A Decade of Indecision

The Effects of Federal Student Financial Aid Policy from 1987 to 1996 on Within-Year Persistence of Four-Year Undergraduate Students

A proposal submitted to

the Graduate School of the

University of Arkansas at Little Rock

by

James E. Cofer

July, 1998

CHAPTER 1

Introduction

The original goal of federal student financial aid was to make college affordable and accessible to low-income students through a series of grants, and assist lower middle-income students through the use of government subsidized loans. Since its inception in 1965, through a series of philosophically unrelated amendments, federal student financial aid now attempts to serve the twin goals of access and choice. However, in the 1990s, Loans have replaced grants as the subsidy of choice. In 1975-76, grants peaked at 80 percent of all federal aid (College Entrance Examination Board, 1996). By 1992, loans accounted for 52 percent of all federal aid, and by 1997 loans accounted for 79 percent of the federal student aid dollar (College Entrance Examination Board, 1997). Schershel (1998) reports that of the $239.2 billion borrowed since the inception of the federal loan programs in 1966, one-third of this amount, $79.1 billion, was borrowed in fiscal years 1994, 1995, and 1996. Due to the dearth of research on the effects of student loans on access, persistence, attainment, and retention, one can only speculate about the greater reliance on loans and student decisions.

This study examines the effect of debt on within-year persistence on students at four-year universities and colleges, and the changes in those effects resulting from the changes in federal financial aid policy over the period 1987 to 1996. Chapter One presents background information on federal initiatives for student financial aid with special emphasis on the Higher Education Act, and the growing research base on the causes of student persistence or withdrawal from college. This chapter also delineates the problem statement, purpose of the study, the significance of the study, and the research questions. Chapter One concludes with definitions and an overview of the study's organization.

Background on Student Financial Aid

In order to analyze the policies of the last ten years, one needs to review federal involvement in student financial aid. The Higher Education Act of 1965 (HEA) (20 U.S.C. § 1001 et seq.), the watershed event for federal student financial aid, was born of the "Great Society" of President Lyndon B. Johnson, although it was the progeny of the piecemeal efforts of the National Defense Education Act of 1958, (NDEA) (20 U.S.C. § 401), and the Serviceman's Readjustment Act of 1944 (GI Bill) (38 U.S.C. § § 3451-4393, 38 C.P.R. 21.1020). Prior to HEA, there does not appear to be a general policy established by the Federal government for financing higher education students. There were a number of actions preceding HEA, such as the Morrill Land-Grant Acts, the War Loan Programs, and National Youth Administration programs, that provided federal funding, directly and indirectly, to students in higher education. These individual actions laid the foundation, not by major, sweeping legislation, but in an incremental and unrelated fashion common to public policy development in the United States (Cates, 1979; Gill & Saunders, 1992).

Early Federal Student Aid Initiatives

The first concrete expression of the federal role in financing higher education, the Morrill Land-Grant Act of 1862 (Pub. L. 97-98), gave public land, or its equivalent, for the "support of at least one college in every state" (Rudolph, 1990, p. 252). These funds were designated for training in agriculture and the mechanical arts. The Morrill Act concept, that a trained work force was important to the general economic prosperity of the nation, became the major objective of higher education financing, through subsequent federal programs. The Morrill Act philosophy of indirect benefit to students involved in higher education was followed by two refinements to the Act in 1890 and 1917. The Smith-Lever Act of 1914 (7 U.S.C. § 341) provided additional funds for agriculture and home economics, and the Smith-Hughes Act (20 U.S.C. §11) three years later provided for college-level training of vocational education teachers. The Reserve Officers Training Corps (ROTC) was established on college and university campuses with the passage of the National Defense Act of 1916 (39 Stat. 191) (Finn, 1978). All of these measures were responses to specific national priorities, and not a part of an overall strategy to assist higher education.

Depression Era Programs

The Depression years were hard on colleges and universities in the United States with thousands of students forced to leave school for financial reasons. Rural residential colleges in the East and North were the hardest hit and state universities limited enrollment due to financially strapped state budgets (Levine, 1986). Levine notes a disillusionment with college and universities at the depth of the Depression.

By the winter of 1932-33, most institutions hd begun to suffer from losses in income and attendance. Middle-class students found their checks from home had diminished; the radical student leader James Wechsler noted the steady and irresistible decline and fall of the middle-class members of 'collegiate country clubs' across the nation. Poorer students were truly engaged in a life-or-death struggle to stay in school. (p. 30)

The Roosevelt administration considered both a federal loan program, proposed by Chicago's Robert Maynard Hutchins, and a scholarship program for needy students; both were rejected (Axt, 1952; Levine, 1986). The Federal Emergency Relief Administration (FERA) funded a pilot program of student work-study at the University of Minnesota, and announced its support of the work-study concept if the Minnesota experiment worked well. The program in Minnesota was successful, and in 1934, a national program of work-study was implemented under the direction of FERA. The work-study program was to be the centerpiece of the New Deal youth strategy. "By encouraging young people to stay in college, it reduced the number of untrained, unemployed youth as well as the number of young people on relief " (Levine, 1986 p. 197). The National Youth Administration (NYA) absorbed the work-study program in 1935 (Axt, 1952; Quattlebaum, 1956).

The NYA extended the program to all public school students over the age of 16 years. The estimates of the number of college students assisted by the program ranged from 188,000 (Quattlebaum, 1956) to 620,000 (Axt, 1952; Levine, 1986). It is agreed, however, that the program was a success, despite the inadequate funding and lack of meaningful jobs. In the eight years the program was in existence, the NYA spent approximately $93 million for the program in the first substantial federal program to assist needy students to attend colleges and universities.

Roosevelt felt that the work-study program was a relief measure. This was particularly evident when the program was transferred to the NYA over the objection of the National Education Association. The concept of direct assistance to college students had not germinated in the American psyche. Levine (1986) notes;

In its ten years the FERA and the NYA work-study programs helped over two million young people, including more than 600,00 college students. In an emergency, it had enabled many of the best students of the generation to attend college, yet it died a quiet death. The program's limitations confirmed that the preservation of economic and social privilege remained a higher priority than the principle of access in American higher education policy in the 1930s. Educational benefits in the 1930s--like the benefits provided veterans after World War II--were justified not as a tool to increase access to prestigious occupations and professions, but as a means to limit unemployment. (p. 201)

Student Assistance During World War II

The Student War Loan Program, started in 1942, was established to provide a technically trained labor force as quickly as possible. The War aggravated an already critical shortage of a technically trained workforce in the United States, and the War Loan Program provided assistance in the form of loans to students in physics, engineering, chemistry, medicine, pharmacy, dentistry, and veterinary medicine who were within two years of graduation. The students who received loans agreed to accept employment as directed by the War Manpower commission. The basic law provided $500 in any twelve-month period at an interest rate of 2.5 percent. The loans were to be repaid at a rate of 25 percent of the total obligation per year one year after graduation. (Van Dyke 1949; Rivlin, 1961).

Congress appropriated $5 million for the program of which approximately $3.5 million was lent to 11,044 students (Flynt, 1946). Table 1 (see Appendix A) shows that the most popular disciplines under the war loan program. During the two-year period in which loans were made, 34 and 36 percent of the graduates in medicine and veterinary medicine respectively received loans. Dentistry was third in popularity with 32 percent of the graduates receiving loans. During the two year period the loans were in existence, more engineers graduated from college than Dentists and Veterinarians. While only 13 percent of the engineers that graduate received loans they accounted for 36 percent of the dollar amount of the loans.

GI Bill

The most important piece of legislation during the 1940's was the Servicemen's Readjustment Act (GI Bill) (38 U.S.C. §§ 3451-4393, 38 C.F.R. 21.1020). The GI Bill was enacted to reward veterans who had served their country during wartime, and to help them catch up to their peers whose lives had not been interrupted by military service. Educational benefits were not considered the centerpiece of the legislation, a distinction reserved for the 52-20 Club. This provision provided all veterans unemployment benefits of $20 per week for fifty-two weeks. However, the educational benefits extended higher education, and vocational opportunities to thousands of men and women of all races, creeds, and religious affiliations, who might otherwise not have had the opportunity for additional training (Finn, 1978; Gladieux & Wolanin, 1976; Gladieux, 1995).

Origins and Motives of the GI Bill

After World War I, President Wilson and the Congress failed to enact legislation to assist able-bodied veterans in their return to society. The American Legion, formed immediately after World War I, lobbied the White House and Congress for appropriate treatment of veterans. It was not until 1924 that a bill was enacted to provide compensation to veterans; payable in 1945. This reluctance to assist veterans led to the Bonus March on Washington in 1932. The March was dispersed after President Hoover, who used the military, headed by General Douglass MacArthur and his aide-de-camp Major Dwight D. Eisenhower, to disperse the marchers. Instead of bonuses, the veterans got guns and tear gas. Four people died (Bennett, 1996).

President Roosevelt's Administration was cognizant of the problems of unemployed American veterans after World War I, and they also recognized that the rise of socialism and fascism in Europe was directly attributable to unemployed European WW I veterans. Hitler's Brown Shirts, who took power in Germany in the 1930's, were primarily unemployed WW I German veterans (Bennett, 1996). President Roosevelt had been Secretary of the Navy during World War I and ". . .often remarked about the lack of planning for that earlier period and wanted to avoid repeating that neglect" (Olson, 1974, p. 10).

Government concern for the postwar welfare of veterans began with the first peacetime draft in 1940, some 18 months before the United States entered the war in Europe. The peacetime conscription act (50 U.S.C. § 301, repealed) contained a provision guaranteeing post-war re-employment rights to veterans. Jobs were hard to obtain during the early 1940s, with unemployment around 15 percent and young men were reluctant to spend a year in the military without assurance of employment upon discharge. The re-employment rights were an attempt by the Congress to reduce hostility toward the peacetime draft (Olson, 1974).

In 1939, President Roosevelt charged the National Resources Planning Board (NRPB) headed by his uncle, Frederick A. Delano, to study the highways, riverways, airports, and natural gas distributions systems, and to plan for their evolution in the next decade. One of their primary concerns was the matter of postwar problems and the demobilization of individuals whose lives had been directly affected by the war. The NRPB established the Post-war Manpower Conference (PMC) to analyze problems of the demobilization of men and women in the armed forces.

The PMC used as a basis for their eventual recommendations the Wisconsin Educational Bonus Law and the Canadian Demobilization Plan. The Wisconsin legislation was created after WW I, and awarded grants for four academic years of schooling, on any level, to qualified veterans from Wisconsin. The Canadian Plan granted veterans; who had their schooling interrupted by the war, tuition and a monthly allowance (Olson, 1974). In 1943, the PMC recommended a demobilization plan that called for, in addition to vocational/technical training programs, a two-part educational plan, modeled substantially after the Canadian model. The first part of the plan provided twelve months of education to train veterans for new jobs in the civilian economy. The second part provided for a competitive system of scholarships for veterans who had their schooling interrupted by the war (Ross, 1969).

The PMC had taken an inordinate amount of time to prepare what Roosevelt considered preparatory work for the demobilization of veterans. Therefore, the President established, in conjunction with the Army and the Navy, a committee to examine the full aspect of educational benefits for veterans. The so-called Osborn Committee, named after its chairman, Brigadier General Frederick H. Osborn, was instructed to work closely with the PMC and the NPRB, and the results, understandably, did not differ substantially from the report of the PMC. The American Council of Education (ACE) worked closely with the Osborn Committee and the PMC to draft legislation, including post-war educational benefits, for presentation to Congress in 1943.

The study committees were concerned primarily with the effect of 15 million veterans returning home to few, if any, peacetime jobs. The PMC assumed that the postwar economy would be depressed and that the number of unemployed might reach as high as 9 million individuals. Olson (1974) describes the underlying motives for the GI Bill in the following fashion;

But the persons responsible for the legislation clearly indicated by their statements and testimonies that the primary problem lay with the economy, not the veteran. Almost everyone realized that war spending had ended the depression that had lasted throughout the 1930s and that during the war the nation enjoyed the rarity of full employment. Politicians and other leaders had little faith that the economy could sustain this full employment as it moved from war to peace, yet they were convinced that the United States could not safely allow extensive veteran unemployment. If the man in the uniform went from the battle line to the breadline, the common theme ran, he would probably demand radical economic and political changes. The fear of unemployed veterans, not the fear of maladjusted veterans motivated the persons who enacted the GI Bill.

At its root, the Serviceman's Readjustment Act of 1944 was more an antidepression measure than an expression of gratitude for veterans. (p. 23-24)

GI Bill Provisions

The language of the original GI Bill limited educational benefits to those whose education had been interrupted by the war. Veterans who were under 25 years old when they entered the war were presumed to have had their education interrupted, but those over 25 years of age at the time of enlistment had to prove educational interruption. These provisions were changed within a year of the passage of the Act. The new provisions eliminated the age limitation and provided all veterans with more than 90 days of service educational benefits (Rivlin, 1961).

Eligibility for benefits was not based on ability, field of study, or type of institution. Eligibility for educational benefits rested solely on duration of military service, and a veteran's ability to be accepted and maintain passing grades in any school, college, or university recognized by the veteran's state of residence or the Veterans Administration (VA) (Rivlin, 1961). Veterans could take any vocational, technical, agricultural, or academic course of study. Over 7.8 million WW II veterans took advantage of the GI Bill educational benefits between 1944 and the expiration of the Act in 1956, almost one in every two veterans of WW II (Quattlebaum, 1956). The Veterans Administration classified the programs as; On the Farm, On the Job, Below College, and College. Only 29 percent of the veterans taking advantage of the educational provisions of the GI Bill went to college (see Table 2 in Appendix A). In addition, the youngest group of veterans tended to select the college benefits, and the oldest group returned to the farm. The largest group of veterans, 44 percent, obtained training below the college level.

Dual Payments under the GI Bill. Payments under the 1944 GI Bill were made to both the institution and the veteran. The legislation specified that the VA would pay institutions up to $500 per year for tuition, fees, books, and supplies. This provision worked to the disadvantage of some state and municipal institutions that were under a legal obligation to provide free or reduced cost education to all state residents. To accommodate these institutions, they were allowed to treat all veterans as non-resident students or were allowed "fair and reasonable compensation" for educating a veteran (Olson, 1974; Axt, 1952).

In addition to tuition payments to the school, veterans received a subsistence allowance from the Veterans Administration each month they were enrolled in school. The original act designated that a full-time student would receive $50 a month if he had no dependents, and $75 a month if he had one or more dependents. These amounts were raised to $65 and $75 respectively in December 1945. After significant pressure by veterans, the subsistence allowance was raised to a rate of $75 a month for single veterans, $105 if the veteran had one dependent, and $120 per month for veterans with more than one dependent (Olson, 1974; Axt, 1962).

Impact of the GI Bill

The World War II GI Bill influenced two later bills, for Korean and Vietnam War veterans. The Korean GI Bill eliminated the payment to the institution, but increased the subsistence allowance and made the veteran responsible for all costs. The Vietnam GI Bill was slightly less generous than the previous legislation. Vietnam veterans were required to have 180 days of creditable service, not 90 days as previous bills, and Vietnam veterans received one day of training for each day served, not one and a half days like Korean veterans.

The GI Bill's objective of keeping unemployed veterans occupied immediately after the war was a success. The college benefits under the GI Bill were a method to readjust the veteran into the American economy, and also proved an unqualified success. Veterans found jobs and because many had additional training, they held higher paying jobs than they had held before they left. The GI Bill's educational benefits tied to the housing provisions inexorably changed the landscape of the American economy (Olson, 1974).

The impact on higher education was, however varied. When veterans brought their spouses to campus it forced an acceptance of married students. Prior to the war, married students were not welcome, and in the case of many married women students, signaled their departure from higher education. In addition, because of the sheer numbers of veterans, the "uncritical acceptance of largeness became a major legacy of the GI Bill" (Olson, 1974, p. 103).

Previous federal funding of higher education had been characterized by linking funds to specific national programs and academic disciplines. With the GI Bill, resources were placed directly into the hands of the students, and gave those individuals the power to decide what was best for themselves. "The GI Bill gave people the economic resources to train their mind, their most important property, and they went out and used it to pursue happiness and safety" (Bennett, 1996).

The Sputnik Effect

The first large-scale federal student loan program was born in response to the launch of Sputnik in 1957. Fearful that the United States was lagging in science and technical education, President Eisenhower proposed spending $1.6 billion to improve education in science and foreign languages. Congress passed the National Defense Education Act of 1958 (20 U.S.C. § 401), which created a number of education efforts to promote education in science, including the National Defense Student Loan Program (now Perkins Loans), and a fellowship program for graduate study (20 U.S.C. §§ 1987aa-1087hh, 34 C.F.R. Part 674).

The loan program provided students exceptionally low interest rates on loans, with a 90/10 split between the federal government and the institution on the capital for the loans (Mumper, 1996). Undergraduates and graduate students were eligible to borrow up to $1000 per year, or a total of $5000 as long as they were full-time students. Institutions were free to choose recipients but were requested to give special consideration to those students who desired to teach in elementary or secondary schools, and those students who had a superior capacity in the sciences or a foreign language. If a graduate taught at the elementary or secondary school, 50 percent of the loan could be canceled at the rate of 10 percent for each year they taught (Rivlin, 1962).

The NDEA also created two limited, specifically directed grant programs. National Defense Fellowships were intended to increase the supply of college and university faculty as well as provide a broader geographical distribution of those faculty. The Act provided for 1,000 fellowships during the first year of the program and 1,500 fellowships in each of the preceding three years. The fellowships were outright grants amounting to between $2,000 and $2,400 depending on year of study. The institutions were also compensated $2,500 per student as a reimbursement for the additional cost of the program. The Act specifically stated that the fellowships were for "new" or "expanded" programs. In addition to the National Defense Fellowships, a second, smaller program, the National Defense Language Fellowships, was established by the Act. The Commissioner of Education selected 85 languages in which trained linguists were needed (Rivlin, 1962).

The NDSL program has been modified over the years, but has philosophically remained the same. Forgiveness of loans was phased out in the mid-1970s, and the name was changed to Perkins Loans (after its sponsor Senator Carl Perkins) in 1972.

Genesis of a Federal Education Policy

Federal funding for loan and grant programs for higher education had been, up to 1965, responses to specific national objectives or emergency conditions. The Land-Grant Act, while benefitting students, states, agriculture, technology, and the nation as a whole, was essentially a method of disposing of excess federal land. The NYA programs and the GI Bill were designed to avoid flooding the private sector with large numbers of unemployed youth and veterans. The National Defense Loan Program was a response to a national defense emergency and embarrassment caused by the advances in science made by the Soviet Union. It was not until the Johnson Administration that several pieces of legislation came together to form the first comprehensive education policy for the federal government.

Higher Education Act

Under the rubric of the "Great Society," the Higher Education Act (HEA) (20 U.S.C. § 1001 et seq.), and the companion Elementary and Secondary Education Act (ESEA) (20 U.S.C § 6301 et seq.) formed the cornerstone of federal policy in education. The Higher Education Act of 1965 consolidated several previously enacted anti-poverty measures with programs whose major attention was the provision of access to higher education for the poor and talented students (Keppel, 1987; Kimberling, 1995 ). The Higher Education Act of 1965 merged the National Defense Education Loan Program of NDEA (20 U.S.C. § 401), the College Facilities Act of 1963 (20 U.S.C. § 701), the College Work-Study Program (42 U.S.C. §§ 2751-2756(a), 34 C.F.R. Part 675) implemented in the Economic Opportunity Act (42, U.S.C. § 2701 et seq.), and created two new financial aid programs; the Educational Opportunity Grant (EOG) (20 U.S.C. §§ 1070b-1070b-3, 34 C.F.R. Part 676) and the Guaranteed Student Loan Program (GSL) (20 U.S.C. §§ 1087aa-1087hh, 34 C.F.R. Part 674) (Kimberling, 1995).

The original legislation creating the Higher Education Act in 1965 was the result of two competing ideologies for federal funding of higher education. The first was that of assisting higher education by funneling funds directly to institutions through some all-encompassing formula based on student counts. This first method would clearly benefit institutions. The second was that of assisting needy students directly through grants to the poorest students, and subsidized loans to lower-middle class students. This alternative provided a system of portable aid that provided both access and choice to the student. Even though the 1965 legislation was indeed a compromise between these two competing interests, the primary objective was one of equalizing educational opportunity for socioeconomically disadvantaged students by providing equal access to higher education for all students.

Foundation for future student financial aid

The 1972 amendments to HEA (Pub. L. No. 92-318) broadened the policy objectives to access and choice, and established the foundational role of the federal government in student aid. Direct portable aid to needy students was introduced in the 1972 amendments through the Basic Educational Opportunity Grants (BEOG) (20 U.S.C. §§ 1070a-1070a-6, 34 C.F.R. Part 674). The new Basic Educational Opportunity Grant (BEOG - was later renamed Pell Grants after Senator Claiborne Pell) was a portable voucher-like system for needy students. The BEOG differed from the earlier EOG Program in that the students applied directly to the federal government instead of the institution. These new grants were awarded solely on the basis of need, and that need was determined by a formula which was administered at the federal level.

Changes to HEA in 1972 not only broadened student choice, but broadened eligibility for a whole new group of institutions. The access and choice question is essentially a public school and private school dichotomy. Public schools offered low tuition and very little institutional based aid. Low tuition implies that most of the aid at public schools will go to low-income students, since middle-and high-income students will have no need. Therefore, federal student aid acts to supplement the low-tuition strategy of public institutions, and promote access. Private schools represent the choice part of the access and choice equation. In general, private institutions spend a large portion of their resources on student aid through tuition discounting. Therefore, federal student aid "federalizes" some of the costs of student aid that these institutions would incur in the natural course of business. Secondly, the need-based federal financial support focuses more on the lower-income students than the college would without the federal subsidy. Is it fair to subsidize private higher education with public dollars? This question is best answered in relation to one's position on the question, "Should the federal government bear part of the cost of student choice?" Coupled with the funding of choice, the proprietary institution sector developed rapidly after the change of "higher education" to "postsecondary education" in the definition of eligible training in the 1972 amendments.

Middle Income Student Assistance Act (aid for everyone)

During the 1970s, 76 percent of federal student financial aid was in the form of grants and approximately 20 percent in loans. By the mid-1980s, that ratio had almost reversed with loans accounting for 67 percent and grants 29 percent of federal financial aid (Hannah, 1996). Further, Pell grants paid less than 50 percent of the cost of education (College Board, 1992). The passage of Middle Income Student Assistance Act (MISSA) (Pub. L. No. 96-49) in 1978 was a fundamental shift in federal financial aid policy and philosophy. Easily accessible, non-need based aid became available to large numbers of middle- and upper-income students on a quasi-entitlement basis. Federal student aid outlays increased by 59 percent between 1977-78 and 1980-81 (Hearn, 1993). The lower-income targeted, need-based, grant oriented federal policy period came to an end. While the stated goals of MISSA were to promote educational choice, persistence, and access for both middle- and lower-income families, upper-income families took advantage of the program in substantially larger than expected numbers. Figure 1 (Appendix A) illustrates the changes in the percentage distribution between federal grants and loans from 1971 to 1997.

The push for the changes came from two sources; the middle-class constituencies of politicians, and the intense lobbying efforts of banking officials (McPherson, 1989). Politics and power, based on a perceived threat to the middle-class status quo, combined to reverse the original intent of federal financial aid policies. After the passage of MISSA, enrollments of lower-income students declined (Mortenson, 1990a).

Policy Drift During the Regan and Bush Administrations

In 1981, President Reagan pushed through Congress an amendment to the Higher Education Act that essentially limited student loan borrowing for all income levels. From 1982 through 1985, Reagan advocated further reductions in federal financial aid, however, Congress refused to act on his recommendation.

By the mid-1980's, college costs were increasing and median family income was falling. The budget situation was casting a "dark shadow" over all federal programs, and expectations were not very high for the 1986 Reauthorization of the Higher Education Act (Pub L. No. 99-498) (Mumper, 1996). The emphasis on student financial aid at the federal level was "to put the brakes on." The federal budget situation, an intransigent president, and a growing awareness that financial aid programs were not achieving their intended objectives all preceded the consideration of the 1986 Reauthorization of HEA. Proposals were advanced, primarily by Congressman Ford, to expand the programs so that low-income students could cover the increasing costs of college attendance. Opponents of Congressman proposed increased loan limits, for a limited number of middle-income students ("Five year education," 1986).

The final compromise, according to Mumper (1996),

. . . did nothing to reverse the declining value of Pell grants to low-income students. It increased dependence of many students on borrowing for college. It increased the long-term costs of the program, and it accelerated the shifting character of federal aid toward loans away from grants. (p. 98)

Dean (1994) notes that during the early 1980s there was a willingness by some lenders to provide loans to students regardless of their institution's default rates. Lenders were insured against defaults in the GSL program, and institutions were constrained by costs from becoming debt collectors under the NDSL program. The 1986 Reauthorization of HEA did little to change the shifting the responsibility for federal financial aid policy from the federal government to the families of students. A new Congressional Methodology (CM) was enacted, effective in the 1988-89 school year, that standardized the need assessment mechanism. The needs test for loan eligibility was reinstated, family assets, as well as income, were considered in the needs assessment, with emphasis more heavily weighted toward income. The new Congressional Methodology replaced the Uniform Methodology which had been used as a guide for needs assessment since 1976 for all programs except Pell Grants. In addition, the definition of independent student was changed, and under CM, independent students with dependents were treated differently than dependent students. The differences generally increased the expected family contribution (EFC) for independent students with no dependents, and decreased the EFC for independent students with dependents (Lee, 1988; Mortenson, 1990a).

Two measures were included in the 1986 reauthorization bill aimed at cutting program costs. First, as mentioned earlier, family assets were added to the needs analysis calculation under the new CM, which had the effect of lowering the amount of aid for which a student could qualify. Second, a 3 percent insurance fee was charged to all borrowers to ease the growing default problem. However, limits on middle-income students were relaxed so that the aggregate loan amounts for undergraduates went from $17,500 to $25,000. The cumulative amount that a student could borrow for both undergraduate and graduate school was $54,750, more than twice the previous limit of $25,000 (Gladieux & Wolanin, 1976). With the 1986 reauthorization, the transformation was complete: loans replaced grants as the primary method of financing college costs, and even these loans were limited.

Expansion of Loan Programs (again)

The 1992 reauthorization of HEA (Pub. L. No. 102-325) cemented the shift in federal policy from a commitment to promote access through need-based grants to a broader strategy of loans regardless of family income or need. Congress proposed making more loans available to the middle-class student, decrease the reliance on loans by increasing grant support, and restore the integrity to the loan program.

The relaxed eligibility for government subsidized loans resulted in an increase of 2 million additional students receiving loans between 1990 and 1996, with a concomitant 92 percent increase in money borrowed. Table 3 (see Appendix A) illustrates the loan limits before and after the 1992 Reauthorization. With only a 16 percent increase in the Pell grant program, the imbalance between loans and grants worsened (Hartle, 1996). The efforts to restore integrity to the program have been successful with defaults rates dropping from an aggregate 22.4 percent in 1990 to 10.7 percent in 1994 ("FTEL Default Rates," 1997).

Summary of Federal Funding for Student Financial Aid

From its beginnings in 1965, through several philosophically incoherent amendments, several key principles of federal student financial aid policy have evolved. Johnstone (1995) and McPherson, Shapiro, and Winston (1993) generally agree that these principles are: the costs of higher education are shared by the student, the student's parents, and taxpayer; and the ability of the parents to pay for education should be objectively measured. In addition, the federal government believes that it makes grants available to low-income students so that they can afford moderately-priced public institutions, and makes loans available to middle-and upper-income students so that they can afford higher-priced private institutions; and aid to students is given without regard for course of study or academic promise.

Student Persistence in Higher Education

This section summarizes the research on student persistence in higher education. Persistence has been an area of interest to researchers since the 1920s (Edgerton, 1929; Hoke, 1922; McNeely, 1938; Minnesota University, 1924; Tallman, 1927). Tallman (1927) examined retention at the University of Iowa using the Class of 1920 as the sample. Tallman attempted to link intelligence as measured by the Thordike Intelligence Examination to student persistence, but found that the primary reason given for leaving school was financial. Tallman also concluded that just as many students with high intelligence scores dropped out for financial reasons as those with low intelligence scores. Likewise, Hoke (1922) concluded that intelligence as measured by four different intelligence tests popular during that decade did not furnish a satisfactory basis for prediction of academic success in a group of entering freshmen at Hood College.

As the model for higher education in the United States changed from elite to mass, attrition, and the causes of it, became a significant issue (Trow, 1979). At the same time, federal higher education policy defined persistence as a goal (Gladieux & Wolanin, 1976). Spady (1971) and Tinto (1975) adapted Durkheim's theory of suicide (Durkheim, 1951) to explain college student dropout. Durkheim proposed that suicide was much more likely to occur when individuals were insufficiently integrated into Society. Tinto (1975) applied this analysis to college student attrition:

When one views the college as a social system with its own value and social structures, one can treat dropout from that social system in a manner analogous to that of suicide in the wider society. One can reasonably expect, then, that social conditions affecting dropout from the social system of the college would resemble those resulting in suicide in the wider society; namely insufficient interactions with others in the college and insufficient congruency with the prevailing value patterns of the college collectivity. Presumably, lack of integration into the social system of the college will lead to low commitment to that social system and will increase the probability that individuals will decide to leave college and pursue alternative activities. (p. 91-92)

Tinto described the conditions under which attrition occurs, and proposed that background characteristics (e.g., social status, high school experiences, neighborhood, etc.) were important, but so were the expectations and motivational attributes individuals brought to college. The latter were termed educational and goal commitments, and introduced psychological variables into models of student attrition. In addition, a student's commitment to a particular institution (institutional commitment) was important.

The early research of Pascarella and Terenzini (1977, 1979, 1980) supported the Tinto model. Their 1979 study examined the predictive validity of measures of social and academic integration. To assess the dimensions of such integration, a series of items measured intellectual development, peer-group interactions, interactions with faculty, and institutional and goal commitments. The study controlled for various background characteristics, first-year grade point average, and the extent of involvement in extracurricular activities.

Later studies have incorporated financial aid variables in persistence models. Bean's model (Bean, 1980; Hossler, Bean & Associates, 1990) included social and academic integration variables, but also addressed the issue of financial support in background variables. Cabrera and Nora (Cabrera, Castanada, Nora, & Hengstler, 1992; Nora & Cabrera, 1993) compared the Tinto and Bean models using a questionnaire given to all full-time students enrolled at a large southwestern university. They concluded that a model that included integration and financial variables explained more of the variance in persistence decisions. St. John and Somers (St. John, 1989, 1991, 1992, 1994, St. John & Andrieu, 1995; St. John, Paulsen, & Starkey, 1996; Somers & St. John, 1996; Somers, 1992, 1993b, 1994a, 1995) developed and tested a model to study first-time enrollment and persistence using existing institutional data. They found that aid could have a significant impact on student attendance and persistence decisions; likewise, the lack of aid could also influence decisions.

National Postsecondary Student Aid Survey

The research on student persistence was given a big boost by the National Postsecondary Student Aid Study (NPSAS) of 1987. Prior to the development of NPSAS, much of the research had been at the institutional level or using small, select national samples. NPSAS provided an opportunity to study persistence using a large, national sample with an extensive set of economic, demographic, and college experience variables.

A series of studies using NPSAS:87 and NPSAS: 90 were done by St. John and his associates. For instance, St. John, Paulsen, and Starkey (1996) examined the influence of finances on college choice and persistence decisions. Financial choices had direct and indirect influences on college persistence, and market-based, monetary measures of financial aid, tuition, housing costs, and other living costs had substantial direct effects on persistence. Another study (St. John, 1992) focused on persistence by traditional college students. Their persistence was negatively influenced by the increasing tuition levels. Grants were positively associated with persistence in private colleges but negatively associated with persistence in public colleges.

Several studies (Andrieu, 1990, 1991; Andrieu & St. John, 1993; St. John & Andrieu, 1995) examined within-year persistence of graduate students. The conclusions were that tuition charges had a substantial negative effect on persistence and comprehensive aid packages (grants, loans, and assistantships together) were the most effective means of subsidizing costs.

DeAngelis (1997) extended the work of Andrieu and St. John on graduate student persistence using NPSAS:93. DeAngelis found that financial aid significantly and positively influenced within-year persistence of graduate and professional students. Further, students receiving aid packages including loans, grants, and assistantships were more likely to persist.

A few studies have used NPSAS:93 to examine enrollment and persistence. Perna found that financial aid allowed students to attend more expensive institutions (Perna, 1996). In a recent study using NCES's Beginning Postsecondary Student Survey, Perna (1997) examined the contribution of financial aid to the retention to graduation of full-time, four-year, financially dependent students. Perna found that students who received financial aid were more likely to graduate than those who did not. Using path analysis, Perna concluded that the contribution to retention from the receipt of financial was in the middle range of factors in all the models tested. All types of financial aid were positively related to retention, with the exception of loans. Loans had an indirect negative effect on grades, and grades were the single most important contributor to retention in this study.

Debtload

Almost from the inception of the student loan program in the 1960s, there has been speculation about the impact of debtload on student decisions about attending college, persisting in college, and life decisions after graduation. The common assumption was that high debt burden influences student choices, including institution, major, and a career with a high expected income. In the mid-1980s, attention was focused on the choices of law, medical, and dental students (Brotherton, 1995; Chambers, 1992; Dial, 1987; Petersdorf, 1991). This interest was prompted by two major concerns: the perceived large numbers of these graduates who filed for bankruptcy shortly after graduation to escape a heavy educational debtload and the concern that these graduates shunned lower-paying jobs, often in public service, in favor of positions in higher paying specialities. The research of this era was quantitative and sampled primarily law, medical, and dental students.

A large group of studies examined debtload more generally. The higher education literature on student debtload can be classified into two eras, both of which correspond to change in federal policy. First, are the articles and studies from the mid-80's to 1992, which reflected the growing concern over the consequences of shifting from grants to loans. The literature of the second era, since 1992, has become increasingly alarmed over the size of student debtload as a result of the increased availability of unsubsidized loans.

Educational Record devoted the Summer 1987 issue to the theme of student aid. Henderson (1987) compared the debtload of graduates in 1977 ($2,700) to the comparable, inflation-adjusted figure in 1984 ($5,470). She concluded that:

. . .there may be "pockets" of loan burden, particularly for some women and persons majoring in education or the humanities. Their average college debts are not significantly greater than those of other graduates, but their burden of repayment is greater because, on average, their salaries are lower (Henderson, 1987, p. 29).

Atwell (1987) observed, "Yet another fallout of loading up students with loans is that the current generation is leaving a legacy of profligacy for the next one" (Atwell, 1987, p. 8). Fisher (1987) described howls of protest against increased student borrowing, including this statement by Howard Swearer, "We are not only asking the next generation to pay for most of the mistakes of the past; we are equipping them less well to do so" (Fisher, 1987, p. 20).

In 1996, Educational Record again devoted an entire issue to student financial aid. One article focused on student debtload (Baum, 1996). She cautioned that:

[just because] average borrowing levels for undergraduates are well within reason does not mean that we should stop worrying about student debt. It would be irresponsible not to emphasize the real danger that the prospect of high debt may discourage vulnerable population groups from participating in higher education. (p. 35)

She notes that graduate and professional students can borrow considerably more than undergraduates with the Stafford limit of $138,500.

Judith Eaton (1995) of the Council for Aid to Education described federal policy for financing higher education as follows:

Suppose that a country wanted to create a public subsidy program for college and university attendance. Suppose it made ability to pay a major factor in determining who could attend college, thereby excluding the less affluent from the benefits of higher education. Suppose higher education institutions and the country created a pricing structure which ensured that incurring debt was an inevitable byproduct of college or university attendance. Suppose the country decided to reward students who attained degrees from the most expensive institutions with the best jobs, highest incomes, and most elevated status. . . .How perverse that we can ensure access to collegiate education for students who are already less affluent only through a public policy that forces them into indebtedness. (p. 12)

According to Eaton, student borrowing has become the primary subsidy for higher education in the U.S. She talks of a future of "educational mortgages" and the intergenerational shift in the responsibility of funding higher education. With each generation shifting the debt burden to students, they are in turn less able to pay for their children's college education.

Much of the recent research on the issue of debtload can be classified into two areas: descriptive statistical reports generated by loan guarantee agencies and federal contractors, and empirical research using data such as NPSAS:87, 90, and 93. The majority of the empirical research precedes the substantial increase in borrowing associated with the reauthorization of HEA in 1992. Loan guarantee agencies are primarily concerned with the mounting amount of debt and the students ability to repay those loans. Scherschel (1998) of the USA Group, a loan guarantee agency, observes that:

Between the inception of the federal loan program in 1966 and the end of fiscal year 1996, students and parents borrowed $239.2 billion. They borrowed fully one third of this amount-$79.1 billion-in just the past three fiscal years (FY1994, FY1995, FY1996). (p. 2)

Figure 1 (see Appendix A) illustrates the growth in loan commitments from 1970 to 1996.Problem Statement

Even though the amount of debt carried by college students has been escalating rapidly, there is limited research (Cofer & Somers, 1997; Cofer & Somers, 1998; DeAngelis, 1997; St. John, 1994a; Somers & Bateman, 1997; Somers & Cofer, 1998; Somers, Cofer, DeAngelis, & Cook, 1997) on the role debtload plays on student decision making. Indeed, the broader issue of student loans and educational decision making has not been explored. There are several potential explanations for a lack of research on the impact of loans on student decisions (Somers & Bateman, 1997). First, much of the research agenda has been based on the priorities generated by Congress. Over the past decade, there has been increasing concern over default rates. Further, the concerns over balancing the federal budget in recent years has shifted the discussion to the need and funding levels for loans, and away from the impact that loans have on students. Finally, the increased borrowing levels authorized by the Higher Education Act of 1992 beg for a study of the impact of loans on college student choices. However, there are little data available on this time frame

Mortenson (1990), a critic of the adequacy of federal financial aid programs, said;

The reallocation of student financial aid from poor to middle income and affluent students has been the most pervasive, persistent, and least discussed trend in student aid policy during the last dozen years. Beginning with the federal Middle Income Assistance Act in 1978, virtually every major change in federal, state, and institutional aid has worked for the benefit of middle income and affluent students. (p112)

Changes in federal financial aid policy starting with MISSA and continuing through the reauthorization of 1986 and 1992, resulted in a significant shift in federal policy from a commitment to promote access through need-based grants to a broader strategy of loans regardless of family income or need. Mumper and Vander Ark (1991) believe that Pell grants suffered in relation to loans because they appealed to lower-income students, and middle-income students, banks, and private colleges, all represented by well-funded lobbying groups, favored loans. Wilensky (1975) notes "The greater the educational opportunity (measured by enrollment rates or fraction of the children of lower strata who go on to higher education), the more chance the middle-class parents of college students as well as large numbers of highly educated will resist the push from below for equality" (p 54-55).

The descriptive studies on student debtload (Baum & Saunders, 1998; Choy, 1998; King, 1998; Scherschel & Behymer, 1997) present statistical evidence of increasing debtload on college students, their spouses, children, and families. For the most part, they conclude that based on the descriptive statistics, the amount of debt is not excessive, and that students can repay these loans. These same studies generally reason that students can handle even more debt. On the other hand, the empirical evidence through a series of studies by Somers and her associates (Cofer & Somers, 1995, 1997, 1998; Somers, Cofer, DeAngelis & Cook, 1997; Somers & Cofer, 1998; Somers & Bateman, 1997) conflict with conclusions drawn from the descriptive studies.

Fossey (1998) in a provocative article entitled "The Dizzying Growth of the Federal Student Loan Program: When will Vertigo Set In?" notes;

Research and policy analysis have not kept up with this explosion in student loan debt. Consequently we have no clear picture how a lending program of such enormous size will affect the national economy in the coming years, not to mention the financial planning decisions of millions of student loan debtors. (p. 10)

What is the effect of increased availability of debt on the persistence/withdrawl decision of college students? Does accumulated debt have an effect on the persistence/withdrawl decision? Does the effect of accumulated debt differ by income? These questions remain unanswered while the accumulated debt of college students inexorably increases.

Research Questions

The primary focus of this study is to determine the effect of accumulated debt on the within-year persistence/withdrawl decision of four-year college students by income. This study looks at the difference in effects of background, achievement and aspiration, college experience, and price variables across a time period from 1987 to 1996 on the student persistence/withdrawal decision in an attempt to isolate the federal financial aid policies that contributed to the changes. The following research questions are addressed:

• How do background, achievement, college experiences, price, and debtload influence within-year undergraduate persistence at four-year colleges across income strata?
• Has the influence of background, achievement, college experiences, and price on within-year undergraduate persistence at four-year colleges across income strata changed between 1987 and 1996?
• What is the effect of debtload on within-year undergraduate persistence at four-year colleges across income strata?
• Has the influence of debtload on within-year undergraduate persistence at four-year colleges across income strata changed between 1987 and 1996?
• What implications do these findings have for federal financial aid policy?

Purpose of the Study

The economic value of a college education is firmly established in the literature (Astin, Green, Korn, & Schalit, 1985; Leslie & Brinkman, 1988; Leslie, Johnson, & Carlson, 1977; Mattila, 1982). It is evident (College Entrance Examination Board, 1996) that students and their families are willing to invest time and money, and assume debt to obtain an education. Can the "system" handle the increasing amount of debt? Leslie and Brinkman (1987) began their article "Student Price Response in Higher Education" with "What happens to enrollments when colleges and universities raise their prices?" The interaction of prices, subsidies, demographics, culture, and policy make that question harder to answer now.

Attitudes toward loans, amounts of subsidies, and net cost of college attendance all changed in the intervening years between the NPSAS:86 and NPSAS:96 data collection. The assumption is that these and political forces changed the financial aid paradigm indelibly. The release of NPSAS:96 enables researchers to examine at the effect of the 1992 Reauthorization of HEA, including the changes because of the switch to the Congressional Methodology in calculating financial need, and the increased amounts of loans.

There is a growing body of evidence that students' responses to tuition increases are not as negative as once assumed, and there is a growing reluctance to accumulate debt for a college education. Clearly college students and their families are willing to assume greater and greater amounts of debt, but there is a limit. Who broaches that limit first, the federal government or the college student, will direct federal financial aid policy in the future. The response to a constantly changing federal financial aid policy has not "fallen into place," obscuring important patterns and relationships, and continued research is required to make meaning of the effects of these changing policies.

The purpose of this study is to determine the effect of debt on the persistence/withdrawal decision faced by college students, and to determine if that effect has changed from 1986 to 1996. Specifically, the study examines the effects of debt across income strata, and attempt, to determine how federal financial aid policy changes in the last decade have affected student persistence, and if those effects differ by income strata.

Significance of the Study

From the NDEA in 1958 to the HE of 1965, and then through several continuous philosophically incoherent amendments, federal student financial aid policy has evolved. This evolutionary process appears to have done more harm than good to those individuals who fought for the expanded choice in federal financial aid programs (Cofer & Somers, 1998).

The primary objective of federal financial aid programs under Title IV of the Higher Education Act of 1965 was to ensure access and opportunity for students to attend college, regardless of their ability to pay. The shifts in policy from 1965 to 1997 resulting from the patchwork of amendments makes one question whether this objective has been met.

Definitions

A number of important financial aid terms are used throughout the study. To ensure proper communication and understanding, these terms are defined as follows:

EOG - Educational Opportunity Grants were initiated by the Higher Education Act of 1965. The country's first general grant program for college students limited grants to $800 per year for the purpose of making available the benefits of higher education to qualified high school graduates of "exceptional" financial need. EOG was merged into the SEOG program in the 1972 Reauthorization.

Pell Grant/BEOG - Basic Education Opportunity Grants were first authorized under the Educational Amendments to HEA in 1972, and were renamed to Pell Grants in 1980 to honor Sen. Claiborne Pell, sponsor of the original legislation. Under BEOG, grants were to be made to every student who was determined eligible under a formula used to assess a family's ability to contribute to the student's cost of education.

Stafford Loan - The federal Stafford Loan program offers low-interest loans to students for college and post-baccalaureate study. Borrowers with subsidized loans have the interest on the loans paid by the federal government while they are in school and through a grace period before repayment begins. Students with unsubsidized Stafford loans (a program created in the 1992 Reauthorization) are responsible for the interest on their loans from the date of issue.

Perkins Loan

NDSL Loan - The National Defense Student Loan program was originally created by the National Defense Act of 1958. The statute provide for a long-term, extended repayment, and low-interest loan program on each campus. The federal government contributes nine of every ten dollars needed for the program.

PLUS Loan - The Parent Loans for Undergraduate Students (PLUS) issues loans to parents and spouses for the education of their children and spouses. The loans are not subject to annual or aggregate borrowing limits, effective with the 1992 Reauthorization.

FFELP - Federal Family Education Loan Program, known as the Guaranteed Student Loan (GSL) program until 1992, the FFELP consists of the subsidized and unsubsidized Stafford Loans and the PLUS loans.

Ford Direct Loan Program - Created as a pilot program in 1992 and established as a full program in 1993, the FDSL ("direct lending") issues federal direct loans directly through colleges and universities, thereby eliminating the bank as a middleman. Capital for the loans are provided by the U. S. Treasury.

SEOG - Supplemental Educational Opportunity Grant was first established as the EOG in HEA of 1965. The 1972 amendments modified the EOG somewhat and retitled it SEOG. The SEOG was rigidly targeted to the most needy and disadvantaged, but restrictions were eventually relaxed because of escalating tuition costs, primarily in the private college sector.

GSL - See FFELP

Congressional Methodology - The Congressionally mandated needs analysis system for all federally funded student financial aid programs. The CM was effective with the school year 1988-89, and replaced the Uniform Methodology.

SLS - Supplemental Loans for Students was created to offer loans to independent students without an interest subsidy. The program was eliminated in the 1992 Reauthorization and replaced by the unsubsidized Stafford Loan program.

CWS - The College Work-Study program was originally created as part of the Economic Opportunity Act of 1964, and later transferred to the Department of Education in HEA of 1965. The CWS provides part-time employment at a postsecondary institution for low-income students. Wage rates, assignments, hours, and general supervision are given by the institution. The federal government provides 80 percent of the cost of the program. In addition to CWS, some schools have their own work-study program funded entirely through institutional funds.

Debtload - Debtload is defined for the purpose of this study as the total accumulated debt, from all sources, that a student or a student's family uses for educational purposes.

Scope and Delimitations of the Study

This study will focus on two groups four-year college and university students included in the National Postsecondary Student Aid Survey of 1986-87 and the National Postsecondary Student Aid Survey of 1995-96 (NPSAS:87and NPSAS:96). NPSAS is well suited for research on within-year persistence because the sampling procedure included college students in the fall of the year with a follow-up survey in the spring of the same academic year. NPSAS:87 contains a sample of 59,886 students enrolled at 1,074 postsecondary institutions as of October 1986. NPSAS:96 contains a sample of 48,300 students at 836 institutions enrolled during the 1995-96 academic year.

Limitations

While the National Postsecondary Student Aid Surveys provide an excellent opportunity to study student persistence, the surveys do have a number of limitations. First, NPSAS:87 was a fall sample and may not have been representative of all college students. Even though NPSAS:96 collected data several times during the year to overcome the limitation in 1986, NPSAS:96 will be adjusted to a fall sample for comparability purposes with NPSAS:87. This limitation is less of a problem with four-year college students, since the majority of four-year college students enroll in the fall.

Second, NPSAS:87 and NPSAS:96 did not include all the information necessary for an "ideal" persistence model. In particular, high school experiences including grades and rank are not included in the data. In addition, the NPSAS:87 data did not include achievement test scores. Additionally, neither of the studies include adequate data on certain emotional or psychological factors that have proved important in prior persistence research. This study is concerned primarily with the effect of financial variables on student persistence, and the presence of these missing variables would make the model better. Their absence however, is not considered fatal.

Third, there are a number of missing values for variables in both surveys. It was assumed that the missing values were randomly distributed, and where the missing values appeared to created a problem, dummy variables were inserted in the model.

Fourth, both surveys over sampled freshmen and seniors to accommodate sub-surveys (Beginning Postsecondary Students and Baccalaureate and Beyond). This situation is discussed in the Method section.

Fifth, the delta-p statistic for a given independent variable is usually applicable to the otherwise "average" sample member. This measure is of questionable validity when sample members are on the extremes on the variable. This condition limits the applicability of the delta-p statistic in extreme cases.

Organization of the Study

This study is organized into seven chapters. Chapter One serves as an introduction to the study. It includes the background of the research problem, the problem and purpose of the study, research questions, delimitations and limitations of the study, and definition of critical terms.

Chapter Two contains the review of related literature to the study. It contains a discussion of the process used to complete the literature review, a history of the Higher Education Act and subsequent amendments, a review of research on student persistence, and a review of the research using the National Postsecondary Student Aid Surveys (NPSAS).

Chapter Three contains a description of the methods and procedures used in the study. Included is a description of the research design, the model, the sample, and the statistical technique the study uses.

Chapter Four has the findings obtained from the data from the 1987 Survey, and an analysis of these findings. The findings from the 1996 survey are included in Chapter Five, with accompanying analysis.

Chapter Six compares the findings and analysis of the two surveys, and highlights the similarities and the differences. Chapter Six concludes with policy recommendations for future federal financial aid policy actions. The study concludes in Chapter Seven with an overview of the study, a summary of the results, conclusions, policy recommendations, and suggestions for future research.

CHAPTER TWO

Review of the Related Literature

Introduction

Chapter Two contains the review of literature related to the study. Included are a history of the Higher Education Act and subsequent amendments to the Act, a review of research on student persistence, a review of the research using the National Postsecondary Student Aid Surveys (NPSAS), and a review of the literature dealing with the effect of debt on student persistence.

Higher Education Act

Financing of higher education in the 1950's and early 1960's was marked by a resistance toward general scholarships to undergraduates. The breakthrough for students came with the groundswell accompanying the "Great Society" (Gladieux & Lewis, 1987). President Johnson enjoyed an enormous amount of success on the domestic front in 1965 with historic legislation such as the Voting Rights Act (42 U.S.C. § 1971 et seq.), and the Elementary and Secondary Education Act (ESEA) (20 U.S.C. § 6301 et seq.). The Higher Education Act of 1965 (20 U.S.C. § 1001 et seq.) was passed in the wake of ESEA, and featured the first program of federal scholarship assistance to undergraduates (Gladieux & Wolanin, 1978). Every few years HEA is reauthorized, in an extremely complicated process. New programs are added, old programs are extended, changed, or deleted, in a process that affects six million college students. This section will explore the provisions of the Higher Education Act of 1965, and its subsequent revisions through 1992.

Higher Education Act of 1965

The Higher Education Act (HEA) of 1965 embraced two approaches to federal support for higher education: aid to the students and aid to institutions (particularly predominantly black institutions in Title III). Aid to institutions took the form of construction of academic facilities, and funding of academic program improvement efforts. Student aid was focused primarily on providing access for needy students (Aaron, 1978). The focus of this section is the student aid programs.

The debate surrounding federal support for higher education in 1965 was centered around two major policy issues: "should the government henceforth concentrate its higher education efforts on programs that assisted institutions or on those that provided aid primarily to students?" (Finn, 1978, p. 61). Many issues, including some that will have a common thread through subsequent proposals, were debated in 1965. The advocates of institutional aid favored a formula funded approach of general support for higher education institutions. They envisioned this approach as the first step to universal, free, public, higher education financed in large part by the federal government. Those in favor of student aid considered it a more efficient mechanism for equalizing opportunities in higher education (Rivlin, 1987). The original bill was a compromise with provisions for institutional aid in Title III, and student aid in Title IV. Although institutional aid received the lion's share of the first authorization, the distinction was short-lived (Kimberling, 1995).

For college student aid, HEA was designed to increase the pool of college trained employees, and correct "the appalling frequency with which a student is forced to forego the opportunity of post-secondary education because of inability to meet the costs" (S. Rep. No. 673, 1965). Title IV of the program created two new types of student assistance: opportunity grants for poor students, and federally guaranteed loans for low- and middle-income students. HEA also transferred an expanded College Work Study program, established in 1964 under the Economic Opportunity Act (42 U.S.C. § 2701 et seq.), from the Office of Economic Opportunity to the Office of Education.

Title IV was the first explicit federal attempt to equalize college opportunities for needy students. The centerpiece of Title IV was the Educational Opportunity Grants (EOG), but it also contained Talent Search and Upward Bound programs. Both of these programs were designed to identify and foster access to college for poor minority students. Many of the specific provision of the EOG were unclear, and colleges were left with the guidance to "vigorously" identify and recruit students of "exceptional financial need" (Gladieux, 1995; Mumper, 1996).

According to a number of authors (Gladieux & Wolanin, 1976; Gladieux, 1995; Mumper, 1996) the Guaranteed Student Loan Program (GSLP) was a compromise solution developed to assist the middle-class student attend college. The competing proposal was a tax credit for parents with students in college, and the GSLP was considered a less costly option. The GSLP provided low-interest loans that were not repayable until twelve months after a student left school. The Act provided $1,500 a year up to a maximum of $7,500 for graduate students, and $1,000 per year up to a maximum of $5,000 for undergraduates (Pub. L. No. 98-326).

College Work-Study was reauthorized in the 1965 Act, even though it was originally considered an anti-poverty measure, rather than a student-aid program. The Perkins Loan Program originally established in the 1958 National Defense Education Act (20 U.S.C. § 401) was also reauthorized in HEA.

1972 Reauthorization of the Higher Education Act

The intervening years between the original passage of HEA and the 1972 reauthorization saw a continued growth in student aid which eclipsed the other provisions of the original bill, including the construction of academic facilities. With the backdrop of a proposal by the Nixon Administration to consolidate all student aid programs into a single basic grant, and a strong legislative contingent advocating grants led by Senator Pell, the resulting 1972 reauthorization established the framework of the current student aid system.

As stated in the previous section, the original legislation was a compromise between the institutional- and student-aid proponents, however, by 1972, student aid was the predominant method of financing higher education at the federal level. Congress refused to repeal the campus-based programs like College Work Study, and Perkins Loans, but did adopt a new, portable voucher-like system for needy students as the foundation for all forms of aid. These new Basic Educational Opportunity Grants (BEOG - later renamed Pell grants after Senator Pell) authorized a maximum $1,400 grant for needy students. The Pell grants differed from the earlier EOG Program in that the students applied directly to the federal government instead of the institution. "This insured that students with similar economic backgrounds, and attending similar institutions would receive the same size grant" (Mumper, 1996, p. 82). Pell grants would be awarded solely on the basis of need, and that need was determined by a formula which was administered at the federal level.

It was envisioned that Pell grants would establish a minimum level of resources to help assure access to higher education, and that other programs would provide supplemental aid to students and institutions. In that vein, Congress authorized the State Student Incentive Grant (SSIG) program and the Student Loan Marketing Association (Sallie Mae) (Fenske & Boyd, 1981). SSIG provided matching dollars to states that established their own need-based aid programs. Sallie Mae was a complimentary agency that was established to repurchase existing GSLs to provide additional capital for new loans (Hannah, 1996; Mumper, 1996).

The debate on the 1972 reauthorization of HEA included not only the Congress and the White House, but a plethora of higher education associations and interest groups (Hearn, 1993). Their impact was felt in the legislation by the substitution of "postsecondary education" for "higher education." The intent was to broaden the options available to students further than the traditional baccalaureate degree from a four-year institution. The 1972 HEA amendments extended federal financial aid to vocational education, community colleges, trade schools, proprietary schools, and part-time students (Mumper, 1996).

The GSL program was also amended during the 1972 reauthorization to play a secondary role to Pell grants. The 1972 reauthorization mandated a needs test, and students whose families earned less than $15,000 now found themselves having to prove need. Because of the timing of the effective date of the changes in the GSL program, thousands of students had their applications delayed or denied. Congress passed emergency measures that took the GSL program regulations back to pre-1972 status in order to avoid a pending crisis. Seven months later Congress amended the law so that families with less than $15,000 annual income would qualify automatically, and families making above $15,000 would qualify for subsidized loans based on need (Gladieux & Wolanin, 1976).

1976 Reauthorization of the Higher Education Act

The foundation of student financial aid programs were firmly in place with the amendments of 1972, and student aid passed through a period of increased funding with each reauthorization with little or no conflict. The 1976 reauthorization provided incentives for states to establish loan guarantee agencies in an attempt to get banks to lend more money for higher education. In addition, the intent of the program was changed to help students with "the ability to benefit" (ATB) from postsecondary education. Hearn (1993) suggested that this reauthorization was more notable for the "refinement, and expansion of policy, not for dramatic conflict" (p. 111).

The 1976 amendments raised the maximum Pell Grant award to $1,800, and relaxed the eligibility requirements slightly. Congress, however, provided only enough funds to cover $1,600 per year for each student. The GSL program limit was increased so that a student could borrow $15,000, up from the previous limit of $10,000 (Finn, 1978; Mumper, 1996).

1978 Amendments to the Higher Education Act

In an off-year of the reauthorization cycle, a controversy arose over the perceived plight of middle-income students. This period in the development of student aid policy was highlighted by considerable conflict among the White House, Republican members of Congress, public and private schools, and financial aid associations. As was the case in 1965, the idea of tuition tax credits for parents of college students was seen as a method to assist middle- and upper-income families. The White House and proponents of the general approach to financial aid embodied in the Higher Education Act, were able to structure an alternative way to provide federal assistance to middle-class students. The Middle-Income Student Assistance Act (MISAA) maintained the basic features of the then-current programs, but adjusted them to make many more students eligible (Mumper, 1996; Hearn, 1993, Gladieux, 1995: Brademus, 1987).

Both the tax credit plan and MISAA would have substantially increased federal spending for financial aid. Mumper (1996) stated;

The politics that unfolded during the debate over student aid in 1978 amounted to little more than a bidding war between supporters of the two approaches. Proponents on each side continued to expand their package to appeal to even larger numbers of middle- and upper-income families. (p. 88)

Ultimately, the supporters of the tax credit plan could not agree on size and scope, and the proposal died in committee. MISAA allowed the award of Pell grants to students from families making less than $25,000, up from $15,000. In addition, a large number of independent students were now eligible for Pell grants under new asset exclusion policies. Independent students with dependents had the same asset exclusions under MISAA that dependent students had before 1978. The most significant change, however, was the removal of the income ceiling (previously $25,000) for subsidized loans under the GSL program. The removal of the income ceiling made guaranteed subsidized loans available to all students (Mumper, 1996), initiating a philosophical shift in financing student aid programs to more of a reliance on loans.

1980 Reauthorization of the Higher Education Act

The administration of student aid programs was described by Califano (1981) as a monumental mess. Student loan records were kept on index cards, and alleged fraud was widespread. Over a third of the 3.4 million applicants for aid in 1978-79 were disqualified because of inaccuracy, incompleteness, and inconsistencies in the government's first attempt to screen applicants (Califano, 1981). The 1980 reauthorization was accomplished in the midst of reports of mismanagement, budget shortfalls, and a presidential election campaign. For the first time, both the Congress and the White House agreed to the curb the growth of student financial aid programs.

The 1980 amendments increased the interest rates on GSLs from 7 to 9 percent, and shortened the after-school grace period before repayment started. A new program, Parents Loans for Undergraduate Students (PLUS), was started. This program was an attempt to shift some of the federal government's cost to the parents by allowing parents to borrow $3,000 for each dependent child. However, the effect of these cost cutting moves will never be known due to the election of President Reagan, and a Republican majority intent on reducing domestic spending. Many of the provisions of the 1980 reauthorization were never implemented, with the act being amended in 1981 to further curb the growth in federal spending. The needs test for families earning more than $30,000 was reinstated, and the loan amount was limited to only the difference in educational costs and expected family contribution (EFC). In addition, an origination fee of 5 percent was allowed for PLUS loans (Mumper, 1996; Hearn, 1993; Hannah, 1995).

In 1982, 1983, 1984, and 1985, President Reagan proposed dramatic cuts in all student aid programs. Congress rejected all the proposals with the backing of the higher education community. The pattern of a stalemate between the White House and Congress was firmly established.

1986 Reauthorization of the Higher Education Act

The federal budget situation, an intransigent president, and a growing awareness that Title IV programs were not achieving their intended objectives all preceded the consideration of the reauthorization of HEA in 1986. Proposals were advanced, primarily by Congressman Ford, to expand the programs so that low-income students could cover the increasing costs of college attendance. Alternative plans were advanced to direct existing funds more toward the low-income students; while limiting availability, but, with increased loan limits, for middle-income students (Five Year Bill Cleared, 1986).

The final compromise, according to Mumper (1996),

. . . did nothing to reverse the declining value of Pell grants to low-income students. It increased dependence of many students on borrowing for college. It increased the long-term costs of the program, and it accelerated the shifting character of federal aid toward loans away from grants (p. 98).

The amendments included the reinstatement of the needs test for loan eligibility, and it limited loans to the amount of need for educational costs. In an effort to cut costs, family assets were added to the needs analysis calculation, and a 3 percent insurance fee was charged to all borrowers to ease the default problems. Loan limits were increased so that freshmen and sophomores could borrow up to $2,650 annually, and upperclassmen could borrow $4,000. These amounts were increased from a pre-1986 level of $2,500. The aggregate loan amount for undergraduates was raised from $17,500 to $25,000. The maximum annual amount for graduate school was increased to $7,500. The cumulative amount that a student could borrow for both undergraduate and graduate school was $54,750, more than twice the previous limit of $25,000 (Gladieux, 1986).

1992 Reauthorization of the Higher Education Act

Prior to 1992, the 1988 and 1990 Budget Reconciliation Acts had raised loan limits and interest rates. In addition, income caps were reinstated for GSL, and restrictions added to deal with the alarming increase in default rates (Hannah, 1996). The atmosphere surrounding the 1992 reauthorization was pregnant with anticipation. Expectations that the "policy drift" of the preceding 12 years were about to end were matched by the rhetoric of Congress (Gladieux, 1986). The dilemma, however, was that there was substantial disagreement over which of the problems were the most critical: loan defaults, program costs, borrowing for middle-and upper-income students, or access for low-income students (Mumper, 1996). What resulted was the most far-reaching changes since the HEA 1965 legislation.

Congressional leaders, particularly Congressman Ford, wished to make Pell grants an entitlement. This effort was in response to the lack of funding for the grant program. For example, in 1991, the maximum authorized Pell grant was $3,100, but Congress only appropriated enough to fund $2,400 of that amount (Mumper, 1996). However, an unexpected budget shortfall, Education Department overruns, and White House resistance killed the entitlement movement. While the entitlement idea failed, Congress did change Pell grant eligibility by eliminating home and farm equity in the need calculation, thereby combining Pell Grant needs-analysis formula with the Congressional Methodology formula that was used for other programs. This created a single formula for all federal financial aid programs. The maximum grant was increased to $3,700 for the 1993-94 academic year, $3,900 for 1994-95, $4,100 for 1995-96, $4,300 for 1996-97, and $4,500 for 1997-98. In addition, Pell grants were extended to part-time students on a pro-rata basis (The Higher Education Amendments of 1992, 1992).

The 1992 legislation created an unsubsidized loan option not restricted by student or family income. This new loan program was created with the same limits as the need-based GSLs. Loan limits were increased to $3,000 for undergraduates and $5,000 for graduate students at eligible institutions. The loan limits were increased for both undergraduates and graduates by $1,000 for students attending institutions that had low default rates, and had agreed to increase their contribution to the program. The reauthorization also increased loan limits to $3,500 for lower level undergraduates, and $5,500 for other undergraduates, effective on July 1, 1993. Graduate student loan limits were increased to $8,500 effective October 1, 1993. Maximum debt was established at $23,000 for undergraduates, and $65,500 for graduate students, which included debts incurred as undergraduates.

The 1992 legislation retained the $4,000 limit for freshmen and sophomores in Federal Supplemental Loans for Students program, but increased the limit to $5,000 for other undergraduates, and to $10,000 for graduate students. The amendments also specified that undergraduates were permitted to borrow a total of $23,000 under the supplemental loan program, and graduate students were limited to a total of $73,000, which included their undergraduate debts. In addition, the $4,000 limit on PLUS loans was eliminated, but the legislation specified that loans cannot exceed the difference between college costs and other forms of aid received. Interest on most programs was pegged at 3.1 percent above specified Treasury Bills ("The Higher Education Amendments of 1992," 1992; Mumper, 1996).

The effect of the changes was to make larger sums of money available to a larger number of students, mainly in the form of loans. The appearance of additional Pell grants was labeled a "Charade" by Mumper (1996).

The Congress pretended to provide lower-income students with more resources to attend college, but it failed to provide the money to make those grants available. It pretended to provide middle-income students with more aid, but that aid was noting more than access to larger loans which the student will be required to repay with interest. For both groups, then, what had appeared to be much-needed federal help to cover rising college prices, was in reality only the opportunity to take on additional debt, and to finance an even greater portion of their educational costs. (p. 104)

As was to be expected, the student loan volume soared as a result of the 1992 reauthorization. Total educational borrowing increased by $10 billion between academic years 1992-93 and 1994-95 (Gladieux, 1995).

Post-1992 Changes in Student Financial Aid Policies

The 1992 reauthorization created a direct lending demonstration project for the GSLP. This program eliminated the middlemen (primarily banks) in the loan making process. It was envisioned that by administering the program on campus, institutions could become more efficient. President Clinton, shortly after the election, proposed immediate implementation of the direct lending program. The compromise was a gradual phase-in of the direct lending concept through 1998-99.

Clinton also proposed the National and Community Service Trust, a program of national service as a means of providing funding for college or job training. The program was coupled with income contingent loan repayment plans, and a loan forgiveness plan. The premise was that students could take low-paying public service jobs without worrying about how to repay after college. Supporters of student aid do not see the National and Community Service Trust as a student aid program. Those monies spent on national service do not reduce Title IV funds thereby providing new opportunities for Title IV to help additional needy students (Mumper, 1996).

The 1992 Reauthorization created State Postsecondary Review Entities (SPREs) to help the federal government determine institutional eligibility for the GSLP. Congress expected the Department of Education in conjunction with the accrediting agencies and the states to tighten eligibility criteria for institutions ("The Higher Education Amendments of 1992," 1992). This experiment was a failure, and SPREs went quietly out of existence shortly after their creation.

Philosophical Foundations of the Higher Education Act

Federal support for student aid has been, and continues to be, motivated by a number of various unconnected objectives. Legislation has passed under a variety of philosophies, including equalizing opportunity for disadvantaged students, protecting the interests of the middle-class, rewarding deserving students and institutions, protecting private institutions, fostering the development of proprietary institutions, and furthering national economic interests (Gladieux, 1983; Leslie & Brinkman, 1988; Hearn, 1993). Hearn (1993) sates that the path the legislation has taken represents a lack of "philosophical coherence" (p. 95). The development of student financial aid policy resembles the type of compromise, and reaction to public concerns typically found in a democratic society. The following sections will explore this lack of "philosophical coherence."

Initial Philosophy of the Higher Education Act

The original legislation creating the Higher Education Act in 1965 was the result of two competing ideologies for federal funding. The first was that of assisting higher education by funneling funds directly to institutions through some all-encompassing formula based on student counts. The second was that of assisting needy students directly through grants to the poorest students, and subsidized loans to the lower- and lower-middle class students. Even though the 1965 legislation was indeed a compromise between these two competing interests, the federal financial aid policy has, since its inception, been powerfully shaped by the "needs" ideology.

The clash of institutional aid proponents and student aid proponents can be viewed in the sense of the status quo versus the oppressed. In 1965, the primary objective was one of equalizing educational opportunity for socioeconomically disadvantaged students with the primary goal of HEA to provide equal access to higher education for all students. The idea of leveling of society and the power of the oppressed (poor students) is clearly a Marxist philosophy. While Marx did not particularly dwell on education in his writing, his proponents have been "major advocates of making resources available to everyone, and public education is one of the social goods they have advocated." (Ozman & Craver, 1992, p. 346).

Choice and Access

The 1972 Act broadened the policy objectives to access and choice, and established the foundational role of the federal government in student aid. Direct portable aid to needy students was solidified in the 1972 amendments. Many authors (Gladieux, 1983, 1987; Hearn, 1993; McPherson, Schapiro, & Winston, 1993) see 1972 as policy refinement and expansion. Changes in 1972 not only broadened student choice, but broadened eligibility for a whole new group of institutions. The access and choice question is essentially a public school and private school dichotomy. Public schools offer low tuition and very little institutional based aid. Low tuition implies that most of the aid at public schools will go to low-income students, since middle- and high-income students will have no need. Therefore, federal student aid acts to supplement the low-tuition strategy of public institutions, and promote access. Private schools represent the choice part of the access and choice dichotomy. Private institutions, in general, spend a large portion of their resources on student aid through tuition discounting. Therefore, federal student aid "federalizes" some of the costs of student aid that these institutions would incur in the natural course of business. Secondly, the need-based federal financial support focuses more on the lower-income students than the college would without the federal subsidy. Is it fair to subsidize private higher education with public dollars? This question is best answered in relation to one's position on the question, Should the federal government bear part of the cost of student "Choice?" Coupled with the funding of choice, the proprietary institution sector developed rapidly after the change of "higher education" to "postsecondary education" in the definition of eligible training in the 1972 amendments.

Broadening the eligibility requirements furthered the initial philosophy of need-based aid for needy students, and allowed them to take that aid to additional institutions and programs. Changes in 1972 clearly established the philosophy that equal opportunity underlies all federal student financial aid, and that student needs take precedence over institutional needs. These changes came from Congressional initiatives, and involved little or no White House pressure, or conflict. What started as a small perceived change in policy resulted in substantial constituent pressure on Congress to address the perceived plight of the middle class student (Hearn, 1993; Brademus, 1987). While the Marxist philosophy of aid, and education for the underprivileged was expanding to include a more equalizing situation, the very act of broadening "access" and "choice" worked to change the philosophical focus of federal student financial aid.

MISAA and a Further Philosophical Shift

The passage of MISAA was a fundamental shift in federal financial aid policy and philosophy. Easily accessible, non-need based aid became available to large numbers of middle- and upper-income students on a quasi-entitlement basis. Federal student aid outlays increased by 59 percent between 1977-78 and 1980-81 (Hearn, 1993). The lower-income targeted, need-based, grant oriented federal policy period came to an end. While the stated goals of MISAA were to promote educational choice, persistence, and access for both middle- and lower-income families, upper- income families took advantage in substantially larger than expected numbers.

The push for the changes came from two sources; the middle-class constituencies of politicians, and the intense lobbying efforts of banking officials (McPherson, 1989). Politics and power, based on a perceived threat to the middle-class status quo, combined to reverse the original intent of federal financial aid policies. After the passage of MISAA, enrollments of lower-income students declined (Mortenson, 1990).

The philosophical shift was as dramatic as the policy shift, from one of an ideal Marxist philosophy where the powerless are benefitted to one more akin to Foucault's philosophy of the power of the powerful. Clearly, the shift of federal financial aid from the lower-income students to the middle- and upper-income students assisted by the banking industry represent a shift from the powerless to the powerful. Although Foucault uses a prison as an example in Discipline and Punish (1979), the prison is a structural model of how power is distributed to and internalized by its subjects in a larger culture. Any institution which has some sort of production as its goal, for examples schools, prisons, factories, and legislatures can be organized in the spirit of the Panopticon, where the exercise and dissemination of power is controlled by a small, privileged group over a large unprivileged populous. The Foucauldian vision is one where power is expressed, disseminated, and reproduced in its subjects to stratify its subjects. Wilensky (1975) notes "The greater the educational opportunity (measured by enrollment rates or fraction of the children of lower strata who go on to higher education), the more chance the middle-class parents of college students as well as large numbers of highly educated will resist the push from below for equality" (p 54-55).

The continuing conflict of the powerful and the powerless combined with a severe fiscal crisis in the 1980's to de-stabilize federal policy. Many of liberalizing effects of MISAA were scaled back by 1991, and direction, operation, and control were to be the themes of federal financial aid during the next decade. The period starting around 1980-81 has been called a period of policy drift (Hearn, 1993; Gladieux, 1995). The stalemate between the White House and Congress, and the federal budget became the major domestic issues, and federal financial aid policies became only background music to the Washington melodrama.

Reauthorization efforts in 1986 began with serious consideration of efforts to reform the GSL program, but ended in frustration and stalemate (Mumper & Vander Ark, 1991). Attempts at trimming programs met with strong resistance, and the GSL program continued to grow unabated, while at the same time the purchasing power of Pell grants declined. The major occurrence during this period was the dominance of loans over grants. Mumper and Vander Ark (1991) believe that Pell grants suffered in relation to loans, because they appealed to lower-income students, and middle-income students, banks, and private colleges, all represented by well-funded lobbying groups, which favored loans.

Loans Replace Grants

The 1992 reauthorization of HEA was in a significant shift in federal policy from a commitment to promote access through need-based grants to a broader strategy of loans regardless of family income or need. Congress planned to make more loans available to the middle-class student, decrease the reliance on loans by increasing grant support, and restore the integrity to the loan program. The results were mixed.

The relaxed eligibility for government subsidized loans resulted in an increase of 2 million additional students receiving loans between 1990 and 1996, with a concomitant 92 percent increase in money borrowed. With only a 16 percent increase in the Pell grant program, the imbalance between loans and grants worsened (Hartle, 1996). The efforts to restore integrity to the program has been successful with defaults rates dropping from an aggregate 22.4 percent in 1990 to 10.7 percent in 1994 ("FTEL Default Rates," 1997).

The philosophical shift from need-based grants to promote access for low-income students to loans for every student regardless of income can be analyzed in several ways. From a Marxist view, this shift is a classic example of bourgeois domination of the lower classes. Western Marxist, like Feriere, would contend that the shift was the result of the capitalist bankers, politicians, and middle-class marginalizing and de-humanizing workers, and that everything could be explained as forms of oppression towards race, ethnicity, gender, and socioeconomic status. Using Wilensky's (1975) analysis, the upper- and middle-class white, male, educated majority do not want non-white, female, or poor students attending college. These under represented groups might become educated and therefore powerful. The Foucauldian philosophy of the powerful over the powerless are equally as applicable in 1992 as it was in 1978.

Student Persistence

Early research on persistence (Edgerton, 1929; Hoke, 1922; McNeely, 1938; Minnesota University, 1924; Tallman, 1927) tended to be cross-sectional in design, the results of which were difficult to interpret. Many times the research showed what had happened and not necessarily why, with the major thrust of early research centering on the relationship between intelligence and persistence (Bragg, 1956; Hoke, 1922; Iffert, 1957; Tallman, 1927). In one of the earliest studies, Hoke (1922) concluded that intelligence as measured by four different intelligence tests popular during that decade did not furnish a satisfactory basis for prediction of academic success in a group of entering freshmen at Hood College. Tallman (1927) examined retention at the University of Iowa using the Class of 1920 as the sample. Tallman attempted to link intelligence as measured by the Thordike Intelligence Examination to student persistence, but found that the primary reason given for leaving school was financial. Tallman also concluded that just as many students with high intelligence scores dropped out for financial reasons as those with low intelligence scores.

Iffert (1957) examined 13,669 college students from a wide range of institutions ranging from technical to elite colleges for the U.S. Department of Health, Education, and Welfare in 1956. Iffert found that the primary reason for withdrawal from school in the first year was poor grades. Financial difficulties was the primary reason for withdrawal after the first year. Several studies (Grace, 1957; Holmes, 1959; Iffert, 1957; Maxwell, 1960) examined the role of psychological and sociological factors associated with college persistence. These factors tended to be slightly better predictors of persistence than general intelligence, but as with the earlier studies the causative inference was difficult to discern from the analysis (Astin, 1964). Again, the studies tended to report the results of what had happened and not why.

Persistence Theory Development-Sociological

Astin (1964) examined 6,600 high aptitude students drawn from the 1957 National Merit Scholarships competition over a 4-year period on their tendency to drop out of college before completing a degree. Data were collected from two surveys, and high school and college records. Each student completed a survey shortly after entering college in 1957, and a second survey was administered late in 1961 about the time the class was expected to graduate.

It was found that students who drop out came from lower socioeconomic backgrounds, had lower ranks in high school, planned initially to get lower degrees, and applied for relatively fewer scholarships than do students who do not drop out. Personality measures suggested that dropouts tended to be more aloof, self-centered, impulsive, and assertive than non-dropouts. There were no significant effects found in the institutional variables for men, but women were more prone to drop out if they attended a college with a high proportion of men.

Spady (1971) postulated that the decision to leave college is a complex social process that includes family and previous educational background, academic potential, friendship support, intellectual development, grade performance, social interaction, satisfaction, institutional commitment, and a term he called normative congruence. Normative congruence referd to the general degree of compatibility between the disposition, attitude, interests and expectations of the student and the set of behaviors, expectations, and demands that they may be exposed as a result of interaction in the college environment.

Applying Durkeim's (1951) theory of social integration, Spady hypothesized that the relationships among family, previous educational background, academic potential, friendship support, and normative congruence jointly affected intellectual development, and grade performance. All of the above factors had a direct influence on social interaction which had a direct influence on satisfaction, and satisfaction led to institutional commitment. Institutional commitment and grade performance directly informed the dropout decision.

Spady concluded that the model was much more complicated that he originally hypothesized, and that there were different variations of the model according to gender. Spady also concluded that intellectual growth is unrelated to previous high school performance, greater commitment can be generated by providing students experiences that affect the intrinsically meaningful spheres of their lives rather than treating them just as a student, and that persistence and academic achievement become more synonymous over time.

Tinto (1975) theorized that attrition resulted from interactions between a student and the educational environment during college, and observed that persistence is a function of an individual's integration into the academic and social systems of the college given the individual's characteristics, family background, and commitments. Tinto (1975) applied Durkheim's (1951) theory of suicide to college student attrition in the following manner:

When one views the college as a social system with its own value and social structures, one can treat dropout from that social system in a manner analogous to that of suicide in the wider society. . .social conditions affecting dropout from the social system of the college would resemble those resulting in suicide in the wider society; namely insufficient interactions with others in the college and insufficient congruency with the prevailing value patterns of the college collectivity. . .lack of integration into the social system of the college will lead to low commitment to that social system and will increase the probability that individuals will decide to leave college and pursue alternative activities. (p. 91-92)

Tinto proposed that background characteristics (e.g., social status, high school experiences, neighborhood, etc.) were important, but so were the expectations and motivational attributes individuals bring to college. The latter were termed educational and goal commitments, and introduced psychological variables into models of student attrition. In addition, a student's commitment to a particular institution (institutional commitment) was important. Other things being equal, the higher the degree of integration of the individual into college systems, the greater was his commitment to the specific institution and to the goal of college commitment. Low goal or institutional commitment led to dropout. The interaction between and among goal and institutional commitment, and integration can led to different forms of dropout, transfer, and "stick it out behavior."

In a series of studies testing Tinto's student-integration model, Pacscarella (1980) and Pascarella and Terenzini (1977, 1978, 1979, 1980, 1983) confirmed the critical role of faculty-student interaction in student persistence, and supported the predictive validity of Tinto's (1975) model of student attrition. In their early studies, Pascarella and Terenzini (1977, 1978, 1979) suggested that those students persisting had more positive perceptions of their academic and non-academic lives than those withdrawing. Persisters reported significantly more contacts with faculty members than those students leaving college. Pascarella and Terenzini (1979) concluded;

Various measures of frequency and quality of student-faculty relationships made significant contributions to the prediction of male and female persistence with the influence of twelve entering characteristics and five other measures of social and academic integration were held constant (p. 209)

In 1983, two studies (Pascarella & Chapman, 1983; Pascarella & Terenzini, 1983) tested Tinto's theoretical model with a path analysis. Each study used different population groups. Pascarella and Terenzini tested the validity of Tinto's model for first-year students at a residential college. Their model correctly predicted 80 percent of the persisters and voluntary withdrawals. Pascarella and Chapman collected data from eleven institutions, including residential, commuter, and two-year colleges, and found that Tinto's original model explained approximately 75 percent of the variance in persistence/withdrawal decisions.

Persistence Theory Development-Financial

Beginning as early as 1980, and accelerating in the 1990s, researchers began to incorporate financial aid variables in persistence models. Bean's model (Bean, 1980; Hossler, Bean & Associates, 1990) included social and academic integration variables, but also addressed the issue of financial support in background variables. Cabrera and Nora (Cabrera, Castanada, Nora, & Hengstler, 1992; Nora & Cabrera, 1993) compared the Tinto and Bean models using a questionnaire given to all full-time students enrolled at a large southwestern university, and concluded that a model that included integration and financial variables explained more of the variance in persistence decisions. In one of the earliest works to view persistence as a dichotomus process, Cabrera, Nora, and Castaneda (1992) studied how finance influences college persistence using a variant of Tinto's student integration model. In addition to the sociological, organizational, and psychological influences advanced by Tinto and Bean, this study incorporated the variables related to financial aid. The study concluded that receiving some form of financial aid facilitated students' interactions with other students, because they did not need to secure employment. The availability of financial aid gaves the student more time to become integrated into the social and academic realm of the institution, which according to Tinto, enhances student persistence. Cabrera, et al. concluded further that financial aid may have a two fold effect. On one hand, by reducing the burden of meeting financial costs the attractiveness of alternative activities, such as transferring or entering the labor market is reduced. Secondly, the institution may be seen as an instrument in securing additional financial aid, thereby increasing the commitment to the institution.

This interaction increased the student's satisfaction, which the authors hypothesized had the same effect as receiving financial aid. That is, in receiving financial aid a student's satisfaction with overall finances may reduce concerns about finances, allowing students to spend more energies and efforts to academic development and intellectual endeavors. In sum, the results appeared to suggest that financial aid, and it effects on satisfaction, commitment and attitude, was important not only because it equalized opportunities between affluent and low-income students, but because it facilitated the integration of students into the academic and social components of the institution.

Somers and St. John (St. John, 1989,1991, 1992,1994, St. John & Andrieu, 1995; St. John, Paulsen, & Starkey, 1996; Somers & St. John, 1996, Somers, 1992, 1993b, 1994a, 1995) developed and tested a model to study first-time enrollment and persistence using existing institutional data. They found that aid could have a significant impact on student attendance and persistence decisions; likewise, the lack of aid could also influence decisions.

St. John (1989) found that both loans and grants had a positive impact on persistence during the first three years of college. However, no form of aid was significant in the fourth-to-fifth-year transition (graduation). This implied that the closer a student came to graduation, the less aid influenced persistence. St. John et al. concluded that social and educational background had differing effects on persistence at different points in college. The college experience (full-time attendance and grades) and financial aid had a positive effect on persistence for each year of attendance. Finally, they found that federal grants and loans had an impact on persistence, especially long-term persistence.

In another national study, St. John (1990) examined how the amounts of financial aid and tuition charges affected year-to-year persistence. St. John found that persistence decisions were more responsive to increases in aid (grants, loans, and work study) than increases in tuition. In this study, students were more responsive to changes in tuition only between the second and third years of college.

Somewhat outside the mainstream of current persistence research (Cabrera, Nora, & Castaneda, 1992; Cini & Harden Fritz, 1996; Okun, Ruehlman, & Karoly, 1991; Rusbult, 1980) is a series of articles using Rusbult's (1980) Investment Theory to explain student persistence decisions. Rusbult (1980) proposed a theoretical model of investment in order to explore continuation of close relationships. This investment theory posited that the departure from or persistence in an organization was influenced by commitment, investment, satisfaction, and alternative value. Accordingly, students' persistence at a particular institution should increase if they experience more rewards than costs, have few or no good alternatives, and have invested many resources, time, and money. Cabrera et al. (1992) concluded that receipt of financial aid in the form of scholarships or grants may have been viewed as a form of recognition by the recipient, thereby increasing their satisfaction with their current institution. In addition, the Cabrera et al. study suggested that financial aid decreased the burden of meeting costs, thereby decreasing the attractiveness of alternatives, such as transferring or quitting school. In addition, Cabrera et al. concluded that receipt of financial aid eliminated the need to secure employment, and if they had employment, receipt of financial aid allowed them to spend less time at that job. This allowed more time for social integration. When combined, these factors reinforced the students' commitment to a particular institution, in that the current institution was viewed as instrumental in securing future aid. These factors all tended to increase motivation to perform at a high level. According to Rusbult's Investment Theory, the larger the commitment and satisfaction, and less attractive the alternatives, the more one is willing to invest in maintaining the relationship.

According to the results of Cabrera et al. (1992) and confirmed by Cofer and Somers (1997), students who intend to complete their degrees at both the bachelor's and master's level were more likely to persist, representing an intent to persist. Second, as the student classification changed from sophomore to senior, the likelihood of persistence increased. This commitment to persist is coupled with the assumption that as the grant amount increased, the satisfaction level increased. The increase in social integration, which influenced satisfaction and commitment, was increased not only by continued progression through the class levels, but also by the finding that on campus residency increased the likelihood of persistence. On the other hand, working full-time, which decreased social integration, and therefore satisfaction, have a negative influence, and accordingly students leave the relationship.

Using the Rusbult theory, several scenarios - both positive and negative - are suggested. Non-persistence would cause the repayment of larger amounts of accumulated debt in the face of a tough job market, and the loss of expected grants (i.e., income). Persistence would enable the student to capture those future grants and continue to postpone repayment of accumulated debt.

Intent to persist, satisfaction with and commitment to the institution, and lack of viable alternatives, according to the Rusbult Theory, would encourage students to increase their investment, via increased loans, tuition, and living expenses to maintain their relationship with the institution.

Persistence Theory Development- NPSAS

Research on student persistence was given a big boost by the National Postsecondary Student Aid Study (NPSAS) of 1987. Much of the research prior to NPSAS had been at the institutional level or using small, select national samples. NPSAS provided an opportunity to study persistence using a large, national sample with an extensive set of economic, demographic, and college experience variables. NPSAS contains a wealth of data of interest to institutional researchers, policy makers, and college administrators. However, previous versions of the database, especially NPSAS:93, have been underutilized. The new licensing procedure mandated by Congress with strict security and the complexities of NPSAS:96 will make this database even less accessible to researchers and policy makers.

The National Center for Educational Statistics (NCES) hires independent contractors to collect the data for NPSAS. Several of these contractors and NCES analyze the data in the Surveys using descriptive statistical techniques. The non-empirical studies (Byce & Schmitt, 1992, 1993; Choy, 1994, 1995; Choy & Premo, 1996; Edgerton & Toops, 1929; Horn & Maw, 1995a; Horn & Maw, 1995b; Horn, 1995; Lee & Clery, 1995; McCormick, 1995; Pelavin Associates, 1992; Tuma, 1994; Tuma, 1995) produced by NCES and their contractors will not be reviewed in this study. Several other studies examined student loan default (Dynarski, 1991; Dynarski, 1994; Volkwein, 1995; Volkwein & Szelest, 1994).

Missing from these surveys is data associated with social and academic integration of the student, and variables representing institutional and goal commitment. Therefore, the models developed using NPSAS approach persistence research from a different theoretical perspective than the previously accepted models. The definitive model for use with NPSAS data was primarily developed through a series of studies using NPSAS:87 and NPSAS: 90 done by St. John and his associates (Andrieu, 1990, 1991; Andrieu & St. John, 1993; St. John, 1991a, 1992, 1994; St. John & Andrieu, 1995; St. John, Paulsen, & Starkey, 1992; St. John & Starkey, 1995; Starkey, 1993; Trammell, 1994).

St. John, Paulsen, and Starkey (1996) examined the influence of finances on college choice and persistence decisions. Financial choices had direct and indirect influences on college persistence, and market-based, monetary measures of financial aid, tuition, housing costs, and other living costs had substantial direct effects on persistence. Another study (St. John, et al., 1992) focused on persistence by traditional college students. Their persistence was influenced by the tuition amount and grants were positively associated with persistence in private colleges but negatively associated with persistence in public colleges.

Several studies (Andrieu, 1990, 1991; Andrieu & St. John, 1993; St. John & Andrieu, 1995) examined within-year persistence of graduate students. The conclusions were that tuition charges had a substantial negative effect on persistence and comprehensive aid packages (grants, loans, and assistantships together) were the most effective means of subsidizing costs. Proprietary school students were the subject of one study (St. John & Starkey, 1995). African Americans, Hispanics, and students without high school degrees were more likely to persist, although tuition charges had a substantial negative influence on student persistence. St. John, Oescher, and Andrieu (1992), using a sample drawn from NPSAS: 87 of traditional age four-year college students, concluded that tuition did influence within-year persistence decisions, and in a negative direction. Financial aid had less of an influence on persistence than tuition, but in most cases the effect was negative. Higher tuition had a larger effect on public school persistence than private school Grants had a positive effect on persistence in private schools and a negative effect in public schools. Loans also had a negative effect on persistence decisions in public schools. The explanation offered by the authors was that grants did not keep pace with tuition increases in relation to tuition increases, due to modest increases in Pell grants. In other words, the amounts of unmet needs could explain the negative association between grants and within-year persistence in public colleges. These findings provided insight into the effect of the Bush and Regan administrations' policies on financial aid. Student aid at the time of the study was decreasing, and it appears that private schools increased their grants to account for the changes. Public schools, however, did not and their persistence rates indicate the consequences.

In an article following his study of traditional age four-year college students, St. John and Starkey (1995) explored the concept of net price as it applies to student demand for higher education. They concluded that students respond to a set of prices and subsidies rather than a single net price, and that;

This finding leads to the conclusion that there could be differences in the effects prices have on different types of persistence. Specifically, it appears that from this study that the within-year persistence decisions made by students from all income groups are more sensitive to tuition charges than to student aid. (p. 178-179)

Two other studies are of interest here. A study of graduate students (Millett & MacKenzie, 1995) found that minorities were more likely to receive fellowships, but less likely to receive assistantships. In research on the intergenerational effects of paying for college, the results indicated that part of the intrafamily process of deciding on how students should pay for college was the college financing experience of their parents (Flint, 1995).

A few studies have used NPSAS:93 to examine enrollment and persistence. Perna (1996) found that financial aid allowed students to attend more expensive institutions. In a recent study using NCES's Beginning Postsecondary Student Survey, Perna (1997) examined the contribution of financial aid to retention to graduation of full-time, four-year financially dependent students. Perna found that students who received financial aid were more likely to graduate than those that did not. Using path analysis, Perna concluded that the contribution to retention from the receipt of financial was in the middle range of factors in all the models tested. All types of financial aid were positively related to retention, with the exception of loans. In addition, financial aid had an indirect effect, negatively for loans, on grades, and grades were the single most important contributor to retention in this study.

Persistence Theory Development - Accumulated Debt

With the shift from grants to loans starting in 1978, questions were raised about what was an excessive level of debt. Adopting an existing formula for student debtload (Horch, 1978), calculated that the percentage of total consumption expenditures available for repayment ranged from 5.8 percent to 9 percent, depending on the student's budget. From this, Horch estimated that at most, no more than 4 percent of financially dependent and 5 percent of dependent students had excessive debt.

Several articles explored the impact of debt on students at specific institutions. A study of Sweet Briar College alumni (Bodfish, 1989) found that they had paid 56 percent of their loans and that parents paid 70 percent of the student's loan. A study of graduating students at Iowa State University (Holland & Healy, 1989) showed a lack of knowledge about their debt repayment schedule. This confusion would be echoed in studies in the 1990's.

A study of graduates of "elite" colleges and universities found that debt did not inhibit graduate school attendance (Schapiro, O'Malley, & Litten, 1991). However, certain individual variables (gender and ethnicity) and institutional attributes, such as women's colleges and universities, had significant effects on the rate of progression to graduate school.

The New England Student Loan Survey II explored how student debtload influenced post-graduate decisions (Pedalino, Chopick, Saunders, & McHugh, 1992). A total of 2,200 borrowers were investigated. Seventy-five percent had not changed career plans because of their loan, although those who had borrowed over $20,000 were more likely to have reported a career change as a result of their debt. Over 70 percent felt that educational borrowing was worth the burden of repayment, although the 30 percent who disagreed were more likely to have larger debtload and monthly payments.

A study of 1,000 graduating students who had borrowed through the GSL program at Iowa State University examined the influence of background variables (excluding income) on debt volume (Hira, 1992). Age, major, residency status, housing location, employment status, GPA, and knowledge of loans were all positively correlated with the amount of loans. The researchers suggested that length of time in school (age), higher non-resident tuition rates, living in more expensive off-campus housing, having a higher financial need (and therefore working) could explain much of the correlation between debtload and demographic variables. They also concluded that students with a greater knowledge of student loan programs were more likely to have borrowed more.

St. John (1994a) used the High School and Beyond '80 cohort to examine the influence of debt on choice of major. He found that in the mid-Eighties, the choice of major was influenced by social background, high school achievement, high school major choice, and college experience, and that debt burden was not significantly associated with choice of major. He concluded;

Decisions to majors with high expected earnings are appropriately viewed as an integral part of the attainment process. Students who come from families with high incomes, achieve well in high school, and persist and achieve well in college are more likely to choose majors with high expected earnings. ( p. 10)

Another study (Topper, 1994) also concluded that the shifts in choice of major (away from the liberal arts) were due to more older, part-time, and financially independent students than to debtload.

In 1996, there was a surge of publications on student debtload, triggered in part by concern over the higher debtload in the 1992 Reauthorization of HEA. One study (Greiner, 1996) had an alarming finding: 26 percent of the students surveyed (Iowa student loan recipients) had unmanageable debt. This compared to 6.5 percent in 1986 (Decision Resources Corporation & Westat, 1992) and 8.3 percent in 1990 (Westat, 1993). Both of these studies used national samples.

In a provocative article, an anonymous author (Keynes, 1995) argued that increasing costs and decreasing aid have caused students to assume more debt. While Keynes believed that the situation was still manageable, the author made recommendations on monitoring debt and improving institutional practices. Cofer and Somers (1997) found, using NPSAS:93, that accumulated debtload was negatively associated with within-year persistence of all four-year, undergraduate students; and in a separate study, Cofer & Somers (1998) found a negative association with accumulated debt and within-year persistence for both public and private four-year undergraduate students. In addition, they found that private school undergraduates were more sensitive, to price, subsidy, and debtload variables than the public school undergraduates

Two qualitative studies examined how undergraduate (Somers, 1997) and graduate (Somers, Cofer, DeAngelis, & Cook, 1997) student decisions in college choice, major, career, personal decisions, and consumer decisions were influenced by debtload. They found the students tremendously angry. Students were bitter that they did not have the same low-interest and forgivable loans as their parents, the early baby boomers. Students also blasted their institutions for "bait and switch" financial aid tactics, paying less than a working wage for graduate assistantships, and not paying for graduate student tuition waivers. Their greatest wrath, however, was reserved for Congress, which the students described as out of touch, feathering their own retirement nest, and completely ignoring investment in higher education. The phrase "term limits" was raised frequently in these interviews.

Conclusion

The research from theses studies suggests that the appropriate model for the study of student persistence includes a number of factors. Demographic factors such as gender, ethnicity, socioeconomic status, and pre-college performance have been shown by prior research (Astin, 1964; Cabrera, 1988; Cabrera, Nora & Castenada, 1992; Pascarella & Chapman, 1983; Nettles, Thoeny, & Gosner, 1991; Nora & Cabrera, 1993; Pascarella & Terenzini, 1977, 1979,1980, 1983; Pascarella & Chapman, 1983) to have an influence on student persistence. Tinto's (1975, 1987, 1993) studies on the theory of voluntary departure indicate that activities that facilitate a student's academic and social integration into the institution increase the likelihood of persistence. Bean (1980, 1982) further modified Tinto's model with the concept of "intent to persist" through exploration of goal and institutional commitment. Cabrera, Nora and Castenada (1992) modified Tinto's model to examine the effect of financial variables.

Prior research (Astin, 1975; Bean, 1980; Cabrera, Nora & Castenada, 1992; Nora & Cabrera, 1993; St John, 1990b, 1992) has shown that educational aspirations, institutional characteristics, and institutional control have shown both direct and indirect effects on student persistence. Persistence is generally found to be grater at four-year rather than two-year institutions ( Pascarella & Chapman, 1983; Pascarella & Terenzini, 1991; St. John, 1990) and generally higher at private rather than public institutions (Pascarella & Terenzini, 1991).

While some researchers, using institutional data (Murdock, 1987; Pascarella & Terenzini, 1991; Stampen & Cabrera, 1988; Stampen & Fenske, 1988) have shown that recipients and non-recipients of financial aid persist at comparable rates, others (Somers, 1992, 1993, 1994, 1995, 1996) found a positive effect on persistence from the receipt of financial aid. St. John (1990a), using a national sample of high school seniors found that persistence increased with financial aid, and using early versions of national databases (Terkla, 1985; Cabrera, Nora, & Castenada, 1992) concluded similarly. The work of St. John and associates and Somers and associates previously cited demonstrates the impact, both positively and negatively, of price and subsidy variables on the student persistence process.

CHAPTER THREE

Methods and Procedures

Introduction

The purposes of this study were to 1) examine the effect of debtload by income strata on the within-year persistence/withdrawal decisions of four-year college students using a previously validated model of student persistence using the National Postsecondary Student Aid Survey data (St. John & Starkey, 1994; Cofer & Somers, 1997); 2) determine if the effect of debtload on within-year persistence/withdrawal decisions has changed over the time period 1987 to 1996; and 3) attempt to determine how federal financial aid policy changes in the last decade have affected student persistence, and if those effects differ by income strata. This chapter describes the methods that were used.

Method

This study uses the National Postsecondary Student Aid Survey of 1995-96 (NPSAS:96) and 1986-87 (NPSAS:87) to explore how debtload and other variables affect within-year persistence at public and private colleges in undergraduate programs, and to study how the effect of debtload has changed over that period. Since the decision to persist is a dichotomous outcome (either a student persists or withdraws) the data from the samples drawn from the two surveys was analyzed using logistic regression. The following sections describe the sources of data for both of the surveys (NPSAS:87 and NPSS:96), the sample for this study, the model specifications and the statistical techniques used in the study. Special coding techniques necessary to analyze data using a logistic regression technique are also explained in the concluding section of the Method section.

Sources of Data

This study used data collected by the National Postsecondary Student Aid Survey of 1995-96 (NPSAS:96) and 1986-87 (NPSAS:87) to explore how debtload and other variables affect within-year persistence at four-year public and private colleges in undergraduate programs, and to study how the effect of debtload has changed over that period. NPSAS is a "comprehensive nationwide study to determine how students and their families pay for postsecondary education," (U. S. Department of Education, 1996). The original survey was designed and implemented by the National Center for Educational Statistics (NCES) of the Department of Education in 1986-87 with subsequent surveys conducted triennially in 1989-90, 1992-93, and 1995-96. This study used the restricted-use NPSAS:96 and the non-restricted NPSAS:87 database. The raw data were drawn from public use in the 1990 cycle due to privacy problems identified by the Department of Education (DOE). The restricted-use NPSAS:96 and the public-use NPSAS:87 databases allow the researcher to investigate individual records which are not available in the non-restricted version (Knapp, 1996).

NPSAS:87

NPSAS:87 sampled a total of 59,886 students enrolled as of October 15, 1986, at 1,074 two-year and four-year postsecondary institutions, and follow-up surveys were conducted in the spring of 1987. The resulting sample included 47,000 students including 6,500 graduate, and 4,280 first professional students (U. S. Department of Education, 1987). A geographic-area-clustered, three stage sampling designed was used. The stages were: 1) Zip code constructed geographic regions; 2) institutions within Zip code areas; 3) students within the selected institutions. Students jointly enrolled in high school, and those taking only GED courses were excluded form the sample.

NPSAS:96

The NPSAS:96 survey was a two stage sample design. The geographic-area stage was eliminated form the NPSAS:96 survey for cost reasons. Institutions were stratified into clusters based on institutional control and highest level of offering. A stratified sample of 973 two-year and four-year institutions were selected. NPSAS:96 identified an original sample of 59,593 students eligible for sampling. The response rate of 85.9 percent resulted in a initial survey sample of 51,195 students, including 5,064 graduate students, and 2,186 first-professional students. However, NPSAS:96 included students from Puerto Rico for the first time, and sampled students that were enrolled at any time during the survey period.

Study Sample

The NPSAS:96 and NPSAS:87 databases were adjusted in four phases to arrive at the study sample. The first phase consisted of eliminating all two-year colleges, two-year college students, and graduate students. The second phase eliminated all records that indicated a "missing value" for the total amount borrowed variable. NPSAS:96 differed from NPSAS:87 in two fundamental aspects. First, the 1987 sample did not include students in Puerto Rico; the 1996 sample did. Secondly, the 1987 sample was essentially a fall sample. Students were selected and sampled in the fall of the academic year, and follow-up surveys were conducted in the spring. The 1996 survey contained students that were enrolled in any term beginning between May 1, 1995 and April 30, 1996. Therefore to assure compatibility between the surveys, only fall semester enrollees, and students not from Puerto Rico were selected from the 1996 survey. Finally, to adjust for the over sampling of seniors in both surveys, a random sample of approximately 50 percent of the seniors remaining after the first three phases was taken to arrive at a more uniform distribution by class level. These adjustments yielded a total sample of 21,448 students from NPSAS:87 (10,230 private and 11,218 public college students), and xxxx students form NPSAS:96 (xxxx private and xxxx public college students).

Demographic Characteristics of Sample

The descriptive statistics for NPSAS:87 and preliminary estimates for NPSAS:96 are shown in Appendix B2. The dispersion by income level showed a marked difference between the two surveys. How much of that difference was attributable to the change in definition of a "dependent student" in federal government aid calculations is not known. In the background statistics, it is interesting to note the reduction in the number of mothers who had a college degree. However, a review of the data indicated that a greater number of mothers had advanced degrees in 1996 than 1987.

Between the two surveys, private college student bodies became more diverse in terms of ethnicity, gender, and age; the change for public institutions was smaller. Income distribution and dependency were substantially different between the two surveys, again owing to the definition change by the federal government. In both the private and the public school statistics, a smaller percentage of students with disabilities were represented in the 1996 data.

Academic achievement appears to have fluctuated between 1987 and 1996. More students had a grade point average (GPA) higher, and more students had a GPA lower in 1996 than they did in 1987. This could be partially attributable to the increase in diversity among students. Fewer students at both private and public institutions were taking remediation, and fewer students were attending doctoral institutions in the later survey. Concomitant with the increase in age in the 1996 survey, more students were working full-time at both types of institutions.

Prices showed the most dramatic change of any of the variables in both models. Private school tuition increased 87 percent, and public school tuition increased 128 percent between 1987 and 1996. Financial aid and accumulated debt variables all increased between the two surveys, but not at the rate of increase in tuition. In conjunction with the absolute level of accumulated debt rising, more students borrowed, and more students had higher levels of debt. Those students with high debt (more than $7,000) increased 43 percent and 57 percent respectively for private and public school students. In addition, the percentage of students who had no accumulated debt decreased between 1987 and 1996 for both private and public schools. The highest percentage of students with debt were the private school students, of whom 59.5 percent had accumulated debt, compared to 48.2 percent of public school students.

Model Specifications

The model used in this study (see Appendix B1) drew on the previous NPSAS research (Andrieu, 1990; Andrieu, 1991; Andrieu & St. John, 1993; St. John, 1991a; St. John, 1992; St. John, 1994b; St. John & Andrieu, 1995; St. John, Oescher, & Andrieu., 1992; St. John & Starkey, 1995; Starkey, 1993; Trammell, 1994; Cofer & Somers, 1997, 1998 ). The study focused exclusively on within-year progression of students from the fall to the spring semester. In addition to including the variables consistent with prior studies, the amount of debt was added to the model as three separate thresholds of debt, consistent with prior research including debt in persistence models (Cofer & Somers, 1997, 1998; DeAngelis, 1997).

Variables were carefully selected from NPSAS:87 and NPSAS:96 for consistency of the models. In some cases, variables that one would have liked to have included in the analysis were not present in both datasets. In those cases, variables were excluded from both models.

Fifteen dichotomous background variables were included in the model: three ethnicity variables (African American, Hispanics, and "Other" were compared to White students), gender (males were compared to females), two age variables (under 22 and over 30 were compared to students 22-30), two income categories (under $11,000 and over $60,000) were compared to students whose income fell between $11,000 and $60,000. The income categories represented the parent's income if the student was dependent and the student's income if the student was independent. In addition, dependency status, whether the student was married, and two variables dealing with the educational attainment of the student's parents (father had a college degree and mother had a college degree), having a disability, and two variables dealing with high school completion (no high school degree and GED/certificate holders) were compared to students holding a high school degree, were included in the model.

Three variables related to student aspirations and achievement were examined: aspiration for a college degree, aspiration for an advanced degree, and aspiration for completion of some college were compared to those students not aspiring to a college degree. Normally, achievement scores would be included in the aspiration and achievement variables, however, no achievement scores were collected for NPSAS:87. Therefore, achievement scores were eliminated from the models.

Eleven college experience and institutional characteristic variables were included in the analysis: three class academic variables (juniors, sophomores, and seniors were compared to first-year students), Whether a student lived on campus, received any hours of remediation, attended school full-time, or worked full-time were included as college experience variables. One institutional characteristic variable was included in the analysis: doctoral institution compared to non-doctoral granting institutions.

Careful analysis of the data in both datasets revealed that a large number of cases were missing the GPA variable. To adjust, a method developed by Somers (92), which used a series of dummy variables was used. Three grade point variables (High GPA, Low GPA, and No GPA) were compared to GPA between High and Low. High GPA represented those students having a GPA of 3.5 or higher, and Low GPA represented those students having a GPA of 2.0 or less. The GPA score cutoffs were determined by examining a frequency distribution of the scores in the NPSAS:96 data set, and dividing the scores into thirds. For consistency among the models, the same cutoff scores were used for the NPSAS:87 and the NPSAS:96 models.

Four price and financial aid variables were included in the model: full-time, full-year tuition, current year grants/scholarships, current year loans, and current year work-study. Each of the price and financial aid variables were continuous and were divided by 1,000 for compatibility with prior studies.

The total amount borrowed for education was divided by 1,000 to arrive at debtload statistics, and this same variable was separated into high debt, medium debt, and low debt. As with other categorical variables in the analysis, a frequency distribution of the NPSAS:96 and NPSAS:87 variable was examined, and divided into thirds. High debt was represented by total amount borrowed for education over $7000, low debt represents total amount borrowed less than $3000, and medium debt were those amounts between $3000 and $7000. Students with these threshold amounts were compared to students with no debt.

Statistical Method

To describe the relationship between an outcome (dependent) variable and one or more explanatory (independent) variables, statistical regression methods are used. Regression techniques are used to find the "best fit" between the explanatory variables and the outcome variable. In a linear regression model, two assumptions are important. The first is that the variables are continuous The second is that the relationship between an outcome variable and independent variables is expressed by a straight line. Both of these assumptions are violated when the outcome is dichotomous (Cabrera, 1994).

For a model where the outcome variable is dichotomous (such as this study), the standard regression formula (Ordinary Least Squared method) can seriously mis-estimate the dependent variable. For a model where the outcome variable is dichotomous, a technique known as logistic regression is used. Since a student chooses to persist or not, the outcomes are dichotomous: either yes or no (coded as 1 or 0). The resulting graph of the relationship is not a straight line, but a curved line bounded by 0 and 1.

The basic logistic regression equation is:

exp(₀ + ₁X₁ + ₂X₂ + ... + _nX_n)

P = E(Y|X) =

1+ exp(₀ + ₁X₁ + ₂X₂ + ... + _nX_n)

Regardless of the values of the constants _i or the variables X_i, this equation still results in values between zero and one, because of the properties of the natural logarithm. The value P can also be thought of as a probability measure that the outcome variable will be 1 (yes). This is precisely what a dichotomous model requires. Logistic regression uses maximum likelihood techniques to maximize the value of the function P = E(Y|X). A log-likelihood function indicates how likely it is to obtain observed values of Y given the values of the independent variables. Logistic regression solutions are found indirectly through an iterative process of comparing two log-likelihood functions, beginning with a tentative solution, and repeating the process with small changes to see if the solution can be improved. The process is halted when the next step in the iteration provides a negligible improvement in the solution (Menard,1995).

Coding Variables for Logistic Regression

Logistic regression generally functions best when the dependent variables are either dichotomous or have relatively few data points (Aldrich & Nelson, 1984). This is because a large range of data values can produce either very large or very small numbers in the exponential log function step of the regression. Thus, numeric variables like grade point average, that have a large range and many values for any given sample, may prohibit the use of logistic regression. In these cases, the preferred method for coding is to use dichotomous variables. For example, a student was coded as either a male (1) or a female (0), or was married (1) or single (0).

A strength of logistic regression is that the model may contain many variables, some of which may be on different measurement scales (Hosmer & Lemeshow, 1989). If some variables are nominal, it is inappropriate to code them as if they were interval scaled. This requires the use of design or dummy variables. For example, if the data for ethnicity include four choices (African American, Caucasian, Hispanic, and Other), three design variables would be created. They are African American, Hispanic, and Other (all coded as 0 = no and 1 = yes). All subjects would be coded as either "yes" on one of the variables or "no" on all of the variables (for the "Caucasian" group). In general, if a nominal variable has k possible values, then k-1 design variables will be needed.

Model Statistics

In a linear regression, the evaluation of the model is based on the sum of the squared errors of prediction. A smaller sum of the squared errors represents a better fit to the actual data of the regression line. The multivariate F test is used to determine if the predicted value for Y could be attributable to random sampling variation. In addition, the coefficient of determination, or R², is the proportion of the variance explained by the model (Menard, 1995). In logistic regression, with two finite values for the dependent variable, the evaluation of the model is based not on the amount of variance explained between the predicted and the actual values, but on whether the prediction is correct or incorrect. The log-likelihood function in logistic regression is analogous to the sum of the squares in linear regression, it is the "criterion for selecting parameters." The smaller the log-likelihood value the better the fit of the model. As a test of goodness of fit of the model the -2 log-likelihood function will be used. When -2 is multiplied by the log-likelihood function, the resulting value approximates the ² (chi-square) distribution (Menard, 1995).

A number of statistics analogous to R² have been suggested for logistic regression (Agresti, 1990; Aldrich & Nelson, 1984; Hosmer & Limeshow, 1989; Knoke & Burke, 1980).

Aldrich and Nelson (1984) proposed a "pseudo-R^2" with the following general form;

pseudo-R² = c / (n+c), where c is the chi-square statistic for the overall fit of the model, and N is the total sample size. (P. 57)

"The measure is equivalent to the squared contingency coefficient used in calculating associations for nominal variables" (Menard, 1995, p. 22). The pseudo-R² statistic is equal to zero when the independent variables are unrelated to the dependent variable, and approaches 1 when there is a perfect relationship (Menard, 1995). Care should be used when analyzing pseudo-R² since its value can never reach 1.

The beta coefficients in the logistic regression solution are converted to a delta-p statistic using a method recommended by Petersen (Peterson, 1984).

exp(L₁) exp(L₀)

Delta P = -

1+exp (L₁) 1+exp(L₀)

Where L₀ is the logit score before the change in the ith variable, and L₁ = L₀ + _n, is the logit score after the unit change in X_n.

The delta-p measures change in the dependent variable. For dichotomous variables, the delta-p provides a measure of the extent to which the outcome was likely to change if a student had the specified characteristic. For example, a delta-p of 0.050 for females is interpreted as increasing the probability of enrollment by 5.0 percentage points for this group. With continuous variables, the delta-p is interpreted as meaning that a change in a unit measure will change the probability of the outcome by a certain percentage. For example, a delta-p statistic of .0450 per $1000 of financial aid indicates that the probability of attendance or persistence increases by 4.5 percent per $1000 of financial aid awarded.

Time line

The study should be completed by the end of October, 1998. NPSAS:87 variables were previously analyzed and the sample drawn from this survey. Preliminary logistic regression analysis has also been performed on the NPSAS:87 sample. The NPSAS:96 database has been completed by NCES, and I am currently waiting distribution. NCES has advised that the database will be distributed in July. Assuming a July distribution date, the NPSAS:96 sample will be draw the beginning of August. Comparison of the variables for NPSAS:87 and NPSAS:96 and logistic regression analysis on NPSAS:96 will be performed in early August. Comparisons of the output and tabular displays will be completed in late August. The data will be analyzed and findings and conclusions completed by the end of September. Final presentation of the study should be completed by mid to late October.