Why has for-profit colleges’ share of higher 
education expanded so rapidly? Estimating 
the responsiveness to labor market changes
Authors: Gregory Gilpin, Joseph Saunders, & Christiana Stoddard
NOTICE: this is the author’s version of a work that was accepted for publication in Economics of 
Education Review. Changes resulting from the publishing process, such as peer review, editing, 
corrections, structural formatting, and other quality control mechanisms may not be reflected in 
this document. Changes may have been made to this work since it was submitted for publication. 
A definitive version was subsequently published in Economics of Education Review, [VOL. 45, 
(April 2015)] DOI# 10.1016/j.econedurev.2014.11.004
Gilpin, Gregory A., Joseph Saunders, and Christiana Stoddard. “Why Has for-Profit Colleges’ 
Share of Higher Education Expanded so Rapidly? Estimating the Responsiveness to Labor 
Market Changes.” Economics of Education Review 45 (April 2015): 53–63. 
doi:10.1016/j.econedurev.2014.11.004
Made available through Montana State University’s ScholarWorks 
scholarworks.montana.edu 
Why has for-profit colleges’ share of higher education 
expanded so rapidly? Estimating the responsiveness to labor 
market ch
Gregory A. Gilpin & Christiana Stoddard: Department of Agricultural Ec
ashingt
1Joseph Saunders: Accenture LLP, 800 Connecticut Avenue, Ste. 600, W
Abstract
Over the last two decades, for-profit colleges (FPC
of the higher education market. One potential expl
sponsive to labor market changes than public comp
degree students, we examine the effects of chang
ous employment fields on enrollment and degree
indicate that enrollment and degree completion in
ployment growth and wages in related occupation
unresponsive. Heterogeneity analysis reveals that t
of students by gender and ethnicity. Furthermore
public institutions are non-responsive to changes
an Associate’s or Bachelor’s degree.
. Introduction
For-profit colleges (FPCs) are one of the most rapidly ex-
panding types of higher education providers in the United
States. Many of these schools focus on two-year programs,
and over the past two decades they have accounted for much
of the growth in Associate’s degrees. FPCs doubled their
enrollment of Associate’s degree seeking students between
1995–1996 and 2010–2011, a rate of increase that was three
times faster than in public community colleges. The num-
ber of Associate’s degrees awarded grew even faster, with a
growth rate at FPCs thatwas six times faster than that at com-
munity colleges, resulting in FPCs increasing their share of
Associates degrees awarded from 10% to 22% between 1995–
1996 and 2010–2011.anges
onomics and Economics, Montana State University, Bozeman, MT 59717
on, DC 20006s) have substantially increased their share
anation is that FPC sector may be more re-
etitors. Using panel datasets of Associate’s
es in labor market conditions across vari-
completion in related majors. The results
the FPC sector is positively related to em-
s, while public institutions remain largely
hese relationships are similar across groups
, the results also indicate that students in
in labor markets associated with requiring
Costs and the type of education services provided by FPCs
differ from their public competitors. Knapp, Kelly-Reid, and
Ginder (2011) report the average annual tuition difference
between public community colleges and for-profit colleges
to be approximately $12,000 (see Kofoed, 2014 for analysis
on financial aid allocation). Typically, for-profit institutions
have hadmore limited program offerings, withmany smaller
FPCs specializing in only one or two programs. In contrast,
public two-year institutions typically offer a wider spectrum
of programs, with greater availability of general studies and
liberal arts programs. For example, during the 2011–2012
academic school year, FPCs awarded 31% of all degrees in
health fields, 25% in business fields, and 9% in computer and
information sciences fields, while 42% of public two-year
0
60
2
-year
ut
io
ns
tions, 1
ta Syste0
20
4
1998 2000 200
For-profit 2
N
um
be
r o
f n
ew
 in
st
it
Fig. 1. Number of new public and for profit two-year degree granting institu
Source: Author calculations based on Integrated Postsecondary Education Da
degrees were awarded in liberal arts, humanities, or gen-
eral education fields (Digest of Education Statistics, 2012,
Table 292).
FPCs have also rapidly expanded into new fields. Little lit-
erature exists to rigorously compare the barriers to entry in
the higher education market and how these barriers com-
pare across for-profit and public institutions. Data on new
institutions and programs suggest that FPCs are much more
likely to be entrants. Fig. 1 shows the number of new two-year institutions from1998 through 2008. Although there are
substantially more public institutions than private, between
35 and 60 new two-year FPCs opened each year during this
decade, on average at least twice as many as the number of
public entrants in the same year. Fig. 2 displays this infor-
mation by field of study.2 A new field of study is defined as
one where no related majors existed in previous years. Fig. 2
shows that in each year between 1998 and 2008, roughly 5%
of the fields of study at public institutions are new introduc-
tions, while on average between 10 and 20% of the fields of
study are new introductions at for-profit institutions.3
For-profit colleges have structural features that may al-
low them to quickly adapt to changing economic conditions,
generating new entry. For example, FPCs typically have a dif-
ferent governance structure than public institutions, with
more defined stakeholder interests, fewer tenured faculty,
and physical and financial structures that allow more flex-
ibility. One potential explanation for the expansion of FPCs
may be greater program flexibility related to labor market
2 Majors are aggregated into 12 broad fields of study (e.g., business, edu-
cation, legal studies, etc.), as defined in more detail below.
3 Schools must establish a Program Participation Agreement with the US
Department of Education to be a Title IV institution by meeting eligibility
requirements of state licensure, institutional accreditation, standards of fi-
nancial responsibility and administrative capability. A participating school
must also comply with various state and federal laws.2004 2006 2008
Public 2-year
998–2008.
m (IPEDS).
conditions. This could allow them to capture more of each
programs’ enrollment share, while public institutions may
be not be as nimble in adjusting their program offerings.
Does some of the growth of FPCs represent differences in
responsiveness to labor market conditions, or does this sim-
ply reflect an overall shift towards this new institutional op-
tion? To evaluate this, we examine enrollment and degree
completion in various Associate’s degree programs across
states and over time in response to labor market conditionswithin related occupations. Although there is a growing body
of research on individual’s choice to enroll at FPCs versus
community colleges, with particular emphasis on student
characteristics and tuition at competing schools (see Cellini,
2009; Chung, 2012; Deming, Goldin, & Katz, 2012; Turner,
2003), to our knowledge there is little research exploring
how changes in the labor market conditions affect the share
of students at FPCs.4
Following the literature, we examine the Associate’s de-
gree level and restrict the comparison group to two-year pub-
lic institutions (i.e., community colleges).5 We focus on Asso-
ciate’s degrees for several reasons. First, certificate programs
are much more heterogeneous, making it more difficult to
isolate the relative roles of for-profit or public institutional
status. Focusingon two-yeardegreegranting institutions lim-
its the analysis to institutions that are fairly comparable along
many other dimensions. Second, many of the concerns about
4 The closest research is Turner (2003) who studies the effect of local
unemployment rates on the rate of enrollment at FPCs. The research finds
that enrollment at for-profit institutions are sensitive to changes in the local
economic conditions, but does not provide empirical analysis on whether
public competitors are also influenced. Thus, the relative responsiveness is
not estimated.
5 The FPC literature uses community colleges as a comparison group to
reduce selection bias (see Chung 2008; Cellini 2009,2012; Deming, Goldin,
and Katz 2012).
2
ye
-year
rofit and
ta Syste0
.0
5
.1
.1
5
.2
F
ra
ct
io
n 
of
 fi
el
ds
 o
f s
tu
dy
 th
at
 a
re
 n
ew
1998 2000 200
For-profit 2
Fig. 2. Fraction of fields of study that are new introductions, two-year for-p
Source: Author calculations based on Integrated Postsecondary Education Da
FPCs relate to high levels of debt their students incur, levels
that are obviously higher for two-year programs than shorter
duration certificates. These concerns have received increas-
ing attention as FPCs have expanded intomore advanced lev-
els of education and their students have become recipients
of federal financial aid. Accordingly, we only examine Title IV
eligible institutions, as other institutions are not eligible for
the same types of federal financial aid. However, the penetra-
tion of FPCs into four-year and graduate programs has beenrelatively recent, limiting our ability to use trends over time
for these institutions.
We study the impacts of labor market conditions on both
enrollment and degree completion. One concern with study-
ing enrollment is that the results may be susceptible to the
confounding influence of community college students tak-
ing courses for non-employment reasons. Studying degree
completion provides additional evidence that a confounding
effect is not driving the results.We capture contemporaneous
and recent state-level labor market conditions across various
employment fields using the average employment growth
and the real wage level over the previous three years prior
to enrollment or degree completion. Using the average over
a three-year time span permits an evaluation of the changes
in employment conditions when individuals were deciding
to enroll and also smoothes out transitory shocks. These two
measures capture different potential labor market attributes
that may affect students. Betts (1996), Freeman and Hirsch
(2008), Webbink and Hartog (2004) have examined student
responsiveness to occupational employment opportunities
as well as wage opportunities and find both to be distinctly
relevant measures.
The panel data analyses indicate that enrollment in Asso-
ciate’s degrees and degree completions are more responsive
to changes in employment conditions of related occupations
at FPCs than at community colleges. Specifically, a 1% increase2004 2006 2008
ar
Public 2-year
public Degree Granting Institutions,1998–2008.
m (IPEDS).
in employment growth in an occupational field is correlated
with expanding enrollment shares in related majors at FPCs
of about 3.2 percentage points and an increase in degree com-
pletion at FPCs of 1.5 percentage points. While these appear
to be large responses, it is important to note that the mag-
nitude of the results indicate a change in flow of new hires
within occupations; although new hires are a small propor-
tion of total employment, entrants with new degrees may be
a substantive share of new hires.6 The effect of employmentgrowth at community colleges is much smaller, only 0.5 per-
centage points, and is negatively related. Additionally, a 1%
increase in anoccupational field’s realwage is correlatedwith
an expanded enrollment share in majors in related fields of
study of about 5.0 percentage points at FPCs. Wage effects at
community colleges are about two-thirds as large.
We explore the robustness of these results using a variety
of specifications accounting for permanent differences in un-
observed characteristics of FPCs, states, and time periods as
well as differential effects across occupational fields, preva-
lence of FPCs, availability of program offerings, labor market
conditions by level of education, and student demographic
characteristics.We find the pattern of greater responsiveness
at FPCs is preserved.
2. Background: differences in for-profit colleges and
public community colleges
Currently there are approximately 3.3 million students
enrolled at FPCs (see Digest of Education Statistics, 2013,
Table 301.1). The FPC sector contains a wide range of schools.
Best known are the largest institutions that offer a wide array
of programs, such as the University of Phoenix owned by the
6 We thank a referee for pointing out this interpretation.
Apollo group. The largest institutions account for themajority
of enrollment at FPCs, with the top 15 institutions enrolling
approximately 60% of all FPC students (Bennett, Lucchesi, &
Vedder, 2010). At the other end of the spectrum, there are
hundreds of smaller FPCs that offer limited sets of programs.
In a study of California FPCs, on average, each college offered
four degrees or certificate programs and enrolled approxi-
mately 71 students (Cellini, 2005).
FPCs differ from public institutions in several key ways.
First, the overall objectives of public colleges and FPCs are
different. For example, 13 of the 15 largest for-profit college
firms are publicly traded (Bennett et al., 2010),7 and therefore
operate similarly to other publicly-traded firms who pursue
profits and distribute dividends to shareholders. Second, the
colleges are also more likely to hire part-time and non-
tenured adjunct instructors, while faculty lines at community
colleges are more likely to be full-time tenured positions.
Kinser (2007) finds that only 3% of faculty members at
publicly-traded for-profit collegeshave tenure,while 17.5%of
community colleges’ faculties are tenured and 42% of newly
hired full-time faculty is tenure track. As a result, FPC boardsvast majority of FPCs’ revenue (88%) is derived from student
tuition (Hentschke, 2011). While students do receive federal
grants or subsidized loans to reduce their share of these tu-
ition costs, only 7% of revenue is derived from government
grants, contracts, and appropriations that flow directly to the
institution. In contrast, direct government support accounts
for approximately 53% of revenues at public community col-
leges in the form of state and local appropriations, while
35% is derived from student tuition, which may include tu-
ition payments provided by federal grants or subsidized loans
(“Community College Fact Sheet”, 2012).8 Because public dol-
lars for community colleges are specifically tied to the indi-
vidual institution, as opposed to following the student to a
range of institutions, community colleges have a mandate
from voters to provide services for their local area. Conse-
quently, community colleges typically have mission state-
ments with diverse goals that address the broader interests
of voters and area residents, including aiding developmental
education, lifelong learning, university transfer, cultural enrich-
ment, civic, personal, and cultural growth, community service,
community prosperity, and community vitality.9 For-profit col-
leges have a more prominent objective of maintaining rela-
tionships with local businesses, often using Employer Ad-
visory Boards or Industry Advisory Councils, and marketing
these relationships to potential students (Hentschke, 2011).
The two school types also have different management
structures. For-profit colleges typically have boards of di-
rectors that dictate decision making and operations to the
instructors. In contrast, community colleges typically have
shared faculty governance for decision making on curricu-
lum development, peer evaluation, hiring, marketing, and
compensation (Bailey, 2006; Hentschke, 2011). For-profit
7 The University of Phoenix is the largest for-profit college and currently
enrolled 319,400 undergraduate students and a total of 380,800 students
acrossmany campuses and through online programs in 2011 (“Apollo Group,
Inc.” 2011). By contrast, ArizonaStateUniversity, the largest public university
by enrollment, had 58,404 undergraduates and a total of 72,254 students
(About ASU, 2012).
8 While almost all students at FPCs take advantage of Federal student aid,
revenue at FPCs is dictated by enrollment and not simply provisioned in
legislation.
9 These mission statement excerpts are taken from various community
colleges, including City College of San Francisco (CA), Cleveland State Com-
munity College (TN), Cloud County Community College, (KS), Collin College
(TX), Cuyahoga Community College (OH), Flathead Valley Community Col-
lege (MT), Manchester Community College (CT), Nassau Community College
(NY), PiedmontVirginia Community College (VA), and State College of FloridaManatee-Sarasota (FL).of directors have fewer constraints for reducing or expand-
ing programs, changing course offerings, or removing poor
performing instructors.
The differences in funding sources may cause community
colleges to be more susceptible to changes in the business
cycle. Given the heavy reliance on local and state funds, rev-
enues for community college are more likely to fall when
local and state revenues decrease (Berger & Kostal, 2002).
This pro-cyclical business cycle effect is exacerbated by the
fact that the demand for higher education can be counter-
cyclical (Betts & McFarland, 1995). If tuition increases are
insufficient to cover the marginal costs of offering additional
programs, thismay reduce faculty size (e.g., throughattrition)
or cap program enrollment and courses, precisely when the
demands for these programs are greatest. Since FPCs receive
little direct government funding, instead receiving funds pri-
marily through student tuition, revenues are more likely to
increase during economic downturns. This is because student
enrollment tends to be counter-cyclical (Betts & McFarland,
1995; Black & Sufi, 2002; Dellas & Sakellaris, 2003; Sakellaris
& Spilimbergo, 2000). As a result, instructor and enrollment
caps are less likely to be binding.10
For-profit colleges typically have a physical plant struc-
ture that is more adjustable than that of community colleges.
For-profit colleges are more likely to lease space (often in
commercial areas) at multiple locations throughout a region
than to have a large central campus, which is more typical of
a community college. Moreover, FPCs have expanded more
rapidly online, holding 42% of the online higher education
market share (Bennett et al., 2010).
As discussed above, research on the student choice to
enroll at a FPC or a competing public institution has been
established, with particular emphasis on the effects of stu-
dent characteristics and costs of competing schools. Chung
(2012) studies the decision to attend a FPC versus attend-
ing community colleges and finds that the students’ choice
is heavily influenced by student’s socioeconomic background
and parental involvement. Chung (2008) also finds that the
local concentration of FPCs is also important. Cellini (2009)
finds that expansion of federal grant programs, such as the
Pell Grant program, increases entry of FPC institutions. How-
ever, these gains do not come at a loss to community college
enrollment. Rather, both FPC and community college enroll-
ment increase. Deming et al. (2012) find that, relative to com-
parable students who attended community colleges or other
public or private non-profit institutions, FPCs educate a larger
fraction of minority, disadvantaged, and older students.11
10 See Berger and Kostal (2002) for more detail on state appropriate supply
constraints.
11 There is also a recent line of research studying whether the returns to a
FPC education is different than that of a public competitor. The results are
mixed with some finding a negative effect (Deming, Goldin, & Katz, 2012;Lang & Weinstein, 2012; Liu & Belfield, 2014), no statistically significant
3. Empirical specification
The previous literature suggests that both demand and
supply characteristics of for-profit and community colleges
maydiffer. Themain empirical analysis examines the reduced
form relationship between labor market conditions in a spe-
institutions) formajors classified intofield j in state s at time t.
The vector ECjst includes employment condition variables,
employment growth and the real wage level, in log form.
These measures appear to be the most widely available to
prospective students, and career counselors typically advise
students on which fields are expanding and what average
wages are in related occupations. We average both employ-
ment conditions over the current year and the previous two
years. The variable FPCi is a dummy variable for whether
school type i is a FPC. The vector Xst contains state-level con-
trol variables and μj is employment-field specific fixed ef-cific occupational field (e.g., health occupations) and Asso-
ciate’s degree enrollment and degree completion in related
majors at different types of institutions. Specifically, the anal-
ysis examines whether enrollment and degree completion
change when field-specific employment conditions in a state
change, and if so, whether the change is statistically different
between FPCs and community colleges. For example, when
employment conditions in medical fields improve, does en-
rollment in medically-oriented Associate’s degrees increase,
and is the responsiveness greater at FPCs than at public insti-
tutions?
We assume that there is a market for educational pro-
grams in state s at time t that leads to gainful employment in
occupational field j. To conduct the analyses, we restrict our
attention to FPCs and public counterparts that are as com-
parable as possible. We compare Associate’s degree-seeking
students at FPCs to those at two-year public institutions,
predominantly community colleges.While some public four-
year degree universities and non-profit private higher edu-
cation institutions offer two-year degrees, they are excluded
as students in these programs since they are less likely to
be similar to Associate’s degree-seeking students at FPCs.12
Additionally, certificate granting schools are muchmore het-
erogeneous, are not Title IV eligible, and are not well rep-
resented in national surveys of higher education. The focus
on Associate’s degree seeking students allows us to examine
populations that are as similar as possible across public and
for-profit entities.
The supply of educational program j is a function of the
costs of producing a set enrollment capacity. The production
function includes the physical plant, instructors, and other
human resources necessary to provide the program. Simi-
lar to standard production functions, inputs may be sticky in
the short-run, but are fully adjustable in the long-run. Fur-
thermore, the time to change inputs may be different across
school types.When employment conditions becomemore fa-
vorable in employment field j, this affects demand for slots in
educational programs in related majors. The resulting quan-
tity change is related to the underlying demand and supply
conditions.
The main empirical specification is a reduced form esti-
mate:
lnYijst = β0 + ECjst × β1 + ECjst × FPCi × β2 + Xstβ3
+β4 × ln Tst + FPCi + μj + εijst (1)
where the dependent variable, ln Yijst,, is either the log of
total enrollment or degree completion (specified in more de-
tail below) at school type i (either a FPC or two-year public
difference (Darolia, Koedel, Martorell, & Perez-Arce, 2014; Lang &Weinstein,
2013), or a positive effect (Cellini, 2012; Cellini & Chaudhary, 2012).
12 While not reported, we also extend the analysis to include all two-year
institutions (public and private) as well as all two-year and four-year public
institutions. These analyses do not control for any potential selection biashowever and are available from the authors upon request.fects. These fixed effects account for permanent differences
in the proportion of an employment field’s majors. For ex-
ample, business-related majors tend to attract higher pro-
portions of students. In addition, we include the log of total
enrollment (or degree completions) for a state, Tst, to permit
a straightforward interpretation of the main coefficients of
interest, β1 and β2. Lastly, εijst is the error term, for which
we use a variety of specifications as discussed below.13
This structure permits an examination of the aggregate
of the school type choice and major choice of individuals
by studying enrollment and degree completion at different
institutions types across time, states, and fields. The aggre-
gate approach of using totals within a state has a number
of empirical advantages. First, the level of aggregation on
the left-hand side and right-hand side are consistent. Given
the fact that mobility, particularly for college-aged students,
is relatively high within a state, it is likely that students re-
spond tooverall employment conditionswithin a state, rather
than, for example, MSA-level conditions. Additionally, occu-
pational field-specific employment conditions are measured
most completely at the state level.14 Second, state-level anal-
ysis permits us to exploit a panel framework to control for
selection using institution type fixed effects. The FPC choice
literature suggests that students who enroll in FPCs likely
differ from students who enroll in public institutions in sys-
tematic ways and these systematic differences can be con-
trolled using appropriate statistical methods, i.e., institution-
type fixed effects. Without the use of these fixed effects, even
using comparison students that are most similar to FPC stu-
dents, such as community college students, does not assure
that selection is properly controlled.15 Additionally, using oc-
cupational field-specific employment conditions along with
state-fixed effects instead of general employment conditions
assures that this effect is not confounded with other state-
level influences.
The coefficients of interest in this specification are β1 and
β2. The coefficient β1 represents the percentage change in
enrollment or degree completion in majors at community
13 An alternative specificationwould be to use the share of total enrollment
(or degree completions) as the dependent variable. This essentially restricts
β4 to be equal to one and moves ln Tst to the left-hand side. We have run re-
gressions under both specifications and find that we generally cannot reject
the null of a coefficient equal to one. This implies that the interpretation of
the other coefficients remains unchanged under either specification.
14 Some metro-level employment conditions are available but not for all
occupational fields and only for large metropolitan areas. An analysis at a
more disaggregate level would lose much data and would not be represen-
tative of the national sample.
15 For other discussions of aggregation and selection, see Card and Krueger(1996) and Moffit (1995).
colleges in response to a change in employment conditions in
the related occupational field, while β1 + β2 captures the ef-
fect of changes in employment conditions at FPCs. Theparam-
eter β2 is the differential effect. If this coefficient is positive,
it indicates that student enrollment or degree completion in
2003–2004, and 2007–2008 school years.17 The survey is a
stratified random sample of students enrolled in Title IV eligi-
ble postsecondary institutions throughout the United States.
Title IV eligible institutions are those that meet federal eligi-
bility requirements to receive federal student financial aid.18
We use the NPSAS sampling weights to construct state-level
representative samples. We restrict the sample to students
pursuing an Associate’s degree at either FPCs or two-year
public institutions. Given that we study the Associate’s de-
gree level, requiring institutions to be Title IV eligible is not
binding, i.e., almost all higher education institutes that offer
Associate’s and Bachelor’s degrees are Title IV eligible.
The data include demographic, background, and financial
information for students and their families. Importantly, the
geographic data is based on the residence of the student,a field of study at FPCs is more likely to expand in response to
positive employment conditions in the related occupations
than is the case at public two-year institutions. An F-test
statistically validates whether the effect at FPCs is statisti-
cally different from zero, i.e., whether β1 + β2 is statistically
different from zero.
The specification then captures overall shifts of students
across for-profit and public institutions, as well as shifts
acrossmajorswithin each institution’s type. Given that states
clearly differ in the penetration of FPCs and in overall trends
in enrollment and that students may becomemore or less re-
sponsive to employment conditions over time in their choices
of majors, we report results from two alternative specifi-
cations. The first specification includes occupational field
fixed effects, school-type-by-year fixed effects, school-type-
by-state fixed effects, and state-level control variables. The
school-type-by-state fixed effects account for permanent dif-
ferences in community colleges and FPCs across states. For
example, in the data, FPCs in rural states tend to have less
penetration than inmore populous states. Additionally, these
fixed effects also address differences in state-level financing
of public schools, average tuition differences across states,
and average tuition differences across the type of institutions.
The school-type-by-year fixed effects account for any com-
mon economy-wide effects in a given year. For example, the
effect of the U.S. business cycle would be absorbed by these
fixed effects. In addition, they account for common shocks to
all FPCs in a given year or common shocks to public institu-
tions in a given year. Thus, only occupational field-specific
changes in employment condition at the state-level remain.
The error term, εijst, is clustered at the state level.
16
The second specification includes the most flexible fixed
effects, school-type-by-state-by-yearfixedeffects, FPCi ×μst.
In this specification, state-level variables are excluded due
to collinearity with the fixed effects. These fixed effects as-
sure that the results identify the effect of changes in rela-
tive employment conditions within occupational fields in a
given state and year. Consequently, any potential omitted
variable bias is mitigated due to the large number of perma-
nent and time-varying variables that can be accounted for
in this framework. The error term is again clustered at the
state-level. This is our preferred model specification.
4. Data
Threemain sources of data are used to construct the panel
datasets. First, the enrollment regressions rely on data from
the restricted-access versions of the National Postsecondary
Student Aid Study (NPSAS) conducted by the National Cen-
ter for Education Statistics collected during the 1999–2000,
16 Given the empirical specification, cluster robust standard errors address
correlation within each state. Clustering reduces the likelihood of commit-
ting a Type I error, where the null hypothesis that a variable’s effect equalszero is rejected when it is actually true for the population (see Kézdi, 2004).as opposed to the location of their university, a distinction
that is particularly important given the prominence of online
instruction at for-profit colleges. Overall, the NPSAS sample
includes 53,800 students enrolled during the sample period,
with 46,400 students at public community colleges and 7,390
students at FPCs.
Second, the degree completion data are provided by the
Integrated Postsecondary EducationData System (IPEDSData
Center (2012)) for similar school years as the NPSAS. This is
a comprehensive survey of Title IV higher education institu-
tions in the United States. The IPEDS data have the advan-
tage of reporting the universe of college completers at these
institutions, which is particularly useful in the early 2000s
when the penetration of FPCs was lower. However, demo-
graphic data is limited to race and gender compositions of
institutions and is compiled based on the institution location
rather than student location. Overall, the IPEDS data include
1,581,264 studentswho completed an Associate’s degree and
whose fields of study were reported, with 1,418,909 earned
at community colleges and 162,355 earned at for-profit col-
leges.
Third, occupationdata are providedby theBureauof Labor
Statistics (BLS)’ Occupational Employment Statistics (OES)
from 1997 through 2008.19 The OES reports state-level an-
nual employment counts in designated occupations and the
mean hourly wage. These wages are deflated to 2008 dollars
using the GDP deflator provided by the Bureau of Economic
Analysis. Additional informationon local areaunemployment
rates are providedby theBureauof Labor Statistics’ Local Area
Unemployment Statistics (LAUS) and real per capita income
17 The sampling framework is based on the 2000–2001 Integrated Postsec-
ondary EducationData Systemfiles and stratifies institutions by institutional
control, institutional level, highest level of offering, Carnegie classification,
and state. Weights are then constructed based on probabilities proportional
to institutions sampled. Because the NPSAS sampling frame is based on the
IPEDS files, neither represent non-Title IV institutions well. See National
Postsecondary Student Aid Study (2000, 2004, and 2008) for further details.
18 Enrollment observations are rounded to the nearest 10s place as per
restricted-use data license.
19 Prior to 1999, the BLS classified occupations on a slightly less disaggre-
gate level. Data from 1997 and 1998 have been aggregated to the current
classification (see http://www.bls.gov/oes/1999/g_itsoc.pdf for further de-
tails). Two data anomalies appeared during the construction of the employ-
ment conditions. The total employment in the Arts/Communications and
Community, Social Services, and Protective Services fields during 1999 and
the Physical Science field during 2004 experienced unrealistic increases. Inboth cases, the average of the previous and next year are substituted.
(in 2008 dollars) is provided by the Bureau of Economic Anal-
ysis (BEA). Lastly, we construct the annual number of public
institutions per 1 million residents in each state using higher
education institutional data from IPEDS and population data
from the U.S. Census.We use this as ameasure of community
college access.20
To capture the responsiveness of enrollment and degree
completion in majors across FPCs and community colleges,
ideally the analysiswouldmapAssociate’s degree collegema-
jors to the most common subsequent occupations chosen by
studentswith thosemajors.However, no suchmappingexists
21for education attainment at the Associate’s level. Instead,
similar to Lang and Weinstein (2013), we use the Classifica-
tion of Instructional Programs (CIP 2000) from the National
Center for Education Statistics (NCES). The NCES and BLS
jointly provide a crosswalk for these CIP codes and SOC occu-
pation codes, where the program of study is a direct prepa-
ration for an occupational group.22 We use this crosswalk to
construct a correspondence between two-year college ma-
jors and broad occupational fields.We aggregate occupations
into 12 related employment fields using the two-digit SOC
occupation codes given by the BLS and then map majors
into these fields using the related two-digit CIP codes. The
complete mapping is provided in Supplementary Appendix
Table A1. A few of these fields map a single CIP code to a
single two-digit SOC occupation code. For example, the edu-
cation field mapsmajors with CIP code 13 (education) to SOC
code 25 (education, training, and library occupations); the
legal field maps CIP code 22 (legal professions and studies)
with SOC code 23 (legal occupations). Others, like the Life and
physical science field, map a group of majors (CIP codes 3, 26,
40, and 41) to a single SOC occupational group (SOC 19 life
and physical science occupations). Other fields are defined
as a combination of SOC occupations when students might
seek employment in various occupations (e.g., the business
field maps majors with CIP code 51 (business, management,
marketing, and related supporting services majors) with four
related SOC codes (11: management, 13: business and finan-
cial operations, 41: sales and related, and 43: office and ad-
ministrative support occupations). Most of these mappings
are fairly unambiguous. It is important to note that a student
in a given major might not obtain employment in an occu-
pation related to their degree, but arguably the major and
institution choices were related to their ideal career aspira-
tions when entering college.
One challenge with the mapping is that the CIP code for
liberal arts degrees corresponds with liberal arts teaching
occupations, but liberal arts students are typically employed
in a broader range of occupations. Accordingly, the liberal
20 This measure is constructed with the entire population as the denom-
inator, rather than the population of young people. This is due to the fact
that community colleges and FPCs tend to attract high proportions of non-
traditional aged students—both students still of high school age as well as
students returning to school at later stages in their career.
21 One potential mapping is provided by the U.S. Census’ American Com-
munity Survey (ACS), but this is only for Bachelor’s degrees. This mapping
would require a critical assumption that individuals earning an Associate’s
degree and aBachelor’s degreewould go into similar occupations. This seems
highly unlikely.
22 See the 1998 Standard Occupational Classification Files (1998) andOccupation-to-Training Classification for further details.arts majors are assigned to a liberal arts field and the
corresponding occupations are those in the social sciences,
community and social services, legal, education, and services,
as reported in Supplementary Appendix Table A1. This field
is potentially problematic given the substantial share of
community college students enrolled in liberal arts or
general education programs. Empirical robustness checks
verify that the main results remain even after removing this
substantial group of students (see Supplementary Table A4
for a comparison of results).
Given the mapping of majors and occupations to employ-
ment fields, we aggregate individuals using the NPSAS data
to obtain the annual total enrollment in each field of study
by school type for each state and aggregate institutions using
the IPEDS data to obtain the annual total degrees completed
in each field of study by school type for each state, Yijst. This
aggregation incorporates both the choice of college (for profit
or community college) as well as the choice ofmajor.We also
construct similarly the annual total enrollment (and degrees
completed) for each state, Tst. Using these aggregates allows
an analysis of the distribution of students across schools and
majors in response to changes in related labor market condi-
tions while controlling for overall changes in the number of
students seeking or who have completed an Associate’s de-
gree. For example, the specifications provide evidence as to
whether increases in enrollment in health-related majors is
related to increases in employment in the health field in Cali-
fornia and how the rate of increase differs between for-profit
colleges and community colleges.
Themain specifications only include the observationwith
positive enrollment or degree completions and exclude the
zero observations. That is, the estimated effects are condi-
tional on there being some enrollment or degree completions
in majors related to a given field at a specific school type. The
assumption is that if there are noprogramsoffered in apartic-
ular field in a state at a given school type, enrollment and de-
gree completions cannot respond to employment conditions.
Subsequent robustness checkswill extend themain specifica-
tion and analyze how including zeros effects the estimation.
Supplementary Appendix Table A2 provides descriptive
statistics on the total enrollment and degrees completed dur-
ing the last decade across school types and fields at the na-
tional level. This table indicates that Liberal Arts majors at
community colleges are the most popular two-year degree.
Community colleges and FPCs both enroll substantial shares
of their student bodies in health related majors, while busi-
ness majors have become increasingly popular at FPCs.
The empirical analysis examines the effect of employ-
ment conditions in occupations that typically require an As-
sociate’s degree. The BLS’ 2012–2013 Occupational Outlook
Handbook reports educational attainment by SOC occupa-
tion code (see Occupational Outlook Handbook (2012)). In
the main analysis, occupations are included if less than half
of the workers hold a Bachelor’s degree or higher but more
than half hold a high school diploma. This excludes occupa-
tions with high educational attainment requirements (e.g.,
doctors, lawyers, and teachers) and those with low educa-
tion attainment requirements (e.g., shampooers and garbage
collectors). Included occupations are those that a graduate
with an Associate’s degree is likely to pursue, such as nurs-ing, recreational therapy, paralegal, and teacher’s assistant.
ion Data
Maxim
dent var
255,9
55,7
dent var
1,138,9
90,8
conditiTable 1
Descriptive Statistics.
Source: Author calculations based on Integrated Postsecondary Educat
Community college
Mean Standard Minimum
deviation
Depen
Enrollment per field 9456.4 19,925.2 0.0
Degree completion
per field
779.8 2693.7 0.0
Indepen
Total enrollment 162,096.0 191,826.5 1.8
Total degrees
completed
10,376.5 13,454.6 24.0
Unemployment rate
(in %)
4.67 1.12 2.53
Community colleges
per 1 million
residents
5.05 3.13 0.00
Log real income (per
capita)
0.32 4.18 0.03
% Asian 3.99 10.28 0.00
% Black 8.91 8.35 0.00
% Hispanic 5.39 7.09 0.00
% Female 61.72 6.54 37.31
Labor market
Employment growthAA occupations 2.95 10.31 −44.18
BA+ occupations 2.0 12.3 −48.0 1
All occupations 2.4 4.5 −28.8
Hourly Wage
AA occupations 19.03 5.34 8.63
BA+ occupations 29.81 9.90 11.63
All occupations 23.63 7.75 8.90
Notes: Observations are total number of students in a particular state and yea
school type (community college or for-profit college).
a State-level data. AA occupations: Includes occupations where at least 50% o
a Bachelor’s degree or higher. This is the baseline specification group. BA± occ
a Bachelor’s degree or higher. All occupations: All occupations regardless of wo
That is, the employment conditions (namely, employment
growth and the real wage level) in a field are calculated as
ECjst =
K∑
k=1
ωkjst × o¯kjst × I(k ∈ j)
where ωk
jst
is the share weight in employment field j in
occupation k in state s during year t, t, o¯k
jst
is the three-year
average employment growth or real wage level during year
t, t − 1, and t − 2 in occupation k in state s, and I is an indica-
tor function to whether occupation k is an element of the set
j. The set j is defined as the subset of occupations in field
j that satisfy the above educational attainment rules. Using
theaverageover a three-year time spanpermits anevaluation
of the changes in employment conditions when the individ-
ual was deciding to enroll in the two-year program. Annual
national averages are provided for employment growth and
hourly wages in 2008 dollars across all fields in Supplemen-
tary Appendix Table A3.
Subsequent robustness analyses study whether FPC and
community college students respond differently to labor
market conditions by including occupationswheremore than
half of all workers hold a Bachelor’s degree or higher and an-
other with all occupations regardless of workers’ educational
attainment. This will be especially important for communitySystem (IPEDS).
For-profit college
um Mean Standard Minimum Maximum
deviation
iables
59.0 788.8 2421.5 0.0 27,677.3
75.0 83.0 254.8 0.0 3269.0
iablesa
46.0 162,096.0 191,826.5 1.8 1,138,946.0
57.0 10,376.5 13,454.6 24.0 90,857.0
7.67 4.67 1.12 2.53 7.67
14.18 5.05 3.13 0.00 14.18
67.19 0.32 4.18 0.03 67.19
73.95 3.99 10.28 0.00 73.95
37.31 8.91 8.35 0.00 37.31
36.16 5.39 7.09 0.00 36.16
80.00 61.72 6.54 37.31 80.00
on measures81.51 2.95 10.31 −44.18 81.51
48.9 2.0 12.3 −48.0 148.9
30.8 2.4 4.5 −28.8 30.8
41.04 19.03 5.34 8.63 41.04
68.84 29.81 9.90 11.63 68.84
55.50 23.63 7.75 8.90 55.50
r in an occupational field (see Supplementary Table A2) at a specific
f workers hold a high school degree and less than 50% of workers hold
upations: Includes occupations where more than 50% of workers hold
rkers’ educational attainment. Hourly wages are in 2008 dollars.
college students as some of them use two-year programs as
an initial step towards obtaining a Bachelor’s degree and con-
sider the employment conditions of occupations requiring a
Bachelor’s degree when enrolling in programs at the Asso-
ciate’s degree level.
5. Results
The descriptive statistics summarizing all variables in the
regressions are provided in Table 1 and indicate that the aver-
age annual number ofmajors enrolled in a given field of study
at community colleges is 9456 per state, while the average
is 789 at FPCs. This is due to the overall larger share of stu-
dents enrolled in community colleges. [Note that throughout,
student sampling weights are used for the NPSAS sample to
ensure representativeness.] Similarly, the table indicates that
the average annual number of degrees completed is 780 at
community colleges and 83 at FPCs. The table also indicates
that there is substantial variation in these totals across fields.
Table 2 reports the coefficients of interest from the esti-
mation approaches outlined above for student enrollment.
Table 3 reports parallel regressions with degree comple-
tion as the dependent variable. Because employment growth
and wages are likely related, we report specifications that
include each separately, as well as jointly. The first three
Table 2
Effect of changes in employment growth and wages on student enrollment.
1
−0.00
(0.00
0.03
(0.01
2188
0.69
0.00Employment growth
Employment growth × FPC
Real wage
Real wage × FPC
N
R2
P-value from F-test that employ. growth + employ. growth × FPC = 0
P-value from F-test that real wage + real wage × FPC = 0
Occupational field fixed effects Yes
School type-by-year fixed effects Yes
School type-by-state fixed effects Yes
State-level control variables Yes
School-type-by-state-by-year fixed effects No
Notes: Asterisks ∗ , ∗∗ , and ∗∗∗ indicate significance at the 10%, 5%, and 1% statisti
beneath the coefficient estimates. Control variables include the unemployment rat
female, and percentages of students by race (whites excluded). Baseline specificati
degree and less than 50% of workers hold a Bachelor’s degree or higher are used
employ. growth + employ. growth × FPC = 0 and real wage + real wage × FPC =
Table 3
Effect of changes in employment growth and wages on degree completion.
1
Employment growth −
(
Employment growth × FPC
(
Real wage
Real wage × FPC
N 2
R2
P-value from F-test that employ. growth + employ. growth × FPC = 0
P-value from F-test that real wage + real wage × FPC = 0
Occupational field fixed effects Y
School type-by-year fixed effects Y
School type-by-state fixed effects Y
State level control variables Y
School-type-by-state-by-year fixed effects N
Notes: Asterisks ∗ , ∗∗ , and ∗∗∗ indicate significance at the 10%, 5%, and 1% stati
listed beneath the coefficient estimates. Control variables include the unemp
income, percent female, and percentages of students by race (whites exclude
workers hold a high school degree and less than 50% of workers hold a Bache
columns include occupational field fixed effects, school-
type-by-year fixed effects, school-type-by-state fixed ef-
fects and a variety of state level control variables. The fi-
nal three columns report results containing occupational
field fixed effects and school-type-by-state-by-year fixed
effects.23
The first three columns of Tables 2 and 3 indicate a very
small negative correlation between a field’s employment
growth and the enrollment and degrees completed in ma-
jors for related fields of study at community colleges, but the
23 Alternative specifications using various subsets of these fixed effects
produce consistent results.2 3 4 5 6
6∗∗ −0.004 −0.006∗∗ −0.005∗
3) (0.003) (0.003) (0.003)
9∗∗∗ 0.037∗∗∗ 0.039∗∗∗ 0.037∗∗∗
1) (0.0111) (0.010) (0.010)
0.040∗∗∗ 0.040∗∗∗ 0.033∗∗ 0.032∗∗
(0.010) (0.010) (0.012) (0.012)
(0.012) (0.012) (0.012) (0.012)
(0.012) (0.012) (0.012) (0.012)
2188 2188 2188 2188 2188
0.69 0.69 0.74 0.74 0.74
0.00 0.00 0.00
0.00 0.00 0.00 0.01Yes Yes Yes Yes Yes
Yes Yes No No No
Yes Yes No No No
Yes Yes No No No
No No Yes Yes Yes
cal levels, respectively. Standard errors are clustered by state and listed
e, community colleges per 1million residents, per capita income, percent
on group of occupations where at least 50% of workers hold a high school
to construct employment growth and wages. P-values from F-tests that
0 in all cases are 1% or less.
2 3 4 5 6
0.002 −0.002 −0.001 −0.001
0.002) (0.002) (.003) (0.003)
0.013∗ 0.013∗ 0.016∗ 0.016∗
0.007) (0.007) (0.009) (0.009)
0.002 0.001 0.002 0.001
(0.015) (0.016) (0.016) (0.017)
0.002 0.001 0.003 0.001
(0.013) (0.014) (0.014) (0.014)
397 2397 2397 2397 2397 2397
0.70 0.70 0.70 0.72 0.72 0.72
0.10 0.12 0.10 0.11
0.83 0.92 0.81 0.90
es Yes Yes Yes Yes Yes
es Yes Yes No No No
es Yes Yes No No No
es Yes Yes No No No
o No No Yes Yes Yes
stical levels, respectively. Standard errors are clustered by state and
loyment rate, community colleges per 1 million residents, per capita
d). Baseline specification group of occupations where at least 50% of
lor’s degree or higher are used to construct employment growth.
estimate is statistically imprecise for degrees completed. The
point estimate for FPCs, however, is positive and significant: a
1% increase inemploymentgrowth inafield is correlatedwith
a 3.3 percentage point increase in student enrollment and a
1.1percentagepoint increase indegree completion.Addition-
ally, F-tests indicate that the correlation between employ-
ment growth and FPC enrollment is statistically significant at
1%, while the correlation between employment growth and
FPC degree completion is statistically significant at 11%.
Real wages are moderately related to student enrollment,
although the effect of wages on student enrollment is larger
at FPCs than at public colleges. Specifically, a 1% increase in
real wages is associated with a 4.0% increase in enrollment
at public institutions, but a 5.9–6.4 percentage point increase
at FPCs. When both wages and employment growth are in-
cluded in the same regression, the additional effect at FPCs is
not estimated as precisely. Wages are not statistically associ-
ated with degree completions at either type of college.
As discussed in previous sections, these specifications ac-
count for a number of fixed effects and state level control
variables, but do not allow for common shocks to school
types in a year and state—for example, a common increase
in all FPCs in California in 2008. The final three columns in-
clude school-type-by-year-by-state fixed effects to allow for
these common unobserved characteristics. For example, they
control for differences in the penetration of FPCs relative to
community colleges in different states and in different years.
The variation remaining that identifies the model is the vari-
ation in the relative growth of specific fields of study across
institution types.
The last three columns of Tables 2 and 3 presents results
of the second specification and are consistent with the first
specification (results provided in the first three columns of
Tables 2 and 3). There is a significantly negative but small
effect of employment growth in a field on the share of en-
rollment in related majors in community colleges, and no
effect on majors’ shares of degree completion. Similar to
previous specification estimates, there is a much larger and
positive correlation between employment growth and en-
rollment and degree completion at FPCs: a 1% increase in
employment growth in a field is correlated with an approx-
imate 3.2 percentage point increase in enrollment and a
1.5 percentage point increase in degree completion in re-
lated majors. The effect of wages also are similar in these
specifications, with about a 2 percentage point larger impact
on enrollment at FPCs, although this is not always statistically
significant. Again, while these estimated responses appear to
be large, the magnitude of the results indicate a change in
flow of new hires within occupations; although new hires
are a small proportion of total employment, entrants with
new degrees may be a substantive share of new hires.
Themain results suggest that enrollment anddegree com-
pletion in majors at FPCs are more responsive to changes
in related employment conditions than at community col-
leges. We test the strength of this relationship using a se-
ries of robustness checks. These are all detailed in the online
Supplementary Appendix 2.
First, as noted, the liberal arts field is much larger at com-
munity colleges than at for profit colleges, and has a broader
match to related occupations. We drop the liberal arts field
to see if this affects the findings. The results from this estima-
tion are provided in Supplementary Table 4. A second concern
is that we have a high proportion of observations with zero
enrollment or degree completions at either community col-
leges or FPCs. Supplementary Table 5 provides alternative
estimation approaches for dealing with this problem, again
described in the online appendix. Third, we examinewhether
the results are drivenby sensitivity to different labormarkets.
Community college programs are often designed to enable
students to advance to four-year public institution to obtain
a Bachelor’s degree. There is lower responsiveness at com-
munity colleges to labor market conditions for occupations
requiring an Associate’s degree, but is this true for general la-
bor markets conditions or occupations requiring a Bachelor’sdegree? The online appendix Table 6 reports these robust-ness checks. Finally, we examine whether the results vary by
students’ race or for men and women, with results in Sup-
plementary Tables 7 and 8. In general, the baseline pattern
of results is confirmed in all of these instances, with larger
effects of labor market conditions at FPCs, and more muted
or statistically insignificant results at community colleges.
6. Conclusion
For-profit colleges have been increasingly scrutinized for
their high tuition rates, low graduation rates, and their stu-
dents’ high debt burdens upon graduation (see Allison, 2014;
Editorial, 2012; Matthews, 2012; Porter, 2014; Stiglitz, 2012;
Webley, 2012).However, these institutionshavegrownmuch
more rapidly than other institutions of higher education. This
is particularly true in the two-year higher education market,
where student enrollment rates at FPCs have grown six times
faster than at public community colleges andwhere FPCs cur-
rently account for almost a quarter of all Associate’s degrees
awarded.
The rapid shift in enrollment has occurred evenwithmore
limited program offerings at many two-year FPCs. However,
FPCs may have an advantage over community colleges in
terms of structural features that may allow them to more
easily adjust their course offerings in response to chang-
ing employment conditions. Community colleges are more
likely to have decentralized faculty governance systems with
tenured professors, to have financial support that is tied to lo-
cal voter support, and to have large campuseswith high capi-
tal costs. For-profit colleges tend to have more part-time fac-
ulty, more centrally administered decision making, to lease
or rent space, to rely more on online courses, and to have
fewer constraints when it comes to expanded and removing
courses of study.
The empirical analysis examines Associate’s degree stu-
dent enrollment and degree completion in various majors
at community colleges and FPCs. The results indicate that
FPCs have expanded their enrollment and degree completion
rapidly in majors where employment conditions in related
occupations are improving and have decreased enrollment
in majors where employment conditions in related occupa-
tions are diminishing. The estimates suggest that a 1% higher
employment growth rate in occupations in a given employ-
ment field in a state is correlated with a 3.2 percentage point
increase in students enrolled in FPCs in the state in related
majors. The student enrollment share at community colleges
majors is much less responsive. Additionally, a 1% higher real
wage is associatedwith approximately a 3.2 percentage point
higher enrollment in related majors at community colleges
but an effect that is about a third larger at for-profit col-
leges. Finally, the effects on degree completions are some-
what smaller for both FPCs and community colleges in re-
sponse to changes in employment growth and wages, but
again show the same pattern of larger positive effects at FPCs
and smaller, sometimes negative, and insignificant effects at
community colleges. These results are robust to a variety of
flexible specifications of fixed effects for employment fields
and school types, differential penetration of FPCs, separate
segments of the employmentmarket based on education lev-els, and for students of different demographic characteristics.
Community College Fact Sheet. (2012). American Association of
Community Colleges. http://www.aacc.nche.edu/AboutCC/Documents/
FactSheet2012.pdf.
Dellas, H., & Sakellaris, P. (2003). On the cyclicality of schooling: Theory and
evidence. Oxford Economic Papers, 55(1), 148–172.
Deming, D., Goldin, C., & Katz, L. (2012). The for-profit postsecondary schoolThe results are unable to distinguish whether the higher
responsiveness is due to the unobserved attributes of the
students who are most attracted to FPCs or due to the in-
stitutional features of the FPCs themselves. Any change in
enrollment requires action of both students and institutions.
Furthermore, even if FPCs are more nimble in expanding
enrollment than public competitors, there may be other
changes occurringwhen institutions adjust their program of-
ferings, such as a decline in quality. This quantity-quality
trade-off may be particularly salient when expanding en-
rollment rapidly, especially when increasing enrollment in
non-existent or smaller programs. Future work should eval-
uate the institutional nuances and other dynamic inter-
actions that may have occurred during periods of rapid
expansion.
Supplementary materials
Supplementary material associated with this
article can be found, in the online version, at
doi:10.1016/j.econedurev.2014.11.004.
References
About ASU. (2012). Arizona State University. http://about.asu.edu/
facts.html.
Allison, T. (2014, May 14). Why the Feds should regulate for-profit colleges.
Huffington Post.
Apollo Group Inc. (2011). Reports fiscal 2011 fourth quarter and
year end results Apollo Group, Inc. http://investors.apollogrp.edu/
phoenix.zhtml?c=79624&p=irol-newsArticle&id=1618736.
Bailey, T. (2006). Increasing competition and growth of the for-profits. In
T. Bailey, & V. S. Morest (Eds.), Defending the community college equity
agenda (pp. 87–109). Baltimore: Johns Hopkins University Press.
Bennett, D., Lucchesi, A., & Vedder, R. (2010). For profit higher education:
Growth, innovation, and regulation (Policy paper) Center for College
Affordability and Productivity.
Berger, M. C., & Kostal, T. (2002). Financial resources, regulation, and
enrollment in US public higher education. Economics of Education Re-
view, 21(2), 101–110.
Betts, J. (1996).What do students knowaboutwages? Evidence froma survey
of undergraduates. Journal of Human Resources, 31(1), 27–56.
Betts, J., & McFarland, L. (1995). Safe port in a storm: The impact of labor
market conditions on community college enrollments. Journal of Human
Resources, 30(4), 741–765.
Black, S., & Su, A. (2002). Who goes to college? Differential enrollment by
race and family background (NBER working paper no. 9310).
Card, D., & Kruegar, A. (1996). School resources and student outcomes: An
overview of the literature and new evidence from North and South Car-
olina. Journal of Economic Perspectives, 10, 31–50.
Cellini, S. (2005). Community colleges and proprietary schools: A compar-
ison of sub-baccalaureate postsecondary institutions (Working paper
05-012) California Center for Population Research (CCPR).
Cellini, S. (2009). Crowded colleges and college crowd-out: The impact of
public subsidies on the two-year college market. American Economic
Journal: Economic Policy, 1(2), 1–30.
Cellini, S. (2012). For-profit higher education: An assessment of costs and
benefits. National Tax Journal, 65, 153–180.
Cellini, S., & Chaudhary, L. (2012). The labor market returns to a for-profit
college education (NBER working paper no. 18343).
Chung, A. (2008). For-profit student heterogeneity (Unpublished
manuscript). Ann Arbor, MI: University of Michigan.
Chung, A. (2012). Choice of for-profit college. Economics of Education Review,
31, 1084–1101.sector: Nimble critters or agile predators? The Journal of Economic Per-
spectives, 26(1), 139–163.
Darolia, R., Koedel, C., Martorell, P., & Perez-Arce, F. (2014). Do employers
prefer workers who attend for-profit colleges? Evidence from a field
experiment (RAND working paper WR-1054).
Digest of Education Statistics (2012). U. S. Department of Education, Na-
tional Center for Education Statistics, NCES 2012001. http://nces.ed.gov/
pubs2012/2012001.pdf.
Digest of Education Statistics (2013). U. S. Department of Education, Na-
tional Center for Education Statistics, NCES 2014015. http://nces.ed.gov/
pubs2014/2014015.pdf.
Editorial (2012). Editorial false promises. New York Times.
Freeman, J., & Hirsch, B. (2008). College majors and the knowledge content
of jobs. Economics of Education Review, 27, 517–535.
Hentschke, G. (2011). For profit sector innovations in business models. In
B. Wildavsky, A. Kelly, & K. Carey (Eds.), Reinventing higher education:
The promise of innovation (pp. 159–196). Cambridge: Harvard Education
Press.
IPEDS Data Center. (2012). U.S. Department of Education, National Center for
Education Statistics. http://nces.ed.gov/ipeds/datacenter/.
Kézdi, G. (2004). Robust standard error estimation in fixed-effects models.
Hungarian Statistical Review, 9, 95–116.
Kinser, K. (2007). Dimensions of corporate ownership in for-profit higher
education. The Review of Higher Education, 30(3), 217–245.
Knapp, L., Kelly-Reid, J., & Ginder, S. (2011). Postsecondary institutions
and price of attendance in the United States: 2010–2011, degrees and
other awards conferred: 2009–2010, and 12-month enrollment: 2009–
2010. U.S. Department of Education, National Center for Education
Statistics.
Kofoed, M. (2014). For-profit and traditional colleges: Institutional control
and financial aid allocation (Working paper).
Lang, K., & Weinstein, R. (2012). Evaluating student outcomes at for-profit
colleges (NBER working paper 18201).
Lang, K., & Weinstein, R. (2013). The wage effect of not-for-profits and
for-profit certifications: Better data, somewhat different results. Labour
Economics, 24(C), 230–240.
Liu, V., & Belfield, C. (2014). The labor market returns to for-profit higher
education: Evidence for transfer students (CAPSEE working paper).
Matthews, D. (2012, July 30). Harkin: For-profit colleges a terrible deal. The
Washington Post.
Moffit, R. (1995). Selection bias adjustment in treatment-effects models as a
method of aggregation. Proceedings of the American Statistical Association,
234–238.
National postsecondary student aid study, 2000, 2004, and 2008. U.S. De-
partment of Education, National Center for Education Statistics.
Occupation-to-training classification. Crosswalks: SOC to CIP. National
Crosswalk Service Center. http://www.xwalkcenter.org/index.php/
classifications/crosswalks.
Occupational outlook handbook. (2012). U.S. Department of Labor, Bureau of
Labor Statistics. Table 1.11. http://bls.gov/emp/ep_table_111.htm.
Porter, E. (2014, February 25). The bane and the boon of for-profit colleges.
New York Times.
Sakellaris, P., & Spilimbergo, A. (2000). Business cycles and investment in
human capital: International evidence on higher education. Carnegie-
Rochester Conference Series on Public Policy, 52(1), 221–256.
Standard occupational classification files. (1998). National Crosswalk
Service Center. http://webdata.xwalkcenter.org/ftp/download/soc1998/
socoes98.xls.
Stiglitz, J. (2012, July 3). Debt buries American dream. USA Today.
Turner, S. (2003). For-profit colleges andnon-traditional students: Responses
to cyclical shocks (Working paper) National Center for the Study of Pri-
vatization in Education.
Webley, K. (2012, January 9). For-profit schools: ‘Agile predators’ or just
business savy?. TIME Magazine.
Webbink, D., & Hartog, J. (2004). Can students predict starting salaries? Yes!.
Economics of Education Review, 23(2), 103–113.