Is per capita energy use stationary? Panel data evidence for the EMU countries

Abstract

The aim of the study is to examine the stationary properties of per capita energy use in the 19 Eurozone member countries by using yearly data in the 1960–2013 period. First and second generation of panel unit root tests are applied. Empirical findings show mixed results, since the evidence favors the panel stationarity of energy use per capita in all groups of countries for the specification with the constant only; on the contrary, in the specification with constant and trend, the null hypothesis that all series are non-stationary cannot be rejected. Therefore, policymakers should be cautious, paying attention to the energy use series’ properties, as different results conduct to opposite policy implications.

Keywords

Energy use EMU cross-section dependence unit root panel data

Introduction

The stationarity and unit root properties of per capita energy or electricity series are explored by an increasing amount of studies, which use different methodologies. Examining the time series properties of energy series is crucial both for researchers and for the policymakers, given the close link between energy and the real economy. In fact, if energy exhibits the presence of a unit root, this suggests that this series does not revert to its equilibrium level after being hit by a shock (Kula et al., 2012). Thus, if energy is a non-stationary process, then any shock to energy is likely to be permanent (Chen and Lee, 2007). On the contrary, if per capita energy is a stationary process, then the effect of the shock is transitory, and it is possible to forecast future movements in energy based on the past behaviors of the series. The series will be consistent and stable with path dependency. Path dependency of energy implies that world energy markets innovation will have permanent impacts. There is a large literature examining the causal relationship between energy and real output. Therefore, testing whether the energy use has a unit root is essential to any effective and sustainable energy policy. In the bargain, the issue of whether energy use is stationary has important implications for modelling. The findings from these studies have important policy implications (Mishra et al., 2009).

In addition, in this literature, an interesting aspect is related to the presence of cross-section dependence. In fact, only the more recent studies deal with it. Yet, taking cross-section dependence into account can completely change the tests results, conducting to opposite policy implications.

Moreover, to our knowledge, this is the first study that analyzes the topic of stationarity of energy series in the case of Eurozone member states in a panel context. In early 2007, the Commission put forth its ambitious energy-climate package, which unambiguously illustrated the role of the EU in leading the effort to create a climate-compatible energy system. The “20/20” package included the goals for 2020 of reducing GHG emissions by 20%, improving energy efficiency by 20%, achieving a 20% share of renewable energy, and a 10% share of biofuels (Eurostat, 2009).

Besides the “Introduction” section, the remainder of this article is organized as follows: “Literature review” section show a concise survey of the literature; “Samples selected and analytical methods” section presents the econometric strategy and data; and “Conclusions and policy implications” section gives the conclusions and policy implications.

Literature review

The topic of the stationarity properties of energy series is receiving a growing attention by applied analysts, because of its relevant implications on energy policies established by national authorities. Thus, testing for the unit root and cross-section dependence of energy variables emerged as a new branch of research in energy economics. Notwithstanding, empirical analyses for the Euro area member states are very few.

In Table 1, we sum up the applied studies that concern Eurozone countries. For an overview, see Magazzino (2014, 2015).

Table 1.

Comparison of empirical literature on stationarity of energy or electricity series in a panel data framework.

Author(s)	Country	Study period	Method(s)
Lee (2005)	18 developing countries	1975–2001	Panel unit root and cointegration tests
Al-Iriani (2006)	6 GCC countries	1970–2002	Panel unit root and cointegration tests
Chen and Lee (2007)	104 countries	1971–2002	Panel unit root tests with structural breaks
Chen et al. (2007)	10 Asian countries	1971–2001	Panel unit root and cointegration tests
Joyeux and Ripple (2007)	7 East Indian Ocean countries	1971–2001	Panel unit root and cointegration tests
Lee and Chang (2007)	40 countries	1965–2002	Panel unit root tests with structural breaks and GMM
Mahadevan and Asafu-Adjaye (2007)	20 countries	1971–2002	Panel unit root and cointegration tests
Mehrara (2007)	11 oil exporting countries	1971–2002	Panel unit root and cointegration tests
Narayan and Smyth (2007)	182 countries	1979–2000	Univariate and panel unit root tests
Hsu et al. (2008)	84 countries	1971–2003	Panel SURADF
Huang et al. (2008)	82 countries	1972–2002	GMM
Lee and Chang (2008)	16 Asian countries	1971–2002	Panel unit root and cointegration tests
Lee et al. (2008)	22 OECD countries	1960–2001	Panel unit root and cointegration tests
Narayan and Smyth (2008)	G-7 countries	1972–2002	Panel unit root and cointegration tests
Narayan et al. (2008)	60 countries	1971–2003	Panel unit root tests with and without a structural break
Apergis and Payne (2009a)	11 Commonwealth of Independent states	1991–2005	Panel unit root and cointegration tests
Apergis and Payne (2009b)	6 Central American countries	1980–2004	Panel unit root and cointegration tests
Mishra et al. (2009)	13 Pacific Island countries	1980–2005	Panel unit root tests with and without a structural break
Lee and Lee (2010)	25 OECD countries	1978–2004	Panel unit root and cointegration tests
Ozturk et al. (2010)	51 countries	1971–2005	Panel unit root and cointegration tests
Agnolucci and Venn (2011)	UK industrial subsectors	1970–2004	Univariate and panel unit root tests
Belke et al. (2011)	25 OECD countries	1981–2007	Panel unit root and cointegration tests
Narayan and Popp (2012)	93 countries	1980–2006	Panel unit root and cointegration tests
Shahbaz et al. (2012)	103 countries	1971–2010	First and second generation panel unit root tests

Source: Our elaborations.

GMM: Generalized Method of Moments; SURADF: Seemingly Unrelated Regressions Augmented Dickey-Fuller test.

Empirical analyses that concern also one or more Euro area member countries with a panel data approach are those of Lee and Chang (2007), Narayan and Smyth (2007), Hsu et al. (2008), Huang et al. (2008), Lee et al. (2008), Narayan and Smyth (2008), Lee and Lee (2010), Belke et al. (2011), Narayan and Popp (2012), and Shahbaz et al. (2012).

Samples selected and analytical methods

The econometric literature has proposed several unit root tests in a panel framework (Baltagi, 2013; Frees, 2004; Hsiao, 1989; Wooldridge, 2010).

The first generation of panel unit root tests here considered include Fisher-type test, IPS test (Im et al., 2003), and MW test (Maddala and Wu, 1999). The second-generation test is CADF (Pesaran, 2003). The main difference between two generations of tests lies in the cross-sectional independence assumption. In fact, first generation tests assume that all cross-sections are independent, whereas second-generation tests relax this assumption. Analogous to IPS test, Pesaran’s CADF is consistent under the alternative hypothesis that only a fraction of the series are stationary. Pesaran (2007) suggested modified IPS statistics based on the average of individual CADF, which is denoted as a cross-sectional augmented IPS (CIPS). In addition, second-generation tests are more useful when co-movements are observed in the national business cycles in a sample of countries in the same economic area (Hurlin and Dumitrescu, 2012). Cross-section dependence in macro panel data may arise from globally common shocks with heterogeneous impact across countries. Alternatively, it can be the result of local spillover effects between countries or regions (Eberhardt and Teal, 2011; Moscone and Tosetti, 2009). In this study, we employed several tests. The Pesaran (2004) CD-test employs the correlation-coefficients between the time-series for each panel member. The test is robust to nonstationarity, parameter heterogeneity, or structural breaks, and was shown to perform well even in small samples (Pesaran, 2006). Moreover, we present results of semiparametric Friedman (1937) test, as well as the parametric testing procedure proposed by Pesaran (2004), which test the hypothesis of cross-sectional independence in panel-data models with small T and large N.

The energy use series are derived from the World Development Indicators (WDI) database by World Bank. The 19 Eurozone member countries considered in this study, for the 1960–2013 period, are Austria, Belgium, Cyprus, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Luxembourg, Luxembourg, Malta, Netherlands, Portugal, Slovakia, Slovenia, and Spain.

Figure 1 gives an initial graphical inspection of these series, showing the evolution of selected data: an upward trend seems to emerge.

Figure 1.

Energy use in the EMU countries (1960–2013, kilograms of oil equivalent per capita).

Results and discussion

Descriptive statistics are given in Table 2. In general, energy use has a negative value of skewness, indicating that the distribution is left-skewed, with more observations on the right. Moreover, the values of mean, median, and 10% trim mean are close, as well as those of Standard Deviation and Pseudo Standard Deviation, suggesting a normal distribution. While the Inter-Quartile Range (IQR) analysis shows the absence of outliers in our sample.

Table 2.

Exploratory data analysis.

Area	Mean	Median	St. Dev.	Skewness	Kurtosis	IQR	10-Trim
Eurozone	7.9468	8.0135	0.6254	−0.5863	4.1168	0.6914	7.978

Source: Our calculations on WB data.

Table 3 shows the results of first-generation panel unit root tests (IPS and Fisher-type). The level models have been specified without subtracting the cross-sectional averages from the series, while the Hannan–Quinn information criterion is used to determine the number of lags used to remove higher order autoregressive components of the series. Both tests reject the null hypothesis that all the panels contain a unit root, in favor of the alternative that at least one series in the panel is stationary.

Table 3.

First generation panel unit root tests.

Im, Pesaran and Shin (IPS) test
Area	PCEU
Eurozone	−9.4888 (0.0000)
Fisher-type test
Eurozone	208.1957 (0.0000)

For the IPS test, the W-t-bar statistic and the P-value are reported; for the Fisher-type test, the Inverse chi-squared statistic and, in parentheses, the P-value are reported. Panel unit root tests include the intercept.

More in detail, in Table 4, the results for Maddala and Wu (1999) as well as the Pesaran (2007) panel unit root tests for multiple variables and lags are reported. As one can note, specifications without trend suggest, for both tests, the rejection of the hypothesis of nonstationarity in all panel members (H₀). On the contrary, specifications with trend are both in favor of the alternative hypothesis of stationarity (H₁).

Table 4.

MW and Pesaran’s CIPS panel unit root tests.

MW: specification without trend
Area	Lags	PCEU
Eurozone	Lag 0	225.624 (0.000)
	Lag 1	197.581 (0.000)
	Lag 2	114.813 (0.000)
MW: specification with trend
Area	Lags	PCEU
Eurozone	Lag 0	33.423 (0.681)
	Lag 1	80.838 (0.000)
	Lag 2	30.358 (0.807)
CIPS: specification without trend
Area	Lags	PCEU
Eurozone	Lag 0	−3.140 (0.001)
	Lag 1	−3.434 (0.000)
	Lag 2	−3.257 (0.001)
CIPS: specification with trend
Area	Lags	PCEU
Eurozone	Lag 0	−0.072 (0.471)
	Lag 1	−0.050 (0.480)
	Lag 2	0.479 (0.684)

Critical values at the 5% significance level in parentheses. For Maddala and Wu test, the χ² statistics and, in parentheses, the P-values are reported; for Pesaran CIPS test, the Z-t-bar statistics and, in parentheses, the P-values. MW test assumes cross-section independence. CIPS test assumes cross-section dependence is in form of a single unobserved common factor. Panel unit root tests include the intercept.

MW: Maddala and Wu; CIPS: cross-sectional augmented IPS.

Nevertheless, a standard assumption in panel data models is that the error terms are independent across cross-sections. Empirical results in Table 5 show that, at a 5% significance level, the null hypothesis of cross-sectional independence in our panel should be rejected for our series.

Table 5.

Panel cross-section dependence tests.

Area	1	2	3
Eurozone	22.173 (0.0000)	31.47 (0.0000)	98.191 (0.0000)

: Pesaran (2004) cross-sectional dependence in panel data models test; 2: Pesaran (2004) CD test for cross-section dependence in panel time-series data; 3: Friedman (1937) test for cross-sectional dependence by using Friedman’s χ² distributed statistic. P-values in parentheses. Tests include the intercept.

To eliminate this type of dependence, the standard DF (or ADF) regressions are augmented with the cross-section averages of lagged levels and first-differences of the individual series (CADF statistics). Controlling for cross-section dependence, previous results are partially confirmed (Table 6). In fact, the null hypothesis that all series are non-stationary holds only for the specification with constant and trend, in line with results MW and Pesaran’s CIPS tests.

Table 6.

Panel unit root test in presence of cross section dependence tests.

Pesaran’s CADF test
Area	PCEU
Area	Constant	Constant and trend
Eurozone	−3.257 (0.001)	0.479 (0.684)

Z-t-bar or t-bar statistics; P-values in parentheses.

Therefore, after we take the cross-sectional correlations into account, our results substantiate that the panel datasets for the 19 country-based panels are non-stationary. In fact, in the specification with constant and trend, the null that all series are non-stationary cannot be rejected, which is a clear sign of the non-stationary properties of the series; otherwise, it means that any shock may have a transitory effect on energy use. If most shocks to energy use were temporary, then a stabilization energy policy would have long-lasting effects in the selected panels according to our results. When energy use temporarily deviates from the mean value, the administrative policy of the government should not be to adopt unnecessary targets.

However, and this is relevant to note, when we consider the specification with the constant only, the evidence favors the panel stationarity of energy use per capita in all groups of countries. The fact that energy use shows stationarity indicates that it should be possible for the series to forecast future movements in energy use based on past behavior. Other macroeconomic variables linked to energy demand via flow-on effects, such as income, will not inherit that non-stationarity and transmit it to major economic variables, such as employment (Narayan and Smyth, 2007).

Conclusions and policy implications

In this study, we analyzed the properties of per capita energy use in 19 Eurozone member countries, using annual data for the 1960–2013 period. As regards the empirical strategy, first and second generation of panel unit root tests have been applied. As discussed above, we found mixed results. In fact, for the specification with constant and trend, we reach the conclusion that the series are non-stationary, meaning that any shock may have a transitory effect on energy use. On the other hand, the alternative specification with the constant only is in favor of stationarity, indicating that it should be possible for the series to forecast future movements in energy use based on past behavior.

Important policy implications emerge from our empirical results. Sustainable energy policies rely heavily on the forecasts of energy. In this regard, determining whether shocks to energy are permanent or transitory is important for setting feasible goals for sustainable energy policies. When energy use is non-stationary, then past behavior is of no value in forecasting future demand and one would need to look at other variables explaining energy use to generate forecasts of energy demand into the future. In fact, if shocks to energy use have permanent effects, such shocks will be transmitted to the other sectors of an economy, making energy policies based on forecasts invalid. Thus, for the policymakers, it is necessary to pay attention to energy use series.

Suggestions for future researches

Given the little amount of studies devoted to the analysis of the properties of energy series in the Euro area countries, it should be investigated the same properties with a time series approach. In fact, a limitation of most existing studies of the stationarity properties of energy, however, is that they do not adequately address the problem of unit root and/or allow for structural breaks in the data. Moreover, stationarity and cross-section dependence should be explored also for disaggregated data (i.e. different sources of energy) (Magazzino, 2012).

Footnotes

Acknowledgments

The author would like to thank two anonymous referees and the Editor for their valuable suggestions and helpful comments, which have greatly enhanced the quality of this article. Any remaining errors, of course, belong to the author.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Agnolucci

Venn

(2011) Industrial energy intensities in the UK: is there a deterministic or stochastic difference among sectors. Applied Economics 43(12): 1447–1462.

Al-Iriani

(2006) Energy-GDP relationship revisited: an example from GCC countries using panel causality. Energy Policy 34(17): 3342–3350.

Apergis

Payne

(2009a) Energy consumption and economic growth: Evidence from the Commonwealth of Independent States. Energy Economics 31(56): 641–647.

Apergis

Payne

(2009b) Energy consumption and economic growth in Central America: Evidence from a panel cointegration and error correction model. Energy Economics 31(2): 211–216.

Baltagi

(2013) Econometric Analysis of Panel Data, Chichester: Wiley, pp. 287.

Belke

Dobnik

Dreger

(2011) Energy consumption and economic growth: New insights into the cointegration relationship. Energy Economics 33(5): 782–789.

Chen

Lee

(2007) Is energy consumption per capita broken stationary? New evidence from regional-based panels. Energy Policy 35(6): 3526–3540.

Chen

Kuo

Chen

(2007) The relationship between GDP and electricity consumption in 10 Asian countries. Energy Policy 35(4): 2611–2621.

Eberhardt

Teal

(2011) Econometrics for grumblers: A new look at the literature on cross-country growth empirics. Journal of Economic Surveys 25(1): 109–155.

10.

Eurostat (2009) Panorama of Energy, Luxembourg: European Communities, pp. 252.

11.

Frees

(2004) Longitudinal and Panel Data: Analysis and Applications in the Social Sciences, Cambridge: Cambridge University Press, pp. 162.

12.

Friedman

(1937) The use of ranks to avoid the assumption of normality implicit in the analysis of variance. Journal of the American Statistical Association 32(200): 675–701.

13.

Hsiao

(1989) Analysis of Panel Data, Cambridge: Cambridge University Press, pp. 187.

14.

Hsu

Lee

(2008) Revisited: Are shocks to energy consumption permanent or temporary? New evidence from a panel SURADF approach. Energy Economics 30(5): 2314–2330.

15.

Huang

Hwang

Yang

(2008) Causal relationship between energy consumption and GDP growth revisited: A dynamic panel data approach. Ecological Economics 67(1): 41–54.

16.

Hurlin

Dumitrescu

(2012) Testing for Granger non-causality in heterogeneous panels. Economic Modelling 29(4): 1450–1460.

17.

Pesaran

Shin

(2003) Testing for unit roots in heterogeneous panels. Journal of Econometrics 115(1): 53–74.

18.

Joyeux

Ripple

(2007) Household energy consumption versus income and relative standard of living: a panel approach. Energy Policy 35(1): 50–60.

19.

Kula F, Aslan A and Ozturk I (2012) Is per capita electricity consumption stationary? Time series evidence from OECD countries. Renewable and Sustainable Energy Reviews 16: 501--503.

20.

Lee

(2005) Energy consumption and GDP in developing countries: a cointegrated panel analysis. Energy Economics 27(3): 415–427.

21.

Lee

Chang

(2007) Energy consumption and GDP revisited: a panel analysis of developed and developing countries. Energy Economics 29(6): 1206–1223.

22.

Lee

Chang

(2008) Energy consumption and economic growth in Asian economies: a more comprehensive analysis using panel data. Resource and Energy Economics 30(1): 50–65.

23.

Lee

(2010) A panel data analysis of the demand for total energy and electricity in OECD countries. The Energy Journal 31(1): 1–24.

24.

Lee C, Chang C and Chen P (2008) Energy-income causality in OECD countries revisited: The key role of capital stock. Energy Economics 30(5): 2359--2373.

25.

Maddala

(1999) A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin of Economics and Statistics 61(S1): 631–652.

26.

Magazzino

(2012) On the relationship between disaggregated energy production and GDP in Italy. Energy & Environment 23(8): 1191–1207.

27.

Magazzino

(2014) Electricity demand, GDP and employment: Evidence from Italy. Frontiers in Energy 8(1): 31–40.

28.

Magazzino

(2015) Energy consumption and GDP in Italy: Cointegration and causality analysis. Environment, Development and Sustainability 17(1): 137–153.

29.

Mahadevan

Asafu-Adjaye

(2007) Energy consumption, economic growth and prices: A reassessment using panel VECM for developed and developing countries. Energy Policy 35(4): 2481–2490.

30.

Mehrara

(2007) Energy consumption and economic growth: The case of oil exporting countries. Energy Policy 35(5): 2939–2945.

31.

Mishra

Sharma

Smyth

(2009) Are fluctuations in energy consumption per capita transitory? Evidence from a panel of pacific island countries. Energy Policy 37(6): 2318–2326.

32.

Moscone

Tosetti

(2009) A review and comparison of tests of cross-section independence in panels. Journal of Economic Surveys 23(3): 528–561.

33.

Narayan

Popp

(2012) The energy consumption-real GDP nexus revisited: empirical evidence from 93 countries. Economic Modelling 29(2): 303–308.

34.

Narayan

Smyth

(2007) Are shocks to energy consumption permanent or temporary? Evidence from 182 countries. Energy Policy 35(1): 333–341.

35.

Narayan

Smyth

(2008) Energy consumption and real GDP in G7 countries: new evidence from panel cointegration with structural breaks. Energy Economics 30(5): 2331–2341.

36.

Narayan

Smyth

(2008) Are oil shocks permanent or temporary? Panel data evidence from crude oil and NGL production in 60 countries. Energy Economics 30(3): 919–936.

37.

Ozturk

Aslan

Kalyoncu

(2010) Energy consumption and economic growth relationship: Evidence from panel data for low and middle income countries. Energy Policy 38(8): 4422–4428.

38.

Pesaran

(2003) A simple panel unit root test in the presence of cross section dependence. Cambridge Working Papers in Economics 22(2): 265–312.

39.

Pesaran MH (2004) General diagnostic tests for cross section dependence in panels. University of Cambridge-IZA Discussion Paper, No. 1240.

40.

Pesaran

(2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74(4): 967–1012.

41.

Pesaran

(2007) A simple panel unit root test in the presence of cross-section dependence. Journal of Applied Econometrics 22(2): 265–312.

42.

Shahbaz

Tiwari

Khan

(2012) Is Energy Consumption Per Capita Stationary? Evidence from First and Second Generation Panel Unit Root Tests, MPRA Paper, University Library of Munich, Germany.

43.

Wooldridge

(2010) Econometric Analysis of Cross Section and Panel Data, Cambridge, MA: The MIT Press.