Академический Документы
Профессиональный Документы
Культура Документы
net/publication/5105114
CITATIONS READS
7 187
1 author:
Fumitaka Furuoka
University of Malaya
105 PUBLICATIONS 589 CITATIONS
SEE PROFILE
All content following this page was uploaded by Fumitaka Furuoka on 23 October 2014.
Introduction
In the process of transforming Malaysia into a fully-industrialized nation by the year 2020, the
government has been playing an increasingly important role in the country’s industrialization
drive. The Malaysian Government has launched various grand-scale development projects.
Among them are, setting up the national car maker Proton (Perusahaan Otomobil Nasional),
establishing the Multimedia Super Corridor (MSC), building the North-South highway, the
Kuala Lumpur International Airport (KLIA), and the new administrative capital Putrajaya.
The planning, implementation and overseeing of these government initiated development
projects could have resulted in setting up new government agencies and offices, and the
expansion of bureaucracy. In other words, Malaysia may be one of the developing countries
where government expenditure grows as the country becomes wealthier.
Generally, in the early stages of the industrialization process, the governments may attempt
to diversify their activities by substituting private sector activities and services. As a result, the
size of government expenditure tends to expand. This inevitable increase of government
expenditure during the industrialization process is known as ‘Wagner’s Law’. According to
Adolph Wagner, the public sector expenditure grows as a country develops economically and
its national income increases.
* Lecturer, School of Business and Economics, Universiti Malaysia Sabah, Sabah, Malaysia.
E-mail: fumitakamy@gmail.com
Wagner
© 2008 ’sThe
L aw
IcfaiinUniversity
Malaysia: A New
Press. Empirical
All Rights Evidence
Reserved. 33
Wagner’s Law is a much discussed topic in the field of applied economics. However,
the majority of previous research studies have tested the existence of Wagner’s Law in the
context of developed countries. There is still a lack of systematic research involving the
developing countries, such as Malaysia, Thailand and Indonesia. Empirical testing of Wagner’s
Law in the former could offer interesting insights, as economic reality in developing countries
nowadays closely resemble the situation in European countries and Germany, when those
nations were rapidly industrializing their economies at the end of the 19 th century. At that
time, the German Government had also implemented various development projects in order
to catch up with its more economically advanced European neighbors and to promote
industrialization.
This paper chooses Malaysia as a case study to test Wagner’s Law. To analyze the hypothesis,
it employs a three-stage procedure proposed by Oxley (1994). Firstly, unit root test is used to
examine the stationarity of the datasets. Secondly, cointegration test is done to examine the
long-run movement of variables. Finally, the paper runs the Granger causality test based on
the Vector Error Correction Model (VECM).
Next, the paper briefly reviews important literature on Wagner hypothesis. Then, it discusses
the research methodology employed in this study, and reports the results of the empirical
tests. Finally, it offers concluding remarks.
Literature Review
‘Wagner’s Law’ has been attracting the attention of researchers, economists and applied
econometricians, since the end of the 19 th century. Wagner (1883) proposed the existence of
an inevitable increase in public expenditure as a country’s national income increases.
He predicted that as a country becomes wealthier, the size of the government and its expenditure
would tend to increase as well.
Wagner gave the following three reasons as to why public expenditure tends to expand
when a country undergoes the industrialization process (Bird, 1971). Firstly, the administrative
and protective functions of the state would substitute public for private activities. Secondly,
economic development would lead to an increase in ‘culture and welfare’ expenditure.
Finally, intervention from the government would be required to manage and finance natural
monopolies.
Research Methodology
In the current paper, several econometric analyses are done to examine the relationship between
public expenditure and economic growth in Malaysia. This study uses time series datasets for
the period 1970-2005. The main source of the data is Malaysia Economic Statistics – Time
Series published by the Department of Statistics, Malaysia (2006). Wagner’s hypothesis could
be estimated using the following equation (Thornton, 1999):
where ln is natural log, GDPt is real Gross Domestic Product (GDP) in year t , GE t, is real
government expenditure in year t , POPt is population size in year t, and μ is a random
disturbance term. Real government expenditure could be estimated using the GDP deflator
(Thornton, 1999). Support for Wagner’s Law would require that the elasticity of government
expenditure with respect to domestic product exceeds unity, i.e., β 0 (Oxley, 1994).
The presence of Wagner’s Law can be examined in three stages. In the first stage, unit root
test is used to examine the stationarity of datasets (Oxley, 1994). This paper uses the augmented
Dickey-Fuller (ADF) unit root test to investigate the stationarity (Dickey and Fuller, 1979 and
1981). The ADF test is based on the following regression,
n
y t μ βt 1t γ y
i 1
i t i εt ...(2)
RRS
AIC q T ln 2q
n q
where T is the sample size, RRS is the residual sum of squares, n is lag length, and q is the
total number of parameters estimated.
In the second stage, this study would employ the Ordinary Least Squares (OLS) regression
model, if the variables are integrated of order zero. On the other hand, if the variables are
integrated of order one, the Johansen cointegration test would be employed to check the
long-run movement of the variables (Johansen, 1988 and 1991). The Johansen cointegration
test is based on maximum likelihood estimation of the K-dimensional Vector Autoregressive
(VAR) model of order p,
Z t μ A 1 Z t 1 A 2 Z t 2 A K 1 Z t p 1 εt
Z t μ 1 Z t 1 2 Z t 2 k 1 Z t p 1 π Z t 1 εt
For example, if r =1, then the relationship between the variables could be written as:
ln GE t μ1 k 1
i ,11 i ,12 i ,13 ln GE t i
ln GDPt μ 2 i , 21 i , 22 i , 23
ln GDPt i
ln POPt μ 3 i 1
i , 33 ln POPt i
i , 31 i , 32
α1 ln GE t 1 ε1
α 2 β1 β 2 β 3
ln GDPt 1 ε 2
α 3 ln POPt 1 ε 3
The vector β represents r linear cointegrating relationship between the variables. This paper
uses the Trace (Tr ) eigenvalue statistics and Maximum (L-max) eigenvalue statistics (Johansen,
1988; and Johansen and Juselius, 1990). The likelihood ratio statistic for the trace test is:
where λˆr 1, , λˆp are the smallest eigenvalues of estimated p – r. The null hypothesis for the
trace eigenvalue test is that there are at most r cointegrating vectors. On the other hand, the
L-max could be calculated as:
L max T ln 1 ˆr 1
The null hypothesis for the maximum eigenvalue test is that r cointegrating vectors are
tested against the alternative hypothesis of r + 1 cointegrating vectors. If trace eigenvalue test
and maximum eigenvalue test yield different results, the results of maximum eigenvalue test
should be used, because the power of maximum eigenvalue test is considered greater than
the one of trace eigenvalue test (Johansen and Juselius, 1990).
The major problem of the Johansen cointegration test is that the test statistics are highly
sensitive to the choice of model specification and lag length. As shown in Table 1, five
different model specifications are used for the Johansen cointegration test.
The optimal model specification and lag length are determined by minimizing the AIC.
In the third stage, this paper runs the Granger causality test based on the following VECM:
n n n
ln GE t b 1
i 1
b 2i ln GDPt i
i 1
b 3i ln POPt i b
i 1
4i ln GE t i b 5 u t 1 εt
The current study is different from Sinha’s (1998) research, as it uses the Granger causality
test based on the VECM instead of the standard Granger causality test employed by Sinha.
There are two advantages of using this method rather than the standard Granger causality
test: 1) The F -test of the independent variables indicates short-run causal effect, and
2) Significant and negative error correction term indicates long-run causal effects.
Empirical Results
In the first stage, the ADF unit root test was done to examine the stationarity of the variables.
The results from the ADF unit root test are shown in Table 2. Despite minor differences in the
findings as reported in the table, the obtained results indicate that the three variables—lnGDP,
lnPOP and lnGE—are integrated of order one, I(1).
Secondly, the AIC was used again to determine the most appropriate model specification
for the Johansen cointegration test. As Table 4 shows, the best model specification is Model
2 and the number of cointegrating equation is one.
Number of
5% Critical 1% Critical
Eigenvalue Trace Statistic C oi nt eg r at in g
Val u e Val u e
Equations
Number of
M a xi m u m 5% Critical 1% Critical
Eigenvalue C oi nt eg r at in g
St ati st ic Val u e Val u e
Equations
The findings indicate that there exists long-run relationship between the three variables
(i.e., lnGDP , lnPOP and lnGE ), which means that these variables are cointegrated. In other
words, although the variables are not stationary at levels, in the long run, they move closely
with each other. Long-run cointegration when the variables are normalized by cointegrating
coefficients could be expressed as:
Finally, the Granger causality method based on the VECM was employed to examine the
long-run and short-run casual relationships between the three variables.
Firstly, the AIC was used to determine the optimal length for the causality test. As Table 7
shows, the optimal lag length for causality test is six which minimizes the AIC.
Next, results of F-test and t-tests are reported in Table 8. The findings show that the error
correction term is statistically significant, but positive. This means that there is no long-run Granger
Var i a bl e F - st at i st ic s p-value
lnGDP 3.845 0.029
Coefficient t - s ta t i st i c
u t –1 0.133 4.854**
N o t e : To test for causality when variables are cointegrated, the following Granger causality test
based on the VECM could be used:
n n n
ln GE t b1
i 1
b 2i ln GDPt i
i 1
b 3i ln POPt i b
i 1
4i ln GE t i b 5 u t 1 εt
Short-run causality: The joint significance of the coefficients is determined by the F-test;
Long-run causality: The level of significance for error correction term is determined by the
t-statistics; The results are based on a VECM with six lags; and ** indicates significance
at 1% level.
In a nutshell, empirical findings of the present study imply that there is a long-run
relationship—but no causality—between Malaysia’s GDP and public expenditure. These findings
partially confirm the results of the previous research on Wagner’s Law in Malaysia, done by
Sinha (1998). On the other hand, in the present inquiry, short-run causality has been detected
between Malaysia’s GDP and the public expenditure, which contradicts Sinha’s findings. This
difference in results could stem from the differences in methodology adopted by the two
studies. While Sinha employed the standard Granger causality test, the present study used the
Granger causality test based on the VECM.
To conclude, empirical findings of the present study indicate that there is long-run
relationship and short-run causality between Malaysia’s GDP and public expenditure. These
findings confirm that Wagner’s hypothesis may be valid in the context of an Asian developing
country, such as Malaysia. In other words, the current study provides an additional empirical
evidence to support the existence of Wagner’s Law.
There have transpired some contradictions in the course of the current research, such as
the absence of long-run causal relationship between national income and government
expenditure despite the existence of cointegrating relationship between the two. To address
this issue, future research studies may choose to employ different model specifications
which would incorporate variables such as per capita GDP or government expenditure per
person.
References
1. Afxentiou P C and Serletis A (1991), “An Empirical Re-examination of Wagner’s Law”,
Economics Letters, Vol. 26, pp. 169-173.
3. Akaike H (1974), “A New Look at the Statistical Model Identification”, IEEE Transaction on
Automatic Control, Vol. 19, pp. 716-723.
5. Bird R M (1971), “Wagner’s Law of Expanding State Activities”, Public Finance, Vol. 26,
pp. 1-26.
6. Chang T, Liu W and Caudill S (2004), “A Re-examination of Wagner’s Law for Ten Countries
Based on Cointegration and Error-Correction Modelling Techniques”, Applied Financial
Economics, Vol. 14, pp. 577-589.
9. Dickey D and Fuller W (1979), “Distribution of the Estimators for Autoregressive Time
Series with a Unit Root”, Journal of the American Statistical Association, Vol. 74, No. 366,
pp. 427-431.
10. Dickey D and Fuller W (1981), “Likelihood Ratio Statistics for Autoregressive Time Series
with a Unit Root”, Econometrica, Econometric Society, Vol. 49, No. 4, pp. 1057-1072.
11. Engle R and Granger C (1987), “Co-integration and Error Correction: Representation,
Estimation and Testing”, Econometrica, Vol. 55, No. 2, pp. 251-276.
12. Hondroyiannis G and Papapetrou E (1995), “An Examination of Wagner’s Law for Greece:
A Cointegration Analysis”, Public Finance, Vol. 50, pp. 67-79.
14. Johansen S (1991), “Estimation and Hypothesis Testing of Cointegrated Vectors in Gaussian
VAR Models”, Econometrica, Vol. 59, No. 6, pp. 1551-1580.
15. Johansen S and Juselius K (1990), “Maximum Likelihood Estimation and Inference on
Cointegration – With Application to the Demand for Money”, Oxford Bulletin of Economics
and Statistics, Vol. 52, pp. 169-210.
16. Kolluri B R, Panil M J and Wahab M S (2000), “Government Expenditure and Economic
Growth: Evidence from G7 Countries”, Applied Economics, Vol. 32, pp. 1059-1068.
17. Krzyzaniak M (1974), “The Case of Turkey: Government Expenditures, the Revenue Constraint
and Wagner’s Law”, Growth and Change, Vol. 5, No. 2, pp. 13-19.
18. Lin C A (1995), “More Evidence for Wagner’s L aw ”, Public Finance , Vol. 50,
pp. 262-277.
19. Nagarajan P and Spears A (1990), “An Econometric Test of Wagner’s Law for Mexico”,
Public Finance, Vol. 45, pp. 165-168.
20. Oxley L (1994), “Co-integration, Causality and Wagner’s Law: A Test for Britain
1870-1913”, Scottish Journal of Political Economy, Vol. 41, No. 3, pp. 286-298.
21. Ram R (1986), “Causality Between Income and Government Expenditure: A Broad
International Perspective”, Public Finance, Vol. 41, pp. 392-413.
22. Ram R (1987), “Wagner’s Hypothesis in Time-Series and Cross-Section Perspective: Evidence
from Real Data for 115 Countries”, Review of Economics and Statistics , Vol. 62,
pp. 194-204.
24. Thornton J (1999), “Cointegration, Causality and Wagner’s Law in 19th Century Europe”,
Applied Economics Letters, Vol. 6, pp. 413-416.
25. Vatter H G and Walker J F (1986), “Real Public Sector Employment Growth, Wagner’s Law,
and Economic Growth in the United States”, Public Finance, Vol. 41, pp. 117-137.
27. Wagner R E and Weber W E (1977), “Wagner’s Law: Fiscal Institution and the Growth of
Government”, National Tax Journal, Vol. 30, pp. 59-68.
Reference # 05J-2008-07-03-01