WPS4245
A Poverty-Focused Evaluation of Commodity Tax Options·
B. Essama-Nssah
The difficulties faced by many developing countries in raising revenue from direct taxes
have forced them to rely heavily on indirect taxes to finance development interventions.
The purpose of this paper is to illustrate how to identify socially desirable options for
commodity taxation in the context of a poverty reduction strategy. Within the logic of
social evaluation the author assesses tax options on the basis of value judgments
underlying members of the additively separable class of poverty measures. The criterion
hinges on both the pattern of consumption of each commodity and the price elasticity of
the poverty measure used. An application of this methodology to data for Guinea shows
that many components of food expenditure (particularly cereals, grains and roots) would
be good candidates for exemption from VAT. Even though expenditure on health and
education is distributed in favor of the non-poor, their importance for human capital
development argues for a program of targeted subsidies in a broader context of cost
recovery.
Keywords: Poverty reduction, Commodity taxation, Social evaluation, Guinea
JEL Classification: D12, D63, H22, H31, I32, I38
World Bank Policy Research Working Paper 4245, June 2007
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the
exchange of ideas about development issues. An objective of the series is to get the findings out quickly,
even if the presentations are less than fully polished. The papers carry the names of the authors and should
be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely
those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors,
or the countries they represent. Policy Research Working Papers are available online at
http://econ.worldbank.org.
·Forthcoming in the Journal of International Development. The author is grateful to
David Coady, Jean-Yves Duclos, and Peter J. Lambert for useful comments on an earlier
version of this paper.
Poverty Reduction Group, MSN: MC4-415, World Bank,1818 H Street, NW,
Washington, D.C.20433, USA; Email: bessamanssah@worldbank.org; Telephone: (202)
473-7564; Fax: (202) 522-3283
1. Introduction
It is commonly believed that maintaining and improving the living standard of a
population is the ultimate objective of public policy and a fundamental expectation of the
governed (Sen et al. 1987). Every policymaker should therefore strive to make people
capable of choosing to live well. This Aristotelian view of the world underscores the
importance of public finance for development. Indeed, the pursuit of this basic policy
goal entails raising revenue and reallocating it to potential beneficiaries through public
spending. Furthermore, the way public spending is financed and allocated is bound to
affect the behavior of socioeconomic agents and the distribution of economic welfare
within society (World Bank 1988).
Developing countries confront a host of macroeconomic challenges in the design
and implementation of development strategies and policies. These challenges often stem
from a poor policy response to exogenous shocks or from careless fiscal and monetary
policies. In the 1980s and much of the 1990s, developing countries experienced a severe
economic crisis linked to the two oil price shocks of 1973 and 1979, and the sharp
decline in the world price of non-oil commodities between 1980 and 1987. These shocks
left many countries saddled with unsustainable imbalances in terms of current account
deficits, excessive debt and fiscal deficits. The crisis also brought fiscal adjustment to the
forefront of the development agenda. Such an adjustment entails a modification in the
way the public sector raises resources from the private sector through taxes or charges on
current or future economic activities, and redistributes these resources in the form of
public spending.
1
Ultimately, public spending is constrained by the ability of the tax system to
produce revenue beyond what may be available from other sources. When faced with
mounting budget deficits after significant expenditure cuts, many developing countries
have had to reform their tax systems to increase revenue. This invariably entails a
restructuring of the tax base and a modification of rates. This side of fiscal adjustment is
the focus of our paper with particular emphasis on the reform of commodity taxation.
This emphasis stems from the fact that developing countries have serious difficulties
raising revenue from direct taxes on individuals and firms due to the fact that many
potentially taxable transactions take place in a sizable informal sector of the economy.
These countries are thus forced to rely heavily on indirect taxes on goods and non-factor
services (such as sales, excise, and value added taxes)1. It is estimated that developing
countries derive about 40 percent of their tax revenue from indirect taxes and 25 percent
from income tax (Keen and Simone 2004). Over time, there has also been a significant
shift from trade to domestic taxes as a result of the process of trade liberalization.
Finally, in the 1990s, many developing countries adopted the value-added tax or VAT as
a preferred means to tax consumption.
The purpose of this paper is to illustrate how to identify socially desirable options
for commodity taxation in the context of a poverty reduction strategy. The perspective of
development as empowerment has prompted the international community to declare
poverty eradication a basic objective of development and therefore a benchmark measure
1 Indirect or commodity taxes generally include taxes on domestic production and consumption, and on
international trade. Production taxes are placed on goods before they enter the distribution chain while
consumption taxes are levied on the sale of final goods only (World Bank 1988). A pure consumption tax
is levied at the retail stage for all domestically consumed goods whether imported or produced locally. The
value added tax or VAT is applicable to all transactions within the production-distribution chain all the way
to the retail point. However, all intermediate purchasers can claim taxes paid against taxes due on sales.
This is what makes the VAT a commodity tax on consumption (the final purchaser has no way to
recuperate the amount paid in taxes).
2
of the performance of socioeconomic systems. In this context, poverty must be seen as
the deprivation of basic capabilities to lead the kind of life that one has reason to value
(Sen 1999). A well designed and implemented public finance policy can therefore be a
powerful tool for promoting poverty-reducing economic growth.
We frame our approach within the logic of social impact analysis understood as
an assessment of variations in individual and social welfare attributable to the
implementation of a policy. We maintain that, the observed poverty in a given society
depends fundamentally on individual behavior and endowments (internal e.g. will and
ability, and external such as ownership of assets and access to social support), and the
socio-political arrangements that govern social interaction. We can therefore restate the
focal policy problem as one of restructuring a system of indirect taxes to maintain or
increase yield, reduce distortions and minimize the burden on the poor. This is basically
a problem of optimal commodity taxation, the solution of which falls within the
normative approach to policymaking which seeks to maximize a social welfare function
subject to the economy's resource, technology and institutional constraints (Dixit 1996,
Diamond and Mirrlees 1971a, b).
To identify desirable changes from the status quo, we need to compute the
consequences of the policy reform and use explicit value judgments to assess these
consequences. This is the standard approach underlying the modern theory of tax reform
(Newbery and Stern 1987). Within this general framework, the computation of
individual gains or losses relies on two basic assumptions: (1) the principle of
optimization is used to characterize individual behavior, and (2) competitive markets
3
govern interactions2 among socioeconomic agents. The optimal behavior of a
socioeconomic agent is represented by a maximum value function (e.g. indirect utility)
indicating the maximum value of the (individual) objective function in terms of various
parameters underpinning the optimization problem (e.g. prices and the fixed budget).
One can then invoke the envelope theorem to calculate the impact of changes in these
parameters on the objective function (Dixit 1990).
In the public finance tradition (Drèze and Stern 1987, Newbery and Stern 1987,
Slemrod and Yitzhaki 2001, Yitzhaki 2003), social desirability of a policy is judged on
the basis of the variation of a social welfare function with respect to changes in individual
utility functions induced by changes in policy instruments. This function is usually
assumed to be a member of the Bergson-Samuelson family where social welfare is
defined in terms of individual (indirect) utility functions. Thus the social impact of a
policy change depends on the specification of this function on the basis of a set of value
judgments.
The poverty focus of this paper requires a social evaluation criterion that pays
special attention to the outcomes of the poor. We therefore base the assessment on
poverty measures satisfying Sen's focus axiom. These measures are invariant with
respect to the welfare of the non-poor (Sen 1997). In other words, evaluation is
conducted on basis of a truncated distribution of outcomes. Truncation occurs at the
poverty line. Our approach is consistent with Besley and Kanbur's (1988) analysis of the
poverty impact of a reallocation of food subsidies at the margin. This approach is also
similar to that of Bibi and Duclos (2007) or Son and Kakwani (2006). In particular it
2Within this setting, each person's claims to available goods and services are limited by the amount of
income obtainable by that person's successful sale of something of value on the market. Lindblom (2001)
calls this type of interaction, the rule of quid pro quo.
4
relies on the logic underlying a measure of the pro-poorness of economic growth
proposed by Essama-Nssah and Lambert (2006).
The outline of the paper is as follows. Section 2 presents the evaluative
framework. It explains the computation of changes in individual welfare attributable to
the tax reform, and the ranking of social states on the basis of poverty outcomes. Section
3 applies the method to data for Guinea. Concluding remarks are presented in Section 4.
2. Evaluation Framework
According to Sen (1999), any evaluative approach is characterized by its
informational basis. This entails a distinction between the information required in
passing judgments within the chosen approach and that which has no direct evaluative
role. Our approach is essentially welfarist to the extent that it focuses on both individual
and social welfare. Policy evaluation therefore requires the identification of impact on
individuals as well as the specification of an aggregation rule to translate individual
outcomes into social impact.
Individual Outcomes
Assessing the impact of policies on individual welfare and poverty status requires
an explicit linkage between policy instruments and the determinants of individual living
standard. In particular, commodity taxes affect prices at which such commodities are
sold. We therefore need an expression of individual (or household) welfare as a function
of relevant prices. Under the simple assumption that the consumer has an exogenous
budget x to spend on a set of m commodities at fixed prices, the indirect utility is the
maximum level of utility attainable given the outlay and the prevailing prices. This
5
function can be formally written as (Deaton and Muellbauer 1980):
v(x, p) = max[u(q); p q = x] , where q is an m-dimensional vector of quantities of
q
commodities and p the corresponding vector of consumer prices. We assume that
producer prices are not affected by the reform. They can therefore be normalized to unity
so that we express the consumer price of commodity k as a function of the producer price
and the applicable tax (or subsidy)3 rate tk. That is: pk = 1+ tk ; k = 1,2, ..., m.
We can invoke Roy's identity to compute the welfare impact of a marginal
change in the price of commodity k induced by a marginal change in the applicable tax.
According to this identity, the Marshallian demand function of a commodity is equal to
the negative of the first-order derivative of the indirect utility function, divided by the
marginal utility of income. Formally we write.
qk (x, p) = - v/tk = - v/pk (2.1)
v/x v/x
This expression implies that d*(x,p) = qk(x,p)dpk measures the welfare loss (in
monetary terms) induced by a marginal increase4 (dtk ) in the price of commodity k. In
this setting, note that dpk = dtk .
To assess the poverty implications of this price change, we need to define a
measure of poverty at the individual or household level (depending of the unit of
analysis). Let z be the poverty line and (x|z) an indicator of poverty (or deprivation) for
an individual with total level of expenditure (or real income) equal x. Following
Kakwani (1999), we assume that: (i) the poverty indicator is equal to zero when x is
3In this context, a subsidy is considered a negative tax on the relevant commodity.
4A rise in a subsidy would be associated with a welfare gain. Also, *(x,p) is equal to the indirect utility
normalized by the marginal utility of income.
6
greater or equal to the poverty line; (ii) the indicator is a decreasing convex function of x,
given z. Thus the first-order derivative with respect to x is negative, given z. In other
words, (x | z) = (x | z) < 0.
x
The observed demand functions are a consequence of optimal behavior. Hence,
the level of expenditure x represents the minimum required to attain utility *(x,p). Thus
when the price of commodity k increases by a small amount, the consumer faces a loss of
real income dx = qk(x,p)dpk . We compute the corresponding marginal poverty increase
as follows.
(x | z) = (x | z) x (2.2)
pk x pk = - (x | z)qk
In the case of a subsidy, the marginal reduction in poverty is equal to the absolute value
of expression (2.2).
Aggregation and Ranking
The ranking of potential tax options is based on the impact of the tax reform on
the overall poverty. We focus here on members of the class of additively separable
poverty measures defined by the following expression, where f(x) stands for the density
function of real income x within the population.
= (x | z) f (x)dx
z
(2.3)
0
Members of this class are additively separable in the sense that the deprivation felt by an
individual depends only on a fixed poverty line and the individual economic welfare not
on the welfare experienced by others in society. They include such poverty indices as the
Foster Greer Thorbecke (1984) family and the Watts measure. The Foster Greer and
Thorbecke (FGT) family is characterized by the following deprivation function:
7
(x | z) ={max[(1- x/ z),0]} . When is equal to zero, one or two, equation (2.3)
leads respectively to the headcount ratio (H), the normalized poverty deficit (or poverty
gap), and the squared poverty gap. The deprivation function underlying the Watts index
is similarly defined: (x | z) = maxln , 0 .
z
x
Equations (2.2) and (2.3) imply that the change in overall poverty associated with
a marginal increase in the price of commodity k is equal to:
z
(2.4)
pk = - qk (x | z) f (x)dx
0
It is more convenient to couch the analysis in terms of elasticity. Thus, we write the price
elasticity of poverty for commodity k as:
k() = pk = - 1 z
(2.5)
pk pkqk (x | z) f (x)dx
0
Furthermore, the price elasticity of real income with respect to commodity k is equal to
the negative of the budget share of commodity k (Deaton and Muellbauer 1980).
x pk = - pkqk (2.6)
pk = -wk(x)
x x
This expression means that, when the price of commodity k increases by 1 percent, real
income x declines by an amount equal to the corresponding budget share, wk(x).
Equation (2.5) is therefore equivalent to the following.
k() = - xwk(x)(x | z)f (x)dx
1 z
(2.7)
0
This elasticity is positive for a price increase and negative for a price reduction. It is a
weighted sum of individual budget shares (interpreted as price elasticities of real
8
income5). Whether or not a change in the price of commodity k is favorable to the poor
(i.e. pro-poor) depends on the distribution of these budget shares over the entire
population.
We need a decision rule for declaring a price change pro-poor. For our purpose,
an increase (reduction) in the price of commodity k is considered pro-poor if it induces
an absolute increase (reduction) in poverty smaller (greater) than it would in a benchmark
case6. As a benchmark we select a situation where a one percent change in the price of k
would have the same relative impact on real income x. We construct such a hypothetical
by assuming that everybody assigns the same proportion of real income, w0k, to the
purchase of commodity k7. This constant budget share can be computed as the average
budget share for commodity k. That is:
mx pkqk (x) f (x)dx mxxwk (x) f (x)dx
w0k = 0 = 0 (2.8)
mx mx
xf (x)dx xf (x)dx
0 0
where mx represents the highest level of real income in the population. The
corresponding price elasticity of poverty is:
0k() = - w0k z
(2.9)
x (x | z) f (x)dx
0
Since budget shares do add up to unity, the price elasticity defined by equation (2.9) can
be rewritten as: 0k () = - w0k
z m m
() = w0k() ,
0x wk (x) (x | z) f (x)dx = w0k
k
k=1 k=1
5 Interestingly, this price elasticity is analogous to the growth elasticity of poverty proposed by Essama-
Nssah and Lambert (2006) in the context of measuring the pro-poorness of economic growth.
6The choice of a benchmark is analogous to selecting a level of significance in statistical analysis.
7 This idea is similar to the use of a Lorenz curve to assess inequality. Indeed, within the Lorenz
framework, the amount of observed inequality is inferred from a comparison of the actual distribution of
economic welfare (as represented by income or expenditure) with a hypothetical distribution where
everybody would have equal share.
9
m
where () = k () is the percentage change in poverty induced by a marginal
k=1
increase in all prices (Son and Kakwani 2006).
A 1 percent increase in the price of commodity k would increase poverty by k
under the observed distribution of budget shares, and by 0k in the benchmark case. To
declare the price increase pro-poor one can make either an additive comparison or a
comparison in ratio form. Consider the ratio form first. This leads to the following
indicator.
zxwk (x) (x | z) f (x)dx z xk (x) (x | z) f (x)dx
k() =
0 = 0 (2.10)
z z
w0k x (x | z) f (x)dx x (x | z) f (x)dx
0 0
Where k (x) = wk (x) is the budget share of commodity k at income level x normalized
w0 k
by the aggregate budget share. The ratio measure of pro-poorness is thus a weighted
average of normalized budget shares among the poor (i.e. along the distribution of real
income up to the poverty line). An increase in the price of k would be considered pro-
poor if this indicator is less than one. In other words, an increase in the price of
commodity k would hurt the poor less than the non-poor. Commodities for which this
indicator is greater than one are good candidates for subsidization.
It is instructive to consider the additive comparison based on the following
measure.
k() =(k -0k) = - x(x | z)[wk(x)- w0k]f (x)dx
z
(2.11)
0
This additive measure of pro-poorness is a weighted sum of the deviation of budget
shares for commodity k from their mean, up to the poverty line. When this measure is
10
negative, an increase in the price of commodity k would be pro-poor (i.e. it would hurt
the non-poor more than the poor). In other words, k () < 0. A poverty-focused fiscal
reform would not advocate a subsidy on such a commodity. Expression (2.11) also
reveals that wk (x) > w0 x < z is sufficient for an increase in the price of commodity k
k
to hurt the poor more than the non-poor, regardless of the poverty measure used among
those defined in (2.3).
We now consider expressions for the price elasticity and measures of pro-
poorness for specific poverty measures that we will use in the empirical section. The
headcount ratio is equal to the cumulative distribution function of x evaluated at the
poverty line. That is H=F(z). The price elasticity of the headcount can be computed in
terms of the equivalent variation in the poverty line as follows.
k(H) = - H z pH k = f (z) pkqk (z)= zf (z)wk (z) (2.12)
z pk H H
This expression implies that when all prices increase by one percent, the headcount ratio
increases by (H) = zf (z)
. The corresponding ratio measure of pro-poorness is equal
H
to:
k(H) = wk (z) = k(z) (2.13)
w0k
While the additive measure of pro-poorness is equal to the following expression:
k(H) = zf (z)[wk(z)- w0 ] (2.14)
k
When using the headcount ratio to assess the pro-poorness of a price change, expressions
(2.13) and (2.14) make it clear that such a judgment is based only on the budget shares of
individuals located at the poverty line. This indicator does not use available information
11
efficiently to the extent that it does not take into account the pattern of consumption of
commodity k among the poor.
For the Watts index of poverty, the price elasticity is equal to:
k(W ) = 1 z
W wk (x) f (x)dx (2.15)
0
The corresponding additive measure of pro-poorness is:
k(W ) = [wk(x)- w0k]f (x)dx
z
(2.16)
0
Thus a reduction in the price of commodity k (e.g. a subsidy) would be pro-poor if the
average budget share for the poor is greater than the overall average budget share. That
1 z
is: wk (x) f (x)dx > w0k . Note that the multiplicative measure of pro-poorness for
H
0
the Watts index is equal to:
k(W ) = 1 zwk (x) f (x)dx = 1 zk(x)f (x)dx (2.17)
w0k H
0 H 0
Generally, when 1, the price elasticity of poverty associated with members of
the FGT family is equal to8:
x1- x -1
k( ) = z z z
wk (x) f (x)dx (2.18)
0
This expression leads to the following measures of pro-poorness:
8 Given that x/z=[1-(1-x/z)], this elasticity can also be written as:
k( ) = z x -1 z x
This decomposition
0 wk (x)1- f (x)dx - wk (x)1- f (x)dx .
z 0 z
implies that the overall impact associated with the same change in all prices can simply be written as
m
follows: ( ) = ( ) = [-1 - ] (Son and Kakwani 2006).
k
k=1
12
-1
z k(x)f (x)dx
0
k() = x1- xz
z (2.19)
-1
z
0x1- xz
z f (x)dx
and
-1
k( ) = z [wk (x) - w0 ] f (x)dx (2.20)
0 x1- xz
z k
The Social Cost of Public Funds
In general, the concept of social cost of public funds establishes a relationship
between the funds collected through taxation and the change in social welfare induced by
the collection process. In the context of our poverty-focused framework, the poverty cost
per additional dollar raised by increasing the tax on commodity k is defined by the
following expression.
k() = /tk (2.21)
R/tk
where R stands for government revenue from commodity taxation. We refer to this
indicator as the marginal poverty cost of public funds. Given two commodities j and k
such that j()