Elasticity and buoyancy of the tax system in Pakistan.
Bilquees, Faiz
This paper examines the elasticity and buoyancy of the tax system
for the period 1974-75-2003-04. The elasticity of the total tax revenue
both with respect to the total GDP and the non-agricultural GDP base is
less than unity. Overall, sales tax takes the lead by way of improving
revenues. The high coefficient of income tax inclusive of withholding
tax, which is an indirect tax, is high. Excluding the withholding tax
leads to a lower coefficient. Sales tax with respect to imports and
manufacturing also takes care of loss of revenue due to lowering of
tariff and excise duties. However, the sales tax coefficient with
respect to the GDP base reflects the inclusion of service sector and
utilities in the sales tax net, which has serious implications for the
poor. The estimates of buoyancy suggest that tax changes did not lead to
significant revenue augmentation. The low buoyancy of income tax
exclusive of the withholding taxes implies that imposition of massive
withholding taxes coupled with an increase in the taxable income limits
is working at cross purposes.
INTRODUCTION
Tax responsiveness to changes in income is a crucial variable in
projecting the tax revenues, and is a basic criterion for a good tax
system. This response is measured by two concepts: tax elasticity which
measures the automatic response of revenue to income changes, net of
discretionary changes; and tax buoyancy which measures the total
response of tax revenue to changes in income. In developing countries
generally, the major taxes tend to have low elasticity and sometimes
even the buoyancy is low. This is mainly due to the inherent weaknesses
in economic structure where a large majority remains out of the tax net
due to low average income levels, and unorganised nature of most
economic activities, which erode the income tax base. However, an
equally important factor has been the provision of massive tax
incentives and exemptions to the manufacturing sector over extended
periods in most of these countries. As a result, the levels of budget
deficits and borrowings, and/or aid requirements become unsustainable
over time. Domestic resource mobilisation and reduction of budget
deficits then become the major targets of the Structural Adjustment
Programmes in the short run, and in many cases even in the long run.
In the case of Pakistan, while the initial stages of rapid economic
growth were characterised by massive tax concessions, the
nationalisation in the 1970s resulted in a massive shift to the informal
sector, which became synonymous with the parallel or underground economy
in a very short period. (1) The continued reduction in formal
employment, as a policy measure since the 1990s, has further expedited
the expansion of the informal sector. As compared to 20 percent in 1974,
after the nationalisation in 1972, and to 25 percent in 1990-91, it
accounted for 54 percent of the GDP in 1998 [see Kemal (2003)]. As a
result of the distortions created in the economy over time, Pakistan has
had a chain of stand-by and structural adjustment and stabilisation
programmes from 1973-74 until 2003. All the programmes had a special
focus on tax reforms including improved tax governance, increasing the
share of direct taxes, expanding the tax net, imposition of sales tax on
a wider scale, and improving the tax elasticity and buoyancy. However,
the implementation of the reforms particularly in the fiscal sector was
visible only for the last programme of 2000-03. The expanded enforcement
of the sales tax, rationalising the exercise of power by the tax
officials, and some expansion in the tax net has been achieved. However,
the overall tax revenues as a percentage of G DP still average less than
15 percent and the share of indirect taxes still exceeds 60 percent of
total tax revenues.
When the elasticity of major revenue sources remains low despite
tax reforms either due to low base, or due to evasion or avoidance, the
governments raise additional resources through discretionary measures.
Then, the growth of tax revenues comes through high buoyancy rather than
through elasticity. The objective of this paper is to measure the
buoyancy and elasticity of the tax system in Pakistan over the period
1974-75 to 2002-03 by using the Divisia Index Approach, and analyse the
factors responsible for the resulting size of elasticity coefficients.
The coefficient of elasticity depends on the level of tax rates, the
progressivity of the rate structure, and the responsiveness of the tax
base to changes in income. This makes it possible to break up the value
of elasticity into two components--the response of the tax base to a
change in income, and the response of the tax yield to a change in the
tax base of individual taxes through decomposition of elasticities [see
Musgrave (1959)].
The value of base to income elasticity does not depend on the
progressivity of tax rates; it simply relates the responsiveness of the
tax base to a change in income. The growth of the base depends on the
way the structure of the economy changes with economic growth. The
tax-to-base elasticity depends on the tax rates; if the rate structure
is progressive or if there is an improvement in tax administration, the
tax-to-base elasticity will be raised by preventing evasion. The
decomposition of elasticity in this manner permits us to identify the
source of growth of tax revenues.
The paper is structured as follows. Section Ii very briefly reviews
the trends in direct and indirect taxes, and highlights the policies
adopted to improve the tax system. Section III describes the data
sources and the methodology used for the estimation of elasticity in
this paper. Section IV analyses the results based on the Divisia Index
Estimates of elasticity and buoyancy, and the decomposition of
elasticities to describe the prevailing situation in the tax system and
its implications for the economy of Pakistan. Finally, Section V
concludes the paper with some policy recommendations.
I. TRENDS IN DIRECT AND INDIRECT TAXES (2)
Pakistan's tax system in its ability to raise adequate tax
revenues has not been any better than many other developing countries.
It has not been able to generate more than 14 percent of tax revenues in
relation to GDP. This lacklustre performance is largely attributed to
three inherent weaknesses of Pakistan's tax system, i.e., narrow
and distorted tax base, over-reliance on indirect taxes, and weak tax
administration. For example, in the case of personal income tax, the
fringe benefits of the employees remained exempt for a very long period
of time. Similarly, the exemptions from taxation of gross income have
been quite generous, amounting to one hundred thousand rupees in the
budget for 2003-04. The agriculture tax, mainly a provincial tax, has
not been fully implemented due to the loopholes provided in the
legislation for agriculture tax. On the other hand, the industrial
sector and the external sector have enjoyed massive concessions and tax
holidays in the 1960s and 1970s. A large number of commodities were
exempted from customs and excise duties without adequate monitoring to
prevent the abuse of this concession. The most important element in this
"exemption exercise" has been the "ad hocism" of the
policy measures. All these special departures result in the loss of
revenue and reduce the elasticity of tax revenues in relation to GDP.
Furthermore, taxes have failed to increase as a share of GDP as the
widespread tax avoidance/evasion--and the inability of the tax
authorities to enforce tax laws--has undermined the confidence of the
taxpayers in the efficacy of the tax system.
The resulting inadequacy of revenues is reflected in Table 1, and
total tax revenues average 13 percent of GDP over the period 1974-75 to
2002-03. (3,4)
Direct taxes comprise incomes of the non-corporate and the
corporate sector including the withholding taxes. (5) The share of
direct taxes remained very low until late 1980s, averaging 2 percent of
the GDP, largely due to the extended tax holidays and exemptions in the
1960s and early 1970s, but picked up in the early 1990s. These increased
by more than 3 percent of the GDP in the early 1990s and averaged 3.5
percent of GDP between 1995 and 2003. The share of indirect taxes
averaged between 11-12 percent from late 1970s to mid-1990s, and
registered a decline thereafter to around 9 percent of GDP.
However, over time, the shift in the shares of direct and indirect
taxes as a percent of total taxes has been quite significant. The share
of direct taxes increased from 13 percent in 1974-75 to 35 percent in
2001-02, but declined to 32.3 percent in 2002-03, while that of the
indirect taxes declined from 86.8 percent to 68 percent over the same
period.
It is important to note that the increase in the share of direct
taxes in the 1990s has come from the massive increase in the withholding
taxes. In the late 1960s, withholding taxes were levied only on three
sources of income: salaries, interest on securities, and payments to
non-residents. Until 1979, six kinds of payments were subject to
withholding tax, which increased to 19 in 1994-95 and to 25 in 1999-00
[CBR (March 2003)].
Among the indirect taxes, significant changes have been observed
over time. Customs duties which formed the major chunk of tax revenues
until early 1980s, have been rationalised from a maximum rate of around
125 percent in 1987-88 to 25 percent in 2002-03. Consequently, the share
of customs duties in total indirect taxes has declined from 54 percent
in 1990-91 to 19 percent in 2002-03. It has been decided that they will
now be used only for protection purposes. However, the decline in the
customs duties and the excise duties has been picked up by the sharp
increase in the sales tax revenue on imports and domestic production.
Until late 1980s, sales tax on domestic production and imports was
administered in such a way that it was no different from excise duties
on domestic production and customs duties on imports. As such, it served
no useful propose but affected the distribution of tax revenues between
the provinces and the federal government. The generalised sales tax was
introduced very comprehensively in 1989-90 and its rates have been
revised quite frequently to increase tax compliance by the unregistered
taxpayers and to generate additional tax revenues. Between 1990-91 and
1998-99 general sales tax revenues increased from 1.65 percent of GDP in
1990-91 to 2.77 percent in 1994-95, but these declined to 2.33 percent
in 1998-99. However, as a result of the increase in the sales tax base,
by extending the sales tax net in 1999-00, sales tax revenues increased
to 3.7 percent of GDP in that particular year. After averaging 4.6
percent in the following two years, sales tax revenues increased to 5.9
percent of GDP in 2002-03. The slow growth in sales tax revenues was due
to the revisions in tax rates for some commodities. In the budget for
2003-04, sales tax has been fixed at 12.5 percent for all entities.
Frequent revisions in the tax rates have been adopted as a measure
of fiscal policy to increase revenues. The personal income tax rates for
salaried and non-salaried persons have been unified; the minimum
threshold for income tax was raised from Rs 40,000 in 1988-89 to Rs
60,000 in 1998-99, and five tax rates ranging from 7.5 percent to 35
percent were introduced. In the budget for 2002-03 and 2003-04, the
minimum thresholds were raised to 80,000 and 100,000 rupees with the
objective to enhance revenue collection by lowering the taxable income
levels. Similarly, in the corporate sector, the banking company's
rates have been continuously slashed from 60 percent in 1992-93 to 50
percent in 2002-03. This decline at the rate of 3 percent per annum will
continue until the reform target of 35 percent is achieved within a
five-year period.
In the late 1990s, a number of important tax reforms were
announced, including the introduction of the book-keeping requirements
into the Income tax Ordinance in February 2000, whereby 25 percent of
the 1.2 to 1.5 million returns under the self-assessment schemes were to
be subjected to random audit. The announcement of 25 percent random
audit under self-assessment was meant to encourage the public at large
to file their returns without fear of undue harassment by the tax
officials. In the fiscal year 2002-03 only two percent of the returns
were audited, and while the CBR said it was not a random audit, it did
not specify any other basis for it. This was essential to mitigate the
general complaint against the tax authorities--that they unnecessarily
harassed the taxpayers in the private sector in the name of detailed
scrutiny. In this regard, the powers of the tax inspectors were also
severely curtailed; they are not allowed to open the returns filed.
However, the elimination of various tax exemptions, such as the
withdrawal of tax-whitener schemes that guaranteed immunity from tax
probes, making the interest income of national saving certificates
taxable, bringing in-kind benefits of the employees within the ambit of
income tax, and measure taken to make agricultural income tax fully
operational have helped improve tax collection and the tax structure to
some extent. These measures, it is expected, would lead to an increase
in the number of taxpayers in the coming years, provided, the continuity
of the policies is maintained.
III. DATA AND METHODOLOGY
The relevant data for this study for the period 1974-75 to 2002-03
are taken from the Central Board of Revenue Yearbook and the Pakistan
Economic Survey. Contrary to the earlier studies [Chowdhury (1962); Khan
(1973); Jeetun (1978); Gillani (1986); Kemal (1995)] the data used in
this study were tested for stationarity with the objective to use
co-integration technique rather than the simple OLS. The details of the
requisite data transformation/adjustments and the outcomes are reported
in Appendix I. It will be seen from the appendix table that the series
are non-stationary because the variables are integrated with different
orders, so the data fail to meet the pre-requisites of the
Co-integration technique for estimation. Therefore, we use the Vector
Auto-regressive (VAR) technique to estimate elasticity and buoyancy for
this study.
To estimate the built-in elasticity of a tax system, the historical
revenue data need to be adjusted for the effects on revenue from
discretionary tax changes as applied from time to time. The three common
methods adopted to eliminate the effects of discretionary changes in
taxes include the proportional adjustment method; the constant rate
structure; and the dummy variable method. However, a complete adjustment
of historical revenue series is not possible in any of the methods. The
first method requires use of budget estimates of tax yields resulting
from discretionary changes. Not only are such data difficult to obtain,
their reliability is questionable as the actual discretionary outcomes
may differ significantly from the changes proposed in the budget. The
data on discretionary revenues provided by the Central Board of Revenue
are not only incomplete, these underscore two limitations of the data.
Firstly, some of the measures proposed at the time of budget-making were
never implemented, and therefore the financial effect could be different
than reported in the data provided. Secondly, the calculation of
financial effect of tax measures is based on the 'static'
model. No dynamics have been captured, therefore ex-ante and ex-post
differential can exist.
The second method--the constant rate structure method--is not used
very commonly because it places heavy demands on the availability of
data. It requires data on effective tax rates and on the changing
composition of the bases. Provided these data are available and both the
tax and its base are defined narrowly enough to permit application of
the reference year rates to later year tax bases, this method generates
the most accurate revenue series. However, such data are hard to come by
particularly in the less developed countries.
The dummy variable method does not require the use of disaggregated data on taxes, but it cannot be used properly when discretionary tax
changes are quite frequent in the past. Furthermore, even if the
discretionary changes are not very large, the specification of the
estimation equations can be problematic unless there is information on
the nature of the tax changes and the extent to which their effects are
independent of one another. In the case of Pakistan, there have been
changes in taxes almost every year, and the Central Board of Revenue has
generally been overestimating tax revenues.
Therefore, we choose to estimate elasticity by the Divisia Index
(DI) Method--since it does not involve the traditional adjustment of the
historical revenues to eliminate the effects of discretionary tax
measures. The measurement of elasticity by the DI approach involves
three steps: first, the effects of discretionary tax measures on
revenues are estimated by an index that isolates the automatic growth of
revenues from the total growth; second, the buoyancy of tax revenue is
estimated with respect to GDP by a standard regression technique; and
third, the estimated buoyancy is adjusted by a suitable transformation
of the index of discretionary revenue, estimated in step one, to obtain
the estimate of elasticity of the tax yield. (6)
The DI method uses only historic data and does not require the
collection of specific information on revenue effects of discretionary
tax changes, or on the frequency of past discretionary tax changes.
However, two caveats may be noted: first, it may
underestimate/over-estimate the revenue effects of discretionary tax
change; and secondly, in case of large revenue effects, it may give
unsatisfactory results.
III. (a) Divisia Index Method (7)
The Divisia Index, derived from a weighted sum of growth rate of
factor inputs, is an index of factor inputs, for the measurement of
technical change. The index of technical change is the ratio of an index
of total productivity to an index of factor productivity, the latter
measured by the Divisia index. This measure implies that the percentage
increase in total productivity caused by technical progress is equal to
the percentage increase in output divided by the percentage increase in
factor inputs. The appropriateness of this measure is based on its
property of invariance, i.e., if there is no technical change, the
growth in total productivity is entirely due to factor inputs. A change
in the Divisia index, therefore, gives a measure of the change in total
productivity that shifts the production function due to all sorts of
factors that are jointly termed as 'technical change'.
Intuitively, the effects of technical change are analogous to the
effects of discretionary tax measures. Discretionary tax measures
produce changes in tax yields over and above those caused by automatic
growth in tax bases as technical changes induce changes in total
productivity over and above that induced by factor inputs. This analogy
can be explained more specifically as follows: assume that there exists
a tax function that shows the tax yields resulting from k bases. This is
analogous to the production function that shows the aggregate output
given by n factor outputs. For the given tax structure and the given
configuration of the tax bases, tax yield will not change in the absence
of any discretionary tax measures, just as, for the given level of
factor inputs, there is no change in output in the absence of technical
change. On the other hand, for a given set of tax bases, if we assume a
discretionary tax measure is adopted that alters tax rates and/or
exemption levels of one or more categories of taxes, the revenues
produced are different from what they would be in the absence of such an
action. This difference or change in the tax yield arises due to the
induced shift (of either the intercept, the slope, or both) in the
aggregate tax function due to the discretionary tax action analogous to
that caused by technical change in the production function, which
produces a different output from that produced without technical change.
A Divisia index of discretionary tax change, therefore, can be
considered as analogous to the index of technical change. This index
should be equal to the percentage increase in total tax yield divided by
the percentage increase in total tax yield owing to the automatic
increase in the bases. Similarly, like the index of technical change, a
change in this index should reflect the overall revenue effects of
discretionary tax measures.
The applicability of this index of discretionary tax change is
subject to two conditions: it must be derived from an aggregate tax
function analogous to a production function; and it must posses the
invariance property.
The necessary and sufficient conditions that ensure the invariance
property of the Divisia index are as follows: there exists a
well-defined continuously differentiable aggregate function, f
([x.sub.l] (t), ..., [x.sub.k](t)); the function f is linear homogenous (implying constant returns to scale).
The existence of condition (a) is fundamental to the existence of a
relationship between the tax yields and the tax bases, and the concepts
of elasticity and buoyancy. In the absence of an underlying aggregate
tax function, there is a fundamental indeterminancy about the tax yield
and tax bases. The "continuously differentiable" character of
the aggregate tax function ensures the regularity of such a function,
and prevents the erratic behaviour of the tax yield. Indeed the
existence of condition (a) is central to the derivation of the Divisia
index.
The requirement of linear homogeneity is more restrictive because
in a progressive rate structure, such as that in case of income tax, it
is clear that an increase in per capita incomes will produce more than
proportional increase in revenues. However, the elimination of this
restriction by Hulten (1973) (8) made possible the application of the DI
to aggregate functions with non-constant returns without violating the
invariance property. It is important, however, to point out here that
the assumption of a homogeneous aggregate tax function is justified in
case of LDCs where tax ratios (tax/GDP) have increased relatively faster
than in the developed countries but the average increase has been rather
small. (9) These trends in aggregate revenues can be written as a
homogeneous function of GDP (x)
T = [ax.sup.[mu]] ... ... ... ... ... ... ... (1)
When x rises over time, the tax ratio remains constant or rises as
the value of [mu]. (buoyancy) equals or exceeds unity.
This homogeneous tax function has been extensively used in
empirical studies for estimating the buoyancy or the elasticity of tax
revenues.
Derivation of the DI of Discretionary Tax Revenues
Using the continuously differentiable aggregate tax function at
each point in time:
T(t) = f[[x.sub.i](t), ... [x.sub.k](t); t] ... ... ... ... ... (2)
Where T is the aggregate tax yield; x denotes the proxy tax base
for k categories of taxes; and t, the time variable is the proxy for
discretionary tax measures. The effects of discretionary tax measures in
Equation (3) are obtained by taking the logarithm of the tax function,
differentiating with respect to time and re-arranging as:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
Setting [f.sub.t](t)/f(t) = [??](t)/D(t) and
[f.sub.i](t)[x.sub.i](t)/f(t) = [[beta].sub.i](t) where D(t) is the DI
of discretionary tax change; Equation (3) is rewritten as
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)
Where; [??](t)/D(t), represents the growth of tax revenues owing to
discretionary tax measures.
To obtain the index of discretionary tax revenue over the time
interval [0,n] Equation (4) is integrated as:
D(n)/D(0) = [T(n)/T{0)]exp(-[k.summation over
(l)][[integral].sup.n.sub.0][[beta].sub.i](t) [[??].sub.i]/[x.sub.i](t)
dt) ... ... ... (5)
Normalising by setting D(0 = 1), D(n) represents the index of
revenue growth only due to discretionary tax measures over time.
Star and Hall (1976) simplified the right-hand side of Equation (5)
by replacing the fluctuating [[beta].sub.i](t) by a constant
[[??].sub.i] (t) which is some form of weighted average of the
[[beta].sub.i](t).
This transformation yields Equation (6) of the form:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)
integrating the left-hand side of Equation (6) leads to
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)
In Equation 7 the growth rate of tax revenues is divided by the
index of automatic growth of tax revenues as measured by the
denominator.
If the left-hand side of Equation (7) is put into the right-hand
side of Equation (5), we get
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)
In Equation (8) the growth of discretionary tax revenues is the
difference between the growth rates of total tax revenues and automatic
tax revenues, which is a sum of the growth rates of the (proxy) bases
where the weight [[??].sub.i](t) is obtained from Equation (7).
In log form Equation (8) can be written as:
log D(n) = log(T(n)/T(0)) - [k.summation over (l)[[??].sub.i]
log([x.sub.i(n)]/xi(0) ... ... ... (8a)
The index D(n) does not require any adjustment of historical
revenue data, and it is the exact index of discretionary tax revenues as
derived from Equation 2, subject to the limitations of over- and
under-estimation as pointed out earlier.
Having derived the index of discretionary change, the estimation of
the buoyancy of tax revenues is simple: for the simple homogenous
aggregate tax function when the homogeneity is assumed to be r > 0,
it can be shown that if the growth rates of all the bases are equal to
that of GDP, then the tax function takes the form
T(t) = ax[(t).sup.r][D.sup.*](t) = ax(t)[mu] ... ... ... ... ...
(9)
Where x denotes GDP, [D.sup.*] denotes an index of revenue growth
due to the discretionary changes in the time interval [0, n], and [mu]
denotes the buoyancy of tax yield. The index [D.sup.*] is a special case
of index D, and, for the time interval [0, n] it has the same form
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (10)
Where [[??].sup.*] = 1/n [[integral.sup.n.sub.0][beta](t)
[rho](t)/[bar.[rho]] dt, [rho] being the growth rate of GDP.
Also from Equation (9) it follows the index [D.sup.*] for the time
interval [0, n] can be written as:
[D.sup.*](n) = [[[x.sub.i](n)/[x.sub.i](0)].sup.[mu] - r] (11)
Index [D.sup.*] is invariant. If there is no discretionary tax
change in the time interval, the elasticity of the tax system r must
equal the buoyancy [mu] of the tax yield, hut this implies that
[D.sup.*](n) = 1.
Under the Dl method, the estimation of elasticity from the
historical data involves two steps:
We first estimate the buoyancy p from the unadjusted historical
revenue data for the time interval [0, n], by estimating the tax
function T = [ax.sup.[mu]]. Secondly, since index [D.sup.*] is derived
from the underlying tax function f as index D, the latter can be
substituted in Equation (11), which yields
D(n) = [[x(n)/x(0)].sup.[mu] - [??]] (12)
Taking logs, Equation 12 gives
[??] = [mu] - logD(n)/log[x(n)/x(0)] ... ... ... ... ... (13)
Equation 13 provides the estimate of the elasticity r from the
unadjusted historical revenue data, subject to the limitations of over-
or under-estimation.
IV. EMPIRICAL RESULTS
With reference to Equation 2, the main categories of federal taxes
and their relevant bases, besides the GDP, are reported in Table 2.
Until the mid-1970s, the income tax and corporate tax were reported
separately, but it was decided to merge the income tax statistics of the
corporate and the non-corporate sector. Since the agriculture tax is a
purely provincial subject, it is not included; but the withholding tax
of the corporate sector is included in the income tax. The relevant base
for customs duty is the value of imports. Prior to the policy of tariff
reduction, customs duty was the major source of tax revenues, but now it
is only used for protection purposes. Similarly, prior to mid-1990s,
excise duty was levied on the manufacturing sector, retail businesses,
financial services, and other services. However, with the enlargement of
the sales tax net, and its effective implementation, excise duties are
now levied mainly on domestic manufacturing. Sales tax is levied on
domestic output including services and imports.
Following the DI approach, the buoyancy is estimated by estimating
the tax function T = [ax.sup.[mu]], with respect to total GDP, and the
elasticity estimates are obtained by adjusting buoyancy as given by
Equation 13. Since the overall effect of discretionary change is shown
to increase revenues, the elasticity of tax revenues is expected to be
smaller than the buoyancy. However, in Table 3 we see that the
coefficients of elasticity slightly exceed those of the buoyancy for
both the customs duties and the sales tax in the long run. In case of
customs duties, however, the elasticity coefficient is considerably
lower than unity. The high elasticity of sales tax with respect to the
total GDP base may clearly be attributed to the extension of sales tax
to electricity, gas, and petroleum products. These items constitute the
basic input to all the production and distribution network in the
economy. The inelastic demand for these inputs makes the revenues from
the sales tax more elastic with reference to GDP base.
The short-run elasticities are included with a view to examining if
the measures announced have full impact during the year or not.
Elasticity estimates for all tax categories with respect to their
relevant bases have also been estimated both for the short run and the
long run. However, only the results for the long-run estimates with
reference to relevant bases are discussed here since they are more
relevant from the policy point of view.
It will be seen from Table 4 that the elasticity of total tax
remains below unity mainly due to a sharp decline in the elasticity
coefficient of customs duties. The income tax accounts for only 35
percent of the total tax revenue, and income tax elasticity though
highly significant is only 1.15 despite the inclusion of withholding
taxes. (10)
Withholding taxes accounted for more than 70 percent of net income
tax collections, and for almost the entire revenue augmentation in the
1990s. Since most of the withholding taxes constitute a final discharge
of tax liability if these were excluded then the direct taxes would be
around 1-2 percent of the GDP in the 1990s [IMF (2001)]. The withholding
taxes in Pakistan are essentially indirect taxes because
firms/individuals paying a fixed 5 percent withholding tax are not
required to file an income tax return. In this case, a person pays a 5
percent withholding tax to import a commodity, but when he earns 10
percent or more profit on selling it onwards, that profit is not taxable
because he does not file an income tax return. Therefore, accounting
them as direct taxes has serious implications. While the government
loses on revenue, the tax authorities are content that a fixed amount of
tax is ensured without getting into the hassle of tax returns, which
involves scrutiny, disputes, claims, refunds, etc. This has prevented
the modernisation of the tax administration to improve the income tax
collection on more rational grounds for a long time. Currently, the CBR
is reported to have introduced the filing of tax returns by the WHT payers on a limited scale.
Sales tax accounts for 64 percent of the revenues from indirect
taxes. Disaggregating by relevant bases we see that the elasticity of
sales tax on domestic output with respect to non-agricultural GDP is the
highest--l.81. (11) However, as pointed out earlier, this increase has
been made possible by extending the sales tax net to the petroleum
products, gas, and electricity since 1999-2000. These commodities are
basic inputs to all the production and distribution network, thus the
production and distribution of the basic food items has been affected by
the taxation of these items, affecting the consumption levels of the
already poor. The increased expenditure on fuel and lighting-despite
lowering its consumption, has led to lower consumption of other
essential food items. However, the full increase in the prices of these
three items is not fully reflected in the consumer price index [Kemal
(unpublished)]. The report of the State Bank (12) has revised its
end-of-the-year target of inflation upwards due to the rapid increases
in the prices of the petroleum products on fortnightly basis for the
past two months. The report relates the decline in sales tax revenues
from kerosene oil to the decline in the purchase of kerosene oil by the
poor, who have turned to cutting the already thin woodlands for fuel.
Recent estimates by different groups show that the government gets Rs 13
per litre of petroleum products in the form of customs and excise
duties, a fixed sum is earmarked as revenue from the petroleum
development levy, in addition to a 15 percent sales tax. The profits of
the oil marketing companies are over and above these and other costs
deductions.
The coefficient of elasticity of sales tax for imports is 1.2Y
However the elasticity of customs duties on imports is exceptionally low
at -1.25 in the presence of sales tax, which picks up the loss in
revenue due to the reduction in tariffs. Similarly, although excise
duties account for 18 percent of the indirect tax revenues and are
levied on inelastic items including beverages, cigarettes, and natural
gas, the elasticity of excise duty is less than unity due to the
imposition of the sales tax.
Decomposition of Elasticities
The elasticity of a given tax [DELTA][T.sub.k]/[DELTA]Y x
Y/[T.sub.k] comprises two components, i.e., the tax-to-base elasticity
[DELTA]T/[DELTA][B.sub.k] x [B.sub.k]/[T.sub.k] or the elasticity of the
tax collected relative to base; and the base-to-income elasticity
[DELTA][B.sub.k]/Y x Y/[B.sub.k]
Therefore [[DELTA]T/[DELTA][B.sub.k] x [B.sub.k]/[T.sub.k]]
[equivalent to] [[DELTA]T/[DELTA][B.sub.k] x
[B.sub.k]/[T.sub.k]][[[DELTA][B.sub.k]/Y x Y/[B.sub.k]].sup.14]
Where
T = tax revenue,
Y = GDP, and
[B.sub.k] = base related to individual categories of taxes.
The decomposing of tax elasticities is helpful in identifying the
dynamic and the lagging components of the tax system. Furthermore, the
governments can influence the tax to base component to improve the
elasticity of a particular tax. The results of decomposition of four
major taxes have been related to the relevant bases, which are then
related to GDP as shown in Table 5.
In case of income tax, the base-to-GDP elasticity is low whereas
the tax-to-base elasticity is higher. The vast scale of tax exemptions
and evasion both in the formal and the informal sectors, and the
unchecked growth of the informal sector, which is synonymous with the
underground economy in Pakistan, are responsible for the erosion of the
tax base. The current policy of loans through the Small and Medium
Enterprise Development Agency (SMEDA) bank is a step in the right
direction for the informal sector in the small scale manufacturing
sector, where firms are constrained to expand due to credit constraints.
However, the policy of micro credit in general, mainly through the
non-government organisations as well as through some special banks to
promote self-employment is a step towards the expansion of the informal
sector. The amounts of loans in this case are so small that it allows
individuals to engage only in subsistence self-employment in the
informal sector. The higher tax-to-base elasticity for income tax
reflects the inclusion of the withholding taxes in income tax, which, as
pointed out earlier, blunts the incentive to broaden the base of income
tax.
The low coefficient of the base-to-GDP elasticity for customs
duties is due to the exemption from customs duties for a wide range of
raw materials and machinery. The low tax-to-base elasticity captures the
effect of the reduction in tariffs over time. The high tax-to-base
elasticity of the sales tax, resulting in a high overall elasticity
coefficient (1.81), implies that the tax net has been widened
significantly. This includes the imposition of the sales tax on
electricity, gas, and petroleum products, with consequent adverse
effects on the already poor. Elasticities of the three components for
excise duties are less than unity. Low tax collection relative to the
base is attributable mainly to replacement of excise duty by the sales
tax. However, at the same time, cheaper smuggled as well as imported
goods have flooded the domestic markets, resulting in loss of revenues
to the government on account of excise duty and sales tax.
Overall, the low base-to-income elasticities, as well as the
relatively lower tax-to-base elasticities, explain the low elasticities
of the various taxes. They reflect the failure of the successive
governments to improve the tax administration. Efforts to improve the
tax imposition and implementation have been beset with loopholes to
allow evasion in one form or another. Furthermore, "ad
hocisim" in policy measures to enhance revenues has resulted in
distortions and narrow bases with consequent decline in revenues.
The findings of this study can be compared with earlier studies to
show that the low elasticities have been inherent in the Pakistani tax
system for a long time. Of the six studies reviewed here, four have
similar findings, which are attributed to low tax-to-base elasticities.
The study by Khan (1974) for the period 1960-61 to 1971-72 estimates the
elasticity and buoyancy of the tax system using the Dummy Variable
method. In this study, elasticites exceed buoyancy for total tax
revenues, excise duties, and income tax. However, elasticities for sales
tax, customs duties, and income tax from manufacturing are less than
unity but still exceed the buoyancy coefficients. It appears that this
study is beset with all the problems associated with this method as
described in Section III. The coefficient of customs duties is puzzling
because it was the main source of revenues until late 1980s. The study
by Jeetun (1978) uses Khan's data-set extended by four years to
estimate the buoyancy and elasticity by the Proportional Adjustment
Method, and his findings differ significantly from those of Khan. He
reports less than unity elasticity for all taxes with different bases
except for customs duties with GDP base. Furthermore, they are less than
the buoyancy estimates. He attributes these results to low tax-to-base
elasticities particularly for income tax, which is supposed to be
progressive. In the case of Pakistan, Jeetun rightly argues that
progressivity is lost to various concessions and exemptions, besides a
weak tax administration and widespread evasion. These findings are
confirmed by the present study despite the use of a different
methodology and the different time-period covered. This confirms that
the distortions prevailing in the tax system in the 1960s still persist.
Gillani (1986) uses both the proportional adjustment and the
Divisia Index methods to estimate the elasticity and buoyancy of taxes
for the period 1972 to 1982. The results of this study are closer to
those of Khan's study, it shows high elasticity coefficients for
all taxes except for export taxes on account of the two methodologies
used. However, the gaps in the methodologies make a systematic analysis
difficult. The estimates of elasticity and buoyancy by Kemal (1995) over
the period 1971 to 1992, based on the Proportional Adjustment Method,
coincide with those of Jeetun and the present study. The buoyancy
coefficients exceed the elasticity coefficients by a wide margins. As
compared to this study, the coefficients of elasticity are relatively
lower because they do not capture the impact of the reforms initiated in
the early 1990s.
V. CONCLUSIONS AND POLICY RECOMMENDATIONS
The estimates of buoyancy and elasticity based on the Divisia Index
approach show that overall the use of discretionary measures has been
relied upon significantly as a source of revenue augmentation in
Pakistan. The low elasticites of the tax system reported by earlier
studies using a different methodology confirm the existence of continued
exemptions, allowances, and loopholes for evasion. All these factors
contribute to distortions in the tax system, preventing the tax base to
broaden as the economy expands. The reforms in the tax structure since
late 1990s--in terms of a relatively cleaner administration facilitating
the tax-payers--and the broadening of the sales tax base are visible.
However, the efforts to increase the share of direct taxes are at best
extremely limited. Inclusion of the withholding tax as a part of direct
tax reflects an artificial increase in the share of income tax, which
has adverse implications for a genuine increase in income tax revenues,
as for expanding the tax net.
Broadening of the sales tax base in a relatively short period is a
positive development. However, at the same time, the imposition of sales
taxes on petroleum products and utilities ignores its adverse impact on
the already poor segments of the population.
The reduction in the tariff rates over time affects the coefficient
of elasticity for imports significantly. However, the decline in the
customs duty coefficients is picked up by the sales tax revenue from
imports. The decline in revenues from the excise duties is also
predictable considering the extension of sales tax to a large number of
sectors/commodities. This decline is captured by the coefficient of
sales tax on domestic output.
Overall, it appears that taxes are being levied without due
consideration to equity issues. For example, the capital gains on equity
remain exempted from tax on the pretext of developing stock markets;
therefore the billions being earned from the stock market bubble remain
untaxed, while sales tax on petroleum products and utility prices has
overburdened the common man. This has directly affected the common man
with limited resources.
The easy way to realise revenues--through sales tax on petroleum
products and rising utility prices--has equity implications; a higher
sales tax rate on luxury items could be used to prevent, or at least
reduce, the taxation of the essential and basic inputs to the production
and distribution processes. This would help increase production, output,
and hence employment. High cost of production has been a major factor
inhibiting investment, besides procedural and institutional delays. This
calls for a better coordination between the revenue department,
investment ministry, and other relevant departments. In this regard,
serious note should be taken of the State Bank of Pakistan report which
has not only revised the end-of-fiscal-year target of inflation due to
the massive increase in oil price (which includes levy of customs,
excise, and sales tax besides the petroleum development levy). It also
points out that the decline in revenues of sales tax from kerosene oil,
which is mainly used as fuel by the poor, is due to the decline in its
sale. The report maintains that these people are turning to woodlands
for fuel. This reflects the lack of communication between the Ministry
of Environment and the tax authority.
The revenue department, working in isolation, is focussing only on
more revenue generation from the direct taxes by including withholding
tax (WHT) in the income tax, and from indirect taxes by levying a
uniform sales tax on all items regardless of the possible negative
outcomes. For an effective outcome of its reform package, it needs to
consider the consequences of its policies for other departments, as well
as the long-run consequences of short-term measures to achieve a rapid
increases in direct and indirect taxes. This calls for a consultative
process in policy formulation, which is generally lacking at all levels.
APPENDIX I
TESTS FOR STATIONARITY
The first step in time-series analysis is the test for
stationarity. If a variable is stationary, i.e., it does not have a unit
root, it is said to be 1(0)--integrated of order zero. If a variable is
not stationary on level but stationary in its first-differenced form, it
is said to be integrated of order one, or (1). The presence of unit root
in a univariate time-series is tested by the Augmented Dicky-Fuller
regression (1979 and 1981) of the form:
[DELTA][Y.sub.t] = a0 + [beta]t + [gamma][Y.sub.t - 1] +
[[summation].sup.P.sub.i = 2][[beta].sub.i][DELTA][Y.sub.t - i + 1] +
[[epsilon].sub.t]
Where
[gamma] = -[1 - [[summation].sup.P.sub.i = 1][a.sub.i]],
[[beta].sub.i] = [[summation].sub.j = i.sup.P][a.sub.j]
[DELTA] is first difference of [Y.sub.t], [a.sub.0] is the
stochastic term that follows the classical assumptions--it has zero
mean, constant variance, and is non-auto-correlated, or is white noise.
The null hypothesis [H.sub.0] : [gamma] = 0 implies that the
time-series is non-stationary;
The alternative hypothesis, [H.sub.0] : [gamma] = 0, implies that
the time-series is stationary.
We estimate the equation by the OLS and compare the t-ratios of the
estimated co-efficient of [Y.sub.t - 1] with the Dicky-Fuller table. If
the computed absolute value of tau statistics exceeds the DF critical
tau ([tau]) values, we reject the hypothesis that [gamma] = 0, and in
this case the time-series is stationary. On the other hand, if the
computed [absolute value of [tau]] is less than the critical [absolute
value of [tau]] values, we do not reject the null hypothesis. The
results of the unit root test performed on each variable are reported in
the following table.
Appendix Table
Augmented Dickey-Fuller Test Results for Unit Roots
(Divicia Index Approach for Sample 1974-75-2002-03)
Variables ADF-stats ADF-stats
Level Level First Difference
LADD -2.359382 (c, t) -2.699034 (c) *
LY -2.626483 (c, t) -2.565092 (c)
LTDI -2.777864 (c, t) -3.603557 (c) *
LADIDT -2.743358 (c) -3.588622 (c) *
LTIDT -2.799894 (c, t) -3.359485 (c) *
LADC -3.551917 (c) *
LCUS -5.723694 (c) *
LET -3.330872 (c) *
LADE -2.858721 (c) -2.237782 *
LS -4.496973 (c, t) *
LADS -4.102213 (c, t) *
Variables ADF-stats
Level Second Difference Result
LADD 1 (1)
LY -6.843948 * 1 (2)
LTDI 1
LADIDT 1 (1)
LTIDT 1 (1)
LADC 1 (0)
LCUS 1
LET 1
LADE 1 (1)
LS 1 (0)
LADS 1 (0)
Note: * Denotes significance at 5 percent, ("c") indicates that the
constant term is significant; (c, t,) indicate that both the constant
and the trend are significant; 1 (1), indicates the unit root in
levels, and stationary after first differencing; 1 (0) denotes
stationary at level; and I (2) denotes stationary after second
differencing.
Author's Note: I am grateful to two anonymous referees of the
PDR for their very useful comments and suggestions on the paper. I would
also like to thank Dr A. R. Kemal, Director, PIDE, for his valuable
guidance at every stage in the preparation of this paper. Thanks are
also due to Mr Imtiaz Ahmad, PhD Fellow at PIDE, for his computational
assistance.
REFERENCES
Central Board of Revenue (2002) CBR Quarterly Review 2:2, Oct-Dec.
Central Board of Revenue (2003) CBR Quarterly Review 2:2, Oct-Dec.
Central Board of Revenue (2003) CBR Quarterly Review 2:3,
Jan-March.
Chelliah, R. J., H. J. Bass, and M. R. Kelly (1975) Tax Ratios and
Tax Efforts in Developing Countries, 196 I-71. IMF Staff Papers 22.
Choudhry, N. N. (1979) Measuring the Elasticity of Tax Revenues: A
Divisia Index Approach. IMF Staff Papers 26:1.
Chowdhury, A. H. M. N. (1962) Predictability and the Flexibility of
Tax Revenues in Pakistan. The Pakistan Development Review 2:2, 189-214.
Gilllani, Syeda Fizza (1986) Elasticity and Buoyancy of Federal
Taxes in Pakistan. The Pakistan Development Review 25:2, 163-174.
Hulten, C. R. (1973) Divisia Index Numbers. Econometrica 41, Nov.,
as reported in Choudhry (1979).
IMF (2001) IMF Country Report No. 01/1178. International Monetary
Fund, Washington, D. C.
Jeetun, Azad (1978) Buoyancy and Elasticity of Taxes in Pakistan.
Applied Economics Research Centre (AERC), Karachi. (Research Report No.
27.)
Kemal, A. R. (1995) Pakistan: A Profile of the Tax System. In
Issues and Experiences in Tax System Reforms in Selected Countries of
the ESCAP Region. New York: United Nations.
Kemal, A. R. (1999) Elasticity and Buoyancy of the Tax System and
the Incidence of Taxes. (Unpublished.)
Kemal, M. Ali (2003) Underground Economy and Tax Evasion in
Pakistan: A Critical Evaluation. Pakistan Institute of Development
Economics, Islamabad. (Research Report No. 184.)
Khan, M. Z. (1973) Responsiveness of Tax Yield to Increases in
National Income. The Pakistan Development Review 12:4, 416-432.
Musgrave, Richard A. (1959) The Theory of Public Finance. New York:
McGraw Hill.
Pakistan, Government of (Various Issues) Pakistan Economic Survey.
Islamabad: Economic Advisers Wing, Ministry of Finance.
Pakistan, Government of (Various Issues). CBR Yearbook. Islamabad:
Directorate of Research and Statistics, Central Board of Revenue.
Star, Spencer, and Robert E. Hall (1976) An Approximate Divisia
Index of Total Factor Productivity. Econometrica 44, March, as reported
in Choudhry (1979).
State Bank of Pakistan (2001-02) Annual Report of the State Bank of
Pakistan. Karachi.
State Bank of Pakistan (2005) The State of Pakistan's Economy:
Second Quarterly Report for the Year 2004-2005 of the Central Board of
State Bank of Pakistan. Karachi.
(1) Government policies to exempt the four units of weaving from
taxes, and in general the policies that favoured micro enterprises over
the large-scale enterprises led to a massive shill to the informal
sector.
(2) All the tax revenues are the consolidated Federal and
Provincial taxes. Local taxes, however, are not included.
(3) Even though there has been a re-basing of national accounts, a
large number of enterprises may have been outside the GDP estimates. And
in any case, informal enterprises included in the GDP have not been
paying taxes due to concessions as well as tax evasion.
(4) The disaggregated financial statistics for present Pakistan,
after the creation of Bangladesh, begin in 1974-75, as announced by the
State Bank of Pakistan.
(5) Other direct taxes including wealth tax, gift tax, and estate
duty have been abolished, and the share of workers' welfare fund is
very small in the direct taxes.
(6) As explained in Section III. (a) from Equation 8a to Equation
13.
(7) This section is based on Choudhry (1979).
(8) As reported in Choudhry (1979).
(9) While Choudhry 0979) quotes Chelliah, et al. (1975). reporting
that the average tax ratios of developing countries has increased by
13.6 percent for 1961-6g, to 15.1 percent for 1969-70, it will be seen
from Table I that for Pakistan this ratio averages only 14 percent for
1990-91 to 2002-03.
(10) When Writ is excluded, the coefficient of the income tax is
reduced and its significance level also drops. However, it is not
reported here due to gaps in information on WHT in the earlier period.
(11) This may seem at odds with the coefficient of sales tax with
GDP base in Table 3. However, it reflects the fact that a large
proportion of the poor population who live in the rural areas do not
have access to gas and electricity, and hence do not pay sales tax.
However, by its weight in the total count it pulls down the estimates,
and when it is excluded the coefficient rises.
(12) State Bank of Pakistan (2005).
(13) This identity is strictly true when the function is perfectly
estimated, i.e., when [[bar.R].sup.2] 2 level is 1.00. It does not hold
strictly in this case because there are differences between the overall
elasticities of tax and the product of the decomposed elements, as
follows: for total taxes (0.961 vs. 0.948), income tax (1.208 vs.
1.165), customs duties (0.432 vs. 1.232), excise duties (0.435 vs.
0.690), and sales tax (1.498 vs. 1.838).
Faiz Bilquees is Joint Director at the Pakistan Institute of
Development Economics, Islamabad.
Table 1
Composition of Consolidated Federal and Provincial Taxes in Pakistan
Percent Share in
Percentage of GDP Total Taxes
Tax Direct Indirect Direct Indirect
Period Revenue Taxes Taxes
1974-75 8.03 1.12 6.91 13.2 86.8
1976-77 11.94 1.78 10.16 17.4 82.6
1978-79 12.94 1.71 11.23 14.6 85.4
1980-81 14.12 2.52 11.60 20.7 79.3
1983-84 13.18 2.19 10.99 17.4 82.6
1986-87 14.48 1.94 12.54 16.4 83.6
1987-88 13.84 1.84 12.40 13.0 87.0
1990-91 12.70 2.03 12.00 18.0 82.0
1993-94 13.22 2.80 10.67 25.1 74.9
1995-96 14.35 3.72 10.41 29.1 70.9
1997-98 13.26 3.91 10.64 35.0 65.0
1999-00 12.90 3.67 9.33 32.5 67.5
2000-01 12.90 3.75 9.14 31.8 68.2
2001-02 13.22 4.06 9.14 35.3 64.7
2002-03 13.80 2.89 9.91 32.2 67.8
Percent Share of Individual Taxes
in Indirect Taxes
Customs Sales Central Excise
Period
1974-75 50.0 11.3 38.7
1976-77 47.5 10.5 42.0
1978-79 50.7 9.7 34.6
1980-81 51.7 10.5 37.8
1983-84 51.5 11.1 37.4
1986-87 58.9 9.9 31.2
1987-88 47.0 10.8 19.1
1990-91 54.9 17.6 27.5
1993-94 49.7 23.5 26.9
1995-96 46.8 26.3 26.9
1997-98 39.1 28.3 32.6
1999-00 26.4 49.9 23.7
2000-01 24.3 57.4 18.3
2001-02 18.3 63.7 18.0
2002-03 18.9 65.9 15.2
Source: Pakistan Economic Survey (Various Issues).
Table 2
Consolidated Federal and Provincial Tares and
Corresponding Relevant Bases
Categories of Taxes Relevant Bases
1. Direct tax/total income
tax of the corporate and
non-corporate sector Non-agricultural GDP
excluding the tax on
agricultural income but
including the withholding
tax
2. Customs Duties Imports
3. Excise Duties Manufacturing sector output
4. Sales Tax Domestic output, services, and imports
Table 3
Elasticity and Buoyancy of Consolidated Federal and
Provincial Taxes (with GDP Base)
Total Income Customs Excise Sales
Tax Tax Duty Duty Tax
Short run 1974-2003
Buoyancy 0.44 0.40 -0.06 0.48 0.42
Elasticity 0.33 0.31 -0.20 0.06 0.38
Long run 1974-2003
Buoyancy 0.92 1.23 -1.19 0.48 1.41
Elasticity 0.88 1.21 0.43 0.44 1.50
Table 4
Elasticity Estimates of Federal and Provincial Taxes
Using only Relevant Bases
Elasticity
Tax Relevant Base Coefficients
Total Tax 0.94 (0.42) (d)
Direct Tax/Income Tax Non-agriculture GDP 1.16 (0.16)
Sales Tax on Domestic Output Non-agriculture GDP 1.81 (0.32)
Sales fax on Domestic Output GDP 1.85 (0.58)
Sales Tax on Imports Imports 1.23 (0.49)
Customs Duty Imports -1.25 (-0.23)
Excise Duty Manufacturing Output 0.71 (0.25)
Weight of
Tax Tax in 2003 [R.sup.2] T-test
Total Tax 100.00 1.00 2.37
Direct Tax/Income Tax 35.3 (a) 0.99 6.91
Sales Tax on Domestic Output
Sales fax on Domestic Output 63.7 (b)(c) 0.99 7.28
Sales Tax on Imports 0.99 3.77
Customs Duty 18.3 (b) 0.99 12.31
Excise Duty 18.0 (b) 0.99 3.08
(a) Percent of total taxes; (b) percent of indirect tax; (c) refers to
total sales tax; (d) short-run elasticities are reported in
parentheses.
Table 5
Decomposition of Tax Elasticities
Taxes Tax to GDP Base to GDP Tax to Base
Total 0.961 1.017 0.934
Direct 1.208 1.017 1.146
Customs 0.432 0.986 -1.250
Excise 0.435 0.984 0.706
Sales 1.498 1.017 1.808
