Munich Personal RePEc Archive
On the dynamics of energy consumption
and employment in public and private
ICFAI University, Tripura
11. May 2010
Online at http://mpra.ub.uni-muenchen.de/24076/
MPRA Paper No. 24076, posted 23. July 2010 / 13:23
On the dynamics of energy consumption and employment in
public and private sector
Aviral Kumar Tiwari
Research scholar and Faculty of Applied Economics,
Faculty of Management, ICFAI University Tripura,
Kamalghat, Sadar, West Tripura, Pin-799210
Email-Id: email@example.com & firstname.lastname@example.org
On the dynamics of energy consumption and employment in public
and private sector
This study intended to analyze the direction of Granger-causality between energy consumption
and employment in public and private sector. We have adopted DL approach for Granger-
causality analysis. We found from the whole analysis that there is evidence of bidirectional
causality between energy consumption and employment in organized public and private sector.
Therefore our study supports for our third testable hypothesis i.e., “feedback hypothesis”.
Keywords: Energy consumption, employment in public and private sector, Granger-causality.
JEL classification code: C22, J45, J48.
The relationship between energy consumption, economic growth and employment and
the policy implications of the empirical findings has been comprehensively examined within the
energy economics literature. Griffin and Gregory (1976), Berndt and Wood (1979), and Berndt
(1980, 1990) emphasize the substitutability or complementarity between energy and the factors
of production and the interplay with technical progress and productivity within a neoclassical
(1993), Kemfert and Welsch (2000), and Smulders and de Nooij (2003), among others, explore
the role of energy within a general equilibrium framework. While the work cited above has been
important in understanding the role of energy in the economy, there has been a growing literature
on the causal relationship between energy consumption and economic growth utilizing a variety
of time series econometric techniques. This study has also made an effort in the direction of
examine the role played by energy in employment sector of India in bivariate framework.
Rest of the paper is organized as follows. Second section presents a comprehensive literature
review followed by data source objectives and estimation methodology in section third. Section
forth presents the data analysis and results followed by conclusions in the fifth section.
II. Literature review
We can classify the studies to date into four groups on the basis of their findings. First, a
large number of studies find unidirectional causality running from electricity or energy
consumption (both aggregate and disaggregate level) to GDP or employment. Studies worthy of
mention include those by Altinay and Karagol (2005) for Turkey, which find strong evidence for
the period 1950-2000, Lee and Chang (2005) in Taiwan for the period 1954-2003, Shiu and Lam
(2004) in China for 1971-2000, and Soytas and Sari (2003) for Turkey, France, Germany and
Japan, Wolde-Rufale (2004) in Shanghai for the period 1952-1999, Morimoto and Hope (2004)
in Sri-Lanka for the period 1960-98.
Second, those that finds unidirectional causality running from economic growth or gross
domestic product to electricity or energy consumption. These include Ghosh (2002) in India for
the period 1950-1997, Cheng (1999) in India for the period 1952-1995, Fatai et al. (2004) in
New Zealand and Australia for the period 1960-1999, and Hatemi and Irandoust (2005) in
Sweden for the period 1965-2000, Cheng and Lai (1997) in Taiwan for the period 1954-1993,
Chang and Wong (2001) in Singapore for the period 1975-1995 and Aqeel and Butt (2001) in
Pakistan for the period 1955-1996.
A third group comprises studies that find bi-directional causality. This include Soytas and
Sari (2003) for Argentina, Oh and Lee (2004) for Korea in 1970-1999, Yoo (2005) also for
Korea in 1970-2002 and Glasure (2002) in South Korea for the period 1961-1990, Jumbe (2004)
in Malawi for the period 1970-1999, Ghali and El-Sakka (2004) in Canada for the period of
1961-1997, Hwang and Gum (1992) in Taiwan for the period 1961-1990.
And the last group comprises studies that find no causal linkages between energy or
electricity consumption and economic growth, such as Cheng (1995) in US for the period 1947-
1990, and Stern (1993) in USA for the period 1947-1990, Akarca and Long (1980) in US for the
period 1950-1968 and 1950-1970, Yu and Hwang (1984) in US for the period 1947-1979.
III. Objective, data, hypothesis and estimation methodology
The first subsection of this section presets about the objective set for this study and the
source of data followed by hypothesis formulation in second sub section and in third subsection
methodology to be used for estimation has been presented.
III.I. Objectives and data
In this we have tried to estimate the direction of causality among private sector
employment, public sector employment and energy consumption. This objective is justified as
best of my knowledge this kind of study has not been conducted so for in India.
We have sourced data from Hand Book of Statistics of Indian Economy by Reserve Bank
of India (RBI). Time period of this study is 1971-2006.
III.II. Testable hypothesis formulation
The direction of causality between energy consumption and economic growth, measured by
either employment or real output, can be summarized in four testable hypotheses mentioned as
The first, hypothesis is the “growth hypothesis” which suggests that energy consumption
contributes directly to economic growth within the production process. In this case, the policy
implication is that energy conservation policies which reduce energy consumption may possibly
reduce real output. The growth hypothesis is supported if there is unidirectional Granger-
causality running from energy consumption to real output or employment. Example of this types
of studies are Altinay and Karagol (2005), Lee and Chang (2005), Shiu and Lam (2004), and
Soytas and Sari (2003), Wolde-Rufale (2004), Morimoto and Hope (2004).
The second, hypothesis is the “conservation hypothesis” which implies that energy
conservation policies designed to reduce energy consumption and waste may not reduce real
output. Unidirectional Granger-causality running from real output or employment to energy
consumption would lend support for the conservation hypothesis. Examples of such kind of
studies are Ghosh (2002), Cheng (1999), Fatai et al. (2004), Hatemi and Irandoust (2005), Cheng
and Lai (1997), Chang and Wong (2001) and Aqeel and Butt (2001).
The third, hypothesis is the “feedback hypothesis” which asserts that energy consumption
and real output or employment are interdependent and act as complements to each other. The
existence of bidirectional Granger-causality between energy consumption and real output or
employment substantiates the feedback hypothesis. Examples of this hypothesis are Soytas and
Sari (2003), Oh and Lee (2004), Yoo (2005), Glasure (2002), Jumbe (2004), Ghali and El-Sakka
(2004), and last but not least Hwang and Gum (1992).
Finally, the fourth hypothesis is the “neutrality hypothesis” which suggests that energy
consumption as a relatively minor factor in the production of real output in which case energy
conservation policies may not adversely impact real output and hence employment. The absence
of Granger-causality between energy consumption and real output or employment is supportive
of the neutrality hypothesis. Examples of this hypothesis are Cheng (1995), Stern (1993), Akarca
and Long (1980), and last but not least Yu and Hwang (1984).
III.III. Estimation methodology
In the present study energy consumption has been measured by Electric power consumption
(kWh per capita) as % of GDP, and employment (in millions) has been considered in two sectors
private and public organized sector. All variables have been analyzed by making them in natural
logrthism form as it minimizes the fluctuations in the series and makes the series of less order of
autoregressive. To know the causality among the test variables the standard test to be used in the
study is Engle-Granger approach in VECM framework. But this approach requires certain pre-
estimations (like testing the stationarity of the variables included in the VECM analysis and
seeking the cointegration of the series) without which, conclusions drawn from the estimation
will not be valid. Granger non-causality test in an unrestricted VAR model can be simply
conducted by testing whether some parameters are jointly zero, usually by a standard (Wald) F-
test. This approach in integrated or cointegrated systems has been examined by Sims et al.
(1990) and Toda and Phillips (1993). These studies have shown that the Wald test for non-
causality in an integrated or cointegrated unrestricted VAR system will have nonstandard limit
These results have given rise to alternative testing procedures, such as Toda and Phillips
(1993) and Mosconi and Giannini (1992), but they require sequential testing and are
computationally burdensome. Toda (1995) has shown that pretesting for cointegration rank in
Johansen-type error correction mechanisms (ECMs) are sensitive to the values of the nuisance
parameters, thus causality inference based upon ECM may be severely biased. Toda and
Yamamoto (1995) and Dolado and LÄutkepohl (1996) propose a method of estimating a VAR
for series in levels and test general restrictions on the parameter matrices even if the series are
integrated or cointegrated. This method is theoretically simpler and computationally relatively
straightforward in causality tests. They develop a modified version of the Granger causality test
which involves a modified Wald (MWALD) test in an intentionally augmented VAR model.
Once the optimal order of the VAR process, p, is selected, Toda and Yamamoto (TY) (1995)
propose estimating a VAR(p + dmax) model where dmax is the maximal order of integration that
we suspect might occur in the true generation process. Linear or nonlinear restrictions on the first
p coefficient matrices of the model can therefore be tested using standard Wald (F-) tests
ignoring the last dmax lagged vectors of the variables. Dolado and LÄutkepohl (DL) (1996) also
propose estimating an augmented VAR with the difference that they add only one lag to the true
lag length of the model. One estimates the VAR(p+1) model and perform the standard Wald (F-)
tests ignoring the last lag of the vector. The advantage of DL and TY are that they are
computationally relatively simple and do not require pretesting for integration or cointegration of
the data series. These tests are especially attractive when one is not sure whether series are
stationary or integrated of order one. Toda and Yamamoto (1995) proves that the Wald (F-)
statistic used in this setting converges in distribution to a χ2 random variable, no matter whether
the process is stationary or nonstationary. The preliminary unit root and cointegration tests are
not necessary to implement the DL test, since the testing procedure is robust to the integration
and cointegration properties of the process. Consider the following VAR( p) model:
Y = γ + AY − + ... + A Y − + ε ......... .........( )
p (t p)
Where Yt, γ, and εt~(0, ) are n-dimensional vector and Ak is an n×nmatrix of parameters
for lag k. to implement the TY test the following augmented VAR(p+d) model to be utilized for
the test of causality is estimated,
Y = ˆ
γ + AY − +...+ A Y − + A Y
+ ˆε ...................( )
p (t p)
p d )
Where the circumflex above a variable denotes its Ordinary Least Square (OLS)
estimates. The order p of the process is assumed to be known, and the d is the maximal order of
integration of the variables. Since the true lag length p is rarely known in practice, it can be
estimated by some consistent lag selection criteria. In the present study we have used SIC
(preferably) and AIC. It is important to note that if the maximal order of integration is d=1, then
TY test becomes similar to DL test. The jth element of Yt dose not Granger-cause the ith element
of Yt, if the following null hypothesis is not rejected:
Ho: The row i, column j element in Ak equals zero for k= 1,…,p.
The null hypothesis is tested by Wald (F-) test which is named modified Wald
(MWALD) test in case of the augmented VAR outlined above.
For example, in a bivariate VAR model with the optimal lag length, suppose it is 3, Eq.
(2) is re-estimated by OLS setting the lag length 4 (3+1) as suggested by DL test.
4 + 1t
Where L denotes logarithms of X and Y variables. The hypothesis that X variable dose
not Granger-cause Y can be constructed as:
Ho : 1
a = a = a = 0
Whereas the hypothesis that Y variable does not Granger-cause X can be constructed as:
Ho : 1
a = a = a = 0
and these joint hypothesis can be tested by MWALD test.
Finally, stability of VAR analysis has been performed as in order to draw valid
conclusions from the above system, it is necessary that the VAR be stable or stationary. If the
estimated VAR is stable then the inverse roots of characteristics Autoregressive (AR) polynomial
will have modulus less than one and lie inside the unit circle. There will be kp roots, where k is
the number of endogenous variables and p is the largest lag.VAR stability has been checked by
ignoring last lag from the analysis as to test the joint hypothesis last lag is ignored.
IV. Data analysis and results interpretation
To proceed for analyzing Granger-causality in DL framework we require a prior
knowledge of lag length to be included in VAR framework. Since we do not have any idea about
that therefore we have carried out lag length selection test for max 3, max 4 and max 5 lags.