SRDC
Series #196
School Choice Policy: Estimates
of Supply and Demand Response in Private Education
by
Warren Kriesel, Associate Professor
Andrew G. Keeler, Assistant Professor
Marc White, Graduate Research Assistant
Agricultural and Applied Economics Department
University of Georgia
Athens, Georgia
View PDF Version
Executive Summary
Policies that would facilitate households' switching to private
schools have been proposed in response to a perceived crisis in
public education. Proponents argue that vouchers or tuition tax
credits would encourage competition that would serve as a catalyst
in improving public education. However, necessary conditions for
effective competition are (a) that private schools have sufficient
excess capacity to absorb new students and (b) that students exit
public schools in adequate numbers to cause teacher layoffs and
school closure. Research results for the state of Georgia indicate
that private schools' current excess capacity is about 4,650 vacancies.
A demand model indicates that a $1,000 income tax credit would
encourage nearly 2,000 students to switch to private schools.
In other words, the proportion of school-aged children in private
schools would increase from 4.4 percent to 4.58, for an increase
of 0.18 percent. Thus, it is concluded that while the private
schools can take on additional students, the demand response by
households will probably fall far short of fostering effective
competition between private and public schools. Finally, the demand
model indicates that parents are sensitive to indicators of public
school quality in making their schooling decisions. This means
that public school administrators can decrease the loss of pupils
by improving their school's performance.
Introduction
The United States spends a greater share of its gross national
product (7.5 percent) on education than any other country except
Israel (Wallis). Despite this fact, students in the U.S. are outperformed
in math and science by students in more than 10 nations. From
1950 to 1990 the proportion of school spending that went to classroom
instruction decreased from two-thirds to less than one-half. Meanwhile,
administrative outlays have doubled from 4 to 8 percent (Wallis).
Many believe that statistics such as these illustrate that an
educational crisis exists and that the US is falling behind the
rest of the industrialized world in both academic achievement
and labor productivity.
Frustration with the quality of American schools has increased
since the 1983 release of A Nation at Risk. That report,
produced by the National Commission on Educational Excellence,
emphasized the plight of the educational system. Education reform
became a top priority for many states and spurred such actions
as increasing graduation requirements, lengthening the school
year and day, requiring more science and math for graduation,
and mandating more testing of students (Uchitelle). Many observers
view school choice as the catalyst that will bring change and
reform needed to improve education. Proponents argue that school
choice would become a powerful force for improving public schools.
For this reason, a voucher or tuition tax credit that would give
students more access to private schools has been a popular policy
proposal.
The effect of any tax credit is to give a dollar-for-dollar
tax reduction that is linked to any approved activity. In the
case of education, a tax credit for tuition effectively reduces
the price of private schooling by a dollar-for-dollar amount.
The amount of the credit can vary, depending on the policy proposed.
A tuition voucher may be used to purchase educational services
from a school of the family's choosing. This policy would allow
parents to use tax dollars, in the form of vouchers, to pay for
tuition at private schools. If the amount of the voucher is below
the total cost of tuition, the parents would make up the difference.
Proponents believe that an effective voucher or tax credit policy
will inject competition into the market for educational services
that will increase pressure for all schools to improve in response
to new market forces. School choice would enable parents and students
to choose the school they believe best fits their needs. Schools
would be placed in the position of competing for student enrollments
in the open market through the quality of education they provide,
or face the prospect of losing students and essentially be forced
to close their doors.
Opponents of school choice policy question how many private
schools would actually participate in a choice program. They argue
that many private schools are already filled to capacity and they
would not have the additional space and staff needed to accommodate
any enrollment increases. If this were the case, then the use
of public monies to support private schools would amount to a
subsidy to families that are currently in private school. Others
argue that private schools would take the "best and brightest"
public school students and not be receptive to public school children
who are more academically challenged. This would result, they
argue, in public schools that may suffer further deficiencies
if parents switch better students to private schools. Also, the
loss of pupils to private schools may reduce the average voter's
commitment to high quality public education. Perhaps one of the
most controversial questions is whether encouraging private school
enrollments would increase educational diversity or would instead
possibly increase segregation along racial, wealth, and religious
lines (Sandy). Further, higher income families would be better
able than low-income families to make choices where additional
tuition and transportation necessities are required. In summary,
opponents do not believe that market competition alone will increase
student achievement nor accomplish necessary school reform.
Central to the school choice policy debate is the question of
how many pupils would switch to private school. If an empirical
analysis indicated that large numbers of households would leave
public schools (and if private schools could absorb them), then
a school choice policy would have the potential to be an effective
instrument in improving education. This outcome could also have
potentially devastating fiscal effects on public schools. On the
other hand, if few households responded, then a school choice
policy would not be effective, simply because public schools would
still face little competition, and the policy may have the side
effect of acting as a subsidy to households that currently use
private schools.
Given these research questions, the first objective of this
research is to estimate empirically how many additional pupils
private schools in Georgia can absorb. The second objective is
to estimate a demand model of private education and perform simulations
of how many households would leave the public schools. These objectives
are met by analyzing primary survey data from private schools
as well Census tract-level data for households and additional
data that describe public school conditions in Georgia.
The results indicate that private schools' current excess capacity
is about 4,650 vacancies. The demand model indicates that a $1,000
income tax credit would encourage nearly 2,000 students to switch
to private schools. In other words, the proportion of school-aged
children in private schools would increase from 4.4 percent to
4.58, for an increase of 0.18 percent. Thus, it is concluded that
while the private schools can take on additional students, the
demand response by households will probably fall far short of
fostering the kind of competition between private and public schools
that might lead to improvements in education. Finally, the demand
model indicates that parents are sensitive to indicators of public
school quality in making their schooling decisions. This means
that public school administrators can decrease the loss of pupils
by improving their school's performance.
Review of Previous Research
Much of the theoretical research on this topic has focused on
the ways that taxpayer/voter preferences about school finance
map into choices about educational expenditure and quality. Expenditure
on education has been used as a proxy for school quality (Stiglitz);
he hypothesized that private schools are better than public schools
because per-pupil expenditure is higher. Flowers focused on the
relationship of private and public school qualitylowering the
cost of private education decreases the number of children in
the public schools but also decreases the willingness of the average
resident to fund public education. She found that while a tax
credit would tend to increase enrollment in private schools, it
could conceivably increase the quality of public education. This
would happen if the marginal cost of educational quality increased
with the number of students enrolled while the tax credit decreased
enrollment. Although higher expenditures cause higher educational
quality in Flowers' work, she recognized that private and public
education had different cost functions.
Given this emphasis on the relationship of cost differences
and school choice in the literature, the actual behavior of parents
when confronted with differing circumstances is an essential input
for public debate. There have been five empirical studies on this
topic. Frey (1983) performed a national analysis, with states
as the units of observation. Using six independent variables,
he found that all variables had their hypothesized signs and estimated
the price elasticity of demand to be between 0.4 and 2.1. Frey's
work is based on statewide data with the explanatory variables
associated with private schools averaged to obtain state-level
data; thus, his work is unavoidably limited to highly aggregated
variables and thus loses some of the flavor of local choice. West
and Palsson also estimated a model with state-level variables.
Using eight independent variables, they estimated the price elasticity
to be from 1.5 to 3.0.
Hamilton and Macauley used school district-level data from New
Jersey and applied the log transformation to the dependent variable,
so it is not bounded at zero. They did not use a price variable
in their study because they selected homogeneous school districts
where the variance of private tuition was very low. They focussed
special attention on the standard deviation of household income
because, they argued, the variable is a proxy for the effect of
student peer groups on achievement. Lankford and Wyckoff applied
a logit analysis to a unique data set for 28,000 individual students
in New York State. The decision to attend a religious school was
modeled as a function of income and tuition, plus five school
characteristics, five environmental factors, and five demographic
indicators.
A study by Keeler and Kriesel is similar to Hamilton and Macauley's,
but their data set represents 105 school districts in rural Georgia.
Their demand model explains the proportion of children in private
schools as a function of tuition (adjusted for travel costs) income,
percent minority population, and three variables that describe
the district's public schools: test scores, pupil-teacher ratio
and per-pupil expenditures. Their regression results imply a demand
elasticity of -1.07. However, the interpretation of this elasticity
(as well as those from the other studies) is awkward because the
dependent variable is a proportion. Therefore, they emphasize
the importance of running a simulation on effect of new school
choice policies. In particular, they found that a $1,000 tuition
tax credit would increase the proportion of children in private
school from 5.5 percent to 6.82 percent for an increase of 1.32
percent.
This research extends the work performed by Keeler and Kriesel
by disaggregating the data from the school district level to the
census tract/BNA level. In doing so it will more closely approach
decision-making at the individual household level. Also, Frey
(1991) and Martinello and West introduce two important considerations
for this analysis. First, in situations where the supply of private
education is not perfectly elastic a tuition tax credit might
be expected to (a) increase the tuition charged by private schools
or (b) result in more queuing by prospective students. If supply
is not perfectly elastic then the results of a single equation
demand analysis may be affected by simultaneity bias. Therefore,
this study reports the results of a survey quantifies the supply
response by private schools to changes in school choice policy.
A second consideration is that a tax credit of the magnitude
offered in policy proposals might be of such large magnitude it
might cause out-of-sample changes in private school enrollment.
This study avoids the problem by applying a micro-level analysis
that has substantial variation in both enrollment and in the explanatory
variables, providing a wider range of latitude for policy simulation
before running into out-of-sample problems.
Private School Survey
We are concerned with determining how school choice policies
may encourage more parents to send their children to private school
or, alternatively, what enrollment declines the public schools
may expect. Previous studies have estimated a demand response
by parents, but in doing so they explicitly assumed that the supply
of private education is perfectly elastic. Perfect elasticity
means that, subsequent to a school choice policy, private schools
could expand their enrollment without increasing tuition. If supply
is not perfectly elastic, then in order to take more students
private schools would either have to raise tuition or put prospective
students on a waiting list. Either of these two outcomes would
mean that previous studies have overestimated the number of students
who would leave the public school system.
The condition of perfect elasticity could be met if it can be
demonstrated that private schools are currently operating below
their maximum student capacity. Furthermore, they should be sufficiently
below capacity so they could accept all of the additional students
that a school choice policy would encourage to leave the public
school system. This research makes this determination by comparing
private school excess capacity estimated from the survey results
(reported below) with the number of additional private school
pupils estimated from the demand analysis that is reported in
a subsequent section. Our conclusions are reported in the final
section.
Survey Results
The private school survey was conducted through an interview
process targeting the entire state of Georgia. For a sampling
frame, a list containing all private schools in Georgia was obtained
from the Georgia Department of Education that detailed the location
of schools by county and provided enrollment figures. From that
list the state was divided into six target areas: Albany, Savannah,
Macon, Augusta, Athens, and Atlanta. The number of schools to
be used for the survey was taken from a percentage of the total
number of private schools located in that area. To obtain an accurate
representation of each target area, they were further divided
between metropolitan and rural schools. The schools in each target
area were notified by telephone to schedule a time and date for
an interview.
Forty-five schools were eventually interviewed. Analysis reveals
that the average total enrollment for all schools was 519 students
with a standard deviation of 302 (Table 1). Average school capacity
was approximately 93.4 percent. Average high school enrollment
was 199 students and the average high school capacity was approximately
90.5 percent. The average SAT score of the previous year's graduating
class was 1015. Out of the 45 schools surveyed, two schools located
in Atlanta did not report SAT scores. One of the schools stated
it did not release SAT information for fear that they are used
for the wrong purpose. The other school was about to graduate
its first senior class and did not yet possess any SAT information.
Two variables more difficult to quantify were the acceptance
rate and the student/teacher ratio. The accuracy of the acceptance
rate is questioned because of reporting in consistencies by the
schools surveyed. The smaller, rural schools do not keep records
of acceptance rates. In addition, many schools will discourage
or encourage a student to apply based on either the initial phone
conversation or interview. Many larger schools have very detailed
numbers on inquiries, applications, and on the number of students
admitted, for accreditation purposes, but they still acknowledge
acceptance rates may suffer from inaccurate reporting. Because
of these differences in accuracy these results may be biased.
According to this analysis, the average acceptance rate was 59.4
percent with a 33 percent standard deviation. Fifteen schools
(33 percent of those surveyed) reported an acceptance rate of
90 percent or above, and 15 schools reported an acceptance rate
below 50 percent.
To estimate the supply response by private schools to a school
choice policy, administrators were asked to quantify what effect
a tuition-tax credit would have on the tuition they charged and
on the number of new students admitted. For many administrators,
these questions were quite abstract and difficult to answer. They
responded that many factors affect such decisions and all would
need to be considered with such a question. Not only do physical
and economic limitations exist, but also, limitations placed by
school boards, school philosophies, teachers, parents, and students.
Of the 45 schools in the survey sample, only three administrators
said that they would increase their tuition in response to a tuition
tax credit. Interestingly, one replied that if a $1,000 tax credit
were introduced then he would increase tuition by $1,000, but
this was the only evidence of monopolistic pricing. The vast majority
replied that their tuition increases would keep pace with expected
inflation only.
The administrators' ability to estimate the number of new students
admitted was quite limited. They were very uncertain about their
future admissions because significant increases would entail new
construction programs, and these decisions were out of their hands.
Therefore, to estimate the schools' supply response, their estimates
of current excess capacity is relied upon. As reported in Table
1, the schools are currently at 93.5 percent of full capacity.
From the sampling frame, the 465 private schools in Georgia enroll
an average of 150 students. This implies that the average school's
full capacity is 150/0.935=160 students, or that it can take on
10 additional students. Extrapolating this to all private schools
implies there are currently 4,650 vacancies that could be filled
by students who may leave public schools following implementation
of a school choice policy. This figure of 4,650 vacancies is compared
with an estimate of increased private school demand presented
in the next section.
Private Education Demand Model
The purpose of this section is to analyze how economic incentives,
school choice policies in particular, affect the demand for private
education. Parents who select a private school for their children
must pay the required tuition plus continue to pay the property
taxes also provide for public school operation. In essence, they
are foregoing the subsidy already available to them to send their
children to public schools. The desire for a private school education
by a household is influenced by the school's attributes, by financial
considerations, and by the tastes and preferences of individual
families (Lankford and Wycoff). A wide range of private schools
exist to serve a wide variety of needs and interests. For example,
parochial schools are sponsored by local churches, college preparatory
academies stress academic excellence, and Montessori schools apply
their own philosophy of education.
By selecting a private school, the family will have less income
to allocate to other goods and services. Parents who are willing
to incur the cost of a private school education, presumably, believe
the attributes that characterize the quality of education in the
private school exceed the attributes provided for by the local
public school. The decision between public and private schools,
therefore, involves a tradeoff between the attributes that characterize
the quality of education received by the family and the family's
consumption of other goods.
The household's decision to switch to private education can
be modeled within a random utility framework (Keeler and Kriesel).
The utility of the ith household if private education is selected
(subscripted by v) is given by Uiv (qv, cv, v) where qv is a vector
consisting of private school characteristics, CV is a composite
of all other goods consumed, and v is an error term. The utility
associated with public education (subscripted by p) for the ith
household is given by Uip (qp, cp, p ). Parents evaluate whether
the characteristics associated with private education (qv and
CV) provide higher utility than the relevant public school alternative.
The probability of any individual's attending a particular educational
alternative (for example, a private school) is the expected value
of a random variable Pi that takes on the value of 1 if Uiv >
Uip and 0 if Uiv < Uip. The presence of the error term dictates
that Uiv and Uip are random variables. The dependent variable
in this demand analysis is the proportion of children enrolled
in private school, which is given by j = (Pvj)/Nj , where N is
the number of school age children and the subscript j denotes
the census unit of observation. This random utility specification
has been found to be generally appropriate for explaining individual
school choice decisions.
The dependent variable in the following regression is the proportion
of children who attend a private school. This variable is obtained
from the 1990 Census estimate of children aged three and over
who attended a private elementary or high school and dividing
that number by the total number of school aged children in the
census tract.
There are 2,733 individual census tract/BNA's located within
the state of Georgia. Out of this number approximately 299 are
invalid observations because they do not report any population.
Therefore, the demand model contains 2,434 observations. The average
tract/BNA had a mean private school attendance for school aged
children of 4.4 percent with a standard deviation of 11.88 percent
(see Table 2).
Selection of the variables in q, the vector of school
attributes, depends on the context of the study and the available
data. The variables and their summary statistics are listed in
Table 2. School quality is obviously an important characteristic
higher quality public schools or lower quality private schools
can be expected to decrease Pi. In this study we restrict
ourselves to the former because of data limitations. As noted
by Monk, there is no single best measure of school quality. One
measure is students' test scores; the hypothesis is that better
scores reflect higher school quality. Test scores can also reflect
a student's household characteristics and other influences as
well as school quality; nonetheless it remains a favored explanatory
variable of educational researchers. The school district's average
score for the tenth-grade criterion reference test of reading
ability is used to measure the educational quality of public schools.
We expect that higher test scores will make public schools more
attractive and thus decrease private school enrollment.
Expenditure on education has also been used as a proxy for school
quality; the hypothesis is that higher expenditures result in
better education. Although Monk has pointed out the difficulties
of equating expenditure with quality, many studies have used expenditure
measures. Therefore, we use the school district's total per-pupil
expenditures. The student/teacher ratio is another commonly accepted
indicator of quality; lower ratios might be expected to indicate
more individual attention and thus better education.
Recent literature has proposed that low teacher pay has resulted
in a decline in the number of young people seriously considering
teaching as a career alternative. Therefore, this study incorporates
the average teacher salary as a proxy for school quality. This
variable is expected to have a negative effect on the demand for
private education. Higher teacher salaries are hypothesized to
attract higher quality teachers and result in higher quality public
schools. The data for these four public school district variables
were for 1990 and were obtained from the Georgia Department of
Education (see Table 2).
Another characteristic we examine is the percentage of African-American
school aged population by school district. In much of the South,
including rural areas, private education boomed in the 1960's
as the integration of public schools mandated by Brown versus
Topeka Board of Education was implemented. We include the racial
composition variable to investigate the effects of racial composition
on school choice. The final variable that describes school district
conditions is the percentage of adults, older than 25 years, without
a high school diploma. Lankford and Wycoff state that past studies
on school choice have consistently shown poorly educated parents
tend to have lower expectations of school quality for their own
children, and therefore they are less likely to become dissatisfied
with the public schools. Therefore, this variable is hypothesized
to negatively impact private school attendance.
The variables that indirectly determine the household's non-public
education consumption are family income and the cost of private
education. Higher levels of income can be expected to increase
the probability of sending a child to private school, and this
variable was quantified as the Census tract's average household
income. On the other hand, higher private school tuition charges
should decrease the probability of sending a child to private
school. Since the financing of public education is separate from
individual choices to send children to public or private schools,
the individual's marginal cost of sending a child to public school
is zero and the cost of public schooling can be ignored.
Each tract/BNA is assigned an adjusted tuition variable that
takes into account the tuition of surrounding private schools
and associated travel costs. Information was obtained through
a telephone survey that produced a data set containing the price
variable of interest, tuition. A total of 128 private schools
were telephoned. Additional tuition information was also obtained
for Chattanooga, Tallahassee, and Jacksonville as these are metropolitan
areas located near the state borders that also influence the decision
of where one may choose to attend a private school. Therefore,
the tuition sample contains 131 observations. The information
compiled from the survey produced a data set that included tuition,
student enrollment, and the school's zip code. The average tuition
in the data set was approximately $3,287 with a standard deviation
of $1,904. The average enrollment of those schools surveyed was
317 with a standard deviation of 318.
Two location variables obtained from the 1990 census were the
latitudinal and longitudinal internal points of each tract/BNA.
This information facilitated the measurement of a distance variable
between the internal point of each tract/BNA to the location of
the private school. A separate data set that contained the latitudinal
and longitudinal internal points for each zip code in the state
allowed a program to be executed which then assigned that point
measurement for each private school's zip code. The distance measured
between each tract/BNA and private school was calculated by applying
the Pythagorean theorem.
The adjusted tuition variable was estimated by determining which
private schools are located within a 40-mile radius of each tract/BNA.
The travel cost is obtained by multiplying the distance between
the tract and private school by 35 cents per mile, times 180 school
days, times 2 trips per day. Tracts that had several schools located
within a 40-mile radius were assigned an adjusted tuition variable
equal to the weighted average travel-cost adjusted tuition of
all the schools. Tracts that were located more than 40 miles from
the nearest school were simply assigned the closest, least-expensive
private school plus travel cost. Therefore, the adjusted tuition
variable is the weighted average tuition plus the average travel
cost. The average adjusted tuition was found to be $6,833 with
a standard deviation of $1,587.
Statistical Methods
For this regression model, the dependent variable is a proportion
bounded between 0 and 1. If ordinary least squares is applied,
the model will yield inefficient estimates due to problems of
heteroskedasticity that are caused by the bounding (Gujarati).
Therefore, to correct for heteroskedasticity, a log transformation
is applied to the dependent variable. This transformation causes
the dependent variable to be bounded between zero and negative
infinity.
Further examination of the dependent variable reveals that out
of 2,434 total observations available, 653 observations, nearly
27 percent, do not report any children that attend private school.
This is referred to as censoring of the dependent variable, and
the disturbance term's distribution is a mixture of discrete and
continuous parts. Conventional regression methods fail to account
for the qualitative difference that exist between the limit (zero)
and nonlimit (continuous) observations. The total probability
is still one, as required, but instead of scaling the continuous
observations, the censored region is simply assigned the full
probability at the censoring point, which in this case is zero
(Greene). Therefore, a tobit log-linear model is appropriate for
this research.
The results of the tobit log-linear model are presented in Table
3 (Model 1). All explanatory variables had their expected signs
except expenditures per pupil. The coefficient for per capita
income is positive as expected and significant at the one percent
level. This supports the hypothesis that wealthier populations
are more likely to send a higher percentage of their children
to private schools. The racial composition variable was also positive
and significant at the one percent level. As racial heterogeneity
increases so does the demand for private education. The educational
attainment variable was negative and significant at the one percent
level. This supports the hypothesis that as the percentage of
adults who have less than a twelfth grade education increases,
the demand for private education decreases.
Regarding the proxy variables for public school quality, the
student/teacher ratio was positive as expected, and the student
test score variable was negative as hypothesized. However, neither
of these explanatory variables was statistically significant.
The variable that measured the average teacher salary for public
school districts was negative as expected, and it was the only
variable that proxied public school quality, which was statistically
significant. Average teacher salaries are significant at the one
percent level and support the hypothesis that as salaries for
public school teachers increase, private school enrollments decrease.
This indicates higher quality teachers are associated with higher
quality public schools. The adjusted tuition variable was negative
as hypothesized and significant at the 5 percent level. This supports
the claim that higher tuition results in decreased enrollments.
The prevalence of insignificant coefficients with wrong signs
raises the possibility that model 1 may be affected by damaging
collinearity. Collinearity occurs when two or more explanatory
variables are linearly related. An examination between expenditures
per pupil and the average teacher salary reveals that they have
a correlation coefficient of 0.81. A correlation coefficient above
0.70 is evidence that collinearity may exist among the variables.
Expenditures are also negatively related with student test scores
and the student/teacher ratio.
Since expenditures is a questionable proxy for school quality,
model 2 is the demand model without the expenditures variable.
The magnitude of the beta coefficients is relatively identical
to Model 1; however, the standard errors are lower in Model 2.
The following variables were positive and significant: per capita
income and the percentage of African-American school aged children.
The following variables were negative and significant: percent
educational attainment, tuition, and the average teacher salary.
As in Model 1, the variables for student test scores and the student/teacher
ratio were negative and positive as hypothesized respectively;
however, neither was statistically significant. An examination
of possible multicollinearity that may exist among the variables
revealed the highest correlation existed between income and the
educational attainment variable of 0.61. Therefore, these results
are probably not affected by damaging collinearity.
The price elasticity describes how changes in prices for educational
services will affect the household's decision of attending a private
school. Whereas past studies estimated price elasticities of anywhere
between -0.5 to -3.3, Keeler and Kriesel's price elasticity, with
Georgia school districts as the unit of observation, was on the
lower end of the scale at 1.07. The following formula approximates
demand elasticity:
Ed = (Y/ X ) * (X / Y)
where Y and X refers to the change in Y divided by the change
in X multiplied by the mean value of X divided by the mean value
of Y. The change in Y divided by the change in X refers to the
slope coefficient for the tuition variable which is multiplied
by the mean adjusted tuition to obtain an elasticity of demand
approximation of 0.18, quite a bit lower than previous studies.
This elasticity is interpreted as follows: a one percent increase
in tuition will decrease the proportion of children in private
schools by 0.18 percent. The low magnitude of this elasticity
indicates that the demand for private education is relatively
unresponsive to price.
Of these existing studies, only Lankford and Wyckoff and Keeler
and Kriesel use the estimated model to simulate the effects of
public interventions that change the effective price of private
education. Lankford and Wyckoff present an experiment that predicts
the change in private school enrollment if tuition were effectively
reduced to zero. They find that such a policy would double the
proportion in their sample from 12 percent to 25 percent. Keeler
and Kriesel examine the effect of more modest public subsidies
to private education. They calculate that a $1,000 tuition tax
credit would cause demand for private education in Georgia to
rise by 1.32 percent of the total number of children. According
to the 1990 Census, there are about 1,040,000 school aged children
in Georgia and 1.32 percent of 1,040,000 represents 13,728 more
private school students.
The estimated parameters of model 2 are used to simulate the
effect of vouchers or tax credits for private education. If people
view these policies as a dollar-for-dollar private school tuition
reduction, then a credit's effect can be projected by substituting
a new tuition value reflecting the subsidy and then calculating
the resulting . The average private school tuition in this sample
is $6,833 per year. A $500 tuition tax credit (or voucher) would
increase the proportion of children in private school education
by 0.092 percent of the total student population of 1,040,000,
or 957 children.
A $1,000 credit would cause demand for private education to
rise by 0.19 percent of the total enrollment, or 1,976 new private
school students. A $2,000 credit would increase demand by 0.38
percent of the total student population, or 3,952 new private
school students. A $2,600 voucher the amount contained in California's
Proposition 174, which was narrowly defeated in November, 1993
would increase private school enrollment by 0.499 percent of the
total student population, or 5,190 new private school students.
Finally, attention is turned to the question that was raised
in the last section: If a school choice policy were implemented,
would the current excess supply of private school places be sufficient
to absorb the estimated number of students who would leave the
public school system? In the last section it was calculated that
private schools' current excess capacity equals 4,650 vacancies.
Comparing this estimate of vacancies with this study's projections
of new private school students indicates that the effective tuition
reduction would have to be quite large, above $2,000, before there
would be a shortage of places for the new students. This is an
important finding because it means that this school choice policy
would have its desired effect without causing higher tuition and/or
longer waiting lists for school admission. Thus, it is concluded
that while the private schools can take on additional students,
the demand response by households (i.e., only 2,000 to 5,000 students
statewide) will probably fall far short of fostering the kind
of competition between private and public schools that might lead
to improvements in education.
We recognize the possibility that this study's estimates of
new private school pupils may be underestimated. Again, Keeler
and Kriesel's results for Georgia, using a very similar model
but with school districts as the unit of observation, indicate
that the $1,000 tuition tax credit would result in 13,728 new
private school pupils. It is felt that additional research into
the construction of the tuition variable should be performed before
more definite answers can be obtained.
Summary and Conclusions
Many fear that an educational crisis exists in U.S.public schools
as evidenced by falling test scores, lack of discipline, violence,
and other indications that the public educational system is not
delivering the level of service parents and pupils demand. Policies
that would make it easier for households to switch to private
schooling have been advocated as a way to inject competition into
the market for educational services. Proponents of school choice
argue that if households could freely choose which school to patronize,
unattractive schools would either improve themselves or lose students
and face the prospect of closing their doors. However, a precondition
for competition is that households must respond to a choice policy
in numbers large enough to place pressure on the public school
system, and private schools must be able to absorb additional
students. This research attempts to answer these empirical questions.
This study analyzes primary survey data from private schools
as well Census tract-level data for households and additional
data that describe public school conditions. Statewide data from
Georgia are used. Two features make this study unique. The first
is the incorporation of a private school survey to analyze private
school supply. The second is the level of aggregation employed
in the data set that is at the tract/BNA level.
The results indicate that private schools' current excess capacity
is about 4,650 vacancies. The demand model indicates that a $1,000
income tax credit would encourage nearly 2,000 students to switch
to private schools. In other words, the proportion of school-aged
children in private schools would increase from 4.4 percent to
4.58, for an increase of 0.18 percent. Therefore, private schools
would be able to absorb all of the additional students who would
take advantage of this school choice policy, and there would be
no increase in tuition or queuing by prospective students.
However, these estimates of additional students and vacancies
represent extremely small proportions of the 1,040,000 school
age population. These are proportions that, if accurate, mean
the school choice policies we have simulated cannot lead to effective
competition between public and private schools. Our intuition
says that public schools would have to experience enrollment declines
that would lead to teacher layoffs and building closures before
administrators would experience competitive pressure. This pressure
may be felt when enrollment declines are in the range of five
to ten percent. Further research could review the literature in
school consolidation to obtain better estimates of how enrollment
decline causes school closure.
Finally, the demand model indicates that parents are sensitive
to indicators of public school quality in making their schooling
decisions. The model indicates that improvements in the local
school district's student/teacher ratio, standardized test scores
and teacher salaries lead to fewer students switching to private
schools. Public school administrators should recognize that their
clientele indeed pay attention to these and other measures of
school performance.
References
Flowers, Marilyn R. "Tuition Tax Credits and the Public Schools."
National Tax Journal. 41 (1988): 87-96.
Frey, Donald. Tuition Tax Credits for Private Education:
An Economic Analysis. Ames, Iowa: Iowa State University Press,
1983.
Greene, William H. Econometric Analysis. New York, New
York: MacMillan, 1990.
Gujarati, Damordar N. Basic Econometrics. New York, New
York: McGraw Hill, 1988.
Hamilton, Bruce W. and Molly K. Macauley. "Determinants and
Consequences of the Private-Public School Choice." Journal
of Urban Economics. 29 (1991). 282-294.
Keeler, Andrew and Warren Kriesel. "School Choice in Rural Georgia:
An Empirical Analysis". Journal of Agriculture and Applied
Economics. 26 (1994). 526-534.
Lankford, Hamilton and James Wycoff. "Primary and Secondary
School Choice Among Public and Religious Alternatives." Economics
of Education Review. 11 (1992). 317-337.
Matson, Barbara Smith. "School Choice: What Guides an Adolescent's
Decision?" Contributed paper to American Educational Research
Association, Atlanta, April, 1993.
Monk, D. Educational Finance: An Economic Approach. New
York: McGraw-Hill, 1990.
Sandy, Jonathan. "Evaluating the Public Support for Educational
Vouchers: A Case Study."
Economics of Educational Review. 11 (1991). 249-256.
Uchitelle, Susan. "School Choice: Issues and Answers." Phi Delta
Kappa Educational Foundation, Bloomington, Indiana.(1993).
Stiglitz, J. "The Demand for Education in Public and Private
School Systems," Journal of Public Economics 3(1974): 349-385.
Wallis, Claudia. "A Class of Their Own." Time. 144 (1994).53-63.
West, Edwin G. and Halldor Palsson. "Parental Choice of School
Characteristics: Estimation Using State-Wide Data." Economic
Inquiry . 26 (1988):725-740.
US Bureau of the Census. 1990 Census of Population and Housing:
Summary Tape File 3 Technical Documentation. Prepared by the
US Department of Commerce, Bureau of the Census. Washington, D.C.,
1992.
Table 1. Means and standard deviations for selected
variables from the survey of 45 private schools in Georgia, 1995.
| Variable
| Mean
| Standard Deviation
|
| Total enrollment
| 518
| 302
|
| High school enrollment
| 199
| 182
|
| Total operating capacity
| 93.5%
| 10
|
| High school operating capacity
| 90.5%
| 14
|
| Average SAT score
| 1015
| 108
|
| Tuition
| $4,141
| 1968
|
| Student/teacher ratio
| 11.28
| 2.83
|
| Acceptance rate for new students
| 59.4%
| 33
|
Table 2. Means and standard deviations for dependent
and explanatory variables in the private education demand model.
| Variable
| Mean
| Standard Deviation
| Expected Sign
|
| Proportion of children attending private school
| 7.41
| 11.88
|
|
| Adjusted tuition
| $6,833
| $1,587
| Negative
|
| Household income
| $11,640
| $7,988
| Positive
|
| Percent African-American
| 29.44
| 18.02
| Positive
|
| Percent adults without diploma
| 35.98
| 18.54
| Negative
|
| Expenditures per pupil
| $3,162
| $521.76
| Negative
|
| Student/teacher ratio
| 18.18
| 1.25
| Positive
|
| Criterion Reference Test
| 325.29
| 36.13
| Negative
|
| Average teacher salary
| $22,040
| $1,908
| Negative
|
Table 3. Tobit regression results for the private
education demand mode (Dependent variable is the log of the proportion
of children within a Census tract that are enrolled in private
school).
|
| Model 1
| Model 2
|
| Variable
| Coefficient (std. error)
| Coefficient (std.error)
|
| Intercept
| -1.8404
0.3881
| -1.6425
0.3680
|
| Household income
| 0.000052
0.000003*
| 0.000053
0.000003*
|
| Percent African-American
| 0.0112
0.0013*
| 0.0119
0.0012*
|
| Percent adults without diploma
| -0.0088
0.00009*
| -0.0087
0.0014*
|
| Adjusted tuition
| -0.0000278
0.0192*
| -0.000027
0.000013*
|
| Student/teacher ratio
| 0.0205
0.0005
| 0.00277
0.0157
|
| Criterion Reference Test
| -0.000207
0.0002
| -0.00058
0.00045
|
| Average teacher salary
| -0.00093
0.0014*
| -0.00063
0.00011*
|
| Expenditures per pupil
| 0.000014
0.0109
| N/A
N/A
|
Number of observations =2,433
* Significant at alpha < 0.05
Model 1 Log Likelihood = -1824.01
Model 2 Log Likelihood = -1825.29 |
