CO2 emissions and economic activity : heterogeneity across countries and non stationary series

CO2 emissions and economic activity: heterogeneity across countries and non stationary series Matías Piaggio Emilio Padilla 10.09 Facultat d'Economia i Empresa De p artament d'Economia A p licada Aquest document pertany al Departament d'Economia Aplicada. Data de publicació : Departament d'Economia Aplicada Edifici B Campus de Bellaterra 08193 Bellaterra Telèfon: (93) 581 1680 Fax:(93) 581 2292 E-mail: [email protected] http://www.ecap.uab.es Desembre 2010 CO2 EMISSIONS AND ECONOMIC ACTIVITY: HETEROGENEITY ACROSS COUNTRIES AND NON STATIONARY SERIES1 Matías Piaggio* and Emilio Padilla Department of Applied Economics, Univ. Autónoma de Barcelona, Edificio B, Campus de Bellaterra, 08193, Bellaterra, Spain E-mails: [email protected]; [email protected] Tel. : +(34) 935814572 Fax: + (34) 935812292 *Corresponding author 1 We are deeply grateful to Prof. J.L. Raymond, for his advice, patience, and generosity in sharing his knowledge. M. Piaggio wants to thank to the Agencia Española de Cooperación Internacional from the Ministerio de Asuntos Exteriories for the financial support provided. The authors also acknowledge support from projects ECO2009-10003 (Ministerio de Ciencia e Innovación), 2009SGR-600 and XREPP (DGR). CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series Abstract This paper explores the homogeneity of the functional form, the parameters, and the turning point, when appropriate, of the relationship between CO2 emissions and economic activity for 31 countries (28 OECD, Brazil, China, and India) during the period 1950 to 2006 using cointegration analysis. With a sample highly overlapped over time between countries, the result reveals that the homogeneity across countries is rejected, both in functional form and in the parameters of long term relationship. This confirms the relevance of considering the heterogeneity in exploring the relationship between air pollution and economic activity to avoid spurious parameter estimates and infer a wrong behavior of the functional form, which could lead to induce that the relationship is reversed when in fact it is direct. Keywords: Bound testing, cointegration, CO2 emissions, environmental Kuznets curve, heterogeneity. JEL codes: C32, O13, Q53, Q56. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 1 1. Introduction The Environmental Kuznets Curve (EKC) hypothesis comes from Kuznets (1955), where an inverted-U shaped relationship is supposed between income inequality and income level. The EKC hypothesis suggests the existence of an inverted-U shaped relationship between environmental degradation and income level. Grossman and Krueger (1991) argued that there are three channels that explain this path. In early stages of economic growth, the greater requirement of natural resources and waste generation increases environmental degradation (scale effect). This growing path might lead to changes in the economic structure towards less polluting activities (composition effect), which along with the increase in the capacity of higher income countries to face technological substitution towards less polluting processes (technological effect) would lead to a turning point in the relationship and to the decreasing section of the curve. Therefore, the transition from the increasing to the decreasing section of the curve in the relationship between environmental degradation and economic activity would arise when the composition and technological effects worked in the indicated direction and overcame the scale effect2. 2 The existence of composition and technological effects do not necessarily imply a result as the one suggested by the EKC hypothesis. For this to be the case, it is required that the composition effect involves a reduction of polluting sectors in absolute and not only in relative terms. As for the technological change, it might sometimes involve new processes with new (and sometimes unknown) pollutants or efficiency improvements leading to the increase of extractive or other environmentally damaging activities (Roca and Padilla, 2003). Therefore, it depends on the type of technological and composition change that these effects compensate or reinforce the scale effect for a specific pollutant. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 2 However, an EKC can be driven by different underlying factors, so that the relation behind the hypothesis can be generated by different structural models (Perman and Stern, 1999). The literature highlights the distribution of power (Torras and Boyce, 1998), income-elasticity of the demand for environmental quality (McConell, 1997; Dasgupta et al. 2002), environmental regulation and international agreements (de Bruyn, 1997) or structural transitions, like the oil price shocks in the 1970s (Moomaw and Unruh, 1997). Also, an EKC can be reached by individual countries through the pollution haven hypothesis (Stern et al., 1996; Cole et al., 1997). In this way, although an inverted-U relationship can be empirically shown, this can be a statistical result stemming from other factors, which might imply that the observed relationship between environmental degradation and economic growth is spurious. Moreover, these factors might vary across countries and be different for different pollutants. Earlier works ignored that the relationship between environmental degradation and income can be heterogeneous across countries (or regions), both in the functional form as well as the parameters and the turning point (Grossman and Krueger, 1991 and 1994; Shafik and Bandyopadhyay, 1992; Selden and Song, 1994; Carson et al. 1997; Cole et al. 1997 and Vincent, 1997). This issue was first studied in the late 1990s and early 2000s (Perman and Stern, 1999 and 2003; List and Gallet, 1999; Dijkgraaf and Vollebergh, 2001; Martínez-Zarzoso and Bengochea-Morancho, 2003 and 2004 and Dijkgraaf et al., 2005). Following the same concerns, a series of analyses of the EKC at national level has emerged, (among them Vincent, 1997; de Bruyn et al., 1998; Moomaw and Unruh, 1998; Friedl and Getzner, 2003; Lekakis, 2000; Roca et al., 2001; Decon CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 3 and Norman, 2004; Egli, 2004; Hung and Shawn, 2004; Shen, 2006; Halicioglu, 2008; Piaggio, 2008; Song et al., 2008; and Wang, 2009). Moreover, until the study of Perman and Stern (1999), the statistical properties of the data employed were not taken into consideration. The analysis using nonstationary series has to be carried out taking into account this characteristic. The traditional EKC approach not only ignores that economies with the same level of activity might present heterogeneous functional forms with respect to the relationship between income and environmental degradation, but also assumes parameter homogeneity in this relationship across countries. An EKC estimated from cross-section or panel data when the series are not or are hardly overlapped over time across countries can simply reflect the juxtaposition of a positive relationship between environmental degradation and income in rich countries with a negative one in developing countries, and not a relationship operating for both kinds of countries (Vincent, 1997). This problem can be solved if the panel data set has overlapped observations for large periods (Egli, 2004). However, this would not solve the problem of assuming homogeneity in the functional form of the relationship between environmental degradation and income among countries. In light of the above, the analyses that assume homogeneity in the functional form and in the parameters across countries might in fact not reflect the behavior of the relationship between environmental degradation and income for these at the individual level. So, the conclusions that, after certain point, CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 4 environmental degradation decreases with greater economic activity for the more developed countries might be wrong. Consequently, more attention should be paid to individual countries behavior in order to assess the possible benefits of the increase in economic activity on environmental quality for each country (de Bruyn et al., 1998). To impose a priori the constraint of homogeneity between countries in the functional form and the parameters might be a statistical device more than a model that appropriately approximates reality. Carson (2010) argues that the analysis should distinguish between a “weak” version of the EKC hypothesis, for a particular political jurisdiction, and a “strong” one, applying for the different political jurisdictions. The objective of this paper is to analyze the homogeneity in the functional form and parameters among countries in the long-run relationship between carbon dioxide emissions (CO2) and economic activity. The analysis is carried out for 31 countries (28 OCDE countries, Brazil, China and India) over the period 1950–2006; such a period presents a high degree of overlapping across the series. First, the functional form homogeneity will be tested through the estimation of the relation for each individual country. The time period considered in this paper is longer than the one from previous studies. This is very important, because a longer period increases the overlap among countries that have similar economic activity level but might have heterogeneous functional forms. For those countries with homogeneous functional forms the homogeneity in the parameters of the long run relationship would be tested, allowing variations among them in both short term adjustments and in the rate of convergence to the long run relationship. Also, homogeneity in the turning CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 5 point among the countries that presents one would be tested. The use of cointegration techniques would avoid the possibility of a spurious relationship between CO2 emissions and economic activity. In the next section, the conceptual framework of the EKC hypothesis and the relationship between economic growth and environmental degradation adjusted to our analysis is presented. Section 3 presents the methodology and data used. Section 4 details the analysis results. Section 5 presents the final remarks. 2. Conceptual framework The EKC hypothesis arises from a reduced model specification. Therefore, it can be the result of one or more different structural relationships, because it is an empirical phenomenon. So, this is in fact an apparent relation analysis between environmental degradation and economic activity. In line with previous works, the reduced form model relates environmental degradation level with economic activity for each country, which can follow a lineal, quadratic or cubic functional form: (1) where E denotes the indicator of environmental degradation (per capita) and Y is income (per capita). Subscript i=1,…, N indicates subjects (countries), subscript t = 1, …, T is the time period indicator, and ε is the error term normally ititititiit YYYE εβββα ++++= 3 3 2 21 CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 12 particular criteria (Hall, 1991)5. The term within brackets represents the error correction term (ECT). The interpretation of its parameters should be cautious because when the term is normalized with respect to variable E, the sign of the other variable coefficients is opposite to the expected one. Besides the improvement in the consistence provided by the estimation method, this specification, presents three more advantages: i) it allows to identify the long run relationship, the short run dynamic and the coefficient of adjustment to the equilibrium relationship (α), ii) if the series in levels are cointegrated, the ECM is a linear combination of stationary variables. Then, estimations are robust, and conventional inference procedures can be applied, and iii) this specification allows testing different restrictions among individuals (Perman and Stern, 1999 and 2003). Cointegration analysis and the estimation of the long run relationship by means of the ECM should be reiterated for the cubic specification (equations (2) and (3)), quadratic (when 0 and 0  1…)) and linear (when 0 and 0  1… and 1…). That way, the best functional form of the long run relationship between CO2 emissions and income level for each single country will be determined (if one exists). Those countries that do not satisfy the BT cointegration test, or that the model estimated is not satisfactory for the functional form that the BT indicates, a unit roots analysis through the Augmented Dickey-Fuller test (ADF) and the cointegration analysis through Engel-Granger test (1987) should be carried out (Enders, 2004). Then, when the series are I(1) and are cointegrated the ECM may be estimated for each 5 A general model for a given p, p1, p2 and p3 value, large enough, is specified. Then, the lag is reduced, in an independent way for each of them, determining the value of each of them for the lag of greater degree statistically significant. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 13 specification. Engle-Granger cointegration test is seen as the most appropriate one for the present analysis, because a priori we explore the existence of only one cointegration relation. The test proposed by Johansen and Juselius (1990) and Johansen (1991) becomes complex in the presence of non linear transformations of one of the variables, as it allows for the existence of more than one cointegration relationship. The present specification does not tackle the omission of relevant variables problem. List and Gallet (1999) argue that a reduced form model allows to measure the direct and indirect relationship between economic activity and environmental degradation, so that the inclusion of additional variables would distort the analysis. Therefore, it is not possible to make causality conclusions based on a reduced form model. So, it is not possible to assess what causes the relationship to exist. This kind of analysis allows the study of apparent elasticities, not being an analysis of the determinants of environmental pollution. As it is a uniequational specification, it does neither solve the problem of a possible feedback between the variables. However, as it is developed through a cointegration analysis, the estimated parameters will be superconsistent, not being affected by the endogeneity bias of the variables (Veerbek, 2005). The specification of the ECM for the analysis of this relationship is employed by Perman and Stern (1999 and 2003) for SO2 emissions, and Martínez-Zarzoso and Bengochea-Morancho (2003 and 2004) and Dinda and Condoo (2006) for CO2 emissions, all of them working with panel data. Egli (2004) for diverse kind of contaminants and Iwata et al. (2009 and 2010) for CO2 emissions employed CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 14 it for individual countries, and Haciglou (2008) and Piaggio (2008), who study CO2 emissions for individual countries but in a multi equation specification. Once the correct functional form is specified and the long run relationship through the ECM is estimated, the homogeneity of parameters among countries with equal functional form will be studied, allowing the short run coefficients to be different among countries, as well as the quantity of lags in each one of them. This will be tested computing confidence intervals (CI)6 for the parameters of the long run relation, grouping those countries with same functional form the CI of which overlap. The same exercise is carried out with respect to the coefficient of adjustment of disequilibria from the long run relationship (α). A similar strategy is followed for testing the turning point homogeneity. The turning point for countries with a quadratic functional form equation (3) is given by θ   Normal󰇡θ ,Vθ 󰇢7, given the distribution of parameters β1 and β2. From this, the turning point CI will be computed for the turning point of those countries whose best adjustment is the quadratic functional form. A similar procedure might be developed with respect to those with cubic functional form. 6 IC: 󰇣󰆹/  √󰇤, where   is the standard deviation associated to the estimated parameter 󰆹, 󰇛1󰇜 is the confidence level, and  is the sample size. 7 Employing the Delta Method, following Hayashi (2000: pp. 93–94) and Greene (2003, p. 70), . 󰇡  󰇢󰇡 󰇢󰇭   󰇮 CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 15 2.2. Data The analysis takes into account 31 countries (28 OECD countries8, Brazil, China and India) between1950–20069. This time period is longer than the one from previous studies on the homogeneity of the parameters for CO2 emissions, which increases the possibility of taking into account countries with overlapped income levels but heterogeneous paths. Moreover, the sample contains almost all countries (except Iceland and Luxembourg) committed to quantitative limits in CO2 emissions through Annex B of the Kyoto Protocol (United Nations, 1998). CO2 emission data is published by the Carbon Dioxide Information Analysis Center (CDIAC) (Boden et al., 2009). It is consistent with the one of the World Bank (2005) for the period 1960–2005, allowing to take into account ten more years. CO2 emissions are measured in metric tons of CO2. Logarithmic transformation of emissions per capita (co2pc) is employed. Economic activity at national level employed are estimated and transformed to 1990 Geary-Khamis dollars (which corrects by purchasing power parity, PPP) by Madison (2003), updated to 2005 by the same author for 155 countries. The 8 Australia, Austria, Belgium, Canada, former Czechoslovakia (after 1992 the values for Czech Republic and Slovakia are added), Denmark, Finland, France, Germany (for the period 1950– 1990 the information for the German Federal Republic and the German Democratic Republic are added), Greece, The Netherlands, Hungary, Ireland, Italy, Japan, South Korea, Mexico, Norway, New Zealand, Poland, Portugal, Spain, Sweden, Switzerland, Turkey, UK, USA, and former Soviet Union (from 1992 the values of Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine and Uzbekistan are added). Two OECD countries, Iceland and Luxembourg, are excluded due to lack of information for the entire period. 9 Except for Belgium, for which we took the period 1962–2006, as it presented atypical values for the two first years of the sample. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 16 National Accounts System was set up in 1950 in various countries, which allows having reliable information. Logarithmic transformation of per capita growth domestic product for the variable in levels, and its quadratic and cubic transformation are used (gdppc, gdppc2, and gdppc3, respectively). 4. Results In this section, first the cointegration analysis through the BT test is carried out to determine the existence of a long run relation between the variables and the more adequate functional form for each of the countries. Second, the analysis of the parameters of the long run relation homogeneity between countries, of the turning point and of the ECT coefficient is performed through confidence intervals construction. 4.1. Cointegration analysis Following Pesaran et al. (2001) we will carry out the contrast several times, including up to four lags, due to the sensitiveness of the analysis to the quantity of lags included. Though the quantity of lags seems high when working with annual data, the length of the series allows it. Table I summarizes the results of the F-statistic of the Wald test for the linear, cubic and quadratic specification of equation (2). Some countries of the sample allow for the existence of a long run relationship for the variables of interest for more than one functional form. This might result, CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 17 for example, from quadratic forms that have not achieved the maximum, or that have just surpassed it, or cubic forms with tiny decreasing sections might be approached through linear models. Therefore, the adequate functional form for each country would be determined from the cointegration analysis jointly with the estimation of the equation (3) for each one of the functional forms in the countries confirming the existence of a long run relationship10. As shown in Table I, BT rejects the null hypothesis of no cointegration for Australia (linear and quadratic specification), Austria (quadratic and cubic), former Czechoslovakia (linear and quadratic), Denmark (quadratic and cubic), Germany (cubic), Greece (linear and quadratic), Hungary (linear and cubic), Ireland (linear, quadratic and cubic), Italy (linear, quadratic, and cubic), Japan (quadratic), South Korea (linear, quadratic, and cubic), Poland (linear), Portugal (quadratic and cubic), Switzerland (linear, quadratic and cubic), Turkey (quadratic and cubic), former Soviet Union (quadratic) and China (linear and cubic). When the BT is inconclusive, Iwata el al. (2009 and 2010) argue that the non existence of a cointegration relationship may be rejected or not according to the test of significance of the parameter of adjustment (α) of equation (3). The BT is not conclusive for Belgium (quadratic), Canada (quadratic and cubic), former Czechoslovakia (cubic), Finland (linear, quadratic and cubic), Greece (cubic), the Netherlands (quadratic and cubic), Hungary (quadratic), Japan (cubic), 10 For the choice of the functional form we employed different statistical and analytical tools, such as the t-statistic significance of the parameters, the Schwartz information Criteria, and taking into account if the turning point estimated is lower than the maximum level of income reached by each country. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 18 Mexico (linear), Norway (quadratic and cubic), New Zealand (quadratic and cubic), Spain (quadratic), Sweden (linear and cubic), former Soviet Union (linear and Cubic), China (quadratic) and India (quadratic). Finally, the test indicates that there is not a long run relationship for any functional form for France, United Kingdom, USA and Brazil. Lags 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 AUS 4.09b7.66** 1.02 0.80 0.72 5.12** 4.93** 1.35 0.72 0.84 2.31 0.51 2.03 1.82 NA AUT 1.50 0.86 1.17 1.25 0.78 4.25* 2.87 3.92b2.15 1.62 3.31b2.10 4.11* 3.86* 2.93b BEL 4.91* 2.96 2.37 1.73 0.89 9.78*** 3.20b2.07 1.65 1.36 9.31*** 1.70 2.17 1.14 0.93 CAN 0.65 0.49 0.78 1.50 2.14 2.57 3.08 1.54 1.88 3.53b2.35 3.40b2.19 1.54 2.81b CZE 8.01*** 3.37 3.12 3.96 5.1* 4.63* 2.87 2.08 1.96 1.99 3.14b1.73 1.39 0.36 0.56 DEN 2.16 2.14 1.75 1.73 1.42 9.66*** 5.53** 5.84** 7.23** 3.27b7.05*** 4.61** 4.98** 5.79*** 3.58b FIN 2.82 2.63 2.81 4.54b3.71 3.22b2.58 2.19 1.61 1.25 2.81b2.46 2.95b1.90 1.32 FRA 1.21 1.49 1.55 2.17 1.69 1.84 1.79 2.07 1.78 1.01 2.22 1.53 2.51 3.03b2.75b GER 1.39 0.36 0.54 0.62 0.75 1.33 1.92 1.67 1.23 1.26 3.89* 2.62 2.26 2.23 0.80 GRE 4.27b5.07* 6.27** 7.58** 7.45** 5.64** 6.15** 6.35** 4.43** 5.30** 2.88b 2.69 2.18 1.37 1.16 HOL 1.23 0.63 0.55 0.85 0.93 3.16 2.45 3.50b2.60 2.07 2.88b2.01 2.92b2.52 2.07 HUN 13.01** * 8.69*** 2.80 4.73b4.43b3.84b3.29b0.63 0.49 0.51 3.15b3.87* 1.10 2.49 1.81 IRE 1.94 2.20 4.73b5.73** 8.32** 7.12*** 3.86b2.75 1.87 3.25b7.80*** 5.14** 3.34b2.08 1.90 ITA 6.50** 2.85 2.75 2.54 1.78 3.67b5.38** 1.75 1.98 2.31 3.10b4.82** 1.72 3.17b3.36b JAP 1.16 3.26 1.95 1.73 1.84 2.01 4.35** 2.65 1.10 1.79 1.54 2.85b1.82 2.36 2.73b KOR 24.53** * 12.57*** 19.36** * 8.56** 3.97** 21.19*** 8.41*** 10.20** * 7.06** 3.39b15.58*** 6.28** 9.75*** 3.50b1.95 MEX 1.09 0.67 0.13 0.49 0.66 1.67 1.50 2.36 1.49 1.52 1.67 1.53 3.48b2.64 3.59b NOR 2.20 1.60 1.51 1.89 2.53 3.54b1.30 1.62 0.82 1.54 3.22b1.86 2.77b2.42 1.41 NZL 2.70 2.50 1.11 2.22 2.23 2.52 2.41 2.68 2.21 3.20b3.12b1.72 1.63 1.80 1.96 POL 5.23** 3.26 1.72 1.54 1.46 2.38 0.82 0.50 0.03 0.06 1.93 1.59 1.06 0.42 0.46 POR 0.04 0.13 0.11 0.06 0.06 6.68*** 3.69b2.50 2.82 3.73b6.95*** 3.75b2.25 3.01b4.59** SPA 0.14 0.09 0.02 0.02 0.12 3.24b1.19 1.62 1.00 1.20 2.25 1.06 1.17 0.91 1.27 SWE 4.41b3.58 2.83 4.43b3.60 1.48 0.74 0.89 0.83 1.23 2.12 2.46 2.55 3.26b2.45 SWI 1.44 0.89 2.76 4.60b6.42** 5.83** 8.05*** 3.82b3.37b2.20 5.67*** 7.02** 1.57 3.14b1.67 TUR 0.64 1.16 2.65 2.31 1.65 7.51*** 6.86*** 3.28b2.17 2.88 5.14** 4.62** 2.38 2.19 3.39b UK 3.97 1.61 1.89 1.70 1.33 2.61 1.65 1.86 1.36 0.83 2.16 1.43 1.70 1.22 0.70 USA 0.53 0.90 1.71 1.74 3.10 1.14 1.11 0.78 0.96 1.17 1.98 1.26 0.54 0.71 0.47 USS 4.53b2.21 3.27 2.33 1.06 4.57* 1.38 2.83 1.76 0.54 3.29b0.92 2.58 2.04 1.76 BRA 3.75 3.48 2.18 3.63 0.81 2.50 1.93 1.61 1.46 1.57 2.65 1.83 1.70 1.65 1.25 CHN 6.70*** 3.88 4.67b7.22** 5.25* 2.64 2.99 2.49 3.46b3.51b3.63b4.39** 2.86b2.78b3.80* IND 0.10 0.58 2.59 1.71 1.66 4.25* 3.82* 1.16 1.14 0.75 2.50 2.39 0.59 0.86 0.37 3 1% CV (4.29;5.61), 5% CV (4.35;3.23) and 10% CV (3.77;2.72) ***, **,* significant at 1%, 5% and 10% respectively b inconclusive at 1% Table I - CO 2 emissions and economic activity bound testing cointegration test Lineal 1 Quadratic 2 Cubic 3 1 1% CV (6.84;7.84), 5% CV (4.98;5.73) and 10% CV (4.04;4.78) 2 1% CV (5.15;6.36), 5% CV (3.79;4.85) and 10% CV (3.17;4.41) CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 19 From the analysis above, when BT does not reject the existence of a long run relationship equation (3) is estimated. Therefore, the preferred functional form for each country is determined. Table A1 of Annex A summarizes the ECT estimation of equation (3) for each one of the possible functional forms. The results indicate the existence of a long run relationship between CO2 emissions and economic activity, both in per capita terms, in a cubic path for Sweden, quadratic for Australia, Austria, Belgium, Canada, Denmark, Finland, The Netherlands, Ireland, Italy, Japan, Norway, Switzerland, China and India, and lineal for South Korea, Greece and Brazil. Finally, there is no long run relationship between the variables involved for any functional form for former Czechoslovakia, Hungary and the former Soviet Union. From the 17 countries for which a quadratic specification is possible, 14 present the turning point within the sample, which confirms an inverted-U path. The other 3 are very close to achieving it. Sweden also presents the turning points within the values of the sample. The functional form specification for 18 countries of the sample has been determined, and the ECM for each one of them has been estimated. Moreover, for those countries that BT did not indicate the existence of a cointegration relation (France, United Kingdom, USA and Brazil), and for those that BT did not reject it for one of the specifications but was not possible to estimate a satisfactory long run relationship (Germany, Mexico, New Zealand, Poland, Portugal, Spain and Turkey), a unit roots analysis through the ADF statistic and a cointegration analysis through the Engle-Granger test are implemented. All the series for all the countries are I(1), and the existence of a long run CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 20 relationship is not rejected for any of the specifications for France and Spain, the linear and quadratic specifications for Germany, Mexico, USA and Brazil, the cubic for Poland and the linear for Portugal and Turkey. The analysis rejects the existence of a long run relationship for New Zealand11. There is a long run quadratic relationship for France, Germany and USA, and linear for Mexico, Portugal, Spain, Turkey and Brazil. Poland and United Kingdom do not present any satisfactory specification, as equation (3) shows estimations for the specifications that do not reject the existence of a cointegration relationship. Table II summarizes the results, 25 of the 31 countries of the sample do not reject the existence of a long run relationship between economic activity and CO2 emissions (7 linear, 17 quadratic and 1 cubic). The result obtained confirms the heterogeneity among countries of behavior patterns for similar activity levels. Comparing these results with other analyses for the same pollutant for individual countries, they are consistent with the ones of Iwata et al. (2009) for France (for the period 1960–2003), and Iwata et al. (2010) for Finland (1977– 2003) and Japan (1966–2003). The last one tests —and obtains positive evidence of— the existence of a quadratic path for South Korea (1977–2003) and Spain (1968–2003), in contrast with the linear model supported by our results. Both works quoted take into account the share of nuclear power in total 11 The results from the unit roots and cointegration tests are available from the authors upon request. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 21 energy generation for each country. However, the linear specification for Spain is consistent with Roca and Padilla (2003) for the period 1980–2000, who also included factors referred to the energy sources structure. In contrast with our results, Friedl and Getzner (2003) found a cubic relationship for Austria (1960–1999), introducing the weight of imports and industry in total income. Haciloglu (2008) also found a different path from ours for Turkey Model Country Decision Method BRA EG GRE BT KOR BT MEX EG POR EG SPA EG TUR EG AUS BT AUT BT BEL BT CAN BT CHN BT DEN BT FIN BT FRA EG GER EG HOL BT IND BT IRE BT ITA BT JAP BT NOR BT SWI BT USA EG Cubic SWE BT CZE BT HUN BT NZL EG POL EG UK EG USS BT Table II - Summary of long term relationship estimation LinearQuadraticNo relation CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 28 economic activity growth in environmental quality, the study should focus on the analysis of the relationship between these factors at single country level. The results of the present research are consistent with Dijkgraaf and Vollebergh (2001) and Dijkgraaf et al. (2005) on the problematic assumption of parameters homogeneity of the long run relation between CO2 emissions and economic activity level, both per capita, employing a longer period sample, which allows a greater degree of overlapping of the series among countries. At the same time, this greater overlapping reinforces the result of rejecting the homogeneity in the functional form among countries (Perman and Stern, 1999 and 2003; List and Gallet, 1999; Dijkgraaf and Vollebergh, 2001; Martínez-Zarzoso and Bengochea-Morancho, 2003 and 2004 and Dijkgraaf et al., 2005). This is highlighted by the fact that heterogeneous functional forms are found for countries with similar level of economic activity. The existence of a general relation for all the countries between CO2 emissions and GDP per capita is clearly put into question. Following Carson (2010), this result rejects the optimistic view of the EKC, where developing countries might ignore environmental problems until they become developed. Developed countries can and have to consider this problem, since nothing guarantees a path as the one of the EKC for all countries (and neither the existence of a common path for them) (Dasgupta et al., 2002). Finally, the turning point homogeneity is rejected for the whole sample of countries. However, there are groups of countries that present parameter CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 29 heterogeneous paths but for which turning points homogeneity is not rejected. Although this is not strong evidence in favor of the optimistic view of the EKC, it suggests that it would be interesting to analyze the determinants of these countries. Moreover, the importance of considering the environmental degradation level from which the damage turns extremely high or irreversible must be highlighted, as while different countries might present a turning point for a similar level of activity, the environmental degradation achieved in this point might be different. 6. References Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Helling, C. S., Jansson, B.-O., Levin, S., Mailer, K.-G., Perrings, C., Pimentel, D. (1995) “Economic growth, carrying capacity, and the environment”, Science, Vol. 268, pp. 520–521. Boden, T.A., Marland, G., Andres, R.J. (2009) Global, Regional, and National Fossil-Fuel CO2 Emissions, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tenn., U.S.A. doi 10.3334/CDIAC/00001 Carson, R. (2010) “The Environmental Kuznets Curve: seeking empirical regularity and theoretical structure”, Review of Environmental Economics and Policy, Vol. 4:1, pp. 3–23. Carson, R, Jeon, Y., Mccubbin, D. (1997) “The relationship between air pollution emissions and income: US Data”, Environment and Development Economics, Nº 2, pp. 433–450. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 30 Cole, M.A, Rayner, A.J., Bates, J.M. (1997) “The environmental Kuznets curve: an empirical analysis”, Environment and Development Economics, Nº 2, pp. 401–416. Dasugpta, S., Laplante, B., Wang, H., Wheeler, D. (2002) “Confronting the environmental Kuznets curve”, Journal of Economic Perspectives, Vol. 16, pp. 147–168. de Bruyn, S.M., van den Bergh, J.C., Opschoor, J.B. (1998) “Economic growth and emissions: reconsidering the empirical basis of environmental Kuznets curves”, Ecological Economics, Vol. 25, pp 161–75. Dijkgraaf, E., Melenberg, B., Vollebergh,H. (2005) “Environmental Kuznets Curves for CO2: heterogeneity versus homogeneity”, Environmental and Resource Economics, Vol. 32, pp. 229–239. Dijkgraaf, E., Vollebergh, H. (2005), “A Test on parameter heterogeneity in CO2 panel EKC estimations”, Environmental and Resource Economics, Vol. 32, pp. 229–239 Dinda, S., Coondoo, D. (2006) “Income and emission: A panel data-based cointegration analysis”, Ecological Economics, Vol. 57, pp. 167–181. Egli, H. (2004) “The Environmental Kuznets Curve – Evidence from Time Series Data for Germany”, ETH Zurich, Http://ideas.repec.org/p/eth/wpswif/0328.html. Enders, W. (2004) Applied Econometric Time Series, Wiley, 2nd Edition. Engle, R. F., Granger, C. W. J. (1987) “Co-integration and Error Correction: Representation, Estimation, and Testing”, Econometrica, Vol. 55, pp. 251– 276. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 31 Friedl, B., Getzner (2003) “Determinants of CO2 emissions in a small open economy”, Ecological Economics, Vol. 45, pp. 133–148. Granger, C.W., Hallman, J. (1991) “Nonlinear transformations of integrated time series”, Journal of Time Series Analysis, Vol. 12, pp. 207–224. Greene, W. (2003) Econometric Analysis, Prentice Hall. Grossman, G., Krueger, A. (1991) “Environmental impacts of a North American free trade agreement”, NBER working paper, Nº 3914. Halicioglu, F. (2009) “An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey”, Energy Policy, Vol. 37, pp. 1156–1164. Hall, A. (1991) “Testing for a Unit Root in Time Series with Pretest Data Based Model Selection”, Department of Economics, North Carolina State University. Mimeo. Hayashi, F. (2000) Econometrics, Princeton University Press. Hung, M.F., Shaw, D. (2004) “Economic Growth and the Environmental Kuznets Curve in Taiwan: a simultaneity model analysis”, en in Boldrin, M., Chen, B.L. y Wang, P. (eds.), Human Capital, Trade and Public Policy in Rapidly Growing Economies: From Theory to Empirics, pp. 269–290. Iwata, H., Okada, K., Samreth, S. (2010) “Empirical study on the determinants of CO2 emissions: evidence from OECD countries”, Munich Personal RePEc Archive, paper nº 21520. Iwata, H., Okada, K., Samreth, S. (2009) “Empirical study on the Environmental Kuznets Curve for CO2 in France: the role of nuclear energy”, Munich Personal RePEc Archive, paper nº 18997. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 32 Johansen, S. (1991) “Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models”, Econometrica, Vol. 59, pp. 1551–1580. Johansen, S., Joselius, K. (1990) “Maximum Likelihood Estimation and Inferences on Cointegration— with applications to the demand for money”, Oxford Bulletin of Economics and Statistics, Vol. 52, pp. 169–210. Kuznets, S., (1955) Economic growth and income inequality, AER. Lee, C., Lee, J. (2009) “Income and CO2 emissions: evidence from panel unit root and cointegration tests”, Energy Policy, Vol. 37, pp. 413–423. Lekakis, J. (2000) “Environment and Development in a southern European country: which Environmental Kuznets Curves?”, Journal of Environmental Planning and Management, Vol. 43, pp. 139–153. List, J., Gallet, C. (1999) “The Environmental Kuznets Curve: does one fits all?”, Ecological Economics, Vol. 31, pp. 409–423. McConnell, K. (1997) “Income and demand for environmental quality”, Environment and Development Economics, Vol. 2, pp. 383–399. Maddison, A. (2003) The World Economy: Historical Statistics, OECD. Martínez-Zarzoso, I., Bengochea-Morancho, A. (2004) “Pooled mean group estimation of an environmental Kuznets curve for CO2”, Economics Letters, Vol. 82, pp. 121–126. Martinez-Zarzoso, I., Bengochea-Morancho, A. (2003) “Testing for an Environmental Kuznets Curve in latin-american countries”, Revista de Análisis Económico, Vol. 18, pp. 3–26. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 33 Moomaw, W., Unruh, G. (1997) “Are environmental Kuznets curves misleading us? The case of CO2 emissions”, Environment and Development Economics, Vol. 2, pp. 451–463. Perman, R., Stern, D. (2003) “Evidence from panel unit root and cointegration test that the Environmental Kuznets Curves dows not exist”, The Australian Journal of Aricultural and Resource Economics, Vol. 47, pp. 325–347. Perman, R, Stern, D. (1999) “The Environmental Kuznets Curve: implications of non-stationarity”, The Australian National University, Centre for Resource and Environmental Studies, Working Paper in Ecological Economics Nº 9901. Pesaran, M.H., Shin, Y. Smith, R. (2001) “Bounds Testing approaches to the analysis of level relationships”, Journal of Applied Econometrics, Vol. 16, pp. 289–326. Piaggio, M. (2008) “Relación entre la Contaminación Atmosférica y la Calidad del Aire con el Crecimiento Económico y otros Determinantes: Uruguay a lo largo del Siglo XX”, Quantum, Vol. III, Nº 1, pp. 35–54, http://www.quantum.edu.uy/numero2/nota%203.pdf. Roca, J., Padilla, E. (2003) “Emisiones atmosféricas y crecimiento económico en España”, Economía Industrial, Vol. 351, pp. 73–86. Roca, J., Padilla, E., Farré, M., Galletto, V. (2001) “Economic growth and atmospheric pollution in Spain: discussing the environmental Kuznets curve hypothesis”, Ecological Economics, Vol. 39, pp. 85–99. Selden, T., Song, D. (1994) “Environmental quality and development: Is there a Kuznets curve for air pollution emissions?”, Journal of Environmental Economics and Management, Vol. 27, pp. 147–162. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 34 Shafik, N., Bandyopadhyay, S. (1992) “Economic growth and environmental quality: time-series and cross-country evidence”, Policy research working papers, Background paper for the World Development Report. Shen, J. (2006) “A simultaneous estimation of Environmental Kuznets Curve: Evidence from China”, China Economic Review, Vol. 17, pp. 383–394. Song, T., Zheng, T., Tong, L. (2008) “An empirical test of the environmental Kuznets curve in China: A panel cointegration approach”, China Economic Review, Vol. 19, pp. 381–392. Stern, D., Common, M., Barbier, E. (1996) “Economic growth and environmental degradation: the environmental Kuznets curve and sustainable development”, World Development, Vol. 24, pp. 1151–1160. Torras, M., Boyce, K. (1998) “Income, inequality, and pollution: a reassessment of the environmental Kuznets Curve”, Ecological Economics, Vol. 25, pp. 147–160. United Nations (1998) Kyoto Protocol of the Framework Convention of the United Nation on Climate Change, United Nations. Veerbeek, M. (2004) A Guide to Modern Econometrics, Wiley. Vincent, J. (1997) “Testing for environmental Kuznets curves within a developing country”, Environment and Development Economics, Vol. 2, pp. 417–431. Wagner, M. (2008) “The Carbon Kuznets Curve: a cloudy picture emitted by bad econometrics?”, Resource and Energy Economics, Vol. 30, pp. 388– 408 CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 35 Wang, Y. (2009) “Testing the EKC hypothesis: some econometric pitfalls and solutions”, School of Economics, Arndt Building 25a, The Australian National University. World Bank (2005) “World development indicators”, Washington, D.C. CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 36 Annex A Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear ADRL (0,0,0,0) (0,0,0) (1,1) (0,2,2,2) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,3) (1,0,0,0) (0,0,0) (1,1) (0,0,0,0) (0,0,0) (0,1) alfa -0.45 -0.37 -0.07 -0.30 -0.32 -0.04 -0.19 -0.21 -0.13 -0.16 -0.18 -0.09 0.01 -0.02 0.01 -0.53 -0.55 -0.11 t-statistic -3.47 -3.38 -1.92 -2.73 -3.32 -1.29 -2.59 -3.52 -1.89 -2.10 -2.30 -1.73 0.27 -0.79 0.42 -4.52 -4.96 -1.84 C -9.40 -3.13 -9.43 -9.89 -6.08 0.37 -6.48 -6.75 -9.19 21.29 -3.60 -8.52 11.21 -8.60 0.01 5.14 0.69 -8.48 t-statistic -1.95 -5.01 2.04 -3.11 -14.14 1.31 -6.04 -9.22 -39.02 0.99 -1.92 -19.24 2.99 -0.88 0.00 0.63 0.81 -9.23 GDPPC(-1) 3.26 -4.18 0.00 1.84 -1.68 0.00 -2.59 -2.08 -0.06 -33.57 -4.35 -0.32 -7.11 -5.03 -5.60 -12.75 -7.40 -0.19 t-statistic 0.57 -8.43 0.08 0.45 -4.37 0.00 -1.82 -3.49 -0.65 -1.34 -2.91 -1.90 -1.11 -0.59 -0.47 -1.29 -10.84 -0.55 GDPPC^2(-1) -2.25 0.66 -0.72 0.24 0.72 0.43 12.07 0.77 5.81 2.26 3.47 1.36 t-statistic -1.01 6.77 -0.41 2.88 1.01 3.40 1.26 2.64 1.63 0.90 0.89 10.18 GDPPC^3(-1) 0.37 0.07 -0.05 -1.44 -1.34 -0.27 t-statistic 1.31 0.27 -0.43 -1.19 -2.07 -0.54 Interventions Step 1970 Impulse Turning Point 3.07 3.19 5.36 3.46 2.39 2.42 2.61 2.83 0.88 1.11 2.70 2.72 0.95 1.68 7.15 2.97 2.01 5.75 Schwartz iC -3.84 -4.03 -3.86 -2.68 -3.10 -3.86 -3.14 -3.21 -2.97 -3.43 -3.46 -3.49 -3.51 -3.81 -3.71 -2.15 -2.21 -1.86 JB 1.08 0.83 0.58 0.45 0.75 0.75 0.10 0.32 0.72 2.07 1.97 4.72 4.78 5.44 7.56 1.16 0.85 1.13 p-value 0.58 0.66 0.75 0.80 0.69 0.69 0.95 0.85 0.70 0.36 0.37 0.09 0.09 0.07 0.02 0.56 0.66 0.57 BG (4 lags) 0.03 0.18 0.18 1.82 2.02 2.08 1.08 0.76 0.38 0.34 0.90 1.07 0.89 2.07 1.95 2.44 2.09 2.34 p-value 1.00 0.95 0.95 0.14 0.11 0.10 0.38 0.56 0.82 0.85 0.47 0.39 0.48 0.10 0.12 0.06 0.10 0.07 TP in the sample YES NO YES YES YES YES Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear ADRL (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (3,3) alfa -0.33 -0.33 -0.09 -0.31 -0.20 -0.06 -0.26 -0.11 -0.04 -0.26 -0.19 -0.14 -0.20 -0.23 -0.02 -0.16 -0.11 0.01 t-statistic -3.02 -3.02 -1.68 -2.71 -2.47 -1.60 -2.49 -1.75 -1.16 -2.44 -2.41 -1.68 -2.43 -3.10 -0.48 -2.10 -1.90 0.26 C 2.92 -0.90 -7.44 4.10 -3.90 -7.46 -1.64 -2.88 -9.80 -5.28 -4.69 -5.78 -20.04 -1.36 -10.55 21.29 -8.24 6.09 t-statistic 0.60 -0.93 -6.48 0.72 -2.92 -7.07 -0.20 -1.01 -7.94 -6.94 -12.62 -15.05 -1.86 -1.05 -1.59 0.99 -3.43 0.12 GDPPC(-1) -11.44 -5.84 -0.64 -13.61 -4.34 -0.26 -6.55 -5.14 0.36 -1.73 -2.72 -1.35 14.98 -5.94 0.54 -33.57 -1.10 -9.03 t-statistic -1.70 -6.94 -1.44 -1.73 -3.81 -0.86 -0.67 -2.44 0.67 -1.05 -4.68 -9.18 1.16 -5.48 0.21 -1.34 -0.42 -0.28 GDPPC^2(-1) 3.65 1.01 4.64 0.91 1.37 1.03 -0.14 0.39 -6.40 1.10 12.07 0.44 t-statistic 1.21 5.57 1.32 3.86 0.35 2.60 -0.14 2.13 -1.26 4.96 1.26 0.60 GDPPC^3(-1) -0.40 -0.53 0.00 0.09 0.87 -1.44 t-statistic -0.92 -1.05 0.00 0.49 1.32 -1.19 Interventions Step 1982 1982 Impulse 1956 Turning Point imag. 2.90 imag. 2.37 2.40 2.51 3.11 3.45 3.00 2.70 2.61 1.26 imag. imag. 1147.68 -2.08 1.92 2.97 Schwartz iC -1.61 -1.67 -1.59 -1.57 -2.83 -3.02 -3.80 -3.85 -3.91 -2.82 -2.92 -2.81 -2.82 -3.09 -2.77 -3.43 -2.98 -2.84 JB 2.01 1.25 1.13 3.36 1.56 0.34 1.47 0.26 0.48 0.48 1.91 0.45 4.96 7.87 9.06 2.07 7.00 0.26 p-value 0.37 0.53 0.57 0.19 0.46 0.84 0.48 0.88 0.79 0.78 0.39 0.80 0.08 0.02 0.01 0.36 0.03 0.88 BG (4 lags) 0.29 0.07 2.34 2.11 0.49 2.37 1.46 1.26 1.37 1.09 2.19 0.36 1.22 0.87 0.88 0.34 2.59 0.28 p-value 0.88 0.99 0.07 0.09 0.74 0.07 0.23 0.30 0.26 0.37 0.08 0.84 0.32 0.49 0.48 0.85 0.05 0.89 TP in the sample YES YES YES NO YES YES Table A1 - Error Correction Term - ECM cubic, quadratic and linear model AUS AUT BEL CAN CZE DEN FIN FRA GER GRE HOL HUN CO2 Emissions and Economic Activity: heterogeneity across countries and non stationary series 37 Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear ADRL (0,1,1,0) (0,0,0) (3,2) (0,0,0,0) (1,0,0) (0,0) (4,3,3,3) (0,1,0) (1,1,0) (3,3,0,0) (0,0,0) (2,2) (2,1,1,1) (2,0,0) (2,0) (0,2,2,2) (0,0,0) (0,1) alfa -0.70 -0.57 -0.04 -0.15 -0.25 -0.04 -0.37 -0.25 -0.09 0.02 -0.10 -0.08 -0.24 -0.17 -0.12 -0.48 -0.34 -0.11 t-statistic -5.56 -4.58 -0.50 -1.96 -3.68 -1.32 -4.33 -3.64 -2.48 0.32 -2.54 -3.59 -3.30 -2.83 -2.14 -4.34 -3.19 -1.64 C -2.84 -5.94 -11.10 0.00 -2.45 -8.63 -6.96 -5.61 -5.58 7.50 -7.18 -7.44 -12.31 -3.87 -5.80 3.11 -3.49 -7.32 t-statistic -3.51 -28.65 -1.52 0.00 -5.51 -3.29 -7.82 -24.90 -10.14 0.19 -11.31 -17.48 -4.73 -3.28 -20.82 0.51 -3.14 -11.98 GDPPC(-1) -6.52 -2.06 1.44 -9.33 -4.56 -0.04 -1.21 -2.12 -1.13 -23.64 -1.45 -0.54 13.45 -4.34 -1.34 -12.86 -3.65 -0.66 t-statistic -5.52 -10.62 0.34 -1.43 -11.70 -0.04 -0.87 -8.50 -7.29 -0.38 -3.16 -2.80 2.35 -2.51 -8.18 -1.82 -4.03 -3.09 GDPPC^2(-1) 2.29 0.31 3.30 0.80 0.20 0.31 13.01 0.24 -11.06 1.05 4.78 0.60 t-statistic 4.18 6.79 1.04 9.10 0.29 4.91 0.37 1.93 -2.72 1.77 1.76 3.32 GDPPC^3(-1) -0.27 -0.40 -0.02 -2.48 2.67 -0.62 t-statistic -3.35 -0.81 -0.14 -0.36 2.84 -1.80 Interventions Step Impulse 2006 Turning Point imag. 3.36 imag. 2.85 imag. 3.38 imag. 3.05 1.85 2.07 imag. 3.06 imag. imag. imag. imag. 0.91 imag. Schwartz iC -2.44 -2.36 -2.15 -3.52 -3.65 -3.66 -3.10 -3.12 -3.59 -2.75 -2.59 -3.16 -2.83 -2.89 -2.98 -1.62 -1.72 -1.69 JB 1.95 4.18 0.18 2.72 1.37 2.05 0.74 0.39 7.21 0.13 1.94 1.09 5.53 6.38 4.27 0.95 7.80 0.59 p-value 0.38 0.12 0.92 0.26 0.50 0.36 0.69 0.82 0.03 0.94 0.38 0.58 0.06 0.04 0.12 0.62 0.02 0.75 BG (4 lags) 0.85 0.76 1.63 0.60 0.53 0.63 3.11 2.32 1.88 1.31 0.89 0.53 1.79 0.49 1.21 0.88 2.69 2.30 p-value 0.50 0.56 0.19 0.67 0.71 0.64 0.03 0.07 0.13 0.28 0.48 0.71 0.15 0.74 0.32 0.48 0.04 0.07 TP in the sample NO YES YES NO NO YES Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear Cubic Quad. Linear ADRL (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0,0) (1,0) (0,0,0,0) (0,0,0) (0,0) (0,0,0,0) (0,0) (0,0) (1,1,1,1) (0,0,0) (0,1) alfa -0.34 -0.26 -0.25 0.03 0.03 0.00 -0.57 -0.55 -0.31 -0.29 -0.31 -0.16 -0.20 -0.14 -0.06 -0.38 -0.36 -0.04 t-statistic -3.40 -2.66 -2.80 0.78 0.66 0.03 -3.82 -3.79 -3.21 -3.48 -3.93 -2.15 -2.62 -2.00 -1.57 -2.51 -3.80 -0.74 C -64.77 -7.07 -6.50 -2.18 -0.31 93.13 -5.58 -5.82 -5.58 -5.54 -5.94 -6.42 44.63 -0.58 -11.72 41.20 14.38 -8.87 t-statistic -2.64 -2.20 -20.50 -0.44 -0.04 0.02 -14.23 -46.03 -61.00 -6.34 -26.10 -34.58 2.00 -0.14 -4.90 0.85 5.29 -1.72 GDPPC(-1) 70.12 -0.43 -0.87 -6.77 -8.76 -66.08 -1.37 -0.90 -1.20 -2.39 -1.63 -0.95 -61.52 -7.17 1.07 -42.33 -15.80 0.17 t-statistic 2.33 -0.16 -6.91 -1.41 -1.14 -0.03 -1.77 -5.18 -30.44 -1.48 -6.08 -12.61 -2.26 -2.30 1.18 -0.81 -8.00 0.09 GDPPC^2(-1) -28.66 -0.09 0.39 1.63 2.19 0.21 -0.08 0.61 0.19 23.04 1.50 11.32 2.70 t-statistic -2.34 -0.17 1.82 1.23 1.14 0.45 -1.65 0.69 2.63 2.11 2.48 0.61 7.56 GDPPC^3(-1) 3.83 0.33 -0.06 -0.07 -2.82 -0.92 t-statistic 2.32 0.60 -0.64 -0.48 -1.96 -0.42 Interventions Step 1981, 1990 1981, 1990 Impulse Turning Point 2.83 -2.44 1.44 2.00 imag. imag. 4.27 2.33 2.39 2.87 2.92 2.16 -4.72 imag. imag. 3.12 5.36 Schwartz iC -2.83 -2.81 -2.87 -3.72 -3.80 -3.47 -2.66 -2.72 -2.84 -2.74 -2.81 -2.80 -2.23 -2.23 -2.26 -2.61 -2.74 -2.45 JB 3.94 1.55 0.95 4.60 3.76 1.54 1.30 1.67 0.43 2.40 2.22 0.41 0.72 1.01 0.69 0.64 0.67 0.64 p-value 0.14 0.46 0.62 0.10 0.15 0.46 0.52 0.43 0.81 0.30 0.33 0.81 0.70 0.60 0.71 0.73 0.72 0.73 BG (4 lags) 1.08 0.50 1.04 1.80 1.99 2.17 1.38 1.22 0.49 1.76 1.86 1.01 0.93 0.86 0.82 1.55 3.28 2.26 p-value 0.38 0.74 0.40 0.15 0.11 0.09 0.25 0.31 0.74 0.15 0.13 0.41 0.45 0.50 0.52 0.21 0.02 0.08 TP in the sample YES YES YES NO YES YES Table A1 - Error Correction Term - ECM cubic, quadratic and linear model IRE ITA JAP KOR MEX NOR NZL POL POR SPA SWE SWI