Reciprocal -.269 1% 1 Year

Grants Linear .341 1% 4 Years

Admissions Linear -.001 1% 4 Years

Table 7 Specifications, Elasticities, Significance Levels, and Lag Structures for

The Independent Variables in the Full Expenditures per FTE Model (R2 = .36)

Specification Elasticity Significance Level Lag Structure

Academic Support Reciprocal .038 5% 1 Year

Student Services Linear .070 10% 4 Years

Instructional Reciprocal -.258 1% 1 Year

Grants Linear .248 1% 4 Years

Admissions Linear -.001 1% 4 Years


with instructional expenditures explains between 14and 19 percent of the variation in cohort defaultrates, while the linear specification explains onlybetween 12 and 14 percent. Similarly, the reciprocalspecification for academic support explains between2 and 3 percent of the variation in cohort defaultrates, compared with between 1 and 2 percent for thelinear specification.

In the “expenditures per FTE” model, the non-linear approach represented an improvement overlinear specifications for several of the institutionalstrategies. In particular, the reciprocal specificationassociated with instructional expenditures, academicsupport expenditures, and student service expendi-tures all produced models that explained more of thevariation in cohort default rates than did their linearcounterparts. Although the student services strategyseems to have little overall effect on cohort defaultrates, increases in both academic support expendi-tures and instructional expenditures seem to havetheir greatest effect at low levels. This suggests thatinstitutions that spend relatively little in academicsupport and instructional areas should be able to seesignificant reductions in their cohort default rates byincreasing funding in them.

Taken together, the results of the first two partsof the analysis suggest that the strongest effects, bothin terms of explained variation and size of effects, arefor spending on grants and instructional services.However, they operate in different directions. A 1percent increase in the share of the budget devotedto grants causes about half a percentage pointincrease in an institution’s cohort default rate (.53).A similar increase in the share of the budget devotedto instructional expenditures produces slightly lessthan half a percentage point decrease in an institu-tion’s cohort default rate (.39).

Although expenditures in both academic ser-vices and student services are associated with fairlylarge effect sizes, they explain a relatively smallamount of the variation in cohort default rates. Theinstitutional selectivity measure explains a relative-ly large amount of variation in cohort default rates,but its effect size is quite small. As such, it appearsthat increased spending on instructional services,especially at low levels, seems to offer the bestchance for HBCUs to reduce their cohort defaultrates effectively.

However, before making any conclusive state-ments about the size of the effects associated with the

five institutional strategies, it is appropriate to com-bine these effects into a more general model describ-ing all five of the institutional strategies. As wasnoted above, the results from the first two parts of theanalysis were used to narrow the range of lag struc-tures and specifications to arrive at final “budgetaryshares” and “expenditure per FTE” models. Thesemodels have the same lag structures and specifica-tions and are presented in Tables 6 and 7.

Examination of these tables yields some inter-esting and surprising findings. Although spending onboth grants and instructional services still yields thestrongest and most robust effects, their magnitude isdiminished when all of the other institutional strate-gies are statistically controlled for. Specifically, a 1percent increase in the share of the budget devotedto grants increases an institution’s cohort defaultrate by about a third of a percent (.34). At the sametime, a 1 percent increase in the share of the budgetdevoted to instructional services decreases an insti-tution’s cohort default rate by slightly less than athird of a percent (.27). As expected, similarincreases in expenditures per FTE for both of thesestrategies produce smaller changes than those associ-ated with the 1 percent increase in the share of thebudget -- specifically, .25 percent for grants and .26percent for instructional services. However, from apolicy perspective, increasing spending on instruc-tional services produces a declining cohort defaultrate within a year, while increasing spending ongrants takes four years to produce an increase intheir cohort default rate.

The results for the three other institutionalstrategies are mixed. For example, although increas-es in admissions selectivity are responsible for signif-icant decreases in cohort default rates, the size of theeffect is so small as to be of no practical use to insti-tutions struggling with high cohort default rates.Put another way, tightening up admissions criteria isunlikely to significantly cut default rates. Whenexpenditures on academic support are included inthe full model, the effect size washes out. This sug-gests that other institutional strategies are explain-ing the same thing. Finally, the results on studentservice expenditures are mixed, but they do suggestthat increases in this type of spending may indeedlead to increases in cohort default rates four yearsdown the line.

9


Conclusions and Policy ImplicationsFor several year some HBCUs have been strug-

gling with persistently high cohort default rates.Despite the fact that many have been successful inreducing their default rates below the 25 percentthreshold that applies to most other types of institu-tions, some HBCUs remain in danger of losing theireligibility to participate in the federal student loanprograms. Since students who drop out of collegehave been shown to default at much higher ratesthan those who persist, this paper has examined sev-eral retention strategies available to HBCUs to seewhat effects they might have on lowering the insti-tutions’ cohort default rates.

Through the use of both a “budgetary shares” and“expenditure per FTE” model, the effectiveness ofseveral institutional strategies designed to increaseretention were examined to see what effect, if any,they may have on an institution’s cohort defaultrates. Although the results were mixed, severalimportant facts surfaced.

First, increases in such non-academic compo-nents of an institution’s budget as grants and studentservices are statistically associated with increases inthe cohort default rates. These findings suggest thatincreases in grant spending may actually encouragemarginal students heavily dependent on grants andloans to enroll and then subsequently drop out, andthat spending increases on student services may helpproduce a campus climate that actually reduces stu-dents’ chances of completing their degrees.

Second, increases in at least one academic com-ponent of an institution’s budget, instructional ser-vices, are associated with lower cohort default rates,which suggests that spending more on instructionalservices may help motivate students to persist indegree completion. Furthermore, since increases inspending on instructional services have their biggesteffect on cohort default rates at relatively low levelsof spending, the HBCU’s that are struggling withhigh cohort default rates while devoting a relativelysmall share of their budgets to instructional servicesare strongly urged to increase spending in this area.

Finally, the institutional strategies examined inthis study together explained about 40 percent of thevariation in cohort default rates. This suggests thatinstitutional decision-making about allocation ofresources among different alternative functionsaffects cohort default rates. But student-based char-acteristics and other factors, not to mention chance,

are left to explain the majority of the variation incohort default rates.

Taken together, the results of this study shouldprovide new insights for many HBCUs strugglingwith stubbornly high cohort default rates. Althoughonly one institutional strategy, increasing expendi-tures on instructional services, was especially effec-tive for both increasing retention and reducingcohort default rates, the fact that the biggest “bang”for the instructional services “buck” comes at lowlevels of instructional services spending should helpmany HBCUs reduce their cohort default rates.

For policymaking purposes, it is important tonote that all the strategies available to HBCUs toincrease retention explain less than half of the varia-tion in HBCU cohort default rates. This suggeststhat many HBCUs may be doing all they can undertheir present circumstances to keep their students inschool and to handle their default rates. Therefore,it seems that Congress would be well advised to allowthese institutions to continue to participate in thefederal student loan programs, allowing them to servetheir students.

10


APPENDIX A-1Institutional Components for Student Persistence

12



13



14



15



16


APPENDIX B

Instructional Expenditures: This category includes general academic instruction, occupational and vocationalinstruction, special session instruction, community education, preparatory and adult basic education, and remedi-al and tutorial instruction conducted by the teaching faculty for the institution’s students.

Academic Support Expenditures: This category includes expenditures for the support services that are an inte-gral part of the institution’s primary mission of instruction, research, or public service, including expenditures forlibraries, museums, galleries, audio/visual services, academic computing support, ancillary support, academic admin-istration, personnel development, and course and curriculum development.

Student Services Expenditures: This category includes funds expended for admissions, registrar activities, andactivities whose primary purpose is to contribute to students’ emotional and physical well-being and to their intel-lectual, cultural, and social development outside the context of formal instruction, including expenditures forcareer guidance, counseling, financial aid administration, and student health services.

Scholarships and Fellowships: This category includes all expenditures given in the form of outright grants andtrainee stipends to individuals enrolled in formal course work, as well as aid to students in the form of tuitionremission.

17

REFERENCES

Astin, A.W. 1977. Four Critical Years. San Francisco, CA: Jossey-Bass, Inc.

Beal, P.E. and L. Noel. 1980. What Works in Student Retention.Iowa City, IA: American College Testing Program.

Bean, J. P. 1986. “Assessing and Reducing Attrition.” New Directions for Higher Education 14(1): 47-61.

Beanblossom, G. and B. R. Rodriguez. 1989. Characteristics of Stafford Loan Program Defaulters: A National Study.Washington: U.S. Department of Education, Office of Student Financial Assistance.

Conner, D, R. Saab, and K. Cicmanec. 1997. Federal Student Loans Programs Data Book, FY 94 - FY 96. Washington: U. S. Department of Education.

Davis, J. S. 1990. Correspondence with Gisela Vallandigham, August 30.

Davis, J. S. And L. G. Knapp. 1990. “How Federal Subsidies to the Stafford Loan Program Are Distributed Among Pennsylvania Borrowers.” Journal of Student Financial Aid, 20(3): 24-43.

General Accounting Office. 1989. Guaranteed Student Loans: Analysis of Student Default Rates at 7,800 Postsecondary Schools. GAO/HRD-89-63BR. Washington: U.S. General Accounting Office, 1989.

______________________. 1998. Student Loans: Characteristics of Students and Default Rates at Historically Black Colleges and Universities.GAO/HEHS-98-90. Washington: U.S. General Accounting Office.

Merisotis, J.P. 1988. "Default Trends in Major Postsecondary Education Sectors." Journal of Student Financial Aid. 18 (3): 18-28, .

Swail, W. S. 1995. “The Development of a Conceptual Framework to Increase Student Retention in Science, Engineering, and Mathematics Programs at Minority Institutions of Higher Education.” Ed.D. dissertation, The George Washington University.

Tinto, V. 1975. “Dropout from Higher Education: A Theoretical Synthesis of Recent Research” Review of Educational Research, 45: 89-125.

ENDNOTES

1 Data extracted from the National Postsecondary Student Aid Study (NPSAS) 1995-96.2 Although there were 98 HBCUs in existence during this study period, this analysis was lim-

ited to the 80 HBCUs that had four-year undergraduate programs.3 Given the lag structure in the general model, non-cohort default rate data were also collected

for the 1984 to 1987 period.4 The elasticities that appear in Tables 2-7 are from the models with the greatest explanatory

power. However, since the “admissions” variable is categorical, the concept of an elasticity is unde-fined. As such, this number represents the reduction in the default rate associated with movingfrom one level of institutional undergraduate admissions selectivity to the next higher level.

Institutional Retention Strategies at Historically Black Colleges and Universites and Their Effects on Cohort Default Rates

Education

george washington university

educational policy practice

higher education leaders

university of san diego

n o v e

harvard university

university of manitoba

university faculty