Energy Economics 33 (2011) 770–781

Contents lists available at ScienceDirect

Energy Economics
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e n e c o

A dynamic panel study of economic development and the electricity

consumption-growth nexus
Nicholas Apergis a, James E. Payne b,⁎
a
Department of Banking and Financial Management, University of Piraeus, Karaoli and Dimitriou 80, Piraeus, ATTIKI 18534, Greece
b
Department of Economics, Illinois State University, Normal, IL 61790-4100, United States

a r t i c l e i n f o a b s t r a c t

Article history: This study examines the relationship between electricity consumption and economic growth for 88 countries
Received 24 May 2010 categorized into four panels based on the World Bank income classification (high, upper middle, lower middle, and
Received in revised form 21 December 2010 low income) within a multivariate panel framework over the period 1990–2006. The Larsson et al. (2001) panel
Accepted 23 December 2010
cointegration test indicates there is a long-run equilibrium relationship between real GDP, coal consumption, real
Available online 20 January 2011
gross fixed capital formation, and the labor force for the high, upper middle, and lower middle income country
JEL Classification:
panels. The results from the panel vector error correction models reveal (1) bidirectional causality between
C33 electricity consumption and economic growth in both the short- and long-run for the high income and upper-middle
01 income country panels; (2) unidirectional causality from electricity consumption to economic growth in the short-
04 run, but bidirectional causality in the long-run for the lower-middle income country panel; and (3) unidirectional
050 causality from electricity consumption to economic growth for the low income country panel.
Q4 © 2011 Elsevier B.V. All rights reserved.

Keywords:
Electricity consumption
Growth
Panel
Granger-causality

1. Introduction
Electricity serves an important role in both the production and
consumption of goods and services within an economy. The availability
of electricity has been a major contributor to the technological and
scientific advancements that have improved the standard of living across
countries. With population growth, urbanization, and industrialization
of economies, the infrastructure for electricity has emerged as an
important factor in a country's growth prospects. Furthermore, the
growth of the information and communication technologies in advanced
economies confirms the importance of electricity. In a study of over 100
countries, Ferguson et al. (2000) present evidence of a high correlation
between electricity usage and the level of economic development and
growth. However, a high correlation does not necessarily imply a causal
relationship. Thus, understanding the causal relationship between
electricity consumption and economic growth is important in the design
and effectiveness of energy and environmental policies.
The causal relationship between electricity consumption and
economic growth can be broken down into four testable hypotheses.
⁎ Corresponding author. Tel.: + 1 309 438 5669.
E-mail addresses: napergis@unipi.gr (N. Apergis), jepayne@ilstu.edu (J.E. Payne).
First, the growth hypothesis asserts that electricity consumption has a
significant influence on economic growth directly and/or as a
complement to labor and capital in the production process. The
growth hypothesis is supported if there is unidirectional causality
from electricity consumption to economic growth. In this context,
energy conservation policies which reduce electricity consumption
may have an adverse impact on economic growth. Second, the
conservation hypothesis stipulates that electricity consumption is
driven by economic growth. The conservation hypothesis is confirmed
if there is unidirectional causality from economic growth to electricity
consumption. Under this scenario, energy conservation policies
designed to reduce electricity consumption may not have an adverse
impact on economic growth. Third, the feedback hypothesis empha-
sizes the interdependent relationship between electricity consump-
tion and economic growth. The feedback hypothesis is confirmed by
the presence of bidirectional causality between electricity consump-
tion and economic growth. This complementary relationship opens
the possibility that energy conservation policies which reduce
electricity consumption may impact economic growth. Likewise,
such fluctuations in economic growth may very well be transmitted
back to electricity consumption. Finally, the neutrality hypothesis
assumes that electricity consumption plays a relatively minor role in
economic growth. The neutrality hypothesis is supported by the

0140-9883/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.eneco.2010.12.018
Table 1
Summary of the electricity consumption-economic growth literature.
Author(s)
Murray and Nan (1996)
Yang (2000)
Aqeel and Butt (2001)
Ghosh (2002)
Fatai et al. (2004)
Jumbe (2004)
Morimoto and Hope (2004)
Shiu and Lam (2004)
Thoma (2004)
Wolde-Rufael (2004)
Altinay and Karagol (2005)
Lee and Chang (2005)
Narayan and Smyth (2005)
Yoo (2005)
Countries
Canada (1970–1990A)
Colombia (1970–1990A)
El Salvador (1970–1990A)
France (1970–1990A)
Germany (1970–1990A)
Hong Kong (1970–1990A)
India (1970–1990A)
Indonesia (1970–1990A)
Israel (1970–1990A)
Kenya (1970–1990A)
Luxembourg (1970–1990A)
Malaysia (1970–1990A)
Mexico (1970–1990A)
Norway (1970–1990A)
Pakistan (1970–1990A)
Philippines (1970–1990A)
Portugal (1970–1990A)
Singapore (1970–1990A)
South Korea (1970–1990A)
Turkey (1970–1990A)
UK (1970–1990A)
US (1970–1990A)
Zambia (1970–1990A)
Taiwan (1954–1997A)
Pakistan (1955–1996A)
India (1950–1997A)
Australia (1960–1999A)
Malawi (1970–1999A)
Sri Lanka (1960–1998A)
China (1971–2000A)
US (1973:1–2000:1M)
Shanghai (1952–1999A)
Turkey (1950–2000A)
Taiwan (1954–2003A)
Australia (1966–1999A)
Korea (1970–2002A)
Methodology
Granger-causality; VAR
Engle-Granger; no
cointegration; VAR
Engle-Granger; No
cointegration; VAR
Johansen-Juselius; no
cointegration; VAR
Granger-causality;
Toda-Yamamoto; causality;
ARDL bounds test;
Johansen-Juselius;
cointegration; VEC
Engle-Granger; cointegration:
VEC
Engle-Granger; no
cointegration; VAR
Johansen-Juselius
cointegration; VEC
Engle-Granger; no
cointegration; VAR
Toda-Yamamoto causality
Dolado-Lutkepohl causality
Johansen-Juselius;
cointegration; VEC
ARDL bounds test;
cointegration; VEC
Johansen-Juselius
cointegraton; VEC
Main variables Other variables Conclusion(s)
Electricity consumption; Canada, ELC → Y
real GDP Colombia, Y → ELC
El Salvador, Y → ELC
France, ELC ≠ Y
Germany, ELC ≠ Y
Hong Kong, ELC → Y
India, ELC ≠ Y
Indonesia, Y → ELC
Israel, ELC ≠ Y
Kenya, Y → ELC
Luxembourg, ELC ≠ Y
Malaysia, ELC↔Y
Mexico, Y → ELC
Norway, ELC ≠ Y
Pakistan, ELC → Y
Philippines, ELC ≠ Y
Portugal, ELC ≠ Y
N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781
Singapore, ELC → Y
South Korea, ELC↔Y
Turkey, ELC → Y
UK, ELC ≠ Y
US, ELC ≠ Y
Zambia, ELC ≠ Y
Electricity consumption; Taiwan, ELC↔Y
real GDP
Electricity consumption Pakistan, ELC → Y
per capita; real GDP
per capita
Electricity consumption India, Y → ELC
per capita; real GDP
per capita
Electricity consumption; Consumer prices Australia, JJ Y → ELC,
real GDP TY Y → ELC, ARDL Y → ELC
Electricity consumption; Malawi, ELC↔Y
GDP; agricultural GDP;
Electricity production; Sri Lanka, ELP → Y
real GDP
Electricity consumption; China, ELC → Y
real GDP
Total, commercial, industrial, US, IP → ELC, IP → CELC,
other, and residential IP → IELC, OELC ≠ IP, RELC ≠ IP
electricity usage; industrial
production
Electricity consumption, Shanghai, ELC → Y
real GDP
Electricity consumption; Turkey, ELC → Y
real GDP
Electricity consumption; Taiwan, ELC → Y
real GDP per capita
Electricity consumption per Manufacturing Australia, Y → ELC
capita; real GDP per capita employment index
Electricity consumption; Korea, ELC↔Y
real GDP
771
(continued on next page)
 772
Table 1 (continued)
Author(s) Countries Methodology Main variables Other variables Conclusion(s)

Wolde-Rufael (2006) Algeria (1971–2001A) Toda-Yamamoto causality Electricity consumption per Algeria, ELC ≠ Y
Benin (1971–2001A) capita; real GDP per capita Benin, ELC → Y (+)
Cameroon (1971–2001A) Cameroon, Y → ELC (+)
Congo, DR (1971–2001A) Congo DR, ELC → Y (+)
Congo, Rep (1971–2001A) Congo, Rep, ELC ≠ Y
Egypt (1971–2001A) Egypt, ELC↔Y (+)
Gabon (1971–2001A) Gabon, Y → ELC (+), ELC → Y (–)
Ghana (1971–2001A) Ghana, Y → ELC (+)
Kenya (1971–2001A) Kenya, ELC ≠ Y
Morocco (1971–2001A) Morocco, ELC↔Y (+)
Nigeria (1971–2001A) Nigeria, Y → ELC (+)
Senegal (1971–2001A) Senegal, Y → ELC (+)
South Africa (1971–2001A) South Africa, ELC ≠ Y
Sudan (1971–2001A) Sudan, ELC ≠ Y
Tunisia (1971–2001A) Tunisia, ELC → Y (−)
Zambia (1971–2001A) Zambia, Y → ELC (+)
Zimbabwe (1971–2001A) Zimbabwe, Y → ELC (+)

N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781

Yoo (2006) Indonesia (1971–2002A) Johansen-Juselius; no Electricity consumption per Indonesia, Y → ELC
Malaysia (1971–2002A) cointegration; VAR capita; real GDP per capita Malaysia, ELC↔Y
Singapore (1971–2002A) Singapore, ELC↔Y
Thailand (1971–2002A) Thailand, Y → ELC
Yoo and Kim (2006) Indonesia (1971–2002A) Engle-Granger; no Electricity production; Indonesia, Y → ELP, Y → ELC
cointegration; VAR electricity consumption;
real GDP
Chen et al. (2007) China (1971–2001A) Johansen-Juselius; Pedroni Electricity consumption; China, ELC ≠ Y
Hong Kong (1971–2001A) panel cointegration; real GDP Hong Kong, ELC↔Y
Indonesia (1971–2001A) cointegration; VEC Indonesia, ELC → Y
India (1971–2001A) India, Y → ELC
Korea (1971–2001A) Korea, Y → ELC
Malaysia (1971–2001A) Malaysia, Y → ELC
Philippines (1971–2001A) Philippines, Y → ELC
Singapore (1971–2001A) Singapore, Y → ELC
Taiwan (1971–2001A) Taiwan, ELC ≠ Y
Thailand (1971–2001A) Thailand, ELC ≠ Y
Ten country panel, ELC↔Y
Ho and Siu (2007) Hong Kong (1966–2002A) Johansen-Juselius; Electricity consumption; Hong Kong, ELC↔Y
cointegration; VEC real GDP
Mozumder and Marathe Bangladesh (1971–1999A) Johansen-Juselius; Electricity consumption per Bangladesh, Y → ELC
(2007) cointegration; VEC capita; real GDP per capita
Narayan and Singh (2007) Fiji Islands (1971–2002A) ARDL bounds test; Electricity consumption; Labor force Fiji Islands, ELC → Y
cointegration; VEC real GDP
Soytas and Sari (2007) Turkey (1968–2002A) Johansen-Juselius Industry electricity Manufacturing Turkey, IELC → MVA
cointegration; VEC consumption, value employment;
added-manufacturing manufacturing real
fixed investment
Squalli (2007) Algeria (1980–2003A) ARDL bounds test; Electricity consumption Algeria, ARDL Y → ELC, TY Y → ELC(−)
Indonesia (1980–2003A) cointegration; per capita; real GDP Indonesia, ARDL Y → ELC, TY ELC → Y(+)
Iran (1980–2003A) Toda-Yamamoto per capita Iran, ARDL ELC↔Y, TY ELC↔Y (+)
Iraq (1980–2003A) causality Iraq, ARDL Y → ELC, TY Y → ELC(−)
Kuwait (1980–2003A) Kuwait, ARDL ELC → Y, TY Y → ELC(+)
Libya (1980–2003A) Libya, ARDL Y → ELC TY Y → ELC(−)
Nigeria (1980–2003A) Nigeria, ARDL ELC↔Y, TY ELC → Y(+)
Qatar (1980–2003A) Qatar, ARDL ELC↔Y, TY ELC↔Y(+)
Saudi Arabia (1980–2003A) Saudi Arabia, ARDL ELC↔Y, TY
ELC → Y(−) and Y → ELC(+)
UAE (1980–2003A) UAE, ARDL ELC↔Y, TY ELC → Y(−)
Venezuela (1980–2003A) Venezuela, ARDL ELC → Y TY ELC → Y(+)
Table 1 (continued)
Author(s) Countries
Yuan et al. (2007) China (1978–2004A)
Zamani (2007) Iran (1967–2003A)
Abosedra et al. (2008) Lebanon (1995:1–2005:12 M)
Hu and Lin (2008) Taiwan (1982:1–2006:4Q)
Narayan and Prasad (2008) Australia (1960–2002A)
Austria (1960–2002A)
Belgium (1960–2002A)
Canada (1960–2002A)
Czech Republic (1960–2002A)
Denmark (1960–2002A)
Finland (1960–2002A)
France (1960–2002A)
Germany (1960–2002A)
Greece (1960–2002A)
Hungary (1965–2002A)
Iceland (1960–2002A)
Ireland (1960–2002A)
Italy (1960–2002A)
Japan (1960–2002A)
Korea (1971–2002A)
Luxembourg (1960–2006A)
Mexico (1971–2002A)
Netherlands (1960–2002A)
New Zealand (1960–2002A)
Norway (1960–2002A)
Poland (1960–2002A)
Portugal (1960–2002A)
Slovak Republic (1971–2002A)
Spain (1960–2002A)
Sweden (1960–2002A)
Switzerland (1960–2002A)
Turkey (1960–2002A)
UK (1960–2002A)
US (1970–2002A)
Tang (2008) Malaysia (1972:1–2003:4Q)
Yuan et al. (2008) China (1963–2005A)
Akinlo (2009) Nigeria (1980–2006A)
Ghosh (2009) India (1970–2006A)
Narayan and Smyth (2009) Iran (1974–2002A)
Israel (1974–2002A)
Kuwait (1974–2002A)
Oman (1974–2002A)
Saudi Arabia (1974–2002A)
Syria (1974–2002A)
Odhiambo (2009a) Tanzania (1971–2006A)
Odhiambo (2009b) South Africa (1971–2006A)
Methodology Main variables
Johansen-Juselius; Electricity consumption;
cointegration; VEC real GDP
Engle-Granger; Industrial electricity and
cointegration; VEC agricultural electricity
consumption; industrial
valued added; agricultural
value added
Granger-causality; Electricity consumption
VAR growth; real import growth
Hansen-Seo threshold Electricity consumption;
cointegration; VEC real GDP
Bootstrapped Electricity consumption;
Granger-causality real GDP
ARDL bounds test; no cointegration; Electricity consumption per
Toda-Yamamoto causality capita; real GNP per capita
Johansen-Juselius cointegration; VEC Electricity consumption;
real GDP
Johansen-Juselius; cointegration; VEC Electricity consumption;
real GDP
ARDL bounds test; cointegration; VEC Electricity supply; real GDP
Westerlund panel cointegration; Electricity consumption
cointegration; VEC per capita;
real GDP per capita
ARDL bounds test; Electricity consumption
cointegration; VEC per capita, real GDP
per capita
Johansen-Juselius; Electricity consumption
cointegration; VEC per capita; real GDP
per capita
Other variables Conclusion(s)
China, ELC → Y
Iran, IVA → IELC, AVA↔AELC
Temperature; relative Lebanon, ELC → IMP
humidity
Taiwan, Y → ELC
Australia, ELC → Y
Austria, ELC ≠ Y
Belgium, ELC ≠ Y
Canada, ELC ≠ Y
Czech Republic, ELC → Y
Denmark, ELC ≠ Y
Finland, Y → ELC
France, ELC ≠ Y
Germany, ELC ≠ Y
Greece, ELC ≠ Y
N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781
Hungary, Y → ELC
Iceland, ELC↔Y
Ireland, ELC ≠ Y
Italy, ELC → Y
Japan, ELC ≠ Y
Korea, ELC↔Y
Luxembourg, ELC ≠ Y
Mexico, ELC ≠ Y
Netherlands, Y → ELC
New Zealand, ELC ≠ Y
Norway, ELC ≠ Y
Poland, ELC ≠ Y
Portugal, ELC → Y
Slovak Republic, ELC → Y
Spain, ELC ≠ Y
Sweden, ELC ≠ Y
Switzerland, ELC ≠ Y
Turkey, ELC ≠ Y
UK, ELC↔Y
US, ELC ≠ Y
Malaysia, ELC → Y
Capital; employment China, ELC↔Y
Nigeria, ELC → Y
Employment India, Y → ELS
Real exports per capita MENA panel, ELC↔Y
Tanzania, ELC → Y
Employment South Africa, ELC↔Y
773
(continued on next page)
 774
Table 1 (continued)
Author(s) Countries Methodology Main variables Other variables Conclusion(s)

Ziramba (2009)
Acaravci and Ozturk (2010)
South Africa (1980–2005A) ARDL bounds test; Electricity consumption; Employment South Africa, ELC ≠ IP
cointegration; VEC; industrial production
Toda-Yamamoto causality
Albania (1990–2006A) Pedroni panel cointegration; Electricity consumption Panel, ELC ≠ Y
Belarus (1990–2006A) no cointegration per capita; real GDP
Bulgaria (1990–2006A) per capita
Czech Republic (1990–2006A)
Estonia (1990–2006A)
Latvia (1990–2006A)
Lithuania (1990–2006A)
Macedonia (1990–2006A)
Moldova (1990–2006A)
Poland (1990–2006A)
Romania (1990–2006)
Russian Federation (1990–2006A)
Serbia (1990–2006A)

N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781

Chandran et al. (2010)
Ciarreta and Zarraga (2010)
Lean and Smyth (2010)
Ozturk and Acaravci (2010)
Yoo and Kwak (2010)
Slovak Republic (1990–2006A)
Ukraine (1990–2006A)
Malaysia (1971–2003A) ARDL bounds testing; Electricity consumption; Consumer prices Malaysia, ELC → Y
cointegration; VEC real GDP
Austria (1970–2007A) Pedroni panel cointegration; Electricity consumption; Energy prices Panel, ELC(−) → Y
Belgium (1970–2007A) cointegration; VEC real GDP
Denmark (1970–2007A)
Finland (1970–2007A)
France (1970–2007A)
Germany (1970–2007A)
Italy (1970–2007A)
Luxembourg (1970–2007A)
Netherlands (1970–2007A)
Norway (1970–2007A)
Sweden (1970–2007A)
Switzerland (1970–2007A)
Indonesia (1980–2006A) Johansen-Fisher panel Electricity consumption Carbon dioxide emissions Panel, ELC → Y
cointegration; cointegration; per capita; real GDP per capita
Malaysia (1980–2006A) VEC per capita
Philippines (1980–2006A)
Singapore (1980–2006A)
Thailand (1980–2006A)
Albania (1980–2006A) ARDL bounds testing; Electricity consumption Albania, ELC ≠ Y
Bulgaria (1980–2006A) No cointegration per capita; real GDP Bulgaria, ELC ≠ Y
Hungary (1980–2006A) per capita Hungary, ELC ≠ Y
Romania (1980–2006A) Romania, ELC ≠ Y
Argentina (1975–2006A) Johansen-Juselius Electricity consumption Argentina, ELC → Y
Brazil (1975–2006A) cointegration; VEC per capita; real GDP Brazil, ELC → Y
Chile (1975–2006A) per capita Chile, ELC → Y
Colombia (1975–2006A) Colombia, ELC → Y
Ecuador (1975–2006A) Ecuador, ELC → Y
Peru (1975–2006A) Peru, ELC ≠ Y
Venezuela (1975–2006A) Venezuela, ELC↔Y

Notes: definitions of notation: →, ↔, and ≠ represent unidirectional causality, bidirectional causality, and no causality, respectively.
Abbreviations defined as follows: ELC = electricity consumption; ELP = electricity production; ELS = electricity supply; AELC = agricultural electricity consumption; IELC = industrial electricity consumption; CEU = commercial electricity
usage; IEU = industrial electricity usage; OEU = other sector electricity usage; REU = residential electricity usage; AVA = agricultural value added; IVA = industrial value added; MVA = manufacturing value added; IMP = imports; Y = real or
nominal GDP or GNP; IP = industrial production; EMP = employment; NEMP = non-farm employment; MEMP = manufacturing employment. Alternative methodologies other than standard Granger-causality tests: Engle-Granger (1987);
Johansen-Juselius (JJ) (1988, 1990); Hansen-Seo (2002) threshold cointegration; Pedroni (1999, 2004) panel cointegration; Westerlund (2006) panel cointegration; Pesaran and Shin (1999) and Pesaran et al. (2001) ARDL bounds test
(ARDL); Toda-Yamamoto (1995) (TY), and Dolado and Lütkepohl (1996). Abbrevations for models: VAR = vector autoregressive model and VEC = vector error correction model.
 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781 775

absence of causality between electricity consumption and economic Table 2

growth. In this case, the reduction in electricity consumption through Panel unit root tests.

energy conservation policies will not impact economic growth. Variables IPS LLC Fisher-ADF Fisher-PP
It is not surprising that the published studies on the causal
High income country panel
relationship between electricity consumption and economic growth Y − 0.46 − 0.78 23.46 21.37
have yielded mixed results to date, as shown in Table 1.1 The lack of ΔY − 5.99a − 6.83a 84.51a 105.48a
consensus in the empirical results may be attributed to differences in ELC − 0.54 − 0.53 20.39 26.32
ΔELC − 5.05a − 5.63a 69.43a 86.41a
variable selection, model specifications, time horizon, and econometric
K − 0.88 − 0.77 19.44 27.81
approaches undertaken. This study extends the existing literature on the ΔK − 6.01a − 7.13a 78.69a 103.28a
causal relationship between electricity consumption and economic L − 0.62 − 0.67 21.47 32.16
growth along several dimensions. First, the study will include a larger set ΔL − 6.79a − 6.62a 73.68a 121.06a
of countries in the analysis than previous studies; a sample of 88
Upper-middle income country panel
countries is broken down into four panels based on the World Bank
Y − 0.70 − 0.72 24.33 24.33
income classification in order to examine the causal dynamics between ΔY − 6.91a − 6.16a 88.93a 112.37a
electricity consumption and economic development. Second, roughly ELC − 0.59 − 0.64 22.17 22.18
70% of the published studies on the causal relationship between ΔELC − 5.01a − 5.37a 65.42a 76.42a
K − 0.91 − 0.78 15.68 29.54
electricity consumption and economic growth have been conducted
ΔK − 5.15a − 6.75a 74.52a 109.55a
within a bivariate framework. However, a problem inherent with a L − 0.64 − 0.73 24.58 37.64
bivariate analysis is omitted variable bias (Lütkepohl, 1982). Recogniz- ΔL − 6.02a − 6.34a 70.05a 124.58a
ing omitted variable bias, this study investigates the causal relationship
between electricity consumption and economic growth within a Lower-middle income country panel
Y − 0.74 − 0.83 25.41 26.53
multivariate framework by including measures of capital and labor.
ΔY − 6.01a − 6.59a 95.46a 119.87a
Third, with the exception of the studies by Wolde-Rufael (2006), Squalli ELC − 0.85 − 0.83 25.63 27.88
(2007), and Ciarreta and Zarraga (2010), the sign and magnitude of the ΔELC − 5.02a − 5.58a 69.83a 83.21a
respective coefficients will be discussed in relation to the various K − 0.79 − 0.92 18.77 33.47
ΔK − 6.06a − 7.52a 79.80a 117.26a
hypotheses related to the causal relationship between electricity
L − 0.89 − 0.83 25.41 39.25
consumption and economic growth. Fourth, like the studies by Chen ΔL − 5.26a − 6.61a 78.94a 129.06a
et al. (2007), Narayan and Smyth (2009), Acaravci and Ozturk (2010),
Ciarreta and Zarraga (2010), and Lean and Smyth (2010), panel Low income country panel
cointegration techniques are used that recognize the heterogeneity in Y − 0.72 − 0.63 24.94 29.80
ΔY − 6.95a − 5.42a 78.54a 95.22a
the estimated parameters and dynamics across countries. The panel unit
ELC − 0.64 − 0.49 19.04 27.41
root and cointegration tests provide additional power by combining the ΔELC − 5.03a − 4.83a 60.22a 68.73a
cross-section and time series data while allowing for heterogeneity K − 0.99 − 0.54 17.61 37.66
across countries (Harris and Tzavalis, 1999). Unlike previous panel ΔK − 6.05a − 5.21a 71.26a 89.54a
L − 0.62 − 0.53 27.66 42.31
cointegration studies on the electricity consumption-growth nexus, this
ΔL − 5.92a − 5.45a 65.72a 109.37a
study is the first to use the Larsson et al. (2001) panel cointegration
procedure which offers several advantages over residual-based panel Notes: Critical values at the 1% level denoted by “a”: IPS − 4.98, LLC − 0.84, Fisher-ADF
56.09, Fisher-PP 61.15.
cointegration procedures.
Section 2 discusses the data, methodology, and empirical results.
Section 3 provides concluding remarks. The investigation will be undertaken within the context of an
aggregate production function model where electricity consumption,
capital, and labor are treated as separate inputs denoted as follows4:
2. Data, methodology, and results
Yit = f ðELCit ; Kit ; Lit Þ ð1Þ
Annual data from 1990 to 2006 were obtained from the World
Bank Development Indicators, CD-ROM.2 The Appendix lists the 88
The analysis begins with the estimation of several panel unit root and
countries by the World Bank income classification. The multivariate
stationarity tests to infer the time series properties of the variables
framework includes real GDP (Y) in billions of constant 2000 U.S.
before proceeding to test for panel cointegration. Levin et al. (2002)
dollars, real gross fixed capital formation (K) in billions of constant
propose a panel based ADF test that assumes homogeneity in the
2000 U.S. dollars, total labor force (L) in millions, and electric power
dynamics of the autoregressive coefficients for all panel units. On the
consumption (ELC) defined in kilowatt hours.3
other hand, the Im et al. (2003) panel unit root test allows for
heterogeneity in the dynamics of the autoregressive coefficients for all
panel units. Following Maddala and Wu (1999), the nonparametric
1
Note that no distinction is made between short-run versus long-run causality or panel unit root tests are estimated using the Fisher-ADF and Fisher-PP
strong-form causality (both the short-run changes in the lagged variables and error tests. Under the Levin et al. (2002), Im et al. (2003), Fisher-ADF and
correction term are jointly significant) in regards to the results from error correction
Fisher-PP tests the null hypothesis is a unit root while the alternative
models in order to conserve space. Table 1 is an amended version of my survey article on
the electricity consumption–growth nexus, Payne (2010a,b) to also include more recent hypothesis is no unit root. The results of the panel unit root tests
studies. displayed in Table 2 reveal that each variable is integrated of order one
2
The selection of the sample is due to the availability of data on all the variables for for each panel.5
the longest time horizon in order to maintain consistency and comparability of the
data over the time, the decision was made to use the World Bank Development
4
Indicators as done in previous multiple country studies within the literature on the Within the energy consumption–growth literature some studies have essentially
causal relationship between energy consumption and economic growth. estimated energy demand functions with the inclusion of energy prices when
3
Real gross fixed capital formation serves as a proxy for capital in that changes in modeling the causal relationship. However, not uncommon in the literature, this study
investment closely align with changes in the capital stock under the assumption of a estimates the causal dynamics using a production model framework with all real
constant depreciation rate using the perpetual inventory method (Apergis and Payne, variables.
5
2009a,b; 2010 and citations therein). To conserve space, details with respect to the panel unit root tests are not provided.
776 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781

Table 3 Table 4
Panel cointegration tests. Long-run parameter estimates.

Panel A: high income country panel Panel A: high income country panel
Y = 0.655 + 0.083ELC + 0.521K + 0.284L
Cointegrating rank 5% Critical values
(3.71)a (8.71)a (30.71)a (4.52)a
Null hypothesis LR test (Asymptotic)
H0 : r = 0 161.06a 116.93
Adj. R2 = 0.65 LM = 1.13 HE = 1.68
H0 : r = 1 89.85 109.08
[0.32] [0.24]

Homogeneous cointegration vectors 5% Critical values

Panel B: upper-middle income country panel
Null hypothesis LR test (Asymptotic)
Y = 0.582 + 0.223ELC + 0.361K + 0.174L
H0 : b1 = b2 =... = bN 69.84a 34.74
(3.26)a (11.95)a (33.48)a (4.86)a
H0 : B block diagonal 47.41 67.29
Adj. R2 = 0.66 LM = 1.05 HE = 1.94
Panel B: upper-middle income country panel
[0.34] [0.21]
Cointegrating rank 5% Critical values
Null hypothesis LR test (Asymptotic) Panel C: lower-middle income country panel
H0 : r = 0 143.77a 116.93 Y = 0.431 + 0.306ELC + 0.219K + 0.379L
H0 : r = 1 76.28 109.08 (3.55)a (14.86)a (30.28)a (8.57)a

Homogeneous cointegration vectors 5% Critical values Adj. R2 = 0.62 LM = 0.85 HE = 1.36

Null hypothesis LR test (Asymptotic) [0.56] [0.39]
H0 : b1 = b2 =... = bN 56.14a 34.74
Notes: t-statistics and probability values are reported in parentheses and brackets,
H0 : B block diagonal 43.28 67.29
respectively. LM is the Lagrange multiplier test for serial correlation. HE is White's
heteroscedasticity test. Significance at the 1% level is denoted by “a”.
Panel C: lower-middle income country panel

Cointegrating rank 5% Critical values

Null hypothesis LR test (Asymptotic)
H0 : r = 0 131.26a 116.93 generating process can be characterized by the following heteroge-
H0 : r = 1 64.23 109.08 neous VAR(ki) model:
Homogeneous cointegration vectors 5% Critical values k
Yit = ∑ki = 1 Πik Yi;t−k + εit ; i = 1; ::; N ð2Þ
Null hypothesis LR test (Asymptotic)
H0 : b1 = b2 =... = bN 48.47a 34.74
H0 : B block diagonal 33.59 67.29 where for each group i the values Yi,−ki + 1,…,Yi,0 are fixed and the
errors εit are independently identically distributed Np(0,Ωi). The
Panel D: low income country panel heterogeneous error correction model is given as follows:
Cointegrating rank 5% Critical values
k −1
Null hypothesis LR test (Asymptotic) ΔYit = Πi Yi;t−1 + ∑ki = 1 Γik ΔYi;t−k + εit ; i = 1; ::; N ð3Þ
H0 : r = 0 94.25 116.93
H0 : r = 1 38.59 109.08 where ∏i is of the order p × p. If ∏i is of reduced rank, it is possible for
∏i = αiβi′ where αi and βi are p x ri and of full column rank. The reduced-
Homogeneous cointegration vectors 5% Critical values
Null hypothesis LR test (Asymptotic)
rank estimation procedure allows for the estimation of ∏i and
H0 : b1 = b2 =... = bN 25.82 34.74 hypothesis testing on the cointegrating rank as well as for the long-
H0 : B block diagonal 27.49 67.29 run coefficients, βi, and the adjustment parameters, αi.
Notes: Larsson et al. (2001) cointegration procedure with significance at the 5% level is Whether the Π matrix is of reduced rank and all N groups can each be
denoted by “a”. characterized by r cointegrating vectors must be determined. If
cointegration is present, one needs to investigate whether or not the
cointegrating vectors are homogeneous. Larsson et al. (2001) procedure
for cointegration is under the assumption that Πi = αiβ'i = αβ' = Π for
i = 1,…, N with the null hypothesis H0 : rk(Πi) = k for i = 1,…, N. The test
The Larsson et al. (2001) panel cointegration procedure, which
is consistent against the alternative hypothesis that H1 : rk(Πi) = m for a
utilizes a likelihood-based framework for the testing and estimation of
non-vanishing fraction of cross-section members. The test statistic is
cointegrated panels, is employed.6 The Larsson et al. (2001) procedure
similar to Im et al. (2003) in which the test statistic is given by a suitably
has several advantages over residual-based cointegration tests. First,
centered and scaled version of the cross-sectional average of the
this procedure allows for more than one cointegrating vector whereas
individual trace statistics. Using the central limit theorem in the cross-
other methods, such as Pedroni (1999, 2004), assume there is only one
sectional dimension and the appropriate mean and variance correction
cointegrating vector. Second, with the Larsson et al. (2001) procedure,
factors suggests that under the null hypothesis:
no choice has to be made regarding the normalization of variables. Third,
though the cointegrating relations are restricted to each cross-section, qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
the rest of the model is unrestricted which allows for a substantial s
N2 ∑Ni= 1 ðLRsi ðkjmÞ−E LRsi ðk jmÞ
LLL ðkjmÞ = qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  →Nð0; 1Þ ð4Þ
amount of short-run dependence between the groups, in particular, Var LRsi ðkjmÞ
cross-sectional dependence of the error terms.
The Larsson et al. (2001) panel cointegration procedure is outlined in the sequential limit T → ∞ followed by N → ∞. LRsi (k|m) denotes the
as follows. The panel data set consists of N cross-section observed over trace statistic for the null hypothesis of a k-dimensional cointegrating
T time periods. Let i = 1,…,N represent the countries, t = 1,…,T the space for unit i where the superscript s indicates the deterministic
sample time period, and j = 1,…,p the variables in each group. Thus, components. E(LRsi (k|m)) and Var(LRsi (k|m)) represent the mean and
yijt represents the ith group and the jth variable at time t. The data variance of the asymptotic trace statistics. These two moments are
obtained from a stochastic simulation following the procedure
described by Johansen (1995). For T → ∞ the expressions E(LRsi (k|m))
6
The Larsson et al. (2001) panel cointegration procedure is similar to the Johansen and Var(LRsi (k|m)) converge to the limit of the expected value and
(1988) and Johansen-Juselius (1990) methodology. variance of the trace statistic, respectively, corresponding to the
 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781 777

Table 5
Panel causality tests high income country panel.

Panel causality tests

Dependent variable Sources of causation (independent variables)

Short-run Long-run

ΔY ΔELC ΔK ΔL ECT

(5a) ΔY – 37.83 (0.039) 49.88 (0.647) 51.06 (0.324) − 0.234

[0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a
(5b) ΔELC 55.69 (0.244) – 68.14 (0.505) 57.34 (0.164) − 0.387
[0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a
(5c) ΔK 61.25 (0.098) 0.17 (0.081) – 39.26 (0.041) − 0.399
[0.00]a [0.00]a [0.89] [0.76] [0.00]a [0.00]a [0.00]a
(5d) ΔL 58.62 (0.227) 0.26 (− 0.059) 49.06 (0.282) – − 0.327
[0.00]a [0.00]a [0.30] [0.24] [0.00]a [0.00]a [0.00]a

Notes: Partial F-statistics reported with respect to short-run changes in the independent variables. The lag length is two. The sum of the lagged coefficients for the respective short-
run changes is denoted in parentheses. ECT represents the coefficient of the error correction term. Probability values are in brackets and reported underneath the corresponding
partial F-statistic and sum of the lagged coefficients, respectively. Significance at the 1% level denoted by “a”.

deterministic components. Also, finite T correction factors (for different
lag lengths and the numbers of variables) via simulation are undertaken
in this study (see Im et al., 2003).
For each country in the panel the null hypothesis, r = 0, is tested
using the observed trace statistic and, if the null hypothesis is rejected,
then the null hypothesis, r = 1, is examined. This sequential testing
procedure is completed when the null hypothesis, r = ri, is not rejected
which identifies the rank estimate of r. The panel trace test, the statistic
LRsi (k|m) as noted in Eq. (4), is computed by standardizing the average of
the N countries' individual observed trace statistics. If cointegration
is determined, the procedure permits one to test whether the
cointegrating vector is homogeneous across countries. In addition, the
Larsson et al. (2001) procedure allows for a robust test of cointegration
that can be performed with cross-sectional dependence in the error
terms of the panel without arbitrary normalization assumptions.7
As shown in Panels A–D of Table 3, the panel rank (LR) test results
reject the null of no cointegration for the panels of high, upper-
middle, and lower-middle income country panels. However, the panel
rank test fails to reject the null hypothesis of no cointegration for the
low income country panel. This result confirms to some extent the
observation by Ferguson et al. (2000, p. 924) “that wealthy countries
have a stronger correlation between electricity use and wealth
creation than do poor countries”. In those cases in which the null of
hypothesis of no cointegration is rejected in favor of panel cointegra-
tion with one cointegrating vector, the null hypothesis of homoge-
nous cointegrating vectors is examined. The results indicate that the
null of homogenous cointegrating vectors is rejected for the high,
upper-middle, and lower-middle income country panels. Thus, the
Larsson et al. (2001) panel test for cointegration indicates a common
rank, r = 1, between real GDP, electricity consumption, real gross fixed
capital formation and the labor force for the high, upper-middle, and
lower-middle income country panels.8
Panels A–C of Table 4 report the long-run parameter estimates for
the high, upper-middle, and lower-middle income country panels
associated with the cointegrating vector based on the Larsson et al.
(2001) procedure. The coefficients are positive and statistically
significant at the 1% level for electricity consumption, real gross fixed
capital formation, and the labor force, respectively. Given that the
coefficients can be interpreted as elasticity estimates, the results in Panel
A for the high income country panel indicate that a 1% increase in
electricity consumption increases real GDP by 0.083%; a 1% increase in
real gross fixed capital formation increases real GDP by 0.521%; and a 1%
7
The residual based cointegration test of Pedroni (1999, 2004) is sensitive to both
the use of time dummies to account for potential cross-sectional dependence and
variable normalization.
8
The authors thank Dr. Chris Tsoumas for providing the software routine to implement
the Larsson et al. (2001) panel cointegration test.
increase in the labor force increases real GDP by 0.284%. In the case of the
upper-middle income country panel, the results in Panel B show that a
1% increase in electricity consumption increases real GDP by 0.223%;
a 1% increase in real gross fixed capital formation increases real GDP by
0.361%; and a 1% increase in the labor force increases real GDP by 0.174%.
The long-run parameter estimates for the lower-middle income country
panel in Panel C reveal that a 1% increase in electricity consumption
increases real GDP by 0.306%; a 1% increase in real gross fixed capital
formation increases real GDP by 0.219%; and a 1% increase in the labor
force increases real GDP by 0.379%. One observes that the elasticity
estimate with respect to electricity consumption becomes smaller as the
income level associated with the panel increases. This finding may
suggest that with improvements in electricity efficiency and intensity
observed in more developed, advanced economies over time, it is not
necessarily the increase in electricity consumption that contributes to
economic growth, but the efficient use of electricity.
In light of the unique cointegrating vectors for the high, upper-
middle, and lower-middle income country panels, respectively, panel
vector error correction models are estimated to infer the causal
relationship among the variables. Defining the lagged residuals from
the long-run cointegration equations stated in Panels A–C of Table 4,
the following dynamic error correction models are estimated for the
respective panels.
q q q
ΔYit = ξ1j + ∑k = 1
ψ11ik ΔYit−k + ∑k = 1
ψ12ik ΔELCit−k + ∑k = 1 ψ13ik ΔKit−k
+ ∑qk = 1 ψ14ik ΔLit−k + λ1i εit−1 + u1it
ð5aÞ
q q q
ΔELCit = ξ2j + ∑k = 1 ψ21ik ΔYit−k + ∑k = 1 ψ22ik ΔELCit−k + ∑k = 1 ψ23ik ΔKit−k
q
+ ∑k = 1 ψ24ik ΔLit−k + λ2i εit−1 + u2it
ð5bÞ
q q q
ΔKit = ξ3j + ∑k = 1 ψ31ik ΔYit−k + ∑k = 1 ψ32ik ΔELCit−k + ∑k = 1 ψ33ik ΔKit−k
+ ∑qk = 1 ψ34ik ΔLit−k + λ3i εit−1 + u3it
ð5cÞ
q q q
ΔLit = ξ4j + ∑k = 1 ψ41ik ΔYit−k + ∑k = 1 ψ42ik ΔELCit−k + ∑k = 1 ψ43ik ΔKit−k
+ ∑qk = 1 ψ44ik ΔLit−k + λ4i εit−1 + u4it
ð5dÞ
where Δ is the first-difference operator; k is the lag length based on
likelihood ratio tests; and u is the serially uncorrelated error term.
With respect to Eqs. (5a)–(5d), short-run causality is determined by
the statistical significance of the partial F-statistic associated with the
corresponding right hand side variables. Long-run causality is
revealed by the statistical significance of the respective error
778 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781

Table 6
Panel causality tests upper-middle income country panel.

Panel causality tests

Dependent variable Sources of causation (independent variables)

Short-run Long-run

ΔY ΔELC ΔK ΔL ECT

(5a) ΔY – 40.53 (0.087) 46.72 (0.344) 53.39 (0.274) − 0.124

[0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a
(5b) ΔELC 27.13 (0.135) – 61.25 (0.-69) 53.27 (0.058) − 0.066
[0.01]a [0.01]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a
(5c) ΔK 51.38 (0.146) 1.19 (0.019) – 3.35 (0.153) − 0.155
[0.00]a [0.00]a [0.80] [0.94] [0.22] [0.73] [0.00]a
(5d) ΔL 53.47 (0.114) 0.29 (0.017) 0.76 (0.108) – − 0.077
[0.00]a [0.00]a [0.32] [0.93] [0.79] [0.59] [0.00]a

Notes: Partial F-statistics reported with respect to short-run changes in the independent variables. The lag length is one. The sum of the lagged coefficients for the respective short-
run changes is denoted in parentheses. ECT represents the coefficient of the error correction term. Probability values are in brackets and reported underneath the corresponding
partial F-statistic and sum of the lagged coefficients, respectively. Significance at the 1% level denoted by “a”.

Table 7
Panel causality tests lower-middle income country panel.

Panel causality tests

Dependent variable Sources of causation (independent variables)

Short-run Long-run

ΔY ΔELC ΔK ΔL ECT

(5a) ΔY – 44.58 (0.188) 51.25 (0.145) 57.31 (0.065) − 0.254

[0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a
(5b) ΔELC 0.24 (− 0.287) − 0.41 (0.126) 0.56 (0.252) − 0.125
[0.82] [0.56] [0.78] [0.62] [0.54] [0.48] [0.00]a
(5c) ΔK 0.84 (− 0.026) 1.05 (− 0.149) – 1.03 (0.243) − 0.011
[0.95] [0.69] [0.31] [0.53] [0.59] [0.46] [0.41]
(5d) ΔL 0.21 (− 0.186) 0.31 (− 0.045) 0.82 (0.039) – − 0.163
[0.56] [0.67] [0.42] [0.84] [0.63] [0.86] [0.12]

Notes: Partial F-statistics reported with respect to short-run changes in the independent variables. The lag length is one. The sum of the lagged coefficients for the respective short-
run changes is denoted in parentheses. ECT represents the coefficient of the error correction term. Probability values are in brackets and reported underneath the corresponding
partial F-statistic and sum of the lagged coefficients, respectively. Significance at the 1% level denoted by “a”.

correction terms using a t-test. In the case of the low income country
panel, the error correction terms are omitted given the absence of
cointegration as reported in Panel D of Table 3.
For the high income country panel in Table 5, Eq. (5a) reveals that
electricity consumption, real gross fixed capital formation, and the
labor force each have a positive and statistically significant impact on
economic growth in the short-run. With respect to Eq. (5b) economic
growth, real gross fixed capital formation, and the labor force each
have a positive and statistically significant impact on electricity
consumption. For Eq. (5c) economic growth and the labor force each
have a positive and statistically significant impact on real gross fixed
capital formation while electricity consumption has a statistically
insignificant impact. Eq. (5d) shows that economic growth and real
gross fixed capital formation each have a positive and statistically
significant impact on the labor force while electricity consumption
has a statistically insignificant impact. In terms of the long-run
dynamics, based on the statistical significance of the error correction
terms from Eqs. (5a)–(5d), economic growth, electricity consump-
tion, real gross fixed capital formation, and the labor force each
respond to deviations from long-run equilibrium. Furthermore, the
speed of adjustment towards long-run equilibrium is rather fast given
the magnitude of the coefficients of the error correction terms,
ranging from 2.51 years to 4.27 years.9 The results for the high income
9
The speed of adjustment is computed as the reciprocal of the absolute value of the
coefficient on the respective error correction terms.
country panel indicate the presence of bidirectional causality between
electricity consumption and economic growth in both the short-run
and long-run.
Table 6 reports the panel causality results for the upper-middle
income country panel. Eq. (5a) reveals that electricity consumption,
real gross fixed capital formation, and the labor force each have a
positive and statistically significant impact on economic growth in the
short run. In Eq. (5b) economic growth, real gross fixed capital
formation, and the labor force each have a positive and statistically
significant impact on electricity consumption. As for Eq. (5c),
economic growth has a positive and statistically significant impact
on real gross fixed capital formation whereas electricity consumption
and the labor force each have a statistically insignificant impact. With
respect to Eq. (5d) economic growth has a positive and statistically
significant impact on the labor force, but electricity consumption and
real gross fixed capital formation each have a statistically insignificant
impact. In regards to the long-run dynamics, economic growth,
electricity consumption, real gross fixed capital formation, and the
labor force each respond to deviations from long-run equilibrium
given the statistical significance of the respective error correction
terms. The speed of adjustment towards long-run equilibrium is
rather slow, ranging from 6.45 years to 12.99 years. As in the case of
the high income country panel, the results for the upper-middle
income country panel indicate the presence of bidirectional causality
between electricity consumption and economic growth in both the
short-run and long-run.
 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781
Table 8
Panel causality tests lower income country panel.
Panel causality tests
Dependent variable Sources of causation (independent variables)
Short-run
ΔY
(5a) ΔY – 46.55 (0.043)
(5b) ΔELC 1.04 (0.082)
[0.38] [0.88]
(5c) ΔK 59.84 (0.663)
[0.00]a [0.00]a
(5d) ΔL 58.91 (0.316)
[0.00]a [0.00]a
Notes: Partial F-statistics reported with respect to short-run changes in the independent variables. The lag length is two. The sum of the lagged coefficients for the respective short-
run changes is denoted in parentheses. Probability values are in brackets and reported underneath the corresponding partial F-statistic and sum of the lagged coefficients,
respectively. Significance at the 1% and 5% level denoted by “a” and “b”, respectively.
Table 7 displays the panel causality results for the lower-middle
income country panel. Eq. (5a) shows that electricity consumption,
real gross fixed capital formation, and the labor force each have a
positive and statistically significant impact on economic growth in the
short-run. Unlike the results for the high and upper-middle income
country panels, Eq. (5b) reveals that economic growth, real gross fixed
capital formation, and the labor force each have a statistically
insignificant impact on electricity consumption in the short-run.
Likewise in Eq. (5c), economic growth, electricity consumption, and
the labor force each have a statistically insignificant impact on real
gross fixed capital formation. Again, as in the case of the electricity
consumption and real gross fixed capital formation equations, none of
the variables have a statistically significant impact on the labor force,
Eq. (5d), in the short-run. With respect to the long-run dynamics,
economic growth and electricity consumption each respond to
deviations from long-run equilibrium. However, the error correction
terms for real gross fixed capital formation and the labor force
equations are statistically insignificant. The speed of adjustment
towards long-run equilibrium for the economic growth and electricity
consumption equations are respectively, 3.94 and 8 years. Thus, while
there is unidirectional causality from electricity consumption to
economic growth in the short-run, the results suggest bidirectional
causality between electricity consumption and economic growth in
the long-run for the lower-middle income country panel.
Finally, unlike the results reported in Tables 5–7, the absence of
cointegration among the variables in the low income country panel
renders the use of a panel autoregressive framework to infer causal
relations. Table 8 reports the panel causality results for the low income
country panel. In Eq. (5a) electricity consumption, real gross fixed capital
formation, and the labor force each have a positive and statistically
significant impact on economic growth. As in the case of the lower-middle
income country panel, Eq. (5b) reveals that economic growth, real gross
fixed capital formation, and the labor force each have a statistically
insignificant impact on electricity consumption. From Eq. (5c) economic
growth has a positive and statistically significant impact on real gross
fixed capital formation while electricity consumption and the labor force
each have a statistically insignificant impact. The results for Eq. (5d)
indicate that economic growth and electricity consumption each have a
positive and statistically significant impact on the labor force while real
gross fixed capital formation has a statistically insignificant impact. In
contrast to the results from the other country panels, the results for the
low income country panel illustrate unidirectional causality from
electricity consumption to economic growth.
In the cases for the high, upper middle, and lower middle income
country panels, the presence of bidirectional causality provides
support for the feedback hypothesis in which electricity consumption
and economic growth are interdependent. The interdependence
779
ΔELC ΔK ΔL
48.24 (0.154) 63.41 (0.318)
[0.00]a [0.00]a [0.00]a [0.00]a [0.00]a [0.00]a
– 0.64 (0.063) 1.74 (0.019)
[0.72] [0.87] [0.12] [0.91]
0.83 (− 0.043) – 0.31 (0.092)
[0.32] [0.73] [0.52] [0.95]
40.49 (− 0.389) 0.41 (0.251) –
[0.00]a [0.02]b [0.75] [0.57]
between electricity consumption and economic growth suggests
that energy conservation policies which reduce electricity consump-
tion may impact economic growth. Likewise, such fluctuations in
economic growth may very well be transmitted back to electricity
consumption. In terms of the low income country panel, the finding of
unidirectional causality from electricity consumption to economic
growth lends support for the growth hypothesis. Therefore, it is
possible that energy conservation policies which reduce electricity
consumption may have an adverse impact on economic growth.
3. Concluding remarks
Electricity is a vital factor in the growth prospects of countries by
facilitating scientific and technological advancements alongside the
emergence of the information and communication technologies that
have enhanced production and trade across countries. This study
explores the causal relationship between electricity consumption and
economic growth associated with four country panels based on
income classification to discern if any differences based on the level of
economic development. Specifically, the study examines the causal
relationship between electricity consumption and economic growth
including measures of capital and labor within a multivariate panel
error correction model framework for 88 countries categorized into
four income country panels (high, upper-middle, lower-middle, and
low income) over the period 1990 to 2006.
First, the Larsson et al. (2001) panel cointegration tests indicate
that there is a unique long-run equilibrium relationship between real
GDP, electricity consumption, real gross fixed capital formation, and
the labor force in the case of high, upper-middle, and lower-middle
income country panels with the long-run elasticity estimates positive
and statistically significant. For these panels, the elasticity estimated
with respect to electricity consumption decreases in size as the
income level increases. This suggests that with improvements in
electricity efficiency observed in more developed, advanced econo-
mies contributes to economic growth more so than simply an increase
in electricity consumption. However, a long-run equilibrium relation-
ship could not be established for the low income country panel.
Second, the estimation of panel error correction models in the case
of high, upper-middle, and lower-middle income country panels
reveals the presence of bidirectional causality between electricity
consumption and economic growth in both the short-run and long-
run for the high and upper-middle income country panels. In the case
of the lower-middle income country panel, there is unidirectional
causality from electricity consumption to economic growth in the
short-run, but bidirectional causality in the long-run. Finally, using a
panel vector autoregressive model, the results indicate unidirectional
780 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781
causality from electricity consumption to economic growth for the
low income country panel.
The results for the high, upper-middle, and lower-middle income
country panels provide support for the feedback hypothesis. This
interdependence between electricity consumption and economic
growth suggests that energy conservation policies which reduce
electricity consumption may impact economic growth. Likewise, such
fluctuations in economic growth may very well be transmitted back to
electricity consumption. However, energy policies which improve the
efficiency in the production and consumption of electricity may not
have a detrimental impact on economic growth, but may also improve
environmental quality as such policies will curb excessive electricity
consumption and inefficient electricity production methods. The
results for the low income country panel indicate support for the
growth hypothesis. As such, the positive impact of electricity
consumption on economic growth suggests that electricity consump-
tion plays an important role in the growth process. Thus, for low
income countries, improvements in the availability and accessibility
of electricity beyond urban areas to more remote, rural areas would
improve the growth prospects of these countries.
This study is by no means the defining statement on the role of
electricity consumption in the economic development process, but
does highlight the importance of recognizing that the dynamic causal
relationship between electricity consumption and growth may
depend in part on a country's stage of development. Following the
work of Coondoo and Dinda (2002) and Dinda and Coondoo (2006),
among others, future research in this area should incorporate the role
of carbon dioxide emissions into the examination of the causal
relationship between electricity consumption and economic growth
to shed additional light on the environmental impact of electricity
consumption.
Appendix A
List of countries by panel
High income country panel (n= 29): Australia, Austria, Belgium,
Brunei Darassalam, Canada, Czech Republic, Denmark, Finland, France,
Germany, Greece, Hong Kong, Hungary, Iceland, Ireland, Italy, Japan,
Korea, Luxembourg, Netherlands, New Zealand, Norway, Portugal,
Slovenia, Spain, Sweden, Switzerland, United Kingdom, United States.
Upper-middle income country panel (n = 23): Algeria, Argentina,
Belarus, Brazil, Bulgaria, Chile, Costa Rica, Dominican Republic, Gabon,
Kazakhstan, Latvia, Macedonia, Malaysia, Mexico, Panama, Peru,
Poland, Romania, Russia, South Africa, Turkey, Uruguay, Venezuela.
Lower-middle income country panel (n= 25): Armenia, Azerbaijan,
Bolivia, Cameron, China, Ecuador, Egypt, El Salvador, Georgia,
Guatemala, Honduras, India, Indonesia, Iran, Jordan, Morocco,
Nicaragua, Pakistan, Paraguay, Philippines, Sudan, Syria, Thailand,
Tunisia, Ukraine.
Low income country panel (n = 11): Bangladesh, Ethiopia, Ghana,
Kenya, Krygyz Republic, Mozambique, Senegal, Tajikistan, Tanzania,
Uzbekistan, Zambia.
Appendix B. Supplementary data
Supplementary data to this article can be found online at
doi:10.1016/j.eneco.2010.12.018.
References
Abosedra, S., Dah, A., Ghosh, S., 2008. Electricity consumption and economic growth:
the case of Lebanon. Applied Energy 86, 429–432.
Acaravci, A., Ozturk, I., 2010. Electricity consumption–growth nexus: evidence from
panel data for transition countries. Energy Economics 32, 604–608.
Akinlo, A.E. (2009), “Electricity Consumption and Economic Growth in Nigeria:
Evidence from Cointegration and Co-feature Analysis”, Journal of Policy Modeling,
31, 681–693.
Altinay, G., Karagol, E., 2005. Electricity consumption and economic growth: evidence
from Turkey. Energy Economics 27, 849–856.
Apergis, N., Payne, J.E., 2009a. Energy consumption and economic growth in Central
America: evidence from a panel cointegration and error correction model. Energy
Economics 31, 211–216.
Apergis, N., Payne, J.E., 2009b. Energy consumption and economic growth: evidence
from the commonwealth of independent states. Energy Economics 31, 641–647.
Aqeel, A., Butt, M.S., 2001. The relationship between energy consumption and economic
growth in Pakistan. Asia Pacific Development Journal 8, 101–110.
Chandran, V.G.R., Sharma, S., Madhavan, K., 2010. Electricity consumption–growth
nexus: the case of Malaysia. Energy Policy 38, 606–612.
Chen, S.T., Kuo, H.I., Chen, C.C., 2007. The relationship between GDP and electricity
consumption in 10 Asian countries. Energy Policy 35, 2611–2621.
Ciarreta, A., Zarraga, A., 2010. Economic growth–electricity consumption causality in
12 European countries: a dynamic panel data approach. Energy Policy 38,
3790–3796.
Coondoo, D., Dinda, S., 2002. Causality between income and emissions: a country
group-specific econometric analysis. Ecological Economics 40, 351–367.
Dinda, S., Coondoo, D., 2006. Income and emission: a panel-data based cointegration
analysis. Ecological Economics 57, 167–181.
Dolado, J.J., Lütkepohl, H., 1996. Making Wald test work for cointegrated VAR systems.
Econometric Theory 15, 369–386.
Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: representation,
estimation, and testing. Econometrica 55, 251–276.
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. Mathematics and Computers in Simulation 64, 431–445.
Ferguson, Wilkinson, R.W., Hill, R., 2000. Electricity use and economic development.
Energy Policy 28, 923–934.
Ghosh, S., 2002. Electricity consumption and economic growth in India. Energy Policy
30, 125–129.
Ghosh, S., 2009. Electricity supply, employment, and real GDP in India: evidence from
cointegration and Granger-causality tests. Energy Policy 37, 2926–2929.
Hansen, B.E., Seo, B., 2002. Testing for two-regime threshold cointegration in vector
error correction models. Journal of Econometrics 110, 293–318.
Harris, R.D.F., Tzavalis, E., 1999. Inference for unit roots in dynamic panels where the
time dimension is fixed. Journal of Econometrics 91, 201–226.
Ho, C.Y., Siu, K.W., 2007. A dynamic equilibrium of electricity consumption and GDP in
Hong Kong: an empirical investigation. Energy Policy 35, 2507–2513.
Hu, J.L., Lin, C.H., 2008. Disaggregated energy consumption and GDP in Taiwan: a
threshold co-integration analysis. Energy Economics 30, 2342–2358.
Im, K.S., Pesaran, M.H., Shin, Y., 2003. Testing for unit roots in heterogeneous panels.
Journal of Econometrics 115, 53–74.
Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economics
Dynamics and Control 12, 231–254.
Johansen, S., 1995. Likelihood-Based Inference in Cointegrated Vector Autoregressive
Models. Oxford University Press, Oxford.
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on
cointegration with applications to money demand. Oxford Bulletin of Economics
and Statistics 52, 169–210.
Jumbe, C.B.L., 2004. Cointegration and causality between electricity consumption and
GDP: empirical evidence from Malawi. Energy Economics 26, 61–68.
Larsson, R., Lyhagen, J., Löthgren, M., 2001. Likelihood-based cointegration tests in
heterogeneous panels. Econometrics Journal 4, 109–142.
Lean, H.H., Smyth, R., 2010. CO2 emissions, electricity consumption, and output in
ASEAN. Applied Energy 87, 1858–1864.
Lee, C.C., Chang, C.P., 2005. Structural breaks, energy consumption, and economic
growth revisited: evidence from Taiwan. Energy Economics 27, 857–872.
Levin, A., Lin, C.F., Chu, C., 2002. Unit root tests in panel data: asymptotic and finite-
sample properties. Journal of Econometrics 108, 1–24.
Lütkepohl, H., 1982. Non-causality due to omitted variables. Journal of Econometrics 19,
267–378.
Maddala, G.S., Wu, S.A., 1999. Comparative study of unit root tests with panel data and a
new simple test. Oxford Bulletin of Economics and Statistics 108, 1–24.
Morimoto, R., Hope, C., 2004. The impact of electricity supply on economic growth in Sri
Lanka. Energy Economics 26, 77–85.
Mozumder, P., Marathe, A., 2007. Causality relationship between electricity consump-
tion and GDP in Bangladesh. Energy Policy 35, 395–402.
Murray, D.A., Nan, G.D., 1996. A definition of the gross domestic product-electrification
interrelationship. Journal of Energy and Development 19, 275–283.
Narayan, P.K., Prasad, A., 2008. Electricity consumption–real GDP causality nexus:
evidence from a bootstrapped causality test for 30 OECD countries. Energy Policy
36, 910–918.
Narayan, P.N., Singh, B., 2007. The electricity consumption and GDP nexus for the Fiji
Islands. Energy Economics 29, 1141–1150.
Narayan, P.K., Smyth, R., 2005. Electricity consumption, employment, and real income
in Australia: evidence from multivariate Granger causality tests. Energy Policy 33,
1109–1116.
Narayan, P.K., Smyth, R., 2009. Multivariate granger-causality between electricity
consumption, exports, and GDP: evidence from a panel of Middle Eastern countries.
Energy Policy 37, 229–236.
Odhiambo, N.M., 2009a. Energy consumption and economic growth nexus in Tanzania:
an ARDL bounds testing approach. Energy Policy 37, 617–622.
 N. Apergis, J.E. Payne / Energy Economics 33 (2011) 770–781
Odhiambo, N.M., 2009b. Electricity consumption and economic growth in South Africa:
a trivariate causality test. Energy Economics 3111, 635–640.
Ozturk, I., Acaravci, A., 2010. The causal relationship between energy consumption and
GDP in Albania, Bulgaria, Hungary, and Romania: evidence from ARDL bound
testing approach. Applied Energy 87, 1938–1943.
Payne, J.E., 2010a. Survey of the international evidence on the causal relationship
between energy consumption and growth. Journal of Economic Studies 37,
53–95.
Payne, J.E., 2010b. A survey of the electricity consumption–growth literature. Applied
Energy 87, 723–731.
Pedroni, P., 1999. Critical values for cointegration tests in heterogeneous panels with
multiple regressors. Oxford Bulletin of Economics and Statistics 61, 653–670.
Pedroni, P., 2004. Panel cointegration: asymptotic and finite sample properties of
pooled time series tests with an application to the PPP hypothesis: new results.
Econometric Theory 20, 597–627.
Pesaran, M.H., Shin, Y., 1999. An autoregressive distributed lag modelling approach to
cointegration analysis. In: Strom, S. (Ed.), Econometrics and Economic Theory in
the 20th Century: The Ragnar Frisch Centennial Symposium. Cambridge University
Press, Cambridge, UK.
Pesaran, M.H., Shin, Y., Smith, R., 2001. Bounds testing approaches to the analysis of
level relationships. Journal of Applied Econometrics 16, 289–326.
Shiu, A., Lam, P.L., 2004. Electricity consumption and economic growth in China. Energy
Policy 32, 47–54.
Soytas, U., Sari, R., 2007. The relationship between energy and production: evidence
from Turkish manufacturing industry. Energy Economics 29, 1151–1165.
Squalli, J., 2007. Electricity consumption and economic growth: bounds and causality
analyses of OPEC countries. Energy Economics 29, 1192–1205.
Tang, C.F., 2008. A re-examination of the relationship between electricity consumption
and economic growth in Malaysia. Energy Policy 36, 3077–3085.
Thoma, M., 2004. Electrical energy usage over the business cycle. Energy Economics 26,
363–385.
781
Toda, H.Y., Yamamoto, T., 1995. Statistical inference in vector autoregressions with
possibly integrated processes. Journal of Econometrics 66, 225–250.
Westerlund, J., 2006. Testing for panel cointegration with multiple structural breaks.
Oxford Bulletin of Economics and Statistics 68, 101–132.
Wolde-Rufael, Y., 2004. Disaggregated industrial energy consumption and GDP: the
case of Shanghai, 1952–1999. Energy Economics 26, 69–75.
Wolde-Rufael, Y., 2006. Electricity consumption and economic growth: a time series
experience for 17 African countries. Energy Policy 34, 1106–1114.
Yang, H.Y., 2000. A note on the causal relationship between energy and GDP in Taiwan.
Energy Economics 22, 309–317.
Yoo, S.H., 2005. Electricity consumption and economic growth: evidence from Korea.
Energy Policy 33, 1627–1632.
Yoo, S.H., 2006. The causal relationship between electricity consumption and economic
growth in the ASEAN countries. Energy Policy 34, 3573–3582.
Yoo, S.H., Kim, Y., 2006. Electricity generation and economic growth in Indonesia.
Energy 31, 2890–2899.
Yoo, S.H., Kwak, S.Y., 2010. Electricity consumption and economic growth in seven
South American countries. Energy Policy 38, 181–188.
Yuan, J., Zhao, C., Yu, S., Hu, Z., 2007. Electricity consumption and economic growth in
China: cointegration and co-feature analysis. Energy Economics 29, 1179–1191.
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, 3077–3094.
Zamani, M., 2007. Energy consumption and economic activities in Iran. Energy
Economics 29, 1135–1140.
Ziramba, E., 2009. Disaggregate energy consumption and industrial production in South
Africa. Energy Policy 37, 2214–2220.