Educational Policy

Volume 23 Number 5
September 2009 714-731
© 2009 Corwin Press
Does State Funding Affect 10.1177/0895904808321270
http://epx.sagepub.com
Graduation Rates at Public hosted at
http://online.sagepub.com

Four-Year Colleges and

Universities?
Liang Zhang
Vanderbilt University, Nashville, TN

This study uses panel data to examine the direct link between state funding
and graduation rates at 4-year public institutions. We find some evidence for
a positive association between state funding and college graduation rates.
When other factors are held constant, a 10% increase in state appropriations
per full-time equivalent (FTE) student at 4-year public institutions is associ-
ated with approximately a 0.64 percentage point increase in graduation rates.
This positive link appears to hold for both institutions that have enjoyed an
increase in state funding and those that have experienced a reduction. In addi-
tion, the positive association seems to hold for all research/doctoral, master’s,
and baccalaureate institutions. These estimated effects are smaller and, in
general, statistically insignificant when both institution and time-fixed effects
are controlled.

Keywords: state funding; graduation rate; higher education

Introduction

It is no longer a secret that higher education is often regarded as a

discretionary item in many state budgets. At a time of favorable state fiscal
environments, Hovey (1999) predicted that many states would experience
significant difficulties in maintaining their public services over the following
decade. Unfortunately, that forecast has been especially true for public
colleges and universities. Rizzo (2006) documented three major changes in
state funding of education that occurred during the last quarter of the 20th
century: (a) the decline in education’s share of state budgets, (b) the decline

Author’s Note: Any correspondence concerning this article should be addressed to Liang
Zhang at Peabody #414, 230 Appleton Place, Nashville, TN 37203; e-mail: liang.zhang@
vanderbilt.edu.

714

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 715

in higher education’s share of state educational funding, and (c) the decline
in the share of higher education funding that went to public higher educa-
tion. Although for most years during the last quarter of the 20th century the
actual dollar amount of state funding was not reduced—even in real terms—
the share of public institutions’ revenues from state appropriations decreased
from approximately 44% in early 1980s to approximately 32% in recent
years due to increases in college costs and in college enrollments (National
Center for Education Statistics, 2005, tables 171 and 329).
Part of the predicament facing public higher education is a lack of
evidence to show the adverse effects of reduced state funding. Skolnick
(1986) wrote succinctly: “If the cut is so deep, where is the blood?” If higher
education institutions have been as severely hurt by financial limitations as
they claim, why have researchers not uncovered that damage? Analysis of
the influence of public funding is essential if public institutions are to make
the case to legislatures and governors that improved funding will enable them
to better serve the public. Without such evidence, the structural deficits
faced by many states (Hovey, 1999), coupled with growing competition for
public funding among many state functions, will likely result in a continu-
ation of relative decline in state funding for public higher education.
This study uses panel data to address whether state funding influences
college graduation rates at public 4-year institutions. We use eight cohorts
(from academic years 1991-1992 to 1998-1999) of undergraduate students
enrolled at 4-year public institutions in the United States to analyze this issue.
Our data come from various components of the Integrated Postsecondary
Education Data System (IPEDS) and The College Entrance Examination
Board’s Annual Survey of College Standard Research Compilation data files
(henceforth College Board data). This study differs from previous studies on
similar topics in two ways. First, whereas previous studies analyzed the
impact of certain intermediate factors (such as instructional expenditures and
faculty employment) on graduation rate (e.g., Ehrenberg & Zhang, 2005a;
Scott, Bailey, & Kienzl, 2006), the current analysis presents a direct test of
whether state funding affects cohort graduation rates at public institutions.
Second, the panel data approach used in this article alleviates the problem
that the estimated effects of financial resources may have been confounded
by unobserved institutional characteristics.

Literature Review

Research on college retention and degree attainment has often been con-
ducted at the individual level. In particular, much of research along this line

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

716 Educational Policy

has been heavily influenced by Tinto’s (1993) interactionalist model of

student persistence. Indeed, the prominence of Tinto’s model has made it
the de facto starting point for research in this particular area (Braxton &
Hirschy, 2005). According to Tinto’s model, it is the interaction between
student characteristics (including commitments to their educational goals
and institutions) and the academic and social contexts of the institutions
that ultimately determines students’ college persistence and graduation. The
most stripped-down finding is that being female, White, high ability, and
from high-SES families is associated with a higher probability of retention
(Berger & Milem, 1999; Elkins, Braxton, & James, 2000). Over the years,
researchers have reconceptualized and elaborated Tinto’s model to better
capture the impact of increasing students diversity on retention behaviors
(e.g., Braxton & Hirschy, 2005).
This study builds on and extends the established literature on college
graduation by focusing on institutional-level graduation rates rather than
predicting individual success in college. Using institutional-level aggre-
gated graduation rates as an outcome variable is not without risks. For
example, Astin (1993) pointed out that more than half of the variance in
institutional retention rates could be attributed to the variations in one of the
key inputs of higher education, namely, students themselves, rather than to
the kinds of educational experiences they receive on campuses. Nevertheless,
graduation rates remain to be one of the most popular measures of institu-
tional performance and continue to draw an increasing amount of attention
from policymakers in light of the rising issue of institutional accountability.
Although this study focuses on institutional-level graduation rates, it is
important to include important student characteristics that have been shown
to be significant predictors for college graduation in the empirical model after
they are properly aggregated. These characteristics include, for example,
age, gender, race/ethnicity, test scores, attendance status, and whether living
on campus or not.
The institutional-level approach enables a direct test of the link between
state funding and college graduation rates at public colleges and universities.
The impact of financial resources on such institutional performance indica-
tors as graduation rates has recently drawn much attention within the research
community; however, most research in this area focuses on the relationship
between educational expenditures and graduation rates. Educational expen-
ditures are often measured by the standard IPEDS expenditure categories
including instructional expenditures, academic support expenditures, or edu-
cational and general expenditures (see Toutkoushian, 2001, for a summary of
IPEDS basic and derived expenditure categories). Several recent studies used

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 717

cross-sectional samples of 4-year colleges and universities to examine the

impact of institutional expenditures on cohort graduation rates. For example,
Ryan (2004) found a positive and significant relationship between instruc-
tional and academic support expenditures and 6-year cohort graduation rates
for both public and private institutions. Blose, Porter, and Kokkelenberg
(2006) used a similar approach to examine the effect of institutional expen-
ditures on graduation rates at public institutions and found that the adjusted
expenditures (for academic disciplines) per full-time equivalent (FTE) student
had a statistically significant and positive impact on cohort graduation rates.
Scott et al. (2006) sought to determine whether public institutions were
performing better than private institutions, given the same level of financial
resources and similar student populations; they found that for both public and
private institutions, instructional expenditure per student had a positive and
statistically significant effect on cohort graduation rate, although the magnitude
of the effect was rather small.
The evidence of the link between educational expenditures and gradua-
tion rates is important because institutions may wish, accordingly, to redirect
financial resources internally to improve their graduation rates; however,
internal resource allocation is often significantly influenced by revenue
sources. Revenues generated from difference sources (e.g., tuition and fees,
grants and contract, endowment, and state appropriations) have direct bearings
on how these dollars might be spent. Empirical studies have supported the
notion of resource dependence (Pfeffer & Salancik, 1978) in higher educa-
tion finance, which holds that internal organizational activities are influenced
primarily through the actions of external resource providers. For example,
according to resource dependence, higher education institutions respond to
changes in state funding. Changes in state appropriation have direct impact
on expenditures that are related to instruction. Hasbrouck (1997) found that
instructional expenditures were consistently and strongly predicted by state
appropriation and only modestly by gift, grants, and contract revenues. In
other words, a decline in state appropriation would most likely lead to a
reduction in instructional expenditures because it is unlikely that public
institutions could fully compensate for the reduction in state funding from
other revenue resources through internal resource reallocation. In addition,
state appropriations have also been shown to affect other expenditures that
are related to instruction, including, for example, facility maintenance and
operation. More subtly, changes in the resource environment of public insti-
tutions may change their activities. Changes in revenue streams could result
in changes in the composition of faculty (e.g., Ehrenberg & Zhang, 2005a)
and a trade-off in faculty time allocation, particularly to the detriment of

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

718 Educational Policy

undergraduate instruction (Barnett & Middlehurst, 1993; Slaughter &

Leslie, 1997).
It is, therefore, important to examine the impact of state funding on
institutional performances. Somewhat surprisingly, very few studies have
addressed whether reduced state funding for public higher education has
impacted student outcomes, resulting, for example, in less learning, longer
time to degree, and lower graduation rates. We only found a couple of
studies that addressed this issue at the state level. Kelly and Jones (2005)
used cross-sectional state-level data to examine the relationship between
state funding and performance in a variety of areas, including graduation
rates and participation rates. Their study found weak correlation between
state funding and performance. In another study, Bound and Turner (2004)
used almost a half century of state-level data to examine whether financial
resources affect college attainment by exploiting the exogenous variation
generated by cohort sizes. Results indicated that large cohorts within states
have relatively low baccalaureate attainment, suggesting that lower public
subsidies due to cohort crowding could be one explanation.
Although the relationship between public funding and educational
performance at the state level remain unclear, two recent studies by Ehrenberg
and Zhang (2005a, 2005b) suggested a possible link between state funding
and institutional performance. They showed that the growing financial pres-
sures faced by public higher education institutions have led to increases in
the utilization of contingent faculty and reductions in tenured and tenure-track
faculty. In a second study, they found that the increased usage of contingent
faculty adversely affected graduation rates at 4-year institutions, with the
largest impact on students being at public master’s-level institutions. Together,
these two studies suggested promising areas for investigating whether
reduced state funding adversely affects such key institutional performance
indicators as graduation rates.

Data and Methods

Graduation rates of undergraduate cohorts were first added to IPEDS

Graduate Rates Survey in 1997, when 4-year institutions began to report
6-year cohort graduation rates (i.e., 150% of normal time to degree). The
6-year graduation rates reported in 1997 (as of August 31, 1997) were for the
entering freshman cohort in the fall of 1991. For the present study, the year
for which the most recent graduation rate data were available is 2004—for
the entering freshman cohort in the fall of 1998. For each entering cohort

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 719

Table 1
Descriptive Statistics for Main Variables
M SD

Cohort 6-year graduation rate 0.4325 0.1561

Log state appropriation per FTE 6,445 2,970
Log undergraduate tuition and fees 3,150 1,194
Average age of the entering cohort 19.04 1.6457
Mean SAT scores of the entering cohort (100) 10.15 1.1228
Proportion of resident students 0.8524 0.1350
Proportion of minority students 0.2665 0.2726
Proportion of full-time students 0.7159 0.2555
Proportion of male students 0.4150 0.1421

Note: FTE = full-time equivalent; SAT = scholastic assessment test.

from academic years 1991-1992 to 1998-1999, we extracted data on grad-

uation rates for each 4-year institution from IPEDS. Whereas College
Board data provided similar information on the cohort graduate rates, the
data reported by IPEDS had been adjusted for various exclusions (such as
students who died or became permanently disabled and who left school to
serve military, foreign aid, or church missions).
Detailed characteristics of each entering cohort between academic years
1991-1992 and 1998-1999 were available in College Board data. Cohort
characteristics included the average age of the entering freshman cohort,
number of students by gender and attendance status (i.e., full-time vs. part-
time), proportion of students who were minority, proportion of students from
in-state, and 25th and 75th percentile math and verbal scholastic assessment
test (SAT) scores. We then merged this detailed cohort-specific information
by cohort and institution with the data on cohort graduation rates from
IPEDS. Descriptive statistics for the main variables included in this analysis
are provided in Table 1.
We measured the level of state funding at public institutions by state
appropriation per FTE student. Data on state appropriation were available
from IPEDS Finance Survey. We computed the number of FTE students at
an institution by adding the number of full-time students and one-third of
the number of part-time students at that institution. The number of students
by attendance status was reported in IPEDS Enrollment Survey. We then
calculated state appropriation per FTE by dividing the total amount of state
appropriation at an institution by the FTE enrollment at that institution.

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

720 Educational Policy

In estimating the impacts of state funding on cohort graduation rates, it

was important to match up the cohort with financial variables. Previous
studies either used the financial variables in the year freshman cohort started
their colleges (e.g., Blose et al., 2006; Ryan, 2004) or in the year the data on
6-year graduation rates were collected (Kelly & Jones, 2005). Because col-
lege education is a multiyear experience, neither of these measures charac-
terized the financial environment to which a particular cohort was exposed.
We assumed that the relevant financial variables were those during the first
4 years that a particular cohort was enrolled in college. So, for example, the
6-year graduation rates reported by IPEDS Graduation Rates Survey in 2004
were for students who first enrolled as freshmen in the fall of 1998. Hence,
we computed the relevant financial environment that this cohort of students
experienced by averaging the financial variables their institutions reported in
the financial years 1999, 2000, 2001, and 2002 IPEDS Finance Survey.
Tuition was another important financial variable that could have influ-
enced graduation rates. The financial pressure caused by high tuition could
have led to high rates of drop-out or stop-out, both resulting in low gradu-
ation rates. Because of state funding at public institutions, their tuition rates
were kept relatively low; for example, the average tuition charged at 4-year
public institutions was approximately $3,400 in the year 2000. However, as
state funding waned, public higher education institutions might have used
tuition as a buffer. As a result, the impact of tuition increase on graduation
rates could have been regarded as an indirect effect of reduced state funding.
Because the level of tuition was an indication of not only financial burden
for students, but also institutional selectivity to a significant extent, the esti-
mated effect of tuition on graduation rates, which picks up the impact of
students and institutional characteristics when they are not adequately mea-
sured and controlled, is likely to be upward biased when a cross-sectional
sample of institutions is used in regression analysis.
The same logic applies to the estimation of other variables. For example,
because the proportion of nonresident students is probably higher at more
selective institutions than others, the effect of nonresident enrollment could
be confounded by the impact of institutional selectivity when the latter is
not adequately controlled. Similarly, a cross-sectional estimate of the impact
of state funding on graduation rates is likely to be biased when institutional
characteristics are absent from the empirical model. The panel nature of our
data allows us to control for much unmeasured institutional heterogeneity
(e.g., demographic characteristics, institutional selectivity, and other contex-
tual and historical factors) that does not change rapidly within institutions
over the time period for which we have data.

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 721

Our analytical approach is to use panel data to estimate models in which

the 6-year graduation rate of students that entered institution i in year
t(Git) is specified to be a function of the state appropriation per FTE at insti-
tution i averaged over the first 4 years (i.e., year t, t + 1, t + 2, and t + 3)
that the cohort is enrolled at the institution (Sit), the in-state undergraduate
tuition and fees changed at institution i in year t(Tit), characteristics of the
cohort students and of the institution (Xit), institutional fixed effects (ηi),
year-fixed effects (γt), and a random-error term (εit), that is, Git = α0 + α1Sit
+ α2Tit + α3Xit + ηi + γt + εit, where αk (k = 0, 1, 2, 3) are parameters to
be estimated. The characteristics of the students included in the model are
the average age of the entering cohort, the proportion of entering freshmen
who are from in state, the share of underrepresented minority students in
the entering class, the proportion of the entering freshmen who are full-time
students, the share of male students, and the midpoint of the 25th and 75th
percentile SAT scores of the entering class. State appropriation per FTE is
deflated by the Consumer Price Index to 2000 constant dollars.
In our empirical analyses, we use different models including between-
institution estimator, fixed-effects (within-institution estimator), and random-
effects models. We further estimate our preferred model (fixed-effects model
in this particular study) separately for institutions that have experienced an
increase or a decrease in state funding during the period for which we have
data. Finally, we allow the estimated coefficients to vary across different
types of institutions by estimating the model separately for each Carnegie
Classification.

Results

Table 2 presents results of our analyses. Model 1 reports the between-

institution estimator. Essentially, these are cross-sectional estimates, except
that variables in the model are averages over time; thus, the estimated coeffi-
cients should be interpreted across institutions. Turning first to the control
variables of cohort characteristics, we see that institutions with younger
freshman students have higher graduation rates, other factors included in
the model being held constant. On average, an institution with its entering
cohort 1 year younger than others would have a higher graduation rate at
2.7 percentage points higher. Other cohort characteristics can be interpreted
similarly. For example, institutions whose freshman cohort has higher
SAT scores, a higher proportion of nonresident students, a lower propor-
tion of minority students (not statistically significant), a higher proportion of

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 722
Table 2
Estimates for 6-Year Graduation Rates at 4-Year Public Higher Education Institutions (t statistics)
Model 1 Model 2 Model 3 Model 4

Log state appropriation per FTE 0.0695 (5.37) 0.0644 (5.37) 0.0880 (9.31) 0.0212 (1.73)
Log undergraduate tuition and fees 0.0740 (5.56) −0.0043 (−0.60) 0.0299 (4.23) −0.0158 (−2.26)
Average age of the entering cohort −0.0270 (−6.04) −0.0010 (−0.83) −0.0050 (−3.93) −0.0011 (−1.01)
Mean SAT scores of the entering cohort (100) 0.0854 (14.43) 0.0195 (12.21) 0.0247 (14.80) 0.0112 (3.76)
Proportion of resident students −0.1031 (−3.05) 0.0243 (1.53) −0.0382 (−2.39) −0.0025 (−0.16)
Proportion of minority students −0.0114 (−0.53) −0.0163 (−0.66) −0.0940 (−5.51) −0.0245 (−1.01)
Proportion of full-time students 0.3023 (8.03) 0.0601 (4.60) 0.0814 (6.09) 0.0185 (1.42)
Proportion of male students −0.3537 (−6.39) −0.1032 (−4.50) −0.1270 (−5.44) −0.0419 (−1.85)
Number of observations 1,781 1,781 1,781 1,781
R2 0.7668 0.9706 χ2 = 2664.32 0.9732

Note: FTE = full-time equivalent; Model 1 = “Between” estimator; Model 2 = “Within” (fixed-effects) estimator; Model 3 = Random-effects estimator;
Model 4 = Fixed-effects model with year dummies; SAT = scholastic assessment test.

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 723

full-time students, and a lower proportion of male students, have higher grad-
uation rates on average. Turning to the two financial variables, institutions
with greater state funding have higher graduation rates. On average, a 10%
difference in state funding per FTE is associated with a gap of 0.7 percentage
points in graduation rates, other factors being held constant. Not surprisingly,
institutions charging higher tuition and fees also have higher graduation rates.
On average, a 10% difference in resident tuition and fees is associated with a
difference of approximately 0.74 percentage points in graduation rates.
Between-institution estimates in the first column are important because
they explain the variation of graduation rates across institutions; however,
these results are less informative for policymaking if, let us say, an institution
wants to increase its graduation rates. For example, an increase in tuition and
fees in the hope of improving graduation rates would typically not be a good
strategy. To obtain within-institution estimates that give the impact of the
change in independent variables on the change in dependent variables, we
estimate fixed-effects models. The second column (Model 2) presents our
estimates of a fixed-effects panel data model for our sample as a whole.
Results in this column are quite different from those in Model 1. Turning first
to the control variables of cohort characteristics, we see that a decrease in the
average age of the entering cohort does not seem to increase graduation rates
significantly, holding constant other factors in the model including institu-
tional fixed effects. Taken together with the result in Model 1, it suggests that
the relationship between the average age of a cohort and its graduation rates
is primarily a cross-institution phenomenon. It could simply reflect the fact
that younger students are more likely to attend better institutions. From an
institutional point of view, enrolling younger students alone would not be
effective in improving graduation rates. Similarly, the proportion of resident
students becomes insignificant in the fixed-effects model.
The other three cohort characteristics, including the mean SAT scores, the
proportion of full-time students, and the proportion of male students, retain
their significant association with graduation rates in the fixed-effects model,
although the magnitude of their influence has been reduced greatly. For
example, Model 1 indicates that an institution whose entering freshman class
has mean SAT scores 100 points higher than the classes of other institutions
would have more than an 8.5 percentage points advantage in graduation rates.
By contrast, Model 2 suggests that an increase of 100 points in mean SAT
scores at a particular institution would result in approximately a 2% increase
in graduation rates for that institution. Similarly, an increase of full-time
students by 10 percentage points is associated with a 0.6 percentage point
increase in graduation rates. Because the graduation rate is measured only for

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

724 Educational Policy

full-time first-time students, the significant influence of the share of students

that are part-time suggests that more part-time students might create an aca-
demic environment that may adversely affects full-time students. Finally,
results indicate that an increase of male students by 10 percentage points is
related to a 1 percentage point decrease in graduation rates.
Although the effects of most variables have been reduced greatly in the
fixed-effects model, the impact of state appropriation per FTE remains
strong and significant. On average, a 10% increase in state funding per FTE
student at an institution is associated with a 0.64 percentage point increase
in graduation rates at that institution, a coefficient with similar magnitude
as in the between-institution estimates. These results indicate that the posi-
tive relationship between state funding and graduation rates not only exists
across institutions, but also holds for any particular institution on average,
suggesting that increasing state funding is a good strategy to improve grad-
uation rates at 4-year public institutions. By contrast, although there is a
positive association between tuition and graduation rates across institutions,
increasing tuition at a particular institution appears not to be a feasible strat-
egy for improving graduation rates. On average, an increase in tuition at an
institution would result in lower graduation rates, although the effect is not
statistically significant. These results suggest that in a cross-sectional
analysis, tuition is usually correlated with institutional selectivity, whereas
in the within-institution analysis, an increase in tuition probably means an
increase in financial burden for students.
Model 3 reports estimates of a random-effects model. The coefficients in
this column tend to fall between the “between” and “within” estimators.
Model specification test is conducted at this point to identify the most
appropriate model for our data. The Hausman specification test between
fixed- and random-effects models gives a chi-square statistic of 571.66
(significant at .001 level), indicating that the fixed-effects model yields con-
sistent estimates and thus is preferred. Furthermore, because the Hausman
test is asymptotically equivalent to test the equality between the between
estimator and within estimator, we also reject Model 1. Other test statistics
are also informative. The F test for the null hypothesis that all institutional
fixed effects are zero is rejected with an F statistic of 42.84. Finally, results
from the fixed-effects model yield a correlation of .5151 between the
predicted values based on explanatory variables and the institutional fixed
effects, suggesting a high positive correlation between the observed and
unobserved institutional factors. All these tests indicate that the fixed-effects
model is probably the most appropriate model for our data. This exercise
also suggests the cross-sectional analysis is likely to yield a biased estimator

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 725

for the relationship between graduation rates and explanatory variables in

the model because it ignores unobserved institutional characteristics that
are highly correlated with measured characteristics.
In Model 4, the institutional fixed-effects model is extended by adding
year-fixed effects to the model because it is possible that part of the relation-
ship uncovered in Model 2 could be due to the time trend in the graduation
rate. When our panel data models are estimated with time effects removed,
the estimated effect of state appropriation on graduate rate remains margin-
ally significant, but its magnitude is reduced to approximately one third of
that in Model 2. Choosing between Model 2 and 4 is not straightforward,
although a simple F test rejects the null hypothesis that there are no time-
fixed effects. The question is to what extent one is able to distinguish the
impact of explanatory variables on graduation rates and the time trend.
Including the time effects, however, might substantially underestimate the
real effects of explanatory variables on graduate rates because the time trend
could be due to changes in explanatory variables over time. Considering the
possibility of already downward biased estimates in fixed-effects model in
the presence of measurement error, we choose to have Model 2 (i.e., insti-
tutional fixed-effects model without time-fixed effects) as our empirical
model for subsequent analyses. Results based on time-fixed–effect model
are also reported as a comparison.
During the period when data in this analysis were collected, the average
state funding per FTE after adjusting for inflation has actually increased.
Hence, our results suggest that an increase in state funding per FTE is asso-
ciated with an increase in graduation rates, on average. The important ques-
tion is whether the relationship holds for both institutions that enjoyed an
increase and institutions that experienced a reduction in state funding. To
answer this question, we separate our whole sample into two groups based
on the relative average state funding during the first 4 years (financial years
1992, 1993, 1994, and 1995) and last 4 years (financial years 1996, 1997,
1998, and 1999). For the majority of institutions, the relative level of state
funding during last 4 years is greater than that during the first 4 years;
however, approximately 40% of institutions in our sample actually experi-
enced a relative reduction in state funding during the last 4 years compared
with the first 4 years.
Table 3 presents the results of institutional fixed-effects model for these
two groups of institutions. The first column is for institutions that enjoyed an
increase in state funding per FTE, and the second column is for those insti-
tutions that suffered a reduction. Estimated coefficients for state funding on
graduation rates are remarkably similar. For institutions that received a higher

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

726 Educational Policy

Table 3
Estimates for 6-Year Graduation Rates at 4-Year Public Higher
Education Institutions, by Levels of State Funding (t statistics)
Group A Group B

Log state appropriation per FTE 0.0750 (4.94) 0.0564 (2.74)

Log undergraduate tuition and fees −0.0182 (−1.86) 0.0095 (0.88)
Average age of the entering cohort −0.0004 (−0.34) −0.0050 (−1.66)
Mean SAT scores of entering cohort (100) 0.0166 (8.02) 0.0239 (9.21)
Proportion of resident students 0.0773 (2.71) 0.0021 (0.11)
Proportion of minority students −0.0298 (−0.95) 0.0203 (0.50)
Proportion of full-time students 0.0661 (4.33) 0.0435 (1.74)
Proportion of male Students −0.1191 (−4.42) −0.0653 (−1.49)
Number of observations 1,089 684
R2 0.9655 0.9709

Note: Group A shows results for institutions with lower state funding per full-time equivalent
student (FTE) during the first 4 years (1992-1995) than the last 4 years (1996-1999) of our
data period, and Group B shows results for institutions with greater state funding per FTE
during the first 4 years (1992-1995) than the last 4 years (1996-1999) of our data period.
SAT = scholastic assessment test.

level of state funding during a later period, a 10% increase in state funding
per FTE is associated with a .75 increase in graduation rate, on average. On
the other hand, for institutions that received a lower level of state funding dur-
ing the later period, a 10% reduction in state funding per FTE is associated
with a .56 decrease in graduation rates. These results suggest that the positive
relationship between state funding that an institution receives and the 6-year
graduation rate at that institution not only holds for institutions with an
increase in state funding, but also for institutions with a reduction in state
funding. We also estimate time-fixed–effect model by groups of institutions
based on funding change. Not surprisingly, the estimated coefficients for state
funding for these two groups of institutions are very similar to the pooled
model in Table 2, Column 4. Furthermore, due to the reduction in sample
size, neither of these two coefficients is statistically significant.
Finally, Table 4 reports disaggregated results by Carnegie Classification.
In this table, the institutional fixed-effects models are estimated using
separate samples of research/doctoral institutions, master’s institutions, and
liberal arts colleges. In general, the estimated coefficients for different
Carnegie types of institutions are similar to that in the aggregate model,
although in a few cases some coefficients change their directions and/or

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 727

Table 4
Estimates for 6-Year Graduation Rates at 4-Year Public Higher
Education Institutions, by Carnegie Classification
(t statistics in parentheses)
Research/
Doctoral Master’s Liberal Arts

Log state appropriation per FTE 0.1060 (5.18) 0.0561 (3.44) 0.0558 (1.39)
Log undergraduate tuition and fees 0.0092 (1.01) −0.0085 (−0.73) −0.0161 (−0.55)
Average age of the entering cohort 0.0023 (1.38) −0.0029 (−1.70) −0.0015 (−0.35)
Mean SAT scores of entering 0.0183 (7.79) 0.0167 (7.02) 0.0226 (3.97)
cohort (100)
Proportion of resident students 0.0204 (1.06) 0.0185 (0.74) −0.0208 (−0.23)
Proportion of minority students 0.0508 (1.42) −0.0215 (−0.62) −0.3306 (−2.90)
Proportion of full-time students 0.1457 (6.08) 0.0237 (1.28) 0.0752 (2.15)
Proportion of male students −0.2454 (−6.10) −0.0332 (−1.02) −0.1398 (−2.16)
Number of Observations 751 823 207
R2 0.9700 0.9530 −0.9505

See note to Table 1.

become insignificant probably due to smaller sample. The estimated coeffi-

cients for state funding also show some variation across different types of
institutions. For example, a 10% difference in state funding per FTE is asso-
ciated with approximately 1 percentage point difference in graduation rates
at research/doctoral institutions, whereas at master’s institutions, the esti-
mated effect is approximately half as large. Finally, the estimated effect for
state funding at liberal arts institutions is of similar magnitude to master’s
institutions, although it is not statistically significant. We also estimate
time-fixed–effect model by Carnegie Classifications. The estimated coeffi-
cients for state funding for all three groups of institutions are positive, but
only statistically significant for research/doctoral institutions.
A couple of empirical extensions are noteworthy here. In addition to
state governments, federal and local governments also fund public higher
education, especially special institutions such as tribal colleges. For the three
types of institutions included in this analysis (research/doctoral, master’s,
and liberal arts institutions), federal and local appropriations are inconse-
quential relative to state appropriations. For example, in 2000, federal and
local appropriations accounted for less than 2% of total government appro-
priations. When total public funding (appropriations from all levels of
governments) per FTE is used in the analysis, we obtain virtually identical

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

728 Educational Policy

results as when we use state funding per FTE. One might suggest that total
revenues per FTE be used because other sources of revenue might also con-
tribute to student success in college. When we do so, the estimated effect of
total revenues per FTE on graduation rates is approximately one third of the
magnitude as in the case of state funding per FTE. Not surprisingly, when
state appropriations are excluded from the total revenues in the empirical
model, the estimated effect is even lower. These results suggest that, although
our analysis does not reject the idea that other sources of revenues than state
funding also influence student graduation rates, the dominant factor
appears to be state appropriations, which is consistent with what resource
dependence would suggest.
Another topic of interest may be the dynamics between state funding
and tuition charged at public institutions. One hypothesis is that a decrease
in state funding could lead to an increase in tuition, which would further
depress graduation rates. However, for most of the years of our analysis, the
absolute dollar value of state funding per FTE has not decreased on aver-
age; as a result, a negative association between state funding and tuition is
not detected. Alternatively, one could show the relationship between the
relative changes in both variables by removing their time trends. When we
do so, a negative correlation between these two variables emerges; that is,
when state funding per FTE increases slowly (or decreases), tuition charged
by institutions increases fast. By contrast, a relatively fast increase in state
funding is associated with a relatively slow increase of tuition.

Limitations, Conclusions, and Discussion

Our analysis has its limitations. Before getting into the main findings of
our analysis, we offer three limitations for consideration. First, our empiri-
cal model is based on the assumption that many unmeasured institutional
heterogeneities (e.g., demographic characteristics, institutional selectivity,
and other contextual and historical factors) are stable over time (or at least
over the time period for which we have data). Consequently, the results are
subject to omitted variable bias when factors that change over time and that
which are correlated with the state funding variable are not considered in
the empirical model. For example, changes in financial aid policies within
institutions over time might also affect graduation rates. Nonetheless, the
panel data method used in this analysis alleviated the omitted variable bias
significantly when compared with cross-sectional models.
Second, the institutional-level aggregated model might capture a much
more complex relationship among various institutional factors and student

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 729

outcomes. In this analysis, we assume that the relevant financial variables

were those during the first 4 years that a particular cohort was enrolled in
college. This assumption is certainly subject to close scrutiny. Furthermore,
we ignore possible internal resource allocation and program restructuring.
For example, given the variation of graduation rates across fields, there is a
possibility that institutions might have restructured their programs in a time
of financial instability, which could result in changes in graduation rates.
Finally, we recognize the estimated effects of state funding on graduation
rates are small. We feel that these estimated effects are conservative for
several reasons. First, it is well-known that fixed-effects models often yield
downward biased estimates in the presence of measurement errors. Second,
the estimated effects should be viewed as the “total” effect because we ignored
institutional reactions to the changes in state funding. For example, revenue
substitution might occur between state funding and tuition, between instruc-
tion and research, and between short-term and long-term programs (e.g., insti-
tutions may defer facility maintenance due to temporary fiscal difficulty).
Consequently, the results of our analysis should be interpreted in the
context of aforementioned limitations. This study uses panel data to examine
the direct link between state funding and graduation rates at 4-year public
institutions. We find some evidence for a positive association between state
funding and college graduation rates. When other factors are held constant,
a 10% increase in state appropriations per FTE student at 4-year public
institutions is associated with approximately a 0.64 percentage point
increase in graduation rates. This positive link appears to hold for both
institutions that have enjoyed an increase in state funding and those that
have experienced a reduction. In addition, the positive association seems
to hold for all research/doctoral, master’s, and baccalaureate institutions.
These estimated effects are smaller and in general statistically insignificant,
when both institution- and time-fixed effects are controlled.
Our results are largely consistent with recent studies on similar topics
(e.g., Blose et al., 2006; Ryan, 2004). Although these recent studies showed
the positive link between instructional expenditures and graduation rates,
our study makes it clear that it is mainly the state appropriations that have
a positive impact on graduation rates. These results are consistent with the
perspective of resource dependence, which holds that internal organiza-
tional activities are influenced primarily by the actions of external resource
providers. A decline in state appropriation, when other factors are held con-
stant, would most likely lead to a reduction in instructional expenditures. In
other words, it is unlikely that public institutions can fully compensate for the
reduction in state appropriation through internal resource reallocation.

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

730 Educational Policy

Channels through which state funding might affect student success at

colleges and universities need to be delineated. Individual-level analysis
would be helpful in identifying the intermediate factors through which finan-
cial resources exert influences. For example, given that limited financial
resources might force institutions to employ more inexpensive faculty such
as part-time faculty to teach undergraduate classes, what is the implication
with regard to the quantity and quality of student–faculty interaction, which
has consistently been shown to affect students’ persistence and retention
significantly? Another example, given that public institutions have to increase
their tuition and fees to offset the relative reduction in state funding and
increase in production costs, what is the implication with regard to students’
college attendance pattern and financial burden, both of which have been
shown to affect the probability of graduation and time to degree? Establishing
these linkages is essential for us to understand to what extent and in what
ways financial factors might influence students’ success in college, because
this comprehension is the basis for informed public policies.

References
Astin, A. W. (1993, September 22). College retention rates are often misleading. The Chronicle
of Higher Education, p. A48.
Barnett, R., & Middlehurst, R. (1993). The lost profession. Higher Education in Europe,
18, 110-128.
Berger, J. B., & Milem, J. F. (1999). The role of student involvement and perceptions of
integration in a causal model of student persistence. Research in Higher Education, 40,
641-664.
Blose, G. L., Porter, J. D., & Kokkelenberg, E. C. (2006). The effect of institutional funding
cuts on baccalaureate graduation rates in public higher education. In R. G. Ehrenberg (Ed.),
What’s happening to public higher education? (pp. 71-82). New York: Praeger.
Bound, J., & Turner, S. (2004). Cohort crowding: How resources affect collegiate attainment
(Population Studies Center Research Report No. 04-557). Ann Arbor: University of
Michigan.
Braxton, J. M., & Hirschy, A. S. (2005). Theoretical developments in the study of college
student departure. In A. Seidman (Ed.), College student retention: Formula for student
success (vol. 3, pp. 61-87). New York: Praeger.
Ehrenberg, R. G., & Zhang, L. (2005a). The changing nature of faculty employment. In R. Clark
& J. Ma (Eds.), Recruitment, retention, and retirement in higher education: Building and
managing the faculty of the future. Northampton, MA: Edward Elgar.
Ehrenberg, R. G., & Zhang, L. (2005b). Do tenure and tenure track faculty matter? Journal of
Human Resources, 40, 647-659.
Elkins, S. A., Branxton, J. B., & James, G. W. (2000). Tinto’s separation stage and its
influence on first-semester college student persistence. Research in Higher Education, 41,
251-268.

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015

 Zhang / Does State Funding Affect Graduation Rates? 731

Hasbrouck, N. S. (1997). Implications of the changing funding base of public universities.

Unpublished doctoral dissertation, The University of Arizona, Tucson.
Hovey, H. A. (1999). State spending for higher education in the next decade: The battle to sustain
current support. San Jose, CA: National Center for Public Policy and Higher Education.
Kelly, P. J., & Jones, D. P. (2005). A new look at the institutional component of higher educa-
tion finance: A guide for evaluating performance relative to financial resources. Boulder,
CO: National Center for Higher Education Management Systems.
National Center for Education Statistics. (2005). Digest of education statistics, 2005. Washington,
DC: Author.
Pfeffer, J., & Salancik, G. R. (1978). The external control of organizations: A resource depen-
dence perspective. New York: Harper & Row.
Rizzo, M. J. (2006). State preferences for higher education spending: A panel data analysis,
1977-2001. In Ehrenberg, R. G. (Ed.), What’s happening to public higher education?
(pp. 3-36). New York: Praeger.
Ryan, J. F. (2004). The relationship between institutional expenditures and degree attainment
at baccalaureate colleges. Research in Higher Education, 45, 97-114.
Scott, M., Bailey, T., & Kienzl, G. (2006). Relative success? Determinants of college gradu-
ation rates in public and private colleges in the U.S. Research in Higher Education, 47,
249-279.
Skolnick, M. L. (1986). If the cut is so deep, where is the blood? Problems in research in the
effects of financial restraint. The Review of Higher Education, 9, 435-455.
Slaughter, S., & Leslie, L. L. (1997). Academic capitalism: Politics, policies, and the entrepre-
neurial university. Baltimore: Johns Hopkins University Press.
Tinto, V. (1993). Leaving college: Rethinking the causes and cures of student attrition
(2nd ed.). Chicago: University of Chicago Press.
Toutkoushian, R. K. (2001). Trends in revenues and expenditures for public and private higher
education. In M. B. Paulsen & J. C. Smart (Eds.), The finance of higher education: Theory,
research, policy, and practice (pp. 11-38). New York: Agathon Press.

Liang Zhang is an assistant professor of public policy and higher education in the Department
of Leadership, Policy, and Organizations at Vanderbilt University, Nashville, TN.

Downloaded from epx.sagepub.com at University of Manchester Library on May 11, 2015