document resume author garas, walter i. … resume ed 103 979 ea 006 903 author garas, walter i....
TRANSCRIPT
DOCUMENT RESUME
ED 103 979 EA 006 903
AUTHOR Garas, Walter I.TITLE Use of the Lorenz Curve and Gini Index in School
Finance Research.PUB DATE Mar 75NOTE 19p.; Paper presented at the Annual Meeting of the
American Educational Association (60th, Washington,D.C., March 31-April 4, 1975)
EDRS PRICE MF-$0.76 HC-$1.58 PLUS POSTAGEDESCRIPTORS Assessed Valuation; *Educational Finance; Elementary
Secondary Education; *Expenditure Per Student;Expenditures; *Finance keform; Fiscal Capacity;Income; *Statistical Studies; Statistics; Tax Effort;Tax Rr.tes
IDENTIFIERS Gini Coefficient; Lorenz Curve
ABSTRACTThis paper proposes two modifications of the Lorenz
technique for measuring inequalities in financing education. One isto order expenditures (and tax rates) by district fiscal abilitybefore calculating the cumulative percentages. It is possible, as aresult, to get a Lorenz curve that crosses the diagonal line, whichgives some useful insights. The other is to calculate Ginicoefficients for both expenditures and tax rates and to consider thesus of the two coefficients to be a measure of the total inequalityin the system. (Author)
a
USE CF THE LORENZ CURVE AND GINI INDEX IN SCHOOL FINANCE RESEARCH
DE PuTNIE NT OF HIALTN.EDUCATION TEELFARENATIONAL INS MUTE t:
buCA TIENTEIls DOCUMENT t4AS BEEN REPRODuCEO FxACttV t.. RECEIVED FfEox/ME PF. kW's' Ok owC,ANIZA TION ORIGINA T POINTS OT v/E W ore opinmy4sytAtE0 DO NOT NTCrSSARity PEP4ESENT OR t ,AL ATIONAL INS,CTUTi OFECIVC A T.ON POSITION OP POLICY
Walter I. Garms'
University of Rochester
March, 1975
Cr" (Prepared for presentation at the 1975 Annual Meeting of the American
cy. Educational Research Association, Washington, D.C., March 30-April 3, 1975)
C)r.4
C.73 Those interested in school finance have long been interested in
LAI inequalities of educational expenditure and educational effort. Economists
have long been interested in inequalities of income. It is unfortunate
that there has not been more communication between these two groups of
people, for the economists have one standard technique for looking at
income inequalities that can be very useful in examining inequalities in
educational expenditure and effort. That technique is the use of the Lorenz
curve and its associated Gini coefficient. However, the use of the Lorenz
curve in the way it is usually used by economists would give misleading
results as a measure of inequalities in school finance. It is the purpose
of this paper to describe and illustrate a modification of the Lorenz
technique that yields some unique insights for those interested in school
finance research.
As used in the measurement of income inequality, the Lorenz technique
is to array all individuals in order of ascending income. Then at convenient
points (for each one percent of the total individuals, say) the cumulative
percent of individuals and the cumulative percent of total income is
calculated. As an example of how this is done, the information for families
1
2
in the United States in 1969 is given in Table I. The cumulative
percentages are then plotted on a graph called a Lorenz chart. The
horizontal axis is the cumulative percentage of families, running from 0
to 100, and the vertical axis is the cumulative percentage of income, also
running from 0 to 100. The data given above have been plotted on a Lorenz
chart (Figure 1), and are represented by the curved line. The straight
diagonal line represents complete equality of incomes. That is, if all
families received the same income, the "lowest" ten percent of the families
would receive ten percent of the income, the "lowest" fifty percent would
receive fifty percent of the income, and so on. The extent to which the
curved line sags below the diagonal line, then is a measure of the
inequality of incomes. This measure is formalized in the Gini coefficient,
which is the ratio of the area bounded by the diagonal line and the curved
line, divided by the area bounded by the diagonal line, the horizontal axis,
and the right-hand vertical line. A Gini coefficient of zero would represent
absolute equality,and a Gini coefficient of 1.00 would represent absolute
inequality (the situation in which one individual got all of the income and
no one else got any). The Gini coefficient for the family income data
displayed above is about 0.36.
There are, of course, other measures of inequality that can be used.
They .1clude the range of incomes (the highest less the lowest), the
interquartile range (the difference between income at the third quartile and
the first quartile), the interquartile range divided by the mean, the standard
deviation, and the coefficient of variation (the standard deviation divided
by the mean). Each is a better measure than the previous one, at the cost
of increasing difficulty of computation and of explaining to a lay audience.
The range takes into account only two incomes, the lowest and the highest.
Table I
Income
Rank
Family Income in the U.S., 1969
Percent of Cumulative Percent of
Total Income Families Income
Lowest tenth 1 10 1
Second tenth 3 20 4
Third tenth 5 30 9
Fourth tenth 6 40 15
Fifth tenth 8 50 23
Sixth tenth 9 60 32
Seventh tenth 11 70 43
Eighth tenth 12 80 55
Ninth tenth 16 90 71
Highest tenth 29 100 100
100
90
80
70
Percent6o
of 50
Income
40
30
20
10
Figure I
10 20 30 40 50 60 70 80 90 100
4sr
Percent of Families
3
4
Thus this most widely used measure of inequal:ty is only a measure of two
extremes that are almost certain to be unusual cases. The interquartile
range ignores these extremes, thus giving a more stable measure at the
expense of ignoring half of the data. The interquartile range divided by
the mean is simply a way of standardizing the interquartile range so that
distributions with different means may be validly compared. The standard
deviation (and its standardized version, the coefficient of variation)
takes into account all of the data. Its principal drawback is that
condensing all of the data into a single statistic does not allow us to
look at any other characteristics of the distribution. The Lorenz curve
allows visual inspection of the characteristics of the distribution, and
the Gini index condenses the information on inequality into a single
statistic that is comparable across distributions with different means.
The difficulty with applying the Lorenz curve to expenditures per
student in the public schools is that not all of the differences in
expenditure are undesirable. At the same time that states are engaged in
equalization programs to reduce the differences in expenditure caused by
differences in fiscal ability, they are subsidizing programs for special
students that result in higher expenditures for them than for normal
students. Unfortunately, the results of these two kinds of subsidization
are combined in a single statistic called expenditures per student. Part
of the difficulty in trying to measure inequalities of educational
expenditure is the difficulty of separating expenditures for normal students
from expenditures for mentally retarded, physically handicapped, occupational
education, and for children in necessary small schools. Measures such as
the range, interquartile range, and standard deviation are incapable of doing
this. The Lorenz curve, as traditionally used by economists, is also
5
5
Table 2
DistrictNo. Students
Expendituresper Student
A.V. perStudent
Cumulative Percent ofStudents Expenditures
1 1000 800 20,000 10 72 2000 900 40,000 30 21
3 500 1000 60,000 35 25
500 1100 80,000 40 30
56
1000
1000
12001300
100,00090,000
5060
4050
78
2u00500
1400
1500
70,00050,000
8085
7380
9 500 1600 30,000 90 86
to 1000 1700 10,000 100 too
100
90
8o
70Percent
of 6o
Expen-diture 50
40
30
20
10
0
Figure 2
0 10 20 30 40 50 60 70
Percent of Students
80 90 100
Gini = .122
C
6
incapable of making this distinction. Table 2 and Figure 2 give fictitious
data for a sample of ten districts, and the Lorenz curve based on that data.
As with the chart of family incomes, the data have been arrayed in order
of increasing expenditure per student, and the cumulative percentage of
students and of expenditure is calculated. The data on assessed valuation
per student is included for information purposes, but is not used in the
calculations. Since there is no way of knowing how much of the expenditure
in a particular district is the result of differences in fiscal ability,
how much the result of differences in community desires for education, and
how much the result of differences in expenditures for special students,
the chart is interesting, but not especially informative.
If, on the other hand, we choose to array our districts in order of
increasing fiscal ability, the Lorenz chart can then give us information
about the extent to which differences in expenditures are correlated with
differences in fiscal ability. In Table 3 and Figure 3 this has been done.
The data on assessed valuation per student were purposely chosen to be as
uncorrelated as possible with expenditures per student. The result shows
in the Lorenz chart, where the Lorenz curve is much closer to the diagonal
line.
Figure 3 also shows a possible consequence of the use of this technique:
part of the Lorenz curve lies above the diagonal line. In the conventional
use of the Lorenz curve it is impossible for this to happen. But this is
a virtue of this variation of the technique. We can see at a glance, for
example, that the ten percent of the districts that are lowest in fiscal
ability actually spend 14 percent of the total money spent. It can readily
be seen that, with this technique, it is possible to have substantial
differences in expenditure, yet have the Lorenz curve lie close to the
7
Table 3
DistrictNo Students
Expendituresper Student
A.V. perStudent
Cumulative Percent ofStudents Expenditures
10 1000 170o 10,000 10 14
1 1000 800 20,000 20 20
9 500 1600 30,000 25 27
2 2(30 9co 40,000 49 42
8 500 1500 50,000 50 48
3 500 1000 60 ) 55 52
7 2000 1400 7C. dO 75 75
500 1100 80,000 80 80
6 1000 1300 90,000 90 90
5 1000 1200 100,000 100 100
100
90
80
70
Percent
of60
Expen-diture
50
40
30
20
10
0
Figure 3
0 10 20 30 40 50 60 70 80 90 -100
Gini . 0.004
Percent of Students
8
diagonal line. In calculating the Gini Index, it is appropriate to think
of areas between the Lorenz curve and the diagonal line that lie above the
line as negative areas. The result is that there could be substantial
deviation of the Lorenz curve from the diagonal line, but if this were the
result of differences in expenditure that were uncorrelated with differences
in fiscal ability, the Gini Index would be close to zero. Indeed, that is
so here, where the Gini Index is only 0.004, compared with an Index of
0.122 using the same data differently arrayed in Figure 2.
It should be particularly noted that the basic data used in Figures 2
and 3 are identical, and the axes of the Lorenz charts still have the same
labels. Only the method of ordering the data before calculating the
cumulative percentages is different.
We have been examining one side of the coin of fiscal neutrality, that
expenditures should not be a function of community wealth. The other side
is that tax effort should not be a function of community wealth. The Lorenz
curve can be used to examine this in an analogous fashion, although the
meaning of a cumulative percentage of tax rates is not as intuitively obvious.
In Table 4 and Figure 4, tax rates have been calculated for the districts in
our sample assuming a power equalizing system with full recapture that
guarantees an expenditure of $100 per student for each mill of tax rate.
It can be seen that the tax rates have a correlation of 1.00 with expenditures,
and thus the Lorenz curve for tax rates looks just like the Lorenz curve for
expenditures. However, it has an "opposite" meaning. That ;s, in the same
way that high expenditures per student and high family incomes are considered
to be "good", high tax rates are considered "bad." Thus, where an invidious
discrimination would tend to make the Lorenz curve of expenditures fall below
the diagonal, it will make the Lorenz curve of tax rates lie above the
9
9
Table 4
Power Equalized System
Dist. Studs.
Exp./Stud.
AV per Tax
Stud. Rate
Cumulative Percent ofStuds. Expends. Tax Rates
10 1000 1700 10,000 7 10 14 14
1 1000 800 20,000 8 20 20 20
9 500 1600 30,00C 6 25 27 27
2 2000 900 40,000 9 45 42 42
8 500 1500 50,000 5 50 48 48
3 500 1000 60,000 0 55 52 52
7 2000 1400 70,000 4 75 75 75
4 500 1100 80,005 1 80 80 80
6 1000 1300 90,000 3 90 90 90
5 1000 1200 100,000 2 100 100 100
Percent
of
Tax Rates
100
90
80
70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Figure 4
Percent of Students10
10
diagonal. In order to make the Gini coefficients comparable, then,
between exrcr.jitures and tax rates, the sign of the Gini coefficient for
tax rate should be reversed. The Dint coefficient for tax rates in this
case is then -0.004. If the two coefficients, for expenditures and tax
rate, are added together it should give a measure of the total inequality
in the system. In this case the sum of the two coefficients is exactly
zero, confirming the idea of a system that is completely neutral fiscally,
even though expenditures vary.
In Table 5 and Figure 5 the tax rates have been calculated assuming
instead that there is a state flat grant of $500 per student, with no
equalization. The tax rates have been calculated on the assumption that
all of the rest of the money must be raised by local taxes. Of course it
is highly unlikely, in such a situation, that District 10 would continue
to spend $1700 per student, taxing itself at 120 mills. But this has been
allowed to remain to illustrate the difference that the two approaches make
in tax rates when expenditures are not changed. The Gini coefficient for
tax rates in this case is 0.435, and the sum of the Gini coefficients for
expenditures and tax rate is 0.439.
(please see page 11)
11
Table 5
$500 Flat Grant
Dist. Studs.
Exp./Stud.
AV perStud.
TaxRate
Cumulative Percent ofStuds. Expends. Tax Rates
10 1000 1700 10,000 120 10 114 51
1 1000 800 20,000 15 20 20 58
9 500 1600 30,000 37 25 27 66
2 2000 900 40,000 10 45 42 74
8 500 1500 50,000 20 50 48 79
3 500 1000 60,000 8 55 52 80
7 2000 1400 70,000 13 75 75 91
4 500 1100 80,000 8 80 80 936 1000 1300 90,000 9 90 90 97
5 1000 1200 100,000 7 100 100 100
Figure 5
Percent 70
of60
Tax Rates
0 10 20 30 40 50 60 70 8o 90 100
Percent of Students JLZ!
12
It seems possible that the aggregate of decisions of the separate
school districts might be in a direction that would tend to minimize the
sum of the Gini coefficients for expenditures and tax rates. Continuing
with the assumption of a $500 flat grant per student, and using the same
type of table calculation as previously used, the following Gini
coefficients were obtained:
ExpendituresTaxRates
Expendituresplus Tax Rates
Assuming equal expendituresof $1200
(tax rates ranged from 7 to
0.000 0.373 0.373
70 mills)
Assuming equal tax rates of 0.158 0.000 0.15813 mills
(expenditures ranged from $630to $1800)
Assuming an in-between position(tax rates ranged from 10 to
0.104 0.127 0.231
20 mills, and expenditures from$700 to $1500)
The same thing was tried assuming that there was no state aid, and that
all income was raised through local taxes, with the following results:
ExpendituresTax
RatesExpenditures
plus Tax Rates
Assuming equal expenditures of 0.000 0.378 0.378$1200
(tax rates ranged from 12 to120 mills)
Assuming equal tax rates of 20mills
(expenditures ranged from $200to $2000)
0.267 0.000 0.267
Assuming an in-between position(tax rates ranged from 16 to 40
0.149 0.164 0.313
mills, and expenditures from$400 to $1600)
13
13
it appears from this very preliminary investigation that in an
unequalized system the sum of district decisions that results in least
variation in tax rates results in the smallest combined Gini coefficient.
In applying this technique, it is only appropriate to use assessed
valuation per student as an ordering device in states where there has been
almost complete equalization of assessment ratios statewide. Otherwise,
the appropriate measure to use is full value, or equalized value, per student;
the appropriate measure of tax effort, then, is tax rate on equalized value.
Some charts of expenditures per student and of tax rates for Oregon
districts are appended (Figures 7 to 9 ) as an illustration of the use of
the technique with real data. Assessments are carefully equalized in Oregon,
so that assessed valuation can confidently be used as a measure of fiscal
ability. One of the things that may immediately be noted is the nearness
of the Lorenz curves to the diagonal line, compared with the curve in
Figure I for family income in the United States. This is only partially a
result of the state equalization program in Oregon. It is also an artifact
of our method of measurement of educational expenditures which makes the
implicit assumption that all students in a particular school district have
exactly the same amount spent on them. The result of the aggregation is a
vast leveling of expenditure discrepancies. The analogous case is that of
constructing a Lorenz curve of income inequalities by region of the U.S.
If this is done for the four regions into which the country is divided by
the U.S. Statistical Abstract, with the assumption made that all families
within each region receive the same income, the results are as shown in
Table 6 and Figure 6. This Lorenz chart and Gini coefficient should be
compared with Figure I. Of course this aggregation error is not peculiar
to the Lorenz technique, but will affect aro/ measure of inequality one
1 4t
might choose to use.
Table 6
Family Income in the U.S., 1969, by Region
Median Cumulative Percent of
Families (000) Income Families Income
South 19,247 $ 8,105 30 26
Northeast 15,461 10,018 55 52
North Central 17,537 10,020 82 81
West 11,172 10,037 100 100
Percent
of
Income
100
90
80
70
60
50
40
30
20
10
Figure 6
0 10 20 30 40 50 60 70 80 90 100
Percent of Families
Gini = 0.038
14
sipaaM/S AO 11/338gd
N14x3 40 loNVN
* *
000Wt
1
to
/0
I
I
* /
0I
C°I)1(
IISO.° = Xl°14I INIt
* /
IC
0I)X
+I
* /
0I
4
I
_* /
0I
00004Z
3CINT1 v) s3HnliaN34x3 NOA VIVO
$ /
0+ 000114Z
1
* /
I
i
* v/
0i
CD'
-
I
*/
01
1
/C
1gri
I
*-/
0I
000 0
iI'
/0
M/
0 0
+00007.t
S
I
$/
0
1
$/
0
3I
*/
0
*/
0i
3
4/
0+ 000eJ,
I.
v/
0I
v
*/
0I
/0
1
*/
0I
X
*/
4/
00
_
+0000.9
1
*/
0I
*/
0 0
1.0
'/
04640
2 104 X11 oonq
In
*/
0/
01
1
/0
00401t 2 sqlvt1 XVI in 3nNvn
+ 0002L
N
0i
3
0I
3
SP = q0lUIRTO A0 Nnwon
Lbeo r. xlaNg 11119
Is
13
14
I(1 NI1 0) 011Yti XVI ainA viva
+ 00,104q
1
-,
1
n-116t 4,013rtil0I0 1n0032 Linn! !Malin
iiI
s11*:1 xvi (Div I111rilION.3X3
lYNOILTIHIGNI NI S7IIII1VOnJNI 40
3Abn3 zwAmol
+ J009.6
I
+4
++
4I.
I.
000041
0006'0
00000
000L40
000910
00000
0001/40
000E40
0102.0
0001'0
0003.0
C0i3X
L 311115IA
4,
ba1Ia
11
000940I1
.
_
3dX
1i
006069I
.
3II
-
4u11I
'*
1000Z4/
V
h3I1
iv
*/
/
31
0/
00
41
*/
0
3I
*/
0
d00000
*/
0
I/
0
I*
/0
I/
0
I*
/0
I*
00
000016
*0
10
4+
4
b1
a1
_
1000940
II
a1
.
1
3i
dX006069I
.
3II
-
41
u1I
'*
1000Z4/
V
hI
*/
31
iv
/
31
0/ 0 0
41
*/ 0
3I
*/ 0
d00000
*/ 0
I/ 0
I*
/ 0
I/ 0
I* / 0
I* 0 0
000016 * 0
10
4+
4
FIG
UR
E 8
KU()
0,00n0
0,1000
0,20u
0,3000
0,4400
0,5000
Otboou
0,700u
0,8000
+1-
4-
$I.
+t.
1+
p
0,9000
+
9,6000
4LORENZ CURVE OP INEQUALITIES IN INSTRUCTIONAL
EXPENDITURE; AND TAX RATES
00
I
0/
*
t
0/
*
IU4'1;tAX4 ELE.NTAItY PISTRIL15, 1971'12
0C
/*
I
0/
*
I
0/
*
13.0000
it
DATA FUR TAX net:3 to LINE)t
0 0
/*
P1
_.
..0
./ *
£ KI i
0/ *
GINI INDEX m 0,039
NUMICK OF DISTRICTS f 129
0/ *
C1
.0 0
/ 1
EI
0/ 4
N7,21)90
4RANGE OF TAX RAILS a se3,00
0/ *
7I
0.
./ *
.
I0
/ *
U1
NEAR TAX RATE = 413,76
/ *
FI
.0
/ *
I0
/ *
16,i:to°
4/ *
AI
0./ * _
.. _ _ _ .
. -
XI
/*
10 / *
KI
0 / .*.
._
AI
/*
-T
4;0400
+0
£I
S1
/ 0 *
I1 0 *
__
/*
I/
0
3.b000
#/
0
1_
_/
0_
_
1/ 0 0
:
1 i
/ 0 *
,
/ 0 *
-
!.4
ftli,4900
I 4.
/0
/ 0.*
DATA FOR EXPENDITURES (4 LINO;
1/ 0
xtlu
)
1 I
/ 0
/ 0
t;1t4I INDLX = 0,037
I/ 0
I0 0
1,2000
I.
0 *
RANGE, OF EXPENDLIUKES = 42817
I 1
0 *
_o *
10
'1E0 Lxibimoiruhe = $033
I0
I0
0'4,0030
++
+$
++
+t
+$
1,000u
+
01
0 9,6000
I I I I 1 6,4900
P
IE
IR
1C
1C
1(14
7,2000
T
! tn
1F
I IC
6,0000
X
IP
1E
IN
IP
1r
40000
T
1U
Ia
IE
I 1,6004
1 I I 1 I 2,4000
, I IX(10
)
I I ram
/ 1 i 1 I 0,0000
0,0000
0,1000
0.2C00
0,3000
0,4000
0,5000
0,6000
-
0,7000
0,8000
0,9000
1,0000
4.
__
X(10
.)
..
..
PERCENT OF STUDkNIQ
FIGURE
9
+X(10
)
0,0000
0,1000
0,2000
0.3000
0,9000
0,5000
0,6000
9,7000
0,R000
00000
1,0000
++
4+
++
+_
++
++
I0
1
9.6000 +
LORENZ CURVE OF INEQUALITIES 114 INSTRUCTIONAL
EXPENDITURE AND TAX RA7t3
0 * 9,6000
_I
0 *
1
t
0I
IunzotiN ELEMENTARY DISTmIcTO WITH roheR
TITAN 100 UMA.NTO, 1970.72
0*
I
I0
*I
I0
I
6;4000 +
DATA FUR TAX RAILS CO LINOS
0 0
*8,9000
P
Pi
0/
1E
f1
0/*
IR
RI
GINI INDEX = 0,059
NUMBER OF DISTRICTS s 50
0 /
*1
C
c1
.0/
*1
C
Ei
0 0 /
I%
N7,2000 +
RANGE OF TAX RA7E3 a 320,76
0/
*7,2000
T
TI
0/
*1
1
0/
AI
0
UI
MEAN TAX RATE = 39.46
/*
IF
FI
0 / .
Io
/*
*I
E
7V,uoGO +
0 /
6,0000
x
AI
0 /
*1
P
x1
a/
*I
c
I0
/*
I4
1:
I
0 /
*1
D
A1
0 /
*I
I,8
600
T
74,4000 +
0 /
*
E1
0/1(
1U
s1
0 0
/*
IR
100
/A
1E
I0
/*
1
I0
/*
I
3;6000 +
0/
*3.6000
I0_
._/
* «
I_
1/
*I
I0
/*
1
I0
/*
*I
10
/i
:4000 +
0/
*DATA FURixPER0iTOREs (* tame);
2,4000
I0
/.
*1
X +
I0
/*
*I
x(lu
)I
0/
*GINI INDLX = 0,152
ex(
10)
10
/*
i
I/
*1
1,2000 +
0 /
* *
RANGE OF EXPCNolluRc* x
02017
1,2000
10
/A
41
10
/*
I
1/ *
*MEA4 ExPLNDITuRe = $975
1
10 *
I
10 *
1
0,0000 +
++
++
++
++
..
0,01)00
0,0000
0,1000
0,2000
0,3000
0,4000
0,5000
0,600u
0,7000
0,6000
Ol0u00
1,0000
4.
______ _.
X(10 ..). .. ......_
PERCENT OF STOENTG
In summary, this paper proposes two modifications of the Lorenz
techn;que for measuring inequalities in financing education. One is
to order expenditures (and tax rates) by district fiscal ability before
calculating the cumulative percentages. It is possible, as a result,
to get a Lorenz curve that crosses the diagonal line, which gives
some useful insights. The other is to calculate Gini coefficients for
both expenditures and tax rates, and consider the sum of the two
coefficients to be a measure of the total inequality in the system.