For-Profit Higher Education Responsiveness to Price Shocks
An Investigation of Changes in Post 9-11 GI Bill Allowed Maximum Tuitions
Matthew Baird1, Michael Kofoed2, Trey Miller1, Jennie Wenger1
October 2016
Abstract
The Post 9/11 GI Bill represented one of the largest expansions of college benefits for veterans
and their dependents. In the first version of the bill, the Department of Veterans Affairs set the
maximum tuition reimbursement on a state by state level. However, in 2011, the VA set the
maximum tuition benefit to one nationwide amount. In this study we use a difference in
differences estimator to find that in states where the benefit increased, for-profit universities
increased their sticker price tuition by $5 for every extra $100 of benefits. We find however that
these institutions did not increase admissions or enrollments after the policy change.
1
RAND Corporation
United States Military Academy: 607 Cullum Road, West Point, New York, USA, 10996. Tel:
(845) 938-2748, email: michael.kofoed@usma.edu. The views expressed herein are those of the
authors and do not reflect the position of the United States Military Academy, the Department of
the Army, or the Department of Defense.
2
1. Introduction
The rapidly increasing cost of college has caused a great deal of concern in the popular press and
many policymakers. Many factors are cited as potential sources of the increase. Decreases in
state-level aid, an increase in the focus on obtaining outside funds, an increase in the number and
proportion of administrators employed, potential push-back about faculty salaries and teaching
loads, and other reasons have all been suggested as possible reasons.
While policymakers have put forth a number of suggestions for alleviating the effects of
increasing college costs, one of the more popular options is increasing the level and availability
of financial aid (CITE). However, financial aid has the potential to increase the “sticker price”
tuition or published tuition before financial aid.3 William J. Bennett, former United States
Secretary of Education, first proposed this hypothesis, which states that when governments
increase the amount of publicly funded financial aid, universities have an incentive to increase
their tuition to capture this new aid as a form of price discrimination (Bennett 1987). While this
behavior would increase tuition revenues for the university, it would not be optimal for the
student and of course would at least partly defeat the goals of providing additional financial aid.
One public entity that distributes financial aid is the Veterans’ Administration (VA), via
the various iterations of the GI Bill, including the most recent, commonly referred to as the Post
9/11 GI Bill.4
Might be worth mentioning that many students, especially at schools with higher “sticker
prices” do not pay full price, due to a combination of merit-based and need-based financial aid.
4
The PGIB, passed in 2008 and enacted in 2009, is the latest in a series of educational benefits
designed to ease service members’ transitions into civilian life. The PGIB is, on average,
considerably more generous than the previous bill (the Montgomery GI Bill). The PGIB also
includes a unique feature; those who serve for a sufficient amount of time may transfer some or
all of their PGIB benefits to a spouse or child.
3
In this study, we leverage an unanticipated change in the state by state maximum tuition
reimbursement for the Post 9/11 GI Bill (PGIB) in 2011 to a national standard maximum to
measure the effect of changes in financial aid on sticker price tuition. Before the 2011 policy
change, PGIB funds were allocated based on the average cost of attendance for public postsecondary schools within a particular state; students who attended schools that cost no more than
the average had their full tuition costs covered while students attending schools that charged
more than the average had only the average amount covered. As of 2011, the policy changed
and the PGIB covered tuition up to a maximum amount that was the same for all students
regardless of state.
This policy change creates unique exogenous variation among the states because benefits
for some recipients in some states increased while others decreased. In states where the
maximum tuition benefit increased, this policy change created an incentive for schools to
increase the price of their tuition; in other states, schools have faced different incentives as the
maximum reimbursement decreased. In general, for-profit colleges have more flexibility and
incentive to change tuition in response to such policy changes (Deming et al. 2012; Gilpin et al.
2015). We anticipate that, relative to non-profit institutions, sticker price for for-profit
institutions should increase in states where the maximum tuition allowed under the PGIB
increased, whereas it should decrease in states where the maximum allowable tuition decreased.
We also expect the effect to be stronger for schools in particular that enroll a large number of
active duty military, veterans, and their dependents.
For profit colleges are of great concern to the Departments of Defense and Veteran
Affairs. In fact, in the first year of the Post 9-11 GI Bill, nearly 36.5 percent of all benefits were
claimed by students at for-profit universities while enrolling only 23.3 percent of PGIB
beneficiaries (Health Education, Labor, and Pensions Committee, 2010, p. 4; Deming et al.
2012). Interestingly, as the Obama Administration has changed the rules regarding where
Department of Education dollars can be spent to exclude many for-profit universities, these new
rule changes do not affect the PGIB because the Department of Veteran Affairs administers the
program (Hefling 2016). Despite this concern from various federal agencies, very little is known
about how for-profit (and traditional universities as well) respond to exogenous shocks in
veterans’ benefits. If these institutions enroll more veterans (even despite some tuition
increases), then the Departments of Defense and Veteran Affairs will be fulfilling their goals in
helping to re-train military members transitioning to civilian life and labor markets. However, if
for-profit colleges simply increase sticker price tuition without an accompanying increase in
enrollment, then the increase in tuition costs are not accompanied by any social gain.
In this study, we use this variation to examine how universities adjust their sticker price
tuition and if these changes can be explained by shifts in demand (enrollments) or price
discrimination. We find that in states where the maximum tuition payment from the PGIB
increased, sticker price tuition rates at for-profit colleges increased by $5.47 for every $100, and
the increase was concentrated in states that have a larger veteran population. We find no such
increase in traditional non-profit public or private institutions, nor in states where the maximum
tuition payment fell. We also estimate the effect that changes in the PGIB had on fall headcount
enrollment and find no significant effects. This evidence suggests that for-profit colleges may
have used price discrimination to retain additional PGIB tuition funds without increasing
enrollment of recipients (potentially by changing the distribution of their students between
veterans and non-veterans), but that there is no symmetric decrease in sticker price in states
where the maximum PGIB benefit decreased.
The relationship between financial aid and sticker price tuition in the previous literature
is unclear. Among traditional colleges, the evidence of manipulating prices given a change in
financial aid is mixed. Long (2004) examines changes in Georgia college and universities given
the introduction of the HOPE scholarship and finds that four year institutions did increase tuition
pricing after the introduction of HOPE5; however Cornwell et al. (2006) find that HOPE
scholarship also increased the number of students attending Georgia colleges and universities.
Singell and Stone (2007) use panel data to examine variations in financial aid policies and find
that public universities (whose tuition prices are generally regulated by a state board) do not
respond to increases in aid; however private universities and non-resident tuition pricing do
respond to these changes. Turner (2012) shows that universities decrease merit aid when
students report using tax breaks for college education.
For profit universities are a relatively new area of research in this domain. A recent
paper by Turner (2014) uses a regression kink identification and finds that colleges respond to
increases in Pell Grants by lowering institutional merit aid and thus increasing the price of
college that students actually pay (a form of price discrimination). Cellini and Goldin (2012)
compare for-profit colleges that are eligible for federal financial aid to those institutions that are
just below the eligibility cutoff. The authors find that for-profit universities that are eligible for
federal funds have higher sticker price tuition as compared to those colleges not eligible for
programs like subsidized student loans or Pell Grants. Cellini (2010) finds that the number of
for-profit institutions that entered in a given county increased when certain aid programs such as
Pell Grants, Cal Grants, and GI Bill were also increased in California.
5
Founded in 1993, HOPE scholarship is a lottery funded scholarship in Georgia that provides
students with a significant portion of in-state tuition at a state college or university. Students
need to graduate with at least a 3.00 GPA from a Georgia high school to be eligible.
Despite the large numbers of beneficiaries and resources spent on the program there is
surprisingly little research done on the effects of changes in the PGIB on student behavior; this is
likely due to the new and evolving nature of this benefit. However, Barr (2015) showed that the
PGIB increased college enrollment of veterans by fifteen to twenty percent and increased the
number of veterans enrolled at (relatively expensive) four year institutions. On a related note,
Barr (2016) shows that increased state, merit aid programs reduce military enlistments, thus
showing an interesting trade-off between college enrollment and military enlistments. Finally,
Congress and others have expressed concerns about the extent to which PGIB funds are going to
for-profit colleges.
This study contributes to all of these strands of the literature by being among the first
papers that examines the relationship between expanded veterans’ education benefits and the
pricing behavior of for-profit colleges, and by investigating non-symmetric responses to shifts in
financial aid maximums. The structure of the paper is as follows: Section 2 introduces
background information regarding the PGIB. Section 3 discusses the data. Section 4 describes
the empirical strategy and models. Section 5 presents results and Section 6 concludes.
2. Background
Figure 1 presents the timeline of events. In January 2007, Senator Jim Webb introduced
the original bill. After passing through the Senate and House of Representatives, the bill was
signed into law by President George W. Bush in June 2008. The bill went into effect in August
2009. The PGIB automatically enrolled all service members that meet enlistment standards.
Under PGIB, tuition is paid directly to the institutions, and in the first few years the maximum
amount covered differed depending on the maximum in-state tuition at any public university in
the state. This introduced wide variation, with the maximum tuition payment in Delaware as low
as $665, while in Colorado the maximum allowable tuition was set at $43,035.7 Then in 2010, a
revision of PGIB was passed which changed the state-by-state maximum tuition to a universal
national maximum tuition rate of $17,500.8
Figure 2 displays whether a given state saw an increase or decrease in PGIB with the
above changes. Thirteen states saw a decrease in the PGIB, while the Department of Veterans
Affairs increased the maximum reimbursement rates in 37 other states. These events provide
two exogenous shocks to the willingness to pay for a large portion of students in for-profit
schools, namely qualified service members, veterans, and their qualifying dependents. Figure 3
presents the by-state trends in the maximum. With the first PGIB passage, there is wide variation
in the maximum tuition, and for-profit institutions may have incentive to locate, increase
enrollment, and/or increase tuition costs in high-maximum tuition states, and alternatively to
move out of, decrease enrollment, and/or decrease tuition costs in low-maximum tuition states.
On the flip side, when the maximum was set at the national level, states that previously had high
maximum tuitions were now relatively low, while states there were low became relatively high.
The underlying theory of behavior of the for-profit institutions is relatively
straightforward: in an attempt to maximize profits, institutions will attempt to capture increases
in subsidization via financial aid changes through increases in tuition prices. This pricing
behavior is possible because veterans are unable to retain any unspent portions of the benefit, and
have no financial incentive to seek a school that has lower sticker prices as long as those sticker
7
See http://www.benefits.va.gov/GIBILL/resources/benefits_resources/rate_tables.asp for a list of
maximum tuition rates by year
8
There were a few states that had grandfathered maximum tuition rates through 2013 for service
members already enrolled prior to the change. These states are Arizona, Michigan, New Hampshire, New
York, Pennsylvania, South Carolina, and Texas.
prices are covered by PGIB. Indeed, for-profit schools must recruit new students to succeed as a
business.10
The incentives surrounding enrollment and the net effect on enrollment are less clear
theoretically. Schools may face capacity constraints or may be able to garner higher overall
profit through constraining enrollment before any capacity constraints. Even if there is a desire to
increase enrollment in the face of increased benefits, a program such as PGIB may have the
effect instead of changing the distribution of veteran and non-veteran students.
3. Data
We use for our data the Integrated Postsecondary Education Data System (IPEDS). IPEDS
contains data collected from the universe of higher education institutions that participate in
federally funded assistance programs (Title IV), e.g. Pell Grants, subsidized loans, and, relevant
to this paper, PGIB.11 Institutions must report several statistics annually as part of their
qualification for federal funds.
We use IPEDS data from 2003 to 2013. These data span before and after the changes in
PGIB benefits. Figure 4 shows the total enrollment in school types (omitting four-year non-profit
10
Steinerman et al. (2011) in a 2009 survey of for-profit schools found that on average 11% of
revenue was spent on advertising, and the average student costs around $4,000 to recruit, which
can be quickly recaptured in tuition. Veterans may be systematically pursued because of their
financial aid availabilities and there has been at least some success: veterans are about 5 times as
likely to be enrolled in for-profit schools than non-profit schools (Steele, McGovern, and Buryk
2013).
11
Another condition is that the institutions cannot receive over 90% of revenue from Title IV
sources. The average for for-profit schools in 2014 was over 70% of revenue received from Title
IV sources, as opposed to 30% for non-profit schools. Approximately 20% of for-profit schools
receive between 85 and 90% of their revenue from Title IV sources, demonstrating their reliance
and thus likely responsiveness to financial aid changes.
schools, which is relatively steadily increasing over this time span and is an order of magnitude
larger. There is volatility over time, largest for for-profit schools with sharp declines after 2009.
Figure 5 presents the average sticker price for in-state students across all institutions. The
trends are more stable than the total enrollment numbers, generally showing an increase,
although with a decrease after 2009 4 year and 2 year for-profit schools.
4. Econometric Model
We base our identification on the plausibly exogenous changes in maximum reimbursement
rates. This policy change allows for a difference-in-difference dosage type estimator, where the
first difference is across time (before and after the maximum reimbursement changes) and the
second difference is across states that have different levels of changes in the maximum
reimbursement. We use both the initial state maximum in 2008, which we argue is unanticipated
by for-profit colleges (as it is a policy parameter of the new PGIB) as well as the change in
maximum reimbursement rates from 2010 to after (the collapse to the national level shown in
Figure 2). Equation 1 gives the specification for this model:
𝑌𝑖𝑠𝑡 = 𝛼 + 𝛽1 𝑀𝐴𝑋𝑠2008 + 𝛽2 (𝑀𝐴𝑋2011 − 𝑀𝐴𝑋𝑠2010 ) + 𝛾𝑡 + 𝜏1 𝑀𝐴𝑋𝑠2008 × 1(𝑡 ∈ 2008,2010)
(1)
+ 𝜏2 (𝑀𝐴𝑋2011 − 𝑀𝐴𝑋𝑠2010 ) × 1(𝑡 > 2010) + 𝑋𝑖𝑠𝑡 𝛿 + 𝜀𝑖𝑠𝑡
We measure the outcome for institution i in state s in year t. The treatment effects of interest are
captured by 𝜏1 and 𝜏2 , which capture the responsiveness of outcome Y to a dollar increase in
veteran financial aid maximums. We give particular attention to 𝜏2 , which measures the
exogenous shift from the state-level maximums to the converged national maximum, and
measures each additional difference in this change (positive and negative). We additionally
control for other factors, including the log of the state population, the state unemployment rate,
whether the institution is a degree-granting institution, and the level of the degree programs
offered (2 year, 4 year, graduate, etc.). Standard errors are clustered at the state level. We
estimate this model for all school types aggregated as well as separately for 4 year, 2 year, and
less than 2 year schools.
We additionally explore heterogeneity in equation 1. We first allow 𝜏2 to differ
depending on whether 𝑀𝐴𝑋2011 − 𝑀𝐴𝑋𝑠2010 is positive or negative. This allows us to see whether
for-profits increase prices with increases in the maximum as well as whether they decrease their
prices with decreases in the maximum.
We also will examine they hypothesis deeper by trying to identify if schools that have
higher veteran populations are more responsive. We unfortunately do not have data on the
fraction of the students in each institution that are veterans. Instead, we use whether the
institution is in a state that has above-median proportion of the student-age population that are
veterans, as well as a stronger case of whether the institution is above the 75th percentile for the
same measure.
5. Results
Table 1 presents the regression outcomes for listed sticker price in-state tuition rates. Column 1
aggregates all school types for the basic model. We find a small, negative, insignificant
coefficient for the initial state price. We find a positive coefficient for the change from the state
maximum to the converged maximum, and it is quite a bit larger than the coefficient on the
initial state price. It is still insignificant. It finds that a hundred-dollar larger increase in the
change to the maximum is associated with a 1.5 dollar increase in the sticker price. While this
seems quite low, on average only 10 percent of for-profit students are veterans.
We then separate out the effect for positive and negative changes in the maximum. The
effect is largest for the 4-year for-profit schools, and is marginally significant at the 10 percent
level. The coefficient suggests around a 5.5 dollar increase for a 100 dollar increase in the
maximum tuition. With again only 10 percent of for-profit students being veterans, this effect is
a substantial response. Interestingly, the effect is not there for negative changes. Institutions only
respond on average to increases in financial aid, but don’t decrease prices with decreases in their
student maximum benefits. We find small and statistically insignificant effects for two year and
less than two year programs.
Next, we estimate the same model with the natural log of students admitted to the
institution. This variable is important to help distinguish whether the result indicating an
increase in sticker price tuition is a result of increased demand for education or price
discrimination. If the change in PGIB encouraged universities to admit more veterans who now
had more resources to pay for college, then we would also expect the tuition result to be simply a
change in demand. However, if for-profit universities are price discriminating to capture the
increased benefit, then colleges may not admit more students, but simply charge the current
number of students an increased price. Table 2 shows the results for universities using the same
difference in differences routine used previously. We find no evidence that universities admitted
more students in response to a change in PGIB. This result is evidence that for profit universities
may have increased their tuition price in response to the increase in financial aid without
admitting more students.
One possible reason for this increase in tuition pricing could be shifts in demand from
veterans given the increase in PGIB. Also, if the PGIB encouraged more veterans to enroll in
postsecondary education, then one goal of the program would be accomplished. We estimate
the same model with the natural log headcount enrollment. Table 3 displays results this model.
We find that in states where the PGIB was increased to the national maximum, there was no
economic nor statically significant increase in enrollments. This result is important because
while veterans are a small portion of the student population, if colleges were simply responding
to increases in demand, then we would see an accompanying increase in enrollments. However
the result indicates that the increase in sticker price tuition is evidence an increase of price
discrimination by for-profit colleges that are capturing the increased subsidy.
Robustness Check-High Veteran States
Since IPEDS is comprised of institutional level data, one concern could be whether the
result is actually motivated by changes in pricing for veterans. To address this concern, we reestimate the model by interacting our difference in differences estimator with a dummy variable
for whether the state has an above the median veteran population. Table 4 presents the results of
this robustness check. First, we consider institutions that IPEDS has labeled as four year
institutions. While the interaction term with whether a state has an above median number of
student aged veterans is positive but statistically insignificant, our main result (whether PGIB
benefits were increased in a given state) is not only robust but is stronger. The coefficient of
0.104 shows that if the tuition benefit increases by $100, then a for-profit institution will increase
its sticker price by $10.40; roughly capturing 10 percent of the increased benefit.
Next, we estimate the same model, however we interact our main result by with a dummy
variable for whether the state’s student age veteran population is in the top quartile. Table 5
presents the results for this robustness check. We find that the main effect is similar in
magnitude and sign to the previous finding with a $10.50 increase of a change of $100. These
two robustness checks are evidence that states with high student aged veteran populations are
driving the main result in Table 3. Interestingly, as in all specifications, we do not find a
statistically significant effect for institutions that were labeled by IPEDS as two year as or less
than two year programs.
6. Conclusion
The Post 9/11 GI Bill is one of the largest sources of publicly provided financial aid. The
expansion of the GI Bill benefits in the most recent authorization has the potential to provide
opportunities for active duty military members, dependents, and their families. However, one
concern could be whether an increased benefit provides more access to higher education for this
population or if colleges and universities simply increase their sticker price tuition to capture
more financial aid dollars. If the latter is true, then the social welfare benefits of increasing
PGIB benefits would be unclear.
To identify an effect of increasing financial aid benefits on a university’s sticker price
tuition, we use an exogenous policy change to PGIB where various state specific maximum
amounts were changed to one nationwide amounts. This policy change essentially caused an
increase in PGIB benefits in some states while other states saw a decrease. In this study, we
estimate a difference in differences estimator using the exogenous variation in changes to the
PGIB benefit to determine if changes in tuition is price discrimination or a shift in demand for
higher education.
We find that in states where the benefit increased, the sticker price of tuition increased
$5.05 for every $100 of extra benefit. However, we only find this result for colleges that are
labeled as a four year school by IPEDS. We find small and statistically insignificant effects for
two year and less than two year institutions. We also estimate our model using admissions and
enrollments as alternative dependent variables to ascertain whether the increase in sticker price
tuition is price discrimination or allocation of scarce seats to an increase in demand. We find no
evidence that for-profit colleges increased their admissions or enrollments as a result of the
policy change.
As a robust check, we interact our difference in differences estimator with whether the
state population of student aged veterans is greater than the national median. We find that our
results are robust and the magnitude increases such that a $100 increase in PGIB benefit
increased sticker price tuition by $10.10. This magnitude represents that the institution captures
around ten percent of the increased benefit for no additional enrollment of veterans. Thus the
Department of Veteran Affairs may be paying more for the same amount of higher education.
Works Cited
Barr, Andrew. 2015. “From the Battlefield to the Schoolyard: The Short-term Impact of the Post9/11 GI Bill.” Journal of Human Resources 50, no. 2: 580-613
Barr, Andrew. 2016. “Enlist or Enroll: Credit Constraints, College Aid, and the Military
Enlistment Margin.” Economics of Education Review 51: 61-78.
Bennett, William J. 1987. “Our Greedy Colleges.” New York Times, February 18:
http://www.nytimes.com/1987/02/18/opinion/our-greedy-colleges.html.
Cellini, Stephanie Riegg, and Claudia Goldin. 2014. “Does Federal Student Aid Raise Tuition?
New Evidence on For-Profit Colleges.” American Economic Journal: Economic Policy 6:
174-206.
Cellini, Stephanie Riegg. 2010. “Financial Aid and For-Profit Colleges: Does Aid Encourage
Entry?” Journal of Policy Analysis and Management 29, no. 3: 526-552.
Cornwall, Christopher, David B. Mustard, and Deepa J. Sridhar. 2006. “The Enrollment Effects
of Merit-Based Financial Aid: Evidence from Georgia’s HOPE Scholarship.” Journal of
Labor Economics 24, no. 4: 761-786.
Deming, David J, Claudia Goldin, and Lawrence F. Katz. 2012. “The For-Profit Postsecondary
School Sector: Nimble Critters or Agile Predators?” Journal of Economic Perspectives
26, no.1: 139-164.
Gilpin, Gregory A., Joseph Saunders, and Christiana Stoddard. 2015. “Why Has For-Profit Colleges’
Share of Higher Education Expanded so Rapidly? Estimating the Responsiveness to Labor
Market Changes.” Economics of Education Review 45: 53-63.
Health, Education, Labor and Pensions Committee, U.S. Senate. 2010 “Benefiting Whom? ForProfit Education Companies and the Growth of Military Educational Benefits.”
December 8th. http://harkin.senate.gov/documents/pdf/4eb02b5a4610f.pdf
Hefling, Kimberly. 2016. “GI Bill Funds Still Flow to Troubled For-Profit Colleges.” Politico.
February 25. http://www.politico.com/story/2016/02/veterans-education-for-profitcolleges-219758.
Long, Terry Bridget. 2004. “How Do Financial Aid Polices Affect Colleges? The Institutional
Impact of the Georgia HOPE Scholarship.” Journal of Human Resources 39, no. 4: 10451066.
Singell, Larry D., and Joe A. Stone. 2007. “For Whom the Pell Tolls: The Response of
University Tuition to Federal Grants-in-Aid.” Economics of Education Review 26, no. 3:
285-295.
Steele, Jennifer, Geoffrey McGovern, and Peter Buryk. 2013. “Student Veterans’ Outcomes by
Higher Education Sector: Evidence From Three Cohorts of the Baccalaureate and
Beyond.” Working Paper.
Steinerman, Andrew, Jeffrey Volshteyn, and Molly McGarrett. 2011. Education Services Data
Book. J.P. Morgan, North American Equity Research, Business and Education Services.
Turner, Leslie J. 2014. “The Road to Pell is Paved with Good Intentions: The Economic
Incidence of Federal Student Grant Aid.” Working Paper.
Turner, Nicholas. 2012. “Who Benefits from Student Aid? The Economic Incidence of TaxBased Federal Student Aid.” Economics of Education Review 31, no. 4: 463-481.
Tables and Figures
Figure 1: Timeline of Events
December 2010: PGIB
revision passed; sets max.
tuition at national level
June 2008: PGIB
signed into law
2007
January 2007: PGIB
first introduced in
senate
2008
2009
2010
August 2009: PGIB into
effect, max. tuition at state
level
2011
2012
August 2011: Revision into
effect, max. tuition at national
level
Figure 2. States where PGIB increased versus states with a decrease.
Figure 3: Maximum PGIB Tuition Reimbursement Rates by State
Figure 4: Total Student Enrollment over time by school type.
Figure 5: Average in-state sticker price
Table 1. Effects of Change in GI Bill on "Sticker Price" Tuition at For-Profit Colleges
VARIABLES
First State Max
Change in Max to Nat.
(2008-2010) x First State Max
(>2010) x Change in Max. to Nat.
(1)
All
(2)
All
(3)
4 Year
(4)
2 Year
(5)
<2 Year
-0.00734
(0.0109)
-0.00922
(0.0100)
-0.00203
(0.00853)
0.0149
-0.00475
(0.0110)
-0.00889
(0.0101)
-0.00458
(0.00919)
0.00857
(0.0149)
-0.00845
(0.0151)
-0.00690
(0.00976)
-0.0447**
(0.0219)
-0.0129
(0.0134)
0.00254
(0.0245)
0.0603
(0.0393)
0.0273
(0.0230)
0.0336
(0.0870)
0.00640
0.00907
-0.0133
-0.105
(0.0136)
0.0406
(0.0126)
0.0547*
(0.0267)
0.00238
(0.0789)
-0.0190
(0.0268)
1.265***
(0.200)
-0.418***
(0.126)
-1.970
(3.282)
(0.0289)
1.364***
(0.377)
-0.370
(0.278)
-4.534
(5.194)
(0.0481)
1.374***
(0.228)
-0.417**
(0.168)
-5.334
(3.679)
(0.118)
0.0787
(0.415)
-0.763***
(0.223)
14.80*
(8.160)
(0.0104)
Neg. x (>2010) x Change in Max.
to Nat.
Pos. x (>2010) x Change in Max.
to Nat.
Log(Population)
Unemployment Rate
Constant
Observations
R-squared
1.258***
(0.204)
-0.412***
(0.125)
-1.874
(3.366)
11,481
11,481
5,604
4,723
1,154
0.128
0.128
0.070
0.067
0.197
Standard Errors are Clustered at the State Level and in Parentheses
*** p<0.01, ** p<0.05, * p<0.1
Estimates include year fixed effects, whether the institution is degree granting and controls for
the level of the degree programs offered at the institution.
Table 2. Effects of Change in GI Bill on Log Admissions at For-Profit Colleges
VARIABLES
First State Max
Change in Max to Nat.
(2008-2010) x First State Max
(>2010) x Change in Max. to Nat.
(1)
All
(2)
All
(3)
4 Year
(4)
2 Year
(5)
<2 Year
0.00166
(0.00331)
0.00121
(0.00188)
-0.00144
(0.00315)
0.00387
(0.00311)
0.00156
(0.00334)
0.00119
(0.00191)
-0.00135
(0.00322)
-0.00486
(0.00421)
0.00277
(0.00291)
0.00246
(0.00313)
0.00576
(0.00426)
0.000110
(0.00196)
-0.000408
(0.00351)
0.00583
(0.00680)
6.66e-05
(0.00452)
-0.0112
(0.00742)
0.00458
0.00766
-0.00328
0.000272
(0.00460)
0.00220
(0.00606)
-0.00156
(0.00783)
0.00334
(0.00932)
0.00384
0.142***
(0.0531)
0.0638*
(0.0329)
(0.00941)
0.142***
(0.0530)
0.0642*
(0.0325)
(0.00985)
0.0737
(0.0969)
0.0329
(0.0482)
(0.0182)
0.227***
(0.0562)
0.0416
(0.0409)
(0.0264)
0.142*
(0.0770)
0.103**
(0.0445)
1.630**
(0.719)
1.628**
(0.715)
3.950***
(1.363)
-0.0263
(0.780)
1.643
(1.151)
Neg. x (>2010) x Change in Max.
to Nat.
Pos. x (>2010) x Change in Max.
to Nat.
Log(Population)
Unemployment Rate
Constant
Observations
R-squared
8,264
8,264
2,785
2,764
2,715
0.126
0.126
0.029
0.163
0.060
Standard Errors are Clustered at the State Level and in Parentheses
*** p<0.01, ** p<0.05, * p<0.1
Estimates include year fixed effects, whether the institution is degree granting and controls for
the level of the degree programs offered at the institution.
Table 3. Effects of Change in GI Bill on Log Enrollments at For-Profit Colleges
VARIABLES
First State Max
Change in Max to Nat.
(2008-2010) x First State Max
(>2010) x Change in Max. to Nat.
(1)
All
(2)
All
(3)
4 Year
(4)
2 Year
(5)
<2 Year
0.00148
(0.00465)
0.00342
(0.00246)
-0.00316
(0.00255)
-0.000500
(0.00360)
0.00166
(0.00467)
0.00345
(0.00248)
-0.00333
(0.00272)
0.00292
(0.00345)
0.00653**
(0.00259)
0.00128
(0.00291)
0.00487
(0.00601)
0.00336
(0.00350)
0.000446
(0.00389)
-0.00465
(0.0149)
-1.54e-05
(0.00602)
-0.0178***
(0.00658)
-0.00175
-0.00386
-0.00211
0.00560
(0.00484)
0.00243
(0.00478)
0.00543
(0.0113)
0.00169
(0.00760)
-0.0442
(0.0134)
0.173**
(0.0737)
0.0961
(0.0623)
-0.162
(0.952)
(0.0117)
0.0988*
(0.0583)
0.0331
(0.0311)
2.680***
(0.871)
(0.0303)
0.378***
(0.0956)
0.0401
(0.0735)
-3.925***
(1.401)
(0.0274)
0.133
(0.105)
0.183*
(0.0952)
-0.655
(1.553)
Neg. x (>2010) x Change in Max.
to Nat.
Pos. x (>2010) x Change in Max.
to Nat.
Log(Population)
Unemployment Rate
Constant
Observations
R-squared
0.173**
(0.0735)
0.0968
(0.0620)
-0.166
(0.953)
7,658
7,658
2,683
2,604
2,371
0.238
0.238
0.065
0.179
0.052
Standard Errors are Clustered at the State Level and in Parentheses.
*** p<0.01, ** p<0.05, * p<0.1
Estimates include year fixed effects, whether the institution is degree granting and controls for
the level of the degree programs offered at the institution.
Table 4. Robustness Check-Differences in States with Above Median Student Age Veteran
Population on Tuition Results
VARIABLES
First State Max
Change in Max to Nat.
(2008-2010) x First State Max
(>2010) x Change in Max. to Nat
(1)
All
(2)
All
(3)
4 Year
(4)
2 Year
(5)
<2 Year
-0.0198*
(0.0104)
-0.00714
(0.00981)
0.000297
(0.00854)
0.00711
(0.0120)
-0.0146
(0.0109)
-0.00636
(0.00973)
-0.00507
(0.00960)
0.000162
(0.0126)
-0.00743
(0.0143)
-0.00819
(0.0109)
-0.0582***
(0.0214)
-0.00990
(0.0126)
0.0126
(0.0221)
0.0519
(0.0336)
0.0309
(0.0230)
-0.0123
(0.0618)
-0.00342
(0.0146)
0.0543
(0.0331)
-1.550***
(0.366)
-0.00142
(0.0146)
0.104**
(0.0470)
-1.328**
(0.649)
-0.0423***
(0.00994)
-0.00971
(0.0521)
-1.540***
(0.557)
0.127
(0.174)
0.0132
(0.133)
-2.494***
(0.678)
1.788***
(0.218)
-0.340***
(0.0931)
0.00922
(0.0207)
0.0121
(0.0362)
1.813***
(0.213)
-0.349***
(0.0944)
0.0208
(0.0225)
-0.0439
(0.0435)
1.924***
(0.445)
-0.352
(0.218)
0.0658***
(0.0217)
0.0218
(0.0547)
2.050***
(0.205)
-0.360***
(0.102)
-0.298*
(0.171)
0.0463
(0.0530)
0.791**
(0.327)
-0.505***
(0.161)
-10.08***
(3.571)
-10.44***
(3.508)
-12.83**
(6.149)
-15.59***
(3.499)
3.075
(6.323)
11,481
0.138
11,481
0.138
5,604
0.079
4,723
0.078
1,154
0.234
Neg. x (>2010) x Change in Max. to Nat.
Pos. x (>2010) x Change in Max. to Nat.
Above Median Veterans
Above Median x (>2010) x Change in Max. to Nat
-1.503***
(0.345)
0.0205
(0.0137)
Above Median x Neg. x (>2010) x Change in Max. to Nat.
Above Median x Pos. x (>2010) x Change in Max. to Nat.
Log(Population)
Unemployment Rate
Constant
Observations
R-squared
Standard Errors are Clustered at the State Level and in Parentheses
*** p<0.01, ** p<0.05, * p<0.1
Estimates include year fixed effects, whether the institution is degree granting and controls for
the level of the degree programs offered at the institution.
Table 5. Robustness Check-Differences in States with Above 75th Percentile Student Age
Veteran Population on Tuition Results
(1)
(2)
(3)
(4)
(5)
VARIABLES
All
All
4 Year
2 Year
<2 Year
First State Max
-0.016
(0.010)
-0.011
(0.010)
-0.000
(0.009)
0.008
(0.012)
-0.0120
(0.0103)
-0.0107
(0.00997)
-0.00489
(0.00986)
0.00216
(0.0125)
-0.0104
(0.0144)
-0.00757
(0.0118)
-0.0574**
(0.0228)
-0.0147
(0.0143)
0.0126
(0.0220)
0.0531
(0.0414)
0.0253
(0.0229)
-0.0196
(0.0852)
-0.00296
(0.0162)
0.0542*
(0.0312)
-0.711
(0.449)
-0.00257
(0.0158)
0.105**
(0.0456)
-0.886
(0.887)
-0.0423***
(0.0128)
-0.0201
(0.0485)
-0.148
(0.536)
0.0785
(0.207)
0.0271
(0.145)
-1.547**
(0.695)
0.0128
0.0219
0.0773***
-0.241
(0.0248)
0.00590
(0.0312)
-0.0503
(0.0261)
0.0252
(0.201)
0.00128
1.426***
(0.211)
-0.368***
(0.106)
(0.0386)
1.456***
(0.212)
-0.371***
(0.108)
(0.0399)
1.705***
(0.460)
-0.345
(0.235)
(0.0599)
1.458***
(0.280)
-0.391**
(0.163)
(0.0603)
0.467
(0.454)
-0.568***
(0.190)
Constant
-4.521
(3.389)
-4.975
(3.385)
-9.745
(6.306)
-6.702
(4.227)
7.931
(8.330)
Observations
R-squared
11,481
0.130
11,481
0.130
5,604
0.075
4,723
0.069
1,154
0.211
Change in Max to Nat.
(2008-2010) x First State Max
(>2010) x Change in Max. to Nat
Neg. x (>2010) x Change in Max. to Nat.
Pos. x (>2010) x Change in Max. to Nat.
Above 75th percentile Veterans
Above 75th percentile x (>2010)
x Change in Max. to Nat
-0.645*
(0.368)
0.019
(0.016)
Above 75th percentile x Neg. x (>2010)
x Change in Max. to Nat.
Above 75th percentile x Pos. x (>2010)
x Change in Max. to Nat.
Log(Population)
Unemployment Rate
Standard Errors are Clustered at the State Level and in Parentheses
*** p<0.01, ** p<0.05, * p<0.1
Estimates include year fixed effects, whether the institution is degree granting and controls for
the level of the degree programs offered at the institution.