WHOSE PERFORMANCE COUNTS? EQUITY CONCERNS IN PERFORMANCE
FUNDING POLICIES
By
Renata Opoczynski

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Higher, Adult, and Lifelong Education-Doctor of Philosophy
2017

ABSTRACT
WHOSE PERFORMANCE COUNTS? EQUITY CONCERNS IN PERFORMANCE
FUNDING POLICIES
By
Renata Opoczynski
While accountability in higher education has been a topic of debate for decades, in recent
years the discussions have shifted to emphasize efficiency and economic measures of success. A
prominent example of this accountability movement is the increase in popularity of performance
funding policies. These policies connect specific outcomes on state selected metrics to increased
state funding (Goldstein, 2012). Performance funding policies purport to increase efficiency by
rewarding reductions in cost and increases in specific economic outcomes. However, many of
these policies neglect a similar emphasis on maintaining access, which may lead to undesirable
consequences including reducing the enrollment of traditionally underserved students (students
from low socioeconomic status (SES) families and historically underserved students of color)
(Dougherty et al., 2014). Therefore, this study explored whether performance funding policies
have an effect on underrepresented students’ enrollment.
Through a fixed effects panel analysis covering the years 2000 to 2014, this study
explored any changes in enrollment of underserved minority students and Pell Grant receiving
students in public four-year institutions. Findings from this study demonstrate that performance
funding does have the potential to influence enrollment profiles at U.S. public four-year
institutions. Specifically, this study found it changed the enrollment of underserved minority
students. Further, these influences may not be equitable across all institutions, and instead may
effect lower status institutions in a different manner than higher status institutions. Specifically,
those with more flexibility in their enrollment profile may be more likely to change their

enrollment of both Pell Grant students and underserved minority students. These findings have
profound implications for higher education institutions, policy formation, and social equity.

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................................... vii
LIST OF FIGURES ..................................................................................................................... ix
Chapter 1 ...................................................................................................................................... 1
Problem Statement ......................................................................................................................... 1
Push for Efficiency .................................................................................................................. 2
Performance Funding ............................................................................................................... 4
Change in Access ..................................................................................................................... 5
Purpose of Study ............................................................................................................................ 7
Research Questions ........................................................................................................................ 8
Conceptual Framework .................................................................................................................. 9
Chapter 2 ...................................................................................................................................... 13
Higher Education Funding ........................................................................................................... 14
State Funding ............................................................................................................................... 14
State Appropriations .............................................................................................................. 15
Variance Among States.......................................................................................................... 16
Price of Higher Education ...................................................................................................... 17
Financial Aid.......................................................................................................................... 18
Accountability in Higher Education ............................................................................................ 19
Defining Accountability......................................................................................................... 19
History of Accountability ...................................................................................................... 21
Forms of Accountability ........................................................................................................ 24
Push for External Accountability ........................................................................................... 25
Performance Funding ................................................................................................................... 27
Defining Performance Funding.............................................................................................. 27
Spread of Performance Funding ............................................................................................ 30
Concerns with Performance Funding..................................................................................... 33
Best Practices in Performance Funding ................................................................................. 35
Effectiveness of Performance Funding .................................................................................. 38
Immediate impacts ........................................................................................................... 38
Intermediate institutional change ..................................................................................... 38
Ultimate student outcomes ............................................................................................... 39
Unintended outcomes....................................................................................................... 41
Enrollment and Retention Literature ........................................................................................... 43
General Trend Across Race ................................................................................................... 43
General Trend Across SES .................................................................................................... 45
Institutional Factors ............................................................................................................... 46
Student demographics ...................................................................................................... 46
Structural characteristics .................................................................................................. 48
Financial characteristics ................................................................................................... 49

iv

State Factors ........................................................................................................................... 50
Financial........................................................................................................................... 50
Population data................................................................................................................. 52
Conclusion ................................................................................................................................... 53
Chapter 3 ...................................................................................................................................... 54
Sample/Population ....................................................................................................................... 54
Data Sources ................................................................................................................................ 56
Parent-Child Issues ................................................................................................................ 57
Variables ...................................................................................................................................... 58
Dependent Variables .............................................................................................................. 58
Institutional Variables ............................................................................................................ 58
State Variables ....................................................................................................................... 60
Analytical Method ....................................................................................................................... 64
Models.................................................................................................................................... 65
Limitations ................................................................................................................................... 67
Chapter 4 ...................................................................................................................................... 69
Descriptive Statistics.................................................................................................................... 69
Performance Funding and Underserved Student Enrollment ...................................................... 75
Pell Grant Variables ............................................................................................................... 76
Pell Grant percent ............................................................................................................ 77
Pell Grant amount ............................................................................................................ 78
Underserved Minority Student Variables .............................................................................. 81
Total underserved minority students ................................................................................ 81
Percent underserved minority students ............................................................................ 83
Institutional Type and Performance Funding .............................................................................. 88
Carnegie Classification .......................................................................................................... 90
Carnegie classification and Pell Grant variables ............................................................. 91
Carnegie classification and underserved minority student variables ............................... 93
Selectivity .............................................................................................................................. 96
Selectivity and Pell Grant variables ................................................................................. 97
Selectivity and underserved minority student variables .................................................. 98
Conclusion ................................................................................................................................. 101
Chapter 5 .................................................................................................................................... 102
Summary of Study ..................................................................................................................... 102
Discussion of Findings............................................................................................................... 104
Underserved Student Enrollment ......................................................................................... 104
Institutional Type ................................................................................................................. 110
Implications................................................................................................................................ 114
Implications for Policy......................................................................................................... 114
Implications for Research .................................................................................................... 116
Conclusion ................................................................................................................................. 118
APPENDIX ................................................................................................................................ 120

v

BIBLIOGRAPHY ...................................................................................................................... 125

vi

LIST OF TABLES

Table 1

List of States with Performance Funding and Years ..................................... 61

Table 2

Variable Definitions and Sources .................................................................. 62

Table 3

Summary Statistics For All Institutions Across All Years ............................ 71

Table 4

Descriptive Statistics by Performance Funding ............................................. 73

Table 5

Results for the Effects of Performance Funding on Pell Grant Percent ........ 78

Table 6

Results for the Effects of Binary Performance Funding on Pell Grant Amount
........................................................................................................................ 80

Table 7

Results for the Effects of Binary Performance Funding on USM Total ........ 83

Table 8

Results for the Effects of Binary Performance Funding on USM Percent .... 84

Table 9

Results for the Effects of Performance Funding Count on USM Percent ..... 86

Table 10

Transition Rates Across Carnegie Categorical Variable ............................... 89

Table 11

Transition Rates Across Selectivity Categorical Variable ............................. 90

Table 12

Results of Performance Funding Interacted with Carnegie on Pell Grant
Variables ........................................................................................................ 92

Table 13

Results of Performance Funding Interacted with Carnegie on USM Variables
........................................................................................................................ 94

Table 14

Results of Performance Funding Interacted with Selectivity on Pell Grant
Variables ........................................................................................................ 97

Table 15

Results of Performance Funding Interacted with Selectivity on USM
Variables ........................................................................................................ 99

Table 16

Results for the Effects of Performance Funding Count on Pell Grant Percent
...................................................................................................................... 121

Table 17

Results for the Effects of Performance Funding Count on Pell Gant Amount
...................................................................................................................... 122

Table 18

Results for the Effects of Performance Funding Count on USM Total ....... 123

vii

Table 19

Tobit Results for the Effects of Binary Performance Funding on USM Percent
...................................................................................................................... 124

Table 20

Tobit Results for the Effects of Performance Funding Count on USM Percent
...................................................................................................................... 124

viii

LIST OF FIGURES

Figure 1

States with Performance Funding Policies Active and in Transition on
July 2015 ........................................................................................................ 31

Figure 2

Six Year Graduation Rates for Four Year Institutions by Percent of
Pell Recipients ............................................................................................... 47

ix

Chapter 1
Over the last two decades, policymakers and the public have increasingly sounded an
alarm related to the lack of accountability in higher education. While discussions regarding
accountability in higher education have been commonplace since the 1970s, the intensity of their
calls has significantly increased in recent years and taken a marked shift toward efficiency and
demanding production of specific results (Burke, 2005). Further, Fuller, Lebo, and Muffo (2012)
noted that higher education institutions should expect these demands to continue and possibly
even increase in future years. Those sounding the alarm usually claim that higher education has
strayed from its core mission of educating students, citing the fact that despite students attending
higher education at the highest rates ever, the six year graduation rate is at an all-time low (Kelly
& Schneider, 2012). At the same time, the price of participating in higher education is
consistently increasing, leading policymakers and the public to question the value of higher
education (McMahon, 2009).
Problem Statement
Underpinning this accountability movement is a shift in society’s view of its social
contract with higher education. Throughout history, the relationship between the government
and higher education institutions has changed, as has society’s view on the purpose of higher
education (Thelin, 2004). The changing nature and expectations inherent in the social contract
can influence the values and missions of higher education institutions, expectations on and from
both political parties, funding for higher education at the state and federal level, power
relationships between the parties, and views on the purpose of higher education (Edgell, 2009;
Gumport, 2001; Kezar, 2004). Historically, the social contract regarding public higher education
involved state leaders adequately funding autonomous institutions in exchange for those

1

institutions serving the common good of the state (Thelin, 2004). However, beginning in the
1980s, state legislators shifted the conditions of the social contract by imposing a greater degree
of control over higher education institutions, while at the same time decreasing the state financial
support to public colleges and universities (Edgell, 2009; Gumport, 2001).
Often, the new expectations for institutions require a market orientation, which shifts
away from the premise that higher education is a public good focused on meeting the diverse
social needs of a state towards viewing higher education as a private good focused strictly on
private economic outcomes and efficiency (Gumport, 2001; McMahon, 2009; Slaughter &
Rhoades, 2004). As a result, state officials decreased financial support for higher education,
believing that those who directly benefit from the education (the students) should pay a greater
share of the costs of higher education, as opposed to the state significantly subsidizing the cost as
a public investment (Slaughter & Rhoades, 2004). Performance funding policies are a prominent
development within this broader policy shift.
Push for Efficiency
When policy makers adopt a market orientation of higher education, success is measured
through a production function lens, which emphasizes efficiency. Simply defined, efficiency is
the measure of outputs over inputs. State leaders expecting greater efficiency from higher
education pressure institutions to increase outputs, decrease costs, or both. Performance funding
policies are an example of this push for efficiency. In higher education policy, efficiency models,
such as performance funding, emphasize a narrow view of outputs with a focus on market-based
benefits e.g., graduation rates, salary post-graduation, faculty research productivity (McMahon,
2009; Slaughter & Rhoades, 2004). This heightened push for economic focused accountability
and efficiency is the result of a complex set of factors including shifts in state funding models,

2

economic constraints, a growing federal investment in higher education, the public’s demand for
higher education, labor market expectations, and the public’s perception of the cost and value of
higher education (Eaton, 2012; Keller, 2012; NASBO, 2013).
The ability for higher education leaders to respond to efficiency requests is complicated
by the simultaneous decrease in state financial support. Despite three years of increases in per
student appropriations, the State Higher Education Executive Officers (SHEEO) found that
nationally in 2015 per Full Time Equivalent (FTE) funding is still 15.3% lower than levels
before the 2008 Great Recession (SHEEO, 2016). In the last year, most states increased their per
FTE funding, however direct dollar appropriations are still below appropriated amounts in 2008
and 12 states continued to cut their funding per FTE (Mitchell, Leachman, & Masterson, 2016).
The availability of alternative revenue streams, including tuition and fees, makes higher
education an easy target for funding cuts (National Association of State Budget Officers
[NASBO], 2013). Further, the growing pressure from other state needs (e.g., Medicaid, K-12
education, corrections) limits the ability of state leaders to increase funding for higher education
in the near future (NASBO, 2013; SHEEO, 2013).
The push for efficiency is further complicated by the lack of agreement on what output
measures should be used to gauge efficiency (Powell, Gilleland, & Pearson, 2012). Graduation
rates, research output, time to degree, degrees awarded in specific fields, job placement rates and
graduate salaries are often cited as metrics for assessing institutional efficiency. Yet, institutional
leaders often lack the necessary data and benchmarks to know if there is a relationship between
their expenditures and outcomes (Powell et al., 2012). As a result, higher education
administrators tend to increase their expenditures on programs intended to improve outcomes
(e.g., development education, living learning communities, first year seminars) and hope the

3

increase will also improve efficiency (Powell et al., 2012). While it is difficult to define
efficiency in higher education, the term is nonetheless used to motivate many policy discussions.
As McMahon (2009) explains, the term “efficiency” is “thrown around with wild abandon” (p.
12) by policy creators when advocating for higher education accountability. Performance
funding policies help alleviate the challenges associated with this ambiguity by providing
specific definitions of efficiency and outlining what outputs the state legislatures value and
utilize to measure quality.
Performance Funding
Performance funding policies connect specific performance outcomes on state selected
measures to increased state funding (Goldstein, 2012). By tying funding to state specific metrics,
legislators believe institutions will be incentivized to improve efficiency. Through performance
funding policies, institutions are held accountable for meeting outcome targets as part of the state
appropriations process. Performance funding policies allow state officials to emphasize metrics
such as graduation rates, financial stewardship, credit hours earned, wages of graduates, and
research grants that they believe will better incentivize efficiency. The institutions that perform
well on the metrics are financially rewarded through increases in their operating budget. Further,
proponents including the Lumina Foundation, Complete College America (CCA), NASBO and
the National Governors Association (NGA), highlight that performance funding shifts incentives
away from inputs such as enrollments to outcomes and results and helps align higher education
with state priorities (NASBO, 2013). Skeptics however argue that performance funding might
encourage institutions to either relax graduation standards or turn away underrepresented
students by tightening their admission criteria (Dougherty et al., 2014). As of July 2015,
performance funding policies exist in 32 states; five more states were in the process of

4

transitioning (National Conference of State Legislatures [NCSL], 2015).
The belief that performance funding policies will increase efficiency assumes that higher
education operates within a competitive marketplace where the rules and assumptions of market
economics (such as rational choice theory or perfect knowledge) can be met (Gumport, 2001).
However, many higher education researchers believe standard economic market assumptions do
not apply to institutions of higher education (see quasi markets below) and believe policies and
decisions based on these assumptions are flawed (e.g., Gumport, 2001; Marginson, 1997;
McMahon, 2009; Slaughter & Rhoades, 2004). If economic assumptions cannot be met and
higher education does not function as a traditional market, then instituting performance funding
policies would create quasi markets. Market behavior in quasi markets arises from the priorities
of the policy makers, as opposed to true economic efficiency and supply and demand of
resources (Taylor, Cantwell & Slaughter, 2013).
Change in Access
If performance funding policies value efficiency (increased economic outputs at lower
costs) and quasi markets reward institutions that engage in behavior valued by the policy, then
institutions may engage in behaviors to increase their quality and decrease their costs in order to
reap the benefits of the performance funding policy. While there may be advantages to these
behaviors, they may also have unintended consequences. For example, it is possible that
performance funding policies create unintended consequences for access among
underrepresented students. Immerwahr, Johnson, and Gasbarra (2008) noted that many higher
education leaders believe cost, quality, and access are engaged in an “Iron Triangle,” where a
change in one inevitably will affect at least one of the other two. For instance, if an institution
sees an increase in their enrollment numbers through a large first year class, they will either have

5

to increase their costs or their quality may decrease through changes such as larger classes or
increasing the occupants per room in their residence halls. In their study, Immerwahr et al.
(2008) interviewed a diverse group of 30 institutional presidents and found that almost all agreed
that the growing pressures for accountability and decreasing state support increase the price of
higher education, making it less accessible. While legislators and the public may disagree with
this perspective, it demonstrates how college presidents view decision making related to the
elements of the iron triangle.
As previously stated, the belief that cost, quality, and access are fundamentally
interwoven may lead performance funding policies to create unintended outcomes related to
access for underserved students. Specifically, the quasi markets created by performance funding
policies may create incentives for institutions to attend to the metrics measured by policymakers
(efficiency) and give less attention to the metrics for which institutions are not held accountable
(underrepresented student enrollment). In other words, changes to quality and cost may lead
institutional leaders to feel they must change enrollment profiles (Immerwahr et al., 2008).
Research supports this concern; a commonly mentioned possible unintended impact of
performance funding is restricting the admission of less prepared students (Dougherty et al.,
2014). In fact, many institutional administrators acknowledged at least thinking about changing
the institutional makeup of their classes in order to perform better on performance funding
metrics (Lahr et al., 2014). Therefore, in this study, it is hypothesized that performance funding
policies may lead to unintended consequences regarding access for underrepresented students.
Institutions changing enrollment profiles, in particular enrolling less underrepresented
students of color and Pell eligible students, in order to perform better on performance funding
metrics could have serious repercussions to the whole higher education system. Educational

6

attainment in the U.S. for adults aged 25-34 is currently lagging behind 15 other OECD nations
and those nations continue to make progress while the U.S. stagnates (Heller, 2013). In order to
match the level of the highest performing nation, the U.S. would need to increase the number of
students aged 25-64 with college degrees by 7.9% each year until 2020 (Perna, 2013). However,
the high school graduate population (the available pool of students for enrolling in higher
education) is becoming more racially diverse and financially needy (National Center for
Education Statistics [NCES], 2014; Western Interstate Commission for Higher Education
[WICHE], 2012). With these population changes, limiting the enrollment of underrepresented
students of color and Pell eligible students could decrease the number of students eligible to
enroll in and complete higher education. Therefore, if performance funding policies limit
enrollment of underrepresented students, these policies may be antagonistic to the goal of
increasing U.S. higher education participation and completion and decrease the global
competiveness of the U.S. higher education system.
Purpose of Study
Performance funding policies are increasing in prevalence as states push for greater
efficiency and specific outcomes from higher education institutions. These policies connect
specific outcomes on state selected metrics to increased state funding (Goldstein, 2012).
Performance funding policies purport to increase efficiency by rewarding reductions in cost and
increases in specific economic outcomes. However, many of these policies neglect a similar
emphasis on maintaining access, which may lead to undesirable consequences including
reducing the enrollment of traditionally underserved students (students from low socioeconomic
status (SES) families and historically underserved students of color) (Dougherty et al., 2014).
College enrollment and graduation rates differ significantly based on race and family

7

income (Heller, 2013; NCES, 2016). In 2012, 68% of White students enrolled in college directly
after high school compared to 63% of Black students and 62% of Hispanic students (NCES,
2016). Discrepancies in high school graduation rates and lower completion rates in higher
education for underrepresented students of color lead to even greater disparities in bachelor
degree attainment across race. In 2015, only 21% of Black and 16% of Hispanic 25 to 29 year
olds had a bachelor’s degree compared to 43% of Whites in the same age group (NCES, 2014).
The educational gap is more pronounced when looking at SES. Heller (2013) found students
from families in the top quintile of the income distribution were significantly more likely to be
enrolled full time at four-year institutions than students from the bottom quintile (55% verse
33%). The enrollment divide foretells a continuing and systemic issue, as Heller (2013)
concluded that the difference between students from the top and bottom quintile who attained a
bachelor’s degree eight years after high school was 40 percentage points.
Further limiting enrollment of underrepresented students as a response to performance
funding would widen these gaps, which could stand in the way of national policy priorities.
Researchers found that in order to meet the national college completion goals the U.S. must not
only acknowledge the attainment gap, but also actively work to improve the outcomes for
underrepresented students (Heller, 2013; Perna, 2013). Therefore, this study explored whether
performance funding policies have an effect on underrepresented students’ enrollment.
Research Questions
The current educational inequality in the U.S. has led to the increased economic and
social stratification over time and needs to be addressed (Perna, 2013). However, limited
research has explored how higher education institutions are balancing the competing needs of
demonstrating efficient economic outcomes via performance funding while maintaining access

8

and success for all students, particularly in the four-year sector. As such, this study explored how
performance funding influences the enrollment of underserved students and focused on the
following research questions:
1) Are there unintended consequences from performance funding policies related to
underserved students’ enrollment at public four-year institutions?
a. Does performance funding have an influence on the enrollment of Pell Grant
eligible students?
b. Does performance funding have an influence on the enrollment of
underserved minority students?
2) Does institutional type influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
a. Does Carnegie classification influence the effect of performance funding on
underserved students’ enrollment at public four-year institutions?
b. Does selectivity influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
Conceptual Framework
In order to understand how performance funding policies influence underrepresented
students’ enrollment, this study utilizes the Iron Triangle concept introduced by Immerwahr et al.
(2008) and weaves in the concept of Quasi Markets. These theories suggest that the higher
education market does not follow traditional economic principles and instead institutions may
make decisions striving to increase prestige or policy benefits. Further, these decisions may have
an intended or unintended impact on access due to the relationship between access, cost, and
quality.

9

As noted above, institutional administrators believe that cost, quality, and access are
fundamentally interwoven in an Iron Triangle and that a change in one will lead to a change in at
least one of the other two. If performance funding policies value efficiency (increased economic
outputs at lower costs) then institutions may engage in behaviors to increase their quality and
decrease their costs in order to reap the benefits of the performance funding policy. Specifically,
performance funding policies may create incentives for institutions to attend to the metrics
measured by policymakers (efficiency) and give less attention to the metrics for which
institutions are not held accountable (underrepresented student enrollment). In other words,
changes to quality and cost may lead institutional leaders to feel they must change enrollment
profiles (Immerwahr et al., 2008).
The conceptual framework for this study is also informed by quasi markets. Historically,
higher education institutions were autonomous from the market, but during the late twentieth
century institutions began to engage in market like behaviors, (Slaughter & Leslie, 1997).
Institutions engage in market-like behaviors when they compete for funds from external
stakeholders with a focus on financial efficiency (Slaughter & Leslie, 1997). The U.S. now has a
well developed higher education market (Marginson, 1997) and performance funding policies
assume that higher education operates within a competitive marketplace where the rules and
assumptions of market economics (such as rational choice theory or perfect knowledge) can be
met (Gumport, 2001).
However, numerous scholars identified five significant restrictions of the standard
economic viewpoint for higher education, which help explain the behavior of quasi markets.
First, Gumport (2001) explained that the market for higher education is rendered irrelevant due
to regulation by state and federal policies, and admission decisions. In other words, forces and

10

decisions that arise from state and federal policies and institutional decisions supersede standard
market principles. Second, governmental subsidies cause the price of higher education to be
below the actual market price (Marginson, 1997). Therefore, standard market pricing behavior
can not be applied to higher education. Third, the market incorrectly assumes that students are
capable of informed choice with perfect knowledge (Gumport, 2001; McMahon, 2009; Slaughter
& Rhoades, 2004). While many students research higher education extensively, it is difficult to
account for all the unknown factors related to higher education (i.e., how much value added the
student will receive from their degree) and make decisions with true perfect knowledge as
required for market principles to apply. Fourth, perfectly competitive markets would assume
homogeneity of product, however it is clear that not all colleges provide the same rate of return
on a student’s investment (Slaughter & Rhoades, 2004). Finally, higher education institutions are
not necessarily privately owned and most are not focused on maximizing profit (LeGrand, 1991).
Since quasi markets arise out of policy and are not influenced by market price, informed
choice by consumers, or a focus on maximizing profits, they account for the unique
characteristics listed above. In other words, rewards in quasi markets are based on policy
priorities rather than economic efficiency and structured from decisions made by politicians with
the hope of furthering public interest (Le Grand, 1991; Taylor, et al., 2013). Weaving the two
theories together helps us understand how performance funding may influence institutional
enrollment patterns. If certain outcomes are coveted by state legislators through performance
funding, then administrators may strive to increase their prestige and financial appropriation by
responding to the pressures of the performance funding policy instead of standard economic
market pressures, thereby creating a quasi market. In other words, administrators may feel the
need to increase quality or decrease costs in order to perform better in the quasi market created

11

by the performance funding policy. However, since access, quality, and costs are locked in an
iron triangle, then the changes to cost and quality would also impact enrollment decisions.

12

Chapter 2
The new accountability movement shifts the expectations of the social contract with
higher education to one focused on efficiency and limited financial support from the state
(Edgell, 2009; Gumport, 2001; McMahon, 2009; Slaughter & Rhoades, 2004). Performance
funding policies are an outgrowth of this push for efficiency. The quasi markets created by
performance funding policies reward institutions that engage in behavior favored by the policy
and may create incentives for institutions to attend to the metrics favored by policymakers
(efficiency) and give less attention to the metrics for which institutions are not held accountable
(such as underrepresented student enrollment). In other words, performance funding policies
may have unintended consequences related to access and persistence for underrepresented
students due to the iron triangle of cost, quality and access (Immerwahr et al., 2008).
Specifically, changes to quality and cost may lead institutional leaders to feel they must change
access (Immerwahr et al., 2008).
Limited research has explored how higher education institutions are balancing the
competing needs of demonstrating efficient economic outcomes via performance funding while
maintaining access and success for all students, particularly in the four-year sector. As such, this
study will explore how performance funding influences the enrollment and completion of
underrepresented students. This chapter presents an overview of the literature related to higher
education funding (with a specific focus on state funding), the accountability movement, and
performance funding specifically. It then explores the literature on enrollment and retention
from an institutional and state perspective with a focus on underrepresented students of color and
low SES students. It concludes by presenting how this study will enhance and expand literature
on performance funding and underrepresented student enrollment and retention.

13

Higher Education Funding
Higher education institutions, specifically four year public institutions, collect revenue
from several sources including tuition and fees, state and local appropriations, contracts/grants,
donations, and interest from their endowments. Of these, tuition and fees and state and local
appropriations make up the vast majority of institutional revenue. Public institutions collected
$66.9 billion in net tuition revenue in 2015 and $90.9 billion in state and local funding (SHEEO,
2016). On average, state and local appropriations makes up 54% of the revenue for public higher
education institutions (Mitchell, et al., 2016). Tuition revenue makes up 46.5% of all operating
expenses for institutions on average, which is up from 23.8% in 1988; though it should be noted
that there is great variance in funding among the states (SHEEO, 2016).
The makeup of higher education revenue has significantly shifted as higher education
takes a market orientation and perspective shifts to view the sector as more of a private good. In
1998 institutions received 3.2 times as much revenue from state and local government as they did
in tuition, now they receive only 1.2 times as much from state and local appropriations as they do
from tuition (Mitchell et al., 2016). Changes in the funding of higher education have led to a
33% increase (adjusted for inflation) in the net cost of attendance from 2008 to 2015 (Mitchell et
al., 2016). At public four-year institutions, students now cover around half of the costs of their
education (Desrochers & Hurlburt, 2014). As this research focuses on performance funding,
which is a form of state appropriations, this literature review will focus on that area as well.
State Funding
States fund higher education institutions via two primary methods: direct appropriations
and funding students through financial aid. These two methods benefit different numbers of
students as well as different groups of students (Toutkoushian & Shafiq, 2010). State

14

appropriations benefit all in-state students across all different types of income levels by lowering
the price of higher education (Toutkoushian & Shafiq, 2010). In contrast, state financial aid
benefits a smaller number of students as only those who meet eligibility criteria (be it merit or
financial based) receive the funds, which in turn lowers the price they pay for higher education
(Toutkoushian & Shafiq, 2010). Additionally, need based financial aid targets lower income
students as opposed to state appropriations, which benefits all income groups. Both state
appropriations and state financial aid are discussed below.
State Appropriations
The most frequently used option to provide funding to higher education institutions is in
the form of direct state appropriations. More than 90% of state funding for higher education is in
the form of state appropriations (Hauptman, 2011). Policymakers hope that these funds will be
used to reduce the price charged to in state students, allowing more students to attend higher
education (Toutkoushian & Shafiq, 2010). In 2015, state and local funding for higher education
totaled $90.9 billion (SHEEO, 2016). After the great recession hit, many states cut higher
education funding significantly. The last few years have seen a return to funding higher
education, however funding is still 18% lower than it was in 2008, the start of the great recession
(Mitchell et al., 2016).
When looking at funding per full time equivalent students (FTE) the trend remains the
same. In 2015, average state and local funding per FTE was $6,966, an increase of 5.2% (in
constant dollars) from 2014 (SHEEO, 2016). However this increase in funding per FTE is due
mostly to declining enrollment numbers as opposed to increases in higher education funding
(SHEEO, 2016). Additionally, despite three years of increased appropriations per FTE, 2015
funding is still 15.3% lower than prerecession funding (SHEEO, 2016). In fact, in 2016 all states

15

except for Montana, Wisconsin, Wyoming, and North Dakota have lower funding per student
than they did before the recession (Mitchell et al., 2016).
In response to the great recession, the federal government passed the American Recovery
and Reinvestment Act (ARRA) in 2009, which included funding to help stabilize state support
for higher education (SHEEO, 2016). The relative stability of state support for higher education
directly after the great recession is a result of ARRA funds (SHEEO, 2016). When the ARRA
funds ran out in 2012, higher education support decreased by 7% (SHEEO, 2014).
The availability of alternative revenue streams, such as tuition and fees, makes higher
education an easy target for funding cuts (NASBO, 2013). Further, the growing pressure from
other state needs (e.g., Medicaid, K-12 education, corrections) limits the ability of state leaders
to increase funding for higher education in the near future (Mitchell et al., 2014; NASBO, 2013;
SHEEO, 2013). However, state funding still accounts for a large share of the revenue higher
education institutions receive for general operating expenses (SHEEO, 2011).
Variance Among States
There is significant variance regarding state funding for higher education between states
(Weerts & Ronca, 2012). For instance, in 2016, 38 states increased per FTE funding for higher
education, while 12 states decreased funding (Mitchell et al., 2016). Research has found several
factors can influence the amount of money appropriated to higher education including: other
state budgetary demands, general economic health, the business cycle, state higher education
governance structures, state fiscal health, state politics, demographic factors, population changes,
party affiliation of the governor, party control in the legislature; legislative professionalism; and
local funding (Delaney & Doyle, 2011; McLendon, Hearn & Mokher, 2009; SHEEO,
2014;Tandberg, 2010; Tandberg & Grifith, 2013; Weerts & Ronca, 2012).

16

Price of Higher Education
Aligned with this decline in state revenue, there has been an increase in the price of
tuition. Published tuition (often referred to as sticker price) has increased by $2,333 (after
adjusting for inflation) nationally since the 2007-08 school year, a 33% increase (Mitchell et al.,
2016). It should be noted that as states began to restore funding to higher education the increases
in tuition have been more modest. In the 2015-16 school year, the average tuition increase at
public four year institutions across all states (after adjusting for inflation) was only $254 or 2.8%
(Mitchell et al., 2016). From a FTE perspective, net tuition (adjusted for inflation via the Higher
Education Cost Adjustment) per FTE increased 5.0% annually from 2009 to 2011, then by 7.1%
in 2012 and 4.7% in 2013 (SHEEO, 2014). Despite the increases in published tuition, the
revenue gained from tuition only offset about 60% of state and local funding cuts to masters and
doctoral institutions and 30% of funding cuts to bachelors institutions (Mitchell et al., 2014).
This increase in tuition has resulted in students taking on more debt to pay for their
higher education degree. 6o% of the students graduating in 2014 from public four-year
institutions had taken on student loan debt (Mitchell et al., 2016). In the 2014-2015 school year,
the average amount of debt students graduating from public four-year institutions left with was
$25,500, up from $14,300 just three years prior (Mitchell et al., 2014; Mitchell et al., 2016).
Federal financial aid (primarily through the Pell Grant program) has increased to counter this
rising tuition level. The amount of aid distributed through the Pell Grant program increased by
80% between 2007-08 and 2014-15, resulting in a 2.7 million increase in the number of students
who receive Pell grants as well as an increase in the average amount awarded from $3,028 to
$3,673 (Mitchell et al., 2016). On the other hand, state financial aid has decreased by about 4%
since the 2007-08 school year (Mitchell et al., 2016).

17

Financial Aid
The second main funding source state governments provide is through state financial aid.
These funds are given directly to students and can take the form of either need based aid or merit
based aid. Need based aid is awarded to students based on financial need as calculated through
their family’s expected family contribution. Merit based aid is state financial aid that is awarded
based on factors not related to a student’s financial need, for instance performance in high school
or on standardized admission tests such as the ACT or SAT. Historically, the vast majority of
state financial aid was in the form of need based aid and merit aid programs were very small and
specialized programs targeting groups such as widows or children of police officers killed in the
line of duty (Heller, 2011). In 1993 Georgia created the Helping Outstanding Pupils
Educationally (HOPE) program and the popularity of the program inspired many other states to
create similar merit aid programs (Heller, 2011). The differing foci of merit and need based aid
lead to significantly different funding distributions across income brackets. For instance, Heller
(2011) found that 85% of state need based aid went to students in the lower or lower middle
income quartiles, versus only 44% of merit based aid went to students in the lower or lower
middle income quartiles. This raises questions regarding the relationship between income and
measures used to access academic merit (Heller & Marin, 2004).
Total state financial aid in 2014-15 was $12.4 billion (National Association of State
Student Grant and Aid Programs [NASSGAP], 2016). In 2014-2015 state aid constituted 13% of
total state financing of higher education (The College Board, 2016). Once again, there was great
variance amongst the states with a range from 0% in New Hampshire to 38% in South Carolina
(The College Board, 2016). The shift toward merit aid has been increasing over the years and in
2014-2015 24% of all state aid was in the form of merit aid (The College Board, 2016). Yet

18

again, there is great variance amongst the states with 26 states allocating 95% or more of their
aid in a need based manner and 14 states allocating more than half of their aid on merit based
formulas (The College Board, 2016).
Accountability in Higher Education
Discussions regarding accountability have occurred in higher education since the 1970s.
However, in recent years, they have taken a significant shift toward discussions of efficiency and
demanding the production of specific results (Burke-1, 2005). Several factors have driven the
emphasis on efficiency. First, the massification of higher education has shifted the college degree
into a necessity for most Americans and brought the higher education sector into the spotlight
(Burke-1, 2005). Second, economic competiveness and the belief that the United States is falling
behind in the knowledge economy has increased the attention on higher education’s ability to
produce graduates with skills and abilities for today’s labor market (Burke-1, 2005). Third, the
recent economic downturn has strengthened this call for greater accountability due to the
financial pressure placed on states and legislators’ focus on the rising price of higher education
(Powell, et al., 2012). These three factors combined with the shifting social contract has led to a
strong focus on accountability in higher education and a demand for quantified proof of the
economic returns on the public investment in higher education.
Defining Accountability
Burke (2005) stated that accountability is “the most advocated and least analyzed word in
higher education” (p. 1) and that the term is frequently used but with multiple and often unclear
meanings. Burke (2005) outlined a framework for defining accountability, which states that
accountability imposes six demands on higher education institutions. First, institutions must
exhibit proper use of their powers. Second, they must demonstrate work towards achieving their

19

mission. Third, they must have some method of reporting their performance to the public.
Fourth, they must prove efficient use of the resources they receive and demonstrate effective
outcomes from these resources. Fifth, they must demonstrate quality control. Finally, they must
exhibit a commitment to meeting the public needs.
Additionally, several forms of accountability exist, which further complicates the ability
to reach agreement on what accountability should entail. First, upward accountability is the
traditional reporting structure to a superior and covers what is commonly considered procedural,
bureaucratic or legal accountability (Burke-1, 2005). Examples for higher education institutions
include reporting to a Board of Trustees or reporting for state appropriations hearings. Second,
downward accountability centers on a manager’s responsibility to his or her subordinates (Burke1, 2005). An example in higher education institutions is a president responding to requests from
academic senate. Third, inward accountability emphasizes professional or ethical standards and
in higher education it is demonstrated through professional accountability such as accreditation
(Burke-1, 2005). Finally, outward accountability focuses on responding to external constituents,
stakeholders and the public at large (Burke-1, 2005). In higher education institutions this
includes performance reporting, report cards, and assessments. This literature review focuses on
this last form of accountability and specifically performance funding policies.
Higher education accountability is a fine edged sword. If it is too strict and regulatory
higher education leaders view it as bureaucratic oversight and an infringement of academic
freedom (Burke-1, 2005; Burke & Minassians, 2002; Toman, 2008). On the other hand, if the
goals are overly broad or there is not enough “teeth” to the policies, legislators fear it will be
ineffective (Toman, 2008). Rutherford and Rabovsky (2014) noted that a frequent reason
accountability measures struggle is that policy makers fail to grant freedom for experimentation

20

and innovation in the policy and instead they impose accountability without removing any of the
red tape. To understand this dichotomy it is helpful to explore the history of accountability in
higher education.
History of Accountability
How states hold their higher education institutions accountable has changed over the
years from system efficiency, to educational quality, to productivity, to response to market
demands (Burke-1, 2005). The following is a brief summary of the accountability movement in
the United States since the 1970s. The emphasis on accountability in higher education is
buttressed by society’s shifting view of its social contract with higher education. Historically, the
social contract involved state leaders adequately funding autonomous institutions in exchange for
those institutions serving the common good of the state (Thelin, 2004). However, this began to
shift in the 1970s with state revenue declines due to a falling economy. To create more efficient
higher education systems, states created centralized higher education governing boards and
increased regulation on previously autonomous institutions (Burke, 2005). Additionally, the
public began to question the benefits of higher education as the price of education increased
(Toman, 2008).
In the 1980s states continued to increase regulation over higher education institutions,
this time with a focus on student learning assessment (Burke, 2005). More than 30 states
mandated that higher education institutions create plans to assess student learning, however, they
allowed institutions the autonomy to determine the knowledge students should possess, develop
how to best assess student learning, and decide how to use the results to improve institutional
performance (Burke, 2005). The 1980s also saw the beginnings of a larger shift in the social
contract, with the start of significant discussions over the purpose of higher education (Ewell,

21

1994). These discussions led to viewing higher education less as a public utility and more as a
strategic investment, and in turn, accountability began to focus on the return from that
investment (Ewell, 1994). Despite policymakers’ desire for demonstrated return on their
investment, most institutions focused their assessment programs on internal institutional
improvement and neglected the external demands of demonstrated results from accountability
(Burke & Minassians, 2002).
Due to this neglect, the 1990s began a shift beyond just increased regulation for higher
education institutions to legislators connecting increased performance expectations to financial
support from the state (Burke, 2005). Further, state officials perceived the inability to compare
institutions as a significant flaw in previous accountability models and felt that comparing
institutions was the best way to access performance (Burke & Minassians, 2002). From this
desire to benchmark institutions against each other, came the rise in prominence of annual
rankings of institutions such as U.S. News and World Report rankings, which are still prevalent
today (Toman, 2008). This time period also saw the creation of performance funding policies in
many states as legislators sought to press for demonstrated results and the ability to compare
institutions (Burke & Minassians, 2002).
The turn of the century again saw declining state revenues from another recession, which
in turn led to further cuts to higher education (Burke, 2005). Accountability policies shifted to
prioritizing work force needs and other state priorities over academic concerns with state
policymakers seeking to influence the behavior of institutions in order to align institutional
performance with state goals (Burke, 2005; McLendon, Hearn & Deaton, 2006). Within a period
of three years, reports by NGA, NCSL, and SHEEO all advocated for higher education policies
that focused on state needs and economic competitiveness (NGA, 2007; NCSL, 2006; SHEEO,

22

2005). The renewed push for accountability was spurred on by the public’s frustration with the
quality of student learning, higher education’s focus on research and graduate education, the
expanding number of administrative staff on campus, low completion rates and limited
employability of graduates (Burke, 2005). Further, legislators went beyond a push to compare
institutions and began to advocate for the standardization of higher education, believing this
would aid the comparability of institutions and enhance the measure of quality (Eaton, 2012).
The rising price of higher education coupled with the erosion of public trust has renewed
the accountability movement in recent years (Volkwein & Tandberg, 2008). Within this
resurgence, organizations such as the Gates Foundation and Lumina Foundation have
emphasized student completion and success (often defined as employment and salary) post
graduation. To this end, accountability has taken a market orientation focused on economic
outcomes and efficiency (Burke, 2005; McMahon, 2009; Toman, 2008). Simply defined,
efficiency is the measure of outputs over inputs. State leaders desire greater efficiency from
higher education and therefore advocate for institutions to increase outputs, decrease costs, or
both. In higher education policy, efficiency models emphasize a narrow view of outputs with a
focus on market-based benefits (e.g., graduation rates, salary post-graduation, faculty research
productivity) (McMahon, 2009; Slaughter & Rhoades, 2004). Performance funding policies are
an example of this push for efficiency and legislators are once again advocating for and
implementing these policies. The renewed interest in performance funding is to be expected
within the current focus on efficiency as many metrics used in these policies focus on efficiency
(Burke & Serban, 1998). In line with a market orientation of higher education, state legislators
have also decreased financial support for higher education on the premise that higher education

23

is a private good and therefore those who directly benefit from the education (the students)
should pay a greater share of the costs of higher education (Slaughter & Rhoades, 2004).
Forms of Accountability
Forms and measures of accountability have progressed and expanded as the
accountability movement has evolved. However, no clear agreement on the best accountability
measures exists. The following are all mentioned as primary accountability measures or key
principles of accountability: accreditation, rankings, licensure, alumni studies, academic program
review, outcomes assessment, faculty productivity measures, student unit data systems, college
spending measures, rankings such as U.S. News and World Report, end of program assessments
such as the Collegiate Learning Assessment, voluntary assessment measures such as the College
Portrait or the Access to Success initiative, and state report cards (Bogue & Hall, 2002; Carey &
Schneider, 2010; Toman, 2008)
Accreditation is the oldest and most well known form of higher education accountability.
Its central role of providing quality assurance for higher education has existed for over a century.
Accreditation is the primary method for internal accountability and is a process based on self
review and peer assessment which covers an institution’s financial health, governance structure,
faculty and staff relations, achievements, student services and student learning outcomes
(Volkwein, 2009). Mirroring the trends in accountability as a whole, since 1990 accreditation
has seen a growing focus on outputs and outcomes including graduation rates (Volkwein, 2009).
However, such a focus on end of the line results takes away from the whole educational process
and may lead to institutions missing key information that could improve student learning
(Volkwein, 2009). Some accreditation organizations have recently taken steps to alleviate this

24

concern and encouraged institutions to focus on data across the whole institutional timeline
(Volkwein, 2009).
Push for External Accountability
Despite accreditation’s long history, policy makers are pushing for a decreased focus on
internal monitoring and a need for greater external accountability. Volkwein (2009) noted the
struggle accreditation faces between the internal role of self-assessment for improvement and the
external role of accountability to taxpayers, legislators, and students. Legislators fear that
institutions would not be willing to highlight their areas of weakness publically and therefore
believe external oversight is needed. Two main reasons are given for the need for increased
external accountability.
First, institutions are not viewed as credible judges of their own quality and it is believed
that external assessment is needed. The changing social contract has led to an attack on the
current system of accreditation with calls for its overhaul or even complete removal (Eaton,
2012). Accrediting organizations have been around since the early 1900s, but it was only since
the 1950s that they took on the gatekeeping role of determining academic quality for the federal
government in order for institutions to be eligible for federal funds (Eaton, 2012). In recent
years, a belief has emerged that higher education institutions should be held to greater
accountability by the government and this new process has included a more regulatory and
market focus than the traditional collegial model of accreditation (Eaton, 2012). For instance, the
2008 reauthorization of the Higher Education Act included governmental control in areas
previously left entirely to institutions, such as transfer of credit, student achievement, and
distance learning (Eaton, 2012). The results of the accountability movement’s attacks on
accreditation can already be seen in the increased focus throughout the process on detailed

25

outcome measures and a high level of concern around costs and productivity (Bardo, 2009). Part
of this lack of trust in accreditation arises from a perceived secrecy in the process. Regional
accreditation groups are criticized for keeping their discussions private and limiting the amount
of information shared with external stakeholders such as parents and students (Bardo, 2009).
Critics of the current system feel these conversations should be public and that stakeholders
should know the results and changes that arise from an accreditation cycle so that customers
(students and their parents) can make informed decisions (Eaton, 2012).
A second reason given for the need for external monitoring is a discrepancy between how
institutions measure and view quality versus how external stakeholders, in particular state
legislators, measure quality and impose accountability. Institutional leaders tend to measure
success through a resource/reputation model and represent professional accountability (Burke,
2005; Volkwein, 2009). In this structure, an institution’s quality is measured through its
rankings, level of financial resources, faculty credentials, student achievement and fundraising
using methods such as peer review, internal institutional improvement, reputation and rankings
(Burke, 2005; Volkwein, 2009). Policymakers on the other hand focus on a strategic investment
model of success characterized by political accountability (Burke, 2005; Volkwein, 2009). In this
model, quality is measured through a return on investment, efficiency, and productivity of
outputs using methods such as external regulation and performance reporting, budgeting and
funding (Burke, 2005; Volkwein, 2009). Within these competing frameworks, state legislators
often feel that higher education institutions are focusing on the wrong things and wasting
taxpayers’ money, while institutions feel that legislators do not understand higher education’s
structure and values.

26

Performance Funding
The push for external accountability and a focus on efficiency has led to many states
adopting performance funding policies. As states pass performance funding policies, the
literature on the policies has also increased. Literature on performance funding policies has
focused on several areas including defining performance funding, analyzing the spread of
performance funding, identifying concerns with performance funding, recognizing best practices
for performance funding policies, and exploring the effectiveness of performance funding
policies. Each of these topics is explained in detail below.
Defining Performance Funding
Performance funding is part of a growing trend in accountability, which also includes
performance based budgeting and performance reporting. It is important to understand the
similarities and differences between these three accountability measures. All three are
implemented with the purpose of demonstrating external accountability, improving institutional
performance and responding to state needs (Burke & Minassians, 2002). However, the three
differ in the how closely, if at all, institutional performance is tied to state funding. Performance
reporting requires institutions to report their performance on various metrics, but does not tie any
funding to how institutions perform on these metrics. It stems from the belief that institutions
will perform better when they know their outcomes will be made public (Burke, 2005).
Legislators specify a list of metrics that all public institutions need to report annually (Burke &
Minassians, 2002). These reports go to governors, legislators and higher education coordinating
boards, as well as being posted on institution or system wide websites (Burke, 2005).
Performance budgeting allows funding entities to consider performance on various metrics as
one factor when deciding funding amounts, but there is no specific formula connecting funding

27

to performance. In other words, an institution receives its determined state funding allocation
and any additional funds for performance are solely determined by the judgment and discretion
of state legislators (Burke, 2005). Finally, performance funding policies connect specific
performance outcomes on state selected measures to increase state funding (Goldstein, 2012).
The key factor that separates performance funding from performance budgeting is that the
financial relationship is predetermined and specific. A particular score on a metric equates a
specific funding amount.
Through performance funding policies, institutions are held accountable for meeting
outcome targets as part of the state appropriations process. Performance funding policies allow
state officials to emphasize metrics such as graduation rates, financial stewardship, credit hours
earned, wages of graduates, and research grants that they believe will better incentivize
efficiency. Performance funding stems from a resource dependence framework. Legislators
believe that institutions will strive to improve their scores on the various metrics in order to
maximize their funding.
Besides maximizing funding, legislators promote numerous benefits for institutions from
performance funding. State leaders suggest that performance funding policies can provide
administrators with information about their performance as well as how they compare to their
peers (Dougherty & Reddy, 2013). Additionally, state leaders promote the fact that performance
funding focuses on outcomes instead of prescribing behaviors, therefore allowing institutions to
maintain autonomy and decided how they will respond to the policy mandates (Burke &
Minassians, 2002; Rutherford & Rabovsky, 2014). Finally, performance funding policies, with
their formalized and explicitly defined performance metrics, make clear to higher education

28

institutions the values and preferences state legislators will reward and value (Rutherford &
Rabovsky, 2014; Sanford & Hunter, 2011).
Metrics used in performance funding are connected to state priorities and therefore vary
from state to state and even over time in the same state. However, some metrics are commonly
seen across performance funding policies. Burke and Serban (1998) examined metrics across
policies in 1998 and found that the following metrics were most common: graduation and
retention rates, job placement, student transfer rate, faculty workload, and time to degree.
Performance funding policies claim to shift funding (and therefore higher education’s focus)
from input (resources received) and process measures (methods used to deliver services) to state
focused outcome measures (quantity or quality of the products produced) (Burke, 2005).
However, Burke and Serban’s (1998) study also indicated that across all metrics only 12%
focused on outcomes and only 19% on outputs, with 48% focused on process. It should be noted
that this study was done on older performance funding policies, and new research should explore
if this trend is still present in current policies.
Performance funding is often thought of as occurring in two waves, frequently referred to
as performance funding 1.0 and performance funding 2.0. The adoption of performance funding
1.0 took place between1979 (when Tennessee implemented the first policy) and 2000
(Dougherty et al., 2014b). In this version, performance funding is a bonus over regular state
funding and involves small amounts of funding, usually between 1 and 6 percent (Dougherty et
al., 2014a). In performance funding 2.0, starting around 2007 and continuing to the present, the
funding is embedded into the normal state funding for higher education (Dougherty et al.,
2014b). About two thirds of the programs created in the second wave are readoptions of
previously discontinued policies (Dougherty et al., 2014a). The amount of funding in 2.0

29

versions is also usually higher, with between 5 to 25 percent of all state higher education funding
coming from performance funding formulas (Dougherty et al., 2014a).
Spread of Performance Funding
As of July 2015, 32 states have active performance funding policies and five more have
programs in the process of being implemented (NCSL, 2015). Figure 1 shows the states with
performance funding and those in transition. Authors attribute the spread of performance funding
and its fellow accountability measures to diverse factors, including: globalization, pressure to
maximize productivity and efficiency, a shift towards marketization of higher education and
governance, financial pressures from the great recession, changes in political leadership, and
growing frustration with the voluntary and self assessment of higher education (McLendon et al.,
2006; Dougherty et al., 2014b). Several studies have empirically explored the spread of
performance funding. McLendon et al., (2006) conducted an event history analysis model to
explore what sociopolitical factors influence a state’s adoption of performance funding,
performance based budgeting or performance reporting policies across all states. Of the ten
areas they explored, they demonstrated that legislative party strength and higher education
governance arrangements were the primary drivers. For performance funding specifically, a
higher percentage of Republican state legislators and the absence of a consolidated governing
board increased the probability of the state adopting a performance funding policy. The authors
believe that Republicans favor performance funding over performance budgeting or performance
reporting as it grants legislatures the strongest leverage for accountability (McLendon et al.,
2006). Beyond Republican state legislators, governors who identify as Republican also seem to
favor performance funding policies. Since 2007, nine states with Republican governors have
implemented performance funding policies (Dougherty et al., 2014a). Of note, despite the

30

regional nature and spread of performance funding, McLendon et al. (2006) found no evidence
that having neighboring states with performance funding increased the likelihood of a state
adopting a performance funding policy.
Figure 1 States with Performance Funding Policies Active and in Transition on July 2015

Source: National Council of State Legislatures
Dougherty et al., (2014b) examined how political forces influence the adoption of
performance funding through a qualitative analysis in three states (Indiana, Ohio, and
Tennessee). Through interviews with state legislators, higher education administrators, state
governing board members, and business leaders, the authors determined that the main supporters
of the adoption of performance funding policies were the governor, the state higher education
coordination board and to a lesser extent the legislative leaders and business leaders (Dougherty
et al., 2014b). This advocacy coalition arose from the shared beliefs that states needed to
increase the number of residents with college degrees in order to meet economic demands, that
the recession required higher education to become more efficient, and that a performance
funding policy would allow the previous two goals to be achieved (Dougherty et al., 2014b).

31

The role of foundations in policy spread and formation is well documented, and
foundations played a significant role in the spread of performance funding. The Lumina
foundation is a strong supporter of performance funding (Dougherty et al., 2014b) and has
funded many initiatives focused on enhancing or creating performance funding and other
advances in efficiency. Complete College America (CCA) is another strong supporter of
performance funding (Tandberg & Hillman, 2014). In 2009 CCA recommended a series of
reforms for higher education, of which one suggestion is implementing performance funding
(CCA, 2012). Thirty-five states had signed on to these reforms (though not necessarily
implemented performance funding) by 2013 (CCA, 2016).
While performance funding is increasing in prevalence, it has a tumultuous past. Twothirds of performance funding policies that are implemented are later discontinued for at least
some period of time (Dougherty et al., 2014a). In fact, of the 18 states that adopted performance
funding between 1979 and 2012 only a few maintained the policy for more than four consecutive
years (Rutherford & Rabovsky, 2014). Researchers discussed three main forces contributing to
the closure of performance funding polices: opposition from higher education institutions who
felt they were not consulted in the process and had significant concerns with the policies,
economic recessions and the loss of political champions (mainly through election losses or term
limits) (Dougherty, Natow & Vega, 2012; Rabovsky, 2012; Rutherford & Rabovsky, 2014).
Despite performance funding’s spread across states, the policy seems to struggle to be
disseminated across institutions. Senior administrators are familiar with the policy, however
lower level administrators and faculty seem to not be aware of the specifics of the policy (Burke,
2002; Burke, 2005; Dougherty et al., 2014a; Freeman, 2000; Hall, 2000; Schaller, 2004).
Relevant research has attributed this knowledge stratification to two causes. First, there is a

32

perception of role responsibility, with high level administrators seeing their role as responsible
for the institutional implementation and accountability for performance funding, while not
feeling it needs to be the concern of those below them on the campus hierarchy (Freeman, 2000;
Hall, 2000). Second, since performance funding usually goes into the university general fund,
there is a perception that only those connected to and responsible for the general fund (usually
high level administrators) should be concerned with performance funding metrics (Freeman,
2000; Hall, 2000).
Concerns with Performance Funding
Despite performance funding’s spread, several concerns exist including what data is used
in the policies, how to account for institutional differences, and how to incentivize institutional
behavior. First, higher education leaders hold significant reservations regarding the
appropriateness of various metrics used in performance funding policies (Dougherty & Reddy,
2013; O’Neal, 2007; Opoczynski, 2014; Sanford & Hunter, 2011). Most notably, there is great
concern regarding the measures of graduation and retention, which, due to the nature of how
IPEDS measures these metrics, offer only a limited slice of the students attending four year
institutions and completely ignore transfer and part time students (O’Neal, 2007; Opoczynski,
2014). Further, there is debate about how the data should be measured and analyzed
(Opoczynski, 2014; Sanford & Hunter, 2011). For instance, should the metric count the raw
number of students graduating or use graduation percentages, and should institutions be
compared to other institutions in the state, to their peers, or to their historical performance
(Opoczynski, 2014; Sanford & Hunter, 2011)? Additionally, concerns exist regarding what
measures (graduation rates, retention, course completion, etc.) best measure student success
(Sanford & Hunter, 2011). As indicated above, defining success in higher education is difficult

33

and different performance funding policies utilize different definitions and metrics to measure
student success.
Another concern regarding performance funding policies is how to account for
institutional differences (Hall, 2000; O’Neal; 2007). Higher education administrators often feel
that performance funding fails to understand the various missions of higher education institutions
and their varying student demographics (Hillman, Tandberg & Fryer, 2014). During
implementation, higher education institutions, especially open access institutions, frequently
express reservations about the performance funding policies including worry about the policies
understanding their mission (Dougherty et al., 2014b). Further, Rutherford & Rabovsky (2014)
argue that performance funding policies try to apply a one size fits all mentality to institutions
and that the policies ignore many contextual factors, which may influence an institution’s ability
to respond to the performance funding mandate.
A final concern with performance funding is whether the limited funding in many states
can actually incentivize institutions to change their behavior (CCA, 2012; Dougherty & Hong,
2006; Dougherty & Reddy, 2013; Jenkins et al., 2012; Opoczynski, 2014; Rutherford &
Rabovsky, 2014; Sanford & Hunter, 2011; Shin, 2010). In many states, the portion of funding
tied to performance metrics is less than 5% (Sanford & Hunter, 2011). Such small amounts may
not be viewed as meaningful incentives to institutions, especially in relation to other possible
revenues (CCA, 2012; Sanford & Hunter, 2011). In a recent analysis of Michigan, where
performance funding accounts for just 2% of state funding, many administrators commented that
they followed the performance funding metrics to ensure they were in compliance, but that the
possible additional funds did not alter their behavior significantly (Opoczynski, 2014). While
there is agreement that small amounts of funding may not change institutional behavior, there is

34

no clear consensus on an amount that would be viewed as “enough.” In fact, when Tennessee
doubled the money allocated via performance funding in order to increase the incentive for
institutions, there was no change in institutional retention rates (Sanford & Hunter, 2011).
Further, performance funding policies assume that institutions are intentionally not performing at
optimal efficiency or have misaligned priorities and are simply waiting for financial motivation
to change their performance (Rutherford & Rabovsky, 2014). However, Opoczynski (2014)
discussed that administrators stated that the policy did not alter their behavior tremendously
because the behaviors the performance funding policy promoted, such as increased graduation
rates, were values the institution already held and were striving to improve.
Best Practices in Performance Funding
Various researchers and policy organizations have promoted several principles believed
to lead to successful implementation of performance funding, including how to strive to
overcome the concerns indicated above. Several are outlined in this section. The first key
practice leading to successful enactment of a performance funding policy is agreement and
consultation on goals before implementing the policy. CCA (2012) suggested the importance of
creating a clear and concise public agenda that articulate the needs of the state and how
performance funding will address these goals. CCA (2012) also promotes including institutional
representatives throughout the process due to their expert knowledge and the increased chances
of support for the final policy if the representatives are involved from the start. Engaging
institutional leaders could also ensure campus priorities’ are included in the policy and metrics
(Burke, 2005). Dougherty and Reddy (2013) noted that in performance funding 2.0 programs
(which are largely influenced by CCA) higher education administrators are consulted more
frequently as the policy is crafted and that this has proven useful for a smooth implementation.

35

Another beneficial practice is to allow institutions to familiarize themselves with the
metrics and how they would fair in performance funding with limited or no financial impacts
(CCA, 2012; Dougherty et al., 2014b). Two possible structures to accomplish this are a free year
or to phase in the policy over a period of a few years. A free year allows institutions to gain
information on where they fall on the metrics without any financial impacts. Institutions can use
this year to ask any questions they may have on the policy and to make changes to prepare for
performance funding. Alternatively, the policy can be phased in with a small amount at the start
(such as 2%) and increasing each year till the final desired amount is reached (CCA, 2012).
It has also been recommended to take institutional mission into account (CCA, 2012;
Dougherty et al., 2014b). While Burke (2005) believes that performance funding will never be
able to fully account for diverse institutional needs, several options could make steps towards
including institutional mission. First, the performance funding policy could specify different
metrics for different institutional types (CCA, 2012). For instance, Ohio has separate formulas
for their flagship institutions compared with their regional campuses (Dougherty et al., 2014b).
Similarly, institutions could be compared to their peer institution instead of against other state
institutions (Dougherty et al., 2014a). Michigan’s performance funding policy is structured this
way and institutions score points based on how they compare to institutions in their Carnegie
classification (Opoczynski, 2014). Additionally, an institution’s performance could be compared
to itself across time. This would allow performance funding to reward any positive growth or
improvement. Finally, institutions could be allowed to create a “point of pride” metric, which
would allow each institution to have a personalized metric related to their mission (Burke, 2005).
Another recommendation researchers and policy organizations endorse is including
language or policies that protect academic standards and admissions of underserved students

36

(CCA, 2013; Dougherty et al., 2014a; Kelly & Lautzenheiser, 2013). Specifically, an option is to
have a weighted formula where “at risk” students are weighted more heavily (Dougherty et al.,
2014b). States that have weighted metrics often give significant weight to these metrics with
numbers from 40% (Tennessee) to 100% (Texas) (CCA, 2012). Adding this extra weight to at
risk students helps reduce the likelihood that institutions will simply reduce the enrollment of at
risk students to improve their performance on the funding metrics.
It is further recommended that legislators include only a small number (4 or 5) of explicit
and easy to understand metrics that are clearly connected to state priorities and labor market
needs (Burke, 2005; CCA 2012; Kelly & Lautzenheiser, 2013). Limiting the number of metrics
helps ensure that performance funding aids institutions in aligning their goals with state
priorities. With only a few clearly defined metrics, policymakers should be able to articulate to
institutions what the state priorities are and what is expected from the institutions (CCA, 2012).
Having too many metrics can muddle the message regarding state priorities and also potentially
create instances where focusing on improving performance on one metric could negatively affect
performance on another metric (Burke, 2005).
As a large concern with performance funding is the volatility it can create regarding
institutional budgeting, several best practices are suggested to offset these concerns. First,
institutions could include a stop loss provision with would place a limit on how much funding an
institutions could lose (CCA, 2012; Dougherty & Reddy, 2013). This establishes a maximum
loss that an institution could face in any one year. Another option is to calculate performance
data on a three year average in order to avoid sharp shifts in funding (Dougherty et al., 2014b).
Likewise, institutions should be rewarded for any improvement, instead of only being rewarded
if they reach a specific level of performance (CCA, 2012).

37

Effectiveness of Performance Funding
Performance funding advocates believe that following the best practices outlined above
will increase the effectiveness of performance funding policies. Dougherty and Reddy (2013)
describe three levels of impacts of performance funding that institutions must progress through
to see a positive effect of the policy. First, immediate impacts of performance funding are things
such as institutions becoming aware of state priorities and their own performance relative to
these priorities. Next, these immediate impacts must translate into intermediate institutional
change such as modifications of institutional policies, programs, and practices that are designed
to increase performance on state metrics. Finally, these intermediate impacts must result in
changes to ultimate student outcomes such as increased graduation rates. It is these ultimate
outcomes that performance funding policies seek to change. Each of these levels of impacts will
be explored in this section.
Immediate impacts. In their meta analysis, Dougherty and Reddy (2013) discussed that
several studies (mostly in the two-year sector) demonstrated that performance funding policies
did make higher education administrators more aware of state priorities as well as led to a greater
awareness of institutional performance in relation to the state priorities. Further, they found
evidence that performance funding policies increase competition amongst the institutions
(Dougherty & Reddy, 2013). Limited research has explores these changes, with most studies
exploring the effectiveness of performance funding focusing on the next two levels of impacts.
Intermediate institutional change. Some research on performance funding shows that
higher education institutions change their academic and student service policies, programs and
practices when performance funding policies are put in place. For instance, some colleges
reported closing programs with low graduation rate or job placement success (Dougherty et al.,

38

2014a). Institutions also reported changing course content and instruction as well as course
sequences to try and increase course completion rates (Dougherty & Reddy, 2013). Further,
some institutions reported increasing advising, changing retention programs, registration
procedures and student aid in order to increase retention and graduation rates (Dougherty &
Reddy, 2013). These included changes to the reregistration process, changes in financial aid,
improvements in academic advising, and changes to job placement supports (Dougherty &
Reddy, 2013).
Rabovsky (2012) explored intermediate institutional change through a national
quantitative analysis and demonstrated that performance funding policies were positively related
to expenditures on instruction. However, the effects were minimal with the difference between
states with performance funding and those without being less than one percent. When separated
into research universities and non research universities, the effect was more than double for
research universities (1.34 percent verse 0.59 percent) yet still small for both (Rabovsky, 2012).
However, this finding is noteworthy as it highlights the fact that performance funding policies
may have differential impacts based on institutional classification.
Ultimate student outcomes. When exploring whether performance funding policies
increase outcomes such as retention and on time graduation rates the results are more mixed.
Some studies have shown that the number of graduates increases after the implementation of
performance funding (Bell, 2005; Phillips, 2002). However, these studies do not control for the
numerous factors besides performance funding (e.g., increased enrollments, changes to financial
aid) that could lead to an increase in graduation rates (Dougherty et al., 2014a).
Most multivariate studies with extensive controls for other potential causes of graduation
or retention rate increases find little evidence of performance funding policies increasing

39

graduation or retention rates in four year institutions (Rabovsky, 2012; Rutherford & Rabovsky,
2014; Sanford & Hunter, 2011; Shin 2010; Shin & Milton, 2004; Volkwein & Tandberg, 2008).
Two possible exceptions are difference in difference based studies done by Hillman, Tandberg,
and Gross (2014) and Tandberg and Hillman (2014) which found a positive effect of
performance funding on graduation rates in some of their models. In an in depth analysis of
Pennsylvania’s performance funding policy, Hillman, et al. (2014) found performance funding to
have a positive impact in three of their ten models. These effects were noted in the models
comparing performance funding in Pennsylvania to neighboring states, however when compared
to similar states using coarsened exact matching, the authors found no effect of performance
funding (Hillman, et al., 2014).
In a national study which compared states with performance funding to control states
which did not have performance funding, Tandberg and Hillman (2014) demonstrated that
performance funding policies did have a small effect on graduation rates seven, eight, and eleven
years after implementation. The magnitude of these impacts are small, but suggest that
performance funding may have delayed effects or that performance funding may become more
effective over time. However, Rutherford and Rabovsky (2014) conducted a fixed effects time
series analysis with panel corrected standard errors and found that graduation rates may actually
decline over the length of time that performance funding policy is in place. When separated out
between performance funding 1.0 and 2.0 policies, there is a negative relationship between
policy implementation and graduate rates for performance funding 1.0 policies. For performance
funding 2.0 policies, the effect was positive, although the results were not statistically
significant. Other authors have not found any difference based on length of policy

40

implementation (Shin, 2010). More research needs to be conducted to explore the relationship
between length of policy implementation and student outcomes.
In fact, not only has much research failed to demonstrated significant results from
performance funding policies, but some research has shown that the variance in higher education
performance may be best accounted for by variables outside the control of institutions or even
states. Volkwein and Tandberg (2008) demonstrated that in all categories of Measuring Up
(state report cards that measure the performance of higher education in the state) data from 2000
to 2006, state characteristics over which policy makers have little control, such as demographics
and economic conditions, accounted for the greatest share of variance in scores on Measuring Up
data. In other words, things that the state can control such as higher education governance
structures or accountability policies have limited effect, while things beyond their direct control,
such as economic conditions and demographic trends, are the strongest predictors of higher
education performance. In fact, one study found that institutional characteristics account for 76%
of the variance in graduation rates and that state characteristics (of which they included
performance funding as one variable) only account for 15% of the variance (Shin, 2010). This
finding suggests that institutional fixed effects may account for differences in graduation rates.
A few studies (e.g. Rabovsky, 2012; Shin, 2010) explore the effect of performance
funding on research productivity such as grant funding received or faculty research produced.
Since the scope of this study does not address research productivity, they will not be discussed in
this literature review. However, it is worth noting that literature on this topic shows limited
success of performance funding policies in this outcome as well.
Unintended outcomes. Performance funding policies focus on efficiency, which creates
a quasi market where keeping costs low while improving outcomes such as graduation rates is

41

valued over other principles. Due to the iron triangle of cost, quality and access (Immerwahr et
al., 2008), this could create some unintended outcomes related to access. Many authors noted a
concern that performance funding policies could create unintended outcomes, in particular
institutions favoring better prepared students and closing access to other students (Bell, 2005;
Colbeck, 2002; Dougherty et al., 2014b; Lahr et al., 2014; Naughton, 2004; Opoczynski, 2014;
Rutherford & Rabovsky, 2014). For instance, higher education administrators are concerned that
performance funding policies would result in “creaming” or that open access institutions would
lose funding (Dougherty et al., 2014b). Other higher education administrators suggested the
possibility of weakened academic standards and restriction in admissions of underprepared
students as a result of performance funding (Lahr et al., 2014). However, it was only mentioned
as a possibility with no administrators stating it had been an intentional decision at their
institution.
Research exploring whether performance funding policies result in these unintended
outcomes is limited. Kelchen and Stedrak (2016) explored whether performance funding affects
colleges’ financial priorities and found that institutions subject to performance funding may
change their enrollment patterns. Specifically, institutions subject to performance funding
received about $30 less per FTE in Pell Grant revenue than colleges in states without
performance funding. Similarly, Umbricht, Fernandez, and Ortagus (2016) examined any
unintended outcomes from performance funding in the state of Indiana. Utilizing a difference in
difference technique, they found that performance funding changed enrollment patterns and
increased selectivity. Specifically, they found that institutions in Indiana admitted students with
higher ACT scores after performance funding implementation when compared with both similar
states and Indiana’s private institutions. Additionally, the total number of minority students

42

enrolled after performance funding was lower than both institutions in similar states and private
institutions in Indiana.
Enrollment and Retention Literature
In order to address the question of whether performance funding changes the enrollment
profile of underrepresented students over time, it is important to understand what other factors
may account for changes in the enrollment and retention of students. The vast majority of
research exploring the enrollment and retention of students in higher education takes a student
level perspective, exploring factors such as institutional fit, academic readiness, or ability to pay.
While the literature from the student perspective is vast, this research study explores retention
and graduation rates from the institutional and state perspective and therefor this literature
review will focus on these levels.
General Trend Across Race
Higher education enrollment and graduation rates vary considerably across race. Gaps in
college enrollment after high school graduation across race have narrowed over time, yet
significant differences in college attendance still exist (Baum, Ma, & Payea, 2013). In 2014,
68% of White students enrolled in college directly after high school compared to 63% of Black
students and 62% of Hispanic students (NCES, 2016). While these numbers appear similar, they
mask the different enrollment patterns of students by race. Specifically, when limited to
enrollment in bachelor’s degree programs, 39% of students enrolled compared to 24% for Black
students and 18% for Hispanic students. Further, significant differences in high school
graduation rates exist across race resulting in 44% of all White 18 to 24 year olds being enrolled
in higher education compared to 36% for Black young adults and 31% for Hispanic 18 to 24 year
olds (Baum, et al., 2013)

43

Though race discrepancies in preparation for higher education are well documented, it
does not explain the growing racial stratification in higher education (Carnevale & Strohl, 2013).
Black and Hispanic students still earn degrees at lower rates than their equally performing peers.
For instance, 57% of Black students and 56% of Hispanic students who scored at least a 1200 on
the SAT earned certificate, associate or bachelor’s degrees compared to 77% of similar
performing White students (Carnevale & Strohl, 2013). Looking at the next performance group
(those who scored between 1000 and 1200 on the SAT), 47% of Black students and 48% of
Hispanic students earned some form of degree compared to 68% of similar performing White
students (Carnevale & Strohl, 2013).
The racial gap continues throughout the higher education pipeline and large gaps exist in
persistence and completion rates across race. Persistence rates are much higher for Asian
American students and White students (69% and 59% respectively) than Black students and
Hispanic students (46% and 49%) (Chen & St. John, 2011). Graduation rates show similar
results, with National Education Longitudinal Study (NELS) data demonstrating that White
students had a graduation rate of 70% compared to 52% for Black students and 49% for Hispanic
students (Carnevale & Strohl, 2013). When looking at odds ratios, the discrepancies continue.
For Black students, the odds of graduating were 70% compared to White students and for
Hispanic students odds of graduating were 74% compared to White students (Titus, 2006).
Similarly, the odds of dropping out were 1.23 times greater for minority students than White
students (Rhee, 2008).
These inequities result in large gaps in bachelor degree holders aged 25-29. In 2015,
only 21% of Black and 16% of Hispanic 25 to 29 year olds had a bachelor’s degree compared to
43% of Whites in the same age group (NCES, 2016). There is farther stratification in degree

44

attainment by gender with males having lower rates of degree attainment across all races.
Specifically, 11% of Hispanic men, 17% of Hispanic women, 16% of Black men and 24% of
Black women having a bachelor’s degree, compared to 43% for White females and 35% for
White males have attained a bachelor’s degree (Baum, et al., 2013).
General Trend Across SES
Higher education enrollment and graduation rates also vary significantly across income
and SES. Recently, enrollment rates for middle and upper income students have increased, while
enrollment rates for low income students have stagnated, resulting in widening higher education
participation rates by income (Baum, et al., 2013). In 2014 81% of high school graduates from
higher income families went straight to college post high school compared to 64% from middle
income families and 52% from low income families (NCES, 2016).
Higher education completion rates also vary by income quartile. For students who began
higher education in 2003, 58% of students from the highest income quartile had earned a
bachelor’s degree six years later, compare to 26% for those from the lowest income quartile
(with the second and third quartiles having rates of 35% and 44% respectively) (Baum, et al.,
2013). Many factors may be at play creating the stratification by income, however several
studies have demonstrated that after controlling for individual and institutional effects, SES has a
statistically significant impact on retention and graduation (Chen, 2011; Chen & St. John, 2011;
Sparks & Nunez, 2014). For instance, the odds of dropping out for high SES students were 62%
that of low SES students, and the odds for mid SES students were 81% that of low SES students
(Chen, 2011). Similarly, high SES students had significantly higher persistence rates (71%) than
low SES students (44%) and mid SES students (57%) (Chen & St. John, 2011).

45

Institutional Factors
Research has shown that a wide variety of institutional factors may influence enrollment
and retention rates. The three main categories for institutional factors that influence retention
and enrollment rates are student demographics, structural characteristics and financial
characteristics. Each of these is described in turn below.
Student demographics. Several aspects of overall student demographics are believed to
influence institutional graduation rates including the socioeconomic and racial diversity of the
campus as well as the percent of students with non-traditional attributes. The percent of students
who receive Pell Grants (a measure of socioeconomic diversity) at an institution is a strong
predictor of institutional graduation rates (The Advisory Committee on Student Financial
Assistance [ACSFA], 2013; Pike & Graunke, 2014; Titus, 2006). As the percent of students who
receive Pell Grants increases, the six-year graduation rate decreases from a high of 80% for four
year institutions with less than 20% Pell Grant recipients to a low of 25% for four year
institutions with more than 80% of their students receiving Pell Grants (ACSFA, 2013). In fact,
the percent of students who receive Pell Grants accounted for 55% of the variance in graduation
rates (ACSFA, 2013). Figure 2 shows six year graduation rates by percent of Pell recipients for
four-year public institutions in 2010. Additionally, institutional average SES had a significant
positive effect on graduation rates beyond the individual effects of student SES on graduation
rates (Titus, 2006).
Research has also suggested that the percent of students who identify as minorities on
campus negatively influenced retention and persistence rates (Heck, Lam, & Thomas, 2014; Pike
2013; Rhee, 2008; Ryan, 2004; Scott, Bailey & Kienzl, 2006; Sparks & Nunez, 2014; Titus,
2006). When exploring the impact of structural diversity on students who left the institution

46

separated out by those who stopped out, those who transferred to a different institution, and those
who dropped out of higher education completely, structural diversity was significantly related to
stop out, but not transferring to another institution or leaving higher education completely (Rhee,
2008). Yet this finding is not universal; some studies have found that the effect between
proportion of underrepresented minority students and retention or graduation rates was not
statistically significant (Chen, 2011; Pike & Graunke, 2014).
Figure 2 Six Year Graduation Rates for Four Year Institutions by Percent of Pell
Recipients

Six	
  YEar	
  Graduation	
  Rate	
  

100	
  
90	
  
80	
  
70	
  
60	
  
50	
  
40	
  
30	
  
20	
  
10	
  
0	
  
0-­‐20%	
  

20-­‐29%	
  

30-­‐39%	
  

40-­‐49%	
  

50-­‐59%	
  

60-­‐69%	
  

70-­‐79%	
   80%	
  and	
  
above	
  

Percent	
  of	
  Students	
  with	
  Pell	
  Grants	
  	
  

Source: Advisory Committee on Student Financial Assistance
Finally, research also shows that other student demographics significantly influence
graduation rates. The percent of students who are part time verse full time is related to retention
and graduation rates. Several studies have stated that the proportion of students who were full
time had a positive effect on retention and graduation rates (Pike, 2013; Scott, et al., 2006). Pike
and Graunke (2014) also found that the percent of part time students was negatively related to
retention rates, however when they accounted for the unobserved heterogeneity through a fixed
effect model, it was no longer statistically significant. Additionally, the proportion of students

47

living on campus is positively related to retention and graduation (Ryan, 2004; Scott, et al.,
2006). The proportion of students who were non traditional aged had a negative effect on
retention and graduation rates (Pike & Graunke, 2014; Scott, et al., 2006). However, Pike and
Graunke (2014) indicated that the proportion of non traditional aged students was significant in
one of their models, but not the other. This could be from using very strict T-values, and if more
traditional T values had been used than age would have been statistically significant in both
models (Pike & Graunke, 2014).
Structural characteristics. Research also shows that institutional structural
characteristics such as size, classification and selectivity have an effect on institutional retention
and graduation rates. The research on size is inconclusive with several studies finding it has a
significant positive effect on retention and graduation rates (Pike 2013; Pike & Graunke, 2014;
Ryan, 2004) and several others finding no effect (Chen, 2011; Rhee, 2008; Titus, 2006).
Research has also consistently suggested that after controlling for student characteristics, private
institutions have higher retention and graduation rates than public institutions (Ryan, 2004; Pike
2013). Additionally, several studies have mentioned that classification or mission influences
retention and graduation rates (Heck, et al., 2014; Pike, 2013; Shin, 2010). Specifically, doctoral
universities have higher retention and graduation rates than baccalaureate institutions (Heck, et
al., 2014; Pike, 2013) but master’s institutions having lower rates than baccalaureate institutions
(Pike, 2013).
Studies have also demonstrated that institutional selectivity (usually measured by student
average SAT/ACT scores or through Barron’s selectivity index) is a significant predictor of
retention and graduation rates. Several studies noted that selectivity had a positive and significant
effect on graduation and retention rates (Chen & St. John, 2011; Gansemer-Topf & Schuh,

48

2006; Heck, et al., 2014; Pike, 2013; Pike and Graunke, 2014; Scott, et al., 2006; Shin, 2010;
Sparks & Nunez, 2014; Titus, 2006). However, this finding was not unanimous with Chen
(2011) finding that selectivity had no statistically significant effect. Specifically, the likelihood
of retention, persistence and completion increases with institutional selectivity, with students
enrolling in a highly selective institution having a retentions odds ratio of 1.81 (Sparks & Nunez,
2014), a persistence odds ratio of 2.1 (Chen & St. John (2011) and a completion odds ratio of
1.281 compared to other institutions (Titus, 2006). When separating out leaving behavior
between those who stop out, transfer out, and drop out, selectivity had a significant effect on the
odds of stopping out and transferring to another institution, but not on dropping out of higher
education completely (Rhee, 2008).
Financial characteristics. Financial characteristics such as expenditure on instruction,
academic support, and student services is also related to retention and graduation rates, although
the findings have been inconsistent. Some studies have discussed that the amount of money an
institution spends on instruction related activities positively contributes to retention and
graduation rates (Gansemer-Topf & Schuh, 2006; Ryan, 2004; Scott, et al., 2006), while others
found no significant relationship (Chen, 2011; Titus, 2006) and one indicated a negative
relationship (Heck, et al., 2014). Similarly, money spent on academic support is believed to
improve graduation rates in some studies (Gansemer-Topf & Schuh, 2006; Ryan, 2004) and have
no significant relationship in others (Chen, 2011; Titus, 2006). Further, expenditures on student
services were believed to have a statistically significant effect on graduation rates in one study
(Chen, 2011), no effect in others (Ryan, 2004; Titus, 2006) and a negative effect on retention
rates in one study (Gansemer-Topf & Schuh, 2006). Some of these variations could arise from
different methods of computing data, samples that were used or analytical methods. For instance,

49

some studies utilized straight dollar amounts for expenditures, while others used percent across
all expenditures, and others calculated expenditures per FTE student. Additionally, across the
studies, the methods varied including grouped logit regression, structural equation modeling,
OLS longitudinal regression, and event history analysis.
State Factors
Beyond institutional level factors, several state level factors have also been noted to
influence institutional retention rates. While the research in this area is limited, recent studies
have found some financial factors that influence enrollment and graduation rates, as well as a
few state demographic variables. Specifically, the level and type of financial support a state
provides for higher education and the size of college eligible populations as well as
unemployment rates all influence retention and completion rates.
Financial. The amount of money that a state spends on higher education overall and
through appropriations and student financial aid seems to have a positive effect on enrollment
and graduation rates. Toutkoushian and Hillman (2012) demonstrated that the total state financial
assistance per capita had a significant positive effect on students attending higher education.
However, it was no longer statistically significant when isolating out four year institutions.
Most research that explored the effect of state appropriations and higher education
performance across all institutional types found a strong positive relationship (Shin, 2010; Titus,
2009; Toutkoushian & Hillman, 2012). This included research using both direct funding amounts
and analysis of appropriations per capita. Specifically, for every 10% increase in state
appropriations for higher education, bachelor’s degree production in the state would increase by
3% (Titus, 2009). However, other research that disaggregates by institutional type suggests
possible interaction effects between state appropriations and institutional types. For instance, the

50

amount of state appropriations per capita decreased the likelihood of enrollment in both two year
institutions and competitive four years institutions, but had no effect on non competitive four
year institutions (Kim, 2012). Further, the effect of state appropriations varies based on the
state’s political culture (Heck, et al., 2014). In states classified as Individualists (focus on limited
government, social and economic self-improvement, and a marketplace orientation) state
appropriation was negatively related to graduation rates (Heck, et al., 2014). On the other hand,
in Traditionalist states (focus on maintaining established values and structures, and limited
spending on social programs), state appropriation was positively related to graduation rates
(Heck, et al., 2014).
The amount of money a state provides in merit or need based grants also has a positive
effect on enrollment and retention rates (Chen & St. John, 2011; Kim, 2012; Titus, 2009;
Toutkoushian & Hillman, 2012; Zhang & Ness, 2010). A state simply having a merit aid
program increased college going on average by 4.1%, however this is only slightly higher than
comparison groups, which had 3.5%, 2.8% and 2.1% (Zhang & Ness, 2010). Further, most of the
increase in enrollment rates occurs at research and doctoral institutions, which saw an average
increase of 7.6%, while the change in other institutions was not statistically significant (Zhang &
Ness, 2010). The amount of merit based aid also increases college enrollment rates. Specifically,
for each $1000 increase per capita in merit based aid, college going rates would increase by
14.45% for all institutions and 6.54% for four year institutions (Toutkoushian & Hillman, 2012).
Additionally, merit based aid was positively related to persistence with a one standard deviation
increase in merit based aid resulting in a 4% increase in the odds of persistence (Chen & St.
John, 2011). However, when Georgia was excluded (due to it being an outlier in the merit based
aid category) the effect was no longer significant. The amount of tuition the grants covered was

51

also related to persistence. Particularly, for each 1% increase in tuition covered by merit based
odds of persisting increased by 1% (Chen & St. John, 2011). The relationship between
persistence and the percent of tuition covered by the aid remained significant even when Georgia
was excluded.
The results for need based aid are a little more complicated. Some research shows that the
amount of state need based aid per FTE student was significantly positively related to the
number of degrees produced in the state (Titus, 2009). Additionally, state need based aid was
positively associated with enrollment rates in two year institutions (Kim, 2012). However the
same research found no statistically significant relationship between state need based aid and
enrollment in four year institutions (Kim, 2012). Further, Chen and St John (2011) found no
statistically significant relationship between need based aid and the odds of persisting, however,
they did find a statistically significant relationship between the amount of tuition need based aid
covered and the odds of persisting. Specifically, for each 1% increase in tuition covered by need
based aid, odds of persisting increased by 2% (Chen & St. John, 2011).
Population data. A state’s demographic makeup also influences enrollment and
completion rates in the state. Research has shown that the size of a state’s traditional college age
population influences higher education completion rates, but in a negative way (Bound &
Turner, 2007; Bound, Lovenheim, & Turner, 2010). Specifically, a 10% increase in the
traditional college age population leads to a 4% decrease in the bachelor’s degree completion
rate (Bound & Turner, 2007). Similarly, increases in the high school graduate population were
negatively associated with the state’s score on the completion category in Measuring Up data
(Volkwein & Tandberg, 2008). While this may seem surprising, Bound and Turner (2007)
hypothesize that a crowding out effect occurs, where states are not able to fully support their

52

college aged population and therefore completion rates decrease. Relatedly, the percent of
residents with a bachelor degree leads to larger number of students attending higher education
(Toutkoushian & Hillman, 2012).
Some research has also found that a state or county unemployment rate can influence
persistence and graduation (Sparks & Nunez, 2014; Shin, 2010) but this finding is not
unanimous, as some research has found that unemployment rates do not influence degree
production (Titus, 2009). County unemployment rate can influence persistence, with students
more likely to persist if the higher education institution was located in a county with higher than
average unemployment rates (Sparks & Nunez, 2014). More broadly, state unemployment rate
had a negative effect on graduation. Meaning that as unemployment rates went up, graduation
rates went down (Shin, 2010).
Conclusion
This chapter has outlined the literature on higher education funding (with a specific focus
on state funding), the accountability movement, and performance funding specifically as well as
the literature on enrollment and retention from an institutional and state perspective with a focus
on underrepresented students of color and low SES students. While the research in these areas is
vast, limited research has explored how higher education institutions are balancing the
competing needs of demonstrating efficient economic outcomes via performance funding while
maintaining access and success for all students, particularly in the four-year sector. Specifically,
no research has yet quantitatively explored if performance funding policies influence the
enrollment rates of underrepresented students in public four year institutions. This study seeks to
fill this gap in the research on performance funding and underrepresented student enrollment.

53

Chapter 3
This study explored how performance funding influences the enrollment of
underrepresented students through the following research questions:
1) Are there unintended consequences from performance funding policies related to
underserved students’ enrollment at public four-year institutions?
a. Does performance funding have an influence on the enrollment of Pell Grant
eligible students?
b. Does performance funding have an influence on the enrollment of
underserved minority students?
2) Does institutional type influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
a. Does Carnegie classification influence the effect of performance funding on
underserved students’ enrollment at public four-year institutions?
b. Does selectivity influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
This chapter outlines the population studied, data sources, variables used, and analytical method
used to answer these questions.
Sample/Population
This study focused exclusively on public 4-year institutions in U.S. states during the
2000-2014 time period. Criteria for inclusion in the study population were based on the
following considerations. This study included only public institutions, as the performance
funding policy does not apply to the private sector. Institutions were selected in a three-step
process.

54

First, for each year, all four-year public institutions that were degree granting in the U.S.
were identified. This was done using the IPEDS data base by selecting the “Sector of Institution”
variable and identifying all “Public, 4 year or above” institutions as well as the “Degree Granting
Status” variable and selecting “Degree-granting.” Next, the list was narrowed to ensure only
four year, degree granting institutions in U.S. states were included. This was done using the
“Sector of Institution” variable and all institutions listed as “2-year” or “Private nonprofit” were
excluded. Additionally, the Carnegie variable was used to drop any institutions listed as
Associate’s schools or Medical Schools, Health Schools, or Schools of Law. For the years it is
available, the Carnegie Enrollment Profile was used and all institutions listed as two year were
dropped. Further, all institutions with locations in Guam, Northern Marianas, Puerto Rico, Virgin
Islands, or District of Columbia were excluded.
Finally, as performance funding and this study focus on undergraduate students, any
institutions that did not grant undergraduate degrees were dropped. This was done through the
“Institutional Category” variable and all institutions listed as “Degree-granting, graduate with no
undergraduate degrees” or “Degree-granting, not primarily baccalaureate or above” or “Degreegranting, associate’s and certificates” were excluded from the study. Additionally, the Carnegie
undergrad profile were used and all institutions listed as “exclusively graduate/professional” and
“Majority graduate/professional” were dropped.
As the last step, all institutions were merged together across the years and only
institutions that met the above criteria for all years were maintained. This resulted in a final
dataset of 513 institutions. As all institutions that meet these criteria will be used, this study will
focus on the whole population instead of a sample of the population.

55

Data Sources
Data came primarily from the Integrated Postsecondary Education Data System (IPEDS),
which is a dataset collected annually by the Department of Education’s National Center for
Educational Statistics (NCES). All institutions that receive federal funds from the United States
government are required to submit information to the NCES and all the institutions being
analyzed in this evaluation submit IPEDS data annually. Data was gathered from the IPEDS data
center for the years 2000 to 2014. The IPEDS database was selected for five main reasons. 1) It
includes all the institutions in the population and most of the variables of interest. All institutions
selected for this study submit data to the NCES annually. Additionally, most of the variables of
interest are collected by the NCES in IPEDS annually or semiannually. This will ensure limited
missing variable issues; 2) it has specific definitions for all the variables. This guarantees each
institution is using the same criteria in reporting the variables and therefore increases validity; 3)
it is available for all the years of the study. IPEDS data is collected annually since the 1980s and
has yearly reports for the 2000-2014 timeframe of this study; 4) IPEDS is commonly used in
higher education research including numerous studies analyzing performance funding (e.g.,
Rabovsky, 2012; Rutherford & Rabovsky, 2014; Tandberg & Hillman, 2014). It is specifically
noted as beneficial for research questions related to the external environment such as a policy
change (Jaquette & Parra, 2014); 5) IPEDS measures and reports data at the institutional level.
Other databases (e.g., Delta Cost Project) condense multi-campus systems into one measure,
which does not allow proper attention to “parent-child” issues (Jaquette & Parra, 2014).
A few state level variables came from other databases. First, the amount states spend on
higher education appropriations was retrieved from the Grapevine database. Additionally, the
amount the state spends on need based and merit based aid came from the National Association

56

of State Student Grant and Aid Programs (NASSGAP). Finally, the college age population in a
state, and the underrepresented minority population came from the U.S. Census database.
Parent-Child Issues
A parent-child issue arises from states with branch campuses or higher education
systems. In a parent-child relationship data from the “parent” institution (e.g., the main campus)
is used to report for the “child” (e.g., a branch campus) for at least one IPEDS variable (Jaquette
& Parra, 2014). In particular these issues occur in IPEDS Finance data and are prevalent before
2003-2004 (Jaquette & Parra, 2014). As this research study includes states with branch
campuses and systems and includes years before 2003-2004 it is important to discuss how this
concern will be addressed.
Since this study seeks to explore differences between institutional types (Research
question #2) it is important to maintain as many institutions at the child level as possible.
Therefore, to the extent possible, each institution will be treated autonomously in this research
study. In order to decrease the likelihood of parent-child issues the number of IPEDS surveys
used is limited and in particular the IPEDS finance survey is used for only one variable (state
appropriations as a percent of total revenue). Using multiple IPEDS surveys increases the
likelihood of encountering parent-child issues and in particular, the IPEDS finance survey is the
most prevalent for parent-child issues (Jaquette & Parra, 2014). Jaquette and Parra (2014) note
specifically that analyses that uses revenue data but does not use asset/liability measures
typically does not have to be collapsed to the system level because revenue measures are usually
reported at the child level. By limiting the number of IPEDS surveys used and only using a
revenue variable from the finance survey, the likelihood of parent-child issues should be small.

57

Where data was faced with parent-child issues, variables were collapsed to the parent level
through the process outlined in Jaquette and Parra (2014).
Variables
Dependent Variables
Four different dependent variables were used throughout this study. For the research
questions related to Pell Grants, both the percent of First Time Full Time (FTFT) students who
received Pell Grants and the average amount of Pell Grants were used. These present
complimentary views regarding the enrollment of Pell Grant students. The first represents any
changes an institution may make in the number of Pell Grant students it enrolls, while the later
represents the type of Pell Grant student enrolled. For research questions related to underserved
minority students, both the total number of enrolled underserved minority students and the
percent of minority students were used. To create the total underserved minority variable, the
number of FTFT students who identified as Black, Hispanic, Native American, or Alaskan
Native were added together. This number was then divided by the total FTFT enrollment to
create the underserved minority percent variable.
Institutional Variables
Several institutional level variables were included in the models. The rationale for
inclusion for each is explained below and each variable is defined in Table 2. Institutional factors
may influence enrollment rates for underrepresented students of color and Pell eligible students
and need to be controlled for. For instance, research consistently finds that receiving financial aid
has a significant positive effect on student enrollment behavior (McPherson and Schapiro, 1991;
Van Der Klaauw; 2002). In order to control for this, the average amount of institutional aid
granted was added as a control variable. Similarly, the cost of attendance for higher education

58

can influence enrollment behavior, in particular for students with limited financial resources
(Bergerson, 2009). Students frequently make enrollment decisions before tuition and fees are set
for the upcoming year, meaning students deciding whether to enroll for the 2015-2016 academic
year will base their decision on the 2014-2015 tuition and fees listed. Therefore the published
tuition and fees for in state residents the year prior to starting was included as a control. While
some may argue that net price is a more accurate measure for cost of attendance, research has
shown that underrepresented students often do not have the cultural capital to understand net
price calculations and instead make enrollment decisions based on sticker price (Bergerson,
2009).
Institutional factors may also influence how an institution responds to performance
funding. First, an institution’s level of selectivity may influence whether the institution is able to
make changes to its enrollment profile. For example, an institution that accepts only 50% of its
applicants is able to make greater changes to the types of students it admits compared to an
institution that accepts 98% of its applicants. To this end, an institution’s admission rate was
included as a control. Additionally, institutions may react differently to performance funding
depending on how reliant they are on funding from the state. In other words, if an institution is
highly reliant on state appropriations, they may be more likely to make institutional behavior
changes (such as changing the enrollment profile) that could lead to increases in the funding they
receive. To account for this, the percent of institutional revenue that comes from state
appropriations was included as a control variable. For years this variable was not available, it
was calculated by dividing the total institutional appropriations by the institution’s total revenue
and multiplying by 100. Results were compared for the years both the calculated and reported
variables were available and no differences were noted. Therefore for the years that

59

appropriations as a percent of total revenue was not available, the calculated admission rate was
used to decrease the number of missing values.
Research questions 2a and 2b ask whether institutional type influences the effect of
performance funding on the enrollment of underserved students. To answer these questions,
categorical variables were created for Carnegie classification and admission rate. For Carnegie
classification, 10 categories were created (three for doctoral institutions, three for masters
institutions, three for baccalaureate institutions, and one for special focus institutions). For
admission rate, 5 categories were created: open access (admitting 85-100% of applicants),
moderately selective (admitting 75-84.99% of applicants), selective (admitting 50-74.99% of
applicants), highly selective (admitting 30-49.99% of applicants), and very highly selective
(admitting less than 29.99% of applicants).
State Variables
To distinguish between states with and without performance funding a dummy variable
was created that accounts for whether the state has performance funding each year. States that
had a performance funding policy in place during the time frame studied were identified by
looking at various research articles and meta analyses of performance funding research and
compiling a list of states with performance funding. Where there was unanimous agreement the
state was included with the dates provided. Where there was disagreement, the researcher
explored further through state statutes and codes or discussion with state experts and determined
if the state had performance funding and what years the policy was in place. When only some
institutions within a state were subject to performance funding, then only those institutions were
coded as performance funding and the other institutions were coded without performance
funding. A performance funding count variable was also created where the number of years the

60

policy was in place was counted. If a policy was in place before the years of the study, the first
year in the study was marked as the year of the policy. For instance if a policy went into place in
1995 then year 2000 would be marked as “6.” Table 1 lists the states with performance funding
and the years where they had an active performance funding policy.
Table 1 List of States with Performance Funding and Years
State
Arizona
Arkansas
Colorado
Idaho
Indiana
Kansas
Louisiana
Maine
Michigan
Minnesota
Mississippi
Missouri
New Jersey
New Mexico
New York
Ohio
Oklahoma
Oregon
Pennsylvania
South Carolina
South Dakota
Tennessee
Texas
Virginia

Years with Performance Funding
2011-2014
2013-2014
2000-2004
2000-2005
2007-2014
2002; 2005-2010; 2013-2014
2009-2014
2013-2014
2012-2014
2012-2014
2013-2014
1993-2002; 2012-2014
1999-2002
2007-2014
1998-2007(SUNY); 2000-2014(CUNY)
1998-2007; 2009-2014
1998-2000; 2002-2010; 2013-2014
2000-2001; 2007-2014
2000-2014
1996-2004
1998-2002; 2005-2013
1979-2014
2009-2011
2007-2014

Several state level variables were also used as controls. The rationale for including each
variable is explained below and each is defined in Table 2.

61

Table 2 Variable Definitions and Sources
Variable
Percent Pell Grants
Pell Grant Amount
USM Total

USM Percent

Carnegie
Classification
Institutional Aid
Amount
Tuition
Admission Rate
Appropriations as
Percent of Revenue
Appropriations per
FTE
State Need Based
Aid
College Age
Population
College Age USM
Population
College Age USM
Percent
Affirmative Action
Ban

Definition of Variable
The percentage of FTFT degree seeking undergraduate
students who received Pell Grants in the incoming class.
Average amount of Pell Grant received by FTFT
students
The total FTFT degree seeking undergraduate students
who identify as at least one of the following: Black,
Hispanic or Latino, American Indian or Alaskan Native,
and Native American or other Pacific Islander.
The percentage of FTFT degree seeking undergraduate
students who identify as at least one of the following:
Black, Hispanic or Latino, American Indian or Alaskan
Native, and Native American or other Pacific Islander.
An institutional classification coding structure
developed by the Andrew W. Carnegie Foundation for
the Advancement of Teaching.
Average amount of institutional grants
(scholarships/fellowships) received by FTFT students
In-state published tuition and required fees for the prior
academic year
Number of total students admitted divided by the total
applicants. Ratios are converted to percentages and
rounded to the nearest whole number.
State appropriations divided by total core revenues.

Source
IPEDS

The total amount a state spends on tax appropriations
for higher education divided by the total FTE
enrollment minus medical students.
The total amount a state spends on need based aid for
higher education divided by the total FTE enrollment in
the state minus medical students.
The number of residents in the state between the ages of
18 to 24.
The number of residents within a state aged 18-24 who
identify as at least one of the following: Black or
African American, American Indian and Alaska Native,
Native Hawaiian and other Pacific Islander, or Hispanic.
The percent of residents within a state aged 18-24 who
identify as at least one of the following: Black or
African American, American Indian and Alaska Native,
Native Hawaiian and other Pacific Islander, or Hispanic.
Whether a state has an affirmative action ban in place at
any point during the study years

Grapevine

62

IPEDS
IPEDS

IPEDS

IPEDS
IPEDS
IPEDS
IPEDS
IPEDS

NASSGAP
US Census
US Census

US Census

Various
Sources

Financial characteristics have been shown to influence student enrollment in higher
education. Significant research has found a relationship between state appropriation amounts and
students attending and succeeding in higher education (Kim, 2012; Shin, 2010; Titus, 2009;
Toutkoushian & Hillman, 2012). Additionally, the amount of money a state provides in merit or
need based grants also has a positive effect on enrollment and degree production rates (Kim,
2012; Titus, 2009; Toutkoushian & Hillman, 2012; Zhang & Ness, 2010). Both are believed to
represent the overall commitment of the state to higher education. To account for this, the total
state appropriations for higher education was included as a control variable. Additionally, this
study controlled for the amount of money the state spends on higher education need based aid.
Both of these variables were divided by total FTE enrollment (net medical students) in the state
to account for differences in population. As research has shown that the number of college aged
residents in the state can influence the enrollment and completion rates for higher education
(Bound & Turner, 2007; Bound et al., 2010; Volkwein & Tandberg, 2008) this study included a
control variable for the total college age population in the state. A few variables were controlled
for only in select models. It can be assumed that the size of a state’s underserved minority
population would influence the number of underserved minority students enrolled. Therefore,
for the models using percent of students who identify as underserved minorities as the dependent
variable, the state underserved minority population size was used as a control.
In addition to the transformation noted above, variables were transformed or created in
the following ways. First, state appropriations was adjusted for inflation using the Consumer
Price Index for the 2014 year before dividing by FTE enrollment. Second, for the years
admission rate was not available but applicant and admission numbers were available, admission
rate was calculated by dividing the total number of applicants by the total number of students

63

admitted and multiplied by 100. This calculation was also performed on the years admission rate
was available to confirm its accuracy. No difference was noted for the years both calculated and
reported admission rate was available. Therefore for the years missing admission rates, the
calculated admission rate was used to decrease the number of missing values.
Analytical Method
This study examined the effect of performance funding policies on the enrollment and
graduation of underserved students through a linear regression analysis utilizing OLS-based
panel techniques. Panel data refers to repeated observations of the same research unit (in this
case, institutions) over a period of time. Panel methods are applicable for questions about
systematic change over time. Specifically, panel analysis allows a researcher to explore temporal
patterns and identify if any abrupt shifts occurred (Singer & Willett, 2003). Therefore, utilizing
panel analysis for this study allowed the researcher to directly explore whether performance
funding policies influence the enrollment of underrepresented students over time.
Utilizing a panel dataset provides numerous benefits and is more reliable than crosssectional data analysis. First, panel data controls for individual heterogeneity by allowing the
researcher to control for state and time invariant variables (Baltagi, 1995). Second, panel data
gives less collinearity among the variables since the researcher is able to differentiate between
variability between institutions and variability within institutions (Baltagi, 1995). Finally, panel
data is beneficial for policy analysis as it allows for the study of lags in behavior as a result of a
policy change (Wooldridge, 2013).
While this study controlled for the variables mentioned above, it is possible that
unobserved state factors also contributed to the enrollment and completion of underrepresented
students. For instance, a state’s historic culture could promote participation in higher education.

64

In order to account for this in the analysis, state fixed effects estimates were included to account
for the unobserved heterogeneity within states. Additionally, national policies or experiences
could impact all institutions at specific points in time but be unobserved in the study. For
instance, the great recession could influence higher education attendance. In order to account for
this in the analysis, year fixed effects were also included. Finally, unobserved institutional
factors could also influence the analysis, for instance an institution being located in a desirable
place to live. Therefore, institutional fixed effects were also included.
As previous research has shown that the number of years that have passed since
implementation can influence the effect of performance funding (Tandberg & Hillman, 2014),
state specific policy time dummies were created to account for the number of years since policy
implementation with the first year of implementation being year one. For instance, since New
Mexico began their performance funding policy in 2007 all institutions within the state of New
Mexico will have a policy time variable of zero (0) for the years 2000-2006, then a one (1) for
2007, a two (2) for 2008, a three (3) for 2009 and so on. Conversely, since Pennsylvania began
their performance funding policy in 2000 all institutions within that state will have a policy time
variable of one (1) for 2000, a two (2) for 2001, a three (3) for 2002 and so on. This allows the
researcher to explore if the effect of performance funding varies over time.
Models
Several models were estimated to answer the research questions. First, to answer research
question #1, “Over time, are there unintended consequences from performance funding policies
related to underrepresented students’ enrollment at public four-year institutions?” models were
run with four different dependent variables. First, for research question 1a, models were run
with average amount of Pell Grant and the percent of FTFT students who receive Pell Grants as

65

the dependent variables. For research question 1b, models were run with both the total number of
underserved minority students and the percent of FTFT students who identify as underserved
minority students as the dependent variable. Since the effects of policy change may not be seen
till a few years after the policy goes into effect (Hillman, Tandberg, & Fryar, 2014) models were
calculated with one, two, and three year lags. Further, models with both a binary performance
funding variable and a performance funding count variable were used. As outlined above, the
models included controls for institutional and state variables that may influence the dependent
variable as well as institution, state and year fixed effects. A Hausman test was run to determine
if a fixed effects or random effects model was most appropriate and a fixed effects model was
selected for all models.
Next, to answer research question #2, “Over time, does institutional type influence the
effect of performance funding on underrepresented students’ enrollment at public four-year
institutions?” models were run which included interaction effects for both Carnegie classification
and admission rate with performance funding. Models were run for four dependent variables:
Average Pell Grant awarded, percent of FTFT students who receive Pell Grants, total FTFT
underserved minority enrollment, and percent of FTFT who identify as underserved minorities.
These models were run for both a binary performance funding variable and with a performance
funding count variable. As outlined above, the models included controls for institutional and
state variables that may influence the dependent variable as well as institution, state and year
fixed effects. A Hausman test was run to determine if a fixed effects or random effects model
was most appropriate and a fixed effects model was selected for all models. Where significant
findings occurred, tests for robustness were run and are included in the findings below or in an
appendix.

66

When using panel data a few issues may arise and need to be addressed. First, panel data
can cause issues of serial correlation where one observation for an institution is dependent on the
preceding year’s observation at that institution. Serial correlation can bias results by presenting
smaller standard errors and a higher R squared. To account for this, a Wooldridge (2002) test
was conducted and no serial correlation was found.
Limitations
Several limitations in this study deserve mention. First, this study focuses exclusively on
four-year institutions. Despite many performance funding policies affecting two-year
institutions, they are excluded from this analysis due to the variance between the two sectors.
Future research should explore whether performance funding influences the enrollment of
underrepresented students in public two-year institutions. Second, this study focuses on FTFT
students for the dependent variable, which excludes students who enroll part time or transfer into
the institution. As many underrepresented students attend part time this could bias the results.
However, enrollment figures for all undergraduates by Pell Grant recipients and underserved
students of color are not available for all years of the study.
Another limitation in this study is the change in racial categories in IPEDS. In particular,
the category “two or more races,” which was added in 2008-2009 (and was optional until 20102011) presents a challenge. Students who identify as two or more races in the new categories
may have identified themselves previously in an underrepresented student of color category,
which could artificially lower the dependent variable in Research Question 1a “Does a state’s
performance funding policy have an influence on the enrollment of underrepresented students of
color?” As data is available without this category till 2010, students who identify as “two or
more races” were excluded with recognition of the aforementioned bias.

67

A further limitation in this study is the inability to account for variance across
performance funding models, specifically, the various strengths of performance funding models.
As noted above, performance funding models account for anywhere from 5% to 100% of total
state appropriations. It can be assumed that institutions that receive 100% of their state
appropriations from performance funding may respond differently to the policy than those who
only receive a small percent of their state appropriations from performance funding. However,
no reliable data source exists that tracks this information and many state websites and legislative
histories did not include this information. Therefore, this study was not able to control for
strength of performance funding. Similarly, in recent years a few states have included metrics
within their performance funding policies that focus on underserved student enrollment.
However, as many of these metrics only went into effect in recent years they were not included
in this study.
Finally, the performance funding policies analyzed in this study are largely performance
funding 1.0 policies with limited performance funding 2.0 policies. As performance funding 2.0
policies are still relatively new, not enough years of data exists to include most of these policies.
As data becomes available, research should explore whether performance funding 2.0 policies
have similar effects on enrollment of underrepresented students as performance funding 1.0
policies.

68

Chapter 4
This chapter presents descriptive and quantitative fixed effects panel analysis findings for
the following research questions:
1) Are there unintended consequences from performance funding policies related to
underrepresented students’ enrollment at public four-year institutions?
a. Does performance funding have an influence on the enrollment of Pell Grant
eligible students?
b. Does performance funding have an influence on the enrollment of underserved
minority students?
2) Does institutional type influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
a. Does Carnegie classification influence the effect of performance funding on
underserved students’ enrollment at public four-year institutions?
b. Does selectivity influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
First, the descriptive statistics for the variables in the study are presented including any
differences between institutions in performance funding states and institutions without
performance funding. Then the panel analysis results for each of the research questions are
presented in turn. Discussion of the results and implications are presents in Chapter 5.
Descriptive Statistics
Institutions were included in the sample if they were listed as public, four year, degree
granting institutions for all 15 years of the study. However, as performance funding policies
focus on undergraduate education, institutions were removed if they were majority graduate

69

serving institutions. This included institutions listed as medical schools or schools of law under
the Carnegie classification, as well as institutions listed as majority graduate/professional under
the Carnegie enrollment variable. The final result is a sample size of 513 institutions across all
50 states.
Table 3 provides a summary of all variables included in this study for all institutions over
the 15 years. Mean, standard deviation, minimum value and maximum value are included for all
variables. As some variables are skewed (see specifics below), a median is presented for all
variables as well. Skewness is a measure of the symmetry of the data set. If a variable is highly
skewed its data points are unbalanced with a majority on one side or the other. A high skewness
could pull the mean in one direction from a few large outliers or from clumping of the data.
Therefore, the median is a preferred measure of central tendency for variables that are highly
skewed. For the purpose of this study, a skewness greater than 1 or smaller than -1 was
considered highly skewed.
Table 3 demonstrates notable differences across institutions and years in this study. For
instance, states funded higher education at significantly different rates. State appropriations per
FTE (inflation adjusted to 2014 dollars) is measured in thousands of dollars and ranged from
1.892 to 17.152 with a mean of 7.260. Additionally, state need based financial aid per FTE
(measured in thousands of dollars) differed from 0 to 1.867 with a mean of 0.574. Population
variables also demonstrated diversity across all institutions and years. To start, the state college
age population (presented in ten-thousands) skewed to the left (skewness of 1.90) and ranged
from 5.632 to 749.205 with a median of 107.777. For underserved minority residents aged 18-24
(calculated in thousands) there was also a strong skew to the left (skewness of 2.06) and ranged
from 1.771 to 2,165.958 with a median of 183.687. Additionally, the variable has a large

70

variance with a standard deviation of 503.742, demonstrating that there is wide variety across
underserved minority residents aged 18-24 within the dataset. Similarly, the percent of residents
18-24 who identify as underserved minority residents skewed to the left (skewness of 1.13) and
ranged from 1.51% to 67.52% and had a median of 18.18%.
Table 3 Summary Statistics For All Institutions Across All Years
Variable
Median
Mean
SD
State Need Based Aid
.464
.575
.461
Appropriations per FTE
7.151
7.260
1.833
College Age population
107.777
161.592
160.083
College USM Population
183.687
375.063
503.743
College USM percent
18.18
20.29
13.11
Tuition
Institutional Aid Amount
Percent Pell Grants
Pell Grant Amount
USM Percent
USM Total
Appropriations as Percent
of Revenue
Admission Rate

Min
0.00
1.892
5.632
1.771
1.51

Max
1.867
17.152
749.205
2,165.958
67.52

5.300
2.846
36.00
4,078.00
15.48
230.00
34.00

5.755
3.257
38.08
3,974.78
24.79
371.87
34.25

2.630
1.935
16.33
587.51
24.67
420.56
10.78

1.466
0.00
8.00
1,146.00
0.00
0.00
0.00

22.997
15.362
92.00
6,955.00
99.68
3,481.00
81.41

71.00

69.25

17.29

7.00

100.00

The institutional level financial variables also differed across institutions and years in this
study. Published in district tuition (lagged one year and calculated in thousands of dollars)
skewed slightly to the left (skewness 1.08) and ranged from 1.466 to 22.997 with a median of
5.300. To offset increasing tuition rates, institutions often provide institutional financial aid. The
average amount of institutional financial aid provided (measured in thousands of dollars) was
skewed left (skewness of 1.51) and ranged from 0 to 15.362 with a median of 2.846. Pell Grant
aid presented similar patterns. The percent of students who received Pell Grants ranged from 8%
to 92% and had a mean of 38.08%. The average amount of Pell Grant awarded to these students

71

ranged from 1,146 to 6,955 with a mean of 3,974.78. For underserved student enrollment, the
percent of underserved minority FTFT students enrolled skewed left (skewness 1.76) and ranged
from 0% to 99.68% with a median of 15.48%. The total number of underserved minority
enrolled FTFT also skewed to the left (skewness 2.34) with a range from 0 to 3,481 and a median
of 230.
The extent to which institutions relied on state appropriations for their revenue varied
greatly. State appropriations as a percent of total institutional revenue ranged from 0% to
81.41% with a mean of 34.25%. Institutional admissions rates in this study included open access
institutions and highly selective institutions. Overall the mean admittance rate was 69.25% with
the most selective institution having a 7% admittance rate and several institutions admitting all
the applicants who applied (100% admittance rate). While the variety in values across all
institutions is important to note, for this study it is of particular interest if any of the variables are
notably different between institutions subject to performance funding and those without
performance funding.
Table 4 compares institutions with performance funding and institutions without
performance funding for the variables used in this study. If an institution was subject to
performance funding for only a portion of the study period (for instance 2000-2008), they would
be included in the performance funding section for the years they were under performance
funding (2000-2008) and then calculated in the non performance funding section for the years
they did not have performance funding (2009-2014). Some important distinctions in the state
level variables should be noted. First, in regards to institutional funding, two complimentary
trends exist. Institutions with performance funding received slightly more need based state
financial aid per FTE (median of 0.622 for institutions with performance funding compared to

72

0.426 for institutions without performance funding). Additionally, institutions with performance
funding had lower levels of state appropriations per FTE (mean of 6.690 for institutions subject
to performance funding and 7.502 for institutions without performance funding). Together, it can
be inferred that states with performance funding appropriated more funding to need based
financial aid, whereas states without performance funding relied more on direct appropriations to
institutions.
Table 4 Descriptive Statistics by Performance Funding
States With Performance Funding
States Without Performance Funding
Variable
Median
Mean
SD
Median
Mean
SD
State Need
.622
.764
.533
.426
.494
.400
Based Aid
Appropriations 6.609
6.690
1.591
7.395
7.502
1.875
per FTE
College Age
114.557
152.795
110.156
101.736
165.322
176.927
population
College USM
197.343
284.772
305.264
173.522
413.342
562.972
Population
College USM
18.16
19.69
12.64
18.20
20.54
13.29
percent
Tuition
Institutional
Aid Amount
Percent Pell
Grants
Pell Grant
Amount
USM Percent
USM Total
Appropriations
as Percent of
Revenue
Admission
Rate

5.919
2.966

6.484
3.374

2.806
1.999

5.010
2.803

5.439
3.207

2.484
1.904

38.00

39.56

16.24

34.00

37.32

16.33

4,096.00

4,008.57

559.38

4,071.00

3,957.29

600.92

14.81
214.00
32.76

23.12
316.35
32.59

22.80
342.84
9.68

15.81
237.00
35.00

25.50
395.77
34.88

25.40
447.81
11.10

73.00

70.17

17.77

70.41

68.86

17.07

73

Significant differences in state population also existed between institutions with and
without performance funding. Institutions with performance funding resided in states with larger
college age populations and underserved minority college age populations (median of 114.557
and 197.343 respectively) compared to institutions without performance funding (median of
101.736 and 173.522 respectively). Due to this difference in overall numbers, the percent of 1824 who identify as underserved minority residents was also calculated. This variable had limited
variance between institutions with and without performance funding (median of 18.16 for
institutions subject to performance funding and 18.20 for institutions without performance
funding). Together, these measures imply that performance funding is more prevalent in states
with larger 18-24 populations, however the racial diversity is similar for both performance
funding and non performance funding institutions.
Some noteworthy differences in institutional level variables emerged when institutions
were separated by those with and without performance funding. Published in district tuition rates
were significantly higher at institutions with performance funding (mean of 6.484) compared to
institutions without performance funding (mean of 5.439). There were no significant differences
in average amount of institutional aid between institutions subject to performance funding and
those without performance funding. Institutions with performance funding provided an average
of 3.374 of financial aid to students and institutions without performance provided an average of
3.206. The percent of students who received Pell Grants was slightly higher at institutions with
performance funding (39.56%) compared to 37.32% at institutions without performance funding.
The average amount of Pell Grant awarded had negligible difference between institutions with
and without performance funding (4.008 and 3.957 respectively).

74

Overall variables related to institutional cost had little difference outside of published in
district tuition rate, which was notably higher at institutions with performance funding. This
difference in tuition price without a corresponding difference in financial aid implies that
students may pay a higher net price at institutions in performance funding states. Enrollment
variables had some variance, but overall were similar. Underserved minority FTFT student
enrollment was slightly lower for institutions with performance funding (316.35) compared to
institutions without (395.77). However, the percent of FTFT students who identified as
underserved minorities was more similar, with a mean of 23.12% for institutions with
performance funding and 25.50% for institutions without. Institutions without performance
funding were modestly more reliant on state appropriations. Institutions with performance
funding received 32.59% of their revenue from state appropriations, whereas institutions without
performance funding received 34.88% of their revenue from performance funding on average.
In sum, the descriptive statistics established meaningful variance across institutions and
years for variables included in this study. These distinctions demonstrate the importance of
including the variables beyond their already established theoretical significance. Additionally,
significant differences between institutions with and without performance funding are present,
including in some of the variables of interest; signifying the importance of exploring if
performance funding exerts an influence on the dependent variables.
Performance Funding and Underserved Student Enrollment
This section presents the major results for research question one. In order to examine the
influence of performance funding on underserved student enrollment, a panel analysis using OLS
regression was employed. A Hausman test was performed and the null hypothesis was rejected
for all models. Rejection of the null hypothesis indicates that the fixed effects model is

75

consistent and therefor preferred. Robust standard errors were used for all models to address the
potential of correlated standard errors. Institutional and year fixed effects were included for all
models.
For each dependent variable, analysis was performed using performance funding as a
binary variable (1 if the state had performance finding, 0 if it did not) and as a count variable
where each year of performance funding was counted (i.e. 1 the first year, 2 the second year and
so on). If performance funding was in place when the study began the count was started at the
first year of the policy. For instance if performance funding started in 1998 then the count
variable for 2000 would be 3. As the effects of policy change can often take time to appear
models with 1, 2, and 3 year lags were calculated.
Pell Grant Variables
Tables 5 and 6 present the results of research question 1a) Does performance funding
have an influence on the enrollment of Pell Grant eligible students? This study explored the
research question through two dependent variables: 1) the percent of FTFT students who receive
Pell Grants; and 2) the average amount of Pell Grant awarded to students. These present
complimentary views regarding the enrollment of Pell Grant students. The first represents any
changes an institution may make in the number of Pell Grant students it enrolls, while the later
represents the type of Pell Grant student enrolled. Pell Grants are awarded at various levels to
students based on financial need with larger Pell Grants signaling greater need. In 2011 (the
middle year of available data) Pell Grants were awarded in amounts between $555 and $5,550.
Therefore changes in the amount of Pell Grant awarded could signal a change in the level of
need for enrolled Pell Grant students (Klechen & Stedrak, 2016). For example, an increase in the
average amount awarded would demonstrate that the Pell Grant students enrolled that year had a

76

greater financial need and therefore received more Pell Grant aid. As Pell Grant data is not
available in IPEDS until 2008, these results only include variables for 2008-2014.
Pell Grant percent. Table 5 presents the findings for percent of FTFT who receive Pell
Grants with a binary performance funding variable. Across all models, no statistically significant
relationship was present between performance funding and percent of Pell Grant students
enrolled FTFT. Additionally, 95% confidence intervals ranged across positive and negative
variables meaning it is impossible to attribute the direction of any effect that might occur. The
only variables found to effect percent of Pell Grant students enrolled were institutional financial
aid amount and tuition rate. Institutional financial aid amount was significant at the p<.001 level
across all models, however the effect was small. For every $1,000 increase in institutional aid,
there is an expected decrease of .56% of FTFT Pell Grant recipients. Utilizing the overall mean
of FTFT students (1742), this translates to an expected decrease of about 10 Pell Grant students
for every $1,000 increase in institutional aid. Published in district tuition rate for the prior year
was significant at the p< .05 level across all models and again the effect was small. For every
$1,000 increase in tuition there is an expected .64% increase in FTFT Pell Grant students
enrolled. Utilizing the study mean, this translates to an expected increase of 11 Pell Grant
students enrolled FTFT. In sum the variables that presented the strongest effect were cost based
variables (tuition rate and institutional financial aid amount) implying that the cost of higher
education has the strongest effect on enrolling Pell Grant eligible students. This is unsurprising
as Pell Grant students are a proxy for low income students, who may not be able to afford high
cost institutions. Of note, no state variables were statistically significant predictors of the percent
of FTFT students who receive Pell Grants.

77

Little difference was found across models with different lags. The R-squared value shows
that about 62% of variance of Pell Grant students enrolled FTFT can be explained by the
variables in the model. Models were estimated utilizing a performance funding count variable,
which had similar results to the binary performance funding models. Results for the
performance funding count models are included in the appendix.
Table 5 Results for the Effects of Performance Funding on the Pell Grant Percent
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
0.233
0.300
0.203
0.294
funding
(0.292)
(0.316)
(0.356)
(0.452)
(binary)
Institutional
-0.562***
-0.563***
-0.562 ***
-0.564***
Aid Amount
(0.111)
(0.110)
(0.110)
(0.110)
Tuition
0.643*
0.643*
0.638*
0.637*
(0.295)
(0.295)
(0.295)
(0.296)
Appropriations -0.046
-0.046
-0.046
-0.046
as Percent of
(0.031)
(0.031)
(0.031)
(0.031)
Revenue
State Need
0.505
0.521
0.493
0.508
Based Aid
(0.686)
(0.691)
(0.687)
(0.687)
College Age
-0.003
-0.003
-0.003
-0.003
Population
(0.003)
(0.003)
(0.003)
(0.003)
Appropriations -0.1.04
-0.093
-0.093
-0.089
per FTE
(0.229)
(0.229)
(0.229)
(0.229)
Affirmative
-1.604*
-1.508
-1.52
-1.52
Action Ban
(.806)
(0.808)
(0.803)
(0.808)
R-squared
.6216
.6216
.6215
.6216
(within)
n
3324
3324
3324
3324
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Pell Grant amount. As no effect was found in the percent of FTFT student enrollment
that received Pell Grants, this study next explored the average amount of Pell Grant students
received. Table 6 presents the results of this analysis for a binary performance funding variable.
Performance funding was found to have no statistically significant effect on the average amount
of Pell Grant awarded to FTFT students. Additionally, all of the 95% confidence intervals ranged

78

across positive and negative variables, meaning it is impossible to attribute the direction of any
effect that might occur.
Three variables influenced the average amount of Pell Grant awarded at a statistically
significant level: tuition rate, percent of revenue that came from state appropriations, and college
age population in the state. Published in district tuition rate for the prior year was significant at
the p< .05 level for all models. While statistically significant, the effect was small, with every
$1,000 increase in tuition leading to a predicted decrease of $23 in average Pell Grant awarded.
State appropriations as a percent of total institutional revenue was significant at the p<.05 level.
For every 1% increase in the percent of total revenue coming from state appropriations, there
was an expected decrease in average amount of Pell grant awarded of $2.58 to $2.72 depending
on the model. Put another way, institutions that relied more heavily on state appropriations as a
source of revenue, had lower average Pell Grant amounts. The state college age population was a
statistically significant predictor at the p<.05 level for all models. While statistically significant,
the effect was small with limited practical significance. For each 10,000 person increase in
college age population, there was an expected decrease of $0.30 in the average amount of Pell
Grant awarded. In other words, the college age population in a state would have to increase by
over 32,000 people for an expected $1 decrease in average Pell Grant awarded.
The R-squared value shows that about 85% of the variance of average Pell Grant awarded
to students can be explained by variables in the model. One notable difference can be seen
between the models with different lags. The performance funding variable varied from 32.411 to
-2.649 with the highest value occurring with the one year lag and the lowest value from the three
year lag. However, as noted above, none of these were statistically significant. Models utilizing
a performance funding count variable demonstrated similar results with two exceptions in

79

statistical significance. Published in district tuition rate was significant at higher levels for the
performance funding count models with one and two year lag models significant at the p<.01
level and the models with no lag or a three year lag significant at the p<.05 level. Additionally,
the state college age population was significant at the p<.01 level for the two year lag model for
the performance funding count variable. Full results for the performance funding count variable
models are available in the appendix.
Table 6 Results for the Effects of Binary Performance Funding on Pell Grant Amount
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance funding 26.674
32.411
27.799
-2.649
(binary)
(18.503)
(18.034)
(16.603)
(20. 525)
Institutional Aid
3.564
3.403
3.381
3.932
Amount
(7.532)
(7.534)
(7.585)
(7.621)
Tuition
-22.697*
-22.751*
-23.366*
-22.722*
(9.153)
(9.194)
(9.253)
(9.213)
Appropriations as
-2.680*
-2.577*
-2.611*
-2.722*
Percent of Revenue
(1.251)
(1.245)
(1.259)
(1.269)
State Need Based
77.622
79.193
76.708
73.248
Aid
(52.578)
(53.402)
(52.983)
(52.711)
College Age
-0.298*
-0.280*
-0.295*
-0.294*
Population
(0.119)
(0.116)
(0.117)
(0.118)
Appropriations per
8.308
9.615
9.847
8.031
FTE
(7.854)
(8.073)
(7.943)
(7.770)
Affirmative Action
7.944
18.688
19.609
7.865
Ban
(46.583)
(46.84)
47.669
47.418
R-squared (within)
.8546
.8547
.8546
.8544
n
3324
3324
3324
3324
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Taken together, the models described above demonstrated that performance funding had
no statistically significant effect on Pell Grant enrollment through all models, variables, and lags.
This analysis showed that performance funding had no statistically significant effect on the
percent of FTFT students who receive Pell Grants or the average amount of Pell Grant awarded
when not differentiating by institutional type. This finding held true for both performance

80

funding count variables and binary performance funding variables. Additionally, this finding was
consistent across models with no lags and across one to three year lag models. Further, tests of
robustness were conducted which utilized the percent of students who received all federal aid
and the average amount of federal aid received to account for the years of missing data. These
models also found no statistically significant relationship between performance funding and Pell
Grant enrollment.
Underserved Minority Student Variables
Tables 7 through 9 present the results of research question 1b) Does performance funding
have an influence on the enrollment of underserved minority students? This study explored the
research question through two dependent variables: 1) the total number of FTFT students who
identify as underserved minority students; and 2) the percent of FTFT students who identify as
underserved minority students. In order to create the underserved minority variables, students
who identified as Black/African American, Hispanic/Latino, and Native American/Alaska Native
were added together. As data was available across the whole time frame of this study, a fixed
effects panel analysis was conducted over the 2000-2014 time frame.
Total underserved minority students. Table 7 presents the results using a binary
performance funding variable. Through all models, performance funding had no statistically
significant effect on FTFT enrollment of underserved minority students. Four variables
consistently influenced the total FTFT underserved minority enrollment: tuition rate,
appropriations as a percent of total revenue, state need based aid, and the state underserved
minority college age population. Published in state tuition for the prior year was a statistically
significant predictor for all models (p<.01). For every $1,000 increase in tuition, the institution
had an expected increase of 14 underserved minority students enrolled FTFT. Appropriations as

81

a percent of revenue was a statistically significant predictor across all models at the p<.001 level.
For every 1% increase in the share of revenue that came from state appropriations, the institution
had an expected decrease of almost two underserved minority students enrolled FTFT.
State variables also contributed to the enrollment of underserved minority students. The
amount of state need based aid provided per FTE was a statistically significant predictor at the
p<.05 level for all but the no lag model. For every $1,000 increase in state need based aid per
FTE the institution had an expected increase of 31 students who identify as underserved
minorities enrolled FTFT. The state college age underserved minority population was a
statistically significant predictor of underserved minority enrollment at the p<.001 level across
all models. For every 1,000 person increase in underserved minority residents aged 18-24 in the
state, there was a corresponding 1.3 person expected increase in underserved minority students
enrolled at the institution. The models indicate that the factors that were the strongest predictors
of underserved minority enrollment were financial based, for both students and institutions.
Tuition rate and need based aid both increased the number of underserved minority students,
indicating that cost is a significant factor in underserved minority enrollment. Institutional
reliance on state appropriations (a measure of institutional financial need) was a strong negative
predictor of underserved minority enrollment. In other words, institutions that relied heavily on
state appropriations had lower predicted rates of underserved minority enrollment; implying that
institutional revenue and financial need effect enrollment demographics.
The models R squared values show that 40% of the variance in the total number of
underserved minority students enrolled FTFT is explained by the variables included in the
model. Models utilizing a performance funding count variable produced similar results with two
notable exceptions. First, the performance funding count variable produced negative betas for

82

the effect of performance funding whereas the models with a binary variable estimated positive
betas. Second, tuition rate was significant at the p<.05 level for all performance funding count
models. Full results from the performance funding count variable can be found in the appendix.
Table 7 Results for the Effects of Binary Performance Funding on USM Total
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
7.787
3.900
2.765
2.497
Funding (Binary) (8.432)
(8.296)
(8.666)
(8.786)
Institutional Aid
Amount
Tuition

-2.564
(3.060)
13.824**
(5.193)
-1.744***
(0.498)

-2.489
(3.067)
13.795**
(5.204)
-1.756***
(0.498)

-2.469
(3.068)
13.782**
(5.197)
-1.761***
(0.498)

Appropriations
as Percent of
Revenue
State Need Based 29.570
30.722*
30.877*
Aid
(15.199)
(14.922)
(14.877)
College Age
1.297***
1.293***
1.292***
USM Population (0.139)
(0.138)
(0.137)
Appropriations
-1.349
-1.470
-1523
per FTE
(4.183)
(4.172)
(4.148)
Affirmative
3.647
4.413
4.430
Action Ban
23.053
23.385
23.397
R-squared
.3969
.3967
.3966
(within)
n
7016
7016
7016
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001

-2.464
(3.075)
13.797 **
(5.208)
-1.762***
(0.498)
30.914*
(14.865)
1.291***
(0.137)
-1.583
(4.086)
4.309
23.333
.3966
7016

Percent underserved minority students. Table 8 presents the results of a binary
performance funding variable on the percent of FTFT students enrolled who identify as
underserved minorities. Performance funding had a statistically significant negative result at the
p<.05 level for the model with no lag model and at the p<.01 level for the one year lag model.
With no lag, an institution with performance funding had an expected FTFT underserved
minority enrollment level .64% lower than without performance funding. Using the mean FTFT
student enrollment across all institutions in this study (1,742) this translates to about 11 fewer

83

underserved minority students. The one year lag model had an expected decrease of 0.70%, or an
expected decrease of 12 students. Models with two and three year lags found no statistically
significant relationship between performance funding and percent enrollment of underserved
minority students.
Table 8 Results for the Effects of Binary Performance Funding on USM Percent
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
-0.636*
-0.698**
-0.477
-0.332
Funding (Binary) (0.257)
(0.267)
(0.278)
(0.282)
Institutional Aid -0.318**
-0.320**
-0.324**
-0.325 **
Amount
(0.102)
(0.102)
(0.102)
(0.102)
Tuition
0.413*
0.408*
0.412*
0.414*
(0.181)
(0.180)
(0.180)
(0.180)
Appropriations
-0.076***
-0.076***
-0.076***
-0.076***
as Percent of
(0.020)
(0.020)
(0.020)
(0.020)
Revenue
State Need Based 2.444***
2.372***
2.362***
2.364***
Aid
(0.555)
(0.555)
(0.556)
(0.555)
College Age
0.157***
0.161***
0.161***
0.161***
USM Percent
(0.029)
(0.029)
(0.029)
(0.029)
State
-0.147
-0.146
-0.135
-0.126
Appropriations
(0.171)
(0.171)
(0.171)
(0.171)
per FTE
Affirmative
-0.834
-0.922
-0.931
-0.067
Action Ban
(0.630)
(0.635)
(0.631)
(0.176)
R-squared
.2895
.2899
.2886
.2880
(within)
n
7016
7016
7016
7016
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Five variables were statistically significant predictors of percent of underserved minority
students enrolled: institutional aid amount, tuition rate, state appropriations as a percent of
revenue, state need based aid, and the state percent 18-24 underserved minority population. The
amount of institutional aid provided was a statistically significant predictor at the p<.01 level
across all models. For every $1,000 increase in institutional aid, there was an expected decrease
of about 0.32% in percent of underserved minorities enrolled. Utilizing the mean enrollment in

84

this study, this translates to about six less underserved minority students. Published in state
tuition rate was a significant predictor at the p<.05 level across all models. For every $1,000
increase in tuition the percent of students who identified as underserved minorities would
increase by and expected 0.41%, or about seven students. Appropriations as a percent of revenue
was significant at the p<.001 level. For every 1% increase in reliance on state appropriations, the
institution had an expected decrease in the percent of underserved minorities enrolled of 0.076%,
which translates to about one student.
Two state variables were statistically significant predictors at the p<.001 level: state need
based aid and state percent college age underserved minority population. For every $1,000
increase in state need based aid provided per FTE there was an expected increase of 2.4% in the
percent of underserved minority students enrolled. Converted to students, for every $1,000
increase in state need based aid, there was an expected increase of 35 students. For every 1%
increase in the percent of underserved minority college age residents in the state there was an
expected increase of 0.16% in the percent of underserved minority students enrolled, or about
three students. The R-squared value demonstrates that about 29% of the variance in the percent
of underserved minority students enrolled is explained by the variables included in the models.
Conducting the models with a performance funding count variable reveal similar
findings. Table 9 presents the results of a performance funding count variable on the percent of
FTFT students enrolled who identify as underserved minorities. Performance funding is found to
have a statistically significant negative result across all models at the p<.01 level except for the
three year lag model, which was significant at the p<.05 level. With no lag, for each year of
performance funding an institution had an expected decrease in underserved minority enrollment
of 0.17%. Using the mean FTFT student numbers across all institutions in this study, results in

85

an expected decrease of about three students a year. With performance funding lasting an
average of five years across all institutions in this study, that translates to an expected decrease
of 15 students in the last year. Models with one and two year lags lead to similar decreases. The
model with a three year lag had an expected decrease in underserved minority enrollment of
0.14% per year, or 2.4 less students a year and about 12 students by the end of the average five
year length of performance funding.
Table 9 Results for the Effects of Performance Funding Count on USM Percent
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
-0.175**
-0.173**
-0.165**
-0.139*
Funding (Count) (0.058)
(0.059)
(0.060)
(0.059)
Institutional Aid -0.308**
-0.311**
-0.315**
-0.320**
Amount
(0.1003)
(0.100)
(0.100)
(0.101)
Tuition
0.389*
0.392*
0.392*
0.397*
(0.176)
(0.176)
(0.176)
(0.177)
Appropriations
-0.079***
-0.078***
-0.077***
-0.076***
as Percent of
(0.019)
(0.019)
(0.019)
(0.0194)
Revenue
State Need Based 2.455***
2.414***
2.394***
2.350***
Aid
(0.557)
(0.558)
(0.556)
(0.554)
College Age
0.153***
0.155***
0.157***
0.158***
USM Percent
(0.029)
(0.029)
(0.029)
(0.029)
State
-0.155
-0.145
-0.123
-0.107
Appropriations
(0.170)
(0.170)
(0.169)
(0.169)
per FTE
Affirmative
-1.082
-1.122
-0.991
-0.875
Action Ban
(0.635)
(0.639)
(0.629)
(0.621)
R-squared
.2937
.2932
.2925
.2908
(within)
n
7016
7016
7016
7016
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Similar to the binary models, the performance funding count models had five variables
that were statistically significant predictors of the percent of FTFT students who identified as
underserved minorities: institutional aid amount, tuition rate, state appropriations as a percent of

86

revenue, state need based aid, and state percent 18-24 underserved minority population. All had
similar significance levels and betas as the binary performance funding models.
To verify the findings, a Tobit analysis was performed as a check of robustness. Since a
percent variable is limited to values between 0 and 100 percent, the data in this study is
considered left and right censored. Therefore a double bounded Tobit analysis was performed
with lower limits of 0 and upper limits of 100. Performance funding was a significant predictor
at the p<.001 level for most models, with the exception of the binary two year lag model, which
was significant at the p<.01 level and the binary three year lag model, which was significant at
the p<.05 level. All models had similar coefficients for both the variables of interest and the
control variables. However the control variables were significant at higher confidence intervals.
All control variables were significant at the p<.001 level accept for state appropriations which
was significant at the p<.05 level for the binary no lag, 1 year lag, and 2 year lag models as well
as the performance funding count no lag and one year lag models. State appropriations was not a
statistically significant predictor for other models. The Tobit analysis showed only 3 censored
observations and 7013 uncensored observations for all models, implying limited concern with
censored variables. Since Tobit analysis does not allow the inclusion of state or year fixed
effects (though state and year were included as variables in the Tobit models estimated) or robust
standard errors and the Tobit analysis found limited censored observations, the panel analysis is
preferred. Results of Tobit models are included in the appendix.
In summary, performance funding was not a statistically significant predictor of total
underserved minority FTFT enrollment. However, performance funding was found to be a
statistically significant predictor of the percent enrollment of underserved minority students in
almost all models executed. Performance funding was a consistent negative predictor for all four

87

models calculated with a performance funding count variable and was a negative predictor for
both the no lag and one year lag models for the binary performance funding variable. Two year
and three year lag count models had negative betas, but were not statistically significant.
Additionally, using a double bounded Tobit model, performance funding was again a statistically
significant negative predictor for all models calculated. These findings provide strong evidence
that performance funding influences the enrollment of underserved minority students.
Institutional Type and Performance Funding
This section addresses research question 2) Does institutional type influence the effect of
performance funding on underserved students’ enrollment at public four-year institutions? To
answer this research question, two categorical variables were created: one for Carnegie
classification and one for selectivity. As the time period of this study includes three different
Carnegie classification codes (2000, 2005, 2010), recoding the variables allows for analysis
across all years in the study.
The IPEDS Carnegie variable was recoded into a 10 category Carnegie categorical
variable. Three codes cover doctoral institutions: Doctoral A (those listed as “very high research”
or “extensive”), Doctoral B (those listed as “high research” or “intensive”) and Doctoral C (those
listed as “doctoral/research universities”). Similarly, masters institutions are included in three
codes: Masters A (those listed as “Masters I” or “larger programs”), Masters B (those listed as
“Masters II” or “medium programs”), and Masters C (those listed as “smaller programs”).
Baccalaureate institutions also span three codes: Baccalaureate A (those listed as “general” or
“diverse fields”), Baccalaureate B (those listed as “liberal arts” or “arts and sciences”), and
Baccalaureate C (those listed as “baccalaureate/associate”). The final category, Special Focus

88

Institutions, includes “schools of art, music and design,” “schools of engineering,” and “teachers
colleges.”
Table 10 shows the transition rates for the Carnegie categorical variable. Most
institutions did not move across Carnegie classification codes, with between 80.00% to 96.88%
staying in their Carnegie classification across all 15 years of this study. In fact, outside of
special focus institutions (90.41% staying) and Baccalaureate/Associates Colleges (80%
staying), most institutions had a staying rate above 93%.
Table 10 Transition Rates Across Carnegie Categorical Variable
Doct Doct Doct Mast Mast Mast Baccala Baccal
oral
oral
oral
ers
ers
ers
ureate A aureate
A
B
C
A
B
C
B
Doctoral
A
Doctoral
B
Doctoral
C
Masters
A
Masters
B
Masters
C
Baccala
ureate A
Baccala
ureate B
Baccala
ureate C
Special
Focus

3.79

0.00

0.00

0.00

0.00

0.00

0.00

Bacc
alaur
eate
C
0.00

2.48

.31

.21

0.00

0.00

0.00

0.00

.10

0.00

95.5
5
3.24

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

.11

96.7
6
.41

2.30

0.00

.15

.04

0.00

0.00

0.00

0.00

.15

96.3
8
2.67

2.53

.89

.45

0.00

0.00

0.00

0.00

0.00

.35

93.3
1
4.61

.35

1.06

0.00

0.00

0.00

0.00

0.00

.14

.28

93.6
2
1.42

96.88

1.14

.14

0.00

0.00

0.00

0.00

0.00

.26

2.11

1.58

96.04

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

20.00

0.00

80.00

0.00

0.00

2.74

0.00

0.00

2.74

0.00

4.11

0.00

0.00

90.41

96.2
1
1.35

Special
Focus
0.00

To create the selectivity categorical variable, the IPEDS admission rate variable was
transformed into five categories. Institutions that admit less than 30% of the students who apply
were coded as very highly selective institutions. Institutions with admission rates between 30

89

and 49.99% were coded as highly selective. Institutions that admit between 50 and 74.99% of
the students that apply were coded as selective. Institutions with admission rates between 75%
and 84.99% were coded as moderately selective and institutions that admitted at least 85% of the
students that applied were coded as open access institutions. As can be seen in Table 11, the
selectivity categorical variable had higher rates of transfer between categories. Between 62.39%
and 79.75% of institutions remained in their selectivity category across all years. Moderately
selective institutions and very highly selective institutions remained at the lowest rates with
institutions staying 63.21% and 62.39% of the time respectively. Selective institutions had the
highest rates of staying at 79.75%.
Table 11 Transition Rates Across Selectivity Categorical Variable
Open Access Moderately
Selective
Highly
Selective
Selective
Open Access 73.64
17.71
7.03
1.10
Moderately
13.62
63.21
22.28
.89
Selective
Selective
2.97
9.92
79.75
6.95
Highly
2.01
2.55
20.00
71.41
Selective
Very Highly
2.56
0.00
11.97
23.08
Selective

Very Highly
Selective
.51
0.00
.41
4.03
62.39

Carnegie Classification
To address research question 2a: Does Carnegie classification influence the effect of
performance funding on underserved students’ enrollment at public four year institutions, eight
models were estimated. To start, models exploring how Carnegie classification may influence
the effect of performance funding on Pell Grant dependent variables were calculated with both
performance funding binary and count variables. Then models were executed to explore how
Carnegie classification may influence the effect of performance funding on underserved minority
enrollment variables. Each is presented in turn below. All models included the same controls

90

and fixed effects as the main models. As the control variables in these models had similar betas
and significance levels as the main models, only interaction effects are presented below
Carnegie classification and Pell Grant variables. To explore if the effect of
performance funding on Pell Grant variables varies by Carnegie classification, models were
estimated with an interaction effect between Carnegie classification and performance funding
both in a binary and count form. Models were estimated with both average Pell Grant amount as
well at the percent of FTFT students who receive Pell Grants as the dependent variable. The
results are presented in Table 12 with Doctoral A institutions (those classified as very high
research or research extensive) without performance funding as the comparison group.
Utilizing a performance funding binary variable, only one Carnegie classification was a
statistically significant predictor of the average amount of Pell Grant awarded. Doctoral A
institutions that were subject to performance funding had an expected average Pell grant $75.01
lower (significant at the p< .05 level). Utilizing a count variable for performance funding, only
special focus institutions with performance funding were a statistically significant predictor at
the p< .001 level. These institutions had a predicted average Pell Grant amount that was $64.41
lower.
For percent of FTFT students who received Pell Grants, two Carnegie categories were
statistically significant predictors with the binary performance funding variable, and three were
statistically significant predictors with the performance funding count variable. For the binary
performance funding variable, the Doctoral A institutions were a statistically significant
predictor at the p<.05 level and special focus institutions were at the p<.001 level. Doctoral A
institutions subject to performance funding had a predicted enrollment of Pell Grant students that
was 1.47% lower. Utilizing mean enrollment for Doctoral A institutions (4,084) this translates to

91

about 60 fewer students. Special focus institutions with performance funding had a predicted
enrollment of Pell Grant students 4.85% higher or utilizing their mean enrollment (305) about 15
students.
Table 12 Results of Performance Funding Interacted with Carnegie on Pell Grant
Variables
Performance Funding Binary
Performance Funding Count
Pell Grant
Percent FTFT
Pell Grant
Percent FTFT
Amount
Pell Grant
Amount
Pell Grant
Doctoral A
-75.0136*
-1.472 *
-3.367
-0.481*
(29.614)
(0.718)
(5.727)
(0.203)
Doctoral B
22.879
0.174
13.187
-0.104
(93.265)
(1.164)
(7.629)
(0.139)
Doctoral C
-47.094
-0.443
6.642
0.304**
(105.941)
(2.506)
(4.515)
(0.107)
Masters A
-199.416
-3.245
-0.939
-0.024
(121.230)
(2.698)
(4.509)
(0.099)
Masters B
-189.393
-3.200
1.769
-0.162
(128.129)
(2.893)
(5.658)
(0.143)
Masters C
-276.822
-4.694
-11.923
-0.221
(140.783)
(3.069)
(7.501)
(0.125)
Baccalaureate A
-198.222
-3.438
-4.739
0.215
(147.792)
(3.104)
(13.360)
(0.151)
Baccalaureate B
-33.013
-3.495
56.631
0.115
(162.335)
(3.304)
(46.774)
(0.366)
Baccalaureate C
Special Focus
104.311
4.849***
-64.405***
-1.439***
Institutions
(95.876)
(1.225)
(4.867)
(0.149)
R squared
.8565
.6247
.8558
.6255
n
3322
3322
3322
3322
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Note: no institutions classified as Baccalaureate C are included in performance funding states,
therefore those results are blank.
Utilizing a performance funding count variable three Carnegie categories were
statistically significant predictors of percent of FTFT students who receive Pell Grants. Doctoral
A institutions were statistically significant predictors at the p< .05 level and had predicted
enrollments of Pell Grant students .48% lower for each year of performance funding. Utilizing

92

mean enrollment this translates to 20 fewer students each year or about 98 students fewer in the
last year utilizing the average five year length of performance funding. Doctoral C institutions
were statistically significant predictors at the p<.01 level and had predicted Pell Grant enrollment
.30% higher for each year of performance funding or utilizing the enrollment mean (1879) about
6 students per year and 28 students by the end of the average five year length of performance
funding. Finally, special focus institutions were statistically significant predictors at the p<.001
level and had expected Pell Grant enrollment 1.44% lower or utilizing the mean enrollment,
about 4 students fewer per year and 23 students by the end of the average length of performance
funding.
Carnegie classification and underserved minority student variables. To explore if the
effect of performance funding on underserved minority student enrollment varies by Carnegie
classification, models were estimated with an interaction effect between Carnegie classification
and performance funding both in a binary and count form. Models included both the total
underserved minority population and the percent of the FTFT class that identified as underserved
minority students. Table 13 presents the results from these models with Doctoral A institutions
without performance funding as the comparison group.
Utilizing an interaction between a binary performance funding variable and Carnegie
classification, two Carnegie categories were statistically significant predictors for total
underserved minority enrollment. Doctoral A institutions with performance funding had a
predicted enrollment of about 64 more underserved minority students with significance at the
p<.05 level. Doctoral B institutions (those classified as high research activity or research
intensive) with performance funding had predicted enrollment of underserved minority students
almost 99 students higher (significant at the p<.001 level).

93

Table 13 Results of Performance Funding Interacted with Carnegie on USM Variables
Performance Funding Binary
Performance Funding Count
USM Total
USM Percent
USM Total
USM Percent
Doctoral A
63.624*
-0.213
4.406
-0.061
(28.551)
(.644)
(3.819)
(0.094)
Doctoral B
99.098***
0.829
4.597
-0.024
(26.704)
(.604)
(2.807)
(0.069)
Doctoral C
53.460
1.513
4.142*
-0.015
(52.291)
(1.157)
(2.022)
(0.082)
Masters A
62.860
1.132
-2.909
-0.221**
(60.697)
(0.943)
(1.861)
(0.081)
Masters B
16.620
0.054
-4.785**
-0.287***
(63.440)
(1.083)
(1.645)
(0.082)
Masters C
-0.249
-0.870
-6.785**
-0.369
(64.913)
(1.644)
(2.546)
(0.206)
Baccalaureate A
34.455
1.464
-1.489
-0.158
(63.761)
(1.282)
(1.887)
(0.141)
Baccalaureate B
16.090
0.350
-4.818
-0.058
(71.082)
(1.725)
(4.531)
(0.195)
Baccalaureate C
Special Focus
100.568
-0.806
-2.841
-0.839
Institutions
(60.809)
(4.558)
(4.851)
(0.841)
R squared
.4119
.2996
.4009
.2969
n
6992
6992
6992
6992
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Note: no institutions classified as Baccalaureate C are included in performance funding states,
therefore those results are blank.
Utilizing an interaction between a performance funding count variable and Carnegie
classification resulted in three statistically significant predictors of underserved minority
enrollment: Doctoral C, Masters B, and Masters C institutions. Doctoral C institutions were
significant at the P<.05 level and had a predicted underserved minority student enrollment about
4 students higher per year of performance funding or about 20 students by the end of the average
length of performance funding. Masters B institutions (those classified as medium programs or
Masters II) with performance funding had a predicted enrollment rate of underserved minority
students almost five students lower for each year of performance funding or 25 students by the

94

end of the average length of the policy (significant at the p<.01 level). Masters C institutions had
a predicted enrollment rate of underserved minority students almost seven students lower for
each year of performance funding or 35 student in the last year of the average length of the
policy (significant at the p<.01 level).
Exploring the effect of the interaction between performance funding count and Carnegie
classification on underserved minority percent produced similar results, with two statistically
significant predictors at the p<.01 level. Masters A institutions (those classified as larger
programs or Masters I) with performance funding had predicted enrollment of underserved
minority students .22% lower for each year of performance funding or utilizing mean enrollment
(1433) about three less students a year and 15 less a year by the end of the average length of
performance funding. Masters B institutions with performance funding had predicted enrollment
of underserved minority students .29% lower for each year of performance funding or using
mean enrollment (880) about 2.5 fewer students a year and 12 fewer students in the last year of
the average length of performance funding. The interaction between binary performance funding
and Carnegie classification produced no statistically significant predictors for underserved
minority percent.
Overall, the models above demonstrated that Carnegie classification might influence the
effect of performance funding on underserved student enrollment. The percent of Pell Grant
students enrolled varied by institutional type with two Carnegie categories at statistically
significant levels for a binary performance funding interaction and three at statically significant
levels for a performance funding count interaction. Utilizing a binary performance funding
interaction term, Doctoral A institutions with performance funding had predicted enrollment of
Pell Grant students lower than the reference group and special focus institutions with

95

performance funding had higher predicted enrollment of Pell Grant students. Utilizing a
performance funding count interaction, Doctoral A institutions again had lower predicted
enrollment of Pell Grant students per year. Additionally, Doctoral C institutions had higher
predicted enrollment of Pell Grant students and special focus institutions had lower predicted
enrollment of Pell Grant students per year of performance funding.
Institutions again demonstrated divergent responses to performance funding in regards to
enrollment of underserved minority students. A performance funding binary count interaction
resulted in two Carnegie classifications with statistically significant enrollment patterns. Both
Doctoral A and Doctoral B institutions had higher predicted total enrollment of underserved
minority students. Utilizing a performance funding count interaction two institutions had
enrollment patterns statistically significant from the control group. Masters B and Masters C
institutions both had lower predicted enrollment for total underserved minority students per year.
Similarly, Masters A and Masters B institutions with performance funding had lower predicted
percent of underserved minority students enrolled for each year of performance funding. Taken
together, these results revealed that institutions may respond differently to performance funding
based on Carnegie classification.
Selectivity
To answer research question 2b: Does selectivity influence the effect of performance
funding on underserved students’ enrollment at public four year institutions, eight models were
estimated. First, models exploring how selectivity may influence the effect of performance
funding on Pell Grant dependent variables were calculated with performance funding binary and
count variables. Then models were executed to explore how selectivity may influence the effect

96

of performance funding on underserved minority enrollment variables again for both binary and
count performance funding variables.
Selectivity and Pell Grant variables. To explore if the effect of performance funding on
Pell Grant variables varies by selectivity, models were estimated with an interaction effect
between a categorical selectivity variable and performance funding both in a binary and count
form. Models were estimated with both Pell Grant amount as well at Pell Grant percent as the
dependent variable. The results are presented in Table 14 with open access institutions without
performance funding as the comparison group.
Table 14 Results of Performance Funding Interacted with Selectivity on Pell Grant
Variables
Performance Funding Binary
Performance Funding Count
Pell Grant
Percent FTFT
Pell Grant
Percent FTFT
Amount
Pell Grant
Amount
Pell Grant
Open Access
50.060
-0.325
-0.311
-0.063
(35.005)
(0.462)
(4.405)
(0.080)
Moderately
82.640**
-0.559
2.119
-0.034
Selective
(29.846)
(0.552)
(4.058)
(0.090)
Selective
81.434*
-0.792
2.372
-0.072
(32.322)
(0.543)
(4.568)
(0.094)
Highly Selective
61.304
-0.184
0.249
-0.114
(56.668)
(0.831)
(4.883)
(0.116)
Very Highly
233.543*
1.035
12.950
-0.363*
Selective
(94.591)
(1.800)
(8.082)
(0.151)
R squared
.8556
.6163
0.8546
.6158
n
3111
3111
3111
3111
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
The interaction between binary performance funding and selectivity was a statistically
significant predictor of average amount of Pell Grant awarded for three of the five selectivity
levels. The moderately selective institutions with performance funding had a predicted average
Pell Grant $82.64 higher (significant at the p<.01 level). The selective institutions with
performance funding had a predicted average Pell Grant amount $81.43 higher (significant at the

97

p<.01 level). The very highly selective institutions with performance funding had a predicted
average Pell Grant $233.54 higher (significant at the p<.01 level). The interaction between the
performance funding count variable and selectivity did not produce any statistically significant
predictors for the average Pell Grant amount awarded.
The model with an interaction between a binary performance funding variable and
selectivity did not produce any statistically significant predictors for percent of FTFT students
with Pell Grants. Utilizing a performance funding count variable resulted in one statistically
significant predictor at the p<.05 level for the percent of FTFT Pell Grant awardees. Very highly
selective institutions with performance funding had a predicted enrollment of Pell Grant students
.36% lower for each year of performance funding. Utilizing the mean enrollment for very highly
selective institutions (1,840), this translates to about seven less students a year or about 33
students by the last year of the average length of the policy.
Selectivity and underserved minority student variables. To explore if the effect of
performance funding on underserved minority student enrollment varies by admission rate,
models were estimated with an interaction effect between a categorical selectivity variable and
performance funding both in a binary and count form. Models included both the total
underserved minority population and the percent of the FTFT class that identified as underserved
minority students. Table 15 presents the results from these models with open access institutions
without performance funding as the comparison group.
Calculating a model with the interaction of binary performance funding and selectivity
resulted in only one statistically significant predictor of underserved minority student enrollment.
Very high selective institutions with performance funding had predicted enrollment of
underserved minority students 140 students lower (significant at the p<.01 level). Utilizing the

98

same interaction on underserved minority percent enrollment produced similar results, with both
high selective institutions and very high selective institutions being statistically significant
predictors. The highly selective institutions with performance funding had predicted enrollments
of underserved minority students 1.10% lower, with results significant at the p<.05 level.
Utilizing enrollment mean (1,749), this translates to about 19 fewer students. The very highly
selective institutions with performance funding had predicted enrollments of underserved
minority students 5.66% lower with results significant at the p<.001 level. Utilizing the mean
enrollment for very highly selective institutions, this translates to about 104 less students.
Table 15 Results of Performance Funding Interacted with Selectivity on USM Variables
Performance Funding Binary
Performance Funding Count
USM Total
USM Percent
USM Total
USM Percent
Open Access
11.137
0.042
0.885
-0.050
(12.075)
(0.330)
(1.278)
(0.056)
Moderately
15.090
-0.342
0.935
-0.093
Selective
(14.456)
(0.405)
(1.377)
(0.073)
Selective
-1.303
-0.475
1.347
-0.073
(10.767)
(0.384)
(1.467)
(0.056)
Highly Selective -25.195
-1.101*
-0.789
-0.221*
(16.582)
(0.518)
(1.615)
(0.087)
Very Highly
-140.280 **
-5.662***
-17.118***
-0.732***
Selective
(46.965)
(1.311)
(3.899)
(0.080)
R squared
.4255
.3266
.4240
.3308
n
6156
6156
6156
6156
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Estimating the models with an interaction between performance funding count and
selectivity produced similar results. Very highly selective institutions with performance funding
had predicted enrollment of underserved minority students 17 students lower for each year with
performance funding (significant at the p<.001 level). Extrapolated out to the average length of
performance funding, results in 86 fewer students by the last year of performance funding.
Utilizing the same interaction on the percent of underserved minority students, both highly

99

selective institutions and very highly selective institutions were statistically significant
predictors. Highly selective institutions with performance funding had expected underserved
minority enrollment .22% lower for each year of performance funding with results significant at
the p<.05 level. Utilizing mean enrollment, this translates to about four students a year or 19 at
the end of the average length of performance funding. Very highly selective institutions had
predicted enrollment of underserved minority students .73% lower for each year of performance
funding or about 13 fewer students a year or 67 students fewer the last year of the average length
of performance funding, with results significant at the p<.001 level.
Taken together, the models above demonstrated clear evidence that selectivity influences
the effect of performance funding on underserved student enrollment for both Pell Grant eligible
students and underserved minority students. A binary performance funding interaction showed
that three of five selectivity categories were statistically significant predictors for the average
amount awarded in Pell Grants. Additionally, the very highly selective category was a
statistically significant negative predictor for percent of FTFT students who receive Pell Grants.
Further, the very highly selective category was consistently a statistically significant negative
predictor for both total underserved minority enrollment and the percent of underserved minority
students enrolled. This finding held true for both a binary performance funding interaction as
well as a performance funding count interaction. For percent of underserved minority students
enrolled the highly selective category was also a negative statistically significant predictor.
Together these results demonstrated that more selective institutions respond to performance
funding differently than less selective institutions.

100

Conclusion
These findings demonstrate that performance funding does have the potential to influence
enrollment profiles at U.S. public four-year institutions. Specifically, this study found it changed
the enrollment of underserved minority students. Further, these influences may not be equitable
across all institutions, and instead may effect lower status institutions in a different manner than
higher status institutions. Specifically, those with more flexibility in their enrollment profile
may be more likely to change their enrollment of both Pell Grant students and underserved
minority students. These findings have profound implications for higher education institutions,
policy formation, and social equity. The next chapter discusses these findings in context to
previous research, how these findings may affect the goals of performance funding, higher
education institutions, and social equity writ large, and presents implications for both policy
creators and policy researchers.

101

Chapter 5
Summary of Study
Performance funding policies are increasing in prevalence as state legislators push for
greater efficiency and specific quantifiable outcomes from higher education institutions.
Advocates of these policies claim they increase efficiency by rewarding reductions in cost and
promoting desired outcomes. However, this claim assumes that higher education operates within
a competitive marketplace where the rules and assumptions of market economics (such as
rational choice theory or perfect knowledge) can be met (Gumport, 2001). Yet, many higher
education researchers believe standard economic market assumptions do not apply to institutions
of higher education and believe policies and decisions based on these assumptions are flawed
(e.g., Gumport, 2001; Marginson, 1997; McMahon, 2009; Slaughter & Rhoades, 2004). If higher
education does not function as a traditional market, then instituting performance funding policies
could create quasi markets, where market behavior arises from the priorities of the policy
makers, as opposed to true economic efficiency and supply and demand of resources (Taylor,
Cantwell & Slaughter, 2013).
If performance funding policies value efficiency (increased economic outputs at lower
costs) and quasi markets reward institutions that engage in behavior valued by the policy, then
institutions may engage in behaviors to increase their quality and decrease their costs in order to
reap the benefits of the performance funding policy. However, due to the iron triangle of cost,
quality, and access (Immerwahr et al., 2008), there may be unintended outcomes from these
behaviors. Specifically, the quasi markets created by performance funding policies may create
incentives for institutions to attend to the metrics measured by policymakers (efficiency) and
give less attention to the metrics for which institutions are not held accountable (underserved

102

student enrollment). In other words, performance funding may cause institutional leaders to feel
they must change enrollment profiles (Immerwahr et al., 2008).
Limited research has addressed this phenomenon and explored how higher education
institutions are balancing the competing needs of demonstrating efficient economic outcomes via
performance funding while maintaining access and success for all students, particularly in the
four-year sector. As such, this study explored how performance funding influences the
enrollment of underserved students with a focus on the following research questions:
1) Are there unintended consequences from performance funding policies related to
underserved students’ enrollment at public four-year institutions?
a. Does performance funding have an influence on the enrollment of Pell Grant
eligible students?
b. Does performance funding have an influence on the enrollment of underserved
minority students?
2) Does institutional type influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
a. Does Carnegie classification influence the effect of performance funding on
underserved students’ enrollment at public four-year institutions?
b. Does selectivity influence the effect of performance funding on underserved
students’ enrollment at public four-year institutions?
In order to examine the influence of performance funding on underserved student
enrollment, this study utilized OLS based panel analysis for the years 2000-2014. While most
performance funding research addresses state level questions, this study specifically focused on
institutional level behavior. To answer the research questions, models with four different

103

dependent variables were calculated: average amount of Pell Grant awarded, percent of FTFT
students who receive Pell Grants, total underserved minority enrollment, and percent of FTFT
students who identify as underserved minority students. Models with zero, one, two, and three
year lags were calculated as well as models with both a binary and count performance funding
variable. All models included controls for institutional and state variables as well as robust
standard errors. Further, institutional and year fixed effects were included in all models. Models
were calculated without institutional differentiation and with institutions categorized by Carnegie
classification and selectivity. A total of 48 main models were utilized for this study.
Discussion of Findings
Underserved Student Enrollment
Previous research has raised concerns that performance funding may change enrollment
patterns, particularly that institutions may decrease enrollment of underserved students (Bell,
2005; Colbeck, 2002; Dougherty et al., 2014b; Lahr et al., 2014; Naughton, 2004; Rutherford &
Rabovsky, 2014). However, previous research either only raised this concern without
confirming whether this change occurred and was qualitative in nature. For instance, higher
education administrators are concerned that performance funding policies would result in
“creaming” or only enrolling students who would perform well on performance funding metrics
(Dougherty et al., 2014b; Lahr et al., 2014). However, this notion was only mentioned as a
possibility with no administrators stating it had been an intentional decision at their institution.
This study expands on current literature by quantitatively exploring if performance funding does
change enrollment patterns, in particular for underserved students. Findings from this study
justify the concern as results show that institutions do change their enrollment profiles when
subject to performance funding. Institutions with performance funding enrolled underserved

104

minority students at a lower percent than institutions not subjected to the policy, suggesting that
performance funding may indeed have unintended consequences.
This finding supports previous qualitative research that suggested that institutions may
engage in “creaming” or admitting students who are more likely to succeed in higher education
as a response to performance funding (Colbeck, 2002; Dougherty et al., 2014; Lahr et al., 2014).
However, previous qualitative research did not investigate if in fact performance funding does
lead to changes in enrollment profiles. Recently, one quantitative study did explore this specific
question. Umbricht, Fernandez, and Ortagus (2015), studied a variety of unintended outcomes in
relation to performance funding in Indiana. Specifically, through a difference in difference
analysis they found that performance funding decreased the enrollment of minority students in
Indiana. The findings of this current study expand on Umbricht et al.’s (2015) study and suggest
that Indiana is not the only state where these unintended outcomes are occurring.
Decreasing the enrollment of underserved minority students has several implications for
the objectives of performance funding policies, for higher education institutions specifically, and
for social equity broadly. Most performance funding policies include increasing the number of
residents with postsecondary credentials as one of their goals. This requires increasing higher
education completion rates but also often necessitates increasing the number of students
enrolling in higher education as well. If less underserved minority students are enrolling in
higher education institutions this can be detrimental to the state’s goal of increasing higher
education access and completion. Further, decreasing racial diversity in higher education is not
only antithetical to the purpose of higher education but also goes against the desires of many
performance funding policy creators. In fact, a few states have specifically tried to prevent
changes in underserved student enrollment by creating metrics to try to counteract this creaming

105

effect. Up to 13 states have included a metric or measure within the formula specifically
addressing underserved student enrollment (HCM, 2015). For instance, in recent years Michigan
added a metric that measured the enrollment of Pell Grant students. The increasing prevalence of
these metrics in recent years shows that policy creators are cognizant of the creaming effect and
are taking steps to ensure performance funding does not decrease diversity of students enrolled in
higher education.
The potential of decreasing underserved minority students enrollment is also detrimental
for higher education institutions and the goal of exposing students to diverse experiences while
in higher education in order to create beneficial learning experiences for all students. Racial
segregation has increased in the U.S. and most students now live in segregated communities
before entering higher education. Often times, higher education is the first time students are
exposed to the opportunity to learn from peers from different cultures, values, and experiences
(Gurin, Dey, Hurtado, & Gurin, 2011). Research shows that students’ exposure to diversity on
their campus is beneficial for a wide variety of student outcomes across all student racial
identities (Astin, 1993; Gurin et al., 2011; Hurtado, 2001). By decreasing the enrollment of
underserved minority students, performance funding may take away this beneficial learning
experience for many students.
Finally, this finding may have impacts on social equity. With performance funding
spreading across the country (NCSL, 2016) institutions responding by closing access to
underserved minority students could have serious repercussions to the whole higher education
system and educational attainment across the U.S.. Educational attainment in the U.S. for adults
25-34 is lagging behind other nations and in order to match the level of higher performing
nations the U.S. would need to increase the number of adults with college degrees by almost 8%

106

a year until 2020 (Heller, 2013; Perna, 2013). However, in order to do accomplish this,
institutions need to enroll a more diverse group of students as the high school graduate
population is becoming more racially diverse (NCES, 2014; WICHE, 2012). Excluding
underserved minority populations from enrolling in higher education limits the students eligible
to enroll in and complete higher education. Therefore, if performance funding policies limit
enrollment of underserved students, these policies may be antagonistic to the goal of increasing
U.S. higher education participation and completion. Additionally, research shows that a wide
variety of societal benefits (e.g. monetary returns to the state, working in underserved
communities, cultural engagement) are linked with students having diverse experience in higher
education (Gurin et al, 2011). Decreasing racial diversity in higher education would limit these
beneficial diverse experiences and therefore decrease societal benefits associated with higher
education such as increased monetary returns.
The fact that institutions with performance funding decreased their enrollment of
underserved students provides supporting evidence to the hypothesis that when faced with
pressure to increase efficiency (i.e. increase quality and decrease cost) from performance
funding, institutions responded by changing the enrollment of underserved minority students as
predicted by Immerwahr et al.’s (2008) iron triangle of cost, quality, and access. While this study
is not causal in nature and therefore the results can not be interpreted to mean that performance
funding caused this change in enrollment patterns, the finding does present further support and
evidence that institutional leaders feel that cost, quality and access are interrelated. As
institutional leaders feel the pressure of performance funding metrics, which ask for higher
quality without increases in costs, at least some institutions look to decisions around access for
the solutions.

107

One possible explanation for this phenomenon is that as institutions feel the financial
pressure of decreased state funding (SHEEO, 2015) they may feel forced to make decisions to
improve efficiency in order to perform better on performance funding measures. However since
institutional leaders often lack data to know what relationship, if any, exists between institutional
expenditure and outcomes (Powell et al., 2012) they may look elsewhere to determine how to
improve their efficiency. Performance funding policies have explicit metrics that outline how
the state views higher education performance and how they assess improvement in efficiency
(i.e. higher graduation rates). Therefore, this pressure could lead to institutional leaders feeling
the need to enroll students who historically have performed better on metrics measured by
performance funding (such as retention and graduation rates) with the hopes of improving their
score on performance funding metrics and therefore increasing their state funding. As research
has shown underserved minority students have lower retention and graduation rates overall
(NCES, 2014) this may result in institutions changing their enrollment to lower rates of
underserved minority students in order to perform well on the metrics in the performance
funding policy.
If institutions are responding to performance funding due to financial pressure,
institutions that rely on state appropriations more heavily may feel more pressure to perform
better on performance funding metrics in order to not see a decrease in their state funding.
Results from this study give credibility to this theory. In all models of this study, state
appropriations as a percent of institutional revenue is a strong negative predictor on the
enrollment of underserved minority students. The fact that appropriations as percent of
institutional revenue was a strong and consistent negative predictor of enrollment give credence
to Immerwahr et al.’s (2008) belief that cost, quality and access are linked in an iron triangle. As

108

institutions became more reliant on state appropriations (regardless of whether they are subject to
performance funding), they decreased their enrollment of underserved students, with most
models significant at the p<.001 level.
This study also explored if performance funding influences the enrollment of Pell Grant
students. Results from the analysis consistently indicate that performance funding is not a
statistically significant predictor of the percent of Pell Grant students enrolled or the average
amount of Pell Grant awarded by institutions. However, null results do not necessarily indicate
that there is no relationship between Pell Grant variables and performance funding. While no
statistically significant results were found for the models used, it is important to note that several
states included in the model only had performance funding for three years. It may take more
than three years for change to occur based on performance funding policy (Hillman, Tandberg, &
Gross, 2014). For instance, for an institution to respond to their score on a performance funding
model, several years may need to go by where the institution is not only subject to performance
funding, but receives their score and allocation. Without several years of data, an institution may
not have a clear view of the trend that is occurring for their institution.
Complicating the results further, a few states implement performance funding, but
incorporate a soft implementation where for a year or two institutions will receive their scores on
the metrics but no changes to monetary allocations will occur. Therefore, the results of this
study may not fully incorporate any results that may occur five, six or seven years after
performance funding goes into effect. Hillman, Tandberg, and Gross’ (2014) study found no
relationship between institutional performance and performance funding until seven years post
implementation. Therefore, there still exists the possibility that a change in Pell Grant enrollment
could be seen in a longer time frame. Future analysis should continue to examine this

109

phenomenon when institutions have been subject to performance funding for longer time
periods. In fact, other studies that included different years of study, with different states included
and excluded found that performance funding may lead to enrolling students from higher income
families (Klechen & Stedrak, 2016).
These results seem to give credence to the belief that higher education institutions
respond to policy initiatives by creating quasi markets. If enrollment was strictly based on
supply and demand then performance funding should have no effect on enrollment patterns.
Further, state population factors would have a stronger effect. However, this study found that
state population factors either had no statistically significant results or had marginal effects if
statistically significant.
Institutional Type
This study also explored whether the effects of performance funding on underserved
student enrollment varied by Carnegie classification and institutional selectivity. Findings
revealed that institutions do respond differently to performance funding based on Carnegie
classification and selectivity. When interacted with Carnegie classification, performance funding
was found to be a statistically significant predictor of underserved minority enrollment for five
institutional types out of a total of 10 classifications. Similarly, when interacted with Carnegie
classification, performance funding was found to be a statistically significant predictor of Pell
Grant enrollment and funding amount for three institutional types out of a total of 10
classifications. Specifically, Doctoral A institutions with performance funding decreased both the
percent of Pell Grant students they enrolled and the average amount of Pell Grant awarded
(implying they were enrolling less financially needy students), while Doctoral C institutions with
performance funding increased the percent of Pell Grant students they enrolled. In regards to

110

underserved student minorities, Masters A and Masters B institutions with performance funding
both decreased their enrollment of underserved minority students.
Separating institutions into five selectivity categories produced similar results. When
interacted with selectivity level, performance funding was found to be a statistically significant
predictor of Pell Grant enrollment or funding for three selectivity levels. Additionally,
performance funding interacted with selectivity was found to be a statistically significant
predictor of underserved minority enrollment for two levels. Specifically, moderately selective,
selective, and very high selective institutions with performance funding increased the average
amount of Pell Grant awarded. Moderately selective and selective institutions with performance
funding did not change the percent of Pell Grant students at statistically significant levels,
however very high selective institutions with performance funding decreased the percent of Pell
grant students they enrolled at the same time. Implying that very high selective institutions with
performance funding decreased the number of Pell Grant students they enrolled, but that those
they did enroll had greater need than previous enrollment classes. Taken together, these findings
present compelling evidence that not all institutions respond to or are affected by performance
funding in similar ways.
While limited quantitative research on performance funding has explored effects at the
institutional level, this finding is similar to results from Rabovsky’s (2012) study that found
institutions responded to performance funding differently when separated by institutional type
with expenditures on instruction as the dependent variable. In particular the current study’s
finding of diverse policy effects is noteworthy as it could help explain the mixed results research
on the effectiveness of performance funding. If institutions are responding to performance
funding differently, then exploring questions only through state level may not fully capture

111

institutional policy response and effectiveness. For instance, if a state has 14 public four-year
institutions, five of which respond to performance funding by meeting the goals and increasing
graduation rates, but nine institutions have no statistically significant change in graduation rates,
the state level results could potentially show no statistically significant results. Further, the
preponderance of studies utilizing difference in difference design further exasperate this problem
by only matching on state level variables without consideration for the institutional variation that
may be occurring within and between the states.
Results show that institutions that are more selective (and therefore have more flexibility
with their enrollment profile) and more likely to decrease the percent of underserved minority
students enrolled. This finding gives credibility to the PAT framework perspective that
institutions will focus on maximizing their own utility (performing well on performance funding)
as opposed to maximizing state utility (enrolling a diverse population). While performance
funding seeks to decrease this shirking, it seems to have continued the process for those
institutions that are able to edit their enrollment profile. This finding is particularly troubling as
it demonstrates that performance funding may shift underserved students to lower status
institutions.
This finding has implications for performance funding policies, for higher education
institutions specifically, and for social equity writ large. Most performance funding policies
have a goal to increase student completion, in order to reap the benefits that arise from a highly
educated populace. By further segregating students by institutional type, performance funding is
having an unintended outcome counter to the goals of the policy. While this study did not
explore overall student completion, the findings show that underserved student access is
changing with the implementation of performance funding. Shifting underserved student access

112

to lower status institutions may have a detrimental effect on student completion. Less selective
institutions often have fewer resources and lower graduation rates (Schudde & Goldrick-Rab,
2016). In fact, college completion rates at less selective institutions are almost 25 percentage
points lower than rates at very highly selective institutions (Schudde & Goldrick-Rab, 2016).
Shifting underserved students to these institutions may therefore be counterproductive to
performance funding’s goal of increasing student completion.
This finding may also affect institutions and their administration. If underserved students
are being shifted to lower status institutions, this creates a segmentation of higher education by
racial diversity that is antithetical to the purpose of higher education and creating beneficial
learning experiences for students enrolled. For instance, if underserved minority students are
shifted to less selective institutions than more selective institutions will loose diversity, which
would negatively impact all the students enrolled at these institutions (Gurin et al, 2011).
Additionally, less selective institutions, including open access institutions, may then see an
increase in underserved students. While the increase in diversity will benefit these institutions
and students, these institutions are often more limited in their financial resources and may not be
able to provide all the supports required for an increased student body.
Finally, this finding has repercussions for social equity. Access to the most selective
institutions is still unequal in the U.S. and has created a system where institutions with the
greatest access provide the least social benefit, whereas institutions with the potential to provide
the greatest social benefit are still inaccessible to many (Schudde & Goldrick-Rab, 2016).
However, if performance funding may decrease enrollment of underserved students in higher
status institutions, then underserved students may be pushed into lower status institutions
potentially including 2-year institutions. While these institutions can provide an excellent

113

education (often for a lower cost), these institutions also often have fewer opportunities for
students. Further, significant variation in earnings and labor market outcomes occur based on
college type (Schudde & Goldrick-Rab, 2016). Institutional selectivity positively impacts
earnings for bachelor degree recipients and the relationship increases over time (Long, 2010).
Even when controlling for factors that predict enrollment in selective institutions, the positive
effect remain for underserved minority students. Therefore, shifting underserved students to
lower status institutions may decrease these societal benefits.
This study also supports the concern that institutional leaders have raised that
performance funding policies do not take into consideration the diverse missions of institutions
in their state (Dougherty et al., 2014b). Rutherford and Rabovsky (2014) argue that performance
funding policies try to apply a one size fits all mentality to diverse institutions and that the
policies ignore many contextual factors that may influence an institution’s ability to respond to
the requirements or goals of performance funding. The diverse effects of performance funding
found in this study demonstrate that these policies influence institutions differently and may
create significant bifurcation in a state’s higher education system as those who the policy
benefits receive more funds and those who are not able to respond to the metrics of the policy
receive lower appropriations. This study therefore also further supports CCA’s (2012)
recommendation that performance funding policies take into account different institutional types
or be compared only to their peer institutions such as is seen in Ohio and Michigan.
Implications
Implications for Policy
This study suggests several implications for policy makers. First, it is clear that policy
effects are not one size fits all. While future research should explore if other policies also have

114

diverse effects based on institutional type, it is important in the immediate to at least consider the
diverse effects of performance funding based on Carnegie classification and selectivity. It is
clear that institutions are not homogenous and have varying institutional missions and goals.
Institutions may therefore choose to respond differently to performance funding based on these
missions and goals or may even have different ability to respond to performance funding
mandates.
These diverse mission and goals are not always represented in performance funding
policies and may create unequal repercussions from the policy. Further, institutions have
different abilities to respond to policy mandates. This study demonstrated that more selective
institutions appear to have greater ability to change their enrollment profile in response to
performance funding. However, enrollment and student performance are only one area of
performance funding metrics. It can be assumed that institutions may also respond to
performance funding by changing and altering other business practices such as research funding
to perform well on metrics associated with those measures. It can likewise be assumed that
different institution types are able to make these changes to different degrees or with different
amounts of ease. Therefore policy makers should acknowledge these differing abilities to
respond to performance mandates and provide supports and timing that allow institutions to
make the necessary changes. The diverging response to performance funding policies also
suggests the need to not compare diverse institutions within a state, but to compare institutions
within institutional type. Creating a performance funding policy that compares institutions with
their institutional type peers helps ensure institutions are compared to those with similar missions
and ability to respond to performance funding mandates.

115

Another policy implication is that institutional change can be slow and the effects of
policy may not be seen for several years. The slow nature of change at institutions could cause
complications with performance funding policies that often push to see results and change
quickly. Policies that could have financial implications for institutions (such as performance
funding policies) should allow for a few years where institutions are aware of the changes they
need to make and be given the chance to do so without financial repercussions. In fact, the CCA
(2012) argues for a period of time where institutions can familiarize themselves with the metrics
and how they would perform with limited financial impacts. Likewise, it is important to note
that policy effects are often not seen till many years post implementation. Therefore policy
makers should take great care to not make decisions based on information and results seen in the
first few years post policy implementation.
Finally, as performance funding may have an effect on enrollment profiles, it is important
to consider including metrics related to underserved student enrollment. In recent years, several
states have implemented metrics that hope to counteract the desire of higher education
institutions to engage in creaming. These metrics often include weighing underserved students
differently in progression and completion metrics or a separate metric specifically focused on
underserved student enrollment. Findings from this study demonstrate the importance of metrics
such as these, but also raise questions regarding the diverse response based on institutional type.
As states implement these types of metrics, it is important to keep in mind that different
institutions may respond differently, and the metrics should account for this.
Implications for Research
The findings from this study present several areas for potential future research. First, this
study demonstrated the need to separate by institutional type when conducting policy evaluation

116

research. This study presents evidence that performance funding policy effects are not
monolithic and researchers do their study a disservice if the do not explore policy effects by
institutional type. Additionally, as most performance funding research explores outcomes at the
state level, future research should continue to explore performance funding at an institutional
level. For instance, most research that explores the effectiveness of performance funding does so
at the state level. It would be beneficial to explore similar research questions at an institutional
level. In particular, the mixed results arising from research exploring the effectiveness of
performance funding present compelling research questions related to whether there are diverse
effects within the state across institutional types.
Second, state appropriations as a percent of institutional revenue was a strong and
negative predictor of underserved student enrollment in almost all models. Yet, limited research
exploring institutional financial decisions utilizes this control variable. This occurs despite
literature that demonstrates that an institution’s reliance on state appropriations influences
institutional decision making. Future research should consider this relationship carefully and
where appropriate include a measure of reliance on state appropriations as a variable.
Third, as noted above, as more data becomes available and institutions maintain
performance funding policies for longer time periods, research should continue to explore
whether performance funding has an influence on Pell Grant enrollment. Many performance
funding policies did not go into effect until after 2008 so limited data is available on their
effectiveness and any unintended outcomes that may occur. This study found no evidence that
performance funding changed enrollment of Pell Grant students, however several states in the
dataset only had performance funding for three years or less, leaving the possibility of long term
effects unknown.

117

This study found evidence that institutions change their enrollment profiles by decreasing
overall underserved minority enrollment. Future research should explore this further to
determine if this decrease is across all races at random or if certain races are seeing larger
decreases than others. Given the changing demographics of high school graduation cohorts and
the need to increase higher education enrollment in order for the US to stay competitive, it is
vitally important to determine specifically where the enrollment changes are occurring.
A final area for future research is to explore if these changes to enrollment profiles,
assumed to occur to improve institutional performance on performance funding, actually lead to
higher scores on performance funding metrics. If institutions are responding to the pressures of
performance funding by narrowing enrollment profiles with the hope to increase their
performance and therefore increase their funding from state appropriations, it is important to
explore if these enrollment changes do lead to the end goal of increased funding. Further, as most
policy leaders would state that they do not want to see enrollment decrease due to performance
funding, research should explore if steps states have taken to decrease this unintended outcome
actually prevents institutions from changing their enrollment profiles. For instance, in recent
years, a few states have included metrics that focus on maintaining or increasing the admissions
of underserved students. Future research should conduct rigorously controlled quantitative
studies that explore if the inclusion of such a measure does prevent creaming or changing of
enrollment profiles.
Conclusion
As performance funding spreads, researchers have expressed concern regarding potential
unintended outcomes including institutions changing their enrollment profiles. Therefore, this
study examined if performance funding influences underserved students’ enrollment at public

118

four-year institutions in the U.S. through a fixed effects panel analysis for the years 2000-2014.
The findings revealed that performance funding does decrease the percent enrollment of
underserved minority students and that performance funding has diverse effects on underserved
student enrollment by institutional type. These findings have significant implications for
performance funding policies, for higher education institutions and for social equity broadly. The
findings of this study should not be taken as an attack on performance funding policies, but as a
call to think through the unintended outcomes that may occur and create more robust policies
that consider the enrollment of underserved students and the diverse policy effects on institutions
based on mission and ability to respond to performance funding mandates.

119

APPENDIX

120

Table 16 Results for the Effects of Performance Funding Count on Pell Grant Percent
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
-0.133
0.043
0.065
0.078
Funding
(0.079)
(0.080)
(0.078)
(0.069)
(Count)
Institutional
-0.555***
-0.562***
-0.566***
-0.567***
Aid Amount
(0.110)
(0.110)
(0.110)
(0.110)
Tuition
0.658*
0.626*
0.620*
0.623*
(0.295)
(0.296)
(0.296)
(0.296)
Appropriations -0.0489
-0.046
-0.046
-0.046
as Percent of
(0.031)
(0.031)
(0.031)
(0.031)
Revenue
State Need
0.361
0.460
0.394
0.406
Based Aid
(0.669)
(0.691)
(0.695)
(0.688)
College Age
-0.003
-0.003
-0.004
-0.003
Population
(0.003)
(0.003)
(0.003)
(0.003)
State
-0.126
-0.091
-0.087
-0.089
Appropriations (0.227)
(0.229)
(0.228)
(0.228)
per FTE
Affirmative
-2.034
-1.394
-1.410
-1.549
Action Ban
(0.835)
(0.886)
(0.828)
(0.811)
R-squared
.6221
.6216
.6217
.6219
(within)
n
3324
3324
3324
3324
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001

121

Table 17 Results for the Effects of Performance Funding Count on Pell Gant Amount
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
2.118
4.505
4.702
1.543
Funding
(3.796)
(3.195)
(3.130)
(2.621)
(Count)
Institutional
3.830
3.540
3.369
.3.713
Aid Amount
(7.568)
(7.560)
(7.586)
(7.594)
Tuition
-23.250*
-24.454**
-24.462**
-23.315*
(9.298)
(9.431)
(9.307)
(9.208)
Appropriations -2.675*
-2.655*
-2.674*
-2.695*
as Percent of
(1.256)
(1.259)
(1.256)
(1.253)
Revenue
State Need
75.345
.72.488
68.042
72.319
Based Aid
(53.408)
(52.280)
(52.070)
(52.313)
College Age
-0.295*
-0.312**
-0.327**
-0.301*
Population
(0.118)
(0.119)
(0.119)
(0.118)
State
8.511
9.654
.9.551
8.496
Appropriations (7.945)
(7.976)
(7.911)
(7.844)
per FTE
Affirmative
15.520
29.837
22.302
9.569
Action Ban
(48.554)
(49.362)
(48.172)
(46.800)
R-squared
.8544
.8545
.8546
.8544
(within)
n
3324
3324
3324
3324
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001

122

Table 18 Results for the Effects of Performance Funding Count on USM Total
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance
-1.085
-1.319
-1.314
-1.403
Funding
(1.278)
(1.275)
(1.235)
(1.268)
(Count)
Institutional
-2.304
-2.291
-2.317
-2.337
Aid Amount
(3.069)
(3.071)
(3.074)
(3.081)
Tuition
13.450**
13.457**
13.440**
.13.408*
(5.164)
(5.163)
(5.156)
(5.158)
Appropriations -1.791***
-1.788***
-1.778***
-1.766***
as Percent of
(0.498)
(0.498)
(0.497)
(0.497)
Revenue
State Need
31.716*
31.495 *
31.254*
30.739*
Based Aid
(15.012)
(14.946)
(14.911)
(14.870)
College Age
1.288***
1.290***
1.291***
1.294***
USM
(0.138)
(0.138)
(0.138)
(0.138)
State
-1.714
-1.681
-1.513
-.1.349
Appropriations (4.139)
(4.122)
(4.091)
(4.071)
per FTE
Affirmative
2.861
2.289
3.299
4.317
Action Ban
(23.619)
(23.624)
(23.380)
(23.352)
R-squared
.3968
.3969
.3969
.3969
(within)
n
7016
7016
7016
7016
Robust Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001

123

Table 19 Tobit Results for the Effects of Binary Performance Funding on USM Percent
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance Funding -0.635***
-0.699***
-0.477**
-0.337*
(Binary)
(0.149)
(0.150)
(0.153)
(0.155)
Institutional Aid
-0.316***
-0.317***
-0.321**
-0.322***
Amount
(0.0492)
(0.049)
(0.049)
(0.049)
Tuition
0.377***
0.371***
0.375***
0.376***
(0.067)
(0.067)
(0.067)
(0.067)
Appropriations as
-0.073***
-0.073***
-0.073***
-0.073***
Percent of Revenue
(0.010)
(0.010)
(0.010)
(0.010)
State Need Based Aid 2.449***
2.378***
2.368***
2.369***
(0.316)
(0.315)
(0.315)
(0.315)
College Age USM
0.157***
0.162***
0.162***
0.161***
Percent
(0.013)
(0.013)
(0.013)
(0.013)
State Appropriations
-0.159*
-0.158*
-0.148*
-0.138*
per FTE
(0.068)
(0.067)
(0.067)
(0.068)
Affirmative Action
-0.819*
-0.907**
-0.916**
-0.894**
Ban
(0.336)
(0.336)
(0.336)
(0.336)
Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001
Table 20 Tobit Results for the Effects of Performance Funding Count on USM Percent
Variable
No lag
1 year lag
2 year lag
3 year lag
Performance Funding -0.175***
-0.172***
-0.166***
-0.140***
(Count)
(0.023)
(0.024)
(0.025)
(0.025)
Institutional Aid
-0.305***
-0.308***
-0.312***
-0.317***
Amount
(0.049)
(0.049)
(0.049)
(0.049)
Tuition
0.352***
0.355***
0.355***
0.360***
(0.067)
(0.067)
(0.067)
(0.067)
Appropriations as
-0.076***
-0.075***
-0.074***
-0.073***
Percent of Revenue
(0.010)
(0.010)
(0.010)
(0.010)
State Need Based Aid 2.46***
2.420***
2.399***
2.356***
(0.314)
(0.314)
(0.315)
(0.315)
College Age USM
0.153***
0.156***
0.158***
0.159***
Percent
(0.013)
(0.013)
(0.013)
(0.013)
State Appropriations
-0.167*
-0.157*
-0.135*
-0.120
per FTE
(0.067)
(0.067)
(0.067)
(0.067)
Affirmative Action
-1.067**
-1.107**
-0.977**
-0.860*
Ban
(0.336)
(0.336)
(0.335)
(0.335)
Standard Error in parentheses
*Significant at p<.05; **significant at p<.01; ***significant at p<.001

124

BIBLIOGRAPHY

125

BIBLIOGRAPHY
Advisory Committee on Student Financial Assistance. (2013). Measure twice: The impact on
graduation rates on serving pell grant recipients. Washington, DC: ACSFA.
Baltagi, B. H. (1995). Econometric analysis of panel data. Chichester, England: Wiley.
Bardo, J. W. (2009). The impact of the changing climate for accreditation on the individual
college or university: Five trends and their implications. New Directions for Higher
Education, 145, 47-58.
Baum, S., Ma, J., & Payea, K. (2013). Education pays 2013: The benefits of higher education for
individuals and society. New York, NY: The College Board.
Bell, D. A. (2005). Changing organizational stories: The effects of performance-based funding
on three community colleges in Florida (Doctoral dissertation). Available from ProQuest
Dissertations and theses database.
Bound, J. & Turner, S. (2007). Cohort crowding: How resources affect collegiate attainment.
Journal of Public Economics, 91, 877-899.
Bound, J., Lovenheim, M. F., & Turner, S. (2010). Why have college completion rates declined?
An analysis of changing student preparation and collegiate resources. American
Economic Journal: Applied Economics, 2, 129-157.
Burke, J. C. (2005). Achieving accountability in higher education: Balancing public, academic,
and market demands. San Francisco, CA: Jossey Bass.
Burke, J. C. & Minassians, H. P. (2002). The new accountability: From regulation to results.
New Directions for Institutional Research, 116, 5-19.
Burke, J. C., & Serban, A. (1998). Performance funding for public higher education: Fad or trend.
New Directions for Institutional Research, 97.
Carnevale, A. P. & Strohl, J. (2013). Separate and unequal: How higher education reinforces the
intergenerational reproduction of white racial privilege. Georgetown Public Policy
Institute.
Chen, R. (2011). Institutional characteristics and college student dropout risks: A multilevel
event history analysis. Research in Higher Education, 53, 487-505.
Chen, R. & St. John, E. P. (2011). State financial policies and college student persistence: A
national study. The Journal of Higher Education, 82(5), 629-660.
Complete College American (CCA). (2012). Performance funding: From idea to action.
Indianapolis, IN: Jones, D.
126

Complete College American (CCA). (2013). The game changers: Are states implementing the
best reforms to get more college graduates? Indianapolis, IN: CCA
CCA (2016). http://completecollege.org/the-alliance-of-states/ Accessed on November 7, 2016.
The College Board. (2016). Trends in student aid 2015. New York, NY: The College Board
Colbeck, C. L. (2002). State policies to improve undergraduate teaching: Administrator and
faculty responses. Journal of Higher Education, 73(1), 3–25.
Davis, J. H., Schoorman, F. D., & Donaldson, L. (1997). Toward a stewardship theory of
management. Academy of Management Review, 22(1), 20-47.
Delaney, J. A., & Doyle, W. R. (2011). State spending on higher education: Testing the balance
wheel over time. Journal of Education Finance, 36(4), 343–368.
Desrochers, D. M. & Hurlburt, S. (2014). Trends in College Spending: 2001-2011. Washington,
DC: Delta Cost Project.
Dougherty, K. J., & Hong, E. (2006). Performance accountability as imperfect panacea: the
community college experience. In T. Bailey & V. S. Morest (Eds.), Defending the
community college equity agenda (pp. 51–86). Baltimore, MD: Johns Hopkins University
Press.
Dougherty, K. J., Jones, S. M., Lahr, H., Natow, R.S., Pheatt, L., & Reddy, V. (2014).
Performance funding for higher education: Forms, origins, impacts, and futures. The
ANNALS of the American Academy of Political and Social Science, 655, 163-184.
Dougherty, K. J., Natow, R. S., Jones, S. M., Lahr, H., Pheatt, L., & Reddy, V. (2014). The
political origins of performance funding 2.0 in Indiana, Ohio, and Tennessee: Theoretical
perspectives and comparisons with performance funding 1.0. CCRC working Paper No.
68. Community College Research Center, Teachers College, Columbia University.
Dougherty, K. J., & Reddy, V. (2013). Performance funding for higher education: What are the
mechanisms? What are the impacts? ASHE Higher Education Report, 39(2), 1-134.
Dougherty, K. J., Natow, R. S., & Vega, B. E. (2012). Popular but unstable: Explaining why
state performance funding systems in the United States often do not persist. Teachers
College Record, 114(3). Retrieved from http://www.tcrecord.org/Content.asp?Content
Id=16313
Eaton, J. S. (2012). The future of accreditation. Planning for Higher Education.
Edgell, M. S. (2009). Higher education finance policy: A comparative analysis of two dynamics
of social contract in three bologna process countries (Doctoral dissertation). Available
from ProQuest Dissertations and Thesis database. (UMI No. 3395449)
127

Ewell, P. T. (1994). A matter of integrity: Accountability and the future of self-regulation.
Change, 26(6), 24-29.
Freeman, M. S. (2000). The experience of performance funding on higher education at the
campus level in the past 20 years (Doctoral dissertation). Available from ProQuest
Dissertations and theses database.
Fuller, C., Lebo, C. & Muffo, J. (2012). Challenges in meeting demands for accountability. In R.
D. Howard, G. W. McLaughlin, W. E. Knight & Associates (Eds.), The handbook of
institutional research (299-309). San Francisco: Jossey-Bass.
Gansemer-Topf, A. M. & Schuh, J. H. (2006). Institutional selectivity and institutional
expenditures: Examining organizational factors that contribute to retention and
graduation. Research in Higher Education, 47(6), 613-642.
Goldstein, L. (2012). A guide to college and university budgeting: Foundations for institutional
effectiveness (4th ed.). Washington, DC: NACUBO.
Gumport, P. J. (2001). Built to serve: The enduring legacy of public higher education. In P. G.
Altbach, P. J. Gumport & D. B. Johnstone (Eds.), In defense of American higher
education (pp. 85-109). Baltimore, MD: Johns Hopkins University Press.
Hall, K. B. (2000). Tennessee performance funding and the University of Tennessee, Knoxville:
A case study. (Doctoral dissertation). Available from ProQuest Dissertations and theses
database.
Hauptman, A. M. (2011). Reforming how states finance higher education. In D. E. Heller (Ed.)
The States and Public Higher Education Policy. Baltimore: John Hopkins University
Press.
Heck, R. H., Lam, W. S., & Thomas, S. L. (2014). State political culture, higher education
spending indicators, and undergraduate graduation outcomes. Educational Policy, 28(1),
3-39.
Heller, D. E. (2011). The financial aid picture: Realism, surrealism, or cubism? Higher
Education: Handbook of Theory and Research, 26, 125-160.
Heller, D. E. (2013). The role of finances in postsecondary access and success. In L.W. Perna &
A.P. Jones (Eds.), The state of college access and completion (96-114). New York, NY:
Routledge.
Heller, D. E. & Marin, P. (2004). State merit scholarship programs and racial inequality.
Cambridge, MA: The Civil Rights Project at Harvard University.
Hillman, N.W., Tandberg, D. A. & Fryar, A.H. (2014). Evaluating the impacts of “new”
performance funding in higher education. Educational evaluation and policy analysis.
128

Retrieved from
http://epa.sagepub.com/content/early/2014/12/09/0162373714560224.abstract
Hillman, N., Tandberg, D., & Gross, J. P. K. (2014). Performance funding in higher education:
Do financial incentives impact college completions? The Journal of Higher Education,
85(6), 826-857.
Immewahr, J., Johnson, J., & Gasbarra, P. (2008). The Iron Triangle: College Presidents Talk
About Costs, Access, and Quality. The National Center for Public Policy and Higher
Education.
Jaquette, O. & Parra, (2014). Using IPEDS for panel analyses: Core concepts, data challenges
and empirical applications. Higher Education: Handbook of Theory and Research, 29,
467-533.
Keller, C. M. (2012). Collective responses to a new era of accountability in higher education. In
R. D. Howard, G. W. McLaughlin, W. E. Knight & Associates (Eds.), The handbook of
institutional research (371-385). San Francisco: Jossey-Bass.
Kelly, A. P. & Lautzenheiser, D. K. (2013). Taking charge: A state-level agenda for higher
education reform. Washington, DC: American Enterprise Institute
Kelly, A. P. & Schneider, M. (2012). Getting to graduation: The completion agenda in higher
education. Baltimore: Johns Hopkins University Press.
Kezar, A. (2004). Obtaining integrity? Reviewing and examining the charter between higher
education and society. Review of Higher Education, 27(4), 429-459.
Kim, J. (2012). Exploring the relationship between state financial aid policy and postsecondary
enrollment choices: A focus on income and race differences. Research in Higher
Education, 53, 123-151.
Lahr, H. Pheatt, L. Dougherty, K. J., Jones, S. M., Natow, R. S., & Reddy, V. (2014).
Unintended impacts of performance funding on community colleges in three states. New
York, NY: Teachers College, Columbia University.
Lane, J. E. (2012). Agency theory in higher education organizations. In M. N. Bastedo (Ed.), The
organization of higher education: Managing colleges for a new era (278-303).
Baltimore, MD: The John Hopkins University Press.
Lane, J. E. & Kivisto, J. A. (2008). Interests, information, and incentives in higher education:
Principal-agent theory and its potential applications to the study of higher education
governance. In J. C. Smart (Ed.), Higher Education: Handbook of Theory and Research
vol. 23 (141-180). New York, NY: Springer.
LeGrand, J. (1991). Quasi-markets and social policy. The Economic Journal, 101, 1256-1267.

129

Marginson, S. (1997). Markets in education. St. Leonards, Australia: Allen and Unwin.
McPherson, M. S. & Schapiro, M.O. (1991). Does student aid affect college enrollment? New
evidence on a persistent controversy. American Economic Review, 81(1), 309–318.
McMahon, W. (2009). Higher learning, greater good. Baltimore, MD: Johns Hopkins University
Press.
McLendon, M. K., Hearn, J. C., & Deaton, R. (2006). Called to account: Analyzing the origins
and spread of state performance-accountability policies for higher education. Educational
Evaluation and Policy Analysis, 28(1), 1–24.
McLendon, M. K., Hearn, J. C., Mokher, C. G. (2009). Partisans, professionals, and power: The
role of political factors in state higher education funding. The Journal of Higher
Education, 80(6), 686-713.
Mitchell, M., Leachman, M. & Masterson, K. (2016). Funding Down, Tuition up: States cuts to
higher education threaten quality and affordability at public colleges. Washington, DC.
Center on Budget and Policy Priorities.
Mitchell, M., Palacios, V., and Leachman, M. (2014). State are still funding higher education
below pre-recession levels. Washington, DC. Center on Budget and Policy Priorities.
Naughton, B. A. (2004). The efficacy of state higher education accountability programs.
(Doctoral dissertation). Available from ProQuest Dissertations and theses database.
National Association of State Budget Officers (NASBO). (2013). Improving postsecondary
education through the budget process: Challenges and opportunities. Washington, DC:
Pattison.
National Association of State Student Grant and Aid Programs (NASSGAP). (2016). 46th Annual
survey report on state-sponsored student financial aid.
National Center for Education Statistics (2016). The Condition of Education 2016. Washington,
DC: Kena et al.
National Student Clearinghouse (2014). Completing college: A national view of student
attainment rates-fall 2008 cohort. Herndon, VA: Shapiro et al.
National Conference of State Legislatures (2006). Transforming higher education: National
imperative-state responsibility. Washington, DC: NCSL
National Conference of State Legislatures (2015). Retrieved from:
http://www.ncsl.org/research/education/performance-funding.aspx

130

National Governors Association. (2007). Innovation America: A compact for postsecondary
education. Washington, DC: NGA.
O’Neal, L. M. (2007). Performance funding in Ohio’s four-year institutions of higher education:
A case study (Doctoral dissertation). Available from ProQuest Dissertations and theses
database.
Opoczynski, R. (2014). Quantifying the Future of Michigan: Implementation and Short Term
Impacts of a Performance Funding Policy. (Unpublished research).
Perna, L.W. (2013). Conclusions: Improving college access, persistence, and completion:
Lessons learned. In L.W. Perna & A.P. Jones (Eds.), The state of college access and
completion (208-224). New York, NY: Routledge.
Phillips, M. A. (2002). the effectiveness of performance-based funding: Does Florida’s system of
outcome measurements improve community college performance? (Doctoral dissertation).
Available from ProQuest Dissertations and theses database.
Pike, G. R. (2013). NSSE benchmarks and institutional outcomes: A note on the importance of
considering the intended uses of a measure in validity studies. Research in Higher
Education, 54, 149-170.
Pike, G. R. & Graunke, S. S. (2014). Examining the effects of institutional and cohort
characteristics on retention rates. Research in Higher Education, 56, 146-165.
Powell, B. A., Gilleland, D. S. & Pearson, L. C. (2012). Expenditures, efficiency, and
effectiveness in U.S. undergraduate higher education: A national benchmark model. The
Journal of Higher Education, 83(1), 102-127.
Rabovsky, T. (2012). Accountability in higher education: Exploring impacts on state budgets and
institutional spending patterns. Journal of Public Administration Research and theory,
22(4), 675–700.
Rhee, B. (2008). Institutional climate and student departure: A multinomial multilevel modeling
approach. The Review of Higher Education, 31(2), 161-183.
Rutherford, A. & Rabovsky, T. (2014). Evaluating impacts of performance funding policies on
student outcomes in higher education. The ANNALS of the American Academy of
Political and Social Science, 655, 185-208.
Ryan, J. F. (2004). The relationship between institutional expenditures and degree attainment at
baccalaureate colleges. Research in Higher Education, 45(2), 97-114.
Sanford, T., & Hunter, J. M. (2011). Impact of performance funding on retention and graduation
rates. Educational Policy Analysis Archives, 19(33).
Schaller, J. Y. (2004). Performance funding in Ohio: Differences in awareness of Success
131

Challenge between student affairs administrators and academic affairs administrators at
Ohio’s public universities (Doctoral dissertation). Available from ProQuest Dissertations
and theses data-base.
Scott, M., Bailey, T., & Kienzl, G. (2006). Relative success? Determinants of college graduation rates in public and private colleges in the U.S. Research in Higher Education, 47(3),
249–279.
Shin, J. C. (2010). Impacts of performance-based accountability on institutional performance in
the U.S. Higher Education, 60(1), 47–68.
Shin, J. C., & Milton, M. (2004). The effects of performance budgeting and funding programs on
graduation rate in public four-year colleges and universities. Education Policy Analysis
Archives, 12(22), 1–26.
State Higher Education Executive Officers (SHEEO). (2005). Accountability for better results: A
national imperative for higher education. Denver, CO: SHEEO
State Higher Education Executive Officers (SHEEO). (2013). State Higher Education Finance:
FY2012. Denver, CO: SHEEO.
State Higher Education Executive Officers (SHEEO). (2014). State Higher Education Finance:
FY2013. Denver, CO: SHEEO.
State Higher Education Executive Officers (SHEEO). (2016). State Higher Education Finance:
FY2015. Denver, CO: SHEEO.
Singer, J. D. & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and
event occurrence. Oxford, NY: Oxford University Press.
Slaughter, S. & Leslie, L. (1997). Academic capitalism: Politics, policies, and the
entrepreneurial university. Baltimore, MD: Johns Hopkins University Press.
Slaughter, S. & Rhoades, G. (2004). Academic capitalism and the new economy: Markets, states
and higher education. Baltimore, MD: Johns Hopkins University Press.
Sparks, P. J. and Nunez, A. (2014). The role of postsecondary institutional urbanicity in college
persistence. Journal of Research in Rural Education, 29(6), 1-19.
Tandberg, D. A. (2010). Politics, interest groups and state funding of public higher education.
Research in Higher Education, 15(5), 416–450.
Tandberg, D. A. and Griffith, C. (2013). State support of higher education: Data, measures,
findings and directions for future research. Higher Education: Handbook of Theory and
Research, 28, 613-685.

132

Tandberg, D. A. & Hillman, N. W. (2013). State higher education performance funding: Data,
outcomes and policy implications. Paper presented at the Association for the Study of
Higher Education.
Taylor, B. J., Cantwell, B. & Slaughter, S. (2013). Quasi-markets in U.S. higher education: The
humanities and institutional revenues. The Journal of Higher Education, 84(5), 675-707.
Thelin, J. (2004). A history of American higher education. Baltimore: The Johns Hopkins
University Press.
Titus, M. A. (2006).Understanding college degree completion of students with low
socioeconomic status: The influence of the institutional financial context. Research in
Higher Education, 47(4), 371–398.
Titus, M. A. (2009). The production of bachelor’s degrees and financial aspects of state higher
education policy: A dynamic analysis. The Journal of Higher Education, 80(4), 439–468.
Toman, J. K. (2008). Partners or adversaries: A comparative case study of the accountability
relationship between higher education and state government in three states.
(Unpublished doctoral dissertation). University of South Dakota: Vermillion, South
Dakota.
Toutkoushian, R. K. and Hillman, N. W. (2012). The impact of state appropriations and grants
on access to higher education and outmigration. The Review of Higher Education, 36 (1),
51-90.
Toutkoushian, R. K. and Shafiq, M. N. (2010). A conceptual analysis of state support for higher
education: Appropriations versus need-based financial aid. Research in Higher
Education, 51, 40-64.
Van Der Klaauw, W. (2002). Estimating the effects of financial aid offers on college
enrollment: A regression-discontinuity approach. International Economic Review, 43(4),
1249-1287.
Volkwein, J. F. (2009). The assessment context: Accreditation, accountability, and performance.
New Directions for Institutional Research, 3-12.
Volkwein, J. F., & Tandberg, D. A. (2008). Measuring up: Examining the connections among
state structural characteristics, regulatory practices, and performance. Research in Higher
Education, 49(2), 180–197.
Weerts, D. J. and Ronca, J.M. (2012). Understanding differences in state support for higher
education across states, sectors, and institutions: A longitudinal study. The Journal of
Higher Education, 83 (2), 155-185.
Western Interstate Commission for Higher Education (2012). Knocking at the college Door:
Projections of high school graduates. Boulder, CO: Prescott, B. T. & Bransberger, P.
133

Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. The MIT
press.
Wooldridge, J. M. (2013). Introductory econometrics: A modern approach 5th edition. South
Western: Mason, OH
Zhang, L. (2010). The use of panel data models in higher education policy studies. Higher
education: Handbook of Theory and Research, 25, 307-349.
Zhang, L. and Ness, E. C. (2010). Does state merit-based aid stem brain drain? Educational
Evaluation and Policy Analysis, 32 (2), 143-165.

134