Measurement of Tax Progressivity: An International Comparison

Author(s): Nanak C. Kakwani

Source: The Economic Journal, Vol. 87, No. 345 (Mar., 1977), pp. 71-80
Published by: Wiley on behalf of the Royal Economic Society
Stable URL: http://www.jstor.org/stable/2231833
Accessed: 19-07-2016 08:51 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].

Wiley, Royal Economic Society are collaborating with JSTOR to digitize, preserve and extend access to
The Economic Journal

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
The Economic Journal, 87 (March 1976), 7I-80

Printed in Great Britain

MEASUREMENT OF TAX PROGRESSIVITY:

AN INTERNATIONAL COMPARISON1

INTRODUCTION

This paper focuses on the problems of measurement of progressivity in taxation

and public expenditure. The problem has attracted the attention of several
writers. Among the important contributions to this field are those of Dalton
(1955), Slitor (I948), Musgrave and Thin (I948) and Bracewell-Milnes
(97I). The major weakness of the existing literature is that it does not dis-
tinguish between the effects of changes in average tax rates and in pro-
gressivity on the distribution of income. One of the aims of this paper is to show
how the distributional effect of taxation is influenced by changes in the average
tax rate while progressivity is held constant and vice versa. Further the method
enables one to study the contribution of individual taxes and expenditure
items to the overall progressivity of the fiscal system of taxation and public expen-
diture. Finally, the methods developed are applied to the empirical investigation
of fiscal operations in four developed countries - Australia, Canada, the United
Kingdom and the United States.
Section 2 relates to the definition and measurement of tax progressivity.
The major theoretical contribution of this paper is contained in Section 3.
Section 4 and Section 5 summarise the important empirical findings based on
the theoretical measures developed in Section 3.

2. MEASUREMENT OF TAX PROGRESSIVITY

Let T(x) be the tax paid by an individual with income x; the tax system is
proportional when the elasticity of T with respect to x is equal to one for all x.
The tax system is progressive when the elasticity exceeds one and regressive
when the elasticity is less than one. It can be shown that this definition of
progressivity is equivalent to saying that a tax system is progressive, propor-
tional and regressive when the marginal tax rate is greater, equal and less than
the average tax rate, respectively. Using this definition Slitor (I948) proposed
the following measure of progression:
dt(x) _ m(x) -t(X) (2.I)
dx x

where t(x) is the average tax rate at the income level x and m(x) is the marginal
tax tate at the same level of income.
The above measure relates to progression at a given point in the income
scale and therefore does not give an unambiguous index of tax progressivity.
I This paper was written while the author was on the staff of the Development Research Center
World Bank, Washington, D.C. I would like to acknowledge the help given by MrJ. N. Sharma in
drafting and analysis of the numerical results. Malathi Parathasarthy did all the calculations. The
comments from a referee were useful. Responsibility for errors is mine alone.
[ 7' ]

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
 72 THE ECONOMIC JOURNAL [MARCH

In order to arrive at a single measure of progression, Musgrave and Thin

(I 948) compared the inequality of the before-tax and after-tax income distribu-
tions. A progressive tax system is associated with a decrease in income in-
equality, while regressive tax rates will be reflected by an increase in income
inequality. When the income inequalities of the before-tax and after-tax
incomes are identical, then the tax system will be proportional. Thus if the
Gini index is used to measure inequality, the difference, or ratio, of the Gini
indices of the before-tax and after-tax incomes provides a single measure of tax
progressivity.1
The Musgrave-Thin measure of tax progressivity has been widely used. It
appears intuitively appealing but we demonstrate in the next section that by
simply comparing the inequality of pre-tax and post-tax incomes one cannot
arrive at a suitable measure of progressivity.

3. A NEW MEASURE OF TAX PROGRESSIVITY

Suppose income x of an individual is a random variable with mean ,u and

probability distribution function F(x). If F1(x) is the proportion of income of
units having income less than or equal to x, then the relationship between F(x)
and F1(x) is called the Lorenz curve of income x. The Gini index of income is
defined as one minus twice the area under the Lorenz curve.
Let us denote Fl[(T(x)] to be the proportion of taxes paid by the units
having income less than or equal to x. Then the relationship between F(x) and
FJ[T(x)] will be called the concentration curve of taxes (see Kakwani, forth-
coming). The concentration index is equal to one minus twice the area under
the concentration curve.
The concept of " tax progressivity" is related to the concept of "tax elas-
ticity", which is the elasticity of the tax function T(x) with respect to income
x. A measure of tax progressivity shows the deviation of a given tax system
from proportionality. Since the tax elasticity is always unity for proportional
taxes a measure of tax progressivity should depend on the magnitude of the
difference of the tax elasticity from unity.
The distance between the curves F1(x) and FJ[T(x)] depends on tax elas-
ticity. If tax elasticity is unity at all income levels, the two curves coincide. It
follows from Kakwani (forthcoming) that the larger the difference of the tax
elasticity from unity, the greater is the distance between F1(x) and F1[T(x)].
This suggests that a suitable measure of tax progressivity can be derived by
comparing the Lorenz curve of income and the concentration curve of taxes.
If we denote C to be the concentration index of taxes and G the Gini index of

1 The Gini index is the most widely used inequality measure. The measure has recently been criticised
by several authors. Their criticism is mainly based on the fact that the welfare function implied by the
Gini index does not have desirable properties such as additivity and strict concavity (see Atkinson
(1970), Newbery (1970), Das Gupta, Sen and Starrett (I973) and Rothschild and Stiglitz (1973)).
Sen (1974) has recently provided a complete axiomatisation of the Gini index as a measure of inequality.
In this paper we use only the Gini index to measure the distributional effects of taxation. The use of in-
equality measures based on the welfare function would considerably complicate the analysis.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
I977] MEASUREMENT OF TAX PROGRESSIVITY 73

the before-tax income, then we propose P = (C - G) to be a suitable measure

of tax progressivity. It can be seen that P is equal to twice the area between
the curves Fl(x) and F1[T(x)]. P is positive (negative) if the tax elasticity is
greater (less) than unity for all x. P assumes value zero when the tax elasticity is
unity for all incomes. A positive value of P implies a progressive tax system and
vice versa. P increases (decreases) with the increase (decrease) in tax elasticity at
all income levels.
The disposable income of a unit with income x is given by

d(x) = x-T(x), (3.I)

which is a non-decreasing function of x if the marginal tax rate is less than
unity for all incomes. Then it can be shown, using the results of Kakwani
(forthcoming) that

G* = G- tP (3.2)

where G and G* are the Gini indices of before and after-tax income respectively
and t is the average tax rate.
The equation (3.2) shows that the after-tax income inequality is a function of
three quantities, namely the inequality of income before tax, the average tax
rate and tax progressivity. Note that the average tax rate can change without
changing tax progressivity. Therefore one can write the elasticities of the Gini
index of the after-tax income with respect to the average tax rate and progres-
sivity as
Pt
Yt = ( I-t)2G*
and
Pt
= P (I-t) G*
respectively.
The equations (3.3) and (3.4) show that for a progressive (regressive)
tax system, the income inequality decreases (increases) with increases in
t and P. Further, note that the ratio of the two elasticities, It and yp, exceeds
unity in absolute terms. From this we conclude that after-tax income inequality
is more sensitive to the average tax rate than to tax progressivity.
Differentiating (3.2) totally we obtain
dG* G dG dP dt
G* G* =
G #P +5
p tt' t(3.5)
which gives a decomposition of the percentage change in the after-tax income
inequality in terms of the percentage changes in the before-tax income in-
equality, tax progressivity and the average tax rate.
The Musgrave-Thin measure of progressivity is given by (G - G*) which
will be identical to our measure of progressivity P if and only if the average
tax rate of the society is os. It can be seen from (3.3) that the larger (smaller)
the average tax rate for a given tax elasticity or progressivity, the more equal

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
 74 THE ECONOMIC JOURNAL [MARCH

(unequal) will be the post-tax income distribution. If we assume that the

change in the average tax rate does not affect the before-tax income distribution,
the Musgrave-Thin measure may show an increase or decrease in progressivity
even if the tax elasticity has remained constant at all income levels.
This implies that by doubling the tax rates at all income levels the tax
progressivity will increase, according to Musgrave and Thin. This is obviously
unacceptable because progressivity (or regressivity) is supposed to measure the
deviation of a tax system from proportionality. Musgrave and Thin are in fact
measuring the redistributive effects of taxes, which have been shown to be a
function of average tax rate in addition to progressivity.
Now suppose that the total tax function T(x) is equal to the sum of n individual
taxes T1(x), T2(x), ..., T.(x), then using Theorem 4 of Kakwani (forth-
coming) we obtain
n

C Ci) (3.6)

where Ci is the concentration index of the ith tax and ti is the average rate of the
ith tax. From (3.6) it follows that
n

-
it Pi (3*7)
where Pi is the progressivity of the ith tax. This equation shows that the pro-
gressivity of all the taxes together is equal to the weighted average of the
progressivity of individual taxes, the weights being proportional to their
average rates. Equation (3.7) can be used to analyse the percentage contribu-
tion of each tax to the total tax progressivity.
Using (3.7) in (3.2) gives
n
E Piti
G* -G t___
G (i -t) G' (3.8)
which provides the decomposition of the percentage change in inequality
(due to the tax system as a whole) in terms of the contribution of each tax.
If Pi is positive, the contribution of the ith tax is negative and if Pi is negative,
the contribution is positive.

4. INTER-COUNTRY COMPARISON OF FACTORS AFFECTING

INEQUALITIES OF DISPOSABLE INCOMES

In this section we discuss changes in degrees of inequalities of income after tax

in four countries, namely Australia, Canada, the United Kingdom and the
United States. Here after-tax income inequality refers to the inequality of
disposable incomes. The discussion is limited to the effects of personal income
taxes only. In the following section the scope of the discussion is widened to
take into account the effects of all types of taxes (including indirect taxes) and
public expenditures on overall inequalities.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
1977] MEASUREMENT OF TAX PROGRESSIVITY 75

The data for this section were obtained from the official income-tax statistics
of different countries. The specific publications are listed as follows:
(i) Australian Taxation Statistics (Taxation Office), I96I-2 to 197I-2.
(2) Canadian Department of National Revenue, Taxation Division. Taxation
Statistics, I966-72.
(3) Great Britain Board of Inland Revenue, Report of the Commissioners of Her
Majesty's Inland Revenue, 1958- 9 to I 966-7.
(4) U.S. Internal Revenue Service, Statistics of Income: Individual Income
Tax Returns for I958-I970 inclusive.
The published data are available in grouped form. The grouping was done
on the basis of size of income before tax. For each income class, the source
publications reported data on the number of tax payers, the total income and
the total income tax paid.
For the purpose of calculating tax progressivity it was necessary to compute
the Gini index of income before tax and the concentration index of the total
amounts of personal income tax paid. The equation of the Lorenz curve
proposed recently by Kakwani and Podder (1976) was fitted to income and
taxes-paid separately. The concentration indices were then computed by
integrating the area under the concentration curve.
Before we present the numerical results, it is desirable to point out the
several difficulties associated with the comparison of tax progressivity using
the taxation statistics of the different countries. First of all these data refer to
earners or individual income recipients, not to families. Secondly, not all
those who receive income are required to file returns. The proportion of
earners not filing tax returns depends on the exemptions allowed and, therefore,
may vary from one country to another. Thirdly, each country has a different
definition of income. Fourthly, the inaccuracies in the data due to tax evasion
and tax avoidance may not be of the same magnitude in different countries.
The numerical results on tax progressivity are presented in Table I.
The first column in the table gives the Gini index of before-tax income. The
second and the third columns present the tax progressivity index P and the
average tax rate t, respectively. The after-tax Gini index is computed in the
column using the formula (3.2). Elasticities of the after-tax Gini index are
presented in the fifth and sixth columns. Columns 7-IO provide the decomposi-
tion of the percentage change in after-tax income inequality.
Several interesting findings emerge from the numerical results given in the
first table.'
The inter-country comparison shows that there are relatively small differ-
ences in the degrees of income inequality both before and after tax except that
the degree of income inequality is markedly higher in the U.S.A. than in the
other countries. The degree of tax progressivity shows more variation; ranking

1 The empirical findings in the next paragraph are based on the numerical results computed for
about ten years for each country. In Table I we have reported results for only three years in order to
save space.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
76 THE ECONOMIC JOURNAL [MARCH

the countries in descending order of progressivity, the United Kingdom comes

on the top followed by Australia, Canada and the U.S.A. The inter-country
differences in the average tax rate are of a similar size to the year-to-year
variations within each country.
The empirical results bring out clearly the appreciable declines in the
degree of tax progressivity in all countries, the fall being greatest in the
United Kingdom. This decrease in the degree of tax progressivity would
have accentuated the income inequalities in different countries but its effect
was offset by the upward trend in average tax rates. The relative contributions
of changes in different factors to the changes in after-tax income inequality are
brought out in columns 8-io.

5. DISTRIBUTIONAL EFFECTS OF TAXES AND

PUBLIC EXPENDITURES

This section presents our findings with regard to the breakdown of the distribu-
tional effects of taxation and public spending into their individual components.
The data for this section were obtained from different studies by various
authors of the incidence of taxes and expenditures. As such, inter-country
comparisons are rendered difficult because of differences in concepts, methods
and assumptions employed by different authors. Therefore our comments on
inter-country differences should be qualified. Space limitations do not allow
discussion of differences in concepts, methods and assumptions. Interested
readers are referred to the original studies. The Australian data are from the
study by Bentley, Collins and Drane (I974). The source of data for Canada is
the study by Dodge (I975). Finally, United States data were obtained from
two studies conducted by the Tax Foundation (I967) and by Reynolds and
Smolensky (I974).
In this section we only summarise our empirical findings. The detailed
computations are reported in tables 5 to 9 of a discussion paper by the author
(Kakwani, I976). Some salient figures from these tables are, however, pre-
sented in Table 2, which has the same format. The first column gives the
percentages of each individual tax or expenditure item in total revenue or total
expenditure, respectively. The second and third columns give concentration
indices (C) and progressivity indices (P), respectively. The fourth column
contains the relative contributions of individual items to total progressivity in
percentage terms. The last two columns summarise the distributive effects of
individual taxes and expenditure items.
In Australia income tax is substantially more progressive than the tax
system as a whole, and it reduces inequality by 6-9 % while the net effect of
all other taxes is to increase it by I -4 %. The most progressive tax is that on
estates and gifts, but since they account for less than I % of total tax receipts
their effect on income inequality is relatively small. The main regressive element
is the indirect tax system, which increases the income inequality by about 4 %.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
1977] MEASUREMENT OF TAX PROGRESSIVITY 77

U.K

U.SA

Tablei
CAND

AUSTRLI

ProgesivtyIndxfalcmT

%change

I970'32O561-4?8 I960-45372o8iO I9680'435o-217.O I96/7O?345-20o81 I965/0-38o2714O. I964/503o-28iO?7'1. I970O354oi62-18.? I960-35741O2o.8 I9680O357-21oi I971/20O35-i8V64?o I970/O352-486 i96/70O351-2I48o 19680-3542OI_'i.s?7o
(I)2345678910
Yearinc.dx(P)tG*V71bfosvy ofber-taxPgsivyGndpu

GindexAvrag%chuto

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
 78 THE ECONOMIC JOURNAL [MARCH

It is noteworthy that local-government taxes as a whole are significantly

regressive, being second in importance only to the indirect taxes. A further
break-down of overall distributional effects of indirect taxes shows that every
individual commodity tax is regressive. This is an important finding because it
casts doubt on the commonly held belief that indirect taxes can be made
progressive by making a distinction between luxuries and necessities.
The empirical results relating to Canada indicate that the effects of fiscal
policy at both levels of government are favourable to the goal of reducing
income inequality. The significant difference shows up in regard to magnitude
of effects of public expenditures and taxes at two levels of government. Fiscal
operations at the federal level are more progressive than those at the State and
Local level both in regard to taxes and expenditures. It is important to note
that expenditures at both levels of government are more redistributive than
taxes.
The distributional effect of the fiscal operations in the U.S.A. was investigated
for two years, namely I96I and I970. The detailed empirical results brought
out very important features of the fiscal system. While the Federal taxes as a
whole contributed to a reduction in income inequality, the system of taxation
prevailing at the levels of state and local governments resulted in increased
income inequality in both the years studied. The public expenditures at the
Federal level as well as State and Local levels contributed to a reduction in
income inequality both in I96I and I970. Comparing the numerical values
in the two years we note that fiscal operations at the Federal levels were less
progressive in I970 than in I96I both in respect of taxes and in respect of
expenditure. The change during the same period was in the opposite direction
with regard to fiscal operations at the level of State and Local governments.
While taxes became less regressive expenditure became more progressive. The
conclusion that emerges from these changes in the nature of fiscal operations
implies that while the Federal government deviated away from egalitarianism,
the State and Local governments edged closer to egalitarian policies. The
change is most striking in respect of personal income at the Federal level. Its
contribution to reduction in income inequality declined from 595 % to
2-92 %. The net effect of fiscal operations at both levels of government taken
together was less progressive in I970 than I96I. During this period the contribu-
tion of fiscal operations at all levels of government taken together to reduction
in inequality declined from 34 % to little over 27 %-
Table 2 enables us to make immediate international comparisons of the-
nature of fiscal operations at two levels of government. It can be seen that
the tax system as a whole is markedly more progressive in Australia than in
Canada and the U.S.A. Both in Australia and in the U.S.A. the taxes at State
and Local level are regressive whereas in Canada they are progressive. Com-
paring Canada and the U.S.A. we observe relatively small differences in the
degree of progressivity of government expenditure at the Federal level as well
as State and Local level.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
I977] MEASUREMENT OF TAX PROGRESSIVITY 79

Table2
U.SAI970
CAND1970

AUSTRLI96-7

DistrbuonalEfecTypxdP

Netfcoaxsndpiur-827*36 TotalGvernmxpdiuIO-01429357 StaendLoclxpiur42'370-1689I. Fedralxpnitu57-63o82014 TotalGvernmxsI0.417O-986+ StaendLoclxs36-70495+i Fedraltxs63240io5g-OI Netfcoaxsndpiur-0132&8 TotalGv.expndiurIO06-289357 StaendLoclxpiur47'20-381I659 Fedralxpnitu52-80'164o7 TotalGvernmxsIO-038921'5 StaendLoclxs50-2374681QI9 Fedraltxs49f80o7-1I?52.3 TotalGvernmxsI-34809175 StaendLoclxsI7-8602951+3 Fedraltxs82-40361?57i9OI
(I)234567
Typesoftaxndiurgvql
%contribu
taxesorCnciTdp-l%hg %oftalcnribueqydxph ChangeioftxsGv.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms
 80 THE ECONOMIC JOURNAL [MARCH 1977]

6. CONCLUSIONS

In this paper a new measure of tax progressivity has been introduced. This
measure helps to distinguish between different effects of taxation on income
distribution. In particular, it has helped to show clearly that the reduction in
income distribution resulting from taxation depends not only on the degree of
tax progressivity, as is commonly believed, but also on average tax rate. We
have also analysed the effects of various taxes and items of public expenditure at
different levels of Government. This empirical analysis has been done for four
developed countries, namely Australia, Canada, the United Kingdom and the
U.S.A. It would be interesting for future research to extend the empirical
analysis to other developed and developing countries.
NANAK C. KAKWANI

University of New South Wales

Date of receipt of final typescript: August 1976

REFERENCES

Atkinson, A. B. (I970). "On the Measurement of Inequality." Journal of Economic Theory, vol. 2.
Bentley, P., Collins, D. J. and Drane N. T. (I974). "Incidence of Australian Taxation." December.
Bracewell-Milnes, B. (I 97 I). The Measurement of Fiscal Policy. London: Confederation of British Industry.
Dalton, H. (1955). Principles of Public Finance. New York: Frederick A. Praeger Inc.
Das Gupta, P., Sen, A. K. and Starrett, D. (1973). "Notes on the Measurement of Inequality."
Journal of Economic Theory, vol. 6.
Dodge, D. A. (I975). "Impact of Tax, Transfer, and Expenditure Policies of Government on the
Distribution of Personal Income in Canada." Review of Income and Wealth.
Kakwani, N. C. (forthcoming). "Applications of Lorenz Curves in Economic Analysis." Econometrica.
(1976). "Measurement of Tax Progressivity: An International Comparison." Discussion Paper
no. 17, School of Economics, the University of New South Wales, March.'
and Podder, N. (1976). "Efficient Estimation of the Lorenz Curve and Associated Inequality
Measures from Grouped Observations." Econometrica, January.
Musgrave, R. A. and Thin, T. (I 948). " Income Tax Progression I929-48." Journal of Political Economy,
vol. 56 (December), pp. 498-5I4.
Newbery, D. M. G. (1970). "A Theorem on the Measurement of Inequality." Journal of Economic
Theory, vol. 2.
Reynolds, M. and Smolensky, E. (I974). "The Post Fisc. Distribution: I96I and 1970 Compared."
Institute for Research on Poverty, Discussion Paper no. 19 I 74, University of Wisconsin, Madison.
Rothschild, M. and Stiglitz, J. E. (I973). "Some Further Results on the Measurement of Inequality."
Journal of Economic Theory, vol. 6.
Sen, A. K. (1974). "Informational Bases of Welfare Approaches." Journal of Public Economics, vol. 3.
Slitor, R. E. (1948). "The Measurement of Progressivity and Built-in Flexibility." Quarterly Journal of
Economics, vol. 62 (February), pp. 309-13.
Tax Foundation Inc. (1967). "Tax Burdens and Benefits of Government Expenditure by Income
Classes, I96I and I965." Tax Foundation Inc., 50 Rockefeller Plaza, New York.

I This discussion paper is available from the author on request.

This content downloaded from 147.96.1.236 on Tue, 19 Jul 2016 08:51:53 UTC
All use subject to http://about.jstor.org/terms