WPS8077

Policy Research Working Paper 8077

Does Energy Efficiency Promote Economic
Growth?
Evidence from a Multi-Country and Multi-Sector Panel
Data Set

Ashish Rajbhandari
Fan Zhang

South Asia Region
Office of the Chief Economist
May 2017
Policy Research Working Paper 8077

Abstract
This paper examines the causal relationship between energy also finds evidence of long-run bidirectional causality between
efficiency and economic growth based on panel data for 56 lower energy intensity and higher economic growth for
high- and middle-income countries from 1978 to 2012. middle-income countries. This finding suggests that beyond
Using a panel vector autoregression approach, the study finds climate benefits, middle-income countries may also earn
evidence of a long-run Granger causality from economic an extra growth dividend from energy efficiency measures.
growth to lower energy intensity for all countries. The study

This paper is a product of the Office of the Chief Economist, South Asia Region. It is part of a larger effort by the World
Bank to provide open access to its research and make a contribution to development policy discussions around the world.
Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted
at fzhang1@worldbank.org.

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

Produced by the Research Support Team
Does Energy Eﬃciency Promote Economic Growth?
Evidence from a Multi-Country and Multi-Sectoral
Panel Data Set

∗
Ashish Rajbhandari Fan Zhang †
Stata World Bank

Keywords: Energy eﬃciency, energy intensity, economic growth, panel cointegration,
Granger causality, vector autoregressions

JEL Classiﬁcation: C32, C33, O13, Q43

∗
Contact: arajbhandari@stata.com
†
Contact: World Bank, 1818 H Street NW, Washington, DC, USA. fzhang1@worldbank.org
We thank Uwe Deichmann for help conceptualize the research question. We thank Vivien Foster, Yadviga
Viktorivna Semikolenova, Govinda Timilsina, and Maria Vagliasindi for helpful comments on earlier drafts.
Financial support from the Trade and Competitiveness Multi-Donor Trust Fund is greatly appreciated.
1 Introduction
Energy eﬃciency is commonly seen as a key policy option for climate change mitigation. It
also has recently been promoted as an industrial policy to boost economic competitiveness.
For example, the European Union’s 2030 Energy Strategy describes energy eﬃciency as
fundamental in the transition toward a more competitive, secure, and sustainable energy
system.1 And the U.S. government has acknowledged energy eﬃciency as a key part of its
strategy to support trade competitiveness.2
There are several potential channels through which energy eﬃciency policies could spur
competitiveness and growth (Deichmann and Zhang, 2013). First, they could encourage
innovation and technology development. When ﬁrms become more productive by using less
energy per unit of output, they could become more cost-competitive in export markets. Sec-
ond, these policies could create new demand and new markets for energy eﬃcient technologies
and products. The ensuing investment could bring new jobs and growth. Third, by enabling
government to reduce energy-related expenditures, especially in energy-importing countries,
they could allow greater spending in other priority areas that would beneﬁt growth in the
long run, such as health and education. Finally, by producing energy savings for households,
they could boost disposable income and encourage growth-promoting consumption.
Despite many anecdotal accounts of the relationship between energy eﬃciency and eco-
nomic growth, empirical evidence on a causal link is thin. Existing analysis takes a deep
dive into speciﬁc industries to explore productivity beneﬁts of energy eﬃciency measures,
such as in iron and steel manufacturing (Worrell et al., 2003), paper and steel manufacturing
(DOE, 1997), and the glass industry (Boyd and Pang, 2000), At the macro level, numer-
ous studies have investigated the causal relationship between total energy consumption and
economic growth (for example Kraft and Kraft 1978 and Costantini and Martini 2010; see
also Ozturk 2010 for a detailed review). But there has been no broader examination of the
macroeconomic eﬀect of energy eﬃciency policies.
In this paper, we explore the causal relationship between energy eﬃciency and GDP
growth based on a panel dataset of 56 countries spanning the period from 1978 to 2012.
Causality here is conﬁned to Granger causality, in which a variable is said to “Granger-
cause” another variable if it serves as a useful predictor of future values of that variable
after controlling for several lags. Energy eﬃciency is proxied by energy intensity, deﬁned as
energy use per unit of economic output. Because changes in energy intensity can be driven
by both eﬃciency eﬀects (resulted in lower energy use to produce the same amount of a good)
and structural eﬀects (resulted in the tendency of energy intensity to ﬁrst rise as a country
moves from agriculture to industry and then fall as it shifts from industry to services), we
1
European Commission. “Energy Eﬃciency and its contribution to energy security and the 2030 Framework
for climate and energy policy”
2
http://trade.gov/press/publications/newsletters/ita 1009/energy 1009.asp

2
combine macro-level analysis with sector-level analysis of industry and agriculture to control
for change in the sectoral composition of the economy.
Previous studies using country-level time-series data or multicountry panel data to test
for Granger causality between total energy consumption and growth have been mostly con-
ﬁned to a bivariate model with energy use and GDP as the two variables (Ozturk, 2010).
However, inference form bivariate models must be interpreted with caution because of po-
tential omitted variable bias (L¨utkepohl, 1982; Zachariadis, 2007). For example, some recent
studies have underscored the importance of controlling for energy price, as energy price could
have a causal impact on both energy use and output growth (Lee and Lee, 2010; Costantini
and Martini, 2010; Belke et al., 2011). In this paper, we speciﬁcally control for energy price
as an endogenous variable so as to allow more conclusive causal inferences.
We apply a panel vector autoregression (PVAR) approach for the causality analysis. We
ﬁrst test for unit roots in the variables using tests proposed by Pesaran (2007) and Deme-
trescu et al. (2006). Both these tests allow for cross-section dependence between countries.
After establishing unit roots in the variables of interest, we test whether the variables are
cointegrated using the panel conintegration test of Pedroni (2004). We ﬁnd evidence of panel
cointegration, which leads us to estimate long-run cointegrating parameters based on gener-
alized methods of moments (GMM). Finally, we use the lagged residuals estimated from the
cointegrating regression as an exogenous error correction term in a PVAR.
Our sample covers a diverse set countries of diﬀerent income levels. Because energy needs
diﬀer at diﬀerent stages of development, we categorize countries into three groups: high in-
come, upper middle income (UMI) and lower middle income (LMI).3 We ﬁnd evidence of
unidirectional causality from GDP growth to lower aggregate energy intensity and lower
industrial energy intensity for high-income countries in the long run. We also ﬁnd bidi-
rectional Granger causality between economic growth and lower aggregate energy intensity
for lower-middle-income countries and between growth and lower industrial energy inten-
sity for upper-middle-income countries, both in the long run. These ﬁndings suggest that
economic development provides opportunities for countries to become more energy eﬃcient,
possibly through capital and technology upgrades (Deichmann and Zhang, 2013). Moreover,
the bidirectional causality for lower-middle-income countries and for the industrial sector of
upper-middle-income countries implies that energy eﬃciency could be an instrument for ac-
celerating growth and productivity. In other words, beyond climate beneﬁts, middle-income
countries may also earn an extra growth dividend from energy eﬃciency measures.
The rest of the paper is organized in the following ways: Section 2 describes the data.
Section 3 discusses the PVAR approach and presents the results. Section 4 discusses the
policy implications and concludes the paper.
3
The analysis excludes low-income countries because energy price data for these countries are not available.

3
2 Data and Descriptive Analysis
We use three types of data for the analysis: data on total ﬁnal energy consumption and
the sectoral energy consumption of agriculture and industry from the International Energy
Agency (IEA) World Energy Statistics and Balances database; data on GDP and value added
of agriculture, services, and industry from the World Bank World Development Indicators
(WDI) database; and energy price data from the IEA Energy Prices and Taxes database, the
Energy Regulators Regional Association (ERRA) Tariﬀ database, and various government
reports and websites. The IEA reports after-tax industry and household electricity and
natural gas prices for OECD and selected non-OECD countries. The ERRA database reports
after-tax prices for residential and non-residential consumers. We use industry or non-
residential electricity price as a proxy for energy price.4 When electricity price data are not
available, we use the industry natural gas price as a substitute. All price and value added
data are converted to 2005 U.S. prices.5 For countries that have gaps in energy price data,
we linearly interpolate those missing observations using the growth rate in the corresponding
country’s Consumer Price Index.6 Table A1 reports the countries as well as the years for
which energy price data are linearly interpolated.
Our main variable of interest is energy intensity, deﬁned as total ﬁnal energy consump-
tion divided by GDP. Sector-speciﬁc energy intensity is deﬁned as the total ﬁnal energy
consumption of the sector (industry or agriculture) divided by sector value added.
Our yearly dataset consists of an unbalanced panel of 56 countries spanning the period
from 1991 to 2012. We categorize the countries as high income, upper middle income, or
lower middle income based on the World Bank’s income classiﬁcation and the countries per
capita gross national income in 2012. Table 1 lists the countries by their income level in
2012 and by data availability for diﬀerent measures of energy intensity.
Table 2 reports the summary statistics. Figures 1, 2, 3, and 4 show the trends in average
GDP, aggregate and sectoral energy intensity, value added shares, and energy price for each
income group over the sample period, respectively. GDP and aggregate energy intensity are
strongly trending in opposite directions since 1990. While GDP has steadily increased, the
aggregate energy intensity of all income groups has fallen dramatically over the past two
decades.7 The negative correlation between GDP and aggregate energy intensity is highly
signiﬁcant, although it does not indicate causality or the direction of causality. Industrial
4
We use IEA data on industry electricity prices whenever available. We use ERRA data on non-residential
electricity prices when IEA data are not available. The same approach applies to natural gas prices data.
5
GDP data are based on purchasing power parity.
6
Our main conclusions are robust to the alternative approach by which missing observations of energy prices
are not interpolated.
7
There was a signiﬁcant increase in energy intensity in middle-income countries after 1991. This is because
many middle-income countries in the sample are transition economies and experienced a sharp contraction
in output after the end to centrally planned production. Our main conclusions are robust when we restrict
sample to observations between 1991 and 2012.

4
energy intensity shows a declining trend for middle-income countries especially after mid
1990s while it remains ﬂat for high-income countries. Agricultural energy intensity shows no
clear time trend. While the value added share of industry have been on a steady decline for
high income countries, the corresponding shares for middle income countries have been in-
creasing until 2010. The value added share of agriculture are declining for all income groups.
Finally, the average energy prices for high-income and upper-middle-income countries have
more than doubled since 2000. For lower-middle-income countries, it ﬂuctuates around 60
U.S. dollars per MWh between 1991 and 2010 and then declined to 40 USD per MWh in
2012.

3 Model and Results
In this section, we describe procedures for and results of three types of tests: panel unit root
test, panel cointegration test and Granger causality test.

3.1 Panel Unit Root Test
Causality tests are very sensitive to the stationarity of the series. Thus we begin by testing
for data stationarity using panel unit root tests proposed by Pesaran (2007) and Deme-
trescu et al. (2006). These tests do not assume that individual time series in the panel are
cross-sectionally independently distributed. Furthermore, they can be applied to unbalanced
panels and are easy to implement with good power and size properties. The test in Pesaran
(2007), the CIPS test, is an extension of the augmented Dickey-Fuller (ADF) test in which
the standard ADF test is augmented with ﬁrst diﬀerences of individual series and cross sec-
tional averages of lagged levels. The test proposed in Demetrescu et al. (2006), the DHT
test, is based on combining the p -values obtained from a Dickey-Fuller test performed on
individual panels. This test is similar to those in Maddala and Wu (1999) and Choi (2001),
except that the DHT test uses the modiﬁed inverse normal method of Hartung (1999) that
is robust to dependence among cross sections.
For the CIPS test, we ﬁt the following cross-sectionally augmented Dickey-Fuller regres-
sion:
¯t−1 + di ∆¯
∆yit = ai + bi yi,t−1 + ci y yt + eit (1)

where yit is the observation of the ith country at time t, ∆ is the ﬁrst diﬀerence operator,
¯t−1 is the ﬁrst lag of the dependent variable averaged over all panels, and ∆¯
y yt is the ﬁrst
diﬀerence of the average. ai , bi , ci , di are parameters and eit is the idiosyncratic error. Under
the unit root null hypothesis, there is bi = 0 for all i. The alternative hypothesis corresponds
to some panels being stationary.

5
For the DHT test, we ﬁt the following panel version of the augmented Dickey-Fuller test:

∆yit = ai + bi yi,t−1 + eit (2)

where the null hypothesis of unit root is bi = 0 for all i and the alternative hypothesis cor-
responds to all panels being stationary. The individual Dickey-Fuller p-values are combined
to obtain the resulting p-value for the panel unit root test.
To determine the order of integration, we test both the levels and the ﬁrst diﬀerences of
the variables. All variables are transformed into logarithms before testing. Table 3 presents
the test results applied to the levels of the variables. For all variables (except agricultural
energy intensity), the null hypothesis is a random walk with a possible drift, with the al-
ternative hypothesis being stationary around a linear time trend. For agricultural energy
intensity, because there was no clear time trend in the series, we test the null hypothesis of a
random walk against the alternative of stationary with no time trend. We fail to reject the
null hypothesis of a random walk with a possible drift in almost all series. In some cases,
we get mixed evidence of nonstationarity when one test rejects the null hypothesis while the
other one fails to do so.
Table 4 reports test results applied to the ﬁrst diﬀerence of all variables. We ﬁnd that both
tests reject the null hypothesis of a random walk with a possible drift in the ﬁrst diﬀerence
of the variables. The unit root tests provide evidence that all variables are integrated with
order one.

3.2 Panel Cointegration Test
Having established the nonstationarity of all variables in the previous section, we now test
whether the variables are cointegrated. Speciﬁcally, we are interested in testing whether
aggregate and sector-level energy intensity are cointegrated with economic growth after con-
trolling for energy price. The existence of cointegration implies Granger causality.
We use the residual-based tests developed by Pedroni (1999, 2004) to test for panel cointe-
gration that allows for a heteregenous cointegrating vector. The null hypothesis is that there
is no cointegration among the variables in the individual panels. The alternative hypothesis
is that either the variables are cointegrated in all panels with a common autoregressive pa-
rameter or the variables are cointegrated in all panels with a country-speciﬁc autoregressive
parameter. The later case allows for additional heterogeneity among panels.
We consider the following regression for the test:

intensityit = αi + δi t + β1i gdpit + β2i priceit + it (3)

where subscript i denotes country and t denotes year. αi is the individual-speciﬁc intercepts
or ﬁxed-eﬀects, δi is the coeﬃcient on the time trend t, β1i and β2i are the individual slope

6
coeﬃcients. intensityit is aggregate energy intensity and the energy intensity of industry
and agriculture. gdpit is the GDP, and priceit is energy price. Pedroni (2004) provides two
sets of statistics: within-dimension or panel cointegration statistics and between-dimension
or group mean panel cointegration statistics. These test statistics diﬀer in how they pool
information across panels. In both cases, the residual-based cointegration test involves ﬁtting
the following regression of the estimated residuals:

ˆit = γi ˆit−1 + uit (4)

where ˆit is the estimated residuals after ﬁtting the model in (3).
The within-dimension statistics are constructed by pooling the autoregressive coeﬃcient
in (4) across diﬀerent countries. The null hypothesis in this case is H0 : γi = 1 whereas
the alternative hypothesis is Ha : γi = γ < 1. The rejection of the null hypothesis of no
cointegration then implies that the variables in all panels are cointegrated with a common
autoregressive parameter. The null hypothesis for the between-dimension statistics is the
same as that of the within-dimension. However, it diﬀers in the alternative hypothesis which
assume country-speciﬁc autoregressive parameter given by Ha : γi < 1. The rejection of the
null hypothesis in this case implies that the variables in all panels are cointegrated with a
country-speciﬁc autoregressive parameter. Pedroni (1999) provides seven test statistics for
testing cointegration among the variables in a panel setting. The ﬁrst four are the within-
dimension statistics: Panel ν , Panel ρ, Panel t (PP), and Panel t(ADF), where PP and ADF
are test statistics of the Phillips-Perron and augmented Dickey Fuller type. The last three
statistics are the between-dimension statistics: Group ρ, Group t(PP), and Group t(ADF).
We apply these tests to the regression in (3).
Table 5 reports the results of cointegration tests for each of the three country income
groups. Almost all test statistics indicate cointegration between diﬀerent measures of energy
intensity and economic growth for high- and upper-middle-income countries, while all do so
for lower-middle-income countries.

3.3 Estimation of Long-Run Parameters
The existence of cointegration implies a causal relationship between GDP and energy inten-
sity, but it does not indicate the direction of the causality. Before we proceed with Granger
causality tests, we ﬁrst estimate the parameters of the long-run relationship described in
equation (3) using dynamic ordinary least squares (DOLS) to establish the sign of the corre-
lation between GDP, energy intensity, and energy price. The DOLS estimator for univariate
time series was proposed in Saikkonen (1991) and Stock and Watson (1993) and extended
to a panel setting by Kao and Chiang (2000) and Mark and Sul (2003). Using Monte Carlo
simulation, Kao and Chiang (2000) and Wagner and Hlouskova (2010) show that this estima-
tor performs better than the nonparametric fully-modiﬁed ordinary least squares estimator

7
in ﬁnite samples. Similar to the single-equation DOLS, the panel version adds leads and
lags of the regressors to correct for endogeneity of the regressors and serial correlation in the
residuals.
Table 6 reports the estimation results. These results show that for all income groups,
GDP growth and a higher energy price are correlated with lower aggregate energy intensity.
The income elasticity is −0.37 for high-income countries, −0.22 for upper-middle-income
countries, and −0.36 for lower-middle-income ones. The price elasticity is −0.10 for high-
income countries, −0.02 upper-middle-income countries, and −0.18 for lower-middle-income
countries. However, the price eﬀect is not statistically signiﬁcant for upper-middle-income
countries.
Results at the sector level are similar. GDP growth is correlated with a decrease in both
industrial and agricultural energy intensity for all income groups. The eﬀect is especially
large for agricultural energy intensity in high-income countries, with a 1 percent increase in
GDP associated with a 1.20 percent decline in this measure. In general, income elasticity
of high-income countries is larger that of middle-income countries. For industrial energy
intensity, it is −0.67 for high-income countries, and between −0.14 and −0.50 for middle-
income countries.
A higher energy price is associated with lower industrial and agricultural energy intensity
for all income groups. Interestingly, sectoral energy intensity is more responsive to energy
price changes in middle-income countries than it is in high-income countries. The price
elasticity of industrial energy intensity ranges from −0.03 for high-income countries to −0.29
for upper-middle-income countries, and −0.27 for lower-middle-income countries. There is
no statistically signiﬁcant correlation between energy price and agricultural energy intensity
in high-income countries. In middle-income countries, the price elasticity ranges from −0.13
in upper-middle-income countries to −0.06 in lower-middle-income countries.

3.4 Granger Causality Tests
In this section, we test for Granger causality using a vector error correction model (VECM).
The VEC term represents any deviation from the long-run equilibrium between GDP, energy
price and energy intensity described above. Adopting the Engle-Granger method, we use
the ﬁrst lags of the residuals as a proxy for long-run deviations in a VECM. Speciﬁcally, we
consider the following multivariate panel VECM with the ﬁrst diﬀerence of intensity, GDP,
and energy prices as the dependent variables:

8
p p
∆intensity it = θ1i + θ11,l ∆gdp it−l + θ12,l ∆price it−l + γ1 ˆit−1 + νit,1 (5)
l=1 l=1
p p
∆gdp it = θ2i + θ21,l ∆intensity it−l + θ22,l ∆price it−l + γ2 ˆit−1 + νit,2 (6)
l=1 l=1
p p
∆price it = θ3i + θ31,l ∆intensity it−l + θ32,l ∆gdp it−l + γ3 ˆit−1 + νit,3 (7)
l=1 l=1

where ∆ is the ﬁrst diﬀerence operator, θki for k = 1, 2, 3 are country-speciﬁc ﬁxed eﬀects,
and θkk,l are the coeﬃcients corresponding to the lth lag of the endogenous variables. γk are
the coeﬃcients of the error correction terms and νit,k are idiosyncratic errors.
Because all variables are nonstationary I(1), we take ﬁrst diﬀerences to make the system
of equations (5)-(7) stable. However, using lagged diﬀerences of the dependent variable in
this system introduces a bias, and a standard ﬁxed-eﬀects estimator would be inconsistent
(Arellano and Bond, 1991). To obtain consistent estimators, we estimate panel GMM pro-
posed by Holtz-Eakin et al. (1988); Arellano and Bond (1991). The panel GMM estimator
uses further lagged diﬀerences of the dependent variable as instruments to remove endogene-
ity arising from lagged regressors. In our application, we use second and third lags of the
dependent variables as instruments.
A VECM allows testing for both short- and long-run causality. In the system of equations
(5)-(7), the coeﬃcients θkk,l represent the short-run eﬀect of the endogenous variables. A
standard Wald test on these coeﬃcients can be used to test for short-run Granger causality.
Speciﬁcally, we test the null hypothesis H0 : θkk,l = 0 for l = 1, . . . , p. In equation (5),
rejecting the null hypothesis θ11,1 = 0 implies that GDP growth Granger-causes change in
energy intensity in the short run. In other words, the ﬁrst lag of the GDP growth is a
signiﬁcant predictor of changes in energy intensity. Similarly, rejecting the null hypothesis
H0 : θ21,1 = 0 in equation(6) implies that GDP is responding to short-term shocks to energy
intensity. The joint signiﬁcance of the coeﬃcients θ11,1 and θ21,1 implies bidirectional causality
in which the two variables Granger-cause each other in the short run, while the rejection of
only one of the hypotheses implies unidirectional causality.
We can test for long-run Granger causality between variables in our model by examining
the signiﬁcance of the coeﬃcient γk , which represents the speed of adjustment to the long-run
equilibrium in response to any shocks to the system. We test the null hypothesis H0 : γk = 0
for k = 1, . . . , 3. The rejection of the null hypothesis implies long-run Granger causality
running from the error- correcting term to the respective dependent variable. For example,
the signiﬁcance of γ1 in equation (5) implies that changes in energy intensity adjust in the
long run to any temporary deviations from economic growth. Similarly, the signiﬁcance of
γ2 in equation (6) implies that changes in energy intensity directly drive economic growth

9
in the long run.
Tables (7)-(9) report the estimated coeﬃcients of the error correction terms and the chi-
squared statistics for the test of Granger causality. We ﬁnd evidence of long-run causality
from GDP and energy price to energy intensity for all income groups. For lower-middle-
income countries, we also ﬁnd evidence of long-run causality from energy intensity and
energy price to GDP growth. This implies that in these countries, GDP in the long run
responds to shocks to energy intensity and energy prices.
In the short run, we ﬁnd bidirectional causality between aggregate energy intensity and
economic growth in high-income countries. We ﬁnd no Granger causality between energy
intensity and economic growth for middle-income countries.
Tables 7-9 show the results of Granger causality tests at the sector level. For all income
groups, there is long-run causality from GDP and energy price to industrial energy intensity.
We also ﬁnd Granger causality from industrial energy intensity and energy price to GDP
for upper-middle-income countries in the long run, and for high- and lower-middle-income
countries in the short run. Finally, there is evidence that GDP and energy price Granger-
cause agricultural energy intensity for upper-middle-income countries in the long run.
As noted, an aggregate analysis at the country level may not reﬂect variation in sectoral
composition. As a robustness check to control for the eﬀect of structural change on energy
intensity, we include in the system of equations (5)-(7) an additional variable measuring the
ratio of industrial and services value added to GDP. The second panel to the right in Tables
7-9 report corresponding long- and short-run Granger causality test results. Our main con-
clusions on the direction of causality between GDP and energy intensity remain the same
under this alternative speciﬁcation. This is consistent with ﬁndings in the literature on the
decomposition of energy intensity indicating that most of the gains in energy productiv-
ity over the past few decades can be attributed to genuine eﬃciency improvements while
structural change in the mix of economic activities has had less inﬂuence (Zhang, 2013).

4 Conclusion
Energy eﬃciency is recognized as playing an important part in climate change mitigation.
However, the long-term relationship between energy eﬃciency and growth has not been fully
understood. Using panel data for 56 countries from 1991 to 2012, this paper presents a
ﬁrst eﬀort to shed light on this question. We employ panel unit root tests, panel cointegra-
tion analysis, and a multivariate panel vector autoregression framework to investigate the
long- and short-run causal relationship between energy intensity (used as a proxy for energy
eﬃciency) and GDP for a mix of high and middle-income countries. Because changes in
energy intensity can be driven by both changes in sectoral composition and improvements
in eﬃciency, we combine macro-level analysis and analysis of the industrial and agricultural
sectors to diﬀerentiate the eﬀects of these two processes on energy intensity.

10
Our main results indicate unidirectional long-run causality running from economic growth
and higher energy prices to lower energy intensity. This conclusion holds for all income
groups and at the sector level. This ﬁnding corroborates the intuition that higher energy
prices provide incentives for increasing energy eﬃciency, while economic development and
demand growth provide opportunities for achieving eﬃciency gains by replacing old plants
and technologies with new ones.
More interestingly, we ﬁnd evidence of bidirectional long-run Granger causality between
energy intensity and economic growth for middle-income countries, implying a feedback
between GDP and energy intensity. This ﬁnding suggests that encouraging energy eﬃciency
could support higher economic growth in the long run. Thus beyond climate beneﬁts, energy
eﬃciency could also provide long-term economic growth beneﬁts.

11
References
Arellano, M. and Bond, S. (1991). Some tests of speciﬁcation for panel data: Monte carlo
evidence and an application to employent equations. The Review of Economic Studies,
58(2):277–297.

Belke, A., Dobnik, F., and Dreger, C. (2011). Energy consumption and economic growth:
New insights into the cointegration relationship. Energy Economics, 33:782–789.

Boyd, G. A. and Pang, J. X. (2000). Estimating the linkage between energy eﬃciency and
productivity. Energy Policy, 28(5):289–296.

Choi, I. (2001). Unit root tests for panel data. Journal of International Money and Finance,
20:249–272.

Costantini, V. and Martini, C. (2010). The causality between energy consumption and
economic growth: a multi-sectoral analysis using non-stationary cointegrated panel data.
Energy Economics, 32:591–603.

Deichmann, U. and Zhang, F. (2013). Growing Green: The Economic Beneﬁts of Climate
Action. World Bank Publication, Washington, DC.

Demetrescu, M., Hassler, U., and Tarcolea, A. I. (2006). Combining signiﬁcance of correlated
statistics with application to panel data. Oxford Bulletin of Economics and Statistics,
68(5):647–663.

DOE (1997). The relationship between environmental goals, productivity improvement, and
increased energy eﬃciency in integrated paper and steel plans. Technical report, U.S.
Department of Energy, Washington, DC. DOE/PO-0055.

Hartung, J. (1999). A note on combining dependent tests of signiﬁcance. Biometrical Journal,
41:849–855.

Holtz-Eakin, D., Newey, W., and Rosen, H. S. (1988). Estimating vector autoregressions
with panel data. Econometrica, 56:1371–1395.

Kao, C. and Chiang, M. H. (2000). On the estimation and inference of a cointegrated re-
gression in panel data. In Baltagi, B. H., editor, Nonstationary panels, panel cointegration
and dynamic panels, volume 15, pages 179–222, Amsterdam. Elseiver.

Kraft, J. and Kraft, A. (1978). On the relationship between energy and gnp. Journal of
Energy and Development, 3:401–403.

Lee, C. and Lee, J. (2010). A panel data analysis of the demand for total energy and
electricity in oecd countries. Energy Journal, 31:1–23.

12
utkepohl, H. (1982). Non-causality due to omitted variables. Journal of Econometrics,
L¨
19:267–378.

Maddala, G. S. and Wu, S. (1999). A comparative study of unit root tests with panel data
and a new simple test. Oxford Bulletin of Economics and Statistics, 61:631–652.

Mark, N. C. and Sul, D. (2003). Cointegration vector estimation by panel dols and long-run
money demand. Oxford Bulletin of Economics and Statistics, 65:655–680.

Ozturk, I. (2010). A literature survey on energy-growth nexus. Energy Policy, 38:340–349.

Pedroni, P. (1999). Critical values for cointegration tests in heterogenous 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 hypothesis. Econometric Theory, 20:597–625.

Pesaran, M. H. (2007). A simple panel unit root test in the presence of cross section depen-
dence. Journal of Applied Econometrics, 22:265–312.

Saikkonen, P. (1991). Asymptotically eﬃcient estimation of cointegrating regressions. Econo-
metric Theory, 7:1–21.

Stock, J. H. and Watson, M. W. (1993). A simple estimator of cointegrating vectors in higher
order integrated systems. Econometrica, 61:783–820.

Wagner, M. and Hlouskova, J. (2010). The performance of panel cointegration methods:
Results from a large scale simulation study. Econometric Reviews, 29(2):182–223.

Worrell, E., Laitner, J., Ruth, M., and Finman, H. (2003). Productivity beneﬁts of industrial
energy eﬃciency measures. Energy, 28:1081–1098.

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.

Zhang, F. (2013). The energy transition of the transition economies: An empirical analysis.
Energy Economics, 40:679–686.

13
50000
40000
2005 USD PPP
30000 20000
10000

1980 1990 2000 2010 2020

(a) High income
12000
10000
8000
2005 USD PPP
6000 4000
2000

1980 1990 2000 2010 2020

(b) Upper middle income
4000
3000
2005 USD PPP
2000 1000
0

1980 1990 2000 2010 2020

Figure 1: Average per capita GDP

14
.2
toe per thousand 2005 USD PPP
.12 .14
.1 .16 .18

1980 1990 2000 2010 2020

High income Upper middle income
Lower middle income

(a) Aggregate energy intensity
.8
toe per thousand 2005 USD PPP
.2 .4
0 .6

1980 1990 2000 2010 2020

High income Upper middle income
Lower middle income

(b) Industrial energy intensity
.2
toe per thousand 2005 USD PPP
.05 .1
0 .15

1980 1990 2000 2010 2020

High income Upper middle income
Lower middle income

Figure 2: Average aggregate and sectoral energy intensities

15
40
30
Percentage
20
10
0

1980 1990 2000 2010 2020

Services Industry
Agriculture

(a) High income
50
40
Percentage
30
20
10

1980 1990 2000 2010 2020

Services Industry
Agriculture

(b) Upper middle income
50
40
Percentage
30
20
10

1980 1990 2000 2010 2020

Services Industry
Agriculture

Figure 3: Average sectoral value added shares

16
140
120
USD per MWh
80 60
40 100

1980 1990 2000 2010 2020

High income Upper middle income
Lower middle income

Figure 4: Average energy prices

17
Table 1: List of countries

Aggregate energy intensity

High income Australia, Austria† , Belgium† , Canada, Chile, Croatia, Cyprus† ,
Czech Republic, Denmark, Estonia, Finland† , France, Germany,
Greece† , Hong Kong, Ireland, Israel, Italy, Japan, Korea,
Luxembourg† , Netherlands† , New Zealand, Norway† , Poland,
Portugal, Russia, Singapore, Slovak Republic, Slovenia, Spain,
Sweden† , Switzerland, United Kingdom, United States
Upper middle income Albania, Bosnia and Herzegovina, Brazil, Bulgaria, Hungary, Jordan,
Kazakhstan, Macedonia† , Malaysia, Mexico, Romania† ,
Serbia, South Africa, Thailand, Turkey, Venezuela
Lower middle income Armenia, Georgia, India, Indonesia, Moldova, Mongolia† , Ukraine

Industrial energy intensity

High income Australia, Austria† , Belgium† , Canada, Chile, Croatia, Cyprus† ,
Czech Republic, Denmark, Estonia, Finland† , France, Germany,
Greece† , Hong Kong, Ireland, Israel, Italy, Japan, Korea,
Luxembourg† , Netherlands† , New Zealand, Norway† , Poland,
Portugal, Russia, Singapore, Slovak Republic, Slovenia, Spain,
Sweden† , Switzerland, United Kingdom, United States
Upper middle income Albania, Brazil, Bulgaria, Hungary, Jordan, Kazakhstan, Macedonia† ,
Malaysia, Mexico, Romania† , Serbia, South Africa, Thailand, Turkey,
Venezuela
Lower middle income Armenia, Georgia, India, Indonesia, Moldova, Mongolia† , Ukraine

Agricultural energy intensity

High income Australia, Austria† , Belgium† , Canada, Chile, Croatia, Cyprus† ,
Czech Republic, Denmark, Estonia, Finland† , France, Germany,
Greece† , Hong Kong, Ireland, Israel, Italy, Japan, Korea,
Luxembourg† , Netherlands† , New Zealand, Norway† , Poland,
Portugal, Russia, Slovak Republic, Slovenia, Spain, Sweden† ,
Switzerland, United Kingdom, United States
Upper middle income Albania, Bosnia and Herzegovina, Brazil, Bulgaria, Hungary, Jordan,
Kazakhstan, Macedonia† , Malaysia, Mexico, Romania† ,
Serbia, South Africa, Thailand, Turkey, Venezuela
Lower middle income Armenia, Georgia, India, Indonesia, Moldova, Mongolia† , Ukraine
Note: † denotes countries for which gaps in energy price observations are linearly interpolated
using Consumer Price Index for multiple years.

18
Table 2: Summary statistics

GDP Mean Std. Dev. Min Max No. of Obs.

High income 22476.39 17834.08 1330.75 116664.3 1269
Upper middle income 4036.65 3053.55 218.49 15,649.72 503
Lower middle income 1060.31 955.16 97.16 4387.70 340

Aggregate energy intensity Mean Std. Dev. Min Max No. of Obs.

High income 0.12 0.06 0.005 0.404 2104
Upper middle income 0.11 0.08 0.006 0.52 1378
Lower middle income 0.14 0.10 0.036 0.75 1095

Industrial energy intensity Mean Std. Dev. Min Max No. of Obs.

High income 0.13 0.07 0.01 0.52 1106
Upper middle income 0.27 0.24 0.004 1.70 1173
Lower middle income 0.48 0.57 0.002 4.76 915

Agricultural energy intensity Mean Std. Dev. Min Max No. of Obs.

High income 0.11 0.10 0.0003 0.75 1106
Upper middle income 0.11 0.14 0.0001 1.27 1181
Lower middle income 0.07 0.15 0.00007 1.16 915

Energy price Mean Std. Dev. Min Max No. of Obs.

High income 75.86 11.64 42.17 327.78 1056
Upper middle income 64.88 33.25 12.22 169.64 309
Lower middle income 56.47 24.87 7.36 166.71 154
Note: The unit of per capita GDP is 2005 USD PPP. The unit of aggregate and sectoral
energy intensities is toe per thousand 2005 USD PPP. The unit of energy price is USD
per MWh.

19
Table 3: Panel unit root tests in levels

CIPS

Income group GDP Aggregate Industry Agriculture Price

High income −1.70 −2.31 −2.15 −1.94 (rw) −2.92∗∗∗
No. of Obs. 1645 1615 903 856 1050
Upper middle income −2.52 −2.44 −2.37 −2.14 (rw) −2.77∗∗
No. of Obs. 570 562 463 443 309
Lower middle income −2.39 −2.65 −2.90∗∗ −2.08 −1.83
No. of Obs. 211 211 210 210 132

DHT

High income 5.53 2.03 3.20 -0.19 (rw) 3.12
Upper middle income 0.55 1.19 0.62 −0.92 (rw) 1.16
Lower middle income −0.87 0.96 0.30 −0.38 1.13
Note: ∗∗∗ , ∗∗ , and ∗ denotes signiﬁcance at the 1%, 5%, and 10% level respec-
tively. (rw) indicates a null hypothesis of random walk. In all other cases,
the null hypothesis is a random walk with a drift. Number of observations
for DHT test are the same as reported for the CIPS test.

Table 4: Panel unit root tests in ﬁrst diﬀerences

CIPS

Income group ∆ GDP ∆ Aggregate ∆ Industry ∆Agriculture ∆ Price

High income −5.35∗∗∗ −7.22∗∗∗ −5.25∗∗∗ −5.43∗∗∗ −5.11∗∗∗
Upper middle income −4.84∗∗∗ −5.99∗∗∗ −5.90∗∗∗ −6.33∗∗∗ −4.11∗∗∗
Lower middle income −4.66∗∗∗ −5.52∗∗∗ −5.55∗∗∗ −4.85∗∗∗ −3.10∗∗∗

DHT

High income −16.46∗∗∗ −24.92∗∗∗ −17.78∗∗∗ −21.08∗∗∗ −13.43∗∗∗
Upper middle income −9.82∗∗∗ −17.50∗∗∗ −16.45∗∗∗ −20.44∗∗∗ −7.14∗∗∗
Lower middle income −6.33∗∗∗ −11.04∗∗∗ −11.38∗∗∗ −11.96∗∗∗ −1.47∗∗∗
Note: ∗∗∗ , ∗∗ , and ∗ denotes signiﬁcance at the 1%, 5%, and 10% level respectively. Number
of observations are the same as reported in Table 3.

20
Table 5: Panel cointegration test

High income

Test statistic Aggregate Industry Agriculture

Panel ν −4.16∗∗∗ −1.09 3.61∗∗∗
Panel ρ −2.17∗∗ −2.05∗∗ −24.74∗∗∗
Panel t - PP 5.45∗∗∗ 2.95∗∗∗ −2.89∗∗∗
Panel t - ADF 1.73∗ 0.97 −2.70∗∗∗
Group ρ −50.20∗∗∗ −68.37∗∗∗ −149.56∗∗∗
Group t - PP −2.64∗∗∗ −5.80∗∗∗ −14.14∗∗∗
Group t - ADF −0.76 −2.66∗∗∗ −4.60∗∗∗

Upper middle income

Test statistic Aggregate† Industry Agriculture

Panel ν 2.95∗∗∗ 5.55∗∗∗ 7.48∗∗∗
Panel ρ −3.74∗∗∗ −2.94∗∗∗ −15.27∗∗∗
Panel t - PP −1.77∗ 0.56 −3.87∗∗∗
Panel t - ADF −0.83 1.27 −2.81∗∗∗
Group ρ −10.79∗∗∗ −7.22∗∗∗ −28.74∗∗∗
Group t - PP −3.13∗∗∗ −0.86 −8.03∗∗∗
Group t - ADF −2.17∗∗ −0.27 −7.13∗∗∗

Lower middle income

Test statistic Aggregate Industry Agriculture

Panel ν −1.30 −1.54 −0.29
Panel ρ 1.66∗ 0.02 −3.64∗∗∗
Panel t - PP 2.75∗∗∗ 0.85 −3.08∗∗∗
Panel t - ADF 2.84∗∗∗ 3.84∗∗∗ −1.88∗
Group ρ 1.97∗∗ 1.04 −4.08∗∗∗
Group t - PP 1.82∗ 1.24 −3.10∗∗∗
Group t - ADF 1.51 3.11∗∗∗ −2.67∗∗∗
Note: ∗∗∗ , ∗∗ , and ∗ denotes signiﬁcance at the 1%, 5%,
and 10% level respectively. † denotes no deterministic
terms and ‡ denotes a constant and a trend term.

21
Table 6: Estimation of long-run cointegration parameters

High income Upper middle income Lower middle income

Intensity GDP Price GDP Price GDP Price

Aggregate −0.37∗∗∗ −0.10∗∗∗ −0.22∗∗∗ −0.02 −0.36∗∗∗ −0.18∗∗∗
Industry −0.67∗∗∗ −0.03∗ −0.14∗∗∗ −0.29∗∗∗ −0.50∗∗∗ −0.27∗∗
Agriculture −1.20∗∗∗ −0.04 −0.16∗∗∗ −0.13∗∗∗ −0.16∗∗∗ −0.06∗
∗∗∗ ∗∗
Note: , , and ∗ denotes signiﬁcance at the 1%, 5%, and 10% level respectively.

22
Table 7: Granger causality - Aggregate energy intensity

Income group Without control for structure With control for structure

GDP/Price Aggregate/Price GDP/Energy GDP/Price Aggregate/Price GDP/Aggregate
↓ ↓ ↓ ↓ ↓ ↓
Aggregate GDP Price Aggregate GDP Price

Long run

High income −1.19∗∗∗ 0.31∗∗ 0.61 −1.05∗∗∗ 0.12 0.15
(0.28) (0.15) (0.65) (0.21) (0.11) (0.67)
Upper middle income −1.33∗∗∗ −0.01 0.31 −1.31∗∗∗ −0.13 −0.36

23
(0.28) (0.21) (0.92) (0.28) (0.19) (0.77)
Lower middle income −0.88∗∗∗ −0.97∗∗ −3.30∗∗ −0.87∗∗∗ −0.65 −2.73∗
(0.31) (0.47) (1.56) (0.31) (0.44) (1.44)

Short run†

High income 7.74∗∗ 15.79∗∗∗ 1.45 9.38∗∗∗ 4.23 3.03
Upper middle income 0.42 0.35 0.40 0.31 0.19 0.52
Lower middle income 1.18 0.09 1.01 0.59 1.28 3.83

Note: ↓ denotes direction of Granger causality. ∗∗∗ , ∗∗ , and ∗ denotes signiﬁcance at the 1%, 5%, and 10% level respectively.
†
represents the chi squared statistic for the test of Granger causality. Standard errors are reported in parentheses.
Table 8: Granger causality - Industrial energy intensity

Income group Without control for structure With control for structure

GDP/Price Industry/Price GDP/Industry GDP/Price Industry/Price GDP/Industry
↓ ↓ ↓ ↓ ↓ ↓
Industry GDP Price Industry GDP P

Long run

High income -1.70∗∗∗ -0.01 0.66 -1.57∗∗∗ 0.04 0.75
(0.34) (0.11) (0.54) (0.33) (0.10) (0.57)
Upper middle income −1.40∗∗∗ −0.17∗∗ −1.17∗∗∗ −1.32∗∗∗ −0.15∗∗ −1.17∗∗∗

24
(0.19) (0.08) (0.43) (0.22) (0.08) (0.45)
Lower middle income -1.52∗∗∗ −0.07 −0.64 −1.57∗∗∗ −0.05 −0.23
(0.25) (0.09) (0.27) (0.27) (0.10) (0.39)

Short run†

High income 4.64∗ 16.42∗∗∗ 3.36 2.44 3.46 1.43
Upper middle income 1.49 0.04 1.13 0.15 0.30 0.26
Lower middle income 6.11∗∗ 0.35 2.33 8.00∗∗ 1.00 5.42∗

Note: ↓ denotes direction of Granger causality. ∗∗∗ , ∗∗ , and ∗ denotes signiﬁcance at the 1%, 5%, and 10% level
respectively. † represents the chi squared statistic for the test of Granger causality. Standard errors are reported in
parentheses.
Table 9: Granger causality - Agricultural energy intensity

Income group Without control for structure With control for structure

GDP/Price Agriculture/Price GDP/Agriculture GDP/Price Agriculture/Price GDP/Agriculture
↓ ↓ ↓ ↓ ↓ ↓
Agriculture GDP Price Agriculture GDP Price

Long run

High income −0.65∗∗∗ −0.03 0.05 −0.62∗∗∗ −0.03 −0.002
(0.24) (0.02) (0.14) (0.26) (0.02) (0.14)
Upper middle income −1.87∗∗∗ −0.02 −0.12 −1.68∗∗∗ −0.05 −0.22

25
(0.53) (0.04) (0.20) (0.54) (0.05) (0.25)
Lower middle income −1.47∗∗∗ 0.05 −0.27 −1.32∗∗∗ −0.09 −0.0004
(0.12) (0.03) (0.17) (0.13) (0.10) (0.24)

Short run†

High income 2.72 7.10∗∗ 3.36 1.00 2.15 0.28
Upper middle income 0.79 3.91 0.37 1.51 4.79∗ 0.04
Lower middle income 12.70∗∗∗ 2.12 19.65∗∗∗ 22.42∗∗∗ 3.19 4.37
†
Note: ↓ denotes direction of Granger causality. ∗∗∗ , ∗∗ , and ∗ denotes signiﬁcance at the 1%, 5%, and 10% level respectively.
The values represent the chi squared statistic for the test of Granger causality. Standard errors are reported in parentheses.
Appendix

Table A1: Linear interpolation for energy price

Country Years

Austria 2001-2003; 2009-2011
Belgium 2001-2007
Cyprus 2007
Finland 2006
Greece 2006-2007
Luxembourg 1990-2007
Netherlands 2002-2006
Norway 1992 - 1999
Sweden 1998 - 2006
Macedonia 2000 - 2003
Romania 2006 - 2007
Mongolia 2000 - 2002