Testing Greenhouse Gasses in Slovakia for Environmental Kuznets Curve and Pollution Haven Hypothesis

The analytical focus in this paper is to determine the tenability of two theoretical concepts of environmental economics in the Slovak Republic: the hypothesis of the environmental Kuznets curve and the hypothesis of Slovakia as a pollution haven. We have chosen to focus on a few selected sources of environmental pollution in the Slovak Republic, specifically greenhouse gas emissions. First we examine the hypothesis of the environmental Kuznets curves in connection with selected greenhouse gas emissions in the Slovak Republic from the years 1993-2011. Next we analyze the possible causal relationship between the net inflows of foreign direct investment and the volume of selected greenhouse gas emissions in the Slovak Republic. The existence of an environmental Kuznets curve was not confirmed. It is also possible to make the same conclusion in the relationships of carbon dioxide, nitrous oxide, and per fluorocarbons emissions and economic growth (GDP per capita). Only in the case of methane was neither a linear nor a cubic relationship confirmed between the variables of interest. A quadratic relationship was confirmed suggesting that this is the only type of greenhouse gas where estimates indicate a relationship which follows an environmental Kuznets curve. Based on the calculated values of the Granger causality test we conclude that the volume of greenhouse gas emissions per capita in the present period and in any of the previous four periods has no effect on the amount of net FDI inflows as a percentage of GDP in the Slovak Republic.


INTRODUCTION
At an annual meeting of the American Economic Association in 1954, Simon Kuznets, Nobel Prize winner in economics, delivered a speech entitled "Economic growth and income inequality". He presented the theory that up to a certain point, an increase in income per capita is accompanied by an increase in income inequality. According to Kuznets, income inequality regularly rises during the early phases of economic growth, while in subsequent periods of economic development the distribution of income gradually equalizes and plateaus.
is evolution of income inequality can be explained by the concept that in the early stages of capitalism, when investments in physical capital are the most important source of economic growth, income inequality spurs economic growth as resources are allocated to individuals who save and invest the most. In contrast, in developed economies the importance of human capital rises, and income inequality can actually hinder economic growth. is is due to low income households that are situated in imperfect credit markets that do not have su cient funds to nance high quality education. As a result, the quality of education deteriorates which consequently has a negative impact on the quality of human capital and jeopardizes economic growth in developed economies.
Consequently, with further economic growth the skew and kurtosis of income distribution are balanced more evenly. According to Kuznets, income distribution is unequal in the early stages of income growth, but with continuing economic growth and rising income, the measure of income inequality in society is mitigated (Kuznets, 1955). At a certain level of economic development when a high gross domestic product (GDP) per capita is reached, the level of income inequality no longer increases. ere is also a point of in ection beyond which the measure of income inequality actually decreases. e graph below is a curve showing the distribution of income in Great Britain as an inverted letter U, also known as a Kuznets curve. e economic development of Great Britain in the years 1996/7 and 2005/06 is given as an example. In the 1996/7 period the distribution of income is more signi cantly skewed upward with a signi cant peak than in the 2005/06 period ( gure 1).  Bewer, Goodman, Muriel, & Sibieta, 2008. p. 7 Since 1991, the concept of the Kuznets curve has appeared in a new context, speci cally in characterizing the relationship between economic growth and environmental pollution (Stokey, 1998), (Unruh & Moomaw, 1998), (Yandle, et al., 2002), (Lieb, 2003), (Deacon & Norman, 2006). In the early 1990s, discussions about the possible negative impact of economic growth on the quality of the environment arose, especially in light of the emerging free trade agreement between the United States, Mexico, and Canada (North American Free Trade Agreement, NAFTA). In their own empirical research, Grossman and Krueger indicated that if economic growth were to achieve a certain level, a higher income per capita is no longer associated with further deterioration of air quality, but rather with its improvement (Grossman & Krueger, 1991). Another important study regarding the relationship between economic growth and environmental quality was conducted by Sha k et al (1992), which analyzed the relationship between GDP per capita and ten selected indicators of environmental quality. is analysis included data from 149 countries from the years 1960-1986. In regards to the annual rates of deforestation, y ash emission, and river pollution caused by sulfur hexa uoride and fecal coliform bacteria, the results of the empirical analysis con rm the non-linear relationship between GDP per capita and pollution levels, creating an inverted U graph in all cases. Other indicators of environmental quality, such as lack of drinking water, river pollution, and a lack of urban sanitation empirically showed a signi cant linear monotonically decreasing relationship between GDP per capita and each speci c type of pollution. In contrast, solid municipal waste and carbon dioxide emissions demonstrated a positive linear relationship between income and both indicators of pollution.

Net households equivalised income
In 1999, Panayoutou (1999) gave the rst detailed explanation for the relationship between the level of economic development and environmental pollution when represented by an inverted U graph. Figure 2 shows the ideal shape of the environmental Kuznets curve.

Figure 2. Environmental Kuznets Curve
Source: (Yandle, et al, 2002, p. 3) Another theoretical concept of environmental economics and policy is the hypothesis that countries aiming to attract foreign investment lower environmental standards and regulations, especially less developed and developing countries (Cole, 2004), (Cole, et al., 2006). If the standards of environmental regulations in a country are below a level that guarantees e ective protection of the environment, or if countries fail to enforce existing environmental standards, they are known in the literature as "pollution havens", or countries that provide a haven for environmental pollution.
Since it is not possible for all countries to have uniform environmental pollution standards, Eskeland and Harrison de ne pollution havens more narrowly, basing their de nition on the theory of comparative advantage. In a given country, if the cost of compliance with environmental standards in certain sectors is used by foreigners as their own comparative advantage, and if this country o ers investors standards that are connected with lower costs, this country can be described as a pollution haven (Eskeland & Harrison, 1997). However, according to Neumayer this de nition of a pollution haven focuses only on the di erences in the cost of compliance with environmental standards, while ignoring the reasons that lead a country to behave in such a way in the rst place. Neumayer states that countries do not fail to determine and set standards that are necessary for e ective protection of the environment, but rather countries fail to e ectively enforce existing environmental standards because their objective is to attract foreign capital. erefore, Neumayer uses the following de nition of pollution havens for his empirical analysis: "Acountry provides a pollution haven if it sets its environmental standards below the socially e cient level or fails to enforce its standards in order to attract foreign investment from countries with higher standards or countries that better enforce their standards. In." (Neumayer, 2001, p. 148).
is paper aims to investigate presence of relationship represented by environmental Kuznets curve and pollution haven in case of selected greenhouse gases in Slovak republic. Greenhouse gases emissions pollute air and cause global warming e ect. Figure 3 illustrates development of selected types of greenhouse gases emissions in Slovak republic from 1990 to 2011 and gure 4 shows share of selected types of greenhouse gasses emissions in 2011.
Legend: carbon dioxide -left axis, other -right axis. All data are LULUCF excluded.

Figure 3. Development of emissions of greenhouse gasses by type of GHGs (Slovak republic, 1990-2011)
Source: author, based on the OECD Environment Statistics data.
Greenhouse gas emissions in the Slovak Republic have continuously declined since the 1990s. Since 1990, the volume of carbon dioxide emissions in the Slovak Republic dropped from 60,745.229 to 37,671.871 in 2011, or about a 38 percent reduction of the original emissions volume (Figure 3). e largest share of greenhouse gas emissions in the Slovak Republic comes from carbon dioxide (83%) and the second largest from methane (9%) (Figure 4).

Figure 4. Greengasses emissions by type, Slovak republic, 2011
Source: author s own compilation based on the OECD Environment Statistics data e analytical focus in this paper is to determine the tenability of two theoretical concepts of environmental economics in the Slovak Republic: the hypothesis of the environmental Kuznets curve and the hypothesis of Slovakia as a pollution haven. While there are many forms of environmental pollution, we have chosen to focus on a few selected sources of environmental pollution in the Slovak Republic, speci cally greenhouse gas emissions. is paper is structured as follows. e second section examines the hypothesis of the environmental Kuznets curves in connection with selected greenhouse gas emissions in the Slovak Republic from the years 1993-2011. e subject of analysis is the relationship between GDP (adjusted for in ation) per capita and the total greenhouse gas emissions per capita, e.g. emissions of carbon dioxide, methane, nitrous oxide, hydro uorocarbons, and sulfur hexa uoride. We seek to demonstrate either that the growth in GDP per capita was proportional to the rise in each type of gas emission, or that gas emissions grew only to a certain in ection point, beyond which the volume of emissions decreased with an increase in GDP per capita. If the volume of greenhouse gas emissions dropped with the growth of GDP per capita, the hypothesis of a mutual positive relationship between economic growth and the quality of the environment would be con rmed. In the third section we will analyze the possible causal relationship between the net in ows of foreign direct investment (FDI) and the volume of selected greenhouse gas emissions in the Slovak Republic. If the volume of greenhouse gas emissions con rms the causal link with the FDI in ows, we could state that Slovakia is in fact a pollution haven. If the hypothesis of an existing environmental Kuznets curve is con rmed but the hypothesis of the Slovak Republic as a pollution haven is not, we could state that the quality of the environment has improved with economic growth and development in the Slovak Republic, and that less strict environmental policies are not used as a tool for attracting foreign investment.

GREENHOUSE GAS EMISSIONS IN THE SLOVAK REPUBLIC AND THE ENVIRONMENTAL KUZNETS CURVE
e relationship between economic growth and environmental pollution is discussed, among other topics, in the literature dedicated to environmental protection, policy, and regulation. e consensus hypothesis is that economic growth in the early stages of development contributes to environmental pollution. However, if the economic growth reaches a certain level in conjunction with a higher GDP per capita, economic growth contributes to the improvement rather than a degradation of the environment, and is an important prerequisite. e graph of this relationship is known as the environmental Kuznets curve. According to this curve, the relationship between the growth of GDP per capita and environmental pollution is non-linear and is shown as an inverted U-shape. e objective of the economic analysis in this section is to verify the existence of an environmental Kuznets curve in the case of selected greenhouse gas emissions in the Slovak Republic from the years 1993-2011.
Authors of prior research used several mathematical functions for their own analyses. One of the rst studies, Sha k et al (1992) estimated parameters of three functions, the linear, quadratic, and cubic those have the following form: where E is the (volume) of emissions of a speci c pollutant, Y is the GDP per capita, i is the country, and t is the year, (Sha k et al, 1992). Sha k estimated the parameters of all three functional forms for each of the ten selected greenhouse gas emissions. Another study, Stern (2004), estimates functional parameters in several studies dedicated to the veri cation of the environmental Kuznets curve (4) where E represents the emissions, P is the population, GDP is GDP, and is the error (Stern, 2004). Hercegová et al (2012) estimated the functional relationship between carbon dioxide emissions and sulfur dioxide emissions in regards to both per capita a GDP per capita (5).
where E represents the emissions, P is the population, and are a constant and a slope coe cient of the estimated curve, respectively (Hercegová et al. 2012).
Chuku (2011) conducted an empirical analysis for Nigeria through utilization of function (6).
where E is environmental pollution expressed as the volume of carbon dioxide emissions, E/P is the carbon dioxide emissions per capita, P is the size of the population, GDP/P is the real GDP per capita, X t is a vector of variables that often in uence the environment, and t is the deterministic trend as a proxy for technical progress (Chuku, 2011). Khyareh (2012) researched the possible existence of an environmental Kuznets curve for carbon dioxide emissions in Iran. Just as Sha k et al (1992), Khyareh estimated the parameters of three functional forms (7), (8), (9), where T is time (Khyareh, 2012).
Similar to Sha k et al (1992), we estimated the parameters of three functional forms that express the relationship between the dependent and independent variables. e dependent variable was the volume of individual greenhouse gas emissions per capita, the independent variables were the GDP (adjusted for in ation) in 2005 US dollars in logarithmic form, in a logarithm of the variable squared, and in a logarithm of the variable cubed. We estimated the parameters of three functional forms: linear (10), quadratic (11) and cubic (12) ghgtpc a a Y (10) where ghgtpc is the total volume of greenhouse gas emissions per capita, Y is the real GDP per capita, ln is the natural logarithm and is the error. In other estimates we have substituted the speci cally selected greenhouse gases for ghgtpc. Table 1 below gives an overview of the variables and data, Table 2 provides an overview of basic statistics of the individual variables. Table 1 Variables and Data Note: 1) LULUCF stands for Land Use, Land Use Change and Forestry. 2) All types of greenhouse gasses are measured in thousands of tons and do not include LULUCF. Table 2 Review of main statistics involved in models At the beginning of the analysis we calculated the pair-wise correlation coe cients which are shown in the correlation matrix in table 3. Estimates of the parameters for linear, quadratic, and cubic functions for the individual greenhouse gases are shown in table 4. Based on the calculations of the parameters, their p-values, and the coe cients of determination, it can be concluded that there is a statistically signi cant negative linear relationship between total greenhouse gas emissions and the growth of GDP per capita.
Estimated parameters of the quadratic and cubic functions do not indicate a di erent type of functional dependence, and from the graph it is clear that the existence of an environmental Kuznets curve is not con rmed. Based on the data above, we can also make a similar conclusion in the cases of carbon dioxide, nitrous oxide, and per uorocarbons. In the case of sulphur hexa uoride, economic growth and development re ected in the growth of income per capita led to a sharp increase in emissions. In regards to methane neither a linear nor a cubic relationship was con rmed among the researched variables, and this is the only greenhouse gas where estimates suggest the presence of an environmental Kuznets curve. Source: author s own calculation e conclusions from the regression analysis con rmed the graphical representations of individual relationships. In gure 5, scatter-plots are shown in individual panels and their lines of regression were estimated using a bandwidth of 0.8. e line of best t shows a monotonically decreasing trend when comparing the total greenhouse gas emissions of carbon dioxide and nitrous oxide with the evolution of real GDP per capita. is suggests that economic growth and an increase in income per capita led to reductions in emissions of the selected greenhouse gases. e decline of per uorocarbons with the growth of income per capita is signi cant. In the case of sulphur hexa uoride neither economic growth nor an increase in income per capita led to a decrease in emissions, but instead led to an increase. Only in the case of methane does the smoothed regression line indicate the ideal shape, similar to the Kuznets curve. In this case, levels of methane in the Slovak Republic rst rose with the growth of GDP per capita, while methane emissions later declined in connection with further economic growth re ected in income per capita.

TESTING POLLUTION HAVEN HYPOTHESIS IN SLOVAKIA
In this section we examine whether the hypothesis of the Slovak Republic as a pollution haven can be con rmed through analyzing emissions of individual greenhouse gases. Speci cally, we are interested in whether there is a causal relationship between FDI in ows and the volume of greenhouse gas emissions in the Slovak republic, and what direction this relationship takes. Direct foreign investments are measured by the amount of net in ows of direct foreign investments as a percentage of GDP. e data are from the World Bank database. Information concerning other variables is given in tables 1 and 2.
In this econometric analysis, we proceeded by rst calculating the pairwise correlation coe cients of FDI, listed in table 5. ese show a relatively high/strong mutual link between foreign investments and greenhouse gas emissions, showing a negative relationship, with the exception of sulphur hexa uoride. Table 5 Correlation matrix However, the correlation coe cients cannot be interpreted as a causal relationship between variables. ey do not provide enough information about whether the change in emission levels for individual greenhouse gases in past periods signi cantly in uenced the future FDI in ows into the Slovak Republic. For this reason, we analyze the causal relationship between researched variables through the Granger causality test. A prerequisite for testing Granger causality is the stationarity of a time series. erefore, in the rst step we tested the time series for the presence of a unit root.
Granger causality between variables is signi cantly a ected by the number of delays in the model. erefore, in the second step we determined the correct delay. In the third step we estimated the parameters of autoregressive models and their signi cance by the vector autoregressive method. In the last step, we tested for the existence and the direction of a causal relationship between FDI and individual greenhouse gases.
A prerequisite for the application of the Granger causality test is stationarity of the time series, where the time series is stationary if it has no unit root. We tested the presence of a unit root in the times series of variables though the test statistic of an adapted Dickey-Fuller test, ADF test. e individual variables (the net in ows of direct foreign investments and individual greenhouse gas emissions) were tested rst for the presence of a unit root null and the rst di erential in selected cases. e results of testing the time series for the presence of a unit root are in table 6.
Stationarity for fdintinfpr was also tested through the ADF test. e null hypothesis of the ADF test is as follows: the test variable (the time series) has a unit root, meaning that it is not stationary. e rule of thumb is, if the value of the calculated ADF statistic is larger than the table of critical values for a signi cance level of 1%, 5% or 10%, then the null hypothesis is rejected. In our case, the value of the ADF statistic for fdintinfpr null is greater than the table of critical values. We therefore reject the null hypothesis that the time series has a unit root and we accept the alternative hypothesis that the time series does not have a unit root, meaning that the time series is stationary. is variable thus quali es for the application of the Granger causality test. Table 6 ADF test results Z(t) In contrast, the time series for the net FDI in ows at current prices, fdintinfcr, does not meet the stationarity requirement, because the test for the presence of the unit root null indicated the presence of a unit root. Without further adjustments it is not possible to use this variable to verify Granger causality.
rough testing for the presence of a unit root null in the time series we found that only the time series for the variables ghgtpc, coxpc, noxpc, and fdintinfpr are stationary. Conversely, pfcpc, soxpc, mtnpc and even fdintinfcr are non-stationary, i.e. have unit roots. erefore, we chose only the stationary variables, i.e. a time series of the percentage of FDI to GDP, total greenhouse gas emissions of carbon dioxide and nitrous oxide, for testing Granger causality. We performed the analysis for the period 1993 to 2011. e result of the Granger causality test between variables is signi cantly in uenced by the number of delays, which are to be applied in the vector autoregression analysis (VAR). erefore, we rst conducted a preliminary vector auto regression analysis with the selected variables derived from veri ed Granger tests. Using the statistical criteria, we then made the selection of the optimal length of the delay. Econometric science o ers several criteria for the identi cation of the correct length delay, such as the Akaike information criterion (AIC), the Hannan-Quinn information criterion (HQIC), and the Schwarz Bayesian information criterion (SBIC). In order to choose the correct number of delays to be applied in the Granger causality test we calculated the values of both test statistics for the FPE, AIC, HQIC, and SBIC. e resulting values for each criterion are in table 7. For the Granger causality test we chose the number of delays for which the calculated value of the test statistic is the lowest. Values of the AIC, HQIC, and SBIC criteria indicate the suitability of four lags, the FPE value indicates the suitability of two lags. e calculated values for the majority of the criteria suggest that it is best to select four lags when testing Granger causality.
In the next step we performed the vector regression analysis of the selected variables and we chose four lags, staying consistent with the AIC, HQIC, and SBIC criteria. Table 8 Vector autoregression analysis -results Source: author´s own calculation Next, we verify the Granger causality between the net FDI in ows as a percentage of GDP and selected greenhouse gas emissions per capita in the Slovak Republic.
e Granger causality test has the null hypothesis that the delay variable does not cause changes in the second variable. If the p-value is greater than 5%, we cannot reject the null hypothesis and we accept the hypothesis that the delay variable does not share a Granger causal relationship with the second variable. is means that based on the past development of one variable it is not possible to predict the future development of the second variable. e results of the veri cation of Granger causality between net FDI in ows and selected greenhouse gas emissions are in table 9. Table 9 Granger causality test results First we estimated the parameters of the vector auto regression equations for the relationship between carbon dioxide and the percentage of net FDI in ows to GDP. We then applied the Granger causality test. e existence of causality between FDI in ows and the volume of carbon dioxide emissions per capita was not con rmed in either direction. Net FDI in ows into the Slovak Republic as a percentage of GDP do not share a causal relationship with the amount of carbon dioxide emissions in the present period and the past four years. By the same method we examine the potential causal relationship between FDI in ows and greenhouse gas emissions per capita in the present period and the four previous periods. We estimated the parameters of the model of vector regression analysis and then we applied the Granger causality test.

CONCLUSION
A statistically signi cant linear negative relationship between total greenhouse gas emissions and growth of real GDP per capita was con rmed. e estimated parameters of the quadratic and cubic functional forms do not indicate a di erent kind of functional dependence, and from the graph it is clear that the existence of an environmental Kuznets curve was not con rmed. It is also possible to make the same conclusion in the relationships of carbon dioxide, nitrous oxide, and per uorocarbons emissions and economic growth (GDP per capita).
In the case of sulphur hexa uoride, economic growth and development re ected in the growth of income per capita leads to a sharp increase in emissions. Only in the case of methane was neither a linear nor a cubic relationship con rmed between the variables of interest. A quadratic relationship was con rmed suggesting that this is the only type of greenhouse gas where estimates indicate a relationship which follows an environmental Kuznets curve.
We examined the possible causal relationship between net FDI in ows and emissions of select greenhouse gases. e objective was to verify whether the volumes of select greenhouse gas emissions are a signi cant determinant of FDI in ows into the Slovak Republic. e parameters of the coe cients of three models estimated by vector autoregression analysis are not statistically signi cant. Based on the calculated values of the Granger causality test we conclude that the volume of greenhouse gas emissions per capita in the present period and in any of the previous four periods has no e ect on the amount of net FDI in ows as a percentage of GDP in the Slovak Republic. e conclusions of this empirical analysis allow us to conclude that with economic growth in the Slovak republic, the quality of the environment improves, and there is no evidence that the Slovak Republic lowers environmental standards in order to increase its competitiveness and attract foreign investment.