Beruflich Dokumente
Kultur Dokumente
August 2008
Abstract
This paper seeks to analyse whether a nexus exists between central gov-
ernment revenues and expenditures for 1950-51 to 2003-04 by studying the
relationship between total receipts and an array of expenditure measures so
as to determine whether the spend and tax or tax and spend hypotheses
hold for government finances in India. The empirical exercise of the paper
provides us with mixed results. Significant uni-directional causal movements
from expenditures to revenues for three of the seven revenue-expenditure
pairs viz. total expenditures-total receipts, adjusted profile of expenditures-
total receipts and total expenditures-tax receipts indicate that expenditures
lead and revenues follow (spend and tax hypothesis). However, significant
uni-directional causal effects from receipts to expenditures are also seen for
two of the seven revenue-expenditure pairs namely, revenue expenditures-
1
Department of Economics, University of Mumbai.
Email: swatiraju@hotmail.com
2 CIISS August 2008
1 Introduction
Government deficits are one of the important indicators of fiscal health. In-
dia has seen deficits in central government accounts on almost all indicators
– fiscal deficit, primary deficit, revenue deficit - since the 1970s and this
situation worsened in the mid-1980s. The latter half of the 1980s saw fis-
cal deficits in the range of 7-8% of GDP and the gross primary deficit and
revenue deficit were also in the ranges of 4-5% of GDP and 2-3% of GDP re-
spectively. Fiscal consolidation, therefore, was the major focus of the reform
process introduced in 1991-92. The 1990s have seen varied performance on
the deficit indicators. The period 1991-92 to 1996-97 with the exception of
1993-94 had seen improvement in the situation with a decline in the fiscal
Raju: Revenue-Expenditure Nexus - Indian Evidence 3
F i g u r e 1 C e n t r a l G o v e r n m e n t T o t a l E x p e n d i t u r e a n d
R e c e i p t s 1 9 5 1 - 2 0 0 4 ( R s . C r o r e s )
500000 T r e c
T e x p
400000
C r o r e s
300000
R s .
200000
100000
0
1950 1960 1970 1980 1990 2000
Y e a r s
enue receipts and we can observe that both revenue expenditure and revenue
receipts have been showing an increasing trend over the entire period and
the gap between revenue receipts and revenue expenditure which reflects the
extent of saving/dis-saving of the government has widened in the 1990s. Rev-
enue expenditures increased by 4.94 times from Rs.73516 crore in 1990-91 to
Rs.362887 crore in 2003-04 while revenue receipts increased by 4.79 times
from Rs.54954 crore in 1990-91 to Rs. 263027 crore in 2003-04.
F i g u r e 2 C e n t r a l G o v e r n m e n t T o t a l E x p e n d i t u r e a n d T a x
R e c e i p t s 1 9 5 1 - 2 0 0 4 ( R s . C r o r e s )
500000
T R
450000 T e x p
400000
C r o r e s
350000
300000
250000
R s .
200000
150000
100000
50000
0
1950 1960 1970 1980 1990 2000
Y e a r s
F i g u r e 3 C e n t r a l G o v e r n m e n t R e v e n u e E x p e n d i t u r e a
R e v e n u e R e c e i p t s 1 9 5 1 - 2 0 0 4 ( R s . C r o r e s
400000
350000 R E X P
R R E C
300000
C r o r e s
250000
200000
R s .
150000
100000
50000
0
1950 1960 1970 1980 1990 2000
Y e a r s
10 CIISS August 2008
2 Literature Review
Most of the studies have examined the inter-relationship between expendi-
tures and revenues using the standard Granger causality tests and concern
the United States economy. Manage and Marlow (1986) study this relation-
ship for the United States over the period 1929-82. They use annual data
and study the finances of the federal government and conclude that causality
is uni-directional from federal taxes (revenues) to federal expenditures. The
study of Anderson, Wallace and Warner (1986) also for the United States us-
ing annual data for 1946-83 and for the federal government finds that causality
proceeds from expenditures to revenues, a finding contrary to that of Man-
age and Marlow. Ram (1988) in his study also for the United States tried
‘to ascertain which of the above conclusions is more reasonable’ (p. 763) and
12 CIISS August 2008
broadened the scope of his study to the relationship between federal receipts
and expenditures as well as that for state and local governments. Further,
it employed annual data for the period 1929-83 and quarterly data for 1947-
83 for both nominal and real measures of receipts and expenditures. Ram’s
study finds that causality runs from revenues to expenditures at the federal
level while at the state and local level causality proceeds from expenditures to
revenues. Von Furstenberg, Green and Jeong (1986) use a VAR model to anal-
yse the relationship between expenditures and revenues in terms of a tax and
spend or a spend and tax hypothesis. Quarterly data for the period 1954-82
is employed and apart from federal expenditures and taxes they include GNP
gap and inflation as additional variables in their model. The paper examines
4, 6 and 9 variable VAR models and finds no support for the tax and spend
hypothesis (revenues cause expenditures) while it finds support for the spend
and tax hypothesis (expenditures cause revenues) a finding akin to that of An-
derson, Wallace and Warner (1986). Miller and Russek (1990) seek to verify
this relationship using cointegration and error-correction models for federal,
state and local levels of government using annual as well as quarterly data
and both nominal and real measures of revenues and expenditures for 1946-
87. This study finds bi-directional causality between government taxes and
spending for all levels of government when quarterly data is employed. For
annual data, the results show at times uni-directional causality from taxes to
spending for the federal government while bi-directional causality is observed
for state and local governments. Jones and Joulfaian (1991) have examined
this interdependence between revenues and expenditures for the early years
of the American Republic using cointegration and error-correction technique
Raju: Revenue-Expenditure Nexus - Indian Evidence 13
and have concluded that the spend and tax hypothesis prevails in the short
run while the existence of a feedback relation between revenues and expen-
ditures is seen in the long run. Joulfaian and Mookerjee (1991) study the
dynamics between revenues and receipts for 22 OECD countries using a bi-
variate VAR model and later augment the VAR model using the output gap
and inflation variables so as to comprehensively model the nexus between
revenues and expenditures. They find greater support for spend and tax
hypothesis as compared to the tax and spend hypothesis. They further con-
clude that reductions in spending are essential to reducing budget deficits and
government size. Mithani and Khon (1999) study the inter-relation between
revenues and expenditures for Malaysia for 1970-94 using quarterly data and
the seasonal error correction model. This paper too finds uni-directional
causality from government expenditures to revenue – a result similar to that
of Anderson, Wallace and Warner (1986) and Von Furstenberg, Green and
Jeong (1986). Fasano and Wang (2002) test this interdependence for the
Gulf Countries Council (GCC) countries in a cointegration and vector error-
correction model (VECM) framework and find support for the tax and spend
hypothesis. They suggest that the GCC countries could ‘enhance the effec-
tiveness of their fiscal policy by making budget expenditure less driven by
revenue availability’ (p.18). Literature, thus, provides mixed evidence on the
nature of interdependence between government revenues and expenditures.
3 Methodology
A. Tests for Unit Roots
14 CIISS August 2008
p
X
∆yt = γyt−1 + β∆yt−i+1 + ²t (1)
i=2
p
X
∆yt = a0 + γyt−1 + β∆yt−i+1 + ²t (2)
i=2
p
X
∆yt = a0 + γyt−1 + a2 t + β∆yt−i+1 + ²t (3)
i=2
in Step I as the presence of these regressors may have reduced the power of
the test. Hence, in step II test for the significance of the trend term a2 by
testing the hypothesis a2 = γ = 0 using the ϕ3 statistic, that is, the F test in
the ADF test to test a2 = γ = 0 in equation (3) above. If the trend is not
significant proceed to Step III.
Step III Estimate the model without the trend (as given in equation 2
above) and test for the presence of unit root using the τµ statistic, that is,
the t statistic for the ADF test to test if γ = 0 in the presence of a constant
term. If the null of γ = 0 is rejected, conclude that the series does not contain
a unit root. If the null is not rejected, test the hypothesis using a0 = γ = 0
using the ϕ1 statistic, that is, the F test in the ADF test to determine if a
drift exists or test if the restriction a0 = γ= 0 is binding. If the drift is not
significant, proceed to Step IV.
Step IV Estimate the equation without the drift and trend terms (as in
equation 1 above) and test for the presence of unit root using the τ statistic.
If the null hypothesis of γ = 0 is rejected, conclude that the series yt does
not contain a unit root. Else, conclude that yt contains a unit root. (Enders,
2004).
(ii) Zivot and Andrews (ZA) test
The standard ADF test does not take into account the presence of struc-
tural break in the series and this could at times lead to an error when the null
hypothesis is not rejected. Hence, a series could be tested for unit root in the
presence of structural break. A major problem while testing for stationarity
in the presence of structural break is the timing of the break. ZA (1992) pro-
pose a test procedure in which the break point, k, is treated as an outcome
16 CIISS August 2008
k−1
X
∆xt = µA
1 + γ1A t + µA
2 DUt (k)
A
+ θ ∆xt−1 + βj ∆xt−j + ²t (4)
j=1
k−1
X
∆xt = µB
1 + γ1B t + γ2B DTt∗ (k) B
+ θ ∆xt−1 + βj ∆xt−j + ²t (5)
j=1
k−1
X
∆xt = µC
1 + γ1C t + γ2C DTt∗ (k) + µC B
2 DUt (k) + θ ∆xt−1 + βj ∆xt−j + ²t (6)
j=1
Where DUt (k) =1 and DT∗t (k) = t-k if t > k, and 0 otherwise with the
regressions being estimated over the range of the sample (0.15T..0.85T), T is
the total number of observations. Model A allows for a change in the level of
the series, Model B allows for a change in the slope of the trend of a series
and Model C estimates for both change in level and the slope of the trend
(Mills, 1999, pp.92-93).
p−1
X
∆xt = π0 + πxt−1 + Σ πi ∆xt−i + ²t (7)
i=1
condition that the speed of adjustment co-efficient, that is, the ECT in the
corresponding error correction model be equal to zero (Enders, 2004).
4 Empirical Evidence
The paper employs annual data over the period 1950-51 to 2003-04 and tests
for interdependence between central government revenues and expenditures.
The data for all the revenue and expenditure variables are obtained from
various issues of the Reserve Bank of India’s Report on Currency and Finance
and Handbook of Statistics.
The interdependence was examined for total receipts and an array of ex-
penditure measures. While the profile of total expenditures reveals actual
spending of the government, they also consist of defence expenditure and
interest payments which are usually committed expenditures and hence it
was thought that it would be worthwhile to consider a profile for expenditure
that excludes these two components of expenditure and see if this had any
meaningful impact on the interdependence between revenues and expendi-
tures and is given by TE2. Revenue deficits in India have been persistent
and growing since 1982-83, hence the interdependence between revenue re-
ceipts and revenue expenditures was also analysed. Further, expenditure was
further decomposed into development and non-development expenditure so
as to determine whether there exists a causal relation between these expen-
ditures vis-à-vis total receipts (which in turn may help identify whether it
is increases in development and/or non-development expenditure lead total
receipts (and thereby impact the fiscal deficit). Further, fiscal deficits in In-
20 CIISS August 2008
dia have been largely structural in nature, as stated earlier, and hence the
total revenue and total expenditure series were decomposed into their struc-
tural and cyclical components, using the Hodrick-Prescott filter (with) to
determine the trend and we looked at the structural profile of receipts and
expenditures to examine the existence of interdependence between the two.
All the variables are considered in their logarithmic form and denoted by the
suffix ‘L’ and are explained below.
All the variables were tested for the presence of unit roots using the sequential
ADF test and the ZA test which tests for unit roots in presence of a structural
break in the series (tests are detailed in Section III.A above). The lag length
for the tests was determined using the AIC criterion.
Results of the unit root tests are presented in Table1 and the ADF test
(Rows 1-5 indicate results of ADF test for the different variables in levels
Raju: Revenue-Expenditure Nexus - Indian Evidence 21
while Rows 6-7 indicates the results of the ADF test for the different variables
in First Differences) finds all the expenditure and revenue variables to be
difference stationary or I(1) processes except for STEXP and STREC, which
are found to be stationary in levels. Since the drift is significant (Row for
ϕ1 ) and the trend is not significant (Row for ϕ3 ) for most variables in levels,
the first difference ADF statistics (τµ ) reported are those for the presence of
constant, no trend. The Zivot-Andrews (ZA) test (Rows 8-10), with a null
of - series are I(1) without a structural break and a alternative hypothesis
of – series are I(0) with a single structural break, reveals that the variables
LTEXP, LTE2, LTR, LREXP and LRREC reject the null and hence are I(0)
processes with a single structural break while the other variables - LTREC,
LDE, LNDE do not reject the null hypothesis and hence are I(1) processes.
The ZA test indicates a structural break at the end of the 1960s and the
late 1970s and early 1980s. India adopted a model of planned economic
development after independence and accordingly the government played an
active role in economic development and made large investments in areas
it felt private enterprise would not be forthcoming. Public sector outlay
increased by 8.05 times from Rs.1960 crore during the First Five Year Plan
to Rs.15780 crore in the Fourth Five Year Plan and was at Rs.6750 crore
at the end of the three Annual Plans (1966-69). Expenditure increases can
also be attributed during the late 1960s to drought in 1965-66 and 1966-67
and industrial stagnation. Plan expenditures were higher on account of food
subsidies and higher cost of interest on foreign loans due to devaluation. The
drought in 1979 and the oil price shock saw increased expenditures at the end
of the 1970s. The 1980s saw higher growth but was based on excessive internal
22 CIISS August 2008
spending and both internal and external borrowing. In case of STEXP and
STREC since the ADF test itself finds that the variables are I(0) processes
i.e. stationary, we did not proceed to determine if STEXP and STREC were
I(0) processes with a single structural break. Hence the ZA test was not
performed on STEXP and STREC.
23
1 τT -2.31 -2.22 -2.78 -3.22 -1.99 -3.20 -1.79 -3.24 -5.47** -8.03**
2 ϕ3 2.70 2.55 4.14 5.19 2.30 5.26 1.74 5.45 – –
3 ϕ1 5.25 10.09 9.07 14.06 21.74 12.66 9.18 10.53 – –
4 τµ -0.27 0.22 -0.85 -0.31 -0.86 -0.61 -0.69 -0.74 – –
5 Lag(s) 5 2 2 1 1 1 1 3 1 1
6 τµ -3.96** -4.73** -12.72** -5.52** -5.57** -4.41** -4.16** -6.89** – –
7 Lag(s) 2 1 1 1 1 3 3 1 – –
8 ZAA -4.19 -3.98 -4.38 -6.15** -4.45 -6.13** -2.52 -4.66 – –
Lag(s) 0 0 0 1 0 1 0 0 – –
Year (1968) (1968) (1969) (1984) (1983) (1963) (1981) (1968) – –
9 ZAB -7.09** -7.23** -3.41 -3.40 -7.06** -6.13** -2.62 -3.86 – –
Lag(s) 0 0 0 0 0 1 0 0 – –
Year (1979) (1976) (1975) (1975) (1976) (1996) (1991) (1975) – –
10 ZAC -7.39** -7.57** -4.78 -4.79 -6.99** -6.24** -3.83 -4.24 – –
Lag(s) 0 0 0 0 0 1 0 0 – –
Year (1969) (1969) (1968) (1968) (1974) (1995) (1981) (1969) – –
24 CIISS August 2008
In Table 3 below which presents the results of causality and where we find
statistically significant causal interdependence for the above specified three
revenue-expenditure pairs satisfy the condition enumerated in Enders (2004)
and hence the results would hold.
Table 2 reports the results of cointegration using the Johansen technique
for the LDE-LTREC and LNDE-LTREC pairs and finds cointegration for
both these revenue-expenditure pairs. The lag length in the cointegrating
regression from which the lag structure for the Error Correction Model (ECM)
model can be inferred was also determined using the AIC. The ECM can be
employed to study long term causal movements only for the LDE-LTREC and
LNDE-LTREC pairs through the error-correction term (see Section III. B).
Table 3 reports the results of causal interdependence between the different
revenue-expenditure pairs.
26 CIISS August 2008
finances in India.
5 Conclusion
The paper seeks to examine whether there exists interdependence between
total receipts/tax receipts and an array of expenditure measures of the cen-
tral government. The empirical evidence in this paper is mixed and finds
the prevalence of causal interdependence between total receipts and different
categories of government expenditures. Support for expenditures lead rev-
enues - spend and tax hypothesis is observed for three of the seven revenue-
expenditure pairs viz. total expenditures-total receipts, adjusted total expen-
ditures (adjusted for defence and interest payments)-total receipts and total
expenditures-tax receipts. whereas revenues lead expenditure – tax and spend
hypothesis is seen for two of the seven revenue-expenditure pairs namely,
revenue expenditure-revenue receipts and for non-development expenditure-
total receipts. Significant bi-directional short run causal effects are observed
between development expenditures and total receipts. However, stronger
support for expenditures lead revenues can be inferred for development ex-
penditure and total receipts, as a causal relation from expenditures to re-
ceipts (implied by the significance of the ECT) is also observed. While no
causal relation is observed between structural expenditures and structural
revenues. Hence it would be safe to conclude that evidence for four of the
seven expenditure- revenue pairs support expenditures lead revenue i.e. the
spend and tax hypothesis. Hence, we could infer that the government pro-
ceeds with its expenditure plans and then finds means by which revenues can
30 CIISS August 2008
References
Anderson, W., M.S. Wallace and J.T. Warner (1989): “Government Spend-
ing and Taxation: What Causes What?”, Southern Economic Journal, vol.54,
pp.630-639
Barro, R.J. (1979): “On the Determination of the Public Debt”, Journal
of Political Economy, vol.81, pp.940-971.Chelliah, R.J. (1996): Towards sus-
Revenues in the Early Years of the American Republic: Evidence from 1792
to 1860, Journal of Macroeconomics, vol.13, pp.133-55.
Manage, N., and M.L. Marlow (1986): “The Causal Relation between Federal
Expenditures and Receipts”, Southern Economic Journal, vol.54, pp.617-629.
Miller, S.M. and F.S. Russek (1990): “Cointegration and Error Correction
Models: The Temporal Causality between Government Taxes and Spending”,
Southern Economic Journal, vol.57, pp.221-229.
Mithani, D.M. and G.S. Khon (1999): “Causality between Government Ex-
penditure and Revenue: Evidence from the Malaysian Case – A Seasonal
Cointegration Test” in Vasudevan, A., D.M. Nachane and A.V. Karnik (Ed.)
Fifty Years of Development Economics: Essays in Honour of Prof. P.R.
Brahmananda, pp.343-359.
Von Furstenberg, G.M., R.J. Green and J.H. Jeong (1986): “Tax and Spend
or Spend and Tax?”, Review of Economics and Statistics, vol.68, pp.179-188.