Energy Consumption and Economic Growth – The Case of Australia
Hong To a, *, Albert Wijeweera a, Michael B. Charles a
a
Business School, Southern Cross University,
Locked Mail Bag 4, Coolangatta, QLD, 4225, Australia
* Corresponding author. Tel.: +61 7 55893207;
E-mail addresses: [email protected] (H. To), [email protected] (A.
Wijeweera), [email protected] (M.B. Charles).
Abstract:
This study integrates neoclassical growth, endogenous growth, and ecological-economics
viewpoints to examine how energy consumption affects economic growth in Australia. It
utilizes four decades of data ranging from 1970 to 2011 and the bound testing cointegration
approach along with multivariate Granger causality test to examine probable statistical
relationships between the variables. Results based on bound test approach suggest that
energy consumption and Australian economic growth, despite being positively related, are
not statistically significant either in the short run or the long run. The weak relationship
between the two variables is further ascertained by the multivariate Granger causality test
results. These findings suggest, at least ostensibly, that much debated carbon pricing policies
may not necessarily have an adverse effect on Australia’s economic growth.
Key Words: Energy and Growth, Australia, Bound Testing Approach
JEL: O13, Q43, Q48
1
1.
Introduction
Over the years, much research has been carried out to determine the key factors
impacting on economic growth, with energy being a relatively new factor, and one not
included in traditional economic growth models (Stern, 2011; Pirlogea and Cicea, 2012). A
majority of studies have explained economic activity and growth in terms of a production
function. Neoclassical growth models usually regard capital, labour and land as the primary
factors of production, while energy is regarded as an intermediate input eventually produced
by the primary factors of production. Furthermore, neoclassical economists often assume that
energy and capital are perfectly substitutable (Solow, 1974). A decline in energy use does
not, under conditions of economic efficiency, result in a reduction in economic growth. These
viewpoints have led to a focus in the mainstream growth theory on the primary inputs, and in
particular, capital and labour, more so given that land is usually subsumed as a subcategory
of capital. Energy is assumed to have a relatively minor role in economic production in the
mainstream theory of growth. This has been strongly criticised by proponents of ecological
economics, which is grounded in the biophysical theory of the role of energy. The law of
thermodynamics implies that a minimum quantity of energy is required to carry out the
transformation of matter. Since all production involves the transformation or movement of
matter in some way, energy is therefore necessary for economic production and, as a result,
economic growth. That said, there must be limits to the substitution of other factors of energy
production. Furthermore, econometric studies (e.g., Berndt and Wood, 1979; Apostolakis,
1990; Stern, 1993; Frondel and Schmidt, 2002) have employed various functional forms to
estimate elasticities of substitution between energy and capital. These studies have shown
that capital and energy are, at best weak, substitutes, and are quite possibly complements.
As discussed above, energy is an input in the production process, since it is used in other
economic activities. Many countries such as Japan lack energy resources and generally
2
depend on imports of crude oil, natural gas, and coal for their industrial and residential
energy needs, transportation, and electricity generation. In these cases, there is likely to be a
positive relationship between energy consumption and economic growth. Peak oil, energy
security and climate change have become key concerns in recent decades. Given the changes
in energy policies in response to these issues, the causal relationships between energy
consumption and economic growth has become a compelling area of investigation. From an
economic point of view, this relationship lies in two aspects: i) the growing dependence of
economic growth on energy, and ii) economic growth promoting energy technology advances
and large-scale development and utilization of energy. Various studies (e.g., Akarca and
Long, 1979; 1980; Glasure and Lee, 1998; Masih and Masih, 1996; 1997; 1998) have shown
i) that the relationship between energy consumption and economic growth varies depending
on the country, and ii) the relationship varies in the same country at different times.
The discrepancy in results results from a number of factors. These include: i) the
different structures and stages of economic development, ii) the use of different econometric
methods, iii) the varying time horizon of the analysis, and iv) the type and number of
variables employed (Yu and Choi, 1985; Ferguson et al., 2000; Toman and Jemelkova, 2003;
Karanfil, 2009; Payne, 2010). Earlier studies relied on the OLS model of log-linear to
estimate parameters and conduct statistical tests, all without taking into consideration the
special features of time series data. These traditional estimation methods are often associated
with several empirical problems, such as the possible endogeneity of regressors and the nonstationarity of the variables. All of these lead to spurious regressions with misleading
statistical results (Granger and Newbold, 1974). There have been important advances in the
past decade, with new time series econometric techniques such as cointegration, error
correction and vector autoregressive (VAR) methods being developed. As a result, it is
3
necessary to revisit and statistically re-examine the relationship between energy consumption
and economic growth using modern time series analysis.
Furthermore, the existing literature on the relationship between energy consumption and
economic growth has suffered from two major limitations: i) a lack of a synthesis of energybased and mainstream models as a result of different theoretically-based approaches on
economic growth (i.e., mainstream growth theory vs. ecological-economics viewpoints); and
ii) the possibility of omitted-variable biases, which arises when variables known to be
important are omitted from the models. In this study, we attempt to address these issues by
examining the relationship between energy consumption and economic growth in the case of
Australia, where relatively little research using a multivariate approach in this area has been
conducted. Earlier Australian studies are based primarily on bivariate models, which could
hamper an accurate analysis owing to the omitted variable biases (Shahiduzzaman and Alam,
2012). Most recent endogenous growth models hold that investment in human capital,
innovation, and knowledge are significant contributors to economic growth (Aghion and
Howitt, 1997). Furthermore, energy is necessary for economic production and economic
growth from ecological-economics viewpoints. This study adds to the literature by
augmenting the model specification with human capital and energy variables, together with
the classical determinants of growth, i.e., labour force and capital stock. In other words, the
model used herein enables us to integrate neoclassical and endogenous growth with
ecological-economics viewpoints so as to study the relationship between energy consumption
and economic growth in Australia. Although there have been some attempts to integrate
neoclassical growth theory with the ecological-economics approach, such as that of Stern
(2011) and Ayres and Warr (2009), there has been, as yet, no synthesis of these approaches
and endogenous growth theory.
4
The remainder of this paper is organized as follows: the following section provides a
brief overview of the related literature. The third section discusses variables and data sources,
while the fourth outlines the methodology employed in this study. Results are presented in
the fifth section. Some concluding remarks and policy implications complete the article.
2.
Literature Review
There has been a growing literature on the causal relationship between energy
consumption and economic growth. These studies have employed a variety of time series
econometric techniques. This research interest on energy and growth stems from the earlier
oil crisis in the 1970s to the more recent concerns on energy prices, energy security and the
impact of environmental policy to conserve energy and reduce greenhouse gas emissions.
The empirical results on the energy consumption-growth nexus have yielded mixed and
inconsistent results in terms of their causal relationships. In this literature review,
international based studies are discussed first before moving to Australian studies.
2.1. International studies
The first relevant study on energy and growth dates back to the late 1970s. In their
pioneering work, Kraft and Kraft (1978) used annual U.S. data from 1947 to 1974 to study
the relationship between gross national product (GNP) and gross energy inputs. They
employed the Sims causality test procedure to infer the causal relationship, and discovered
that increased GNP leads to increased energy consumption. Using employment to substitute
for economic growth, Akarca and Long (1979) showed that increased energy consumption
leads to higher levels of employment. However, when using different methodology (i.e., Sims
causality test) and different data set (i.e., annual U.S. data from 1950 to 1970), Akarca and
Long (1980) found no causal relationship between energy consumption and GNP. As per
Akarca and Long (1979), Erol and Yu (1987a), together with Murray and Nan (1992), used
employment to substitute for economic growth. Erol and Yu (1987a) applied the Sims
5
causality technique to monthly U.S. data from 1973 to 1984 and found no causal relationship
between energy consumption and employment. Yet Murray and Nan (1992) used the Granger
causality procedure and monthly U.S. data from 1974 to 1988. They found that increased
employment results in increased energy consumption. In other research, Erol and Yu (1987b)
applied both the Sims and Granger causality procedures to examine the causal relationships
between energy consumption and real GNP for Japan, Germany, Italy, Canada, France and
the U.K. The results show that there is bidirectional causality between the two variables in
Japan. For the case of Germany and Italy, increased GNP leads to increased energy
consumption. Increased energy consumption leads to increased GNP in Canada, but there are
no causal relationships between the two in France and the U.K.
The feature of the model specification in the above studies is the reliance on bivariate
causality test of energy consumption and output or employment. However, a common
problem of a bivariate analysis is the possibility of omitted variables bias, which could result
in misleading statistical results (Stern, 2000; Payne, 2010). Recognizing the problem, Yu and
Hwang (1984), together with Stern (1993), incorporated additional variables in their analyses
for the case of the U.S. Yu and Hwang (1984) included employment when examining the
relationship between energy consumption and GNP. They employed both Sims and Granger
causality tests and found that increased employment leads to increased energy consumption,
while there is no causal relationship between energy consumption and GNP. Stern (1993)
incorporated employment and capital in the analysis and found that increased energy
consumption results in growth in real GDP.
In the previous studies, traditional OLS method was usually used to estimate parameters
and to conduct statistical tests. These traditional estimation methods do not take into
consideration the special features of time series data, such as the possible endogeneity of
regressors and the non-stationarity of the variables, both of which could result in spurious
6
regressions, together with misleading statistical results (Granger and Newbold, 1974). With
advances in time series econometrics in the past decade, new time series econometric
techniques such as Engle-Granger (1987) / Johansen-Juselius (1990) cointegration and errorcorrection models have been applied to re-investigate the relationship between energy
consumption and growth.
Results of the studies utilizing the Engle-Granger cointegration and error-correction
model follow. Glasure and Lee (1998) found that there is bidirectional causality between
energy consumption and real GDP in South Korea and Singapore. Francis et al. (2007) found
similar results for the case of Haiti, Jamaica and Trinidad and Tobago. Yet Cheng and Lai
(1997) demonstrated unidirectional relationship from energy consumption to employment
and from real GDP to energy consumption in Taiwan. Taking into account the possibility of
omitted-variable biases, Yu and Jin (1992), Cheng (1996), Paul and Bhattacharya (2004) and
Pirlogea and Cicea (2012) all incorporated measures of capital and/or labour in the context of
a production model framework. Glasure and Lee included wages and energy prices (1995)
and, later, wages and energy prices, real money supply and real government spending (1996)
into their models to examine the relationship between energy consumption and growth. Yu
and Jin (1992) and Cheng (1996) found no long-term cointegration relation and no causal
relationship between the two, while Glasure and Lee (1995, 1996) and Paul and Bhattacharya
(2004), by way of contrast, found bidirectional relationship between energy consumption and
growth. The majority of these studies have focused on the causal relationship between energy
consumption and economic growth using aggregate energy consumption data. Given that the
use of aggregate energy consumption could mask the differential impact associated with
various types of energy consumption, as well as by end use and sector, Yang (2000a, 2000b),
Yoo and Kim (2006), Jinke et al. (2008) and Pirlogea and Cicea (2012) attempted to examine
the impact of various disaggregated measures of energy consumption such as electricity, coal,
7
natural gas, oil and renewables, as well as by sector. Again, there is no consensus on the
causal relationship between the two factors within and across countries.
Johansen-Juselius cointegration and error-correction model has been more widely
employed. The majority of these studies are based on the bivariate model, which includes
only energy and output or employment, as per Masih and Masih (1996), Soytas and Sari
(2003), Yoo (2005, 2006a, 2006b, 2006c), Yoo and Jung (2005), Chen et al. (2007) and
Zachariadis (2007). Other studies included i) measures of capital and/or labour, as per Stern
(2000), Ghali and El-Sakka (2004), Oh and Lee (2004a, 2004b), Paul and Bhattacharya
(2004), Soytas and Sari (2006a, 2007), Yuan et al. (2008); or ii) consumer prices, as per
Masih and Masih (1997, 1998) and Asafu-Adjaye (2000). Glasure (2002), however,
incorporated various variables, including real government expenditure, real money supply,
real oil prices and dummy variable oil price shocks. While most of the studies have used
aggregate energy consumption data, Ghosh (2002), Hondroyiannis et al. (2002), Shiu and
Lam (2004), Yoo (2005, 2006a, 2006b, 2006c), Yoo and Jung (2005), Chen et al. (2007),
Soytas and Sari (2007), Zachariadis (2007) and Yuan et al. (2008) all employed various
disaggregated measures of energy consumption by source and by sector.
Inconsistent and contradictory results are still reported across studies. For example,
Masih and Masih (1996, 1997, 1998) found no causal relationship between energy
consumption and growth in Malaysia, Singapore and the Philippines, while there is a
bidirectional relationship between the two in Pakistan, South Korea and Taiwan. In addition,
they found that increased energy consumption causes growth in India, Thailand and Sri
Lanka, while economic growth leads to increased energy consumption in Indonesia. Stern
(2000) found that greater energy consumption results in growth in the United States, while
Soytas and Sari (2003) discovered i) no causal relationship in Canada, Indonesia, Poland, the
United Kingdom, and the United States; ii) bidirectional causality in Argentina and Turkey;
8
iii) unidirectional causality with greater energy consumption leading to increased GDP in
France, West Germany and Japan; and iv) causality with increased GDP leading to increased
energy consumption in Italy and South Korea. In contrast to Soytas and Sari’s result (2003),
Ghali and El-Sakka (2004) established bidirectional relationship between energy
consumption and growth in Canada. Finally, Oh and Lee (2004a, 2004b) found inconsistent
conclusions for the case of Korea when using different data sets and models.
While the Engle-Granger/Johansen-Juselius cointegration procedures and corresponding
error-correction models have been widely used to study a causal relationship between energy
consumption and economic growth, these methods have been criticized owing to the low
power and size properties of small samples associated with conventional unit root and
cointegration tests (Harris and Sollis, 2003). In response, more recent studies have employed
the autoregressive distributed lag (ARDL) model and bounds testing approach, together with
the Toda-Yamamoto (1995) and Dolado-Lütkepohl (1996) long-run causality tests, which can
be performed irrespective of whether the variables possess a unit root and whether
cointegration exists among the variables. Altinay and Karagol (2005) used the DoladoLütkepohl test of long-run causality between electricity consumption and real GDP for the
case of Turkey and found unidirectional causality, with increased electricity consumption
leading to higher GDP. Lee (2006) employed the Toda-Yamamoto causality test and found
no causal relationship between energy usage and real GDP per capita in Germany, Sweden
and the United Kingdom; bidirectional causality between the two in the United States;
increased energy consumption leading to increases in real GDP per capita in Belgium,
Canada and Switzerland; and increases in real GDP per capita leading to greater energy
consumption in France, Italy and Japan. Soytas and Sari (2006b) also used the TodaYamamoto causality test for their model including energy usage, real GDP, real gross fixed
capital formation and labour force variables to discover the causal relationship between
9
energy consumption and growth in China. Their results showed no causal relationship
between the two. Zachariadis (2007) employed different approaches, including ARDL
bounds test and the Toda-Yamamoto causality test, to study the causal relationship between
the disaggregated measures of energy consumption by sector and income/output measures in
Canada, France, Germany, Italy, Japan, the United Kingdom and the United States.
Inconsistent and conflicting results were found in the research when applying different
econometric methods. Bowden and Payne (2010) also studied the causal relationship between
the disaggregated measure of energy consumption by sector and real GDP in the United
States using the Toda-Yamamoto causality test. They incorporated real gross fixed capital
formation and employment variables in their analysis and found no causal relationship
between commercial/industrial renewable energy consumption and real GDP; bidirectional
causality between commercial/residential non-renewable energy consumption and real GDP;
and unidirectional causality, with residential renewable/industrial non-renewable energy
consumption leading to an increase in real GDP. Another U.S. study reported by Sari et al.
(2008) included the employment variable and employed the ARDL bounds test to investigate
the causal relationship between the disaggregated measures of energy consumption by
sources and industrial production. The results showed unidirectional causality, with increased
industrial production leading to greater energy consumption, except for the case of coal
consumption, which was found to lead growth.
Another approach that addresses the concerns of the low power and size properties of
small samples associated with conventional unit root and cointegration tests is the panel
cointegration tests. Panel unit root and cointegration tests provide additional power by
combining the cross-section and time series data allowing for the heterogeneity across
countries (Payne, 2010). Lee (2005), Chen et al. (2007), Mehrara (2007), Narayan and Smyth
(2007), Lee and Chang (2008) and Lee et al. (2008) employed this approach, while Huang et
10
al. (2008) and Sharma (2010) applied dynamic panel estimation to infer a causal relationship
between energy consumption and economic growth. Lee (2005) included real gross capital
formation in the analysis and found unidirectional causality, with increased energy
consumption leading to real GDP growth for the developing countries panel. Yet Chen et al.
(2007) discovered bidirectional causality between electricity consumption and real GDP for a
ten-country panel including China, Hong Kong, Indonesia, India, Korea, Malaysia, the
Philippines, Singapore, Taiwan, and Thailand. Mehrara (2007), however, found that real
GDP per capita growth led commercial energy usage per capita for the oil-exporting
countries panel. Narayan and Smyth (2007) included real gross fixed capital formation in the
estimation and found that energy consumption per capita causes real GDP growth per capita
for the G7 panel. For OECD countries, Lee et al. (2008) found bidirectional causality
between the two variables in question, while Lee and Chang (2008) incorporated both real
gross fixed capital formation and labor force and found unidirectional causality, with
increased energy consumption leading to real GDP growth for the Asian panel, APEC panel,
and the ASEAN panel. Huang et al. (2008) classified data into four income groups and
discovered i) no causal relationship between energy consumption and real GDP per capita for
the low-income panel; ii) economic growth leading energy consumption positively in the
middle-income group; and iii) economic growth leading energy consumption negatively for
the high-income panel. Mixed results on the impact of electricity and non-electricity
consumption on economic growth for a global panel as well as for four regional panels
(East/South Asian and the Pacific region, Europe and Central Asian region, Latin America
and Caribbean region, and Sub-Saharan, North Africa and Middle Eastern region) were also
found by Sharma (2010). The analysis is based on a model consisting of inflation, capital
stock, labour force, trade, and energy.
11
2.2. Australian studies
Earlier studies using Australian data to examine the causal relationship between energy
consumption and economic growth are based primarily on bivariate models. Using a bivariate
approach, Fatai et al. (2004) applied different time series econometric methods (TodaYamamoto causality, ARDL bounds test, and Johansen-Juselius procedure) to annual data
from 1960 to 1999, and concluded that real GDP growth leads to increased energy
consumption. The authors also studied the impacts of various disaggregated measures of
energy consumption by sources (i.e., coal, electricity, oil, natural gas consumption). Narayan
and Smyth (2005), by way of contrast, used a trivariate model (electricity consumption per
capita, real GDP per capita, and manufacturing employment index) and applied ARDL
bounds test to discover the causality relationship during 1966-1999. The results also showed
that there is unidirectional causality, with growth leading to increased electricity
consumption. Using a bivariate model and Johansen-Juselius test procedure, Chontanawat et
al. (2008) demonstrated causality from real per capita GDP to per capita energy consumption
for the period 1960-2000. These test results are in contrast to those of Narayan and Prasad
(2008), who found a long-run causality from electricity consumption to output in Australia
for the period 1960–2002 using a bootstrapped Granger causality test. To reduce potential
omitted-variable biases, Mahadevan and Asafu-Adjaye (2007) included the consumer price
index as a third variable in their study. They found evidence of cointegration and
bidirectional causality between per capita energy consumption and real per capita GDP for
the period 1971-2002. Shahiduzzaman and Alam (2012) incorporated capital and labour in
their study, in addition to energy consumption and real GDP, and used both JohansenJuselius and Toda-Yamamoto causality tests to determine a causal relationship for the years
1961-2009. They also found evidence of cointegration and bidirectional causality between
GDP and energy usage, consistent with the results of Mahadevan and Asafu-Adjaye (2007).
12
3.
Variable and Data Sources
Neoclassical growth models, such as Solow’s growth model (Solow, 1956), usually
consider capital and labour as the primary factors of production and, therefore, energy is
assumed to have a relatively minor role. Yet most ecological-economics viewpoints consider
only the role of energy and ignore the roles of other classical inputs such as capital and labour
(Stern, 2011). Endogenous growth models have emphasized the role of human capital in
economic growth (Galor and Weil, 2000; Lucas, 2002). To synthesize these approaches, we
use a production function approach, which enables to incorporate capital and labour inputs as
considered in neoclassical growth theory, energy as used in ecological economics models,
and capital input as discussed in endogenous growth models. The production function
approach provides a more comprehensive methodology that avoids the ad hoc selection of
additional variables (Stern, 1993; Stern, 2000; Shahiduzzaman and Alam, 2012).
Following the literature, we use gross domestic product (GDP), real values in $AUD, as
the dependent variable. As explained above, there are four explanatory variables: capital,
labor, energy consumption, and human capital. The capital input (K) in the model is
measured by gross capital formation (real values in $AUD), which consists of outlays on
additions to the fixed assets of the economy plus net changes in the level of inventories. The
labour factor (L) is measured by total labour force comprising people aged 15 and older who
supply labour for the production of goods and services. Energy input (E) refers to the use of
primary energy before transformation to other end-use fuels, which is equal to indigenous
production plus imports and stock changes, minus exports and fuels supplied to ships and
aircraft engaged in international transport. Energy consumption data are aggregated and
measured by kilotonnes of oil equivalent. Human capital refers to expertise or know-how
embodied in people through processes of education and training. The most commonly used
measure of human capital is the level of school attainment in a country. Here, human capital
13
(H) is measured by the total enrollment in tertiary education, regardless of age, which is
expressed as a percentage of the total population of the five-year age group following on
from secondary school leaving. We use annual time series data from 1970 to 2011 sourced
from the World Bank (2012) to estimate the model.
4.
Methodology
The following model is used to examine the relationship between energy consumption
and economic growth.
Log(Yt )    LogLt  LogKt  LogH t  LogEt   t
(1)
where, Yt is Australian gross domestic product in constant Australian dollars. Lt, Kt, Ht, and Et
refer to labour, capital, human capital, and energy consumption as explained in the data
section. In general, we expect that β, γ, λ and θ will all be positive because an increase in
factors of production should, under normal circumstances, lead to a higher output. The model
is in log-log form. Hence, coefficients can directly be interpreted as elasticities. For instance,
β measures the labour elasticity. In specific terms, β shows percentage change in real GDP in
response to a one per cent change in labour force. Other coefficients can also be interpreted in
a similar way. To illustrate, λ shows percentage change in real GDP in response to a one per
cent change in human capital. However, our focus would be on the direction and the
magnitude of θ or the energy elasticity.
One of the limitations of the model given in equation (1) is that it only provides
information on the long-run relationship between the factors of production and national
output in Australia. However, in this paper, we aim to analyse both the short-run and long-run
elasticities, and the energy input elasticities in particular. For that purpose, the study uses the
bound testing cointegration approach suggested by Pesaran et al. (2001). The bound testing
method utilizes the autoregressive distributed lag (ARDL) model, while the ARDL model
used is given in equation (2).
14
n1
n2
n3
k 0
k 0
 log( Yt )      k  log( Yt k )    k  log( Lt k )    k  log( K t k ) 
k 1
n4
n5
k 0
k 0
  k  log( H t k )   k  log( Et k )  LogLt  LogK t  LogH t  LogEt  u t
(2)
Compared to other known methods of cointegration such as Engle and Granger TwoStep approach (1987) and the system-based reduced rank approach of Johansen (1991), the
bound testing approach has several advantages. To illustrate, in other cointegration methods,
researchers are required to know unit root properties of each series before using them in the
estimation. As explained by Pesaran et al. (2001), both the Engle-Granger method and the
Johansen method are concentrated on variables integrated of order one. But in the bound
testing approach, the order of integration (order zero or order one) does not matter. Bound
testing method has a further advantage because it performs better in small samples (Narayan,
2005). More importantly, the ADRL method can be used to estimate both short- and long-run
estimates in one step. To test for cointegration, we should test the null hypothesis of all longrun coefficients being zero. Pesaran et al (2001) advise using a F-test, but with modified
critical values, depending on whether all variables are integrated or order one, or order zero.
5.
Results
Given that the original data sample contains only 42 observations and that the degrees of
freedom is further curtailed by the differencing, we have confined the model to the lag one
first differenced data and the long run relationships. The results are shown in Table 1 below.
Short-run elasticities are given by the estimated coefficients of DLL, DLK, DLH, and DLE,
while the long-run elasticities are given by LL, LK, LH, and LE. As the results show, the
short-run elasticity of the energy consumption has the expected positive sign, but is
statistically insignificant at conventional levels. As far as the other variables are concerned,
the estimate of the coefficient of labour variable is positive and significant at 1 per cent level
of significance. This suggests that there is statistical evidence to support the assertion that
15
labour exerts a positive impact on growth in the short run. It is impossible to comment on the
estimate of the human capital variable or capital variable because both are statistically
insignificant at conventional levels of significance. With respect to the long run coefficients,
all four factors of production have the expected a positive relationship with economic growth,
but only capital and human capital variables coefficients are statistically significant at 5 per
cent level of significance.
Table 1: Short-run and long-run elasticities using ARDL bound test
Dependent Variable: DLY
Sample (adjusted): 1971 2010
Included observations: 40 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C
–5.035383
2.524488
–1.994616
0.0552
DLL(1)
0.921455
0.272004
3.387653
0.0020
DLK(1)
–0.002522
0.039260
-0.064245
0.9492
DLH(1)
0.008440
0.029027
0.290774
0.7732
DLE(1)
0.007450
0.110182
0.067616
0.9465
LL
0.101931
0.135306
0.753335
0.4571
LK
0.081865
0.033083
2.474507
0.0192
LH
0.051259
0.017329
2.957945
0.0060
LE
0.118404
0.088220
1.342153
0.1896
TREND
–0.010115
0.003800
–2.661509
0.0124
R-squared
0.481812
Mean dependent var
0.031562
Adjusted R-squared
0.326356
S.D. dependent var
0.015282
S.E. of regression
0.012543
Akaike info criterion
–5.706989
Sum squared resid
0.004720
Schwarz criterion
–5.284769
Log likelihood
124.1398
Hannan-Quinn criter.
–5.554327
F-statistic
3.099341
Durbin-Watson stat
2.377331
Prob(F-statistic)
0.009417
Now we perform diagnostics tests to demonstrate that our findings are robust. There are
important steps to this diagnostics review. First, as Pesaran et al. (2001) showed, results
based on equation (2) are valid only if the level variables are in fact a part of the estimated
16
model. We perform F-test to test the null hypothesis that β=γ=λ=0 or no level relationship
between the variables under consideration. However, Pesaran et al. (2001) suggest a bound
testing approach for the F-test, which contains two critical values for two bands, upper bound
assuming I(1) variables and lower bound assuming I(0) variables. If the computed F-statistics
fall outside the critical values of these bounds, a conclusive decision can be made without
knowing the order of integration of the variables. However, if the calculated F-statistics falls
within the upper and lower bound, the knowledge of integration is necessary or the inferences
are inconclusive (Pesaran et al. 2001). The upper bound value for our specification is 3.25
and the F- test statistics give 3.51, which is outside this range. As a result, we can make
conclusive inferences from the results based on the ARDL modelling framework shown in
equation (2).
The second diagnostic test involves estimating the ARDL model by substituting an error
correction term for the variables in levels. The significance of the error correction term is
regarded as a further proof for the long-run relationship between the chosen variables. As
shown by Pesaran et al. (2001), this method should be used in the subsequent estimation of
short-run dynamics because it has a more parsimonious specification than the version given
in equation (2). The model with an error correction term is given in equation (3). Here, ECTt-1
is the one period lag residuals saved from equation (1). As shown in Table 2, the error
correction term for the economic activity and growth in terms of a production equation is
estimated as –0.23. This is significant at the 10 per cent level of significance. The value
suggests that, after a shock, economic growth converges to the equilibrium. Approximately
23 per cent of the deviation is therefore corrected within one year.
n1
n2
n3
n4
k 1
k 0
k 0
k 0
 log( Yt )      k  log( Yt k )    k  log( Lt  k )    k  log( K t k )   k  log( H t  k ) 
n5
  k  log( Et  k )  ECTt 1  u t
(3)
k 0
17
Table 2: ARDL bound test with an error correction term (ECT)
Dependent Variable: DLY
Sample (adjusted): 1971 2010
Included observations: 40 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C
0.025523
0.005641
4.524305
0.0001
DLL(1)
0.434682
0.272450
1.595458
0.1199
DLK(1)
–0.043286
0.033986
–1.273625
0.2114
DLH(1)
–0.039493
0.025595
–1.542976
0.1321
DLE(1)
0.046163
0.100996
0.457078
0.6505
ECT(-1)
–0.227068
0.122454
1.854317
0.0724
R-squared
0.301297
Mean dependent var
0.031562
Adjusted R-squared
0.198546
S.D. dependent var
0.015282
S.E. of regression
0.013681
Akaike info criterion
–5.608100
Sum squared resid
0.006364
Schwarz criterion
–5.354768
Log likelihood
118.1620
Hannan-Quinn criter.
–5.516503
F-statistic
2.932313
Durbin-Watson stat
1.958128
Prob(F-statistic)
0.026328
Results based on bound testing cointegration method suggest that energy consumption
and Australian economic growth, despite being positively related, are not statistically
significant either in the short run or the long run. To ascertain this finding, we have also
conducted a multivariate Granger causality test to see whether there are any feed-in effects
between energy use and economic growth. According to the multivariate Granger causality
test, energy (Et) is said to Granger cause GDP, if the prediction error of current GDP
declines, as we include lagged values of energy in addition to lagged values of GDP. In other
words, the coefficients of lagged energy terms are statistically significant. One of the
requirements of Granger causality test is that series are stationary. On account of the fact that
18
all variables are first differenced stationary, we conduct the multivariate Granger causality
test in the following form:
n
n
n
n
n
i 1
i 1
i 1
i 1
i 1
LYt   1Y    iY LYt i    iY LEt i    iY LLt i   iY LK t i   iY LH t i  a1Y LYt 1 
 a 2Y LEt 1  a 3Y LLt 1  a 4Y LK t 1  a 5Y LH t 1   Yt
n
(4)
n
n
n
n
i 1
i 1
i 1
i 1
LEt   1E    iE LYt i    iE LEt i    iE LLt i   iE LK t i   iE LH t i  a1E LYt 1 
i 1
 a 2 E LEt 1  a 3 E LLt 1  a 4 E LK t 1  a 5 E LH t 1   Et
(5)
n
n
n
n
n
i 1
i 1
i 1
i 1
i 1
LLt   1L    iL LYt i    iL LEt i    iL LLt i   iL LK t i   iL LH t i  a1L LYt 1 
 a 2 L LEt 1  a 3 L LLt 1  a 4 L LK t 1  a 5 L LH t 1   Lt
(6)
n
n
n
n
n
i 1
i 1
i 1
i 1
i 1
LK t   1K    iK LYt i    iK LEt i    iK LLt i   iK LK t i   iK LH t i  a1K LYt 1 
 a 2 K LEt 1  a 3 K LLt 1  a 4 K LK t 1  a 5 K LH t 1   Kt
n
(7 )
n
n
n
n
i 1
i 1
i 1
i 1
LH t   1H    iH LYt i    iH LEt i    iH LLt i   iH LK t i   iH LH t i  a1H LYt 1 
i 1
 a 2 H LEt 1  a 3 H LLt 1  a 4 H LK t 1  a 5 H LH t 1   Ht
(8)
Results of the Granger causality test vary according to the lag length used in the
estimation. Optimum lag length is decided by the Akaike information criterion. The test
results using 7 lags are given in Table 3. According to the results, the null hypothesis that
DLE does not Granger cause DLY, or that DLY does not Granger cause DLE, cannot be
rejected. The interrelationship between the two variables seems not strong. This finding is in
contrast to earlier Australian studies, which showed either bidirectional causality between the
two (i.e., Mahadevan and Asafu-Adjaye, 2007; Shahiduzzaman and Alam, 2012), or
unidirectional causality with economic growth leading energy consumption (i.e., Fatai et al.,
2004; Narayan and Smyth, 2005; Chontanawat et al., 2008), or unidirectional causality with
energy consumption leading economic growth (i.e., Narayan and Prasad, 2008).
19
Table 3: Results of Granger causality test
Sample: 1970 2011
Included observations: 39
Dependent variable: DLY
Dependent variable: DLE
Excluded
Chi-sq
df
Prob.
Excluded
Chi-sq
df
Prob.
DLL
1.409789
2
0.4942
DLY
0.061295
2
0.9698
DLK
0.718177
2
0.6983
DLL
7.596891
2
0.0224
DLH
2.410475
2
0.2996
DLK
2.191732
2
0.3342
DLE
1.209791
2
0.5461
DLH
4.850036
2
0.0885
All
5.246247
8
0.731
All
13.35655
8
0.1002
6.
Conclusion and policy implications
In sum, we have applied ARDL bound test to time series data from 1970 to 2011 to infer the
causal relationship between energy consumption and economic growth in Australia. To
reduce potential omitted-variable biases, we have considered a multivariate model including
labour, capital, human capital, in addition to energy consumption and real GDP. The model is
based on the production function framework, which is formulated to synthesize the
approaches from neoclassical and endogenous growth models, as well as from an ecological
economics viewpoint. The main finding is that there is no causality between energy
consumption and economic growth in Australia. The results in this paper support the
‘neutrality’ hypothesis, which views energy consumption as a small component of real GDP
(Payne, 2010). As a result, energy consumption should not have a significant impact on
economic growth. Furthermore, the finding is in line with structural change of the Australian
economy toward a more service-intensive economy, which requires less energy intensity than
an economy relying on a large manufacturing industry. Although energy remains important,
energy-saving technical progress in the manufacturing industry has allowed less energy to be
20
used per unit output and has reduced the constraint that energy resources place on the output
of the economy and economic growth (Stern, 2011). This has important consequences for
energy conservation and climate change policies, especially as Australia grapples with
measures to improve energy security and concomitantly reduce greenhouse gases emissions.
Our results suggest, at least ostensibly, that energy conservation and carbon pricing policies
may not necessarily have an adverse effect on Australia’s economic growth.
References
Aghion, P., Howitt, P., 1997. Endogenous Growth Theory. MIT Press, Massachusetts.
Akarca, A.T., Long, T.V., 1979. Energy and employment: a time series analysis of the causal
relationship. Resources and Energy 2(2–3), 151–162.
Akarca, A.T., Long, T.V., 1980. Relationship between energy and GNP: a reexamination.
Journal of Energy and Development 5 (2), 326–331.
Altinay, G., Karagol, E., 2005. Electricity consumption and economic growth: evidence from
Turkey. Energy Economics 27(6), 849–856.
Apostolakis, B. E., 1990. Energy-capital substitutability / complementarity: the dichotomy.
Energy Economics 12(1), 48–58.
Asafu-Adjaye, J., 2000. The relationship between energy consumption, energy prices, and
economic growth: time series evidence from Asian developing countries. Energy
Economics 22(6), 615–625.
Ayres, R.U., Warr, B., 2009. The Economic Growth Engine: How Energy and Work Drive
Material Prosperity. Edward Elgar Publishing, Cheltenham, Northampton.
Berndt, E. R., Wood, D.O. 1979. Engineering and econometric interpretations of energycapital complementarity. American Economic Review 69(3), 342–354.
Bowden, N., Payne, J.E., 2010. Sectoral analysis of the causal relationship between
renewable and non-renewable energy consumption and real output in the US. Energy
Sources, Part B: Economics, Planning, and Policy 5(4), 400–408.
Chen, S.T., Kuo, H.I., Chen, C.C., 2007. The relationship between GDP and electricity
consumption in 10 Asian countries. Energy Policy 35(4), 2611–2621.
Cheng, B.S., 1996. An investigation of cointegration and causality between energy
consumption and economic growth. Journal of Energy and Development 21(1), 73–84.
Cheng, B.S., Lai, T.W., 1997. An investigation of cointegration and causality between energy
consumption and economic activity in Taiwan. Energy Economics 19(4), 435–444.
Chontanawat, J., Hunt, L.C., Pierse, R., 2008. Does energy consumption cause economic
growth? Evidence from a systematic study of over 100 countries. Journal of Policy
Modeling 30(2), 209–220.
Dolado, J.J., Lütkepohl, H., 1996. Making wald tests work for cointegrated VAR systems.
Econometric Theory 15(4), 369–386.
21
Engle, R.F., Granger, C.W.J., 1987. Co-integration and error correction: representation,
estimation, and testing. Econometrica, 55(2), 251-276.
Erol, U., Yu, E.S.H., 1987a. Time series analysis of the causal relationships between US
energy and employment. Resources and Energy 9(1), 75–89.
Erol, U., Yu, E.S.H., 1987b. On the causal relationship between energy and income for
industrialized countries. Journal of Energy and Development 13(1), 113–122.
Fatai, K., Oxley, L., Scrimgeour, F.G., 2004. Modelling the causal relationship between
energy consumption and GDP in New Zealand, Australia, India, Indonesia, The
Philippines and Thailand. Math. Comput. Simul. 64 (3–4), 431–445.
Ferguson, R., Wilkinson, W., Hill, R., 2000. Electricity use and economic development.
Energy Policy 28(13), 923–934.
Francis, B.M., Moseley, L., Iyare, S.O., 2007. Energy consumption and projected growth in
selected Caribbean countries. Energy Economics 29(6), 1224–1232.
Frondel, M., C. M Schmidt, 2002. The capital-energy controversy: An artifact of cost shares?
The Energy Journal 23(3), 53–79.
Galor, O., D. N. Weil., 2000. Population, technology and growth: From Malthusian regime to
the demographic transition. American Economic Review 90(4), 806–828.
Ghali, K.H., El-Sakka, M.I.T., 2004. Energy and output growth in Canada: a multivariate
cointegration analysis. Energy Economics 26(2), 225–238.
Ghosh, S., 2002. Electricity consumption and economic growth in India. Energy Policy 30(2),
125–129.
Glasure, Y.U., 2002. Energy and national income in Korea: further evidence on the role of
omitted variables. Energy Economics 24(), 355–365.
Glasure, Y.U., Lee, A.R., 1995. Relationship between US energy consumption and
employment: further evidence. Energy Sources 17(5), 509–516.
Glasure, Y.U., Lee, A.R., 1996. The macroeconomic effects of relative prices, money, and
federal spending on the relationship between US energy consumption and employment.
Journal of Energy and Development 22(1), 81–91.
Glasure, Y.U., Lee, A.R., 1998. Cointegration, error correction, and the relationship between
GDP and energy: the case of South Korea and Singapore. Resource and Energy
Economics 20(1), 17–25.
Granger, C.W.J., Newbold, P., 1974. Spurious regressions in econometrics. Journal of
Econometrics 2(2), 111–120.
Harris, R., Sollis, R., 2003. Applied Time Series Modelling and Forecasting. Wiley,
Chichester.
Hondroyiannis, G., Lolos, S., Papapetrou, E., 2002. Energy consumption and economic
growth: assessing the evidence from Greece. Energy Economics 24(2), 319–336.
Huang, B.N., Hwang, M.J., Yang, C.W., 2008. Causal relationship between energy
consumption and GDP growth revisited: a dynamic panel data approach. Ecological
Economics 67(1), 41–54.
22
Jinke, L., Hualing, S., Dianming, G., 2008. Causality relationship between coal consumption
and GDP: difference of major OECD and on-OECD countries. Applied Energy 85(6),
421–429.
Johansen, S., 1991. Estimation and hypothesis testing of cointegrated vectors in Gaussian
vector autoregressive models. Econometrica 59(6), 1551–1580.
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on
cointegration with applications to the demand for money. Oxford Bulletin of Economics
and Statistics 52(2), 169–210.
Karanfil, F., 2009. How many times again will we examine the energy-income nexus using a
limited range of traditional econometric tools?. Energy Policy 37(4), 1191–1194.
Kraft, J., Kraft, A., 1978. On the relationship between energy and GNP. Journal of Energy
and Development 3(2), 401–403.
Lee, C.C., 2005. Energy consumption and GDP in developing countries: a cointegrated panel
analysis. Energy Economics 27(3), 415–427.
Lee, C.C., 2006. The causality relationship between energy consumption and GDP in G-11
countries revisited. Energy Policy 34(9), 1086–1093.
Lee, C.C., Chang, C.P., 2008. Energy consumption and economic growth in Asian
economies: a more comprehensive analysis using panel data. Resource and Energy
Economics 30(1), 50–65.
Lee, C.C., Chang, C.P., Chen, P.F., 2008. Energy-income causality in OECD countries
revisited: the key role of capital stock. Energy Economics 30(5), 2359–2373.
Lucas, R. E., 2002. The industrial revolution: past and future. In Lectures on Economic
Growth. R. E. Lucas: 109–188. Harvard University Press. Cambridge, MA.
Mahadevan, R., Asafu-Adjaye, J., 2007. Energy consumption, economic growth and prices: a
reassessment using panel VECM for developed and developing countries. Energy Policy
35(4), 2481–2490.
Masih, A.M.M., Masih, R., 1996. Energy consumption, real income and temporal causality:
results from a multi-country study based on cointegration and error-correction modelling
techniques. Energy Economics 18(3), 165–183.
Masih, A.M.M., Masih, R., 1997. On temporal causal relationship between energy
consumption, real income, and prices: some new evidence from Asian-energy dependent
NICs based on a multivariate cointegration/vector error correction approach. Journal of
Policy Modeling 19(4), 417–440.
Masih, A.M.M., Masih, R., 1998. A multivariate cointegrated modeling approach in testing
temporal causality between energy consumption, real income, and prices with an
application to two Asian LDCs. Applied Economics 30(10), 1287–1298.
Mehrara, M., 2007. Energy consumption and economic growth: the case of oil exporting
countries. Energy Policy 35(5), 2939–2945.
Murray, D.A., Nan, G.D., 1992. The energy and employment relationship: a clarification.
Journal of Energy and Development 16(1), 121–131.
Narayan, P., Prasad, A., 2008. Electricity consumption-real GDP causality nexus: evidence
from a bootstrapped causality test for 30 OECD countries. Energy Policy 36 (2), 910–918.
23
Narayan, P., Smyth, R., 2005. Electricity consumption, employment and real income in
Australia evidence from multivariate Granger causality tests. Energy Policy 33 (9), 1109–
1116.
Narayan, P.K., Smyth, R., 2007. Energy consumption and real GDP in G7 countries: new
evidence from panel cointegration with structural breaks. Energy Economics 30(5), 2331–
2341.
Oh, W., Lee, K., 2004a. Causal relationship between energy consumption and GDP revisited:
the case of Korea 1970-1999. Energy Economics 26(1), 51–59.
Oh, W., Lee, K., 2004b. Energy consumption and economic growth in Korea: testing the
causality relation. Journal of Policy Modeling 26(8–9), 973–981.
Paul, S., Bhattacharya, R.N., 2004. Causality between energy consumption and economic
growth in India: a note on conflicting results. Energy Economics 26(6), 977–983.
Payne, J.E., 2010. Survey of the international evidence on the causal relationship between
energy consumption and growth. Journal of Economic Studies 37 (1), 53–95.
Pesaran, M.H., Shin, Y., Smith, R., 2001. Bounds testing approaches to the analysis of level
relationships. Journal of Applied Econometrics 16(3), 289–326.
Pirlogea, C., Cicea, C., 2012. Econometric perspective of the energy consumption and
economic growth relation in European Union. Renewable and Sustainable Energy
Reviews 16(8), 5718–5726.
Sari, R., Ewing, B.T., Soytas, U., 2008. The relationship between disaggregate energy
consumption and industrial production in the United States: an ARDL approach. Energy
Economics 30(5), 2302–2313.
Shahiduzzaman, M., Alam, K., 2012. Cointegration and causal relationships between energy
consumption and output: Assessing the evidence from Australia. Energy Economics 34(6),
2182–2188.
Sharma, S. S., 2010. The relationship between energy and economic growth: Empirical
evidence from 66 countries. Applied Energy 87(11), 3565–3574.
Shiu, A., Lam, P.L., 2004. Electricity consumption and economic growth in China. Energy
Policy 32(1), 47–54.
Stern, D.I., 1993. Energy and economic growth in the USA: a multivariate approach. Energy
Economics 15 (2), 137–150.
Stern, D.I., 2000. A multivariate cointegration analysis of the role of energy in the US
macroeconomy. Energy Economics 22 (2), 267–283.
Stern, D.I., 2011. The role of energy in economic growth. Ecological Economics Reviews
1219 (1), 26–51.
Solow, R. M., 1956. A contribution to the theory of economic growth. Quarterly Journal of
Economics, 70 (1), 65–94.
Solow, R. M., 1974. Intergenerational equity and exhaustible resources. Review of Economic
Studies 41: Symposium on the Economics of Exhaustible Resources, 29–46.
Soytas, U., Sari, R., 2003. Energy consumption and GDP: causality relationship in G-7 and
emerging markets. Energy Economics 25(1), 33–37.
24
Soytas, U., Sari, R., 2006a. Energy consumption and income in G7 countries. Journal of
Policy Modeling 28(7), 739–750.
Soytas, U., Sari, R., 2006b. Can China contribute more to the fight against global warming?.
Journal of Policy Modeling 28(8), 837–846.
Soytas, U., Sari, R., 2007. The relationship between energy and production: evidence from
Turkish manufacturing industry. Energy Economics 29(6), 1151–1165.
Toda, H.Y., Yamamoto, T., 1995. Statistical inference in vector autoregressions with possibly
integrated processes. Journal of Econometrics 66(1–2), 225–250.
Toman, T., Jemelkova, B., 2003. Energy and economic development: an assessment of the
state of knowledge. Energy Journal 24(4), 93–112.
World Bank, 2012. The World Bank DataBank, http://data.worldbank.org/.
Yang, H.Y., 2000a. A note on the causal relationship between energy and GDP in Taiwan.
Energy Economics 22(3), 309–317.
Yang, H.Y., 2000b. Coal consumption and economic growth in Taiwan. Energy Sources
22(2), 109–115.
Yoo, S.H., 2005. Electricity consumption and economic growth: evidence from Korea.
Energy Policy 33(12), 1627–1632.
Yoo, S.H., 2006a. Causal relationship between coal consumption and economic growth in
Korea. Applied Energy 83(11), 1181–1189.
Yoo, S.H., 2006b. Oil consumption and economic growth: evidence from Korea. Energy
Sources, Part B 1(3), 235–243.
Yoo, S.H., 2006c. The causal relationship between electricity consumption and economic
growth in the ASEAN countries. Energy Policy 34(18), 3573–3582.
Yoo, S.H., Jung, K.-O., 2005. Nuclear energy consumption and economic growth in Korea.
Progress in Nuclear Energy 46(2), 101–109.
Yoo, S.H., Kim, Y., 2006. Electricity generation and economic growth in Indonesia. Energy
31(14), 2890–2899.
Yu, E.S.H., Choi, J.Y., 1985. The causal relationship between energy and GNP: an
international comparison. Journal of Energy and Development 10(2), 249–272.
Yu, E.S.H., Hwang, B., 1984. The relationship between energy and GNP: further results.
Energy Economics 6(3), 186–190.
Yu, E.S.H., Jin, J.C., 1992. Cointegration tests of energy consumption, income, and
employment. Resources and Energy 14(3), 259–266.
Yuan, J., Kang, J., Zhao, C., Hu, Z., 2008. Energy consumption and economic growth:
evidence from China at both aggregated and disaggregated levels. Energy Economics
30(6), 3077–3094.
Zachariadis, T., 2007. Exploring the relationship between energy use and economic growth
with bivariate models: new evidence from G-7 countries. Energy Economics 29(6), 1233–
1253.
25