Wealth Neutrality in Higher Education: The Effects of Student Grants
Dcpzrtmeot
of Economics.
Gardner
Hall OilA. Hill. NC
University of North 17514. U.S.A.
Carolin:~
BI Chapel
Hill.
Chaprl
Abstract Public policy has been directed toward increasing equ:tl opportunitv of access ro higher education for over two decades. yet there has been little systematic investigation if the cffrctiveness of student financial aid to achieve this goal. This paper examines the effectiveness of publicly provided student grants to achieve wealth neutral college attendance in 1980. where rht: wealth neutral cyuity criterion is defined as an equal probabiliry of college attrndancr across household income. Student grants are seen IO rncouragr a mov~mrnt toward wraith neutrality. but do not complctrly remove the positive effect of income on the probability of collcgc attendance.
INTRODUCTION
Is
A IANDWYKK
California
case.
Supreme
educational the wealth (1975)
aid
secondary refers
the the
education
neutrality
distribution in The
is that
vices, such as education Fundamental services from
interests
should
individual’s particular
the general effect
alike.
them
principle
of
qualify
for
of ser-
from
in
it
Other
out
of wealth services
to have a wealth consumption
students
grants
year.
to student
A
of
higher
income cial
Since the mid-1960s a policy
the federal
to subsidize
government
college
has
aid
Loans.
attendance
both 107
increasing of being
were
college too
yet
the maximum
wealthy
for their
children
income
and made Basic Educational
of
trend
the
has occurred eligibility
Peil
were tightened the number
since
requirements
programs,
for
Pell Grants,
earning
to
for eligible
Loans
families
was
eliminated
for those qualifying
from
to
not
need by period-
legislation
now called
costs
wealthy
response
of financial
in-
aid based on
ceiling
reversal
eligibility
undergraduate
families
education
to
general
families.
financial
government’s
Eventually.
Grants,
neutrality
education. supported
the definition
tunity
neutralizing
higher
The
increasing
to
programs,
broaden
Student
for
steadily
existing
to afford
income
criteria,
that
in the bind
were
provide
lower-income
of student
aid.
ically
loans
edu-
and moder-
what
to
middle-income
without
families.
treat all goods
intergovernmental
individual
to an differs
singles
equity
equity.
of public
enough
of these
relationship
from
1970s.
of low-
initiated
low-interest
the argument
care, are deemed
It
Education
post-secondary
programs
as recipients
caught
categories
that
the
cluded
as
neutrality
The concept
provision the
and
Wealth
criteria
and
students
make
the children
families.
college
aid. The Higher
to
large-scale
criterion
and that consumption
aid intended on
decision
equity
certain
and horizontal
may be extended financial
of the
for
grants in
and
underlying
goods and services.
and services
intergovernprimary
and health
to pay.
equity
such as vertical
of
of
financial
feasible
become
Fetdstein
of
not have a strong
ability
other
light
it
a function
1965 aimed
ate-income
because
district.
student
of
cation
the California
education school
categorical
neutrality”.
wealth
that
consumption
to the implied
“wealth
ruled
of a child’s
for
(1%‘1), the
vs Priest
was unconstitutional
of his parents’
examined
mental
Court
system
made the quality
Serrano
through Act
the
Guaranteed Oppor-
available
to
as much as $25,000 of
increasing
the early
student
1980s. The
for the largest
Grants
and resulted
and
finan-
Guaranteed
in a decrease
and size of student
a
financial
in aid
a\\ard>.
Current
further
admInistration
cutbacks
Slnc2
ail
award>
are
incoms.
in stud2nt
publicly
provided
drtctrmiwd
equal
2ducatwn
a
focal
support
of
LVhils
research
presented
cate
an equity
as a determinant
neutrality
paper
as an equal
across
incomz:
(lY77)
found
wxs
groups.
not attained financial
families.
There
groups a group
to
xction.
The
for
student
individual
ths
rrccnt
programs
future
lev2ls
of
paper
collrge
of
for
of student
attendance. neutrality,
is
section and
attend
variablzs
dstermined.
is divicird of
is siven
into
the ths
for
the
colleg2.
The
on
wealth
neutrality
contains
a brief
third
and
an
model
thr
b2nefits
of
the
Iast
In
order
multi-period d2riv2
to
determine
aid on the the
education.’
utility
influence
nttendanc2
masimiration
determinants The
the
college
model
of
model drmand
hypothesizes
of
a
th2
their
is used
to
for
higher
that th2 hoube-
The 2ith2r
is,
base their uncertainty
not br
mod21
allous
the
collegz
of
at the time
expectations
for
the
for
and
these
expectations
these
choice
realized.
faced with
of each alternative
information
households
behavior
in
nttenclance yrar
to
future
decision
in high
available
to
individual.
school
is
to the housshold
tim2. strcfam
forms
aid
and thr that
the coll2ge
entttrs
exprctations
of the individual
altrrnativcs.
college.
school
of
and indiviclual
or may
use
and
valuation
Households
expectations
howe\,er,
income
dual
year
monetary
single-value
optimal
non-attrndancs
student
decision,
(1980).
s2nior
information.
may
and
the
financial
hIODEL
in
dew-
is assumed to be
that householcls
household
whethsr financial
that
decision
The
ing BEHAVIORAL
subjective.
and costs
non-colles2
summary.
of
uwd
and Abowd
associated
do form
based on ths
the
costs
such as those
by the parents
assum2s
uncertaintizs
at that
section the
and
and cost>. The
on imperfect
and expectations The
and from
including
traditionally
made at th2 2nd of the senior
of student grants
criterion.
highlv
calculations
periods:
results that
and costs
change
empirical
probability
utility
decision
with
benefits
Theorrtically.
behavioral
and ttstimation
future.
determine
four
in the following
contains
data drscription
last,
and the
as policy
description
attendance
y-ants
the
nonmonetary by nature,
the
aid to attain the And
(1976).
and to be one whose effects extend
school
benefits
of IYSO.
aid awards
is examined.
in high
into
is
Second,
er (II. future by
(1979)
attrndance
education. utility
benefits
models
the
(1977).
capital
and Roszn
college
both
the mod21
made final at the end of the individual’s
th2
income
class
financial
of the paper
grants
First.
by Willis
to
by Bishop
ths
ax
the from
single-period
Kohn
lifstimz
maximization
The
programs.
neutrality
peak levcl~. financial
of
current
across
school
whom
criterion
second
the52
ivealth
high
at their
Iev2ls
th2
is threefold.
attendance
degree
general
and
the analyses of the effscts
section
human
opzd
respect
contains
the
in
high
given
wealth
\vill
investments
wealth
aid with
of student
specification,
(1970).
both
for highzr
to capture
(lW),
to the wealth
remainder A
large-scale
and Radner er ctl.
the
attend
utility rnter
dttveloped
vnriablss
th2
mod21 of collrge
before
models
to
for
accrw
pwvious
Th2se
costs of college
sections.
Miller Fuller
may
of
of
not
allows
a
Iex,el
lifetIm
to rsplicitly
estsnds
of uhsther
investment
of the demand
model
maximization
by Bishop
that
college.
for the srnior
promote
The
model
the effects
neutrality
Th2
Neil
attendance and individual
th2
choosing
and
parents
as determinants
;~ttendance
the
utility
is a clear need to esnmin2
the: relative effectiveness direct
study
such.
colleqz
of this
attendance
coll2ge
As
purchase
a hiph2r
o\sr
than
of
of students
wealth
attendance is examined.
1960.
of this
effectiveness
of student
housrhold
collcg2-
to
basis
prwide
valu2
available
were
collrge.
consumption
on thr
uiil
rather
advo-
famllq income
pnxious
th2
loans.
not
of colleez
to
individual.
whether
indl\iduul
presznt
of
drtermincd
and
doss
th2
rducation
lo\v-income
the
and
grants
neutral
of th2s2
purpose
probability
X
in
criterion over
Th2
utility
o\er
for
and a potential
chooses
to
funding
&hate
higher
debate
aid \vas available
financial
neutralitv public
to
the
is defined for the purposes
student
of student
access
in
of parents
indl\idual.
college
of
th2 effscti\,eness
that tvealth
aid
collrge
h2r2
probability
compnwd
farnil\
of neutralizing
of colle~s
Lvralth
financial
hold. going
bv
student
critrrion.
aid in terms
financial
for
dsgrct2
issue:
continued
call
aid prosrams.
studsnt
to somil
opportunity
is
th2
propwals
financial
the direct amount
the
and
alternative
full-time
costs typrs
labor
and
of attrndof student
may
is chosen
assumes
the future
the college
household
alternative the
for
for
receive.
or not. Thr
that the indiviforce
upon
high
graduation. household collzge
selects
or non-coll2~e.
the
optimal basrd
alternative.
on the nlterna-
Wealth ,Ve’eutralit;vin Higher Education tive which yields the highest level of utility. The model yields a demand function of the determinants of college attendance which include parental income, student financial aid, the direct costs of college. the difference between lifetime earnings of the individual under the college and non-college alternatives, and household attributes which serve as proxies for the tastes and preferences of the parents and individual.’ Household and individual attributes include parental education, the individual’s academic achievement level, sex, race. geographic location and the distance to college. ESTIhlATION
AXD RESULTS
The empirical model used to estimate the probability that a household chooses college attendance for the individual is derived directly from the demand function for the discrete choice between the college and non-college alternatives discussed in the previous section. The primary purpose of the model is to examine the effects of student financial aid on the demand for college, while controlling for other variables which may influence the post-secondary decision. Some of the demand determinants require pre-estimation of instrumental variables. In particular, four general types of student financial aid are estimated then entered separately in the demand for college equation.’ The primary source of data used for the empirical estimation is the High School and Beyond (HSB) data set, a nationally representative longitudinal survey of over 28,000 high school seniors sponsored by the National Center for Education Statistics, taken in the spring of 1980. The survey was conducted at a time when student financial aid programs were well established, and proposals to reduce the funding of student financial aid occurred nearly one year after the senior cohort had been sampled. Thus, the sample contains households who were expecting student financial aid programs to continue at 1980 levels, and so provides an ideal base year sample from which to predict the effects of recent decreases in funding on the wealth neutrality criterion. Most of the data required for the analysis is provided by the HSB including socioeconomic and demographic background data, SAT-type test scores, and high school curriculum and grades. An exception is the data required for the estimation of the future return to a college degree. The investment payoff portion of the demand for college is
109
modeled separately. and its estimated value then applied to the observations in the HSB data. The data used for the estimation of age-earnings profiles are provided by the March 1980 Current Population Survey data from the Census Bureau. Student Financial Aid In order to model correctly the demand for college. the availability of student financial aid for college attendance must be included in the estimation. If college attendance is chosen, then student financial aid provides an incentive to attend; if college is not chosen, then financial aid is an opportunity foregone in lieu of some non-college alternative. Student financial aid is available to households from both public and private sources in the form of non-repayable grants and the interest subsidy from low-interest student loans. Four student financial aid variables: (1) public grants, (2) private grants, (3) public loans and (4) private loans are estimated for each individual in the HSB data set. The data used in the estimation of student financial aid is restricted to those individuals who indicated plans to attend college, some of whom received one or more types of financial aid, and many who did not receive financial aid of any type. That is. many observations of student financial aid are zero-valued while others form increasing continuous variables. Due to this truncated nature of the dependent variables, a tobit model is used for each of the four equations used to estimate the student financial aid variables. The independent variables chosen for the estimation of student financial aid are those factors primarily used by the government and higher education institutions to determine financial aid awards. In addition, variables have been included to control for demographic characteristics and for the availability of funds. The independent variables, coefficient estimates and asymptotic f-ratios for the tobit estimation of publicly provided student grants are given in Table 1.’ Public grants are defined as all non-repayable student financial aid from federal and state programs. Since student grants from the government are specifically intended to favor lower-income families, the highly significant and negative coefficient on parental income suggests that households from these groups do receive larger public grant awards than higher-income households as
110
Economics Table
1. Tobit
coefficient
of Education
estimates
Variable
Coefficient
1.6744’
Constant Parental income Direct costs Academic achievement Dependent siblings College siblings Sex (male = 0) Race (white =‘0) Race x sex New England region Middle Atlantic region South Atlantic region East South Central region East North Central region West South Central region West North Central region Mountain region Sample size Limit observations Non-limit observations -2 log likelihood ratio iSignificant fsignificant *Significant
-0.5911i O.llljt 0.1061+ 0.8579+ 0.1413 -3.0334: 5.4922: 0.3637 2.7902’ 3.5473t 0.6680 2.9295t 2.7760+ 3.2662i 2.9211; -t.-t79fIt
provided
student
Asymptotic
grants r-ratio
1.728 -22.706 5.36s 2.820 4.975 0.252 -1.978 8.222 0.275 1.744 3.2Jl 0.603 2.205 2.80s 3.004 2.474 3.221
4.145 2,105 2 .OJO 9-tO.06
at the 0.01 level. at the 0.05 level. at the 0. IO level.
expected. In addition, the direct cost of college, including tuition, books, materials and fees has the expected positive and significant influence on the level of public grants. The household’s financial need is increased as the direct cost of college increases and this is seen to increase the size of the student grant award. Financial need also suggests that family size increases the amount of the student grant, and this is confirmed by the estimated significant and positive coefficient on the number of dependent siblings in the household. Although the distribution of publicly provided student financial aid is not expected to discriminate on the basis of variables unrelated to financial need, the estimation results show that several other demographic variables are statistically significant in the allocation of student grants. These include academic achievement, sex, race and geographic location. The positive coefficient found for academic achievement indicates that there is a systematic bias in favor of higher qualified students, all else equal. The result suggests that colleges favor individuals with higher academic potential in the award process.
of publicly
Review,
Both sex and race are found to be significant factors in the estimation of student grant awards. The negative coefficient on sex indicates a bias against females and the positive coefficient on race indicates a bias in favor of nonwhites in the award process. The estimated coefficients in the dummy variables which correspond to geographic region of the household suggest a bias in favor of the Middle Atlantic, East North Central, West Central and Mountain regions. The result suggests that government grants are not uniformly distributed across regions. Perhaps states with these regions provide a higher level of federal matching student grants. Probability of College Attendance College attendance is defined as the binary dependent variable indicating full-time attendance at a four-year private or public college or university. As such, each individual in the HSB data set is categorized as either a college or non-college attender on the basis of responses to questions indicating post-secondary activities. The estimation technique used to determine the probability of college attendance is a logit model.’ The results of
Wealth Neutrality in Higher Education Table 2. Logit coefficient
Independent
estimates of the probability
111
of college attendance
Coefficient
Asymptotic
estimate
f-ratio
variable
--
-10.~381”
Constant
0.0052’
Parental income Public grants Direct costs
0.0073* -0.0031”
Academic achievement Private grams Public loan subsidy Private loan subsidy Future return Parental education Distance Sex (male = 0; femaIe = if Race (white = 0; nonwhite = 1) Race x sex (other = 0; nonwhite New England region Middle Atlantic region South Atlantic region East South Central region East North Central region West South Central region West North Central region Mountain region -2 log likelihood Estimated R’ Sample size
0.1762’
female
-0.0033 0.0325 -0.6875 0.0(x)3 0.1676” -0.0030’ 0.0198 0.8917’ -0.0434 O.j986* 0.567Ht 0.121tt 0.73X* 0.6971” 0.7061’ 0.6255” 0.53#6*
= 1)
ratio
‘Significant at the O.Ul level. f Significanr at the 0.10 level. $A11 coefficients in the equations
- 10.715 6.500 2.642 -7.750
42.95 1 -0.375 0.235 -0.435 0.056 13.462 -2.727 0.278 5.837 -0.059 -i*128 1.808 1.708 5.742 7.628 3.604 5.666 2.638
-!641.9JY$ 0.2857
I I.500
are jointly significant
the estimation are given in Table 2, which lists the independent variables, coefficient estimates and asymptotic f-ratios. The estimated coefficient found for parentai income is positive and statistically significant. The rest& indicates that, while controlling for other variables, including student financial aid, individuals from higher-income households have a higher probability of attending college than lower-income individuals. Publicly provided grants are found to have a positive and significant effect on the probability of attending college indicating that higher levels of student grants increase the probability that a household chooses to send the individual to college. These results, taken together, suggest that while the probability of college attendance is strongly and positively related to the household’s ability to pay, student grants tend to moderate differences in college attendance rates across in-
at the 0.01 level.
come groups. Lower-income househofds receive larger student grant awards and these awards are seen to increase the probability that the individual attends college. This relationship is explored in detail in the following section. Other results of the estimation indicate that the direct costs of coilege have the expected negative influence on the probability of college attendance and that academic achievement has a significant and positive effect. Higher tuition costs, materials and fees tend to decrease the odds of college attendance, all else equal. Other student financial aid, including privately funded student scholarships, and the interest subsidy from student loans are not found COhave a statistically significant influence on the probabiIity of college atrendance. Other factors, such as academic achievement, parental income and the direct costs of college, which are included in the estimation and which in part determine the levels of
11’
Economics
of Education
these types of student financial aid tend to diminish the direct effects of these awards on the probability of college attendance.
Rmieu
disturbance term. Student grant amounts are estimated for the entire sample. and fitted values, c;,, are used as instrumental variables in the college attendance equation:’
W’EALTH NEUTRALIT In [p,l(l - p,)] = B0 + &Y, + &d, Simply stated. wealth neutrality in higher education requires that the household’s decision of whether the individual will attend college will not be influenced by the household’s ability to pay the costs of college. In a situation of wealth neutrality individuals vvho are equally qualified to attend college. but who are unequally circumstanced with respect to parental income. should have equal opportunity of access to higher education. Operationally, the wealth neutrality criterion is taken to be an equal probability of college attendance irrespective of parental income. while controlling for all other variables. Since in the reported research parental income was found to be positively related to the probability that an individual would attend college, wealth neutral college attendance does not appear to have been obtained by those entering college in 1980. In an attempt to clarify the situation with respect to these students, the following analysis focuses on the relationship between parental income. publicly provided student grants, and the degree of wealth neutrality attained. In addition, the degree of effectiveness of student grants and the reduction of direct costs of college. as policy variables to attain wealth neutrality, are compared. Direct and Indirect Effects The analysis of the effects of student financial aid on the extent of attainment of wealth neutrality is limited to publicly provided student grants, the only type of student financial aid found to be statistically significant in the estimation of the probability of college attendance. In order to determine how the probability of college attendance varies with parental income, both the direct effect of income and the indirect effect of income through its influence on the amount of student grant awards must be considered. The estimation equation for the level of student grants is of the general form: G, = a, + cX]Y, +
+ k,
(1)
where G; is the amount of the publicly provided student grant, Y, is parental income and p_; is the
+
+ E, (2)
where pI is the probability that the ith household chooses college for the individual and E, is the disturbance term. Thus. parental income is seen to influence the probability of college attendance in two ways. The independent and direct effect of parental income on the probability of college attendance. (ap,iak;), is found to be positive and statistically significant. The indirect effect of parental income on the probability of college attendance through student grants. (aP,ia6,)(ad,iaY,). is found to be negative, i.e. higher parental income leads to lovver levels of student grants. Because student grants are positively related to the probability of college attendance, the indirect effect of income through grants is a negative influence on the probability of college attendance for individuals from higher income households, and a positive influence for individuals from lower income households. The full effect of parental income, (aP,IaY,) + (aP,iaC,, (aC;,iaY,), contains both the direct and indirect income effects (which are opposite in direction) on the probability of college attendance. Other independent variables included in the estimation of student grant amounts and in the estimation of the probability of college attendance (such as the academic achievement of the individual and the costs of college) have similar direct and indirect effects on the probability of college attendance. In the case of academic achievement, both the direct and indirect effects are found to have positive impacts on the probability of college attendance. The money costs of attending college are found to have negative direct effects and positive indirect effects on the probability of college attendance. Estimated college attendance probabilities and changes in probabilities for five values of parental income are contained in Table 3. Probabilities are estimated for an individual with sample mean characteristics, and sample mean probabilities (P) and standard deviations (a[,) of the probability distribution are given for comparison purposes. The probability distribution for the base case, presented in column (1). illustrates the degree of wealth
113 Table 3. Estimated
college dttendancs
probabilities
and changes an probahilltier’
Income
effect5 Indlrrst
Direct
(3)
(4)
il~.O33Y +0.0761 +o. 1331 +O. 1797 +0.305
AlI. 1036 TO.1)601 +0.0161 0.0 0.0
+o. 1x5 +o. Ii63 +o. 143’) +u. 17Y7 +0.‘NS
+0.0217 -0.0389
+o. I-123 +o.o-I3
+0.121 +o.os13 *Probabilities
neutrality
of college
The
estimated
ance
is seen
creases.
mean to
attendance
attained of college
increase
probability,
and.
to the extent
of lack of wealth the greater the
ability
The
of
base
distribution
tal income,
income
and
increases
neutrality. shown
This
direct
in column
between
and
the
come is set equal equation.
(2). the direct percent
effect
the dispersion the
direct
probability the effect incomes
effect
of parental
is such that receive
neutrality.
is
in-
income with
individuals tends
12 in
does
shown both
the
student (direct percent
and a smaller
usually
income direct
with than
income
do
grants
effect.
for
student
grant
awards
attendance income
and
The total
parental
on
attendance
are
represent
for each group
effects
the
the when
effects
income
of
effect
for approximately
14
of attendance, income
due
to the
with
lower
which
to
income
of parental
of college
probability
those
tend
groups.
accounts
higher
for
receive
high
of college
are removed.
in probability
in-
grant
they
ineligible
probability income
lowest
is smaller
and
of
higher
student
(4). These estimates
of the mean
increases
the
of student
income
and indirect)
Individuals
are
distribution
direct
grants
The
the equality
attendance
the
parental
awards.
of the full effect
in column
college
grants
induce
because
the allocation
the probability
in
high-middle
rates across parental
but
The result
grant
to increase
college.
higher
incomes
indirect
in the degree
larger
of student
for
who
Thus,
increase
effects.
distribution
only
student
Estimates
of the distribution Not
effect
this
of the student
in the dispersion
grants
receive
higher
no effect
grants.
attendance for over
probabilities
The
2 percent
and an increase
which
grants
of parental
households,
attendance
with
of student effect
causes a decrease Student
groups
have
P. and increases
direct
low-income
that
i.e. when
P. Because
neutrality.
smaller
of college
will attend
those
the largest
probability
of wealth
income
accounts
households.
that an individual
is a less equal degree
income
attend-
the results in column
probability,
college
those
of wealth
parental
wealth
awards.
parental
as the difference when
of income
favor
attendance
for over
of attendance,
of income
in column
are removed.
indirect
is seen to account
come
criterion. of
base probability
and skewness
of higher
of paren-
college
This
of parental
the base case in the
in the estimation
mean probability
influences,
is shown
grants
to zero.
of the distribution,
to zero in the college
of the mean
income
is set equal
effect
grants of
student
prob-
of parental
As can be seen from
parental
indirect from
probability
when
closely
the degree
probability
results
are
(7) and is defined
the estimated
attendance
favor
effect
It is the difference
attendance
of college
decreases
student
effect
effect
the probability
ance and, therefore,
(3).
of the
through
grants
neutrality
direct
estimates
income
and indirect
of the wealth
The
income
neutrality.’
the full effects
the direct
independent
attendance:
neutrality
includes
both
on achievement The
wealth
from
of the degree
of parental
case college
nith sample mznn characteristics
estimated
the mean
of the distribution,
effects
I
income
in-
it differs
in college
of wealth
and indirect
degree
related.
about
that
deviation
the degree
The direct and
neutrality
19SO.
attend-
income
as a measure
the standard
in
of the distribution
of dispersion
zero. may be interpreted the lower
as parental
deviation
the degree
for an individual
probability
The standard
measures
are estimated
Full
(2)
decreases
full
p.
have larger effects
incomes.
of The
the degree
of
Economics
11-I wealth
neutrality.
is
indirect
effect
of
increase
the degree
larger
student
than
grants
of wealth
the
of Educarion
offsetting
which
tends
neutrality
Rer,ieti
For to
each
amount
in college
nation Achievement
Tl-12 essentia! wealth
neutrality
are equally
and Student
requirement criterion
qualified
access of opportunity the academic seen
to
the
grants
positive
to college.
As is noted
the
level
level in
on
It is also
of
the
on
the
Individuals college
are
are not only
but also to receive will
also
larger
of
and
direct
college in-
to
to attend of
The the
measure
of academic
tests,
high
curriculum.x
school
achievement the
high
index
to attend estimated
presented,
level.
represents
college.
student
grants.
are given
The
for
characteristics
course such.
measure
used
In Table increases.
of
grades
th2
how
student
broken
estimated amount
down student
for the other
grants
who
amount
income
available mean
Low Low-middle Middle High-middle High Mean
Table
5 presents
25
61.2S.i 694 103 0.0 0.0
$1.329 738 147 0.0 0.0 __777
*Estimates are obtained by using tobit coefficients repayable awards from federal and state government
by
income
levels
shown
income
in Table
6, reveal
have the largest
attend
from
college.
help
to account
ability
that
these
to attend
for
mean probability
effects
on the parental
effects,
academic that
attend
achievement (Base) 50 5 I.391 SO6 215 0.0 0.0 259
which
achievethe
college
(0.5016).
(percentiles) 75
90
Yl.461 570 279 II.0 0.0
$1.516 925 334 0.0 0.0
353
to of
that student
the lower
the fact
individuals
the mean
effects
These
of average
and
in the
are more likely indirect
of
college level
at or above
that individuals
groups
effects of
are less likely
The
to
achievement
indirect
probability
and income
equal.
of
attributable
student grant awards*
10
178
and
levels
ment.
Academic income
to the
We note that individuals
as expected.
the sample
variables.
as
probabilities
group
the
are largest for students
Estimates
sample
each
individuals
else
achievement.
parental
Table 4. Estimated
Parental
all
probability
college. have
and that
by
college.
direct
sample.
achievement
grants.
grant
for
levels of achievrment
for
student
decrzases
the likelihood
and parental
increase
lowest
attend.
award
attendance
the
college.
achievement
of the award
chooses to attend individuals
well
In Table
relationship
available
both for each parental
for the total
and
inverse and
5. we see that as academic both
attendance
academic
of
academic
for
larger
6 shoas the part of the total probability
attendance
of SAT-type
is the individual.
of five
the annual
if the individual
a
amount
levels income
As
is
at each
index
school
college
individuals are
achievement
is a composite
achievement
4,
achievement
achievement
estimation
prepared
academic college
attendance.
highly
receive
increases
attends
attend
college
better
more
can be said to be effsctive
colleg2
grant awards which
probability
an
its availability
estimated and Table
of
increases.
grant
that
discrimi-
favor
equal,
The 2stimated
that an individual
attendance.
qualified
student
their
extent
achieve-
positive
of college
A student
and
estimated
as academic
note that at each level of
income
incorn
the
students
hand.
achievemrnt
is evident.
in
slse
prepared
par2nral
parental
provided
academic
likely
grants is
All
On the othsr
between
above,
process
students.
academic
equal
a direct
that
better more
increase
has
probability
true
probability
who
publicly
grants.
who
have
of the individual
turn,
has an independent
fluence
individuals
college
which,
influence
attendance. ment
is that
the
Isvel
increases
The results suggest
auard
academically
obtaining
to attend
achievement
increase
studrnt
Grants
for
grant
increases.
in
qualified
income
student
achievement
attendance. Academic
parental
of
409
shown in Table 1. Student grants include all nonstudent financial aid programs.
probis near
Wealth i\;eurralirlr: in Higher Educatmn Table 5. Estimated
college attendance
probabilities Academic
115
by income and academic
achlscement
achievement
(percentiles)
(Ewe) Parental
income
10
25
50
75
90
Low Low-middle hfiddle High-middle High
0.1263 0.1296 0.1331 0.1549 0.1841
0 2371 0:23x o.243q 0.2746 0.3 179
0.1572 0.4949 0.5026 O.j;Y-! 0.5s95
0.7367 0.712 I 0.7479 0.7767 0.8051
0.8774 0.8806 0.8838 0.89-N 0.9125
P UP
0.1326 0.0213
0.2475 0.0358
0.5016 0.0423
0.7472 O.OW
0.883-I 0.0142
Table
6. Effects of student grants on the probability Academic
Parental
achievement (Base) 50
(percentiles)
10
25
Low Low-middIe Middle High-middle High
+O.O368 +0.0217 +O.OO% 0.0 0.0
+0.0635 +0.037s +0.01 I7 0.0 0.0
-CO. 1026 +0.0601 +0.0t61 0.0 0.0
+0.0933 +0.0529 +0.0160 0.0 0.0
+0.0579 +0.032-r +0.0107 0.0 0.0
P
l-0.0060 -0.0133
+0.0120 -0.0231
+0.0217 - 0.0;89
+0.0155 -0.0371
+0.0132 -0.0227
UP
income
of college attendance
The effects of grants for individuals who are below average in academic achievement, and who, therefore, have less than even odds of attending college, are smaller than the effects for their better qualified counterparts at each achievement percentile. In the lowest income level, for example, grants account for 3.68 percentage points of the probability of a student in the 10th percentile of achievement attending college, while for a student in the 90th percentile these effects explain 5.79 percent of the total probability. The availability of student grants for below-average achievement individuals, who are less likely to attend college and who generally receive smaller than average grants, therefore, does little to increase the probability of college attendance. On the other hand, student grants to higher than average achievement individuals, who are more likely to attend college and who tend also to receive larger than average grant amounts provide smaller than average increases in the probability of college attendance. It is those students between the low- and high-achievement
75
90
extremes for whom grants have the most important impact on the attendance decision. If the goal of student grant programs is simply to induce college attendance, the above results suggest that student grants are very effective for average achievement individuals, less effective for below average achievement individuals, and essentially turn out to be gifts to high achievement individuals who generally attend college anyway. Student grants do, however, tend to equalize college attendance probabilities for individuals from different parental income levels when academic achievement percentile is controlled. The changes in the standard deviation of the probability distribution shown in Table 6 indicate the influence of student grants on the movement of the distribution toward wealth neutrality. versus Direct Costs In Table 7 the effectiveness of an increase in student grants to induce college attendance and equalize college attendance probabilities across Student Grants
116
parental
incomes
in the direct estimated
grunt
column
(3)
to that
of college.
college
student from
is compared
costs
Column
attendance
avvards
presents
are the
uould
be estimated
college
increased
individuals levels
who
were
low
to
eligible
Overall,
the
increase
amount
would
be
persion
The
by about
at the
$100
in grants
probability
percent
for
parental
to
and
increase
student
from
when
of change of those
in
IWO.
decrease
the
dis-
0.0-17_.~ to 0.4OlJ.
a
probability (1)
decrease
in
probability
tuition
would
generally awards.
grant
families,
costs
effect
from
anyway.
college
of
odds
of college
the
increased reduction.
probability with
have
sample
of
levels,
the at
cost to
tuition their
touard
wealth
in
would
tend the
wealth
neutralitv.
practices
to negate
the
neutrality.
the
seen
if
in
the
The iead
in
a
tuition
reductions of
wealth
to prevent students
tuition
the that
reductions. be
suggest
to a decrease
in
movement
degree
decreases
results
the
indivi-
that increases
possible with
to be a
aveqe
induce the
of
increase
aid to poorer
along
of
in
result
appears
the
whereas
it were
to occur
would
would
mean probability
for to
that
decreases
increased that
actual
in the deqee
of
For
SUhlhlARY
student
effect
is absent
probabilities
of
of the direct
college
of is
which shows
or
policy
decrease
effect
h&h-income
characteristics
policies
it is noteworthy
to
would
on
estimates
through
suggested
neutrality.
actually
effect
re-
degree
surprisinp.
to substantially
are
reduction
receive
as a result The
grants
expected
reductions. and
dual in the sample.
tend
the
that
the
grants
attendance
by the lower a
two
Neither
means
Only
in
the student
attendance.
neutrality.
groups
to an extent
or no change in the overall
student
result,
while
amounts
because
indirect
attending mean
increase
income
the
the
indirect
the
to cause a decrease
eli$ble
individuals
to
magnitude
of
in
that
by a decrease
these
and
shown
This
to decrease
in
the
and fees on the
caused
tends
not
attendance
the
tuition
of the
and to lessen
tend
of
increases
effective
for an individual
hi%h-middle
are
of the cost
individual
either
cost
at lower
are understood,
comparison
very
is
tend
be offset
is of sufficient
who
these
occur
a
the dispersion
once the mechanisms
the board
indicate
income
For
awards.
direct
individuals
not
attendance
parental
in direct
positive
The
in the probability
grant
attendance
that
results
costs
would
student
effect
attendance
of college
decrease
materials
(5).
to middle
student
neutrality,
tuition
of
grants
to 0.0467,
neutrality,
will
lovvered
effects
reduced
0.0473
wealth
little
of an across
college
and
the increase
and
in tuition.
of
columns
grants
the effects
of $100
the low
A
by
tend to increase
tuition
college
comparison.
decrease
of
result
offsetting from
from
college
attendance. For
levels
changes
this
neutral
income
not illo$cal
by
base
the
resulting
distribution. duced
the
that
reduction
wealth
grants
uealth
from but
degree
income
grants
student
expected
toward
by
the
0.7
for in
costs,
the
terms
middle
of the distribution
movement
in
to increase
attendance
changed
probabilities
result
the base probabilities.
of a decrease (2) shows
indicate for is
an not
For
over
tlvo
college
sidized
decades.
the
attendance
government by
has
providing
sub-
student
orants to individuals from lower-income families c who historicallv have had lower college attendance rates
than
presented
higher-income in this
paper
families. suggest
that
The
results
the policy
has
Wealrh been generally opportunity student
consistent
Srants tended
college income
promote
wealth
in Higher
the
In 19YO
The
analysis
whether
also
shows.
however,
student
neutrality
specific
in
student
grants
of household an individual
higher
lower
income
probability
only
income attends families
of attending
partially
ability
At
offset
the
as a determinant
of
college. are
seen
college
Individuals to have than
income
grant
student
a
indivi-
of
to
reverse
and
grant
Grants
Grants)
will
college
attendance
tend
families. trend
These toward
(Supplemental and State Student
to decrease by
the prob-
individuals
reductions wealth
neutrality
Acknowledgement The author wishes to thank John Alvin for useful comments and suggestions. Any remaining errors are, of course. my own.
An econometric model of the U.S. market for higher education. In Research in (Edited by EHRENBERG, R.G.). Vol. 4. JAI Press. Greenwich. CT. BISIIOP. J. (1977) The effects of public policies on the demand for hi_rher education. J. hum. Resources (1980)
Economics
12. X-307. FELDSTEIN,h1.S. (1975)
Wealth
neutrality
in
education.
REFERENCES
J.M.
from
are likely
I. For those interested in the fully developed theoretical and empirical models, see Schwartz (1985). variables are expressed in present value terms at the time of the college attendance 2. All monetary decision. 3. Pre-estimation is also required for the future return to coller+e. Details of the empirical specification, data, and results for the future return instrument are gjven tn Schwartz (1985). 1. Estimation results for the other three financial aid variables are similar to those found for publicly provided student grants and are given in Schwartz (1985). 5. The model is a reduced form of a structural model of both the household’s decision and college admission. where admission is assumed to be independent of the error term. The logit model was chosen over a probit model for computational ease. 6. Some may argue that a selection bias is present in the data since only those individuals who enroll in college can be observed to receive student grants. However. correction of the potential bias is impossible in the model. See Schwartz (1985). 7. As a measure of the degree of wealth neutrality. the standard deviation is useful for comparison purposes only. A zero standard deviation implies that wealth neutrality requires that college attendance probabilities across parental incomes be set equal to the mean probability, which may not be the intent of student financial aid programs. S. Details of the method used to construct the academic achievement index are contained in Schwartz (19S2).
Labor
in
to eliminate
NOTES
ABOWD,
while
grants. cutbacks
proposals
programs
Opportunity
the
even
of student
supgest that recent
funding
lower-income higher
families.
for the availability
Incentive
wealth
of college
was not achieved.
117
The results further
attendance
that
as an equal probability
across all incomes,
19S0 levels influence
from
Educational
defined
attendance
duals
Educarion
controlling
and thereby
education.
neutrality,
higher
education.
to equalize
across household
are seen to effectively
from
a goal to improve
of access to higher
probabilities higher
with
.\.eurralirl;
and local choice
in public
education.
,4rn.
econ.
Rev. 65,
75-89. FULLER. W.C., MANXI. C.F. and WISC, D.S. (1981) New evidence in the economic determinants of postsecondary schooling choices. J. hrcm. Rrsources 17, -177~495. KOHN, M.G.. ~JANSKI, C.F. and MU&DEL. D.S. (1976) An empirical investigation of factors-influencing college going behavior. Am. econ. sot. ~Wrusrrre 5. 391~-119. MILLER. L. and RADNEK, R. (1970) Demand and supply in higher education: a progress report. Am. econ. Rev. 60. 326-334. SCH~AKTZ, J.B. (19Sj) Student financial aid and the college enrollment decision: the effects of public and private grants and interest subsidies. Econ. ed. Rev. 1. 12% 144. SCHH.AKTZ, J.B. (1982) Wealth neutrality and public subsidies to higher education. Ph.D. dissertation. University of North Carolina at Chapel Hill. SEKRANO vs PKIEST (1971) L.A. 29820, Superior Court No. 93825-1. cited in Hnv. Educ. Rev. 41. 503. WILLIS. R.J. and ROSEN. S. (1979) Education and self-selection. /. polir. Ecorz. 87, S7-S36.