Gooodhali-GovernmentSpendingBudget-2003.pdf

Government Spending, Budget Financing, and Economic Growth: The Tunisian
Experience
Author(s): Khalifa Ghali
Source: The Journal of Developing Areas , Spring, 2003, Vol. 36, No. 2 (Spring, 2003),
pp. 19-37
Published by: College of Business, Tennessee State University
Stable URL: https://www.jstor.org/stable/4192918
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
College of Business, Tennessee State University is collaborating with JSTOR to digitize,
preserve and extend access to The Journal of Developing Areas
This content downloaded from
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00
All use subject to https://about.jstor.org/terms

The Journal of Developing Areas Volume 36 Number 2 Spring 2003
GOVERNMENT SPENDING, BUDGET
FINANCING, AND ECONOMIC GROWTH:
THE TUNISIAN EXPERIENCE
Khalifa Ghali
Kuwait University
ABSTRACT
Many developing countries adopted the IMF debt-stabilization programs without
evidence on whether public sector activities promote or depress economic growt
literature on the subject did not provide a consensus judgment on which a polic
safely be made. In addition, the question of whether the impact of government acti
depends on the source of financing the government budget is still open to debate. T
these issues in the case of Tunisia, which is a small developing country imple
stabilization programs without having any clear guidance on how government a
with macroeconomic variables in affecting its growth process. In contrast to
procedure of estimating a single growth equation, this paper develops a vector e
model and identifies both the direct as well as the indirect channels through w
spending can affect economic growth. These effects are then analyzed depending on
financed or a tax-financed fiscal policy is followed. The empirical results suggest
spending in Tunisia has an important role in shaping the general efficiency o
whereas government reliance on debt financing has adverse effects on econo
INTRODUCTION
In the aftennath of the debt crisis, many developing countries adhered
sponsored debt-stabilization programs, which called for shrinking the size of the
government in order to control the debt and promote economic growth. However, for
most of these countries, there is no evidence on the way government activities interact
with macroeconomic variables in affecting the growth process. Moreover, the existing
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

20 Khalifa Ghali
literature produced conflicting results and raised doubts about the nature of the
relationship between government spending and economic growth.
The economic literature on the subject witnessed the emergence of two conceptually
opposite views about the impact of government spending on economic growth. One point of
view suggests that higher government spending is likely to be detimental to economic
growth. One reason is that government operations are often conducted inefficiently. For
example, state owned firms are often politicized and their main objective is to satisfy
interests of bureaucrats and politicians. To maintain their power and prevent social unrest,
these groups favor excess employment, ineffective locations, and under pricing of output. In
addition, public investment undertaken by heavily subsidized state-owned monopolies has
more often reduced the possibilities for private investment and long-run economic growth.
Another reason is that the financing of public expenditures through external and internal
indebtedness imposes excessive burdens and costs on the economic system resulting in
adverse effects on economic growth. At the other extreme, there are points of view that
attribute to the government a non-negligible role to play in the process of economic
development and argue that a larger govenmment size is likely to have a positive impact on
long-run economic growth. One argument is that government plays a critical role in
harmonizing conflicts between the private and social interests and in providing a socially
optimal direction for growth. Moreover, in countries characterized by the existence of
monopolies, the lack of fully developed markets of capital, insurance, and information,
govemment can make factor and product markets work more efficiently and generate
substantial spillover effects for the private sector.
On the empirical side, most investigations of the impact of government spending on
economic growth were conducted using cross-country models. These studies attempted to
link aggregate measures of fiscal policy with average growth rates of GDP using measures of
the overall size of the govemment in the economy, disaggregate measures of government
spending, or measures of the growth rate of government expenditures. The main
contributions to this literature were made by Rubinson (1977), Feder (1983), Landau (1983),
Kormendi and Meguire (1985, 1990), Ram (1986), Grier and Tullock (1989), Grossman
(1988), Barro (1990, 1991), Romer (1989), Levine and Renelt (1992), and Barro and Sala-i-
Martin (1995), among others.
However, as far as economic policy is concemed, the cross-country literature did not
provide the basis on which a particular country can decide whether an expanding
government size would accelerate or depress economic growth, and that for many reasons.
First, the cross-country literature on the subject produced conflicting results and did not
provide any consensus judgment on the impact of government spending on economic
growth. For example, Landau (1983), Grier and Tullock (1987), Grossman (1988), and
Barro (1990) found a negative and significant relationship between the growth rate of real
GDP and the growth rate of the government share in GDP, while Ram (1986) and Rubinson
(1977) suggest that govenmment size has a positive impact on economic growth. Kormendi
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Government Spending 21
and Meguire (1985), however, found no significant relationship between the share of
government consumption in GDP and economic growth.
Second, since economic policies tend to be country-specific and their success depends
largely on the institutions implementing them, policy recommendations cannot safely be
made by generalizing the results of cross-country models to individual countries. This is
because the pooling process in cross-country regressions imposes strong parametric
restrictions across countries that differ greatly in terms of their economic structure. Pooled
estimates in these models are obtained by pulling the extreme values of the parameters
towards a common mean. This view seems to be consistent with the recent studies by
Maddala and Wu (2000) and Lee et al. (1996) who provided evidence on the problems of
heterogeneity, stability, and interpretation inherent to cross-country models. Moreover, it is
fair to argue that, while a concept such as the long run is a dynamic one, cross-country
models do not include sufficient time and, hence, are not useful for analyzing the dynamic
properties of growth. Therefore, it is hazardous to draw policy conclusions from cross-
country studies and a better strategy would be to investigate the issue in a time-series setting
where the particular characteristics of the country under consideration are taken into
account.
Finally, an important issue that has often been neglected in the literature is the impact of
the source of financing government expenditures on economic growth. In this respect, it is
important to notice that cross-country studies examine the impact of government spending
on economic growth by looking at the sign and significance of the relevant variable in the
growth equation. In this context, government spending would have an accelerating or a
depressing effect on growth depending on whether its sign is positive or negative. Following
this literature, a policy conclusion would be for the government to withdraw from any
activity that has a significant negative impact on growth. However, one could argue that the
negative impact may be the result of the way this activity was financed and not the activity
itself. For instance, a growing govenmment size may affect economic growth in two opposite
ways depending on whether a debt-financed or a tax-financed expansionary fiscal policy is
followed. Theoretically, there is a controversial debate on whether a tax- or a debt-financed
government budget has the same effects on growth. The two competing views known as the
Ricardian equivalence proposition and the traditional view are well documented in the
literature but did not provide any consensus judgment on the issue. Examples are Leiderman
and Blejer (1988), Kormendi and Meguire (1990), and Dalamagas (1992). Consequently,
the question of whether the tax collection policy or the debt accumulation process is
responsible for the distortionary effects of government expenditures on economic growth in
still an open question in countries where the issue has not been addressed before.
The objective of this paper is to investigate whether changes in government spending
constitute an effective policy for promoting economic growth in the small, developing
economy of Tunisia. The analysis focuses on two main issues of interest: (i) investigating the
causal channels through which govenmment spending can affect economic growth; and (ii)
investigating the impact of alternative sources of budget financing on the growth
performance. Upon recommendations by the IMF and the World Bank, Tunisia started
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

22 Khalifa Ghali
large-scale programs of restructuring its public sector in an attempt to reduce the size of the
government in the economy and, hence, accelerate economic growth. However, there is no
empirical evidence that provides policymakers with information concering the particular
causal patterns that link public sector activities to economic growth. Although some of the
past studies have analyzed issues related to growth, economic development, and economic
policymaking in Tunisia, none of them has empirically assessed the impact of government
activities on economic growth.' Examples of these are: Nsouli et. Al. (1993), World Bank
(1996), Ilker and Ghiath (1999) and Ghali (1999). The availability of such infonnation
would provide policymakers in this country with insightful indications on the mechanisms
through which fiscal policy and macroeconomic variables interact in affecting the growth
process. Furthermore, the analysis would help the Tunisian fiscal authorities to assess the
impact of altemative sources of financing the government budget on economic growth.
After reviewing some specification issues raised by past empirical research we model
the dynamic interactions between govermment spending and economic growth in a six-
variable system consisting of real GDP, govemment spending, private investment, exports,
imports, and labor. In particular, we use the Johansen (1988, 1991, 1992) and Johansen and
Juselius (1990) cointegration techniques and develop a vector error-correction (VEC) model
useful for identifying all possible channels through which government spending may affect
economic growth.
The results of the empirical analysis can be summarized as follows: (i) there is a
meaningful, stable long-run relationship that ties the long-run behavior of real output to that
of governent spending, investment, exports, imports and labor; (ii) there are four possible
channels through which government spending can affect economic growth in Tunisia. In
addition to its direct effects, government spending has also indirect effects on growth and
that through investment, imports and labor; (iii) government activities designed to shape the
general efficiency of the economy, and to improve the productivity of labor and imports
seem to have an accelerating effect on growth, whereas government involvement in the
economy as a producer seems to crowd-out private investnent from profitable opportunities,
resulting in a depressing effect on economic growth; and (iv) the government reliance on
debt to fmance its budget seems to induce adverse effects on economic growth. The
remainder of the paper is organized as follows. Section II presents the model and the
method, section III presents the empirical results and section IV concludes.
THE MODEL AND THE METHOD
While the main focus of the paper is on the relationship between govern
and economic growth, we also consider some specification issues raised by past empirical
research, especially those raised in cross-country growth studies. The main reason for this is
the non-existence of a consensus theoretical framework that guides empirical research on
growth. In addition, the existing literature using cross-country growthl models does not
completely specify the variables that should be held constant when inferences are made
about the relationship between economic growth and the variables of interest. For example,
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Feder (1983) and Ram (1986) use an augmented neoclassical production fimction to conduct
their empirical studies, while Romer (1989) and Barro (1990) use endogenous growth
models. Kormendi and Meguire (1985) and Grier and Tullock (1989) use a variety of
models to justify an assortnent of variables that they use in their empirical research. In our
empirical investigations of the causal links between govenmment spending and economic
growth we will also consider variables such as international trade, investment, and labor.
Apart from their theoretical relevance, the inclusion of such variables would also avoid, or at
least minimize, the problem of non-causality due to omitted variables (Lutkepohl, 1982).
The theoretical ties between international trade and economic growth were formalized
by Romer (1986, 1990b), Grossman and Helpman (1990), and Rivera-Baptiz and Romer
(1991), among others. Although theoretical discussions frequently consider the relationship
between international trade and economic growth, the empirical literature has typically
examined the relationship between exports and growhi.2 In our empirical investigations of
the dynamic properties of growth in Tunisia we consider the effects of both imports and
exports on the economic performance of the country.
Another issue that we consider here is the interaction between growth, investment and
trade. Theoretically, the relationship between trade and growth is based on the improved
allocation of resources and not necessarily on enhanced resource accumulation. However,
results from Levine and Renelt (1992) indicate that the relationship between growth and
trade becomes insignificant when investment is introduced in the growth equation, whereas
the relationship between investment and the trade variables is a significant one. Given that
investment was found positive and significant in the growth equation, they suggest the
existence of a two-link chain between trade and growth through investment and conclude
that the relationship between trade and growth is based on enhanced resource accumulation
and not the improved allocation of resources.
Given the importance of these issues, our investigation of the causal links between
government spending and economic growth in Tunisia is made witiin a 6-variable system
containing real GDP, government spending, investment, exports, imports and labor.
The empirical analysis of causality is conducted witiin a framework based on unit root
testing and cointegration. In particular, we develop a vector error-correction model (VECM)
of growth where all variables are allowed to be endogenous to the system. The advantage of
this approach, as opposed to single growth equation estimation, is the possibility to capture
both the direct as well as the indirect effects of govemment spending on economic growth.
Withiin this framework, we first investigate all the possible charmels through which
government spending can affect economic growth, and then we investigate the impact of
alternative sources of financing the govemment budget on economic growth.
A VECTOR ERROR-CORRECTION MODEL OF GROWTH
Consider the set of variables consisting of real GDP and the respective s
private investment, total government expenditures, exports, imports and la
are transformed into their natural logarithm. These variables will be denoted h
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

24 Khalifa Ghali
I, G, X, M and L, respectively. Now consider the vector autoregressive (VAR) model
A- 1 t-I + 2 Zt-2 +- ...+ 4kt-k + + t nt, t1,....,T (1)
where Zj is a 6 x 1 vector containing Y I, G, X, M and L. Sup
I(O) after applying the differencing filter once. If we exploit the
movements of these variables and possibilities that they will tr
run equilibrium state then, by the Granger representation th
following testing relationships that constitute a vector error
growth
AZ = r A 1t-kI +sit, t l. ,T (2)
where AZ4 contains the growth ra
is a difference operator, lit is a v
movements in Z, with flt niid (0
cointegrating vectors (1? r < 6), D
and f3 both 6 x r matrices. , are t
are the adjustment coefficients whic
VEC model.
The Johansen (1988) approach estimates the long-run or cointegrating relationships
between the non-stationary variables using a maximum likelihood procedure which tests for
the cointegrating rank r and estimates the parameters f3 and a.
On expanding out equation 2, the VEC model of growth can be expressed as follows:
r p
AYt =pLi +Jlx,,vkt_p +Y (yi,sAY
k=l s=l
r p
AGt =p2 + Ya2,kvkt-p +(O,sA
k=l s-l
r p
Alt =p3 +EVtp + (8sAYt-s+
k=l s-l
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

r p
AXt =W +Ea4.kvk,t-p +E (4LsAYt-s+
k=l s=l
r p
AMt =ps +Ea5kVkt-p +E (l,sAYt
k=l s=l
r p
AL =p6 + X,kVkt-p +E (ki,sAYt
kl s=l
where aXi,k are the adjustment coeffici
are intercepts.
The advantage of modeling the relationship between government spending and
economic growth in a system of equations as in (3) - (8), instead of a single growth equation,
is the possibility to identify both the direct as well as the indirect links between them. The
direct effect of government spending on economic growth, which is conventionally tested in
a single growth equation, can be detected from equation (3) through the distributed lags in
government spending. The additional equations (4) - (8) offer the possibility to identify those
indirect effects that may result from the impact of govemment spending on any of the
remaining variables that has, in turn, a direct link with economic growth. That is, the
interaction of government spending with investment, trade and labor may result in an
indirect impact on economic growth.
EMPIRICAL RESULTS
Data and Variable Definitions
Data used in the study are annual series for Tunisia and cover the period 1960-2000.
The variables and their definitions where In denotes the natural logarithm are as follows:
Y = ln(real GDP),
I = ln(the ratio of private investnent to GDP),
G = ln(the ratio of total government spending to GDP),
X = ln(the ratio of exports to GDP),
M = ln(the ratio of imports to GDP),
L = ln(labor force),
D = ln(the ratio of government debt to GDP), and
T = In(the ratio of tax revenues to GDP).
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

26 Khalifa Ghali
GDP is gross domestic product, real GDP is GDP deflated by the GDP deflator. Private
investment is measured by the total real gross private-business investment. Total government
spending is measured by government current and capital expenditures. Exports and imports
refer to total exports and total imports. The labor force refers to the total labor force
employed in the different sectors of the economy, including agricultural employment.
Government debt refers to the budget deficit of the government, and tax revenues are the
total taxes collected by the government.
The first difference of Y is taken to be a proxy for economic growth. Data on GDP,
exports, and imports are gathered from various issues of the IMF publication International
Financial Statistics (IFS). Data on private investment, labor, government debt, and tax
revenues were obtained from the Tunisian Ministry of Economy.
TEST RESULTS FOR UNIT ROOTS
Since the VAR specification in (1) requires that some or all variables are integrated of
order one, I(1), we herein investigate the stationarity status of all the series defined above
using the augmented Dickey-Fuller (ADF) and the Phillips-Perron (P-P) tests of stationarity.
These tests are performed both on the level as well as on the first difference of the variables
allowing for a time trend whenever is its significant. In each case the lag-length is chosen
using the Akaike Information Criterion (AIC) after testing for first and higher order serial
correlation in the residuals.
Table 1 reports the results of the unit root tests. The first half of the table contains the
results of testing for unit roots in the level variables. These results suggest that each of the
time-series has a unit root. The second half of the table reports tests for unit roots after
differencing the data once. In this case both tests reject the null hypothesis of unit root.
Because the data appear to be stationary in first differences, no firther tests are performed.
Therefore, we maintain the hypothesis that each of the series is integrated of order one.
Table 1. Test Results for Unit Roots
Variable Y I G X M L D T
ADF -1.182 -1.714 -1.942 -1.642 -1.390 -1.563 -1.992 -1.565
P-P -1.212 -1.699 -1.640 -2.300 -1.558 -1.436 -1.966 -1.601
Variable AY Al AG aX AM AL AD AT
ADF -4.562 -3.829 -3.864 -4.605 -3.752 -3.296 4.514 -3.692
P-P -6.691 4.718 4.628 -5.176 -5.031 -4.672 -6.754 4.930
NOTES: Variables are as defined in the text. The above tests are performed using the following
regression
p
Ayt= po+ plYt-I p2T dsAYt,s+vt,
s=I
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Govenment Spending 27
where yt is the relevant time series, T is a time trend, and vt is a residual term. The time tend in this
regression is allowed only when found to be significant. The lag-length p is chosen using the AIC
criterion after testing for first and higher order serial correlation in the residuals. The critical values for
ADF and P-P are -2.967 and -2.970, respectively.
TEST RESULTS FOR COINTEGRATION
Before applying the Johansen procedure to estimate a and f3, it is5necessary to
determine the lag length, k, of the VAR, equation (1), which should be high enough to
ensure that the errors are approximately white noise, but small enough to alIow estimation.
Since the Johansen procedure is sensitive to the choice of the lag length, we based our
decision on the Akaike' s Final Prediction Error (FPE) criterion and selected k-=3. Using this
lag length, we conducted diagnostic checking tests for normality and serial correlation in the
residuals for each of the 6 equations in VAR. The results of these tests, not reported here,
indicate that with k--3 the residuals are approximately independently identically normally
distributed (niid) for all the equations.
The results of testing for the number of cointegrating vectors are reported in table 2,
which presents both the maximum eigenvalue (X.,) and the trace statistics, the 10 percent
critical values as well as the corresponding X values. This test is performed using an
unrestricted intercept term in VAR, which assumes the existence of a deterministic time
trend in the data. The results of the test indicate the existence of one cointegrating vector
driving the data. The check the robustness of this result, we jointly test the hypotheses that
the rank is equal to one and that the data contain a time trend. We do this using the Johansen
(1992) likelihood ratio test. The results of this test, not reported here, confirm that the rank
of D is equal to one and that the data do contain a time trend component. Consequently, the
intercept term will be associated to the VEC equations and not to the cointegration space.
The estimates of the P and a and, vectors are presented in table 3. From the a vector
we can see that the adjustment coefficient of labor is small and insignificant. Testing that this
coefficient is zero is a test that labor is weakly exogenous and yields a likelihood ratio test =
1.64, which compared to the 5% critical value x2(1) = 3.84 enables us to easily accept the
null hypothesis. Estimates of a and p after imposing the weak exogeneity restriction on L
are presented in table 4. The results of this table indicate that the model is now completely
identified.
The results of cointegration indicate the existence of a meaningful, stable long-run
equilibrium relationship that ties together the long-run behavior of re4 output to that of
investment, government spending, exports, imports, and labor. With this, the short-run
dynamics of output can be seen as an adjustment to this long-run equilibrium state. In each
short-term period, deviations of output from this equilibrium will feed back on its changes in
order to force its movement towards its long-run equilibrium. From the a vector in table 4
we can see that real output in Tunisia has a high speed of adjustment to its long-run
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

28 Khalifa Ghali
equilibrium state. The short-mn behavior of output is being corrected by about 62% of the
imbalance occuring in the previous period between output and its long-rn equilibrium
value.
Table 2. Testing the Rank of n
Trace a
-HO HI Stat. 90% HO HI Stat. 90%
r=0 r> 1 93.99 89.37 r 0 r= 1 35.75 24.63 0.355
r <1 r >2 42.24 64.74 r <i r 2 14.21 20.90 0.288
r:< 2 r > 3 25.03 43.84 r? 2 r = 3 9.03 17.15 0.178
r < 3 r > 4 15.03 26.70 rc3 r = 4 3.98 13.39 0.125
r < 4 r > 5 8.03 13.31 r < 4 r 5 2.05 10.60 0.088
r <5 r >6 1.03 2.71 r<~5 r =6 1.03 2.71 0.064
Table 3. The p and a and Vectors
Variable Yt it Gt Xt Mt h
p 1 -0.359 -0.354 -0.471 0.278 -1.540
(-11.120) (-3.411) (-7,101) (2.856) (-20.078)
-0.622 0.528 0.357 0.441 -0.552 0.026
(4.301) (3.971) (4.020) (3.227) (-2.630) (1.032)
NOTES: The cointegrating vector is normalized on Y and the figure
Table 4. The Restricted 0 and ax Vectors
Variable Yt it G' X Mt Lt
,B 1 -0.367 -0.351 -0.470 0.277 -1.546
(-11.121) (-3.408) t-7.015) (2.856) (-20.078)
-0.620 0.528 0.35"7 0.441 -0.558
(-4.320) (3.970) (4.024T (3.229) (-2.633)
NOTES: The cointegrating vector is normnalized on Y and the figures in p
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Govemment Spending 29
Table 5. Estimates of the VEC Model
Variable AYt Alt AGt AXt AMt ALt
Intercept 1.4003 2.601 -0.455 0.104 1.277 0.563
(3.116) (2.891) (-2.083) (1.998) (2.441) (3.119)
Vt -I -0.620 0.528 0.357 0.441 -0.558
(-4.320 (3.970) (4.024) (3.229) (-2.633)
AYt -i -0.315 0.110 0.522 -0.344 0.237 2.916
(-2.788) (2.661) (2.662) (-0.081) (2.670) (3.530)
AYt -2 0.517 0.088 0.407 0.453 0.102 1.881
(3.116) (1.177) (2.011) (2.013) (2.455) (3.016)
Alt i -0.027 0.322 0.202 0.219 0.417 -0.082
(-0.770) (2.801) (1.708) (2.337) (3.058) (-1.456)
AIt-2 0.319 0.221 0.223 0.045 0.214 1.557
(3.001) (1.701) (1.421) (1.116) (2.309) (1.490)
AGti 0.261 -0.231 0.233 -0.241 0.216 2.056
(2.702) (-3.005) (1.094) (-1.443) (2.913) (5.733)
AGt-2 0.075 -0.066 0.118 -0.070 0.083 1.408
(2.290) (-2.183) (1.002) (-0.844) (1.054) (3.339)
AXt -i -0.007 0.338 -0.127 -0.031 -0.040 -0.976
(-0.113) (2.529) (-0.728) (-0.112) (-0.836) (-1.002)
AXt-2 0.216 0.018 -0.032 -0.012 -0.019 -0.620
(2.768) (1.327) (-0.401) (-0.304) (-0.026) (-0.839)
AMti ,0.026 0.671 -0.004 -0.108 -0.412 -0.540
(1.009) (3.593) (-1.201) (-0.313) (-1.225) (-1.036)
AMt -2 0.418 0.212 -0.013 -0.033 -0.327 -0.033
(3.811) (2.642) (-0.801) (-0.019) (-0.981) (-0.025)
ALT-i -0.075 0.311 0.381 0.015 -0.028 -0.753
(-1.103) (2.818) (2.552) (1.002) (-0.985) (-2.566)
ALt-2 0.251 0.115 0.227 0.012 -0.016 -0.216
(3.716) (2.091) (2.016) (0.974) (-0.335) (-1.499)
R2 0.78 0.69 0.42 0.26 0.21 0.36
a 0.0164 0.0428 0.0477 0.0772 0.0981 0.0544
TSC(10) 6.12 2.701 8.802 6.551 7.733 4.102
N(2) 1.041 0.415 0.453 1.399 1.520 0.836
RESET(1) 0.236 0.813 2.487 0.562 1.644 0.721
NOTES: T-ratios are in parentheses. TSC(10) is a test for up to t
is the Jarque and Bera test for normality, and RESET(1) is a test f
Table 5 reports the OLS estimates of the vector er
cointegration vector in Table 4. Since these estimates are
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

30 Khalifa Ghali
standard assumptions, we subjected the estimated residuals in each equation to a battery of
diagnostic tests. The first test, TSC(10) is a test for up to the tenth order serial correlation in
the residuals, which is distributed X2(10). The second test, N(2), is the Jarque & Bera
normality test whichis asymptotically distributed Z2(2). The third test, RESET(l), is the
RESET test for parameter instability which is asymptotically distributed X2(1). The results
in table 5 suggest that residuals from the 6 equations pass the tests at the 95% significance
levels and, hence, there is no significant departure from the standard assumptions.
TESTING FOR GRANGER NON-CAUSALITY
The recent literature has shown that the conventional testing procedure for gran
causality using the F-statistic has size and power problems due to its, dependen
pretesting for cointegration (Zapata and Rambaldi, 1997). A much accurate and si
procedure was proposed independently by Dolado and Lutkepohl (1996) and Toda and
Yamamoto (1995) and is known as the augmented VAR approach. As shown by Zapata and
Rambaldi (1997) and Giles and Mirza (1999), the augmented VAR testing procedure is very
simple to compute and is independent of the cointegration properties of the data. For the
purposes of this paper, this procedure goes as follows:
1. Since the VAR model contains 3 lags and since the highest order of integration
in the data is one, we first estimate a VAR in levels with 4 lags, then
2. We test jointly that the first 3 lags of the relevant variable are zero using a Wald-
test which has a chi-squared distribution.
Table 6. Test results for Granger non-causality
Variable AY Al AG AX AM AL
AY 24.68* 18.30* 15.67* 17.45* 1 1.21*
Al 18.45* 2.01 13.87* 14.11* 3.17
AG 16.55* 15.67* 2.05 13.81 * 11.77*
AX 21.36* 13.07* 1.19 2.01 2.66
AM 16.47* 10.33* 1.56 0.88 -- 1.16
AL 12.22* 15.09* 13.61* 1.30 2.47
NOTES. The flow of causality in the table is from the v
the first row. A star * indicates that the null hypot
significance level.
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Table 6 reports the results of testing for Granger non-causality between the 6 variables
in the system. The flow of causality in the table is from the variables in the first column to
the variables in the first row. In this setting, table 6 identifies all the possible channels
through which government spending affects economic growth. The main conclusions from
table 6 can be summarized as follows:
1. There are four possible channels through which government spending can affect
the growth process in Tunisia. From table 6, we can see that government spending
affects economic growth directly through the growth equation and indirectly
through its impact on investment, imports, and labor. Since government spending
Granger- causes investment, imports and labor, which in turn Granger-cause
economic growth, then government spending has an indirect causal impact on
economic growth through its effects on these variables. The advantage of
identifying these indirect causal links is the result of modeling growth dynamics
within a system of equation rather than a single growth equation.
2. Looking at the results of tables 5 and 6 together, we can identify three causal
channels through which an expanding government size has an accelerating effect
on economic growth in Tunisia. These are promotion of economic growth,
imports, and labor. This result seems to be consistent with the Tunisian reality.
The Tunisian government's intervention in economic activity has for long played
a critical role in promoting the country's economic growth and development. In
particular, the government was heavily involved in promoting social policies and
economic infrastructure. Government spending on social policies played a
critical role in promoting the standard of living. The provision of free basic
health care, education, and training were necessary for the development of the
country's human capital. Given the relatively low contribution of the private
sector, the government remains the main importer of technology. In addition, the
provision of an adequate infrastructure along with political stability has attracted
foreign direct investment in many key sectors of the economy.
However, government spending is having an indirect negative effect on economic
growth and that through its negative impact on private investment. This could be
the result of the heavy involvement of the Tunisian government in many productive
sectors of the economy, which crowded-out the private sector from profitable
opportunities such as in banking, agriculture, energy, telecommunications, and
transport.
3. As to the relationship between growth, investment and trade, our time-series results
do not support the results of Levine and Renelt (1992). In our case, trade and
investment are both significant in the growth equation. In addition both trade
variables are significant in the investment equation. Hence, the relationship
between trade and growth in the case of Tunisia seems to be based on both the
enhanced resource accumulation as well as on the improved allocation of
resources.
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

32 Khalifa Ghali
4. Finally, regarding the relationship between economic growth and labor, our
results suggest that labor is an important input into production in the case of
Tunisia. Labor is highly significant in both the long-run relationship as well as in
the growth equation. This result contrasts with the results of the panel-data studies
investigating growth convergence. In reference to the study by Islam (1995) as a
typical example of the available panel data studies, one of his main conclusions is
that there is no room for labor accumulation in a growth equation.3 This is
because labor appears to be always insignificant after controlling for the fixed
effects. However, our result seems to be rather supportive of the arguments
advanced by Maddala and Wu (2000) against the assumption that the theory
implicit in Islam's model applies to all countries.
In light of these results, the recommendation by the IMF and the World Bank that
Tunisia should reduce the size of its government in order to promote economic growth does
not seem to be the appropriate one.4 An appropriate policy recommendation would be for the
Tunisian government to enhance its involvement in those activities that shape the general
productive efficiency of the economy and to withdraw from those activities that are hamful to
private investment and economic growth. In particular, Tunisia can build on its relative
strengths in terms of the young profile of its population, its stability, prosperity, and its closer
integration with Europe to further facilitate the country's absorption of new technologies and
new ways of doing business. The government should, however, reorient its development
strategy from one that heavily depends on state-owned monopolies to one that encourages a
greater participation of the private sector in most aspects of economic activities. The
privatization and deregulation of public enterprises along with the enforcement of regulations
encouraging profitable opportunities are the sorts of measures that would stimulate private
investment and allow the economy compete in a global environment.
BUDGET FINANCING AND ECONOMIC GROWTH
We have so far focused on assessing the impact of changes in government si
growth performance in Tunisia. In this respect, we clearly identified the channels
which government activities can promote economic growth. However, given the bur
govemment budget constraint as being a binding factor in formulating governmen
a required additional information is for policymakers to know whether this outcome
the same irrespective of the method used to finance government expenditures. Fo
now turn to investigating the impact of the different sources of financing the g
budget on economic growth.
Given the flow budget constraint of the govemment, G = T + D, we now decom
government expenditures, G, in our VEC model into two components; the ratio
revenues to GDP (T) and the ratio of govemment debt to GDP (D). With this, we now have a
7-variable system. Using the same methodology, we first tested for cointegration between the
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

variables and then estimated a VEC model in order to assess the separate effects of tax
revenues and government debt on economic growth.
Table 7 reports the growth equation from the estimated VEC model and table 8 reports
the results of testing for Granger-causality between economic growth, tax revenues, and
government debt. As shown in these tables, while tax-revenues and debt both Granger-cause
economic growth, they have opposite sign effects in the growth equation. Both coefficients
of the debt ratio are negative and statistically significant. In contrast, the coefficients of the
tax ratio are both positive but only the first coefficient is statistically significant.
Table 7. Effects of Debt and Tax Revenues on Economic Growth
Variable AYt t-statistic
Intercept 1.238 (2.011)
vt, -0.346 (4.706)
AYt- l-0.329 (-1.206)
Yt-2 0.436 (3.078)
Alt-i -0.038 (-0.524)
AIt-2 0.326 (3.886)
ADt- -0.205 (-3.541)
ADt-, -0.134 (-2.007)
ATt-I 0.688 (2.364)
ATt-2 0.256 (0.129)
AXt-i -0.037 (-0.155)
AXt-2 0.016 (3.001)
AMt- l0.103 (1.272)
AMt-2 0.228 (2.344)
A Lt- -0. 103 (- 1.1 05)
Al-t2 0.306 (2.776)
R 2 0.86
a 0.015
TSC(10) 5.32
N(2) 1.778
RESET(1) 0.452
NOTES: T-ratios are in parentheses. TSC(10
is the Jarque and Bera test for normality, a
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

34 Khalifa Ghali
Table 8. Testing The Causal Impact of Tax Revenues and Debt on Economic Growth
Null Hypothesis Test Statistic x2(3)
Debt does not Granger-cause economic 19.45 7.81
growth
Tax revenues do not Granger-cause 14.13 7.81
economic growth
In light of these results, it seems that the method o
not inconsequential on the growth performance in the case of Tunisia. The govemment
reliance on debt seems to induce adverse effects on economic growth, while tax-financed
government expenditures seem to have an accelerating effect on growth. With this, the
positive effects of government spending on economic growth that we found previously (table
5) should be the outcome of tax-revenue effects offsetting the negative effects of government-
debt on growth. Hence, an additional useful policy recommendation is for the Tunisian fiscal
authorities to adopt a strategy of fiancing public outlays by relying more on tax-revenues
than on debt-accumulation.
CONCLUSION
Following the debt crisis, Tunisia was among many developing countries tha
IMF debt-stabilization programs without having any indication on how governme
interacts with macroeconomic variables in affecting its growth process. In addition, the
existing literature did not provide a consensus judgment on which a policy can be safely
formulated. Consequently, this paper attempted to empirically investigate the causal channels
through which government spending can affect economic growth in Tunisia and the extent to
which govemment activities can have an accelerating or a depressing effect on the growth
performance in this.country. An additional issue that we investigate, but which was neglected
in the relevant empirical literature, is whether the impact of government spending on growth
is independent of the source of financing the government budget.
In an attempt to improve upon the conventional procedure of estimating a single growth
equation, the paper developed a vector error-correction (VEC) model that allows to identify
not only the direct effects of government spending on growth but also the indirectly induced
effects resulting from the interaction of government spending with other macro-variables such
as investment, trade and labor. With this, we were able to identify- three channels through
which govenmment spending may have an accelerating effect on economic growth and one
channel through which govemment spending may have a depressing effect on economic
growth.
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

The main conclusions useful for the conduct of economic policy in Tunisia are that
government spending aimed at shaping the general efficiency of the economy and promoting
the productivity of labor and imports have accelerating effects on economic growth. Whereas
government involvement in the productive sectors of the economy has a crowding-out effect
on private investment and, hence, depresses economic growth. In addition, given the adverse
effects of debt-accumulation on growth performance, it is recommended that fiscal
authorities in Tunisia adopt a strategy of financing public activities by relying more on tax-
revenues than on government debt.
ENDNOTES
IFor detailed background descriptions of the Tunisian economy, interested rea
to World Bank (1996).
2Several studies of fiscal policy have excluded trade indicators from their analysis. These
include Landau (1983), Ram (1986), Grier and Tullock (1989), and Barro (1990, 1991).
Other studies have ignored fiscal policy when studying trade policy. These include Feder
(1983) and Edwards (1989). Studies that included both exports and imports include
Kormendi and Meguire (1985), Romer (1990a), and Levine and Renelt (1991, 1992).
3See also Knight et al. (1993), Easterly et al. (1993), and Cashin and Loayza (1995).
4The World Bank (1996) Progress Report referred to the results of cross-country growth
studies and recommended that Tunisia should reduce the size of its govermment.
REFERENCES
Barro, Robert J. (1990), A Government spending in a simple model
growth,@ Journal of Political Economy, Vol. 98, No. 2, pp.103-25.
Barro, Robert J. (1991), A Economic growth in a cross section of countries, @
Quarterly Journal of Economics, No. 106, pp.407-44.
Barro, Robert J. and Xavier Sala -i-Martin (1995) Economic Growth, McGraw-Hill, New
York.
Cashin, Paul and Norman Loayza (1995), A Paradise lost? Growth, convergence and
migration in the south pacific, @ IMF StaffPapers, Vol.42, pp.608-41.
Dalamagas, Basil (1992), A How rival are the Ricardian equivalence proposition and
the fiscal policy potency view? @ Scottish Journal of Political Economy, Vol. 39,
pp.457-76.
Dolado, Juan J. and Helmut Lutkepohl (1996), A Making Wald Test Work for
Cointegrated VAR systems, @ Econometric Reviews, Vol. 15, pp.369-386.
Easterly, William, Kremer M., Pritchett L. and Summers L.H. (1993), AGood policies or
good luck? Country growth performance and temporary shocks, ( Journal of
Monetary Economics, Vol. 1 10, pp. 1 127-70.
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

36 Khalifa Ghali
Edwards, Sebastian (1989), A Openness, outward orientation, trade liberalization and
Economic performance in developing countries,? @ World Bank Working Paper
Series No. 191.
Feder, Gene (1983), A On exports and economic growth, @ Journal of Development
Economics, Vol. 12, No. 3, pp.59-74.
Ghali, Khalifa H. (1999), A Financial development and economic Growth: The Tunisian
Experience, @ Review ofDevelopment Economics, Vol. 3, No. 3, pp.3 10-322.
Giles, Judith and SadafMirza ( 1999), A some Pretesting Issues on Testing for Granger
Noncausality, ( Working Papers in Econometrics. University of Victoria. EWP9914.
Grier, Kevin and G. Tullock (1989), AAn empirical analysis of cross-national economic
growth, 1951-1980, @ Journal ofMonetary Economics, 24 (2), 259-76.
Grossman, P. (1988), A Growth in government and economic growth: the Australian
experience, @ Australian Economic Papers, Vol. 27, pp.3343.
Grossman, P. and Elhanan Helpman (1990), ATrade, innovation, and growth, @ American
Economic Review, Vol. 80, No. 5, pp.86-91.
Ilker, Domac and Shabsigh Ghiath (1999), A real Exchange Rate Behavior and Economic
Growth: Evidence from Egypt, Jordan, Morocco, and Tunisia, IMF Working Paper No.
99/40, Middle Eastem Department, International Monetary Fund, Washington D.C.
Islam, Nazrul (1995), A Growth empirics: a panel data approach, @ Quarterly Journal of
Economics, Vol. 1 10, pp. 1 127-70.
Johansen, S0ren (1988), A Statistical analysis of cointegration vectors?, @ Journal of
Economic Dynamics and Control, Vol. 12, pp.231-54.
Johansen, S0ren (1991), A Estimation and hypothesis testing of cointegration vectors in
gaussian vector autoregressive models?, @ Econometrica, Vol. 59, pp. 1551-80.
Johansen, S0ren (1992), A Determination of the cointegration rank in the presence of a
linear trend?, @ Oxford Bulletin of Economics and Statistics, Vol. 54, pp.383-97.
Johansen, S0ren and Katalina Juselius (1990), A Maximum likelihood estimation and
inference on cointegration with applications to the demand for money?, @ Oxford
Bulletin of Economics and Statistics, Vol. 52, pp. 169-210.
Knight, Malcolm, Loayza Norman and Delano Villaneuva (1993), A Testing the
neoclassical theory of economic growth: a panel data approach, @ IMF StaffPapers,
Vol. 40, pp.51241.
Kormendi, Roger and Philip Meguire (1985), A Macroeconomic determinants of growth:
cross-country evidence @Journal of Monetary Economics, Vol. 16, No. 4, pp.141-
63.
Kormendi, Roger and Philip Meguire (1990), AGovemment debt, government spending and
private sector behaviour: reply and update, @ American Economic Review, Vol. 80,
pp. 04-17.
Landau, Daniel (1983), A Government expenditure and economic growth: a cross-country
study, @ Southern Economic Journal, Vol. 49, No. 3, pp.783-92.
Lee, Michael, Richard Longmire, Laszlo Matyas and Mark Harris (1996), A Growth
convergence: some panel data evidence, @ Applied Economics, Vol. 30, pp.907-12.
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Govermnent Spending 37
Leiderman, Leonardo and Mario Blejer (1988), A Modeling and testing Ricardian
equivalence, @ IMF StaffPapers, Vol 35, pp 1-35.
Levine, Ross, and David Renelt (1992), AA sensitivity analysis of cross-country regressions,
( The American Economic Review, Vol 82, No. 4, pp.942-63.
Lutkepohl, Helmut (1982), A Non-causality due to omitted variables, ( Journal of
Econometrics, Vol. 19, pp.367-78.
Maddala, G. S. and WU, S. (2000) A Cross-country growth regressions: problems of
heterogeneity, stability and interpretation, @ Applied Economics, Vol. 32, pp.635-42.
Nsouli, Salah, Sena Eken, Paul Duran, Bell Gerwin and Zuhtu Yucelik (1993), AThe Path to
Convertibility and Growth: The Tunisian Experience, ( IMF Occasional Paper,
No. 109, International Monetary Fund, Washington D.C.
Ram, Rati (1986), AGovenmment size and economic growth, @ American Economic Review,
Vol. 76, No. 6, pp.191-203.
Rivera-Batiz,L. and Paul Romer, M. (1991), AEconomic integration and endogenous growth
@ Quarterly Journal of Economics, Vol. 106, pp.531-56.
Romer, Paul M. (1989), Ahuman capital and growth: theory and evidence, @ National
Bureau of Economic Research (Cambridge, MA), Working Paper No. 3171.
Romer, Paul M. (1990a), A Capital, labor, and productivity, @ Brookings Papers on
Economic Activity, special issue, pp.337420.
Romer, Paul M. (1990b) A Endogenous technological change, @ Journal of Political
Economy, Vol. 98, No. 2, pp.71-102.
Rubinson, Richard (1977), A Dependency, government revenue, and economic growth, (
Studies in Comparative International Development, Vol. 12, No. 4, pp.3-28.
Toda H. Y. and T. Yamamoto (1995), AStatistical Inference in Vector Autoregressions with
Possibly Integrated Processes, @ Journal of Econometrics, Vol. 66, pp.225-250.
World Bank (1996) Tunisia's Global Integration and Sustainable Development: Strategic
Choicesfo r th e 21st Century, World Bank Middle East and North Africa Economic
Studies, Washington D.C.
Zapata, H. 0. and Alicia N. Rambaldi (1997), AMonte Carlo Evidence on Cointegration and
Causation, @ Oxford Bulletin of Economics and Statistics, Vol. 59, pp.285-298.
197.28.119.186 on Mon, 03 Jul 2023 11:58:16 +00:00

Gooodhali-GovernmentSpendingBudget-2003.pdf

More Related Content

Similar to Gooodhali-GovernmentSpendingBudget-2003.pdf(20)

Recently uploaded(20)

Gooodhali-GovernmentSpendingBudget-2003.pdf