- Email: [email protected]
The causal dynamics between coal consumption and growth: Evidence from emerging market economies
Applied Energy 87 (2010) 1972–1977
Contents lists available at ScienceDirect
Applied Energy journal homepage: www.elsevier.com/locate/apenergy
The causal dynamics between coal consumption and growth: Evidence from emerging market economies Nicholas Apergis a, James E. Payne b,* a b
Department of Banking and Financial Management, University of Piraeus Karaoli and Dimitriou 80 Piraeus, Attiki 18534, Greece College of Arts and Sciences, Illinois State University, Normal, IL 61790-4100, United States
a r t i c l e
i n f o
Article history: Received 10 September 2009 Received in revised form 29 November 2009 Accepted 30 November 2009 Available online 13 January 2010 Keywords: Coal consumption Growth Panel Granger-causality
a b s t r a c t This study examines the relationship between coal consumption and economic growth for 15 emerging market economies within a multivariate panel framework over the period 1980–2006. The heterogeneous panel cointegration results indicate there is a long-run equilibrium relationship between real GDP, coal consumption, real gross fixed capital formation, and the labor force. While in the long-run both real gross fixed capital formation and the labor force have a significant positive impact on real GDP, coal consumption has a significant negative impact. The panel causality tests show bidirectional causality between coal consumption and economic growth in both the short- and long-run. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction Though coal has been a dependable energy source within the world’s energy consumption mix, the environmental consequences of the sustained use of coal has drawn into question the long-term viability of coal in light of the emergence of cleaner, alternative energy sources. Researchers have recently begun to examine the dynamic causal relationship between coal consumption and economic growth and the energy policy implications of their findings.1 The strong economic growth and the rising energy demands of emerging market economies motivate the study of the coal consumption-growth nexus for these countries. Specifically, this study extends the previous research in an examination of a balanced panel of 15 emerging market economies as defined by Morgan Stanley capital international (MSCI) over the period 1980–2006 in the estimation of a panel error correction model to infer the causal relationship between coal consumption and economic growth.2 Table 1 displays some statistics related to the production and consumption of coal along with carbon dioxide emissions attrib-
* Corresponding author. Tel.: +1 309 438 5669. E-mail addresses: [email protected] (N. Apergis), [email protected] (J.E. Payne). 1 See Payne [21–23], for surveys of the literature on the causal relationship between energy consumption and economic growth. Nel and van Zyl [20] discuss energy-based growth models and through model calibration and forecasting to infer the prospects for energy-constrained economic growth. 2 Morgan Stanley Capital International characterizes twenty-two countries as emerging markets. Seven countries are omitted from the analysis due to the unavailability of data for all the variables over the entire period 1980–2008: Colombia, Israel, Pakistan, Poland, Russia, Taiwan, and Turkey. 0306-2619/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.apenergy.2009.11.035
uted to fossil fuels for the 15 emerging market economies. As the table shows, there is a great deal of variation in these statistics across countries. In terms of coal production, China ranks the highest while Morocco ranks the lowest. In the case of coal consumption, again China ranks the highest while Argentina is the lowest. With respect to carbon dioxide emissions, China has the highest rank whereas Peru has the lowest rank. China, India, and South Africa are among the world leaders in both coal production and consumption, not to mention in carbon dioxide emissions from fossil fuel usage. China ranks first in the world in the production and consumption of coal along with carbon dioxide emissions. India ranks third in both coal production and consumption and fourth in carbon dioxide emissions. South Africa ranks sixth in coal production, seventh in coal consumption, and 12th in carbon dioxide emissions. Further examination of Table 1 reveals, though Indonesia is eighth in coal production, it ranks 26th in coal consumption. In fact, Indonesia was the second largest net exporter of coal in the world for 2004.3 Based on the data reported in Table 1, 12 of the 15 countries are net importers of coal. The study extends the literature on the causal relationship between coal consumption and economic growth along several dimensions. First, the study includes a larger set of countries in the analysis than previous studies. Second, with the exception of the studies by Yuan et al. [37], Payne [24], Wolde-Rufael [39], and Apergis and Payne [4], the analysis will be undertaken within a production model framework by including measures for capital
3
Indonesia’s Country Analysis Brief, Energy Information Administration.
1973
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977 Table 1 Coal summary statistics, 2006. Country
Argentina Brazil Chile China Egypt Hungary India Indonesia Malaysia Mexico Morocco Peru Philippines South Africa Thailand
Emerging market economies Coal production
Coal consumption
Carbon dioxide emissions
0.051 (58) 6.490 (32) 0.437 (46) 2620.498 (1) 0.028 (62) 10.970 (28) 500.193 (3) 213.174 (8) 1.164 (44) 12.662 (26) 0.000 (66) 0.118 (53) 2.597 (40) 269.828 (6) 21.022 (22)
0.906 (66) 23.604 (27) 6.445 (47) 2584.246 (1) 1.425 (60) 13.059 (35) 539.486 (3) 24.071 (26) 16.868 (31) 19.876 (29) 6.478 (46) 1.627 (59) 11.117 (37) 195.225 (7) 33.156 (24)
162.19 (28) 377.24 (17) 64.80 (50) 6017.69 (1) 151.62 (30) 58.65 (54) 1293.17 (4) 280.36 (22) 163.53 (27) 435.60 (13) 34.53 (73) 29.93 (74) 72.39 (47) 443.58 (12) 245.04 (24)
Notes: Coal production and consumption denoted in million short tons. Carbon dioxide emissions defined as total from consumption of fossil fuels in million metric tons of CO2. Numbers in parentheses represent world rank. Data obtained from Country Energy Profiles of the Energy Information Administration.
and labor. Third, noting the studies by Hu and Lin [12], Sari et al. [33], Wolde-Rufael [39], and Apergis and Payne [4], the sign and magnitude of the respective coefficient estimates from the panel error correction model will be presented to facilitate the policy interpretation of the findings. Fourth, given the relatively short time horizon of the data, panel unit root and cointegration tests are utilized to provide additional power and size properties over standard unit root and cointegration tests by combining the cross-section and time series data across countries.4 Section 2 outlines the policy implications associated with the causal relationship between coal consumption and economic growth along with a summary of the empirical literature to date. Section 3 discusses the data, methodology, and empirical results. Section 4 provides concluding remarks.
2. Overview of the coal consumption and economic growth literature The identification of the causal relationship between coal consumption and economic growth has a number of policy implications. If an increase in coal consumption causes an increase in economic growth, then energy conservation policies that adversely impact coal consumption may also have an adverse impact on economic growth. On the other hand, if an increase in coal consumption causes a decrease in economic growth due perhaps to the inefficient and excessive use of coal, then energy conservation policies that reduce coal consumption may actually mitigate the adverse impact of coal consumption on economic growth.5 Alternatively, economic growth may cause either an increase or decrease in coal consumption. If an increase in economic growth causes an increase in coal consumption, then energy conservation policies that reduce coal consumption may not have an adverse impact on economic growth. However, the case of an increase in economic growth that causes a decrease in coal consumption may be reflective of an economy that is becoming less coal intensive.6 It is also possible that coal consumption and economic growth 4
The panel error correction methodology employed in this study parallels Apergis and Payne [1–4]. 5 Wolde-Rufael [39] alludes to the possibility that industries dependent on coal may have become less efficient over time. Also, the absence of binding legislative restrictions on carbon dioxide emissions may be a contributing factor to the excessive use of coal. 6 As pointed out by Wolde-Rufael [39], coal as an energy source may be decreasing relative to other energy sources in the production of electricity for example.
exhibit an interdependent relationship as causation may run in both directions. Likewise, it is conceivable that coal consumption plays such a minor role in the economic growth process that no causal relationship can be detected. Previous studies provide a range of results for a relatively small number of countries on the causal relationship between coal consumption and economic growth.7 Yang [34] uses the Engle–Granger bivariate cointegration procedure to find the absence of a long-run equilibrium relationship between coal consumption and real output in Taiwan; however, Granger-causality tests reveal unidirectional causality from economic growth to coal consumption. In a followup study on Taiwan, Yang [35] employs the same methodology to show bidirectional causality between coal consumption and economic growth. Fatai et al. [9] reveal unidirectional causality from economic growth to coal consumption in the case of Australia with both the Johansen–Juselius and Toda–Yamamoto approaches to causality testing; however, the absence of a causal relationship using the autoregressive distributed lag (ARDL) model. Furthermore, Fatai et al. [9] fail to find a causal relationship between coal consumption and economic growth for New Zealand using either the Johansen– Juselius or Toda–Yamamoto procedures. Using the Toda–Yamamoto approach to causality testing, Wolde-Rufael [38] presents evidence of unidirectional causality from coal consumption to real output in the case of Shanghai. In another study of Taiwan, Lee and Chang [16] find bidirectional causality between coal consumption and economic growth within a bivariate error correction model. Yoo [36] also reports bidirectional causality between coal consumption and economic growth from a bivariate error correction model in the case of South Korea. Hu and Lin [12] utilize the Hansen-Seo asymmetric cointegraton procedure to reveal asymmetries in the relationship between coal consumption and economic growth for Taiwan along with bidirectional causality. Jinke et al. [14] use a bivariate error correction model to show unidirectional causality from economic growth to coal consumption for China and Japan, but the absence of a causal relationship for India, South Africa, and South Korea. 7 A summary of the coal consumption-growth studies is provided in Table 1 of Apergis and Payne [4]. Though not explicitly testing for Granger-causality both Sari and Soytas [30] and Ewing et al. [8] employ generalized forecast error variance decomposition analysis. In the case of Turkey, Sari and Soytas [30] show that coal consumption explains up to 8% of the forecast error variance of real GDP. In a study of the US, Ewing et al. [8] find that coal consumption explains up to 10% of the forecast error variance of industrial production. In an examination of coal production in the former Soviet Union, Reynolds and Kolodziej [29] find unidirectional causality from economic growth to coal production.
1974
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977
Sari et al. [33] use the autoregressive distributed lag (ARDL) model to find unidirectional causality from coal consumption to industrial production for the US. In the case of China, Yuan et al. [37] estimate a multivariate vector error correction model to report bidirectional causality between coal consumption and economic growth. In a study of US fossil fuel consumption, Payne [24] fails to find a causal relationship between coal consumption and real output using the Toda–Yamamoto procedure for causality testing. In a multi-country study, Wolde-Rufael [39] finds through the Toda–Yamamoto procedure unidirectional causality from coal consumption to real output for India and Japan; unidirectional causality from real output to coal consumption for China and South Korea; and bidirectional causality for South Africa and the US. In another study of South Africa, Ziramba [40] fails to find a causal relationship between coal consumption and industrial production; however, the evidence does suggest unidirectional causality from coal consumption to employment. Finally, in a panel study of 25 OECD countries, Apergis and Payne [4] find bidirectional causality between coal consumption and economic growth. The next section describes the data, panel unit root and cointegration tests along with the results of a multivariate panel error correction model.8
3. Data, methodology, and results Annual data from 1980 to 2006 were obtained from the World Bank Development Indicators, CD-ROM and the Energy Information Administration for Argentina, Brazil, Chile, China, Egypt, Hungary, India, Indonesia, Malaysia, Mexico, Morocco, Peru, Philippines, South Africa, and Thailand. The multivariate framework includes real GDP (Y) in billions of constant 2000 US dollars, real gross fixed capital formation (K) in billions of constant 2000 US dollars, total labor force (L) in millions, and coal consumption (CC) defined in thousands of metric tons.9 Before proceeding to panel cointegration tests and specification of the panel error correction model, a battery of panel unit root tests are undertaken to infer the stationarity properties of the variables. Levin et al. [17] propose a panel based ADF test that assumes homogeneity in the dynamics of the autoregressive coefficients for all panel units. On the other hand, the Im et al. [13] panel unit root test allows for heterogeneity in the dynamics of the autoregressive coefficients for all panel units. Following Maddala and Wu [18], the nonparametric panel unit root tests are estimated using the FisherADF and Fisher-PP tests. Finally, the Carrion-i-Silvestre et al. [6] heterogeneous panel stationarity test is also examined. Under the Levin et al. [17], Im et al. [13], Fisher-ADF and Fisher-PP tests the null hypothesis is a unit root while the alternative hypothesis is no unit root. On the other hand, the Carrion-i-Silvestre et al. [6] test assumes stationarity under the null hypothesis.10 Panel A of Table 2 reports the results of the panel unit root tests which indicate that each variable is integrated of order one.11 Given the variables are integrated of the same order, the Pedroni [25,27] heterogeneous panel cointegration test which allows 8 See Apergis and Payne [1–4] and citations therein, for additional studies on the use of panel cointegration and error correction modeling in the energy consumptiongrowth literature. 9 Capital is proxied by real gross fixed capital formation in that changes in investment closely align with changes in the capital stock under the assumption of a constant depreciation rate using the perpetual inventory method (see [31] and citations therein as well as [4,39]. 10 Tests of dynamic heterogeneity (i.e. variation of the intercept over countries and time) of the variables in the model were performed following the methodology of Holtz-Eakin et al. [11] and Holtz-Eakin [10]. The results which are available upon request indicate that the relationships exhibit heterogeneity in both the dynamics and error variances across the 15 countries. 11 To conserve space the details of the panel unit root tests are not discussed.
Table 2 Panel unit root and cointegration tests emerging market economies, 1980–2006. Panel A: Panel unit root tests LLC
IPS
Fisher-ADF
Fisher PP
CLB
Y 0.72 0.82 15.38 17.64 19.38b DY 6.47a 7.85a 89.05a 133.89a 2.36 CC 0.56 0.65 20.47 19.21 24.33b a a a a DCC 5.89 6.09 71.28 86.39 1.51 K 0.72 0.78 15.44 25.49 10.79b DK 7.12a 7.53a 90.36a 113.79a 1.32 L 0.97 0.60 20.22 24.38 22.49b DL 8.49a 8.64a 89.45a 146.50a 2.56 Notes: Notation for tests: Levin et al. [17] LLC; Im et al. [13] IPS; Maddala and Wu [18] Fisher-ADF and Fisher PP; Carrion-i-Silvestre et al. [6] CLB. Significance levels at the 1% and 5% are denoted by ‘‘a” and ‘‘b”, respectively. Panel B: Panel cointegration tests Within dimension test statistics Between dimension test statistics 57.35006a Group q-statistic 50.57044a Panel v-statistic Panel q-statistic 50.52008a Group PP-statistic 54.37261a Panel PP55.09101a Group ADF-statistic 4.64055a statistic Panel ADF4.66656a statistic Notes: Of the seven tests, the panel v-statistic is a one-sided test where large positive values reject the null hypothesis of no cointegration whereas large negative values for the remaining test statistics reject the null hypothesis of no cointegration. Critical value at the 1% significance level denoted by ‘‘a”: panel v (24.56), panel q (17.60), panel and group PP (25.59), panel ADF (2.97), group q (21.12), and group ADF (3.18).
for cross-section interdependence with different individual effects is estimated to determine whether a long-run equilibrium relationship exists:
Y it ¼ ai þ di t þ b1i CC it þ b2i K it þ b3i Lit þ eit
ð1Þ
where i = 1, . . . , N for each country in the panel and t = 1, . . . , T refers to the time period. The parameters ai and di allow for the possibility of country-specific fixed effects and deterministic trends, respectively. eit denote the estimated residuals which represent deviations from the long-run relationship. All variables are expressed in natural logarithms so the b parameters of the model can be interpreted as elasticity estimates. The null hypothesis of no cointegration, qi = 1, is tested by conducting the following unit root test on the residuals as follows:
eit ¼ qi eit1 þ wit
ð2Þ
Pedroni [25,27] proposes two types of tests for cointegration: panel and group tests. The panel tests are based on the within dimension approach which includes four statistics: panel v, panel q, panel PP, and panel ADF-statistics. These statistics essentially pool the autoregressive coefficients across different countries for the unit root tests on the estimated residuals. These statistics take into account common time factors and heterogeneity across countries. The group tests are based on the between dimension approach which includes three statistics: group q, group PP, and group ADF-statistics. These statistics are based on averages of the individual autoregressive coefficients associated with the unit root tests of the residuals for each country in the panel. All seven tests are distributed asymptotically as standard normal.12 Panel B of Table 2 reports both the within and between dimension panel cointegration test statistics. All seven test statistics reject the null hypothesis of no cointegration at the 1% significance level. In light of the panel cointegration results, the fully modified OLS (FMOLS) technique for heterogeneous cointegrated panels 12 See Pedroni [25], for details on the heterogeneous panel and heterogeneous group mean panel cointegration statistics.
1975
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977 Table 3 FMOLS long-run estimates and panel causality tests emerging market economies, 1980–2006. Panel A: FMOLS long-run estimates þ 0:219a K þ 0:247L Y ¼ 0:047a 0:251CC a a ð5:09Þ
ð5:86Þ
Adj. R2 ¼ 0:61
ð7:02Þ
LM ¼ 0:85
ð4:50Þ
RESET ¼ 1:53
½0:52
½0:27
HE ¼ 0:93 ½0:47
Notes: t-statistics and probability values are reported in parentheses and brackets, respectively. LM is the lagrange multiplier test for serial correlation. RESET is the misspecification test. HE is White’s heteroscedasticity test. Significance at the 1% level denoted by ‘‘a”. Panel B: Panel causality test Dependent variable
Sources of causation (independent variables) Short-run
Long-run
DY
DCC
(3a)
DY
40.01 [0.00]a
(3b)
DCC
40.08 [0.00]a
(0.208) [0.00]a
(3c)
DK
40.23 [0.00]a
(0.171) [0.00]a
(3d)
DL
0.09 [0.92]
DK (0.023) [0.00]a
DL
ECT
51.30 [0.00]a
(0.341) [0.00]a
50.33 [0.00]a
(0.687) [0.00]a
0.013 [0.01]a
30.09 [0.00]a
(0.192) [0.00]a
0.30 [0.58]
(0.002) [0.57]
0.024 [0.01]a
30.26 [0.00]a
(0.164) [0.00]a
0.038 [0.01]a
(0.005) [0.91]
51.01 (0.507) 0.02 (0.009) 49.11 (0.326) 0.017 [0.00]a [0.90] [0.89] [0.00]a [0.00]a [0.02]b [0.00]a Notes: Partial F-statistics reported with respect to short-run changes in the independent variables. 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% and 5% levels denoted by ‘‘a” and ‘b”, respectively.
is estimated to determine the long-run equilibrium relationship [26].13 The FMOLS results are displayed in Panel A of Table 3. The coefficients are positive and significant at the 1% level for real gross fixed capital formation and the labor force; however, the coefficient for coal consumption is negative and significant at the 1% level. As discussed by Kolstad and Krautkraemer [15], Sari and Soytas [32], and Apergis and Payne [4], the negative impact in the long-run of an increase in coal consumption on real GDP may be the result of the economic costs associated with the environmental impact of carbon dioxide emissions outweigh the immediate economic benefit of coal usage for real GDP.14 Given the coefficients can be interpreted as elasticity estimates, the results indicate that a 1% increase in coal consumption decreases real GDP by 0.251%; a 1% increase in real gross fixed capital formation increases real GDP by 0.291%; and a 1% increase in the labor force increases real GDP by 0.247%. As a comparison, Apergis and Payne [4] find that a 1% increase in coal consumption decreases real GDP by 0.179%; a 1% increase in real gross fixed capital formation increases real GDP by 0.526%; and a 1% increase in the labor force increases real GDP by 0.285% for a panel of 25 OECD countries. Next, a panel vector error correction model is estimated [28]. The Engle and Granger [7] two-step procedure is employed by first estimating the long-run model specified in Eq. (1) to obtain the estimated residuals. The lagged residuals from Eq. (1) will serve as the error correction term in the estimation of the dynamic error correction model as follows:
13 See Narayan and Wong [19], for an example of the FMOLS estimation. The estimates from either the FMOLS or DOLS are asymptotically equivalent for more than 60 observations [5]. The panel data sets of this study contain 405 observations. 14 Sari and Soytas [32] provide several reasons associated with the negative impact of energy consumption on income in the long-run: (1) inefficient use of energy, (2) production of output to less energy intensive service sectors, and (3) improvement in the energy sector crowds out non-energy sectors, known as ‘‘Dutch disease” whereby the abundance of a natural resource has a negative impact on other tradable goods sectors. Though the first two points raised by Sari and Soytas [32] are mentioned in this study, the presence of the ‘‘Dutch disease” phenomenon may not be as relevant for this study as 12 of the 15 countries are net importers of coal.
q X
DY it ¼ n1j þ
q X
/11ik DY itk þ
k¼1
þ
q X
/13ik DK itk þ
k¼1
q X
/14ik DLitk þ k1i eit1 þ u1it
q X
/21ik DY itk þ
q X
k¼1 q X
q X
q X
q X
q X
/31ik DY itk þ
/33ik DK itk þ
q X
q X
q X k¼1
/32ik DCC itk
/34ik DLitk þ k3i eit1 þ u3it
ð3cÞ
k¼1
/41ik DY itk þ
q X
k¼1
þ
ð3bÞ
k¼1
k¼1
DLit ¼ n4j þ
/24ik DLitk þ k2i eit1 þ u2it
k¼1
k¼1
þ
/22ik DCC itk
k¼1
/23ik DK itk þ
k¼1
DK it ¼ n3j þ
ð3aÞ
k¼1
DCC it ¼ n2j þ þ
/12ik DCC itk
k¼1
/43ik DK itk þ
/42ik DCC itk
k¼1 q X
/44ik DLitk þ k4i eit1 þ u4it
ð3dÞ
k¼1
where D is the first-difference operator; k is the lag length set at one based on likelihood ratio tests; and u is the serially uncorrelated error term. With respect to Eqs. (3a)–(3d), short-run causality is determined by the statistical significance of the partial F-statistic associated with the corresponding right hand side variables. Longrun causality is revealed by the statistical significance of the respective error correction terms using a t-test. Panel B of Table 3 reports the results of the short-run and longrun Granger-causality tests. In Eq. (3a), real gross fixed capital formation and the labor force each have a significant positive impact on economic growth in the short-run while coal consumption has a significant negative impact. The negative impact of coal consumption on economic growth may be attributed to an inefficient and excessive use of coal as noted by Wolde-Rufael [39] and Apergis and Payne [4]. With respect to Eq. (3b), both economic growth
1976
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977
and real gross fixed capital formation have a significant negative impact on coal consumption while the labor force has as insignificant impact in the short-run. Wolde-Rufael [39] and Apergis and Payne [4] suggest this negative impact on coal consumption may indicate that the overall energy consumption mix has become less coal intensive. In the case of Eq. (3c), both economic growth and the labor force each have a significant positive impact on real gross fixed capital formation while coal consumption has no impact. For Eq. (3d), both economic growth and real gross fixed capital formation have a significant positive impact on the labor force whereas coal consumption has an insignificant impact. A review of the short-run causality results indicates there is bidirectional causality between coal consumption and economic growth. The long-run dynamics displayed by the error correction terms from Eqs. (3a)–(3d) show that economic growth, coal consumption, real gross fixed capital formation, and the labor force each respond to deviations from long-run equilibrium as noted by the statistical significance of their respective error correction terms. However, given the magnitude of the coefficients on the error correction terms, the speed of adjustment towards long-run equilibrium is rather slow. In summary, the results suggest there is bidirectional causality between coal consumption and economic growth in both the short- and long-run. In the short-run, the coefficient estimates indicate that coal consumption has a negative impact on economic growth and likewise economic growth also has a negative impact on coal consumption. These results are similar to those reported by Wolde-Rufael [39] for the US and Apergis and Payne [4] for a panel of OECD countries.
4. Concluding remarks In light of the economic growth and rising energy demands of the emerging market economies, this study examines the role of coal consumption in the economic growth process for a panel of 15 emerging market economies. Specifically, the causal relationship between coal consumption and economic growth is evaluated using a multivariate panel error correction model over the period 1980–2006. Heterogeneous panel cointegration tests indicate there is a long-run equilibrium relationship between real GDP, coal consumption, real gross fixed capital formation, and the labor force. The long-run elasticity estimates are positive and significant for both real gross fixed capital formation and the labor force; however, the elasticity estimate with respect to coal consumption is negative and significant. As discussed by Wolde-Rufael [39] and Apergis and Payne [4] this negative impact of coal consumption on real output in the long-run may be the consequence of the economic costs of carbon dioxide emissions on the environment outweigh the immediate economic benefit associated with coal usage on real output. The estimation of the panel error correction model shows there is both short- and long-run bidirectional causality between coal consumption and economic growth. However, the short-run causality results indicate that an increase in coal consumption has a negative impact on economic growth, a result that could be attributed to an inefficient and excessive use of coal. In this case, the feasibility of either increasing the efficient use of coal or reducing coal consumption should be assessed in the formulation of energy policy. For example, legislation that would restrict the growth of carbon dioxide emissions might provide an incentive to enhance efficiency or curb excessive coal consumption. Furthermore, greater use of sustainable coal technologies that permit carbon dioxide capture and storage may curtail the environmental consequences of excessive coal consumption. Similarly, the short-run causality tests also indicate that an increase in economic growth has a negative impact on coal consumption. This
result suggests the emergence of alternative energy sources such as renewables may have rendered these economies less coal intensive over time. References [1] Apergis N, Payne JE. Energy consumption and economic growth in Central America: evidence from a panel cointegration and error correction model. Energy Econ 2009;31:211–6. [2] Apergis N, Payne JE. Energy consumption and economic growth: evidence from the commonwealth of independent states. Energy Econ 2009;31:641–7. [3] Apergis N, Payne JE. Renewable energy consumption and economic growth: evidence from a panel of OECD countries. Energy Policy 2010;38:656–60. [4] Apergis N, Payne, JE. Coal consumption and economic growth: evidence from a panel of OECD countries. Energy Policy, in press. [5] Banerjee A. Panel data unit roots and cointegration: an overview. Oxford Bull Econ Stat 1999;61(S1):607–29. [6] Carrion-i-Silvestre JL, del Barrio-Castro T, Lopez-Bazo E. Breaking the panels: an application to GDP per capita. Economet J 2005;8:159–75. [7] Engle RF, Granger CWJ. Cointegration and error correction: representation, estimation, and testing. Econometrica 1987;55:251–76. [8] Ewing BT, Sari R, Soytas U. Disaggregate energy consumption and industrial output in the United States. Energy Policy 2007;35:1274–81. [9] Fatai K, Oxley L, Scrimgeour FG. Modelling the causal relationship between energy consumption and GDP in New Zealand, Australia, India, Indonesia, the Philippines, and Thailand. Math Comput Simulat 2004;64:431–45. [10] Holtz-Eakin D. Testing for individual effects in dynamic models using panel data. NBER technical paper series; 1986. No. 57. [11] Holtz-Eakin D, Newey W, Rosen H. Implementing causality tests with panel data with an example from local public finance. NBER technical working paper; 1985. No. 48. [12] Hu JL, Lin CH. Disaggregated energy consumption and GDP in Taiwan: a threshold co-integration analysis. Energy Econ 2008;30:2342–58. [13] Im KS, Pesaran MH, Shin Y. Testing for unit roots in heterogeneous panels. J Econ 2003;115:53–74. [14] Jinke L, Hualing S, Dianming G. Causality relationship between coal consumption and GDP: difference of major OECD and non-OECD countries. Appl Energy 2008;85:421–9. [15] Kolstad CD, Krautkraemer JA. Natural resources use and the environment. In: Kneese AV, Sweeney JL, editors. Handbook of natural resources and energy economics, vol. 3. p. 1219–65. [16] Lee CC, Chang CP. Structural breaks, energy consumption and economic growth revisited: evidence from Taiwan. Energy Econ 2005;27:857–72. [17] Levin A, Lin CF, Chu C. Unit root tests in panel data: asymptotic and finitesample properties. J Econ 2002;108:1–24. [18] Maddala GS, Wu SA. Comparative study of unit root tests with panel data and a new simple test. Oxford Bull Econ Stat 1999;108:1–24. [19] Narayan PK, Wong P. A panel data analysis of the determinants of oil consumption: the case of Australia. Appl Energy 2009;86:2771–5. [20] Nel WP, van Zyl G. Defining limits: energy constrained economic growth. Appl Energy 2010;87:168–77. [21] Payne JE. On the dynamics of energy consumption and output in the US. Appl Energy 2009;86:575–7. [22] Payne JE. A survey of the electricity consumption-growth literature. Appl Energy 2010;87:723–31. [23] Payne JE. Survey of the international evidence on the causal relationship between energy consumption and growth. J Econ Stud, in press. [24] Payne JE. In press. US disaggregate fossil fuel consumption and real GDP: an empirical note. Energy Source, Part B: Econ, Plann Policy. [25] Pedroni P. Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bull Econ Stat 1999;61:653–70. [26] Pedroni P. Fully modified OLS for heterogeneous cointegrated panels. Adv Economet 2000;15:93–130. [27] Pedroni P. Panel cointegration: asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis: new results. Econ Theory 2004;20:597–627. [28] Pesaran HM, Shin Y, Smith RP. Pooled mean group estimation of dynamic heterogeneous panels. J Am Stat Assoc 1999;94:621–34. [29] Reynolds DB, Kolodziej M. Former soviet union oil production and GDP decline: Granger-causality and the multi-cycle Hubbert curve. Energy Econ 2008;30:271–89. [30] Sari R, Soytas U. Disaggregate energy consumption, employment, and income in Turkey. Energy Econ 2004;26:335–44. [31] Sari R, Soytas U. Energy consumption, economic growth, and carbon emissions: challenges faced by an EU candidate member. Ecol Econ 2009;68:1667–75. [32] Sari R, Soytas U. Are global warming and economic growth compatible? evidence from five OPEC countries. Appl Energy 2009;86:1887–93. [33] Sari RBT, Ewing, Soytas U. The relationship between disaggregate energy consumption and industrial production in the united states: an ARDL approach. Energy Econ 2008;30:2302–13. [34] Yang HY. Coal consumption and economic growth in Taiwan. Energy Source 2000;22:109–15.
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977 [35] Yang HY. A note on the causal relationship between energy and GDP in Taiwanâ€. Energy Econ 2000;22:309–17. [36] Yoo SH. Causal relationship between coal consumption and economic growth in Korea. Appl Energy 2006;83:1181–9. [37] Yuan JH, Kang JG, Zhao CH, Hu ZG. Energy consumption and economic growth: evidence from China at both aggregated and disaggregated levels. Energy Econ 2009;30:3077–94.
1977
[38] Wolde-Rufael Y. Disaggregated industrial energy consumption and GDP: the case of Shanghai, 1952–1999. Energy Econ 2004;26:69–75. [39] Wolde-Rufael Y. Coal consumption and economic growth revisited. Appl Energy 2010;87:160–7. [40] Ziramba E. Disaggregate energy consumption and industrial production in South Africa. Energy Policy 2009;37:2214–20.
Contents lists available at ScienceDirect
Applied Energy journal homepage: www.elsevier.com/locate/apenergy
The causal dynamics between coal consumption and growth: Evidence from emerging market economies Nicholas Apergis a, James E. Payne b,* a b
Department of Banking and Financial Management, University of Piraeus Karaoli and Dimitriou 80 Piraeus, Attiki 18534, Greece College of Arts and Sciences, Illinois State University, Normal, IL 61790-4100, United States
a r t i c l e
i n f o
Article history: Received 10 September 2009 Received in revised form 29 November 2009 Accepted 30 November 2009 Available online 13 January 2010 Keywords: Coal consumption Growth Panel Granger-causality
a b s t r a c t This study examines the relationship between coal consumption and economic growth for 15 emerging market economies within a multivariate panel framework over the period 1980–2006. The heterogeneous panel cointegration results indicate there is a long-run equilibrium relationship between real GDP, coal consumption, real gross fixed capital formation, and the labor force. While in the long-run both real gross fixed capital formation and the labor force have a significant positive impact on real GDP, coal consumption has a significant negative impact. The panel causality tests show bidirectional causality between coal consumption and economic growth in both the short- and long-run. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction Though coal has been a dependable energy source within the world’s energy consumption mix, the environmental consequences of the sustained use of coal has drawn into question the long-term viability of coal in light of the emergence of cleaner, alternative energy sources. Researchers have recently begun to examine the dynamic causal relationship between coal consumption and economic growth and the energy policy implications of their findings.1 The strong economic growth and the rising energy demands of emerging market economies motivate the study of the coal consumption-growth nexus for these countries. Specifically, this study extends the previous research in an examination of a balanced panel of 15 emerging market economies as defined by Morgan Stanley capital international (MSCI) over the period 1980–2006 in the estimation of a panel error correction model to infer the causal relationship between coal consumption and economic growth.2 Table 1 displays some statistics related to the production and consumption of coal along with carbon dioxide emissions attrib-
* Corresponding author. Tel.: +1 309 438 5669. E-mail addresses: [email protected] (N. Apergis), [email protected] (J.E. Payne). 1 See Payne [21–23], for surveys of the literature on the causal relationship between energy consumption and economic growth. Nel and van Zyl [20] discuss energy-based growth models and through model calibration and forecasting to infer the prospects for energy-constrained economic growth. 2 Morgan Stanley Capital International characterizes twenty-two countries as emerging markets. Seven countries are omitted from the analysis due to the unavailability of data for all the variables over the entire period 1980–2008: Colombia, Israel, Pakistan, Poland, Russia, Taiwan, and Turkey. 0306-2619/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.apenergy.2009.11.035
uted to fossil fuels for the 15 emerging market economies. As the table shows, there is a great deal of variation in these statistics across countries. In terms of coal production, China ranks the highest while Morocco ranks the lowest. In the case of coal consumption, again China ranks the highest while Argentina is the lowest. With respect to carbon dioxide emissions, China has the highest rank whereas Peru has the lowest rank. China, India, and South Africa are among the world leaders in both coal production and consumption, not to mention in carbon dioxide emissions from fossil fuel usage. China ranks first in the world in the production and consumption of coal along with carbon dioxide emissions. India ranks third in both coal production and consumption and fourth in carbon dioxide emissions. South Africa ranks sixth in coal production, seventh in coal consumption, and 12th in carbon dioxide emissions. Further examination of Table 1 reveals, though Indonesia is eighth in coal production, it ranks 26th in coal consumption. In fact, Indonesia was the second largest net exporter of coal in the world for 2004.3 Based on the data reported in Table 1, 12 of the 15 countries are net importers of coal. The study extends the literature on the causal relationship between coal consumption and economic growth along several dimensions. First, the study includes a larger set of countries in the analysis than previous studies. Second, with the exception of the studies by Yuan et al. [37], Payne [24], Wolde-Rufael [39], and Apergis and Payne [4], the analysis will be undertaken within a production model framework by including measures for capital
3
Indonesia’s Country Analysis Brief, Energy Information Administration.
1973
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977 Table 1 Coal summary statistics, 2006. Country
Argentina Brazil Chile China Egypt Hungary India Indonesia Malaysia Mexico Morocco Peru Philippines South Africa Thailand
Emerging market economies Coal production
Coal consumption
Carbon dioxide emissions
0.051 (58) 6.490 (32) 0.437 (46) 2620.498 (1) 0.028 (62) 10.970 (28) 500.193 (3) 213.174 (8) 1.164 (44) 12.662 (26) 0.000 (66) 0.118 (53) 2.597 (40) 269.828 (6) 21.022 (22)
0.906 (66) 23.604 (27) 6.445 (47) 2584.246 (1) 1.425 (60) 13.059 (35) 539.486 (3) 24.071 (26) 16.868 (31) 19.876 (29) 6.478 (46) 1.627 (59) 11.117 (37) 195.225 (7) 33.156 (24)
162.19 (28) 377.24 (17) 64.80 (50) 6017.69 (1) 151.62 (30) 58.65 (54) 1293.17 (4) 280.36 (22) 163.53 (27) 435.60 (13) 34.53 (73) 29.93 (74) 72.39 (47) 443.58 (12) 245.04 (24)
Notes: Coal production and consumption denoted in million short tons. Carbon dioxide emissions defined as total from consumption of fossil fuels in million metric tons of CO2. Numbers in parentheses represent world rank. Data obtained from Country Energy Profiles of the Energy Information Administration.
and labor. Third, noting the studies by Hu and Lin [12], Sari et al. [33], Wolde-Rufael [39], and Apergis and Payne [4], the sign and magnitude of the respective coefficient estimates from the panel error correction model will be presented to facilitate the policy interpretation of the findings. Fourth, given the relatively short time horizon of the data, panel unit root and cointegration tests are utilized to provide additional power and size properties over standard unit root and cointegration tests by combining the cross-section and time series data across countries.4 Section 2 outlines the policy implications associated with the causal relationship between coal consumption and economic growth along with a summary of the empirical literature to date. Section 3 discusses the data, methodology, and empirical results. Section 4 provides concluding remarks.
2. Overview of the coal consumption and economic growth literature The identification of the causal relationship between coal consumption and economic growth has a number of policy implications. If an increase in coal consumption causes an increase in economic growth, then energy conservation policies that adversely impact coal consumption may also have an adverse impact on economic growth. On the other hand, if an increase in coal consumption causes a decrease in economic growth due perhaps to the inefficient and excessive use of coal, then energy conservation policies that reduce coal consumption may actually mitigate the adverse impact of coal consumption on economic growth.5 Alternatively, economic growth may cause either an increase or decrease in coal consumption. If an increase in economic growth causes an increase in coal consumption, then energy conservation policies that reduce coal consumption may not have an adverse impact on economic growth. However, the case of an increase in economic growth that causes a decrease in coal consumption may be reflective of an economy that is becoming less coal intensive.6 It is also possible that coal consumption and economic growth 4
The panel error correction methodology employed in this study parallels Apergis and Payne [1–4]. 5 Wolde-Rufael [39] alludes to the possibility that industries dependent on coal may have become less efficient over time. Also, the absence of binding legislative restrictions on carbon dioxide emissions may be a contributing factor to the excessive use of coal. 6 As pointed out by Wolde-Rufael [39], coal as an energy source may be decreasing relative to other energy sources in the production of electricity for example.
exhibit an interdependent relationship as causation may run in both directions. Likewise, it is conceivable that coal consumption plays such a minor role in the economic growth process that no causal relationship can be detected. Previous studies provide a range of results for a relatively small number of countries on the causal relationship between coal consumption and economic growth.7 Yang [34] uses the Engle–Granger bivariate cointegration procedure to find the absence of a long-run equilibrium relationship between coal consumption and real output in Taiwan; however, Granger-causality tests reveal unidirectional causality from economic growth to coal consumption. In a followup study on Taiwan, Yang [35] employs the same methodology to show bidirectional causality between coal consumption and economic growth. Fatai et al. [9] reveal unidirectional causality from economic growth to coal consumption in the case of Australia with both the Johansen–Juselius and Toda–Yamamoto approaches to causality testing; however, the absence of a causal relationship using the autoregressive distributed lag (ARDL) model. Furthermore, Fatai et al. [9] fail to find a causal relationship between coal consumption and economic growth for New Zealand using either the Johansen– Juselius or Toda–Yamamoto procedures. Using the Toda–Yamamoto approach to causality testing, Wolde-Rufael [38] presents evidence of unidirectional causality from coal consumption to real output in the case of Shanghai. In another study of Taiwan, Lee and Chang [16] find bidirectional causality between coal consumption and economic growth within a bivariate error correction model. Yoo [36] also reports bidirectional causality between coal consumption and economic growth from a bivariate error correction model in the case of South Korea. Hu and Lin [12] utilize the Hansen-Seo asymmetric cointegraton procedure to reveal asymmetries in the relationship between coal consumption and economic growth for Taiwan along with bidirectional causality. Jinke et al. [14] use a bivariate error correction model to show unidirectional causality from economic growth to coal consumption for China and Japan, but the absence of a causal relationship for India, South Africa, and South Korea. 7 A summary of the coal consumption-growth studies is provided in Table 1 of Apergis and Payne [4]. Though not explicitly testing for Granger-causality both Sari and Soytas [30] and Ewing et al. [8] employ generalized forecast error variance decomposition analysis. In the case of Turkey, Sari and Soytas [30] show that coal consumption explains up to 8% of the forecast error variance of real GDP. In a study of the US, Ewing et al. [8] find that coal consumption explains up to 10% of the forecast error variance of industrial production. In an examination of coal production in the former Soviet Union, Reynolds and Kolodziej [29] find unidirectional causality from economic growth to coal production.
1974
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977
Sari et al. [33] use the autoregressive distributed lag (ARDL) model to find unidirectional causality from coal consumption to industrial production for the US. In the case of China, Yuan et al. [37] estimate a multivariate vector error correction model to report bidirectional causality between coal consumption and economic growth. In a study of US fossil fuel consumption, Payne [24] fails to find a causal relationship between coal consumption and real output using the Toda–Yamamoto procedure for causality testing. In a multi-country study, Wolde-Rufael [39] finds through the Toda–Yamamoto procedure unidirectional causality from coal consumption to real output for India and Japan; unidirectional causality from real output to coal consumption for China and South Korea; and bidirectional causality for South Africa and the US. In another study of South Africa, Ziramba [40] fails to find a causal relationship between coal consumption and industrial production; however, the evidence does suggest unidirectional causality from coal consumption to employment. Finally, in a panel study of 25 OECD countries, Apergis and Payne [4] find bidirectional causality between coal consumption and economic growth. The next section describes the data, panel unit root and cointegration tests along with the results of a multivariate panel error correction model.8
3. Data, methodology, and results Annual data from 1980 to 2006 were obtained from the World Bank Development Indicators, CD-ROM and the Energy Information Administration for Argentina, Brazil, Chile, China, Egypt, Hungary, India, Indonesia, Malaysia, Mexico, Morocco, Peru, Philippines, South Africa, and Thailand. The multivariate framework includes real GDP (Y) in billions of constant 2000 US dollars, real gross fixed capital formation (K) in billions of constant 2000 US dollars, total labor force (L) in millions, and coal consumption (CC) defined in thousands of metric tons.9 Before proceeding to panel cointegration tests and specification of the panel error correction model, a battery of panel unit root tests are undertaken to infer the stationarity properties of the variables. Levin et al. [17] propose a panel based ADF test that assumes homogeneity in the dynamics of the autoregressive coefficients for all panel units. On the other hand, the Im et al. [13] panel unit root test allows for heterogeneity in the dynamics of the autoregressive coefficients for all panel units. Following Maddala and Wu [18], the nonparametric panel unit root tests are estimated using the FisherADF and Fisher-PP tests. Finally, the Carrion-i-Silvestre et al. [6] heterogeneous panel stationarity test is also examined. Under the Levin et al. [17], Im et al. [13], Fisher-ADF and Fisher-PP tests the null hypothesis is a unit root while the alternative hypothesis is no unit root. On the other hand, the Carrion-i-Silvestre et al. [6] test assumes stationarity under the null hypothesis.10 Panel A of Table 2 reports the results of the panel unit root tests which indicate that each variable is integrated of order one.11 Given the variables are integrated of the same order, the Pedroni [25,27] heterogeneous panel cointegration test which allows 8 See Apergis and Payne [1–4] and citations therein, for additional studies on the use of panel cointegration and error correction modeling in the energy consumptiongrowth literature. 9 Capital is proxied by real gross fixed capital formation in that changes in investment closely align with changes in the capital stock under the assumption of a constant depreciation rate using the perpetual inventory method (see [31] and citations therein as well as [4,39]. 10 Tests of dynamic heterogeneity (i.e. variation of the intercept over countries and time) of the variables in the model were performed following the methodology of Holtz-Eakin et al. [11] and Holtz-Eakin [10]. The results which are available upon request indicate that the relationships exhibit heterogeneity in both the dynamics and error variances across the 15 countries. 11 To conserve space the details of the panel unit root tests are not discussed.
Table 2 Panel unit root and cointegration tests emerging market economies, 1980–2006. Panel A: Panel unit root tests LLC
IPS
Fisher-ADF
Fisher PP
CLB
Y 0.72 0.82 15.38 17.64 19.38b DY 6.47a 7.85a 89.05a 133.89a 2.36 CC 0.56 0.65 20.47 19.21 24.33b a a a a DCC 5.89 6.09 71.28 86.39 1.51 K 0.72 0.78 15.44 25.49 10.79b DK 7.12a 7.53a 90.36a 113.79a 1.32 L 0.97 0.60 20.22 24.38 22.49b DL 8.49a 8.64a 89.45a 146.50a 2.56 Notes: Notation for tests: Levin et al. [17] LLC; Im et al. [13] IPS; Maddala and Wu [18] Fisher-ADF and Fisher PP; Carrion-i-Silvestre et al. [6] CLB. Significance levels at the 1% and 5% are denoted by ‘‘a” and ‘‘b”, respectively. Panel B: Panel cointegration tests Within dimension test statistics Between dimension test statistics 57.35006a Group q-statistic 50.57044a Panel v-statistic Panel q-statistic 50.52008a Group PP-statistic 54.37261a Panel PP55.09101a Group ADF-statistic 4.64055a statistic Panel ADF4.66656a statistic Notes: Of the seven tests, the panel v-statistic is a one-sided test where large positive values reject the null hypothesis of no cointegration whereas large negative values for the remaining test statistics reject the null hypothesis of no cointegration. Critical value at the 1% significance level denoted by ‘‘a”: panel v (24.56), panel q (17.60), panel and group PP (25.59), panel ADF (2.97), group q (21.12), and group ADF (3.18).
for cross-section interdependence with different individual effects is estimated to determine whether a long-run equilibrium relationship exists:
Y it ¼ ai þ di t þ b1i CC it þ b2i K it þ b3i Lit þ eit
ð1Þ
where i = 1, . . . , N for each country in the panel and t = 1, . . . , T refers to the time period. The parameters ai and di allow for the possibility of country-specific fixed effects and deterministic trends, respectively. eit denote the estimated residuals which represent deviations from the long-run relationship. All variables are expressed in natural logarithms so the b parameters of the model can be interpreted as elasticity estimates. The null hypothesis of no cointegration, qi = 1, is tested by conducting the following unit root test on the residuals as follows:
eit ¼ qi eit1 þ wit
ð2Þ
Pedroni [25,27] proposes two types of tests for cointegration: panel and group tests. The panel tests are based on the within dimension approach which includes four statistics: panel v, panel q, panel PP, and panel ADF-statistics. These statistics essentially pool the autoregressive coefficients across different countries for the unit root tests on the estimated residuals. These statistics take into account common time factors and heterogeneity across countries. The group tests are based on the between dimension approach which includes three statistics: group q, group PP, and group ADF-statistics. These statistics are based on averages of the individual autoregressive coefficients associated with the unit root tests of the residuals for each country in the panel. All seven tests are distributed asymptotically as standard normal.12 Panel B of Table 2 reports both the within and between dimension panel cointegration test statistics. All seven test statistics reject the null hypothesis of no cointegration at the 1% significance level. In light of the panel cointegration results, the fully modified OLS (FMOLS) technique for heterogeneous cointegrated panels 12 See Pedroni [25], for details on the heterogeneous panel and heterogeneous group mean panel cointegration statistics.
1975
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977 Table 3 FMOLS long-run estimates and panel causality tests emerging market economies, 1980–2006. Panel A: FMOLS long-run estimates þ 0:219a K þ 0:247L Y ¼ 0:047a 0:251CC a a ð5:09Þ
ð5:86Þ
Adj. R2 ¼ 0:61
ð7:02Þ
LM ¼ 0:85
ð4:50Þ
RESET ¼ 1:53
½0:52
½0:27
HE ¼ 0:93 ½0:47
Notes: t-statistics and probability values are reported in parentheses and brackets, respectively. LM is the lagrange multiplier test for serial correlation. RESET is the misspecification test. HE is White’s heteroscedasticity test. Significance at the 1% level denoted by ‘‘a”. Panel B: Panel causality test Dependent variable
Sources of causation (independent variables) Short-run
Long-run
DY
DCC
(3a)
DY
40.01 [0.00]a
(3b)
DCC
40.08 [0.00]a
(0.208) [0.00]a
(3c)
DK
40.23 [0.00]a
(0.171) [0.00]a
(3d)
DL
0.09 [0.92]
DK (0.023) [0.00]a
DL
ECT
51.30 [0.00]a
(0.341) [0.00]a
50.33 [0.00]a
(0.687) [0.00]a
0.013 [0.01]a
30.09 [0.00]a
(0.192) [0.00]a
0.30 [0.58]
(0.002) [0.57]
0.024 [0.01]a
30.26 [0.00]a
(0.164) [0.00]a
0.038 [0.01]a
(0.005) [0.91]
51.01 (0.507) 0.02 (0.009) 49.11 (0.326) 0.017 [0.00]a [0.90] [0.89] [0.00]a [0.00]a [0.02]b [0.00]a Notes: Partial F-statistics reported with respect to short-run changes in the independent variables. 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% and 5% levels denoted by ‘‘a” and ‘b”, respectively.
is estimated to determine the long-run equilibrium relationship [26].13 The FMOLS results are displayed in Panel A of Table 3. The coefficients are positive and significant at the 1% level for real gross fixed capital formation and the labor force; however, the coefficient for coal consumption is negative and significant at the 1% level. As discussed by Kolstad and Krautkraemer [15], Sari and Soytas [32], and Apergis and Payne [4], the negative impact in the long-run of an increase in coal consumption on real GDP may be the result of the economic costs associated with the environmental impact of carbon dioxide emissions outweigh the immediate economic benefit of coal usage for real GDP.14 Given the coefficients can be interpreted as elasticity estimates, the results indicate that a 1% increase in coal consumption decreases real GDP by 0.251%; a 1% increase in real gross fixed capital formation increases real GDP by 0.291%; and a 1% increase in the labor force increases real GDP by 0.247%. As a comparison, Apergis and Payne [4] find that a 1% increase in coal consumption decreases real GDP by 0.179%; a 1% increase in real gross fixed capital formation increases real GDP by 0.526%; and a 1% increase in the labor force increases real GDP by 0.285% for a panel of 25 OECD countries. Next, a panel vector error correction model is estimated [28]. The Engle and Granger [7] two-step procedure is employed by first estimating the long-run model specified in Eq. (1) to obtain the estimated residuals. The lagged residuals from Eq. (1) will serve as the error correction term in the estimation of the dynamic error correction model as follows:
13 See Narayan and Wong [19], for an example of the FMOLS estimation. The estimates from either the FMOLS or DOLS are asymptotically equivalent for more than 60 observations [5]. The panel data sets of this study contain 405 observations. 14 Sari and Soytas [32] provide several reasons associated with the negative impact of energy consumption on income in the long-run: (1) inefficient use of energy, (2) production of output to less energy intensive service sectors, and (3) improvement in the energy sector crowds out non-energy sectors, known as ‘‘Dutch disease” whereby the abundance of a natural resource has a negative impact on other tradable goods sectors. Though the first two points raised by Sari and Soytas [32] are mentioned in this study, the presence of the ‘‘Dutch disease” phenomenon may not be as relevant for this study as 12 of the 15 countries are net importers of coal.
q X
DY it ¼ n1j þ
q X
/11ik DY itk þ
k¼1
þ
q X
/13ik DK itk þ
k¼1
q X
/14ik DLitk þ k1i eit1 þ u1it
q X
/21ik DY itk þ
q X
k¼1 q X
q X
q X
q X
q X
/31ik DY itk þ
/33ik DK itk þ
q X
q X
q X k¼1
/32ik DCC itk
/34ik DLitk þ k3i eit1 þ u3it
ð3cÞ
k¼1
/41ik DY itk þ
q X
k¼1
þ
ð3bÞ
k¼1
k¼1
DLit ¼ n4j þ
/24ik DLitk þ k2i eit1 þ u2it
k¼1
k¼1
þ
/22ik DCC itk
k¼1
/23ik DK itk þ
k¼1
DK it ¼ n3j þ
ð3aÞ
k¼1
DCC it ¼ n2j þ þ
/12ik DCC itk
k¼1
/43ik DK itk þ
/42ik DCC itk
k¼1 q X
/44ik DLitk þ k4i eit1 þ u4it
ð3dÞ
k¼1
where D is the first-difference operator; k is the lag length set at one based on likelihood ratio tests; and u is the serially uncorrelated error term. With respect to Eqs. (3a)–(3d), short-run causality is determined by the statistical significance of the partial F-statistic associated with the corresponding right hand side variables. Longrun causality is revealed by the statistical significance of the respective error correction terms using a t-test. Panel B of Table 3 reports the results of the short-run and longrun Granger-causality tests. In Eq. (3a), real gross fixed capital formation and the labor force each have a significant positive impact on economic growth in the short-run while coal consumption has a significant negative impact. The negative impact of coal consumption on economic growth may be attributed to an inefficient and excessive use of coal as noted by Wolde-Rufael [39] and Apergis and Payne [4]. With respect to Eq. (3b), both economic growth
1976
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977
and real gross fixed capital formation have a significant negative impact on coal consumption while the labor force has as insignificant impact in the short-run. Wolde-Rufael [39] and Apergis and Payne [4] suggest this negative impact on coal consumption may indicate that the overall energy consumption mix has become less coal intensive. In the case of Eq. (3c), both economic growth and the labor force each have a significant positive impact on real gross fixed capital formation while coal consumption has no impact. For Eq. (3d), both economic growth and real gross fixed capital formation have a significant positive impact on the labor force whereas coal consumption has an insignificant impact. A review of the short-run causality results indicates there is bidirectional causality between coal consumption and economic growth. The long-run dynamics displayed by the error correction terms from Eqs. (3a)–(3d) show that economic growth, coal consumption, real gross fixed capital formation, and the labor force each respond to deviations from long-run equilibrium as noted by the statistical significance of their respective error correction terms. However, given the magnitude of the coefficients on the error correction terms, the speed of adjustment towards long-run equilibrium is rather slow. In summary, the results suggest there is bidirectional causality between coal consumption and economic growth in both the short- and long-run. In the short-run, the coefficient estimates indicate that coal consumption has a negative impact on economic growth and likewise economic growth also has a negative impact on coal consumption. These results are similar to those reported by Wolde-Rufael [39] for the US and Apergis and Payne [4] for a panel of OECD countries.
4. Concluding remarks In light of the economic growth and rising energy demands of the emerging market economies, this study examines the role of coal consumption in the economic growth process for a panel of 15 emerging market economies. Specifically, the causal relationship between coal consumption and economic growth is evaluated using a multivariate panel error correction model over the period 1980–2006. Heterogeneous panel cointegration tests indicate there is a long-run equilibrium relationship between real GDP, coal consumption, real gross fixed capital formation, and the labor force. The long-run elasticity estimates are positive and significant for both real gross fixed capital formation and the labor force; however, the elasticity estimate with respect to coal consumption is negative and significant. As discussed by Wolde-Rufael [39] and Apergis and Payne [4] this negative impact of coal consumption on real output in the long-run may be the consequence of the economic costs of carbon dioxide emissions on the environment outweigh the immediate economic benefit associated with coal usage on real output. The estimation of the panel error correction model shows there is both short- and long-run bidirectional causality between coal consumption and economic growth. However, the short-run causality results indicate that an increase in coal consumption has a negative impact on economic growth, a result that could be attributed to an inefficient and excessive use of coal. In this case, the feasibility of either increasing the efficient use of coal or reducing coal consumption should be assessed in the formulation of energy policy. For example, legislation that would restrict the growth of carbon dioxide emissions might provide an incentive to enhance efficiency or curb excessive coal consumption. Furthermore, greater use of sustainable coal technologies that permit carbon dioxide capture and storage may curtail the environmental consequences of excessive coal consumption. Similarly, the short-run causality tests also indicate that an increase in economic growth has a negative impact on coal consumption. This
result suggests the emergence of alternative energy sources such as renewables may have rendered these economies less coal intensive over time. References [1] Apergis N, Payne JE. Energy consumption and economic growth in Central America: evidence from a panel cointegration and error correction model. Energy Econ 2009;31:211–6. [2] Apergis N, Payne JE. Energy consumption and economic growth: evidence from the commonwealth of independent states. Energy Econ 2009;31:641–7. [3] Apergis N, Payne JE. Renewable energy consumption and economic growth: evidence from a panel of OECD countries. Energy Policy 2010;38:656–60. [4] Apergis N, Payne, JE. Coal consumption and economic growth: evidence from a panel of OECD countries. Energy Policy, in press. [5] Banerjee A. Panel data unit roots and cointegration: an overview. Oxford Bull Econ Stat 1999;61(S1):607–29. [6] Carrion-i-Silvestre JL, del Barrio-Castro T, Lopez-Bazo E. Breaking the panels: an application to GDP per capita. Economet J 2005;8:159–75. [7] Engle RF, Granger CWJ. Cointegration and error correction: representation, estimation, and testing. Econometrica 1987;55:251–76. [8] Ewing BT, Sari R, Soytas U. Disaggregate energy consumption and industrial output in the United States. Energy Policy 2007;35:1274–81. [9] Fatai K, Oxley L, Scrimgeour FG. Modelling the causal relationship between energy consumption and GDP in New Zealand, Australia, India, Indonesia, the Philippines, and Thailand. Math Comput Simulat 2004;64:431–45. [10] Holtz-Eakin D. Testing for individual effects in dynamic models using panel data. NBER technical paper series; 1986. No. 57. [11] Holtz-Eakin D, Newey W, Rosen H. Implementing causality tests with panel data with an example from local public finance. NBER technical working paper; 1985. No. 48. [12] Hu JL, Lin CH. Disaggregated energy consumption and GDP in Taiwan: a threshold co-integration analysis. Energy Econ 2008;30:2342–58. [13] Im KS, Pesaran MH, Shin Y. Testing for unit roots in heterogeneous panels. J Econ 2003;115:53–74. [14] Jinke L, Hualing S, Dianming G. Causality relationship between coal consumption and GDP: difference of major OECD and non-OECD countries. Appl Energy 2008;85:421–9. [15] Kolstad CD, Krautkraemer JA. Natural resources use and the environment. In: Kneese AV, Sweeney JL, editors. Handbook of natural resources and energy economics, vol. 3. p. 1219–65. [16] Lee CC, Chang CP. Structural breaks, energy consumption and economic growth revisited: evidence from Taiwan. Energy Econ 2005;27:857–72. [17] Levin A, Lin CF, Chu C. Unit root tests in panel data: asymptotic and finitesample properties. J Econ 2002;108:1–24. [18] Maddala GS, Wu SA. Comparative study of unit root tests with panel data and a new simple test. Oxford Bull Econ Stat 1999;108:1–24. [19] Narayan PK, Wong P. A panel data analysis of the determinants of oil consumption: the case of Australia. Appl Energy 2009;86:2771–5. [20] Nel WP, van Zyl G. Defining limits: energy constrained economic growth. Appl Energy 2010;87:168–77. [21] Payne JE. On the dynamics of energy consumption and output in the US. Appl Energy 2009;86:575–7. [22] Payne JE. A survey of the electricity consumption-growth literature. Appl Energy 2010;87:723–31. [23] Payne JE. Survey of the international evidence on the causal relationship between energy consumption and growth. J Econ Stud, in press. [24] Payne JE. In press. US disaggregate fossil fuel consumption and real GDP: an empirical note. Energy Source, Part B: Econ, Plann Policy. [25] Pedroni P. Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bull Econ Stat 1999;61:653–70. [26] Pedroni P. Fully modified OLS for heterogeneous cointegrated panels. Adv Economet 2000;15:93–130. [27] Pedroni P. Panel cointegration: asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis: new results. Econ Theory 2004;20:597–627. [28] Pesaran HM, Shin Y, Smith RP. Pooled mean group estimation of dynamic heterogeneous panels. J Am Stat Assoc 1999;94:621–34. [29] Reynolds DB, Kolodziej M. Former soviet union oil production and GDP decline: Granger-causality and the multi-cycle Hubbert curve. Energy Econ 2008;30:271–89. [30] Sari R, Soytas U. Disaggregate energy consumption, employment, and income in Turkey. Energy Econ 2004;26:335–44. [31] Sari R, Soytas U. Energy consumption, economic growth, and carbon emissions: challenges faced by an EU candidate member. Ecol Econ 2009;68:1667–75. [32] Sari R, Soytas U. Are global warming and economic growth compatible? evidence from five OPEC countries. Appl Energy 2009;86:1887–93. [33] Sari RBT, Ewing, Soytas U. The relationship between disaggregate energy consumption and industrial production in the united states: an ARDL approach. Energy Econ 2008;30:2302–13. [34] Yang HY. Coal consumption and economic growth in Taiwan. Energy Source 2000;22:109–15.
N. Apergis, J.E. Payne / Applied Energy 87 (2010) 1972–1977 [35] Yang HY. A note on the causal relationship between energy and GDP in Taiwanâ€. Energy Econ 2000;22:309–17. [36] Yoo SH. Causal relationship between coal consumption and economic growth in Korea. Appl Energy 2006;83:1181–9. [37] Yuan JH, Kang JG, Zhao CH, Hu ZG. Energy consumption and economic growth: evidence from China at both aggregated and disaggregated levels. Energy Econ 2009;30:3077–94.
1977
[38] Wolde-Rufael Y. Disaggregated industrial energy consumption and GDP: the case of Shanghai, 1952–1999. Energy Econ 2004;26:69–75. [39] Wolde-Rufael Y. Coal consumption and economic growth revisited. Appl Energy 2010;87:160–7. [40] Ziramba E. Disaggregate energy consumption and industrial production in South Africa. Energy Policy 2009;37:2214–20.