1 Sana El Harbi (Corresponding author) University of Sousse, and URFQ, Sousse, Tunisie E-mail : [email protected] Gilles Grolleau Montpellier SupAgro and LAMETA, UMR1135 2, place Pierre Viala, Bât. 26, 34060 Montpellier Cedex 1, France. E-mail : [email protected] Are People Living in Countries Emitting Less CO2 Happier? Abstract In this paper we investigate whether people living in countries emitting less per capita CO2 are happier. We begin by considering a panel data approach to address this question. The results we obtain in this framework imply that there is no influence of CO2 emissions on happiness even for the developed countries. However, the multicollinearity problem coupled with this surprising result prompted us to consider a more appropriate technique. We use then the partial least square regression (PLS). In this case when considering the whole sample, per capita CO2 emissions have no effect on the average happiness in countries. However, when considering separately the two sub-samples of developing and developed countries, the PLS results show that per capita CO2 emissions affect the happiness in developed countries, but have no perceptible influence on the happiness in developing countries. This difference of the effect of CO2 emissions on happiness coupled with the importance of revenue can constitute a contentious issue between developing and developed countries. 2 1. Introduction Since decades, economists and psychologists studied the relationship between wealth and happiness. They have generally concluded that ‘wealth increases human happiness when it lifts people out of abject poverty and into the middle class but that it does little to increase happiness thereafter’ (Gilbert, 2006; Layard, 2005). While the impact of several other factors on happiness such as democracy, inflation or personal characteristics has been examined1, very little is know about how polluting emissions of a given country affect the happiness of its inhabitants. The purpose of this paper is to fill this gap by exploring empirically whether countries with lower per capita CO2 emissions experience are happier after controlling for other more conventional factors. At the individual level, Rehdanz and Maddison (2008) found that in Germany, even when controlling for a range of other factors higher local air pollution and noise levels significantly diminish subjective well-being. Except Welsch (2002, 2006) who found, among other results that nitrogen dioxide pollution has a statistical significant and negative impact on overall happiness, empirical evidence is scarce. We investigate the robustness of Welsch’s findings (2002, 2006) over a larger and more diversified sample of countries including developing and developed countries (N=28). Indeed, the Welsch sample included only developed countries (N=10). Nevertheless, the advantage of considering more countries is somewhat offset by stronger limitations on the pollution type we can study. Indeed, availability constraints on other pollutants data lead us to investigate CO2 emissions for which reliable time series are available. In doing so, we follow several previous studies (e.g., Hoffmann et al., 2005 and references therein) which argued that CO2 could be used as a valid proxy for pollution. As the primary greenhouse gas responsible for global warming, regulation of CO2 emissions has been an important inter-governmental issue. A ‘comparison between the top 20 polluters of the world based on CO2 and SO2 emission in 2000 finds 15 countries being in both lists. Further, the correlation coefficient between three major pollutants (CO2 with NO and SO2) among 111 countries in 1990 were 0.9529 and 0.9536 respectively. The high level of correlation between the three pollutants provides sufficient evidence to show that the use of CO2 as a proxy to measure pollution levels is valid (Hoffmann et al., 2005). 1 Interestingly, the effects of weather, climate, noise and other characteristics on happiness have also been investigated and have been found to exert a significant influence on reported life satisfaction (Rehdanz and Maddison, 2005; Brereton et al. 2008; Weinhold, 2009). 3 The remainder of this paper is organized as follows. Data are described in section 2. Section 3 presents our empirical analysis and section 4 discusses our results. Section 5 concludes. 2. Data Our sample includes 28 countries (13 developing and 15 developed) over 9 years covering the period from 2000 to 2008. Several environmental indicators exist for OECD countries (nitrogen dioxide, lead emissions, sulfur dioxide, carbon dioxide, etc.) but this is not the case for developing countries. Our willingness to investigate a larger set of countries and the severe limitations on data lead us to consider a unique type of pollution, i.e., per capita CO2 emissions per year, expressed in metric tons. Per capita CO2 emissions mainly indicate the emissions from fossil-fuel burning, cement manufacture and gas flaring. The choice of per capita emissions is justified by the fact that it is deemed to be an adequate proxy for environmental quality for nations with similar population densities and environmental regulations. The world database of happiness2 offers different measures of happiness drawn from surveys on representative sample of population. Given that happiness is fundamentally an individual issue, the average happiness is about the happiness in a country and not of a country. The most complete happiness database available is the four step verbal life satisfaction where individuals are asked to respond to the following question: How satisfied are you with the life you lead?”. Responses are scaled from (1): not satisfied at all to (4): very satisfied. In addition to our per capita CO2 emissions variable, we introduce some other explanatory variables. While there are several candidates, i.e., factors that are likely to affect happiness, our choice of additional explanatory variables is dictated by related literature, conditioned by data availability for the considered sample and technical constraints namely in terms of collinearity. Concretely, we included the three following variables in our empirical investigation: • Income (GDP per capita): while there are numerous discussions on how income affects happiness, there is a strong consensus among social scientists that income is a major determinant of happiness (e.g., Frey and Stutzer, 2002) 2 Veenhoven, R., World Database of Happiness, Erasmus University Rotterdam. Available at http://worlddatabaseofhappiness.eur.nl/hap_nat/nat_fp.php Assessed at March, 21, 2010. 4 • Inflation rate (INFL) which captures the purchasing power of individuals. As intuitively expected, earlier studies found that inflation affects negatively subjective well being (e.g., Di Tella et al., 2003; Wolfers, 2003, Graham and Pettinato, 2001; Alesina et al., 2004) • Share of exports in total GDP (EXP) which captures the trade openness which is supposed to boost the happiness level (e.g., Ming-Chang, 2009). The source, definition and descriptive statistics of variables are briefly presented in table 1. Table 1. Definition of the variables and descriptive statistics (N=226) Variable and acronyms Happiness level (H) Source World database on Descriptive statistics Mean SD Min Max 2.914 0.395 1.86 3.64 0.835 26.172 Happiness Per capita CO2 emissions World Bank .World (metric tons per capita) development (Poll) indicators Income per capita (in natural FMI, World logarithm) (Y) 8.043 4.876 9.285 1.094 6.913 10.936 46.581 29.955 9.54 181.313 4.549 7.731 -1.319 96.1 Economic Outlook Share of export in total GDP World Bank Data (EXP) Indicators Inflation rate (INFL) FMI, World Economic Outlook As GDP per capita is expressed in thousands of dollars while the endogenous variable (happiness) is a four step measure, then, for reasons of data adequacy, we have carried out the regressions by considering the logarithmic transformation of this independent variable. 3. Empirical strategy In order to investigate the relationship between CO2 emissions and happiness in nations, we at first glance use of a panel data approach. We estimate then the following equation: 5 H it = α i + κ t + β 0 + β 1 Pollit + β 2 Y it + β 3 Z it + uit (1) Where Hit is the average happiness in a country i (i = 1, 2,…, N) at period t (t = 1, 2,…,T). In order to capture the effect of polluting emissions we consider per capita CO2 emissions, Pollit . The α i and κ t represent country and year specific effects, and the uit are the idiosyncratic errors. The Yit is the natural log of per capita GDP at time t in country i. Eq. (1) also includes Z, a vector of additional explanatory variables, namely the inflation rate, and the share of export in total GDP. 4. Results and discussion The basic model in Eq. (1) was estimated using the well known fixed effect and random effect models. Welsch (2002) observes that it is difficult to isolate the effect of pollution because of high correlation between income and pollution. Our data display this problem too. The correlation matrix indicates a correlation coefficient between income per capita and CO2 emissions per capita around 0.68 . Generally, to deal with the collinearity problem, one can either increase the sample size or eliminate the correlated variables. Nevertheless these strategies are not appropriate here. Indeed, given the theoretical importance of income in explaining the happiness in nations, the option of eliminating it is simply not envisageable. The sample extension alternative is also impossible regarding the data availability (see section 2). Though, As long as it is not perfect, multicollinearity does not actually bias results, it rather provides large standard errors in the related independent variables and instable estimated coefficients. Nevertheless, conscious of the multicollinearity problem we begin by estimating Eq (1) with classical panel data tools, hoping that this problem will be mitigated by our data size. The regression results are presented in Table 2. Table2. Empirical results on the whole sample (N=226) H1 H2 H3 Random Fixed Random Fixed Random Fixed effect effect effect effect effect effect Poll -0.0019 0.0062 -0.009 0.0022 -0.0122 -0.00513 GDP 0.3358*** 0.745*** 0.3333*** 0.702*** 0.304*** 0.574 *** Exp 0.0023* 0.015 0.003*** 0.0039 ** Infl - - -0.0081*** -0.0079*** 6 Intercept -0.20072 -4.122*** -0.236 -3.76*** 0.0413 -2.5916*** R2 overall 0.5415 0.5422 0.5448 0.5492 0.537 0.5520 R2 within 0.2162 0.2185 0.2055 0.2184 0.3231 0.3470 R2 between 0.5728 0.5730 0.5702 0.5761 0.5423 0.5652 F tests 28.09 - 27.43 23.91 (0.000) (0.000) (p-value) (0.000) Hausman 20.05 16.52 13.81 (P-value) (0.000) (0.0009) (0.0079) 226 226 Number of 226 observations *, ** and *** stand for significance at the 10, 5 and 1 percent level, respectively Model H1 expresses the average happiness in nations as a function of CO2 emissions per capita and GDP per capita. In this case, the GDP per capita is significant with the positive expected sign that is when the GDP per capita increases, happiness increases too. However, CO2 emissions per capita do not seem to affect happiness in nations. Models (H2) and (H3) include additional explanatory variables. In all these models, the coefficient values have remained relatively stable. The pollution variable has the expected sign but non-significant in H2 and H3 too. By contrast, income, inflation and exports coefficients are found to be significant with the expected signs. Although all these variables are significant, their effect on happiness is largely differentiated: the income per capita has by far the larger effect on the happiness in nations. The fact that we do not find any significant influence of CO2 emissions per capita on happiness is intriguing, especially regarding the results of Welsch (2006) who found that air pollution significantly explains the inter-country and inter-temporal differences in subjective well being. Indeed, this author found that individuals are willing to pay for the improvement of air quality. He found that for the period 1990-1997, valuation of air quality improvements amounts to almost $900 per capita per year for the nitrogen dioxide and more than $1400 per capita per year in the case of lead emissions. Moreover, it is widely admitted that citizens of developed countries are more and more concerned by environmental issues. These considerations encourage us to refine our empirical analysis. Consequently, we analyze whether per capita CO2 emissions affect happiness by considering separately developing and developed countries. 7 The descriptive statistics and correlation matrix related to these sub-samples are provided in the appendix (A.2-A.5). Table3. Empirical results on developed countries (N=131) H1 H2 H3 Random Fixed Random Fixed Random Fixed effect effect effect effect effect Effect Poll 0.0033 0.0041 0.0013 0.009 0.00035 0.0018 GDP 0.553*** 0.614*** 0.536*** 0.603*** 0.494*** 0.5767*** Exp - - 0.0007 0.00044 0.00135 0.0011 Infl - - - -0.012* -0.0134* -2.54** -3.16*** -2.387** -3.058* -1.945* -2.78* R2 overall 0.311 0.3114 0.3281 0.321 0.3185 0.3144 R2 within 0.199 0.199 0.1992 0.200 0.220 0.221 R2 between 0.27 0.2705 0.2903 0.2819 0.2798 0.274 - 128.89 - 122.41 - 111.34 Intercept F tests (p-value) (0.000) (0.000) (0.000) Hausman -0.98 1.47 0.4 (P-value) . (0.689) (0.98) 131 131 131 Number of observations *, ** and *** stand for significance at the 10, 5 and 1 percent level, respectively Surprisingly, even when considering the developed countries, the effect of a polluted air on the reported happiness remains non significant. Obviously, this is inconsistent with Welsh’s (2006) results. However, export coefficient turns out to be not statistically significant. This is probably due to the fact that these developed economies are already open and that the differences in the share of exports in overall GDP are negligible between countries and across years. This is supported by the very small standard deviation indicated in the descriptive statistics (see appendix A.3). As expected, inflation is found to have a negative and significant effect on the average happiness in nations. This finding provides support for inflationcontrolling policies, all other things being equal. In the case of developing countries The random effect model displays a significant negative effect of Poll on the happiness, however the Hausman test strongly supports the fixed effect model where, like the developed countries case, per capita CO2 emissions coefficient is not 8 significant. Meanwhile, the GDP per capita remains the most important factor that influences positively the average happiness. (Table 4) Table 4. Empirical results (developing countries) H1 H2 H3 Random Fixed Random Fixed Random Fixed effect effect effect effect effect Effect Poll -0.351 0.0422 0.035 0.042 -0.049* 0.001 GDP 0.442*** 0.7289*** 0.38*** 0.637*** 0.34*** 0.4723*** -0.00025 -0.0038 -0.0028 0.0093** -0.0077*** -0.0085*** Exp Infl Intercept -0.823 -3.645*** 0.322 -3.031* 0.054478 -1.646 R2 overall 0.344 0.2401 0.2638 0.0965 0.257 0.086 R2 within 0.201 0.226 0.1912 0.2301 0.338 0.39 R2 between 0.419 0.2663 0.3860 0.1388 0.2905 0.087 F tests (p-value) 10.67 7.66 9.5 (0.0000) (0.000) (0.00) Hausman 8.56 14.71 19.14 (P-value) (0.013) (0.002) (0.000) 89 89 89 Number of observations *, ** and *** stand for significance at the 10, 5 and 1 percent level, respectively As expected, inflation has a negative significant effect on happiness. The results regarding openness are not robust over specifications and we cannot draw any clear cut conclusions. One very plausible explanation of the discrepancy of our results compared to Welsh (2002) results can be due to the pollutant choice. Welsh (who considered Nitrogen, Particles, and lead) argues that the positive association he obtained between happiness and environmental quality is due to the fact that the countries of the sample he considered are “on the downward -sloping part of the Environmental Kuznets curve”. Whereas numerous panel and cross-country studies provided evidence in favor of Environmental Kuznets curve (EKC) in some countries and for several pollutants (Grossman and Krueger, 1991; Shafik,1994), there were also some evidence against this hypothesis for 9 a number of measures of environmental degradation (Rothman and de Bruyn, 1998). 3 Indeed, several studies have challenged the robustness of estimated environmental Kuznets curve relationships namely for per capita CO2 emissions (Harbaugh et al., 2002, Millimet et al. 2003, Perman and Stern, 2003). More precisely, if we examine how the per capita CO2 emissions is related to the income (see table 5) we found that our per capita CO2 emissions is significantly positively related to the level of prosperity4. And this is true either for the entire sample or for our two sub-samples. In other words, the countries considered in this paper are, contrary to those considered by Welsh (2002), “still” on the upward – sloping part of the environmental Kuznets curve. Note also that this positive correlation is particularly strong and highly significant. Table 5: Results of the regression of per capita emissions on GDP Entire Sample Developped countries Developping countries GDP 2.63*** 5.33*** 1.55*** intercept -12.98*** -43.69*** -7.67*** R overall 0.51 0.48 0.34 *, ** and *** stand for significance at the 10, 5 and 1 percent level, respectively This encourages us to sacrifice the richness of the panel data structure to use specific techniques that can better deal with the multicollinearity problem. Hence we will use in the following the Partial Least Square Regression. 3 More precisely the application of EKC analyses to greenhouse gas emissions, such as carbon dioxide (CO2), has raised several important questions. First, some empirical studies have estimated an inverted-U shape of per capita CO2 emissions with respect to per capita income, but with a peak in this function occurring well outside the range of incomes in the studies’ samples. 4 We also regressed the percapita emissions on GDP and square GDP (results can be obtained under request). The obtained results show the non existence of Kuznet like curve. 10 5- Partial least square results 5-1 Partial Least Square (PLS) regression PLS regression proposed by Wold et al (1984), is a technique that combines features from factorial analysis and multiple regression. It is particularly valuable when the degree of freedom is very small (it can be also used even when the number of predictors is larger than the observation number) or when the explanatory variables display a high degree of collinearity. The goal of PLS regression is to predict Y from a set of exogenous variables X and to describe their common structure. It determines components from X that are also relevant for Y. More precisely, this regression seeks to find a set of components that performs a simultaneous decomposition of X and Y with the constraint that these components explain as much as possible the covariance between X and Y5. After having determined the components, the PLS regression technique implies the decomposition of X to predict Y. To appreciate the overall model the following statistics (among others) are generally used: • R2X(cum): the cumulative fraction of the variantion of X explained by the components. • R2Y (cum): the cumulative fraction of the variation of the Y explained explained by the component • Q2 (cum): the cumulative predicted fraction of the variation of X. We have run the PLS regression with the software SIMCA-P, which provided the PLS components and coefficients as well as several plots in different plans (t1/t2; t1/t3, w1/w2 ..ect). For the sake of presentation simplicity, we will examine in the following w*(c1)/w*(c2) plan which is described by Tenenhaus (1998) as a privileged tool that allows the interpretation of components t1/t2 related to their ability to predict Y.6 This plot is obtained by the superposition of the loadings (w1/w2) and (c1/c2)7, it discribes the relationship between the endogenous and exogenous variables . 5-2 Developped countries Graph (1) shows that all our considered variables in this case are positively correlated with the first loading w1(c1). However, happiness, GDP per capita and inflation are positively correlated with the second component whereas per capita CO2 emissions and export are 5 Let’s define t : the matrix of scores of X. U the score matrix of Y. The component ci are the weights representing the correlation between Y and t. the matrix W represents the correlation between X and U. 6 The other plots and lists are available under request. 7 All our components are significant according the cross validation rule. 11 negatively correlated with this second loading this negative correlation is much more important for the variable POLL. This configuration allows us to suspect a possible negative relationship between the happiness in nations and per capita emissions. Developed countries w*c[1]/w*c[2] plan 0.60 X Y GDP INFL Happiness w*c[2] 0.40 0.20 0.00 -0.20 Exp -0.40 POLL 0.20 0.30 0.40 0.50 0.60 0.70 w *c[1] Graph 1 The PLS regression coefficients (graph 2) which measure the contribution of each exogenous variable to the construction of the variable Y give support to this contention. Indeed, the PLS equation (expression 2) for the considered developed countries is as follows: ) H = 10.8+0.719GDP+0.32INF+0.04Exp-0.126POll (2) 0.60 0.40 0.20 Var ID (Primary) R2X(cum)=0.877 ,R2Y (cum)=0.4 , Q2 (cum)=0.368. Exp POLL INFL 0.00 GDP CoeffCS[3](Happiness) Developped countries , coefficients plot 12 Graph ( 2 ) The VIP values reflect the importance of terms in the model both with respect to y (its correlation with the responses) and with respect to X (in the projection). Terms with large VIP (larger than 1) are the most relevant for explaining Y. The VIP values of all our considered variables are greater or close to one except for the inflation variable. (see table 6) Variable VIP GDP 1.31 Poll 1.025 Exp 0.92 INFL 0.52 Table ( 6 ) the VIP values This result suggests that prosperity is the most influential variable on happiness. That is when income increased happiness increased too. However per capitaCO2 emissions have a negative, but small, impact on happiness. 5-3 PLS Results for Developping countries 13 Developping countries w*c[1]/w*c[2] plan X Y GDP 0.20 w*c[2] 0.00 happiness INFL -0.20 -0.40 POll -0.60 -0.80 EXP -0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 w *c[1] Graph 3. In the w*(c1)/w*(c2), (graph 3 )and contrarily to the other considered variables, GDP and H have the same position regarding the two components i.e they are positively correlated to both of the two components. EXP is negatively correlated to the two considered components whereas the situation of par capita CO2 emissions is more ambiguous in that it is as happiness positively correlated with the first loading but negatively correlated with the second. Developping countries, Coefficients plot 0.20 0.00 Var ID (Primary) R2X(cum)=0.733 ,R2Y (cum)=0.404. , Q2 (cum)=0.365 Graph (4) Variable VIP GDP 1.436 Poll 0.514 GDP POll INFL -0.20 EXP CoeffCS[2](happiness) 0.40 14 EXP 0.801 INFL 1.01 Table (7 ) the VIP values The PLS regression coefficients (equation 3) are represented in graph (4). ) H = 7.81+0.45GDP-0.26INF-0.33EXP+0.0039Poll ( 3 ) Obviously, GDP is the most influential variable to predict happiness (see graph 4). Moreover the VIP values (see table 7) indicate that only GDP and INF are the most relevant to explain happiness. Per capita CO2 emissions variable do not affect happiness. 5-4 The overall sample Developping and Development countries w*c[1]/w*c[1] plan X Y GDP 0.50 w*c[2] happiness INFL 0.00 POLL -0.50 Exp -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 w *c[1] Graph 5 The points pattern in the plan w*(c1), w*(c2) for the overall sample (graph 5) look very similar to the developping countries case, execpt for the export variable. The PLS equation is indicated in the expression ( 4 ) ) H = 7.37+0.66GDP-0.048INF+0.007Exp+0.092Poll ( 4 ) The PLS regression for the overall sample (see graph 6 ) confirms the primordial importance of the economic prosperity in explaining the happiness in countries. The coefficients of the other variables are by far of a minor importance. The VIP table (table 8) indicates that only the variables GDP and POLL have VIP values greater than one. They are the most relevant in explaining happiness. Notice the positive effect of per capita CO2 pollution that is when per capita emissions increase, the happiness increases too. 15 Developping and Developped countries coeficients plot CoeffCS[3](happiness) 0.60 0.50 0.40 0.30 0.20 0.10 GDP POLL Exp INFL 0.00 Var ID (Primary) R2X(cum)=0.895. ,R2Y (cum)=0. 0.552. , Q2 (cum)= 0.543 Graph 6 Variable VIP GDP 1.41 Poll 1.0049 Exp 0.64 INFL 0.749 Table (8 ) the VIP values 5-5 Discussion of the PLS results Unequivocally, the major determinant of happiness in nations is economic prosperity that prevails in a country. But our PLS results also show that per capita emissions also affect happiness too. More precisely, it • decreases the happiness in developed nations but • increases the happiness in the overall considered countries and • seems to be irrelevant for the developing countries. However, the influence of the openness on happiness is questionable. Indeed, all VIPs values for either the overall sample or developed or developing countries are less than one for the variable EXP. However, for the developing countries inflation is found to be relevant with the expected negative impact on happiness. 16 6. Conclusion We analysed whether per capita CO2 emissions affect the average happiness in countries. Our results suggest that per capita CO2 emissions effect on happiness depends on the development stage of countries. Moreover, even in developed countries the effect of per capita income on happiness is largely more important than any other influence. Our results may help understanding the modest outcomes of the Copenhagen summit. Indeed, beyond technical disputes, our study suggests that the discrepancy between developed and developing countries may originate from the benefits in expected happiness that each party anticipates. Indeed, because the happiness in countries is by far more influenced by income than by any other dimension, developed countries are not willing to forgo economic growth for lower polluting emissions. Developing countries where air quality has no detectible effect on their happiness are reluctant to accept any environmental deal unless it can proved that this arrangement will increase their income per capita. Our happiness approach suggests that the failure of the Copenhagen summit may be explained by this bottleneck situation. It is possible that an equilibrium point will be reached in the future, when the green values will be more rooted in both advanced and less advanced economies. However, the problem is that meanwhile irreversible changes threatening humankind could occur before reaching this equilibrium point. Lastly, a promising extension of our study would be to investigate the robustness of our results by using other reliable proxies for pollution and environmental quality in a larger number of countries, encompassing several environmental dimensions and over a sufficient period. Fortunately several efforts are currently deployed to fill these gaps. Rather than replacing other analyses, we believe that studies using happiness in countries can usefully complement them and provide new insights to policymakers. 17 References Alesina. A., Di Tella, R., MacCulloch, R. , 2004. Inequality and happiness: Are Europeans and Americans different. Journal of Public Economics. 88, 2009–2042. Brereton, F., Clinch JP., Ferreira, S. , 2008. Happiness, Geography, and the Environment. Ecological Economics. 65, 386-396. Di Tella, R., MacCulloch, RJ., Oswald, A., 2003. The macroeconomics of happiness. The Review of Economics and Statistics. 85 ,4 , 809–827. Frey BS and Stutzer, A.,2002. Happiness and economics, University Press, Princeton, USA. Gilbert, D.,2006. Stumbling on Happiness, Knopf, New York, USA. Graham, C., Pettinato, S., 2001. Happiness, markets and democracy: Latin America in comparative perspective. Journal of Happiness Studies. 2, 237–268. Grossman, G. M. & Krueger, A. B. 1991. 'Environmental impacts of a North American Free Trade Agreement'. NBER Working Paper Series, 3914. Hoffmann, R.; Lee, CG., Ramasamy, B., Yeung, M., 2005. FDI and Pollution: a Granger Causality Test Using Panel Data, Journal of International Development. 17, 311-317. Harbaugh, W.T., A. Levinson, and D.M. Wilson. 2002. Reexamining the Empirical Evidence for an Environmental Kuznets Curve. Review of Economics and Statistics 84(3): 541-551. Layard, R. , 2005. Happiness: Lessons from a new science. Allen Lane, London, Great Britain. Millimet, D.L., J.A. List, and T. Stengos. 2003. The Environmental Kuznets Curve: Real Progress or Misspecified Models? Review of Economics and Statistics 85(4): 1038-1047. Ming-Chang ,T., 2009. Market Openness Transition Economies and Subjective Well-being. Journal of Happiness Studies. 10, 523-539 Perman, R. and D.I. Stern. 2003. Evidence from Panel Unit Root and Cointegration Tests that the Environmental Kuznets Curve Does Not Exist. Australian Journal of Agricultural and Resource Economics 47(3): 325-347. Rehdanz, K., Maddison., D., 2008. Local environmental quality and life-satisfaction in Germany. Ecological Economics. 64 ,787-797. Rothman, D. S. & de Bruyn, S. M., 1998. 'Probing into the environmental kuznets curve hypothesis', Ecological Economics, 25, 143-145. Shafik, N., 1994. 'Economic development and environmental quality: An econometric analysis', Oxford economic papers, 46, 757-773. Tenenhaus, M., 1998. La régression PLS. Paris: Technip. 18 Welsch H., 2002. Preferences over prosperity and pollution: Environmental valuation based on happiness studies. Kyklos. 55, 473–494. Welsch H., 2006. Environment and happiness: Valuation of air pollution using life satisfaction data. Ecological Economics. 58, 801–813. Weinhold, D., 2008. How big a problem is noise pollution? A brief happiness analysis by a perturbable economist" LSE Working Paper. Wold, S., Ruhe, A., Wold, H., & Dunn, W. J.,1984. The Collinearity Problem in Linear Regression. The Partial Least Squares (PLS) Approach to Generalized Inverses, 5(3), 735743. SIAM. Retrieved from http://link.aip.org/link/?SCE/5/735/1 Wolfers, J., 2003. Is business cycle volatility costly. Evidence from surveys of subjective well-being. International Finance. 6,1–26. 19 Appendix A.1 Correlation matrix (developing and developed countries) Poll GDP Exp Poll 1 GDP 0.68 1 Exp 0.47 0.19 1 Inf -0.26 -0.42 -0.053 A. 2 Descriptive statistics (developing countries) Variables Mean S.D. Min Max H 2.63 0.336 1.86 3.32 Poll 10.513 1.184 8.417 13.321 GDP 8.256 0.698 6.913 9.5462 Exp 42.35 18.396 10.89 85.38 Infl 7.136 10.736 -1.319 96.1 A.3 Descriptive statistics (developed countries) Variables Mean SD Min Max 3.125 0.28 2.46 3.46 Poll 10.62578 4.69 5.59 26.17 GDP 10.162 0.356 9.305 10.936 Exp 50.057 36.549 9.54 181.31 Infl 2.307 1.134 -0.887 5.254 Exp Inf H A.4 Correlation matrix (developed countries) Poll GDP Poll 1 GDP 0.69 1 Exp 0.51 0.41 1 Inf 0.12 -0.27 0.22 1 20 A.5 Correlation matrix (developing countries) Poll GDP Exp Poll 1 GDP 0.53 1 Exp 0.65 0.27 1 Inf -0.17 -0.27 -0.10
© Copyright 2026 Paperzz