This is the accessible text file for GAO report number GAO-11-4 
entitled 'For-Profit Schools: Large Schools and Schools that 
Specialize in Healthcare Are More Likely to Rely Heavily on Federal 
Student Aid' which was released on October 4, 2010. 

This text file was formatted by the U.S. Government Accountability 
Office (GAO) to be accessible to users with visual impairments, as 
part of a longer term project to improve GAO products' accessibility. 
Every attempt has been made to maintain the structural and data 
integrity of the original printed product. Accessibility features, 
such as text descriptions of tables, consecutively numbered footnotes 
placed at the end of the file, and the text of agency comment letters, 
are provided but may not exactly duplicate the presentation or format 
of the printed version. The portable document format (PDF) file is an 
exact electronic replica of the printed version. We welcome your 
feedback. Please E-mail your comments regarding the contents or 
accessibility features of this document to Webmaster@gao.gov. 

This is a work of the U.S. government and is not subject to copyright 
protection in the United States. It may be reproduced and distributed 
in its entirety without further permission from GAO. Because this work 
may contain copyrighted images or other material, permission from the 
copyright holder may be necessary if you wish to reproduce this 
material separately. 

Report to Congressional Committees: 

United States Government Accountability Office:
GAO: 

October 2010: 

For-Profit Schools: 

Large Schools and Schools that Specialize in Healthcare Are More 
Likely to Rely Heavily on Federal Student Aid: 

GAO-11-4: 

Contents: 

Letter: 

Appendix I: Briefing Slides: 

Appendix II: Scope and Methodology: 

Appendix III: Tables Comparing Schools with Different Characteristics: 

Appendix IV: Technical Appendix on Results from Regression Model: 

Appendix V: GAO Contact and Staff Acknowledgments: 

Tables: 

Table 1: 90/10-Rate Categories: 

Table 2: Percent of Schools, Students Attending, and Federal Student 
Aid Dollars Earned by For-Profit Schools with Specific Characteristics: 

Table 3: Percent of For-Profit Schools with Specific Characteristics 
by Ownership Status: 

Table 4: Odds and Odds Ratios for Very High 90/10 Rates by School 
Enrollment: 

Figures: 

Figure 1: Distribution of For-Profit Schools by 90/10 Rate Category, 
2003-2008: 

Figure 2: Distribution of 2008 90/10 Rates: 

Abbreviations: 

FTE: full-time equivalent: 

IPEDS: Integrated Postsecondary Education Data System: 

OIG: Office of Inspector General: 

OPEID: Office of Postsecondary Education Identification Number: 

PEPS: Postsecondary Education Participants System: 

[End of section] 

United States Government Accountability Office:
Washington, DC 20548: 

October 4, 2010: 

The Honorable Tom Harkin:
Chairman:
The Honorable Michael B. Enzi:
Ranking Member:
Committee on Health, Education,
Labor, and Pensions:
United States Senate: 

The Honorable George Miller:
Chairman:
The Honorable John P. Kline:
Ranking Member:
Committee on Education and Labor:
House of Representatives: 

In the 2008-2009 school year, about 2,000 for-profit schools received 
almost $24 billion in grants and loans provided to students under 
federal student aid programs.[Footnote 1] In the early 1990s, Congress 
was concerned that some for-profit schools receiving federal student 
aid were recruiting students who were not ready for higher education. 
Many of these students left school with no new job skills and few 
employment prospects in their fields of study and many defaulted on 
their federal student loans. 

In response, Congress enacted the 85/15 rule in 1992, which required 
for-profit schools to obtain at least 15 percent of their revenues 
from sources other than federal student aid.[Footnote 2] Proponents of 
the rule believed that for-profit schools offering a quality education 
should be able to earn a minimum percentage of their revenue from 
sources other than federal student aid. In 1998, Congress amended this 
law to create the 90/10 rule, which reduced to 10 percent the 
proportion of revenues schools must obtain from sources other than 
federal student aid.[Footnote 3] These revenues can include cash 
payments from students, private student loans, state educational 
grants, and federal education assistance payments for veterans. For- 
profit schools are required to report in their annual financial 
statements the percentage of their total revenues obtained from 
federal student aid funds.[Footnote 4] This calculation is called a 
school's "90/10 rate" and it is verified each year by an independent 
auditor. Schools that do not comply with the 90/10 rule risk losing 
their eligibility to participate in federal student aid programs. 

As required by the Higher Education Opportunity Act, we are providing 
information on for-profit schools' compliance with the 90/10 rule. 
[Footnote 5] Specifically, this report addresses the following 
questions: 

1. What percentage of for-profit schools is in compliance with the 90/ 
10 rule and to what extent do schools derive their revenues from 
federal student aid funds? 

2. What school characteristics are associated with higher average 
90/10 rates? 

3. What school characteristics are associated with an increased 
likelihood of having a very high 90/10 rate? 

On August 12 and 20, 2010, we briefed cognizant congressional staff on 
the results of this study, and this report formally conveys the 
information provided during these briefings. (See appendix I for the 
briefing slides.) In general, we found that: 

* Between 2003 and 2008, almost 100 percent of for-profit schools 
reported complying with the 90/10 rule. During this period, the 
average percent of revenue received from federal student aid (the 
average 90/10 rate) for all for-profit schools increased slightly from 
62 to 66 percent. 

* In 2008, schools with the following characteristics had 
significantly higher average 90/10 rates than schools without these 
characteristics. Specifically, these schools: 

* Had high proportions of low-income students: 

* Granted degrees no higher than associate's: 

* Specialized in healthcare: 

* Offered distance education: 

* Were large (with 2,000 students or more): 

* Had a publicly traded parent company: 

* Were part of a corporate chain: 

* We found that in 2008, schools that (1) were large, (2) specialized 
in healthcare, or (3) did not grant academic degrees were more likely 
than others to have very high 90/10 rates (above 85 percent), when 
controlling for the effects of other characteristics. Large schools 
and schools that specialized in healthcare both had higher average 
90/10 rates and were much more likely to have very high 90/10 rates 
than other schools. 

We used the following methodology to develop our findings. To 
understand the 90/10 rule and factors that influence schools' 90/10 
rates, we reviewed relevant federal laws, regulations, and program 
guidance. We also interviewed officials from the Department of 
Education (Education), Education's Office of Inspector General (OIG), 
six for-profit schools that represent both large and small schools 
with a mix of ownership types and academic programs, seven independent 
auditors with varying levels of experience in verifying 90/10 rates at 
for-profit schools, and associations focusing on higher education. To 
determine the percentage of schools in compliance with the 90/10 rule, 
we analyzed 90/10 rates from Education's eZ-Audit system for fiscal 
years 2003 through 2008 and augmented this information with data from 
Education and Education's OIG.[Footnote 6] Finally, to learn about 
what school characteristics are associated with higher average 90/10 
rates and with an increased likelihood of having a 90/10 rate above 85 
percent, we conducted descriptive statistical analyses and a 
multivariate regression analysis using 90/10 rates for fiscal year 
2008 reported in eZ-Audit and school year 2008-2009 school 
characteristic data reported in Education's Integrated Postsecondary 
Education Data System (IPEDS). For the purposes of our analyses, we 
considered a for-profit school to be an entity with a unique 
identification number assigned by Education (known as an OPEID) 
because Education requires each for-profit school with an OPEID to 
submit an annual 90/10 rate, and monitors compliance with the 90/10 
rule on an OPEID basis. However, depending on how schools are 
organized, an OPEID may correspond with one or multiple campuses. For 
instance, in one case, five campuses may be organized under one OPEID, 
and in another, five related campuses may all have their own OPEIDs. 
In the first scenario, the five campuses would be part of the same 
school and would collectively report one 90/10 rate, while in the 
second scenario, each campus would report its own 90/10 rate and count 
as a separate school in our analysis. We conducted statistical testing 
and found that this difference did not materially affect the results 
of our regression analysis. We assessed the reliability of each 
Education dataset we used by interviewing agency officials 
knowledgeable about the data, reviewing relevant documentation, and 
conducting additional analyses. We determined that the data from each 
dataset were sufficiently reliable for the purposes of this report. 
For additional information on our scope and methodology, see 
appendixes II, III, and IV. 

We conducted our work from October 2009 to October 2010 in accordance 
with generally accepted government auditing standards. Those standards 
require that we plan and perform the audit to obtain sufficient, 
appropriate evidence to provide a reasonable basis for our findings 
based on our audit objectives. We believe that the evidence obtained 
provides a reasonable basis for our findings. 

We provided a draft copy of this report to Education for review and 
comment. Education did not have any comments on the report. 

We are sending copies of this report to relevant congressional 
committees, the Secretary of Education, and other interested parties. 
In addition, this report will be available at no charge on GAO's Web 
site at [hyperlink, http://www.gao.gov]. 

If you or your staffs have any questions about this report, please 
contact me at (202) 512-7215 or scottg@gao.gov. Contact points for our 
Offices of Congressional Relations and Public Affairs may be found on 
the last page of this report. GAO staff who made key contributions to 
this report are listed in appendix V. 

Signed by: 

George A. Scott: 
Director, Education, Workforce, and Income Security Issues: 

[End of section] 

Appendix I: Briefing Slides: 

For-Profit Schools: Large Schools and Schools that Specialize in 
Healthcare Are More Likely to Rely Heavily on Federal Student Aid: 

Briefing to Congressional Committee Staff: 

Education and Labor: 
House of Representatives: 

Health, Education, Labor, and Pensions: 
United States Senate: 

Overview: 
* Introduction: 
* Research Objectives: 
* Scope and Methodology: 
* Summary of Findings: 
* Background: 
* Findings: 

Introduction: 

For-Profit Schools Receive Billions of Dollars in Federal Student Aid, 
but Face a Limit on How Much They Can Receive: 

During the 2008-09 school year, about 2,000 for-profit schools 
received almost $24 billion in federal student aid.[Footnote 7] 

In the early 1990s, Congress was concerned that some for-profit 
schools that received federal student aid funds were recruiting 
students who were not ready for higher education. Many of these 
students: 

* left school with no new job skills and few employment prospects in 
their fields of study, and; 

* defaulted on their federal student loans. 

In 1992, Congress enacted the 85/15 rule, which required for-profit 
schools to obtain at least 15 percent of their revenues from sources 
other than federal student aid.[Footnote 8] 

Proponents of the rule believed that for-profit schools offering a 
quality education should be able to earn a minimum percentage of 
revenue from sources other than federal student aid. 

In 1998, Congress amended this law to create the 90/10 rule, which 
reduced to 10 percent the proportion of revenues schools must obtain 
from sources other than federal student aid. Schools may now receive 
up to 90 percent of revenues from federal student aid.[Footnote 9] 

For-Profit Schools that Violate the 90/10 Rule Risk Losing Eligibility 
to Receive Federal Student Aid Funds: 

For-profit schools are required to report in their annual financial 
statements the percentage of their total revenues obtained from 
federal student aid funds. 

* This calculation is called a school's "90/10 rate." 

* The 90/10 rate is verified each year by an independent auditor 
according to guidelines issued by Education's Office of Inspector 
General (01G). 

* Schools report their 90/10 rate by their Education identification 
number (known as their OPEID). Schools with multiple locations may 
have one or many OPEIDs, depending on how they are organized.[Footnote 
10] 

If a for-profit school does not comply with the 90/10 rule for: 

* one fiscal year: the school's eligibility to participate in the 
federal student aid program becomes provisional for the next two 
years.[Footnote 11] 

* two consecutive fiscal years: the school becomes ineligible to 
participate in the federal student aid program for at least two fiscal 
years. 

Research Objectives: 

1. What percentage of for-profit schools is in compliance with the 
90/10 rule and to what extent do schools derive their revenues from 
federal student aid funds? 

2. What school characteristics are associated with higher average 
90/10 rates? 

3. What school characteristics are associated with an increased 
likelihood of having a very high 90/10 rate? 

Scope and Methodology: 

To address our objectives, we: 

* reviewed relevant federal laws, regulations, and program guidance; 

* analyzed the Department of Education's data for fiscal years 2003 
through 2008 on for-profit schools' reliance on federal student aid; 
[Footnote 12] 

* interviewed officials from Education, Education's Office of 
Inspector General, six for-profit schools representing both large and 
small schools with a mix of ownership types and academic programs, 
seven independent auditors with varying levels of experience in 
verifying 90/10 rates, and associations focusing on higher education; 
and; 

* conducted additional analyses using:[Footnote 13] 

- 90/10 rates for fiscal year 2008 reported in eZ-Audit, and; 

- school year 2008-2009 school characteristic data reported in 
Education's Integrated Postsecondary Education Data System (IPEDS). 

We determined that these data were sufficiently reliable for our 
purposes. 

Using descriptive statistics, we calculated average 90/10 rates of 
schools with different characteristics to assess, for example, whether 
large schools had a significantly different average 90/10 rate than 
very small schools.[Footnote 14] 

Multivariate regression analysis enabled us to identify 
characteristics that indicated an increased likelihood of having a 
very high 90/10 rate (above 85 percent) while controlling for the 
effects of other characteristics. Through this analysis, we could 
determine whether large schools, for instance, were significantly more 
likely to have very high 90/10 rates than very small schools, when 
controlling for other factors.[Footnote 15] 

We conducted our review between October 2009 and October 2010 in
accordance with generally accepted government auditing standards. 

Summary of Findings: 

Large Schools and Schools that Specialize in Healthcare Are More 
Likely to Rely Heavily on Federal Student Aid: 

Between 2003 and 2008, almost 100 percent of for-profit schools 
reported being in compliance with the 90/10 rule. During this period, 
the average percent of revenue received from federal student aid (the 
average 90/10 rate) for all for-profit schools increased slightly from 
62 to 66 percent. 

In 2008, schools with the following characteristics had significantly 
higher average 90/10 rates than schools without these characteristics. 
Specifically, these schools: 

* Had high proportions of low-income students; 

* Were large (with 2,000 students or more): 
* Granted degrees no higher than associate's; 
* Had a publicly-traded parent company; 
* Specialized in healthcare[Footnote 16]; 
* Were part of a corporate chain[Footnote 17]; 
* Offered distance education. 

We found that in 2008, schools that (1) were large, (2) specialized in 
healthcare, or (3) did not grant academic degrees were more likely 
than others to have very high 90/10 rates (above 85 percent), when 
controlling for the effects of other school characteristics. 

Background: For-Profit Schools: Characteristics of For-Profit Schools 
Have Changed in Recent Years: 

Traditionally, for-profit schools: 

* Were owned by local, sole proprietors; and; 

* Offered certificate and associate's degree programs ranging from
cosmetology to medical assistance and business administration. 

More recently, for-profit schools: 

* Range from small, privately-held schools to large, publicly traded 
companies;[Footnote 18] 

* Have expanded their offerings to also include bachelor's, master's, 
and doctoral level programs; and; 

* Provide course offerings through distance education.[Footnote 19] 

Background: 90/10 Calculation: No More than 90 Percent of a For-Profit 
School's Total Revenues May Be Obtained from Federal Student Aid: 

Only revenues received for a school's educational and institutional 
charges, such as tuition, fees, and certain required course materials, 
are included in its 90/10 calculation.[Footnote 20] 

* Other revenues, such as those from vending machines and parking 
lots, are excluded from the calculation.[Footnote 21] 

The 90/10 rate must be calculated using a cash basis of accounting. 
[Footnote 22] 

No more than 90 percent: 

Only revenues from federal student aid programs authorized by Title IV 
of the Higher Education Act are	included, such as: 

* Pell Grants,	
* Stafford Loans, and, 
* Federal Work Study funds. 

At least 10 percent: 

Revenues counted include: 

* cash payments from students, 

* private student loans (including payments for loans made by schools), 

* state educational grants, 

* federal education assistance payments for military personnel and 
veterans, and, 

* federal and state job training grants. 

Background: Enrollment: Enrollment at For-Profit Schools Has
Increased Substantially in Recent Years: 

Between the fall of 2003 and 2008, student enrollment increased from 
about 1 million to about 1.8 million at for-profit schools and from 
about 16 million to about 18 million at all other schools.[Footnote 23] 

Percentage Growth in Enrollment from Fall 2003 to 2008: 
For-profit schools: 83%; 
All other schools: 9%. 

Source: GAO analysis of IPEDS data on fall enrollments. 

Background: Funding: Schools Have More Than Tripled in Recent Years: 

Between the 2002-03 and 2008-09 school years, federal student aid 
revenues increased from about $8 billion to about $24 billion at for-
profit schools and from about $48 billion to about $82 billion at all 
other schools. 

For-profit schools' share of fall 2008 total enrollment was about 9 
percent, while their share of school year 2008-2009 total federal 
student aid revenue was about 23 percent.[Footnote 24] 

Percentage Growth in Federal Student Aid from School Year 2002-03 to 
2008-09: 

For-profit schools: 210%; 
All other schools: 69%. 

Source: GAO analysis of Education's annual federal student aid funding 
report data. 

Finding 1: 90/10 Compliance--Overview: Almost All For-Profit Schools 
Report Being in Compliance with the 90/10 Rule: 

Between 2003 and 2008, almost 100 percent of for-profit schools 
reported being in compliance with the 90/10 rule.[Footnote 25] 

The average percent of revenue for-profit schools received from 
federal student aid (their average 90/10 rate) increased slightly 
between 2003 and 2008, from 62 to 66 percent. 

Finding 1: 90/10 Compliance: Very Few For-Profit Schools Report Not
Complying with the 90/10 Rule: 

Between 2003 and 2008, an average of 99.8 percent of for-profit 
schools reported being in compliance with the 90/10 rule.[Footnote 26] 

During this period, seven schools lost eligibility for federal student 
aid due to non-compliance with the 90/10 rule.[Footnote 27] 

Table: Percent and Number of For-Profit Schools Reporting Non-
Compliance with the 90/10 Rule, 2003-2008: 

Percent out of compliance: 
2003: 0.1%; 
2004: 0.1%; 
2005: 0.3%; 
2006: 0.2%; 
2007: 0%; 
2008: 0.4%; 
2003-2008: 0.2%. 

Number out of compliance: 
2003: 2; 
2004: 1; 
2005: 6; 
2006: 3; 
2007: 0; 
2008: 7; 
2003-2008: 18[Footnote 28]. 

Source: GAO analysis of Education's eZ-Audit data. 

Table Note: We excluded from our 2005 list one school that did not 
have an auditor-verified 2005 90/10 rate and was non-compliant in 
2006, because Education did not calculate its actual 2005 rate. We 
included one school with a rate below 90 percent in our non-compliant 
list because it received more than 90 percent of its revenue from 
federal student aid during the only part of the year in which it was 
eligible for that aid. 

[End of table] 

Finding 1: 90/10 Compliance: For-Profit Schools' Average 90/10 Rates
Increased Slightly Between 2003 and 2008: 

Between 2003 and 2008, the average 90/10 rate of for-profit schools 
increased from 62 to 66 percent.[Footnote 29] 

The proportion of schools with 90/10 rates above 85 percent rose from 
about 10 percent in 2007 to about 15 percent in 2008.[Footnote 30] 

* During the same period, the proportion of students at for-profit 
schools with 90/10 rates above 85 percent rose from about 8 percent to 
about 19 percent. 

For-profit school representatives stated that increases in 90/10 rates 
may be in part due to increased maximum federal student aid award 
amounts, as well as the effects of the recent recession that are 
making private student loans and state student aid grants less 
available. 

Finding 2: Average Rates—Overview: For-Profit Schools with Certain 
Characteristics Had Higher Average 90/10 Rates than Others: 

For-profit schools with the following characteristics had higher 
average 90/10 rates than schools without these characteristics. 
Specifically, these schools:[Footnote 31] 

Student Population: 
Had high proportions of low-income students. 

Educational Focus: 
Granted degrees no higher than the associate's level; 
Specialized in healthcare; 
Offered distance education. 

School Size, Control, and Management: 
Were large (with 2,000 students or more)[Footnote 32]; 
Had a publicly-traded parent company; 
Were part of a corporate chain. 

Finding 2: Average Rates—Student Population: For-Profit Schools with 
High Proportions of Low-Income Students Tended to Have Higher Average
90/10 Rates: 

Officials from for-profit school groups have indicated that having a 
low-income student population makes it more challenging to comply with 
the 90/10 rule, because such students utilize more federal student aid 
than others. 

For-profit schools with high proportions of students receiving Pell 
Grants (which are targeted at low-income students) had higher 90/10 
rates than other schools.[Footnote 33] 

* Schools with at least 75 percent of students receiving Pell Grants 
had an average 90/10 rate of 77 percent vs. 64 percent for all other 
schools. 

Finding 2: Average Rates—Educational Focus: For-Profit Schools 
Offering Associate's Degrees Had a Higher Average 90/10 Rate: 

Schools offering degrees no higher than the associate's level (about 
18 percent of all for-profit schools) had a significantly higher 
average 90/10 rate than other schools (75 vs. 65 percent).[Footnote 34] 

An Education official suggested these schools may have higher 90/10 
rates because their students are less likely to receive employer 
funding for continuing education than students working on a bachelor's 
degree or higher. 

Finding 2: Average Rates—Educational Focus: For-Profit Schools 
Specializing in Healthcare Had a Higher Average 90/10 Rate: 

For-profit schools that specialized in healthcare had a significantly 
higher average 90/10 rate than other schools (75 vs. 63 percent). 
[Footnote 35] These schools represented nearly one-third of all for-
profit schools. 

A higher education association official we spoke with suggested that 
some for-profit schools specializing in healthcare tend to attract 
lower-income students that rely more heavily on federal student aid, 
leading these schools to have higher 90/10 rates. 

* We found that schools specializing in healthcare had higher 
proportions of students receiving Pell Grants (which are targeted at 
low-income students) than other schools, at 60 vs. 52 percent. 

Finding 2: Average Rates—Educational Focus: For-Profit Schools 
Specializing in Cosmetology Had a Lower Average 90/10 Rate: 

For-profit schools that specialized in cosmetology had a significantly 
lower average 90/10 rate than other for-profit schools (62 percent vs. 
71 percent).[Footnote 36] 

* While 45 percent of schools specialized in cosmetology, they only 
enrolled 9 percent of students and received 4 percent of federal 
student aid dollars earned by for-profit schools.[Footnote 37] 

According to Education officials, independent auditors, and school 
representatives we spoke with, cosmetology schools tend to rely less 
on federal student aid than other schools because they earn additional 
revenue from student-run clinics. 

Finding 2: Average Rates—Educational Focus: For-Profit Schools that 
Offered Distance Education Had a Higher Average 90/10 Rate: 

Schools that offered distance education had a significantly higher 
average 90/10 rate than other schools, at 72 percent vs. 66 percent. 

Schools offering distance education accounted for about 12 percent of 
all for-profit schools, but they enrolled 55 percent of students and 
received 70 percent of federal student aid dollars earned by for-
profit schools. 

The majority of these schools were part of a corporate chain and many 
were owned by a publicly-traded company.[Footnote 38] 

Finding 2: Average Rates—School Size and Control: Large Schools Had a 
Higher Average 90/10 Rate than Very Small Schools: 

Large schools (with 2,000 students or more) had a significantly higher 
average 90/10 rate than very small schools (with fewer than 100 
students), at 74 percent vs. 60 percent. 

* Only about 9 percent of all schools were large, but they enrolled 69 
percent of students at for-profit schools. 

* While 30 percent of schools were very small, they only enrolled 2 
percent of students at for-profit schools.[Footnote 39] 

Schools with 500 students or more also had a significantly higher 
average 90/10 rate than schools with under 100 students. 

Finding 2: Average Rates—School Size and Control: Chain Schools Had 
Higher Average 90/10 Rates than Other Schools: 

Schools owned by publicly-traded companies had a significantly higher 
average 90/10 rate than other schools, at 72 vs. 66 percent. 

* These schools represented about 14 percent of all for-profit schools 
in 2008, enrolled 55 percent of students, and received about 65 
percent of federal student aid dollars earned by for-profit schools. 
[Footnote 40] 

Chain schools (about half of all for-profit schools) had a 
significantly higher average 90/10 rate than other schools, at 70 vs. 
63 percent. 

Finding 3: Very High Rates—Overview: A Few School Characteristics Were 
Associated with an Increased Likelihood of Very High 90/10 Rates: 

In 2008, schools that: 

(1) were large (with 2,000 students or more),[Footnote 41] 
(2) specialized in healthcare, or, 
(3) did not grant academic degrees; 

were more likely than others to have very high 90/10 rates (above 85 
percent—close to the 90 percent limit). 

Unlike with our average 90/10 rate analysis, schools with these 
characteristics had increased likelihoods of very high rates after 
controlling for the effects of other factors.[Footnote 42] 

Finding 3: Very High Rates: Large For-Profit Schools Were Much More 
Likely to Have Very High 90/10 Rates (Above 85 Percent) than Very 
Small Schools: 

Large for-profit schools (with 2,000 students or more) were much more 
likely than very small schools (with less than 100 students) to have 
very high 90/10 rates when controlling for other characteristics. 
[Footnote 43] 

* Large schools had more than three times greater odds of very high 
90/10 rates than very small schools. 

* Schools with 500 students or more had at least two times higher odds 
of very high 90/10 rates than schools with less than 100 students. 

Finding 3: Very High Rates: For-Profit Schools Specializing in 
Healthcare and Schools Not Offering Degrees Were More Likely to
Have Very High 90/10 Rates (Above 85 Percent): 

Schools specializing in healthcare were much more likely to have very 
high 90/10 rates than other schools when controlling for other 
characteristics. 

For-profit schools that did not offer academic degrees were 
significantly more likely to have very high 90/10 rates than schools 
offering at least bachelor's degrees when controlling for other 
characteristics.[[Footnote 44] 

Finding 3: Very High Rates: Schools Owned by Publicly-Traded Companies 
Were Less Likely to Have Very High 90/10 Rates (Above 85 Percent): 

Schools owned by publicly-traded companies were significantly less 
likely to have very high 90/10 rates than other schools when 
controlling for the effects of other characteristics.[Footnote 45] 

* However, they actually had higher average 90/10 rates than other 
schools without controlling for other characteristics. 

Publicly-traded schools may have a lower likelihood of very high 90/10 
rates, but a higher average rate than other schools, because they 
manage their rates more carefully to avoid risking non-compliance. 
School officials we spoke with described strategies to manage their 
90/10 rates, such as: 

* expanding programs that attract students with funding from sources 
other than federal student aid, and; 

* encouraging students to contribute cash payments when possible. 

Finding 3: Very High Rates: A School's Tuition Rate Was Not Associated 
with Its Likelihood of Having Very High 90/10 Rates: 

School officials and auditors told us that schools with low tuition
rates tend to have more trouble complying with the 90/10 rule. 

* Federal student aid is more likely to cover the cost of tuition and 
fees at these schools than at schools with high tuition rates. 

However, we did not find any relationship between a school's tuition 
rate and its likelihood of having a very high 90/10 rate.[Footnote 46] 

* Additionally, we found no correlation between a school's tuition 
rate and its average 90/10 rate. 

* In one exception, schools with tuition rates that did not exceed 
2008-2009 Pell Grant and Stafford Loan award limits did have slightly 
higher average 90/10 rates than other schools, at 68 vs. 66 percent. 
[Footnote 47] 

[End of section] 

Appendix II: Scope and Methodology: 

To address our objectives, we used 90/10 compliance data from the 
Department of Education's (Education) eZ-Audit system to describe 
schools' reliance on federal student aid and analyze trends over time. 
We also reviewed relevant federal laws, regulations, and program 
guidance. Lastly, we interviewed officials from Education, Education's 
Office of Inspector General (OIG), and other stakeholder groups. These 
groups included six for-profit schools that represent both large and 
small schools with a mix of ownership types and academic programs, 
seven independent auditors with varying levels of experience in 
verifying 90/10 rates at for-profit schools, and associations focusing 
on higher education. Additionally, we reviewed Education and OIG 
examinations of schools' compliance with the 90/10 rule. Finally, we 
conducted descriptive and multivariate statistical analyses using 
90/10 rates recorded in eZ-Audit and school characteristic data 
reported by schools through Education's Integrated Postsecondary 
Education Data System (IPEDS) to examine what school characteristics 
are associated with higher average 90/10 rates and a greater 
likelihood of having a very high 90/10 rate. We conducted our review 
between October 2009 and October 2010 in accordance with generally 
accepted government auditing standards. Those standards require that 
we plan and perform the audit to obtain sufficient, appropriate 
evidence to provide a reasonable basis for our findings and 
conclusions based on our audit objectives. We believe that the 
evidence obtained provides a reasonable basis for our findings based 
on our audit objectives. 

Data Sources: 

eZ-Audit: 

Since June of 2003, Education has required for-profit schools to 
submit financial audit data, including their reported 90/10 rates, to 
the eZ-Audit data system. Because a complete set of 90/10 rates for 
fiscal year 2009 was not available at the time of our study, we 
limited the scope of our analysis to fiscal years 2003 to 2008. These 
study years included data on 1,554 schools in 2003 and up to 1,962 
schools in 2008. We assessed the reliability of the eZ-Audit data 
elements we used in our study and determined them to be reliable for 
the purpose of describing 90/10 rates and doing statistical analyses 
on these rates. (See below for a detailed account of our assessment of 
90/10 rate reliability.) 

IPEDS: 

The Integrated Postsecondary Education Data System (IPEDS) is a system 
of interrelated surveys conducted annually, and gathers information 
from every college, university, and technical and vocational 
institution that participates in federal student financial aid 
programs. The Higher Education Act of 1965, as amended, requires that 
institutions participating in federal student aid programs report data 
on enrollments, program completions, graduation rates, institutional 
prices, and student financial aid. Through electronic testing, 
discussions of variables with IPEDS officials, and reviewing past GAO 
uses of IPEDS data, we assessed the reliability of IPEDS data elements 
used in our study and concluded that IPEDS data were sufficiently 
reliable for reporting on and analyzing characteristics of schools and 
their possible associations with 90/10 rates. 

Other Data Sources: 

We used a variety of other Education data sources to supplement our 
data analysis. We used data from the Postsecondary Education 
Participants System (PEPS) to obtain data on schools' changes in 
ownership. PEPS is Education's primary system for tracking the 
eligibility status of schools to receive federal student aid funds. 
Based on discussions with Education officials and our verification of 
date information for a random sample of schools that experienced a 
change in ownership, we concluded that these data were reliable for 
identifying changes of ownership that occurred during the period in 
question. We also used data provided by Education on federal student 
aid funding and schools owned by publicly traded companies. After 
discussions with Education officials about how these data are compiled 
and used, we determined that they were sufficiently reliable for 
reporting aggregate federal student aid amounts for different 
categories of schools and for identifying schools owned by publicly 
traded companies. 

Assessing Reliability of 90/10 Rates Reported in eZ-Audit: 

We undertook extensive efforts to assess the reliability of for-profit 
schools' reported 90/10 rates, which are verified by independent 
auditors and then self-reported to Education. In particular, we 
reviewed OIG evaluations of independent audits of for-profit schools 
and spoke with seven independent auditors of varying experience levels 
about the steps they take to verify their clients' 90/10 rates. We 
also interviewed Education officials about their procedures for 
assessing the accuracy of 90/10 rates reported to be above a certain 
90/10 rate threshold and reviewed all OIG audits and Education 
examinations of schools' compliance with the 90/10 rule to determine 
how frequently they found errors in schools' reported 90/10 rates and 
how large these errors tended to be. 

Because of the importance of independent auditors' role in verifying 
schools' 90/10 rates, we reviewed OIG evaluations of their work to 
determine how commonly the OIG had identified weaknesses in their 
verification practices. Despite the fact that the OIG selects 
independent auditors for review using risk-based selection criteria, 
the OIG found errors in the 90/10 rates verified by selected auditors 
only in a small minority of cases. The OIG found the audit 
documentation did not provide sufficient evidence that the auditor 
appropriately verified a school's calculation in a few additional 
cases. 

We also interviewed independent auditors at seven accounting firms 
with different levels of for-profit school experience. We spoke with 
auditors from three experienced firms, one moderately experienced 
firm, and three less experienced firms.[Footnote 48] All of the 
auditors at the firms with extensive for-profit school experience 
described thorough 90/10 verification practices, while auditors with 
moderate and limited experience had more varied practices. It is 
possible that weaknesses in the practices of less experienced auditors 
could result in inadequate 90/10 rate verification and ultimately in 
reporting errors. We chose to group schools into 90/10 rate categories 
for the purposes of our statistical analyses, instead of using 
schools' actual reported 90/10 rates as a continuous measure, to 
minimize the effect of these errors on our data analysis.[Footnote 49] 

We reviewed a combined total of 42 OIG audits and Education program 
reviews and final audit determinations of for-profit schools with 
90/10 related findings. The OIG and Education selected schools for 
review based on a variety of risk factors. In about a quarter of 
cases, OIG and Education found that a school reporting a compliant 
90/10 rate was actually out of compliance. In about half of cases, no 
calculation errors were found, and in two cases, reviewers had 
insufficient information to recalculate the school's 90/10 rate. In 
other cases, errors were identified that either did not change a 
school's compliance status or resulted in an out-of-compliance school 
being reclassified as in compliance. For all reviews in which the OIG 
or Education did recalculate a school's 90/10 rate, the median 
difference between the school's reported and recalculated rates was 
3.3 percentage points. This level of error would be relatively 
unlikely to result in our miscategorizing a school within our 90/10 
rate categories. 

Lastly, we reviewed the outcomes of Education examinations of all 
schools that reported 90/10 rates above a certain threshold. Officials 
we spoke with described a thorough review process, including using 
auditor workpapers to assist them in their verification efforts. 
Education found errors in the reported 90/10 rates of less than one 
quarter of the cases we reviewed. In less than half of these cases, 
Education found that the school's rate was higher than reported, while 
in the rest, it found that the school's rate was lower than reported. 

Revising the eZ-Audit Data: 

After receiving the original eZ-Audit data file from Education, we 
conducted manual and electronic data testing to identify records that 
were duplicated, were missing from the data set, or contained a likely 
inaccurate 90/10 rate based on our analysis of OIG audits and 
Education's reviews. We then worked with Education to resolve these 
issues. To ensure that the data set was sufficiently reliable for the 
purposes of our analyses, we performed the following tasks. 

Because Education officials do not consistently update the eZ-Audit 
data set when they determine that a school reported an incorrect 90/10 
rate, we manually updated all 90/10 rates for which we had evidence of 
an error that had been identified and corrected elsewhere by 
Education. We reviewed all OIG audits and Education reviews that 
examined for-profit schools' compliance with the 90/10 rule and 
updated the 90/10 rates in our eZ-Audit data file for the schools 
where the OIG or Education found a 90/10 rate error and calculated a 
new rate. Through electronic data testing, we also found a large 
number of schools in the data set reporting 90/10 rates of zero. In 
discussions with Education, we determined that many of these records 
with 90/10 rates of zero were for schools that were not eligible for 
or certified to receive federal student aid for the record year. After 
consulting with Education, we created a set of rules to identify and 
remove records of schools that were likely to fall into these groups. 

Through electronic data testing, we also identified a relatively small 
percentage of records that were duplicated due to a data entry error 
on the part of Education. After receiving clarification from Education 
on all records, we removed the duplicated records. We also identified 
cases in which schools had more than one eZ-Audit record in a single 
calendar year. According to Education, schools changing their fiscal 
year end date are required to submit a financial audit for both their 
old and new fiscal years, which results in two eZ-Audit submissions 
for the calendar year in which the change took place. To limit our 
data to one record per school per year, we retained only the record 
with the fiscal year end date that matched the date in the subsequent 
year.[Footnote 50] 

Unit of Analysis: 

Each for-profit school that disburses federal student aid must report 
a 90/10 rate to Education annually.[Footnote 51] For the purposes of 
the 90/10 requirement and our analysis, a school is an entity that 
corresponds to a unique six-digit Office of Post-Secondary Education 
Identification (OPEID) number. Because schools with multiple locations 
may have one or many OPEIDs, depending on how they are organized, a 
90/10 rate may correspond to a single campus or many campuses. Due to 
this variation, five campuses with only one OPEID will count as one 
school in our study, whereas five related campuses, each with its own 
OPEID, will count as five schools. 

Creating Our Data Set for 2008 Statistical Analyses: 

To prepare the data for our statistical analyses, we merged an eZ-
Audit data set containing schools' fiscal year 2008 90/10 rates with 
various school year 2008-2009 IPEDS survey files, using the OPEID as 
our merger variable.[Footnote 52] In eZ-Audit, each school with a six-
digit OPEID number reports one 90/10 rate per fiscal year. However, in 
IPEDS, a school with a six-digit OPEID number may choose to report 
more than one set of data by using multiple eight-digit subreporter 
OPEIDs. For instance, a school with five campuses could establish five 
separate eight-digit subreporter OPEIDs and break out information on a 
campus-by-campus basis, reporting data on elements such as total 
enrollment, degrees offered, and average tuition rates separately for 
each campus. 

To merge IPEDS data with eZ-Audit data along the OPEID and create 
variables needed for our statistical analyses, we collapsed data from 
IPEDS records with multiple eight-digit subreporters into a single 
record corresponding to each six-digit OPEID. We made decision rules 
to create collapsed variables for schools with multiple subreporters. 
For instance, to measure a school's total enrollment, we simply summed 
enrollment across all subreporters. To identify the highest level of 
education offered by a school, we used the highest level reported by 
any subreporter. We coded a school as offering distance education if 
at least one of its subreporters offered distance education. To create 
school-wide averages (such as average tuition rates), we took into 
account the number of students corresponding to each subreporter. 

While many variables were straightforward to create, we took extra 
steps when creating a few variables. First, we assigned each school a 
specialty based on which of its programs enrolled or graduated the 
highest number of students, depending on the school's reporting 
method. We defined specialties using the Classification of 
Instructional Program codes, which schools use to report their 
programs' educational focuses to Education. Second, due to variation 
in how schools report their tuition rates, we tested three different 
methods for measuring schools' average tuition rates.[Footnote 53] As 
noted in appendix IV, each method we tested produced similar results 
when included in our regression models. Therefore, in our final 
regression model, we chose to use an academic-year equivalent tuition 
measure for all schools. Lastly, we created school size categories to 
reflect the non-linear relationship between school enrollment and 
90/10 rates. We defined large schools as those with 2,000 students or 
more, mid-sized schools as those with 500 to under 2,000 students, 
small schools as those with 100 to under 500 students, and very small 
schools as those with less than 100 students. 

For 31 schools in the eZ-Audit data set, we could not find 
corresponding IPEDS data. Therefore, these 31 schools are excluded 
from the descriptive and multivariate analyses of school 
characteristics. 

Dividing Schools into Categories Based on 90/10 Rates: 

For the purposes of our study, we divided schools into categories 
based on their relative rates of reliance on federal student aid. 
Grouping schools into categories of reliance on federal student aid 
rather than reporting specific 90/10 rates allowed us to mitigate the 
impact of relatively small errors in schools' 90/10 rate calculations 
on our results. In particular, auditors suggested that schools with 
relatively low reported 90/10 rates might be more error-prone. Due to 
limited risk of unidentified non-compliance resulting from errors in 
low reported 90/10 rates, auditors may not undertake extensive efforts 
to resolve small calculation errors in these rates. By grouping 
schools into categories of reliance, if a school's actual 90/10 rate 
should have been 50 percent, but the school mistakenly calculated its 
rate to be 45 percent, the school would still correctly fall into our 
low 90/10 rate category laid out below. 

For the purposes of analysis, we categorized schools into the groups 
listed in table 1. 

Table 1: 90/10-Rate Categories: 

Category: Low 90/10 rate; 
Percent of revenues from federal student aid: 60% or less. 

Category: Medium 90/10 rate; 
Percent of revenues from federal student aid: >60% to 75%. 

Category: High 90/10 rate; 
Percent of revenues from federal student aid: >75% to 85%. 

Category: Very high 90/10 rate; 
Percent of revenues from federal student aid: >85%. 

[End of table] 

Source: GAO. 

The percentages of for-profit schools with 90/10 rates from 2003 to 
2008 in each of these categories are presented in figure 1. 

Figure 1: Distribution of For-Profit Schools by 90/10 Rate Category, 
2003-2008: 

[Refer to PDF for image: vertical bar graph] 

Year: 2003; 
Low (60% or less): 39.6%; 
Medium (>60% to 75%): 25.8%; 
High (>75% to 85%): 22.6%; 
Very high (more than 85%): 12%. 

Year: 2004; 
Low (60% or less): 37.2%; 
Medium (>60% to 75%): 27.2%; 
High (>75% to 85%): 23.2%; 
Very high (more than 85%): 12.5%. 

Year: 2005; 
Low (60% or less): 35.4%; 
Medium (>60% to 75%): 26.6%; 
High (>75% to 85%): 26.4%; 
Very high (more than 85%): 11.6%. 

Year: 2006; 
Low (60% or less): 37.9%; 
Medium (>60% to 75%): 27.6%; 
High (>75% to 85%): 23.8%; 
Very high (more than 85%): 10.9%. 

Year: 2007; 
Low (60% or less): 37.3%; 
Medium (>60% to 75%): 27.8%; 
High (>75% to 85%): 25%; 
Very high (more than 85%): 9.8%. 

Year: 2008
Low (60% or less): 31.7%; 
Medium (>60% to 75%): 27.6%; 
High (>75% to 85%): 25.8%; 
Very high (more than 85%): 14.9%. 

Source: GAO analysis of 2003 to 2008 eZ-Audit data. 

[End of figure] 

Descriptive Statistics: 

We conducted various descriptive statistical analyses to determine 
whether specific kinds of schools had higher 90/10 rates than others. 
First, to assess the statistical significance of the relationship 
between school characteristics and the 90/10 rate category associated 
with schools possessing these characteristics, we conducted chi-
squared tests of association. For example, the chi-squared test 
assessed the distributions of 90/10 rate categories among schools that 
did and did not offer distance education, and how these distributions 
differed from what would be expected if no relationship existed. For 
dichotomous variables, such as whether a school was owned by a 
publicly traded company, we also conducted t-tests to assess the 
statistical significance of differences in mean 90/10 rates. For 
continuous variables, such as the proportion of students that were 
adult learners over the age of 25, we examined the correlation between 
each variable and the 90/10 rate[Footnote 54]. All differences 
reported in the second findings section of this report are significant 
at the 95 percent level unless otherwise noted. This indicates that 
there is less than a 5 percent probability that we would have gotten 
such a result by chance if there were really no difference in average 
rates or in association between variables. Note that a correlation 
between two variables does not indicate that one "causes" the other; 
rather, it shows that the two appear to be associated with each other, 
without controlling for the effects of other variables. 

Multiple Regression Analysis: 

Multiple regression analysis is a method for exploring how a dependent 
variable is related to a number of independent variables, while 
controlling for other factors that could have an impact on the value 
of the dependent variable. Because the distribution of 90/10 rates was 
not normal, with more than half of schools reporting 90/10 rates above 
70 percent but an extremely low percentage of schools reporting 90/10 
rates above 90 percent, it was not appropriate to use ordinary least 
squares regression modeling to assess the average change in 90/10 
rates associated with specific variables. Instead, we used logistic 
regression analysis, a regression method for assessing dichotomous 
outcomes, to estimate the influence of various predictor variables on 
whether or not a school had a very high 90/10 rate of above 85 
percent.[Footnote 55] We tested multiple model specifications and 
definitions of "very high 90/10 rates" to ensure that our estimates 
were stable across different models, and used robust standard errors 
to control for the potential effect of clustering of schools within 
the same corporate entity on our variance estimates. See appendix IV 
for a detailed discussion of our modeling procedures and results. 

Limitations of the Analysis: 

One limitation of our study relates to inconsistency in our unit of 
analysis, which is due to the variation in how schools are organized 
and report their 90/10 rates. As described earlier, in one case, five 
campuses may be organized under one OPEID, and in another case, five 
related campuses may each have their own OPEIDs. In the first 
scenario, the five campuses would count as one unit of analysis (or 
school) in our study, while in the second case, the five campuses 
would count as five separate units of analysis with five 90/10 rates. 
We explored whether we could resolve this inconsistency by combining 
the 90/10 rates of related OPEIDs, such as the five campuses each with 
its own OPEID, into one unit of analysis with an aggregate 90/10 rate. 
However, we determined that doing so was impossible because eZ-Audit 
does not contain sufficient information to allow us to weight 
accurately each 90/10 rate when creating an aggregate rate. We also 
explored whether we could increase the impact of schools with multiple 
campuses, but only one OPEID in our model; it was not possible, 
however, because such schools do not report data that would allow us 
to identify the 90/10 rates for each of their campuses. This 
limitation reduced the influence of these schools in our model. 
However, to determine whether inconsistencies in our unit of analysis 
substantively affected our results, we ran separate models for related 
OPEIDs and stand-alone OPEIDs and compared their results with those of 
our other models. Our results were consistent across each type of 
model, which showed that the limitations of our unit of analysis did 
not materially affect our results. Consequently, we are confident in 
our treatment of the unit of analysis and in our study's results. 

While we are confident that we have adequately handled the limitations 
in our unit of analysis, care should be taken in interpreting some of 
our results. Particularly, differences in how schools are organized 
and report 90/10 rates may affect their size categories. For example, 
a school (or OPEID) with five campuses and 500 students at each campus 
(for a total of 2,500 students) would be considered large because it 
had at least 2,000 students in total. However, five equally sized 
campuses (of 500 students each) with a common owner, but each with its 
own OPEID, would not be considered large, because no one campus had 
2,000 students or more. Additionally, because there are a large number 
of very small for-profit schools in our universe and a relatively 
small number of large schools, it may be difficult to determine the 
prominence of each type of school in the for-profit industry. For 
example, while schools owned by publicly traded companies only 
accounted for 14 percent of all schools in our universe, they enrolled 
55 percent of students and received 65 percent of federal student aid 
that went to for-profit schools. Appendix III includes a table that 
lists school characteristics we examined, and compares the percent of 
schools with each characteristic to the percent of students those 
schools enrolled and the percent of federal student aid they received. 
This table also lists the average 90/10 rate of schools with each 
characteristic. Appendix III also includes a table listing the 
characteristics we examined and showing the percent of schools owned 
by a publicly traded company and the percent of schools not owned by a 
publicly traded company with each characteristic. 

In addition, we could not fully assess the relationship between 
certain school and student characteristics and a school's likelihood 
of having a very high 90/10 rate in our multivariate regression 
analysis because of data limitations. Specifically, we could not 
include student income in our model because the only proxy variable 
available to us through IPEDS--the percent of students receiving Pell 
Grants--was very closely related to our dependent variable, reliance 
on federal student aid.[Footnote 56] Therefore, we decided to leave 
this measure of student income out of our model. However, because our 
model did include other variables that were correlated with a school's 
percent of students receiving Pell Grants, the relationships found 
between these variables and a school's likelihood of having a very 
high 90/10 rate may be overstated. In particular, a school's percent 
of students receiving Pell Grants and percent of minority students 
were correlated (0.39). Thus, our model may have overstated the effect 
of a school's percent of minority students on its likelihood of having 
a very high 90/10 rate. However, variables uncorrelated with student 
income (as indicated by a school's percent of students receiving Pell 
Grants) should not be affected by our inability to directly or 
indirectly control for student income in our model. 

[End of section] 

Appendix III: Tables Comparing Schools with Different Characteristics: 

The tables below provide various comparisons of characteristics 
examined in our study. Table 2 shows the average 2008 90/10 rate of 
schools with each characteristic. It also lists the percent of for- 
profit schools with each characteristic, and compares that number to 
the percent of students attending, and federal student aid funds 
received by for-profit schools with that characteristic. Table 3 lists 
selected characteristics and compares the percent of schools owned by 
a publicly traded company with that characteristic to the percent of 
schools that were not owned by a publicly traded company and had the 
same characteristic. 

Table 2: Percent of Schools, Students Attending, and Federal Student 
Aid Dollars Earned by For-Profit Schools with Specific Characteristics: 

Student population: 

Characteristic: Schools with a high proportion of students receiving 
Pell Grants (75 percent or more); 
Average 90/10 rate (%): 77; 
95 Percent confidence interval (%): Lower bound: 75.2; 
95 Percent confidence interval (%): Upper bound: 77.9; 
Percent of schools (%): 22; 
Percent of students attending (%): 13; 
Percent of federal student aid dollars (%): 10. 

Characteristic: Schools with less than 75 percent of students 
receiving Pell Grants; 
Average 90/10 rate (%): 64; 
95 Percent confidence interval (%): Lower bound: 62.9; 
95 Percent confidence interval (%): Upper bound: 64.8; 
Percent of schools (%): 78; 
Percent of students attending (%): 87; 
Percent of federal student aid dollars (%): 90. 

Educational focus: 

Characteristic: Schools that did not grant academic degrees; 
Average 90/10 rate (%): 65; 
95 Percent confidence interval (%): Lower bound: 63.8; 
95 Percent confidence interval (%): Upper bound: 65.7; 
Percent of schools (%): 70; 
Percent of students attending (%): 24; 
Percent of federal student aid dollars (%): 15. 

Characteristic: Schools that granted degrees up to the associate's 
level; 
Average 90/10 rate (%): 75; 
95 Percent confidence interval (%): Lower bound: 73.5; 
95 Percent confidence interval (%): Upper bound: 76.3; 
Percent of schools (%): 18; 
Percent of students attending (%): 23; 
Percent of federal student aid dollars (%): 16. 

Characteristic: Schools that granted bachelor's degrees or higher; 
Average 90/10 rate (%): 67; 
95 Percent confidence interval (%): Lower bound: 64.9; 
95 Percent confidence interval (%): Upper bound: 69.2; 
Percent of schools (%): 12; 
Percent of students attending (%): 53; 
Percent of federal student aid dollars (%): 69. 

Characteristic: Schools that offered distance education; 
Average 90/10 rate (%): 72; 
95 Percent confidence interval (%): Lower bound: 69.6; 
95 Percent confidence interval (%): Upper bound: 73.8; 
Percent of schools (%): 12; 
Percent of students attending (%): 55; 
Percent of federal student aid dollars (%): 70. 

Characteristic: Schools that did not offer distance education; 
Average 90/10 rate (%): 66; 
95 Percent confidence interval (%): Lower bound: 65.0; 
95 Percent confidence interval (%): Upper bound: 66.9; 
Percent of schools (%): 88; 
Percent of students attending (%): 45; 
Percent of federal student aid dollars (%): 30. 

Characteristic: Schools that specialized in cosmetology; 
Average 90/10 rate (%): 62; 
95 Percent confidence interval (%): Lower bound: 60.6; 
95 Percent confidence interval (%): Upper bound: 63.1; 
Percent of schools (%): 45; 
Percent of students attending (%): 9; 
Percent of federal student aid dollars (%): 4. 

Characteristic: Schools that specialized in construction, mechanics, 
and manufacturing; 
Average 90/10 rate (%): 71; 
95 Percent confidence interval (%): Lower bound: 67.8; 
95 Percent confidence interval (%): Upper bound: 74.2; 
Percent of schools (%): 5; 
Percent of students attending (%): 6; 
Percent of federal student aid dollars (%): 5. 

Characteristic: Schools that specialized in business; 
Average 90/10 rate (%): 64; 
95 Percent confidence interval (%): Lower bound: 59.3; 
95 Percent confidence interval (%): Upper bound: 69.2; 
Percent of schools (%): 4; 
Percent of students attending (%): 32; 
Percent of federal student aid dollars (%): 41. 

Characteristic: Schools that specialized in healthcare; 
Average 90/10 rate (%): 75; 
95 Percent confidence interval (%): Lower bound: 73.4; 
95 Percent confidence interval (%): Upper bound: 75.9; 
Percent of schools (%): 30; 
Percent of students attending (%): 32; 
Percent of federal student aid dollars (%): 27. 

Characteristic: Schools that specialized in computer technology; 
Average 90/10 rate (%): 68; 
95 Percent confidence interval (%): Lower bound: 65.5; 
95 Percent confidence interval (%): Upper bound: 72.9; 
Percent of schools (%): 2; 
Percent of students attending (%): 3; 
Percent of federal student aid dollars (%): 3. 

Characteristic: Schools that specialized in culinary arts; 
Average 90/10 rate (%): 55; 
95 Percent confidence interval (%): Lower bound: 49.1; 
95 Percent confidence interval (%): Upper bound: 60.2; 
Percent of schools (%): 2; 
Percent of students attending (%): 3; 
Percent of federal student aid dollars (%): 3. 

Characteristic: Schools that specialized in visual and performing arts; 
Average 90/10 rate (%): 55; 
95 Percent confidence interval (%): Lower bound: 49.6; 
95 Percent confidence interval (%): Upper bound: 59.5; 
Percent of schools (%): 3; 
Percent of students attending (%): 5; 
Percent of federal student aid dollars (%): 3. 

School size, control, and management: 

Characteristic: Schools that were large (2,000 FTE students or 
more)[A]; 
Average 90/10 rate (%): 74; 
95 Percent confidence interval (%): Lower bound: 71.1; 
95 Percent confidence interval (%): Upper bound: 76.0; 
Percent of schools (%): 9; 
Percent of students attending (%): 69; 
Percent of federal student aid dollars (%): 71. 

Characteristic: Schools that were mid-sized (500 to 1999 FTE students); 
Average 90/10 rate (%): 75; 
95 Percent confidence interval (%): Lower bound: 73.3; 
95 Percent confidence interval (%): Upper bound: 76.3; 
Percent of schools (%): 20; 
Percent of students attending (%): 20; 
Percent of federal student aid dollars (%): 20. 

Characteristic: Schools that were small (100 to 499 FTE students); 
Average 90/10 rate (%): 66; 
95 Percent confidence interval (%): Lower bound: 64.5; 
95 Percent confidence interval (%): Upper bound: 67.1; 
Percent of schools (%): 42; 
Percent of students attending (%): 10; 
Percent of federal student aid dollars (%): 7. 

Characteristic: Schools that were very small (less than 100 FTE 
students); 
Average 90/10 rate (%): 60; 
95 Percent confidence interval (%): Lower bound: 58.8; 
95 Percent confidence interval (%): Upper bound: 62.0; 
Percent of schools (%): 30; 
Percent of students attending (%): 2; 
Percent of federal student aid dollars (%): 1. 

Characteristic: Schools owned by a publicly traded company; 
Average 90/10 rate (%): 72; 
95 Percent confidence interval (%): Lower bound: 70.0; 
95 Percent confidence interval (%): Upper bound: 73.1; 
Percent of schools (%): 14; 
Percent of students attending (%): 55; 
Percent of federal student aid dollars (%): 65. 

Characteristic: Schools not owned by a publicly traded company; 
Average 90/10 rate (%): 66; 
95 Percent confidence interval (%): Lower bound: 64.8; 
95 Percent confidence interval (%): Upper bound: 66.7; 
Percent of schools (%): 86; 
Percent of students attending (%): 45; 
Percent of federal student aid dollars (%): 35. 

Characteristic: Schools that were part of a corporate chain; 
Average 90/10 rate (%): 70; 
95 Percent confidence interval (%): Lower bound: 68.9; 
95 Percent confidence interval (%): Upper bound: 71.2; 
Percent of schools (%): 48; 
Percent of students attending (%): 79; 
Percent of federal student aid dollars (%): 87. 

Characteristic: Schools that were not part of a corporate chain; 
Average 90/10 rate (%): 63; 
95 Percent confidence interval (%): Lower bound: 62.2; 
95 Percent confidence interval (%): Upper bound: 64.6; 
Percent of schools (%): 52; 
Percent of students attending (%): 21; 
Percent of federal student aid dollars (%): 13. 

Characteristic: Schools purchased by another school in the 3 years 
prior to 2008; 
Average 90/10 rate (%): 69; 
95 Percent confidence interval (%): Lower bound: 66.4; 
95 Percent confidence interval (%): Upper bound: 71.1; 
Percent of schools (%): 9; 
Percent of students attending (%): 11; 
Percent of federal student aid dollars (%): 13. 

Characteristic: Schools not purchased by another school in the 3 years 
prior to 2008; 
Average 90/10 rate (%): 66; 
95 Percent confidence interval (%): Lower bound: 65.5; 
95 Percent confidence interval (%): Upper bound: 67.3; 
Percent of schools (%): 91; 
Percent of students attending (%): 89; 
Percent of federal student aid dollars (%): 87. 

Characteristic: Schools with low tuition rates (at or below 
$10,231)[B]; 
Average 90/10 rate (%): 68; 
95 Percent confidence interval (%): Lower bound: 66.3; 
95 Percent confidence interval (%): Upper bound: 70.2; 
Percent of schools (%): 23; 
Percent of students attending (%): 13; 
Percent of federal student aid dollars (%): 9. 

Characteristic: Schools with tuition rates above $10,231; 
Average 90/10 rate (%): 66; 
95 Percent confidence interval (%): Lower bound: 65.2; 
95 Percent confidence interval (%): Upper bound: 67.0; 
Percent of schools (%): 77; 
Percent of students attending (%): 87; 
Percent of federal student aid dollars (%): 91. 

Characteristic: Schools with a failing composite score[C]; 
Average 90/10 rate (%): 66; 
95 Percent confidence interval (%): Lower bound: 64.0; 
95 Percent confidence interval (%): Upper bound: 68.8; 
Percent of schools (%): 13; 
Percent of students attending (%): 13; 
Percent of federal student aid dollars (%): 14. 

Source: GAO analysis of 2008 eZ-Audit data and 2008-2009 IPEDS data. 

[A] A school's full-time equivalent (FTE) enrollment was determined 
using both undergraduate and graduate enrollment totals for the 2007- 
2008 school year. 

[B] $10,231 is the 2008-2009 Pell Grant and Stafford Loan award limit 
for dependent first-year undergraduates. 

[C] Composite scores are calculated by Education using various 
financial ratios and estimate a school's level of financial 
responsibility. 

[End of table] 

Table 3: Percent of For-Profit Schools with Specific Characteristics 
by Ownership Status: 

Student population: 

Characteristic: Schools with a high proportion of students receiving 
Pell Grants (75 percent or more); 
All schools (%): 22; 
Percent of non-publicly traded schools with characteristic (%): 24; 
Percent of publicly traded schools with characteristic (%): 9. 

Characteristic: Schools with less than 75 percent of students 
receiving Pell Grants; 
All schools (%): 78; 
Percent of non-publicly traded schools with characteristic (%): 76; 
Percent of publicly traded schools with characteristic (%): 91. 

Educational focus: 

Characteristic: Schools that did not grant academic degrees; 
All schools (%): 70; 
Percent of non-publicly traded schools with characteristic (%): 75; 
Percent of publicly traded schools with characteristic (%): 40. 

Characteristic: Schools that granted degrees up to the associate's 
level; 
All schools (%): 18; 
Percent of non-publicly traded schools with characteristic (%): 17; 
Percent of publicly traded schools with characteristic (%): 27. 

Characteristic: Schools that granted bachelor's degrees or higher; 
All schools (%): 12; 
Percent of non-publicly traded schools with characteristic (%): 9; 
Percent of publicly traded schools with characteristic (%): 33. 

Characteristic: Schools that offered distance education; 
All schools (%): 12; 
Percent of non-publicly traded schools with characteristic (%): 9; 
Percent of publicly traded schools with characteristic (%): 33. 

Characteristic: Schools that did not offer distance education; 
All schools (%): 88; 
Percent of non-publicly traded schools with characteristic (%): 91; 
Percent of publicly traded schools with characteristic (%): 67. 

Characteristic: Schools that specialized in cosmetology; 
All schools (%): 45; 
Percent of non-publicly traded schools with characteristic (%): 49; 
Percent of publicly traded schools with characteristic (%): 20. 

Characteristic: Schools that specialized in construction, mechanics, 
and manufacturing; 
All schools (%): 5; 
Percent of non-publicly traded schools with characteristic (%): 5; 
Percent of publicly traded schools with characteristic (%): 7. 

Characteristic: Schools that specialized in business; 
All schools (%): 4; 
Percent of non-publicly traded schools with characteristic (%): 4; 
Percent of publicly traded schools with characteristic (%): 6. 

Characteristic: Schools that specialized in healthcare; 
All schools (%): 30; 
Percent of non-publicly traded schools with characteristic (%): 29; 
Percent of publicly traded schools with characteristic (%): 37. 

Characteristic: Schools that specialized in computer technology; 
All schools (%): 2; 
Percent of non-publicly traded schools with characteristic (%): 2; 
Percent of publicly traded schools with characteristic (%): 7. 

Characteristic: Schools that specialized in culinary arts; 
All schools (%): 2; 
Percent of non-publicly traded schools with characteristic (%): 1; 
Percent of publicly traded schools with characteristic (%): 6. 

Characteristic: Schools that specialized in visual and performing arts; 
All schools (%): 3; 
Percent of non-publicly traded schools with characteristic (%): 3; 
Percent of publicly traded schools with characteristic (%): 5. 

School size, control, and management: 

Characteristic: Schools that were large (2,000 FTE students or more); 
All schools (%): 9; 
Percent of non-publicly traded schools with characteristic (%): 4; 
Percent of publicly traded schools with characteristic (%): 36. 

Characteristic: Schools that were mid-sized (500 to 1999 FTE students); 
All schools (%): 20; 
Percent of non-publicly traded schools with characteristic (%): 17; 
Percent of publicly traded schools with characteristic (%): 36. 

Characteristic: Schools that were small (100 to 499 FTE students); 
All schools (%): 42; 
Percent of non-publicly traded schools with characteristic (%): 45; 
Percent of publicly traded schools with characteristic (%): 23. 

Characteristic: Schools that were very small (less than 100 FTE 
students); 
All schools (%): 30; 
Percent of non-publicly traded schools with characteristic (%): 34; 
Percent of publicly traded schools with characteristic (%): 5. 

Characteristic: Schools that were part of a corporate chain; 
All schools (%): 48; 
Percent of non-publicly traded schools with characteristic (%): 40; 
Percent of publicly traded schools with characteristic (%): 98. 

Characteristic: Schools that were not part of a corporate chain; 
All schools (%): 52; 
Percent of non-publicly traded schools with characteristic (%): 60; 
Percent of publicly traded schools with characteristic (%): 2. 

Characteristic: Schools purchased by another school in the 3 years 
prior to 2008; 
All schools (%): 9; 
Percent of non-publicly traded schools with characteristic (%): 7; 
Percent of publicly traded schools with characteristic (%): 19. 

Characteristic: Schools not purchased by another school in the 3 years 
prior to 2008; 
All schools (%): 91; 
Percent of non-publicly traded schools with characteristic (%): 93; 
Percent of publicly traded schools with characteristic (%): 81. 

Characteristic: Schools with low tuition rates (at or below $10,231); 
All schools (%): 23; 
Percent of non-publicly traded schools with characteristic (%): 24; 
Percent of publicly traded schools with characteristic (%): 15. 

Characteristic: Schools with tuition rates above $10,231; 
All schools (%): 77; 
Percent of non-publicly traded schools with characteristic (%): 76; 
Percent of publicly traded schools with characteristic (%): 85. 

Characteristic: Schools with a failing composite score; 
All schools (%): 13; 
Percent of non-publicly traded schools with characteristic (%): 13; 
Percent of publicly traded schools with characteristic (%): 15. 

Source: GAO analysis of 2008 eZ-Audit data and 2008-2009 IPEDS data. 

Note: The percent of schools in each column represents the percent of 
schools with that ownership status that had the characteristic being 
measured. Values across columns do not add up to 100 percent. 

[End of table] 

[End of section] 

Appendix IV: Technical Appendix on Results from Regression Model: 

To explore how 90/10 rates varied among schools with different 
characteristics, we conducted descriptive analysis. Additionally, we 
conducted a series of multiple regression analyses to address the 
question of which school characteristics are associated with an 
increased likelihood of having a very high 90/10 rate, controlling for 
other factors. 

Model Selection: 

Multiple regression analysis is a method for exploring how a dependent 
or "outcome" variable (such as a 90/10 rate) is related to a number of 
independent or "explanatory" variables, controlling for other factors 
that may have an impact on the value of the dependent variable. The 
primary dependent variable of interest in this report is a school's 
90/10 rate, which summarizes a for-profit school's reliance on federal 
student aid as a proportion of its total revenue. The independent 
variables we tested were related to school characteristics, such as 
size, academic focus, ownership structure, and other factors discussed 
below. 

The distribution of 90/10 rates is not uniform across the range of 
theoretical values from zero to 100. Among for-profit schools with 
corresponding IPEDS data in 2008, fewer than 20 percent of schools had 
90/10 rates in 2008 of 50 points or below, while more than 50 percent 
of schools had rates above 70 up to 94 points, the maximum value 
observed in 2008 (see figure 2). 

Figure 2: Distribution of 2008 90/10 Rates: 

[Refer to PDF for image: vertical bar graph] 

Percentage of For-Profit Schools: 

2008 90/10 Rates: 0-5: 0.73%; 
2008 90/10 Rates: 5-10: 0.36%; 
2008 90/10 Rates: 10-15: 0.93%; 
2008 90/10 Rates: 15-20: 1.19%; 
2008 90/10 Rates: 20-25: 0.57%; 
2008 90/10 Rates: 25-30: 1.55%; 
2008 90/10 Rates: 30-35: 2.07%; 
2008 90/10 Rates: 35-40: 3.06%; 
2008 90/10 Rates: 40-45: 3.73%; 
2008 90/10 Rates: 45-50: 4.4%; 
2008 90/10 Rates: 50-55: 5.39%; 
2008 90/10 Rates: 55-60: 6.89%; 
2008 90/10 Rates: 60-65: 6.99%; 
2008 90/10 Rates: 65-70: 9.79%; 
2008 90/10 Rates: 70-75: 11.13%; 
2008 90/10 Rates: 75-80: 11.76%; 
2008 90/10 Rates: 80-85: 14.4%; 
2008 90/10 Rates: 85-90: 14.76%; 
2008 90/10 Rates: 90-95: 0.31%. 

Source: GAO analysis of Education’s eZ-Audit data. 

Note: Figure excludes schools without corresponding data in IPEDS. 

[End of figure] 

This non-normal distribution renders a common multiple regression 
technique, ordinary least squares analysis, inappropriate as the 
primary methodology for regression analysis.[Footnote 57] Instead, we 
chose to use logistic regression analysis with a dichotomous dependent 
variable that categorizes schools as having a very high 90/10 rate or 
not having a very high rate. We defined schools with "very high" 90/10 
rates as those 15 percent of schools with 90/10 rates that exceeded 85 
percent. Because this definition was not based on a natural break in 
the distribution of 90/10 rates, we tested alternative definitions to 
confirm that the results described in our report are robust to 
alternative definitions of "very high" rates; see below for further 
discussion. 

Logistic regression analysis is a technique used in evaluating 
dichotomous, or binary, outcomes. Unlike ordinary least squares 
regression analysis, which provides an estimate of the absolute change 
in the dependent variable associated with each unit change in an 
independent variable, logistic regression estimates the changes in the 
likelihood of having a specific outcome such as a very high 90/10 
rate. Because the specific probability of having an outcome depends on 
the full set of characteristics associated with a school, we often use 
a transformation of logistic regression estimates, or odds ratios, as 
a means of interpreting the results. Odds ratios compare the odds that 
a school with one particular characteristic has the specific outcome 
(here, a very high 90/10 rate), compared to a school that lacks that 
specific characteristic. 

Table 4 illustrates how to construct odds ratios from raw data for one 
specific characteristic, school size. In our data, 167 of the for- 
profit schools in our data set were large, with 2,000 full-time 
equivalent (FTE) or more students, and 579 were very small with fewer 
than 100 FTE students. Of the large schools, 42 schools had very high 
90/10 rates exceeding 85 points, and 125 did not have very high 90/10 
rates. The odds that a large school had a very high 90/10 rate, or the 
"odds on" a very high 90/10 rate, were thus 42 to 125, or .336. In 
contrast, among very small schools, 46 schools had very high 90/10 
rates, and 533 did not. The odds on very high rates for these schools 
were 46 to 533, or .086. The odds ratio compares the odds on very high 
90/10 rates among schools in the two size categories. The odds ratio 
for large schools having very high 90/10 rates compared to very small 
schools is .336 to .086, or approximately 3.9. The odds ratio shows us 
that the odds of having a very high 90/10 rate were nearly 4 times 
higher for large schools than for very small schools. 

Table 4: Odds and Odds Ratios for Very High 90/10 Rates by School 
Enrollment: 

Characteristic: Large schools (2,000 or more FTE students); 
Very high 90/10 rate: 42; 
Not very high 90/10 rate: 125; 
Odds on very high: 42:125; 
Odds on very high 90/10 rate and derived odds ratio: .336. 

Characteristic: Very small schools (less than 100 FTE students); 
Very high 90/10 rate: 46; 
Not very high 90/10 rate: 533; 
Odds on very high: 46:533; 
Odds on very high 90/10 rate and derived odds ratio: .086. 

Characteristic: Odds ratio, large to very small schools; 
Very high 90/10 rate: [Empty]; 
Not very high 90/10 rate: [Empty]; 
Odds on very high: [Empty]; 
Odds on very high 90/10 rate and derived odds ratio: .336/.086, or 
3.91. 

Source: GAO example using 2008 eZ-Audit data and 2008-2009 IPEDS data. 

[End of table] 

Odds ratios of one indicate that the sets of schools within comparison 
categories had similar odds of having the outcome variable. Logistic 
regression analysis enables us to construct odds ratios for individual 
characteristics that "control for," or hold constant, the effect of 
other independent variables on the outcome. 

Model Specification and Estimation: 

Model specification refers to the choice of independent variables to 
include in a model, as well as methodological decisions such as the 
mathematical form of these variables and type of method used to 
calculate standard errors. Based on research, interviews and previous 
GAO reports, we identified a set of school characteristics to use as 
independent variables in our model predicting the likelihood that a 
school would have a very high 90/10 rate exceeding 85 percent. These 
included: 

* Whether a school offered distance education. 

* A school's total FTE enrollment. 

* Whether a school was bought by another school in the three years 
prior to 2008. 

* Whether a school was owned by a publicly traded company. 

* Whether a school was part of a corporate chain. 

* The school's specialty, as designated by the Classification of 
Industrial Programs code associated with the school's largest program. 

* A school's highest level of educational offering. (We divided 
schools into three categories: those that did not grant academic 
degrees, schools that offered degrees up to the associate's level, and 
those that offered bachelor's degrees or above.) 

* A school's number of eight-digit IPEDS subreporters. (This variable 
measured whether the reported 90/10 rate applied to one or more 
locations or campuses.) 

* A school's proportion of students under/over the age of 25. 

* The proportion of students that were minorities, and the proportion 
of "unknowns," or students that did not self-identify as white or 
minority. 

* The proportion of students that were female. 

* A school's average tuition.[Footnote 58] 

* A school's composite score, a numerical score calculated by 
Education using various financial ratios that estimates a school's 
level of financial responsibility. 

Because of the possibility that schools owned by the same corporate 
entity may have systematic similarities in administration, mission, 
and other factors, we accounted for potential non-independence in 
observations by using robust standard errors that allowed for 
clustering within corporate entities. 

We tested a variety of model specifications before selecting a model 
to use for additional sensitivity testing (see below). For example, 
after initial testing revealed high colinearity between a school's 
highest offering and whether a school was categorized as being a 
degree granting institution, we selected the more detailed variable 
listing the specific highest degree for use in the model. However, 
because relatively few schools reported offering masters' degrees, we 
combined those schools into one category with schools that offered 
bachelor's degrees. 

For continuous variables, such as average tuition, total enrollment, 
or the proportion of female students, we tested both continuous and 
categorical versions. In some models we tested distributionally-based 
categories, such as by quartile, in order to identify relative 
outliers (such as in the top quartile only). In others, we used a 
combination of distributional information and substantively based 
categories. For example, we tested our indicator of school size, total 
enrollment, directly as a logged variable to mitigate the effect of 
large outliers on our estimates. After finding that the logged 
enrollment variable had strong predictive power, we generated unequal 
size categories that adequately captured the relationship between the 
log of enrollment and likelihood of having a very high 90/10 rate, 
which also enabled us to illustrate the effect of size in terms of 
size category rather than a log transformation. 

In addition to ensuring that our results were robust across different 
specifications and definitions of the independent variables, we 
examined a likelihood fit statistic, the Quasi-likelihood Information 
Criteria, to check whether inclusion/exclusion of variables or 
inclusion of different versions of the same variables results in an 
increase or decrease in model fit. For sensitivity testing, we 
selected a final model that performed well in terms of model fit 
relative to other models. 

Results: 

Our logistic regression analyses revealed several statistically 
significant predictors of whether a school had a very high 90/10 rate, 
controlling for other characteristics. As discussed below, these 
results were robust to different definitions of the dependent 
variable, different specifications of the model and of independent 
effects, and different subpopulations. Point estimates for odds ratios 
presented here are illustrative of typical results across a variety of 
models and are significant at the 95 percent confidence level unless 
otherwise indicated. 

* Compared to schools that did not specialize in healthcare, schools 
that specialized in healthcare were more likely to have very high 
90/10 rates (estimated odds ratio approximately 2.2 to 2.9). Aside 
from healthcare, schools that specialized in other areas such as 
cosmetology; business; construction, mechanics, and building trades; 
culinary arts; engineering; law enforcement; communications; visual 
and performing arts; or computers and information technology were not 
significant predictors of having a high 90/10 rate when compared to 
schools without these specific specialties. 

* Compared to schools that offered bachelor's degrees or higher, 
schools that did not grant academic degrees were more likely to have a 
very high 90/10 rate, with estimated odds of having a high 90/10 rate 
approximately twice as high as schools that did not offer degrees. 
This result was consistently significant at the 90 percent level and 
frequently significant at the 95 percent level. 

* The odds of having a very high 90/10 rate increased with school size 
as indicated by total enrollment. In categorical terms, large schools 
with 2,000 students and above had much higher odds of having a very 
high 90/10 rate than very small schools with less than 100 students, 
with estimated odds more than 3 times higher. Similarly, mid-sized 
schools with between 500 and 1,999 students had estimated odds of 
having a very high 90/10 rate more than two times higher than very 
small schools. 

* Publicly traded schools had significantly lower odds of having a 
very high 90/10 rate than non-publicly traded schools. The odds that a 
publicly traded school had a very high 90/10 rate were approximately 
40 percent lower than those for a non-publicly traded school 
(estimated odds ratios of approximately 0.5 to 0.6). 

* Several variables appeared to have a statistically significant 
relationship with a school's 90/10 rate in some models, but were not 
consistently significant or were only significant in some models at 
the 90 percent level. Notably, schools that offered associate's 
degrees as their highest level of educational offering appeared to 
have a somewhat higher 90/10 rate than schools offering bachelor's 
degrees or higher in some models, but this result was not consistently 
significant. Similarly, schools that had been purchased by another 
school appeared to have significantly lower odds of having a very high 
90/10 rate than schools not having been purchased by another school in 
some models, but not consistently across most or all models. In some 
specifications, small schools of 100 to less than 500 students had 
somewhat higher odds of having a very high 90/10 rate than very small 
schools (estimated odds ratio approximately 1.5) at the 90 percent 
confidence level. 

* Several student demographic factors were associated with the 
likelihood of having a very high 90/10 rate. For example, the odds of 
having a very high 90/10 rate also increased with the proportion of 
adult learners over the age of 25 (estimated odds ratio approximately 
1.01).[Footnote 59] The odds of having a very high 90/10 rate 
increased with the proportion of minority students (odds ratio of 
approximately 1.03). The odds of having a very high 90/10 rate also 
increased with the proportion of students failing to self-identify for 
race (estimated odds ratio of 1.02). However, as discussed below, our 
inability to control for a key demographic factor, student income, may 
have an impact on the magnitude of our estimates for factors 
correlated with student income, including student demographic factors. 

* There was no apparent relationship between a school's 90/10 rate and 
the following variables across most or all models tested in our 
regression analyses: having a very high proportion of female students, 
offering distance education, being part of a corporate chain, a 
school's number of OPEID subreporters, or being designated as not 
financially responsible under Education's composite score. 
Additionally, there was no apparent relationship between average 
tuition measured in quartiles and the odds of having a very high 90/10 
rate across any of three different tuition measures, including a 
school's tuition as reported, an annualized measure of tuition, and a 
credit-hour measure of tuition. Further, when we tested the model with 
an intercept for having annualized low tuition below the 2008-2009 
Pell Grant and Stafford Loan maximum award amount of $10,231 for 
dependent first-year undergraduates, we found that for-profit schools 
with low tuition rates did not have significantly higher or lower odds 
of having a very high 90/10 rate than schools with higher tuition 
rates, controlling for other factors.[Footnote 60] 

Sensitivity Testing: 

After selecting a final model, we conducted a series of sensitivity 
tests to ensure the stability and robustness of our results. For 
example, based on concerns that chain schools may vary systematically 
from independent schools, we tested a model among separate populations 
of chain schools versus all other schools. Similarly, we also tested 
our model separately among schools reporting multiple subreporters 
under one OPEID. The results of our models were substantively similar 
across these different population categories. 

To ensure the results of the model were not sensitive to the specific 
cut-point for "very high" 90/10 rates, we conducted sensitivity tests 
with a dependent variable consisting of whether a school's 90/10 rate 
was 75 percent or higher, 82 percent or higher, and 88 percent or 
higher. Again, these models produced results that were consistent with 
models that used a cut-off point of above 85 points as our definition 
of "very high." 

Limitations: 

Despite tests to ensure the stability and robustness of our results 
across different model specifications, alternative definitions of the 
dependent variable, and different subpopulations, the results of our 
model are subject to several limitations. Most critically, we were 
unable to include a control for the proportion of a school's students 
with low incomes because our proxy of student income, the proportion 
of students receiving Pell Grants, factors into the calculation of the 
dependent variable.[Footnote 61] As such, odds ratio estimates for 
other variables correlated with student income, such as the proportion 
of adult learners ages 25 and over or the proportion of students that 
self-identified as minorities, may partly reflect the effect of 
student income on a school's 90/10 rates.[Footnote 62] Additionally, 
these models were not designed to explore outcome variables associated 
with the quality of a school's educational offerings, such as job 
placement rates or student loan defaults, and therefore do not provide 
information about the relationship between very high 90/10 rates and 
these factors. 

[End of section] 

Appendix V: GAO Contact and Staff Acknowledgments: 

GAO Contact: 

George A. Scott, (202) 512-7215 or scottg@gao.gov. 

Staff Acknowledgments: 

The following staff members made key contributions to this report, 
Melissa Emrey-Arras, Assistant Director; Michelle St. Pierre, Analyst- 
in-Charge; Ellen Phelps Ranen; Carl Barden; James Bennett; Jessica 
Botsford; Russell Burnett; Susannah Compton; John Mingus; Reina Nunez; 
Anna Maria Ortiz; Sara Pelton; Sal Sorbello; and Shana Wallace. 

[End of section] 

Footnotes: 

[1] These programs are authorized under Title IV of the Higher 
Education Act of 1965, as amended. 

[2] See Pub. L. No. 102-325. For the purposes of this report, by 
federal student aid revenues, we mean revenues from financial aid 
programs authorized under Title IV of the Higher Education Act of 
1965, as amended. 

[3] See Pub. L. No. 105-244. The 90/10 rule is found at 20 U.S.C. § 
1094(a)(24). 

[4] See 34 C.F.R § 668.23(d)(4). 

[5] See Pub. L. No. 110-315, § 1105. 

[4] For the purposes of our analyses, a school corresponds with a 
unique OPEID. Therefore, we consider an OPEID with five campuses to be 
one school. Conversely, we consider five campuses with a common owner, 
but each with its own OPEID, to be five separate schools. 

[5] Education officials said they can more quickly revoke these schools’
eligibility if they do not meet program responsibilities.Prior to 
changes enacted in August 2008 (Pub. L. No. 110-315), a for-profit 
school lost federal student aid eligibility after one year of non-
compliance. 

[6] Through its eZ-Audit system, Education collects information from 
the annual audited financial statements of all schools receiving 
federal student aid funds. 

[7] We define federal student aid as financial aid programs authorized 
by Title IV of the Higher Education Act of 1965, as amended. 

[8] See Pub. L. No. 102-325. 

[9] See Pub. L. No. 105-244. The 90/10 rule is found at 20 U.S.C. § 
1094(a)(24). 

[10] For the purposes of our analyses, a school corresponds with a 
unique OPEID. Therefore, we consider an OPEID with five campuses to be
one school. Conversely, we consider five campuses with a common owner, 
but each with its own OPEID, to be five separate schools. 

[11] Education officials said they can more quickly revoke these 
schools' eligibility if they do not meet program responsibilities. 
Prior to changes enacted in August 2008 (Pub. L. No. 110-315), a for-
profit school lost federal student aid eligibility after one year of 
non-compliance. 

[12] Through its eZ-Audit system, Education collects information from 
the annual audited financial statements of all schools receiving 
federal student-aid funds. The data for this time period represent the 
most recent data available at the time of our study. 

[13] Education's eZ-Audit data are based on a school's fiscal year. 
IPEDS data capture school characteristics at points during its school 
year and Education funding data are based on a school year. 

[14] All differences described in this report are statistically 
significant at the 95 percent level, unless otherwise noted. 

[15] A key limitation in our regression model is that it does not 
account for student income. Specifically, we could not include a 
school's proportion of students receiving Pell Grants (the only proxy 
for low-income students available to us) in our model because Pell 
Grants are included in schools' 90/10 rates. 

[16] All schools were assigned a specialty according to which of their 
programs had either more enrolled students in the fall of 2008 or more 
graduates in the 2008-2009 school year (depending on the school's 
reporting method) than any of its other programs. 

[17] If a school's parent company owned at least one other for-profit 
school, we considered the school to be part of a corporate chain. 

[18] A publicly-traded company is authorized to offer its securities 
(e.g., stocks and bonds) for sale to the general public, typically 
through a stock exchange. The securities of a publicly-traded company 
are typically owned by many investors. 

[19] Distance education is an option for earning course credit at off-
campus locations via the Internet or other means. 

[20] Federal student aid revenues in excess of a school's educational 
and institutional charges must be excluded from a school's calculation. 

[21] For more information, see 20 U.S.C. § 1094(d) and 34 C.F.R. § 
668.28. 

[22] On a cash basis of accounting, revenues are recorded when they 
are received regardless of when they are earned. Outside of the 90/10 
calculation, schools generally track revenues on an accrual basis of 
accounting, where revenues are recorded when they are earned. 

[23] All other schools include public and non-profit schools. 

[24] For-profit schools' share of revenues may be greater than their 
share of enrollment because they generally have higher tuition rates 
than public schools and enroll a larger percentage of low-income 
students, who are eligible for more federal aid than other students. 
For more 12 information on the characteristics of students at for-
profit schools, see GAO, Proprietary Schools: Stronger Department of 
Education Oversight Needed to Help Ensure Only Eligible Students 
Receive Federal Aid, [hyperlink, 
http://www.gao.gov/products/GAO-09-600] (Washington, D.C., Aug. 17, 
2009). 

[25] Data on 90/10 rates are based on a school's fiscal year. As noted 
earlier, schools' 90/10 rates are verified by independent auditors. 

[26] For each year between 2003 and 2008, the number of schools 
subject to the 90/10 rule was 1,554; 1,873; 1,917; 1,941; 1,950; and 
1,962; respectively. In 2008 (the most recent year for which we had 
data), 1,955 schools were in compliance with the 90/10 rule. 

[27] In some other cases, schools that did not comply either closed or 
lost eligibility for other reasons. 

[28] This number represents unique schools and does not equal the sum 
of all cases because one school was non-compliant in two years. 

[29] Schools' average 90/10 rates for each intervening year were 64, 
64, 63, and 64 percent, respectively. 

[30] The number of schools that received at least 85 percent of their 
revenues from federal student aid in 2008 was 292. After removing the 
7 schools that had 90/10 rates above 90 percent and were therefore out 
of compliance with the 90/10 rule, 285 had rates of 85 percent or 
higher. 

[31] We used a descriptive statistical test, which did not control for 
the effects of additional school characteristics, to analyze whether 
certain schools' average 90/10 rates were significantly different than 
those of other schools.	 

[32] Large schools had higher average 90/10 rates than very small 
schools (with under 100 students). Schools with 500 to under 2,000
students also had higher average 90/10 rates than very small schools. 

[33] This result was expected because Pell Grants are included in 
schools' 90/10 rates. It is logical to assume that a school's 
proportion of students receiving Pell Grants would be correlated with 
its 90/10 rate. A school's proportion of students receiving Pell 
Grants was the only proxy for student income available to us. 

[34] Schools offering bachelor's degrees or higher had an average 
90/10 rate of 67 percent and schools that did not offer academic 
degrees had an average 90/10 rate of 64 percent. 

[35] Additionally, schools specializing in construction, mechanics, or 
manufacturing had a significantly higher average 90/10 rate than their 
counterparts: 71 vs. 66 percent. However, only about 5 percent of 
schools were in this category. 

[36] Schools that specialized in culinary arts, about 2 percent of all 
for-profit schools, also had a lower average 90/10 rate than other for-
profit schools, at 55 vs. 67 percent. The 3 percent of for-profit 
schools that specialized in visual and performing arts also had a 
lower average 90/10 rate than other schools, at 55 vs. 67 percent. 

[37] Enrollment data are based on 2007-2008, while funding data are 
based on the 2008-2009 school year. See appendix II for more details. 

[38] Some chain schools were owned by publicly-traded companies. 

[39] For the purposes of our analysis, a school is an entity with a 
unique OPEID that reports an annual 90/10 rate to Education. Schools may
be organized in different ways, and an OPEID may correspond with one 
or many campuses. Thus, an OPEID with five campuses and 500 22 
students at each campus (for a total of 2,500 students) would be 
considered large. However, five equally-sized campuses (of 500 
students each) with a common owner, but each with its own OPEID, would 
not be considered large because no OPEID has 2,000 students. 

[40] A total of 23 publicly-traded companies owned 278 for-profit 
schools. 

[41] These schools were more likely to have very high 90/10 rates 
compared to very small schools (under 100 students). 

[42] We could not include a measure of student income in our model. 
Some characteristics that are correlated with student income, 
specifically, a school's proportion of minority students and students 
over the age of 25, appear significant in our model. Due to concerns
that these characteristics may partially reflect the effect of student 
income on school 90/10 rates, we do not focus on them in this report. 

[43] In conducting this analysis, we used statistical techniques to 
mitigate the effect of large outliers on our model. We determined a
school's total number of students on a full-time-equivalent basis. 

[44] In a few of our models, this finding was significant only at the 
90 percent confidence level. 

[45] Schools that had experienced a change in ownership in the three 
years prior to 2008 had a significantly higher average 90/10 rate than
other schools (69 vs. 66 percent) only at the 90 percent confidence 
level. In some models, they appeared to be less likely to have very 
high 90/10 rates, but this result was not consistent across all models. 

[46] We tested multiple measures of tuition to account for differences 
in how schools report their tuition rates and found no relationship in 
any measure. We did not attempt to assess whether rising federal 
student aid limits had any impact on schools' tuition rates or 90/10 
rates. 

[47] These award amounts were for first-year dependent undergraduates. 

[48] The experienced firms each conducted over 50 audits and 
collectively conducted 22 percent of all fiscal year 2008 audits of 
for-profit schools for which Education had a record of the audit 
firm's name. The moderately experienced firm conducted 10 audits in 
2008, and the least experienced firms conducted less than 5 audits 
each. 

[49] We did, however, use for-profit schools' reported 90/10 rates to 
calculate each year's average 90/10 rate. 

[50] In the few cases where one reported 90/10 rate was zero and 
another rate was greater than zero, we retained the non-zero rate 
regardless of its fiscal year end date. 

[51] Schools that disburse less than $200,000 a year in federal 
student aid funds may request a waiver of this requirement. However, 
once the waiver period is over, schools must report, and the auditor 
must verify, their 90/10 rates for the years in which the waiver was 
in effect. See 34 C.F.R § 668.27. 

[52] In the few cases in which 2008-2009 school year data were 
unavailable, such as for our total enrollment variable, we used IPEDS 
data from the 2007-2008 school year. 

[53] Some schools report their tuition rates on an academic-year 
basis, while others report their tuition rates on a credit-hour basis 
or contact-hour basis, based on the length of their largest program. 

[54] The correlation coefficient measures the direction of and 
strength of association between two variables, where the strength of 
association refers to how the scores on one variable are distributed 
with respect to the scores on other variables. The statistic ranges 
between -1 and 1; with the strength of association between two 
variables increasing the further the statistic is from zero. 

[55] In a 1997 study, GAO conducted several regression analyses to 
determine the impact of a school's level of reliance on federal 
student aid on its completion rate, placement rate, and default rate, 
while holding a variety of other school characteristics constant. That 
study found that schools that relied more on federal student aid 
tended to have worse student outcomes. Our regression analyses, on the 
other hand, measured whether certain school characteristics predicted 
an increased likelihood of having a very high rate of reliance on 
federal student aid (a 90/10 rate of above 85 percent). See GAO, 
Proprietary Schools: Poorer Student Outcomes at Schools that Rely More 
on Federal Student Aid, [hyperlink, 
http://www.gao.gov/products/GAO/HEH-97-103] (Washington, D.C.: June 
13, 1997). 

[56] Our dependent variable was generated using schools' 90/10 rates, 
which are calculated using the dollar value of a school's federal 
student aid disbursements, including Pell Grants. We tested the 
relationship between a school's percent of students receiving Pell 
Grants and its 90/10 rate, and found about a 0.5 correlation. 

[57] Ordinary least squares regression analysis can illustrate the 
specific change in a dependent variable associated with a unit change 
in an independent variable; for example, the predicted increase in 90/ 
10 rates associated with a percentage point increase in the proportion 
of adult learners ages 25 and over. 

[58] We tested three separate measures of tuition; these measures are 
described in appendix II. 

[59] The odds ratio for a continuous variable reflects the change in 
odds associated with a one-unit increase in the independent variable, 
such as a percentage point increase in the proportion of adult 
learners ages 25 and over. 

[60] Schools with tuition rates that did not exceed 2008-2009 Pell 
Grant and Stafford Loan award limits did have slightly higher average 
90/10 rates than other schools, at 68 vs. 66 percent. 

[61] The 90/10 rate is calculated as the proportion of school revenue 
received from federal student aid sources, including Pell Grants. The 
proportion of students receiving Pell Grants, the only proxy for 
student income available to us, was significantly correlated with the 
90/10 rate at r=0.48. 

[62] The correlations between the proportion of students receiving 
Pell Grants and the proportion of adult learners over the age of 25, 
and with the proportion of minority students, were modest but 
statistically significant at r=.11 and .39 respectively. 

[End of section] 

GAO's Mission: 

The Government Accountability Office, the audit, evaluation and 
investigative arm of Congress, exists to support Congress in meeting 
its constitutional responsibilities and to help improve the performance 
and accountability of the federal government for the American people. 
GAO examines the use of public funds; evaluates federal programs and 
policies; and provides analyses, recommendations, and other assistance 
to help Congress make informed oversight, policy, and funding 
decisions. GAO's commitment to good government is reflected in its core 
values of accountability, integrity, and reliability. 

Obtaining Copies of GAO Reports and Testimony: 

The fastest and easiest way to obtain copies of GAO documents at no 
cost is through GAO's Web site [hyperlink, http://www.gao.gov]. Each 
weekday, GAO posts newly released reports, testimony, and 
correspondence on its Web site. To have GAO e-mail you a list of newly 
posted products every afternoon, go to [hyperlink, http://www.gao.gov] 
and select "E-mail Updates." 

Order by Phone: 

The price of each GAO publication reflects GAO’s actual cost of
production and distribution and depends on the number of pages in the
publication and whether the publication is printed in color or black and
white. Pricing and ordering information is posted on GAO’s Web site, 
[hyperlink, http://www.gao.gov/ordering.htm]. 

Place orders by calling (202) 512-6000, toll free (866) 801-7077, or
TDD (202) 512-2537. 

Orders may be paid for using American Express, Discover Card,
MasterCard, Visa, check, or money order. Call for additional 
information. 

To Report Fraud, Waste, and Abuse in Federal Programs: 

Contact: 

Web site: [hyperlink, http://www.gao.gov/fraudnet/fraudnet.htm]: 
E-mail: fraudnet@gao.gov: 
Automated answering system: (800) 424-5454 or (202) 512-7470: 

Congressional Relations: 

Ralph Dawn, Managing Director, dawnr@gao.gov: 
(202) 512-4400: 
U.S. Government Accountability Office: 
441 G Street NW, Room 7125: 
Washington, D.C. 20548: 

Public Affairs: 

Chuck Young, Managing Director, youngc1@gao.gov: 
(202) 512-4800: 
U.S. Government Accountability Office: 
441 G Street NW, Room 7149: 
Washington, D.C. 20548: