Escolar Documentos
Profissional Documentos
Cultura Documentos
a r t i c l e
i n f o
Article history:
Received 26 January 2014
Accepted 11 April 2014
Available online 23 April 2014
Keywords:
ICT
Energy demand
GDP
ARDL
Cointegration
a b s t r a c t
A strongly held belief in Japan is that information and communication technologies (ICT)
contribute to both a reduction in energy use and an increase in economic growth. As this
assertion is presently unproven, the purpose of this analysis is to estimate the long-run
relationship between ICT, energy consumption, and economic growth in Japan. Using an
autoregressive distributed lag (ARDL) bounds testing approach, we estimate two different
multivariate models corresponding to the production function and the energy demand
function, both including ICT investment as an explanatory variable, over the period
19802010. The results reveal the presence of a long-run stable relationship, not only
for the production function, but also for the energy demand function. However, in the
production function, the long-run coefcient estimate for ICT investment is statistically
insignicant, unlike the coefcients for labor, stock, and energy. In the case of the
energy demand function, the coefcients for GDP, energy price, and ICT investment are
statistically signicant. The results also indicate that the long-run ICT investment elasticity
of energy consumption is 0.155. On this basis, we conclude that while ICT investment
could ceteris paribus contribute to a moderate reduction in energy consumption, but not
to an increase in GDP.
2014 Elsevier Ltd. All rights reserved.
1. Introduction
From the end of World War II until the early 1970s, the average annual growth rate of energy consumption in Japan was
higher than the nations remarkable economic growth rate. However, following the 1973 oil crisis, the Japanese economy
was obliged to make some efforts to decrease its energy intensity (the ratio of energy consumption to GDP), and largely succeeded in slowing the growth rate of energy consumption without an accompanying decrease in GDP. Nonetheless, the fact
remains that energy consumption has continued to grow in the long term, with Japanese energy use rising by nearly 50%
from 1973 to 2010. In other words, the decrease in energy intensity has not revealed the concomitance of a reduction in
energy use and economic growth.
However, there is some hope that information and communication technologies (ICT) have the potential to solve the
dilemma of this winwin situation, that is, producing more output from less energy. For example, the Global Information
Infrastructure Commission (GIIC, 2008) emphasizes the role of ICT in protecting the environment without any sacrice in
economic output. Similarly, the Japanese government also relies strongly on ICT, not only to provide the economy with
80
market competitiveness through technical innovation and global entrepreneurship, but also to enable transport, buildings,
and industry to reduce their energy consumption.
Currently, ICT investment represents about 25% of total investment in Japan, with software accounting for about 50% of
total ICT investment and equipment the remainder. ICT has also been one of the fastest growing components of total investment, having almost doubled over the last 15 years. Nevertheless, the average annual growth rate of ICT investment in Japan
is relatively low when compared with the US or many European Union countries. For this reason, the Japanese government
has emphasized the importance of introducing policies and institutions to stimulate additional investment in ICT. However,
on this point, there is insufcient evidence showing whether ICT investment can actually lead to a winwin lower energy
use-economic growth situation, and so it is inappropriate to suggest that ICT investment is an absolute good for society.
In the worst-case scenario, there is even the possibility of a deadweight loss stemming from excessively high ICT investment.
The aim of this paper is to explore the nexuses between ICTGDP and ICTenergy using an autoregressive distributed lag
(ARDL) bounds test approach. Considering the problems associated with the use of a bivariate model pointed out by Stern
(1993), Stern (2000), we employ a multivariate approach in our analysis. For the ICT-growth nexus, we specify a neoclassical
aggregate production function including energy consumption and ICT investment as explanatory variables. For the ICT
energy nexus, we develop a conventional energy demand function, also including ICT investment as an explanatory variable.
The remainder of the paper is organized as follows. Section 2 reviews the literature on related studies. Section 3 describes
the methods and data used in this analysis. Section 4 reports the results of the respective econometric analyses. In Section 5,
we discuss the empirical results and Section 6 concludes.
81
Y f L; K; E; Z
where Y is real GDP, K is the capital stock, L is the labor input, E is the energy input, and Z is ICT investment. Taking the loglinear form of Eq. (1), we obtain the following equation:
ln Y a0 a1 ln L a2 ln K a3 ln E a4 ln Z
where the coefcients ai ; i 1; 2; 3; 4 correspond to the elasticity of GDP with respect to labor, the capital stock, energy
consumption, and ICT investment, respectively.
82
E gP; Y; Z
where P is the energy price. Taking the log-linear form of Eq. (3), we obtain:
ln E b0 b1 ln P b2 ln Y b3 ln Z
where the coefcients bi ; i 1; 2; 3 refer to the elasticity of energy consumption with respect to the energy price, GDP and
ICT investment.
3.3. Stationary test
Given that the conventional augmented DickeyFuller (ADF) test has low power with a short span of data, we apply the
DickeyFuller generalized least squares (DF-GLS) unit root test (Elliott et al., 1996) and the KPSS (Kwiatowski et al., 1992)
test to conrm the stationarity of the variables. As the ARDL bounds test for cointegration is applicable regardless of whether
the variables are I0 or I1, the stationary tests are only to ensure that there are no I2 variables.
3.4. ARDL bounds cointegration test
We employ the ARDL bounds testing approach (Pesaran and Shin, 1999; Pesaran et al., 2001) to examine the long-run
equilibrium relationships between the variables. We estimate the models developed above using the unrestricted error correction model (UECM) approach, where we indirectly estimate the long-run relationship and directly estimate the short-run
equilibrium. The UECM representation corresponding to Eq. (1) is:
D ln Y t k0
q1
X
k1i D ln Y ti
i1
q5
X
q2
q3
q4
X
X
X
k2i D ln Lti
k3i D ln K ti
k4i D ln Eti
i0
i0
i0
i0
where D is the rst-difference operator, the subscript t denotes the time period, and et is a disturbance term assumed normally distributed white noise. The UECM representation corresponding to Eq. (3) is:
D ln Et c0
q1
X
c1i D ln Eti
i1
h4 ln Z t1 et :
q2
X
q3
X
q4
X
i0
i0
i0
c2i D ln Y ti
c3i D ln P ti
The null hypothesis of no cointegration between the variables in Eq. (5) is H0 : l1 l2 l3 l4 l5 0, against
H1 : l1 l2 l3 l4 l5 0, which is denoted as FYjL; K; E; Z. Similarly for Eq. (6), H0 : h1 h2 h3 h4 0, against
H1 : h1 h2 h3 h4 0, which is denoted as FEjY; P; Z.
An F-test with a nonstandard asymptotic distribution is used to test the null hypothesis. To conduct the F-test, we use the
two sets of critical F-values provided by Narayan (2005) as our sample size is only 30. The rst set of critical values refers to
the lower-bound critical values, and the second set of critical values comprises the upper-bound critical values. If the calculated F-statistic exceeds the upper-bound critical value, then the null hypothesis is rejected, inferring that a cointegrating
relationship exists between the variables. If the calculated F-statistic is lower than the lower-bound critical value, then
the null hypothesis cannot be rejected, meaning there is no evidence of cointegration between the variables. If the calculated
F-statistic falls within the critical band between the upper and lower bound, then the result is inconclusive. In such an inconclusive case, applying ECM version of ARDL model will be a useful way of establishing cointegration (Kremers et al., 1992;
Bahmani-Oskooee and Nasir, 2004).
3.5. Data
We employ annual data for Japan over the period 1980 to 2010. The data for real GDP data and the capital stock are from
the Cabinet Ofce, Government of Japan. The data for total labor hours, calculated by multiplying the number of workers and
working time per capita, are from the Ministry of Internal Affairs and Communications and the Ministry of Health, Labor and
Welfare, respectively. The data for total domestic primary energy consumption in joules are from the Agency for Natural
Resources and Energy. Considering that fossil fuels (oil, coal and natural gas) are major sources of the energy supply in Japan,
we specify the price of fossil fuels as the energy price, being a quantity-weighted average of the imported prices (given Japan
imports almost all its fossil-fuel needs) of oil, coal and natural gas. We obtain the data for the prices of imported fossil fuels
83
DF-GLS
ln E
ln Y
ln P
ln L
ln K
ln Z
Dln E
Dln Y
Dln P
Dln L
Dln K
Dln Z
KPSS
Constant, no trend
Constant, no trend
1.620(1)
1.434(1)
1.166(0)
1.881(2)
1.967(1)
1.402(1)
3.697(0)
3.883(0)
3.579(3)
5.173(1)
3.271(3)
3.899(0)
0.868(1)
0.360(1)
1.211(0)
1.331(2)
1.271(1)
0.462(1)
3.376(0)
2.806(0)
4.565(0)
4.457(1)
2.164(0)
2.730(0)
0.183
0.184
0.176
0.168
0.160
0.179
0.122
0.072
0.126
0.127
0.088
0.089
0.612
0.657
0.455
0.187
0.427
0.654
0.318
0.390
0.392
0.286
0.352
0.445
Table 2
Bounds F-test for cointegration for the production model.
Model
F
F
F
F
F
(YL, K, E,
(LY, K, E,
(KY, L, E,
(EY, L, K,
(ZY, L, K,
Z)
Z)
Z)
Z)
E)
F-statistics
Cointegration
4.547
4.119
5.958
8.217
3.783
Inconclusive
No
Yes
Yes
No
from the Institute of Energy Economics, Japan, and calculate the real price using the GDP deator. The data for ICT investment in real terms are from the Ministry of Internal Affairs and Communications. 1
4. Empirical results
Table 1 provides the results of the DF-GLS and KPSS tests. The null hypothesis of the DF-GLS test states that the series in
question has a unit root, whereas the null hypothesis of the KPSS test is that the variable is stationary. As shown, the DF-GLS
test results indicate that all variables are nonstationary in levels, but stationary in rst-differences. The results for the KPSS
test show that all of the variables are nonstationary in levels except for labor (with constant and no trend) and the capital
stock (with constant and no trend). However, with the rst-differences, the KPSS test reveals that all of the variables are stationary. In particular, the results of the two tests conrm that all variables in question are either I0 or I1, namely, not I2.
Table 2 details the results of the ARDL bounds test for the ve-variable model (corresponding to the production model).
The results in Table 2 indicate that the calculated F-value for the equation with real GDP as the dependent variable falls
between the lower and upper bounds of critical values at the 5% level of signicance. Since the result of the cointegration
test is inconclusive, following Kremers et al., 1992, we check the coefcient of the ECM term to determine the long-run relationship among the variables (see Table 4).
Additionally, the calculated F-values for the remaining two equations specifying the capital stock and energy consumption as dependent variables are also higher than the upper-bound critical value at the 5% signicance level. These results
indicate cointegration among the variables. In contrast, the results show that there is no cointegration when labor and
ICT investment serve as dependent variables.
Table 3 provides the results of the ARDL bounds test for the four-variable model (corresponding to the energy demand
model). The results conrm that cointegration is only present when energy consumption is the dependent variable, signifying a long-run energy demand function. Meanwhile, the bounds test indicates that there is no cointegration when real GDP,
the energy price, and ICT investment are the dependent variables.
Next, we proceed to estimate the long- and short-run coefcients of both the production function and the energy demand
function in the long run based on the associated ARDL model and ECM framework. We estimate the ARDL model using the
1
If we consider the consistency of the variables used in the neoclassical production model, it might be more appropriate to use the stock of ICT rather than
ICT investment. However, we use ICT investment for the following reasons. First, the stationarity tests indicate that ICT stock is I2. Second, existing studies
(especially concerning Japan) display a strong tendency to use ICT investment in analyzing the relationship between ICT development and economic growth
and/or energy consumption. The use of ICT investment thus enables us to better compare our results with those of other researchers.
84
(EY,
(YE,
(PY,
(ZY,
P,
P,
E,
E,
Z)
Z)
Z)
P)
F-statistics
Cointegration
8.671
1.716
1.870
3.500
Yes
No
No
No
Table 4
Estimated long-run and short-run coefcients of the production function based on ARDL (1,0,0,1,0) model.
Variable
Coefcient
Standard error
T-statistics
Dependent variable: ln Y
Long-run results
ln L
ln K
ln E
ln Z
1.349
0.138
0.345
0.013
0.434
0.067
0.096
0.043
3.112[0.005]
2.043[0.053]
3.592[0.002]
0.294[0.771]
Short-run results
Dln L
Dln K
Dln E
Dln Z
ECT (1)
0.597
0.061
0.309
0.006
0.442
0.135
0.034
0.089
0.018
0.096
4.415[0.000]
1.781[0.088]
3.453[0.002]
0.301[0.766]
4.558[0.000]
R-bar-square
F-statistic
DW
RSS
Serial correlation (v-square)
RESET (v-square)
Normality (v-square)
Heteroscedasticity (v-square)
0.998
1751.2[0.000]
2.097
0.002
0.081[0.776]
1.717[0.190]
0.449[0.799]
1.419[0.233]
The optimal order of lags in the model are selected based on AIC.
Bracket represents probability values.
Table 5
Estimated long-run and short-run coefcients of the energy demand function based on ARDL (1,1,1,0) model.
Variable
Coefcient
Standard error
T-statistics
Dependent variable: ln E
Long-run results
ln Y
ln P
ln Z
1.075
0.098
0.155
0.336
0.029
0.075
3.204[0.004]
3.287[0.004]
2.061[0.052]
Short-run results
Dln Y
Dln P
Dln Z
ECT (1)
0.845
0.006
0.057
0.369
0.214
0.016
0.027
0.106
3.956[0.001]
0.369[0.715]
2.096[0.047]
3.489[0.002]
R-bar-square
F-statistic
DW
RSS
Serial correlation (v-square)
RESET (v-square)
Normality (v-square)
Heteroscedasticity (v-square)
0.989
371.7[0.000]
2.196
0.004
2.244[0.134]
2.540[0.111]
1.706[0.426]
2.862[0.091]
The optimal order of lags in the model are selected based on AIC.
Bracket represents probability values.
85
Akaike information criterion (AIC) to select the optimum lag order. Table 4 details the results for the production model. As
shown, the long-run elasticities of labor and energy consumption on real GDP are positive and statistically signicant at the
1% level. The long-run elasticity of capital stock on real GDP is also positive and statistically signicant at nearly the 5% level.
However, the long-run elasticity of ICT investment on real GDP is statistically insignicant, even at the 10% level. The shortrun elasticity of ICT investment on real GDP as estimated by the error correction representation of the ARDL model is also
statistically insignicant at the 10% level. Together, these results indicate that ICT investment does not have a positive effect
on economic growth in either the long or short run. As expected, the lagged error correction term in the production model is
negative and statistically signicant, supporting cointegration among the variables (Bahmani-Oskooee and Nasir, 2004). The
estimated coefcient of 0:442 suggests that the convergence to equilibrium following a shock to real GDP takes a little over
two years.
Table 5 provides the results for the long- and short-run estimation of the energy demand model. In line with theory, the
long-run income elasticity of energy demand is positive, and the long-run price elasticity of energy demand is negative. Both
are signicant at the 1% level. More importantly, the long-run impact of ICT investment on energy demand is negative and
almost statistically signicant (p-value = 0.052). The estimated value of the coefcient is 0:155, indicating that a 1%
increase in ICT investment results in a 0:155% decrease in energy demand. Moreover, the short-run elasticity of ICT investment on energy demand is also statistically signicant at the 5% level. Lastly, the lagged error correction term in the energy
demand model is negative and statistically signicant at the 1% level. The coefcient of 0:369 suggests that the convergence to equilibrium following a shock in energy demand takes place within three years.
86
Fig. 2. Plot of CUSUM and CUSUMSQ for the energy demand function.
As shown in Table 4 and Table 5, both of the estimated ARDL models (the production model and the energy demand
model) pass the tests for serial correlation, functional form misspecication, nonnormality, and heteroscedasticity. In addition, we test for the stability of the coefcients in the estimated models by using the cumulative sum (CUSUM) and the
cumulative sum of squares (CUSUMSQ) stability tests. Fig. 1 and Fig. 2 depict the plots for the stability tests for the production model and the energy demand model, respectively. As shown, the estimated parameters for both models appear stable
throughout our chosen sample period.
5. Discussion
One of the focuses of this paper is the impact of ICT development on energy demand in Japan. The empirical results conrm that aggregate energy demand in the long run is determined not only by the energy price and real GDP, but also by ICT
investment. The negative elasticity of ICT development in the energy demand function implies that a ceteris paribus increase
in ICT development will lower energy intensity (as measured by the ratio of energy consumption to GDP).
Needless to say, energy intensity can be lowered through either the same output with less energy or more output with
the same energy. Another of our focuses relates to the latter. Although we nd the impact of ICT investment on real GDP is
not statistically signicant, we note that our ndings do not preclude the possibility of more output with the same energy.
Strictly speaking, the results of our empirical analysis based on the production model indicate that a ceteris paribus increase
in ICT development will not affect economic growth. In other words, more output with the same energy can be achieved if
the input of labor and/or capital stock increase, even without a corresponding increase in energy consumption. These results
87
imply that the role of ICT development on economic growth should not be overestimated. Thus, if we hope to achieve economic growth without an increase in energy consumption, an increase in labor and/or capital stock must accompany developments in ICT.
This role for ICT may then be analogous to, for example, the improvement of fuel efciency in vehicles used for taxi services. Suppose that the fuel efciency of a taxicab doubles. The owner (driver) of the cab could then produce the same
amount of service with half the amount of fuel, or twice as much service (that is, mileage) with the same input of fuel, as
before. However, it is clear that a doubling of the working hours for the driver must accompany the latter. This suggests that
a technology used to improve energy productivity will not necessarily facilitate growth in labor productivity.
Again, our ndings imply that energy could substitute for labor and/or capital (at least to some extent), but not for ICT
investment. Considering the capital stock originates from past output, the results of our analysis imply that long-term economic growth would not be achieved without a continuous increase in either labor or energy consumption. If we avoid the
increase in energy consumption, the pursuit of economic growth would necessitate (more or less) an increase in the labor
force. Although this could contribute to a decrease in the rate of unemployment, it could also lead to an increase in working
hours per capita. Given that the average Japanese worker works longer than average among the OECD countries, the Japanese
government should take deliberate steps to implement a growth-oriented policy.
Conversely, our result that ICT development does not directly affect economic growth also suggests that the rebound
effect, dened as the percentage change in real GDP for a given percentage change in energy intensity, for ICT investment
may not be too serious. If the magnitude of the rebound effect is small, the direct impact and the induction effect may
not be sufciently serious to offset the anticipated energy savings from the decrease in energy intensity associated with
stimulating ICT development. This result is consistent with the ndings in Takase and Murota (2004) concerning the relationship between energy consumption and ICT investment in Japan.
6. Conclusions
This study estimates the impact of ICT investment on economic growth and energy consumption in Japan using annual
time-series data over the period 1980 to 2010. To do this, we develop two different multivariate models, comprising a vevariable model corresponding to the production model, and a four-variable model corresponding to the energy demand
model, both of which incorporate ICT investment as one of their explanatory variables. The results of the ARDL bounds testing approach to cointegration reveal stable long-run production and energy demand functions.
In the energy demand function, we nd the long-run coefcient for ICT investment to be almost statistically signicant at
the 5% level, thereby indicating that ICT investment growth is negatively related to energy consumption, though the impact
is small. The price and income elasticity of energy demand are statistically signicant with the correct theoretically expected
signs. The short-run impact of ICT investment on energy demand is also statistically signicant. In the production function,
however, both the long- and short-run impact of ICT investment on real GDP is not statistically signicant, with the production function explained by labor, the capital stock, and energy consumption, but not ICT investment.
Overall, we conclude that ICT investment directly contributes to a moderate reduction in energy consumption, but not to
an increase in GDP. The Japanese government should then not overestimate the potential of ICT development to realize sustainable growth with lower energy consumption. Adopting the policy instruments to promote heavily ICT development
might create a deadweight loss onto the Japanese economy.
References
Bahmani-Oskooee, M., Nasir, A.B.M., 2004. ARDL approach to test the productivity bias hypothesis. Rev. Dev. Econ. 8, 483488.
Barratt, R.S., 2006. Meeting lifelong learning needs by distance teaching clean technology. J. Cleaner Product. 14, 906915.
Bomhof, R., van Hoorik, P., Donkers, M., 2009. Systematic analysis of rebound effects for greening by ICT initiatives. Commun. Strategies 76, 7796.
Chakraborty, C., Nandi, B., 2003. Privatization, telecommunications and growth in selected Asian countries: an economic analysis. Commun. Strategies 52,
3147.
Chiabai, A., Rubbelke, D., Maurer, L., 2010. ICT applications in the research for environmental sustainability. Basque Centre for Climate Change, Working
Paper Sries 18.
Cho, Y., Lee, J., Kim, T., 2007. The impact of ICT investment and energy price on industrial electricity demand: dynamic growth model approach. Energy
Policy 35, 47304738.
Chu, N., Oxley, L., Carlaw, K., 2005. ICT and causality in the New Zealand economy, in: Proceedings of the 2005 International Conference on Simulation and
Modelling.
Cronin, F.J., Colleran, E.K., Herbert, P.L., Lewitzky, S., 1993. Telecommunications and growth: The contribution of telecommunications infrastructure
investment to aggregate and sectoral productivity. Telecommun. Policy 17, 677690.
Cronin, F.J., Parker, E.B., Colleran, E.K., Gold, M.A., 1991. Telecommunications infrastructure and economic growth: an analysis of causality. Telecommun.
Policy 15, 529535.
Dutta, A., 2001. Telecommunications and economic activity: An analysis of Granger causality. J. Manage. Inf. Syst. 17, 7195.
Elliott, G., Rothenberg, T.J., Stock, J.H., 1996. Efcient tests for an autoregressive unit root. Econometrica 64, 813836.
Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: representation, estimation and testing. Econometrica 55, 251276.
Forge, S., 2007. Powering down: Remedies fo unsustainable ICT. Foresight 9, 321.
Fuchs, C., 2008. The implications of new information and communication technologies for sustainability. Environ. Dev. Sustain. 10, 291309.
Ghali, K.H., El-Sakka, M.I.T., 2004. Energy use and output growth in Canada: a multivariate cointegration analysis. Energy Econ. 26, 225238.
GIIC, 2008. The GIIC Tokyo Declaration. http://www.giic.org
Gilling, E., 1975. Telecommunications and economic development: inter-country comparisons of the catalytic effect of telephone services on development.
Masters thesis. McGill University.
88
Gross, C., 2012. Explaining the (non-) causality between energy and economic growth in the U.S. a multivariate sectoral analysis. Energy Econ. 34, 489
499.
Hardy, A.P., 1980. The role of the telephone in economic development. Telecommun. Policy 4, 278286.
Hilty, L.M., 2008. Information Technology and Sustainability. Books on Demand.
Hilty, L.M., Ruddy, T.F., 2010. Sustainable development and ICT interpreted in a natural science context: the resulting research questions for the social
sciences. Inf. Commun. Soc. 13, 722.
Huber, P., Mills, M.P., 1999. Dig more coal the PCs are coming. Forbes 31, 7072.
Jipp, A., 1963. Wealth of nations and telephone density. Telecommun. J. 30, 199201.
Kremers, J.J.M., Ericsson, N.R., Dolado, J.J., 1992. The power of cointegration tests. Oxf. Bull. Econ. Statist. 54, 325348.
Kwiatowski, D., Phillips, P.C.B., Schmidt, P., Shin, Y., 1992. Testing the null hypothesis of stationarity against the alternative of a unit root: how sure are we
that economic time series have a unit root. J. Econometr. 54, 159178.
Laitner, J., 2002. Information technology and U.S. energy consumption: energy hog, productivity tool, or both? J. Ind. Ecol. 6, 1324.
Madden, G., Savage, S.J., 1998. CEE telecommunications investment and economic growth. Inf. Econ. Policy 10, 173195.
Narayan, P.K., 2005. The saving and investment nexus for China: evidence from cointegration tests. Appl. Eocon. 37, 19791990.
Oh, W., Lee, K., 2004. Causal relationship between energy consumption and GDP revisited: the case of Korea 19701999. Energy Econ. 26, 5159.
Pesaran, M.H., Shin, Y., 1999. An autoregressive distributed lag modeling approach to cointegration analysis. In: Strom, S. (Ed.), Econometrics and Economic
Theory in the 20th Century. Cambridge University Press, Cambridge.
Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approaches to the analysis of level relationships. J. Appl. Econometr. 16, 289326.
Rejeski, D., 2002. E-commerce, the internet, and the environment. J. Ind. Ecol. 6, 13.
Roth, K.W., Goldstein, F., Kleinman, J., 2002. Energy consumption by ofce and telecommunications equipment in commercial buildings, in: Energy
Consumption Baseline, vol. 1. Arthur D. Little.
Sadorsky, P., 2012. Information communication technology and electricity consumption in emerging economies. Energy Policy 48, 130136.
Saunders, R.J., Warford, J.J., Wellenius, B., 1983. Telecommunications and Economic Development. The Johns Hopkins University Press.
Shinjo, K., Zhang, X., 2004. ICT capital investment and productivity growth: Granger causality in Japanese and the USA industries. (mimeo).
Shiu, A., Lam, P., 2008. Causal relationship between telecommunications and economic growth in China and its regions. Regional Studies 42, 705718.
Soytas, U., Sari, R., 2003. Energy consumption and GDP: causality relationship in G-7 countries and emerging markets. Energy Econ. 25, 3337.
Stern, D.I., 1993. Energy and economic growth in the USA: a multivariate approach. Energy Econ. 15, 137150.
Stern, D.I., 2000. A multivariate cointegration analysis of the role of energy in the US macroeconomy. Energy Econ. 22, 267283.
Szklo, A., Schaeffer, R., 2007. Fuel specication, energy consumption and CO2 emission in oil reneries. Energy 32, 10751092.
Takase, K., Murota, Y., 2004. The impact of IT investment on energy: Japan and US comparison in 2010. Energy Policy 32, 12911301.
Toffel, M.W., Horvath, A., 2004. Environmental implications of wireless technologies: News delivery and business meetings. Environ. Sci. Technol. 38, 2961
2970.
Wang, Y., Wang, Y., Zhou, J., Zhu, X., Lu, G., 2011. Energy consumption and economic growth in China: a multivariate causality test. Energy Policy 39, 4399
4406.
Yi, L., Thomas, H.R., 2006. E-business and sustainable development. Int. J. Environ. Sustain. Dev. 5, 262274.
Yuan, J., Kang, J., Zhao, C., Hu, Z., 2008. Energy consumption and economic growth: evidence from China at both aggregated and disaggregated levels. Energy
Econ. 30, 30773094.
Zachariadis, T., 2007. Exploring the relationship between energy use and economic growth with bivariate models: new evidence from G-7 countries. Energy
Econ. 29, 12331253.