10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Received: 25 March 2022 Revised: 12 July 2022 Accepted: 30 July 2022 DOI: 10.1002/sd.2380 RESEARCH ARTICLE The determinants of the inequality in CO2 emissions per capita between developing countries and their implications for environmental policy Emilio Padilla Rosa 1 | Evans Jadotte 2 1 Department of Applied Economics, Autonomous University of Barcelona, Abstract Bellaterra, Spain We analyze the differences in CO2 emissions from fuel combustion per capita 2 Macro Economics, Trade and Investment Global Practice, World Bank, Port-au-Prince, between developing countries and how these are influenced by a series of affluence, Haiti structural, demographic, and climatic variables. First, a regression analysis provides Correspondence new evidence on the determinants of CO2 emissions in developing countries. We Emilio Padilla Rosa, Department of Applied find an N-shaped relationship with GDP per capita and a negative impact of the agri- Economics, Autonomous University of Barcelona, 08193 Bellaterra, Spain. culture share and average daily minimum temperatures, while urbanization and the Email: emilio.padilla@uab.es share of potentially active population would be positively correlated with emissions Funding information per capita. Second, by using the regression-based inequality decomposition method, Global Facility on Growth for Development we indicate the weight of each significant determinant in explaining the inequality in Trust Fund CO2 emissions per capita between developing countries. The main contributors are economic affluence and the potentially active population, in this order. We study the relevance of each factor in the changes experienced by inequality over time. Some of our results contrast with previous evidence for more heterogeneous samples. We derive some relevant implications for environmental and energy policy in developing countries. KEYWORDS CO2 emission drivers, CO2 emission inequality drivers, CO2 inequality, developing countries, environmental policy, regression-based inequality decomposition, sustainable development 1 | I N T RO DU CT I O N contributed to increased economic inequality by slowing, to some extent, the opposite trend that was taking place in recent decades Developing countries have a special interest in the success of global (Diffenbaugh & Burke, 2019). Second, international mitigation agree- policies to mitigate climate change. First, because according to most ments involving transfers of resources and technology to low-income projections (IPCC, 2014; Stern, 2006) they will be among the most countries may also facilitate the development of these countries in a seriously affected and will suffer the worst expected impacts of cli- clean (low-carbon) way. Low-income countries may have the opportu- mate change. Both their geographical location and their lower nity to adopt a low-emission development model, taking advantage of resources and capacity to adapt make several developing countries existing technologies already proven in rich countries, and avoid the more vulnerable to climate change impacts. Moreover, existing climate high-emission development phase experienced by industrialized coun- change has already disproportionally affected poor countries and tries (Padilla, 2017). This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2022 The Authors. Sustainable Development published by ERP Environment and John Wiley & Sons Ltd. Sustainable Development. 2023;31:151–169. wileyonlinelibrary.com/journal/sd 151 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 152 PADILLA ROSA AND JADOTTE The contribution to climate change is highly unequal, with some determinants to the components of the identity used. An alternative high-income countries having CO2 emissions per capita that are twen- appealing method is the application of the regression-based inequality tyfold those of several low-income countries. The unequal contribu- decomposition approach, developed by Fields (2003) for the analysis tion to the problem may lead some representatives of developing of income distribution, which was first applied to the international countries to argue that all the mitigation efforts should be made by CO2 inequality by Duro et al. (2017). The main advantage of this those that are more responsible and capable to act. However, an method with respect to the previous methods used in this literature is international agreement to mitigate climate change can only be effec- that it enables to test the contribution to inequality of all the factors tive with the participation and cooperation of both developed and that are previously identified as relevant determinants of emissions developing nations. The analysis of the inequality in CO2 emissions through an econometric regression. and its driving forces is of great relevance to determine the differ- A few analyses on CO2 emissions distribution have specifically ences in the degree of responsibility, its trajectory, its causes, as well focused on the inequality between developed countries, such as the the fairness and feasibility of different emission distribution agree- OECD or the European Union (Padilla & Duro, 2013), to study its evo- ments. The greater the level of inequality in both emissions and their lution and the determinants of this inequality by using the multiplica- driving forces, the greater the differences that tend to appear in the tive decomposition methods mentioned above. Looking more in detail criteria on the distribution of mitigation efforts and even on the level at regions or income groups of countries allows for more precise of mitigation considered desirable, hampering mitigation agreements. insights into the reasons for the differences in emissions per capita Differences in emissions per capita are driven by factors that follow between countries with a similar level of development (or other com- different trajectories in different countries. The analysis of these mon characteristics) and to identify policies that may facilitate the drivers and their trajectories over time is crucial to enlighten the reduction of these differences and of total emissions. The results of debates on policy design and on the criteria to distribute abatement this type of analyses may have more practical implications for policy efforts. Moreover, the analysis of the causes of inequality may also than focusing on the differences between countries with high differ- shed light on how emissions could be reduced by converging to low ences in development and consumption levels per capita, as there emission per capita levels. may be important heterogeneity in the relevance of the different fac- Analyses in these areas have typically dealt with CO2, the most tors determining emissions for different groups. However, there is no important greenhouse gas. Most studies on the international inequal- analysis specifically looking at inequality in emissions between devel- ity of CO2 emissions per capita have focused on the inequality oping countries and the causes of these differences. A challenge for between world nations. Several of them include a decomposition of coming years is that developing countries advance their development this inequality. Group decomposition of CO2 emissions inequality without leading to an impressive increase in emissions, as experienced tends to signal that most of the inequality is explained by the inequal- by some emerging economies. In this context, to ascertain both the ity between rich and poor groups of countries; either when the groups determinants of CO2 emissions per capita in developing countries as are directly classified according to income criteria or whether they are well as the causes of the differences in emissions per capita between geographic regions with different levels of development (Duro & these countries may help to orientate policies to control emissions Padilla, 2006; Heil & Wodon, 1997; Padilla & Serrano, 2006). Many growth, follow more sustainable development paths, and fulfill their researches also include decomposition of inequality indexes based on pledges to control emissions. identity relationships. In short, various studies apply a multiplicative In short, we want to take the advantages provided by the decomposition to split the inequality index across components in regression-based inequality decomposition method developed by order to study the determining factors of the differences in the levels Fields (2003) to answer the following research questions: to analyze of emissions per capita (such as the Kaya factors in Duro & with an econometric model the economic, structural, demographic, Padilla, 2006; Padilla & Duro, 2009; who do not include the popula- and climate determinants of CO2 emissions per capita in developing tion factor, since it disappears from the original formulation when countries (which in our analysis encompasses low and low-middle emissions per capita are considered). These studies found that the income countries as defined by the World Bank); to find out how major component of the inequality across countries is the inequality these have changed over time; to analyze the contribution of these in income (or GDP) per capita between rich and poor countries. The factors to the inequality of CO2 emissions per capita between devel- importance of other factors, such as energy intensity or emission oping countries; to study the changes in the relative contribution of intensity of energy, is of a lower magnitude, though still significant these factors over time and their causes; to find the contribution of both between and particularly, in relative terms, within some country the different factors to the changes in emissions inequality experi- groups (Duro & Padilla, 2006). Therefore, these commented works enced over the period studied; and to extract relevant conclusions for have aimed both to see the unequal contribution to the environmental environmental policy from the results obtained. We consider CO2 pressures of different countries, regions or groups of countries, as emissions from fuel combustion in our analysis. well as to ascertain the driving forces of the differences in the levels The rest of the article is organized as follows. Section 2 explains of per capita emission deriving interesting implications for environ- the regression-based inequality decomposition method employed for mental policy. These studies provide a first picture of inequality and the analysis of CO2 emissions inequality between developing coun- its driving forces but have the limitation of restricting the tries and describes the variables and data sources used in our analysis. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 153 Section 3 presents and discusses the results. Section 4 gathers the coefficient βk and its variable Xk as the factor components of CO2 main conclusions of the research. emissions per capita. The consistent identity formed allows the traditional decomposition methods (Fields, 2003; Shorrocks, 1982). 2 | METHODS AND DATA We take the variances of both sides of the equation. Actually, the variance of logarithms is a commonly used and easily decomposable As stated above, there are different methodologies to decompose inequality index (Cowell, 2011; Das & Parikh, 1982): inequality indexes into different components. The traditional decomposi- ! tion approach (Shorrocks, 1982) consists in decomposing inequality in X K þ1 varð lnCO2 pcÞ ¼ var βk X k : ð3Þ different additive components. This was later extended to the decompo- k¼0 sition into multiplicative factors using the Theil index. The above cited works have adapted and applied these methods to the analysis of envi- We reorganize it to express the variance of logarithms as the sum of ronmental pressures inequality, particularly providing useful information the covariances between each factor component and the dependent for the analysis of international inequality in emissions (Duro & variable: Padilla, 2006, 2011; Padilla & Duro, 2009, 2013; Padilla & Serrano, 2006). These decomposition methods produce useful insights X K þ1 varð lnCO2 pcÞ ¼ covðβk Xk , lnCO2 pcÞ: ð4Þ on the inequality in CO2 emissions per capita between countries and its k¼0 components. Nevertheless, the techniques described above just assign the contribution to inequality to the components of an identity, which The covariances in (4) are the natural decomposition of the variance. provides a restricted view of the driving factors of emissions. Recent We can then obtain the relative contribution of each factor research widened the field of emission inequality analysis by employing component: the regression-based inequality decomposition method developed by Fields (2003) to study the causes of these inequalities (Duro et al., 2017). cov½βk Xk , lnCO2 pc sk ð lnCO2 pcÞ ¼ , ð5Þ varð lnCO2 pcÞ In contrast to previous decomposition methods, the regression-based inequality decomposition method does not restrict the components of inequality to the elements of an identity but allows to test the contribu- where, sk is the share of the contribution of factor k to the inequality tion to inequality of any set of relevant factors (Fields, 2003). in CO2 emissions per capita. We propose the application of the regression-based inequality The regression can be done for different periods, so that the rela- decomposition methodology to analyze the inequality in CO2 emis- tive contribution of each factor to inequality may change over time. sions per capita between developing countries. This methodology Then, the method also allows to check whether the changes in sk over requires first to estimate a regression with the determinants of CO2 time are caused by changes in the dispersion of factor k (the inequality emissions per capita, from which we can extract interesting conclu- of the factor across countries) or by changes in its coefficient β sions on the factors that lead to greater or lower emissions in differ- (i.e., its importance of the factor in determining CO2 emissions accord- ent developing countries and so compare our results with those of the ing to the regressions): wider literature on the determinants of emissions (such as the envi-     ronmental Kuznets curve or the STIRPAT literature), and particularly cov Zkt , lnCO2 pct cov Zkt1 , lnCO2 pct1 with those works focusing on developing countries. skt  skt1 ¼  varð lnCO2 pct Þ varð lnCO2 pct1 Þ 2  k   3 Following the method described in Fields (2003), we estimate a b k 6cov Z t1 , lnCO2 pc t cov Z t1 , lnCO2 pc t1 7 semi-log linear function of the determinants of our variable of inter- ¼4  5 varð lnCO2 pct Þ varð lnCO2 pct1 Þ ð6Þ est, CO2 emissions per capita: 2    k 3 cov Z k , lnCO pc cov b , lnCO2 pct Z 6 t 2 t t1 7 lnCO2 pc ¼ α þ β1 X 1 þ  þ βK XK þ ε, ð1Þ þ4  5, varð ln CO2 pct Þ varð lnCO2 pct Þ where, CO2pc is the vector of the CO2 emissions per capita in the dif- ferent countries considered, and Xi (i = 1, … , K) the vectors of its bk ¼ β where, Zkt ¼ βkt X kt and Z t1 kt1 X kt . We can denote as dispersion determinants. We can rearrange it: effect the first term of the right-hand side, since the coefficients do not change between t  1 and t (only the dispersion varies). That is, X K þ1 the effect associated to the changes in the factor inequality. We can lnCO2 pc ¼ βk Xk , ð2Þ k¼0 denote as coefficient effect the second term, since the dispersion of vector Xk does not change (only the coefficient varies). That is, the where, X0 is 1, so that β0X0 = α is the constant, and βK+1 is 1, so that effect associated to the changes in the coefficients according to the βK+1 XK+1 = ε. The method considers the product of the estimated regressions estimated. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 154 PADILLA ROSA AND JADOTTE Finally, we can also measure the contribution of factor k to the quite sharper since 2010. We are particularly interested in under- total variation in CO2 emissions per capita inequality between period standing the reasons of this decrease in the inequality in CO2 emis- t  1 and period t as: sions experienced since 2005. The decrease in inequality concentrates in the last years of the period, which clearly contrasts skt Ið:Þt  skt1 Ið:Þt1 with other findings in the literature for the international inequality in δk  , ð7Þ Ið:Þt  Ið:Þt1 CO2 emissions per capita that considered both developing and devel- oped countries, since they found an important decrease much before where, I(.)t is the inequality in CO2 emissions per capita of period t. 2005 (Duro & Padilla, 2006; Heil & Wodon, 1997; Padilla & This expression can be applied to any inequality index. In coherence Serrano, 2006). with the Fields (2003) regression-based inequality decomposition During this same period the emissions per capita of developing methodology employed, we use the logarithmic variance in our facto- countries experienced an important increase. In order to see how this rial decomposition, which is a commonly used inequality index that is increase was distributed among the different countries, we show how decomposable (Das & Parikh, 1982), and is scale invariant the different quartiles changed in Table 2, as the distributive conse- (Cowell, 2011), and so seems an appropriate choice for our analysis. quences of this increase are different whether the greater increases The data required for our analysis have been taken from the were concentrated in lower emitters or in those developing countries World Bank database (https://data.worldbank.org/). Given data avail- emitting above the median. ability, we had to limit our analysis to the period 1995–2014 for Table 2 shows that the reduction in inequality has been particularly which we had a consistent set of developing countries, with minimum driven by the greater increase in emissions per capita in the first quartile variations in the yearly sample, with data for all the required variables of the distribution with respect to the other quartiles. In addition, while (see the list of countries included in the Table A4 of the Appendix A). the minimum value experiences a great increase, there is a reduction in Taking a previous year as the initial year of our analysis involved using the maximum value of the sample. When talking about emissions per a much reduced sample which, as we checked, would have led to a capita, an ideal situation would apparently be one in which all countries pattern in the evolution of inequality for those previous years just converge to lower values, which is not the case, as all quartiles (but the caused by the different countries included in the sample. maximum value) increase in the period. However, we have to take into The dependent variable employed in our analysis, ln CO2pci is account that some low income countries depart from very low levels of measured as the logarithm of the kilograms of CO2 emissions from CO2 emissions per capita, so we cannot see that increase in the lower fuel combustion per capita. As for the independent variables that values as a necessarily negative outcome—given the strong link existing resulted significant and so were included in the final estimated model, between economic development, energy consumption and CO2 these are: GDPpci, the GDP per capita (in constant 2010 US$); Agricul- emissions—but as the result of a parallel economic development process ture_shi, the value added of agriculture, forestry and fishery as a per- experienced during the period. In this sense, our main interest in this centage of country GDP; Urbanizationi, the urban population as a research is to investigate the contribution to CO2 emissions inequality percentage of total population; Pop15–64i, the population of ages 15– of different factors and how has it changed over time. 64 as a percentage of total population; and tmini, the climate normal, The decomposition of emissions inequality with the regression- that is, a 30 years average, for minimum daily temperatures in the based inequality decomposition approach will provide information on country, in  C. Other economic, demographic and climate variables the reasons behind this inequality, including factors beyond those that were tested, but were not found significant, such as the share of could be included in simpler accounting decomposition approaches. industry on total GDP, trade openness, population density, household Moreover, it will provide information on the importance of these fac- members or the average maximum daily temperatures in the country. tors for the particular case of developing countries, which will allow These variables were discarded in the final estimated models to allow for comparison with results for other samples of countries. the subsequent correct decomposition of inequality with the above indicated method into the contribution of all significant variables. Table 1 shows the descriptive statistics of the variables employed in 3.1 | The determinants of CO2 emissions per the estimation. capita in developing countries As stated in the methods section, our method requires, first, to esti- 3 | RESULTS mate the determinants of CO2 emissions per capita. In this respect, our analysis will contribute to the analysis of the determinants of Figure 1 shows the trajectory of the inequality in CO2 emissions per these emissions in developing countries. The variables employed are capita between developing countries for the period analyzed as com- those that the literature has identified as relevant drivers for CO2 puted by the variance of the logarithm in CO2 emission per capita. emissions in the environmental Kuznets curve (Dinda, 2004; There is a decrease of this inequality between 1995 and 2000 Grossman & Krueger, 1991; Shahbaz & Sinha, 2019) or the STIRPAT and a slight increase between 2000 and 2005; then, the inequality literature (York et al., 2003). The comparison with studies which also decreases during the rest of the analyzed period, being this reduction focus on developing countries is of particular interest for our analysis. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 155 TABLE 1 Descriptive statistics of the Variable Mean Median Stand. Dev. Max. Min. Obs. variables 1995 ln CO2pci 5.76 5.93 1.26 9.07 2.75 65 GDPpci 1037.92 870.30 677.95 2938.89 183.55 65 Pop15–64i 53.38 52.46 3.85 66.60 46.62 65 Urbanizationi 32.03 31.84 14.25 66.95 7.21 65 Agriculture_shi 28.07 27.30 12.06 56.54 6.65 65 tmini 16.20 18.90 7.05 25.30 7.70 65 2000 ln CO2pci 5.78 5.82 1.23 8.78 2.85 67 GDPpci 1091.10 845.98 726.13 3117.82 197.43 67 Pop15–64i 54.75 53.84 4.58 69.08 47.43 67 Urbanizationi 33.93 32.98 14.24 67.15 8.25 67 Agriculture_shi 26.14 24.99 12.88 76.07 5.30 67 tmini 16.28 18.90 6.96 25.30 7.70 67 2005 ln CO2pci 5.78 5.82 1.23 8.78 2.85 67 GDPpci 1091.10 845.98 726.13 3117.82 197.43 67 Pop15–64i 54.75 53.84 4.58 69.08 47.43 67 Urbanizationi 33.93 32.98 14.24 67.15 8.25 67 Agriculture_shi 26.14 24.99 12.88 76.07 5.30 67 tmini 16.28 18.90 6.96 25.30 7.70 67 2010 ln CO2pci 6.00 6.01 1.22 8.80 3.20 68 GDPpci 1410.50 1170.28 985.02 4168.51 234.24 68 Pop 15–64i 57.43 55.87 6.10 73.34 47.46 68 Urbanizationi 38.60 36.77 15.34 68.60 10.64 68 Agriculture_shi 22.41 22.01 11.59 52.94 3.83 68 tmini 16.10 18.60 7.02 25.30 7.70 68 2014 ln CO2pci 6.17 6.29 1.13 8.87 3.80 68 GDPpci 1574.82 1297.45 1082.86 4578.29 245.33 68 Pop 15–64i 58.16 56.39 6.22 74.20 47.20 68 Urban popi 40.25 39.39 15.74 69.21 11.78 68 Agriculture_shi 21.09 19.99 10.63 51.79 4.83 68 tmini 16.10 18.60 7.02 25.30 7.70 68 Note: ln CO2pci measured as the logarithm of the kilograms of CO2 emissions from fuel combustion per capita; GDPpci measured in constant 2010 US$; Agriculture_shi, measured as a percentage of country GDP; Urbanizationi and Pop15–64i measured as percentages of total population; and tmini, measured as 30 years average, for minimum daily temperatures in the country, in  C. Source: Prepared by the authors with World Bank data. Table 3 shows the results for the ordinary least squares estima- years that are consistent with those estimated for the rest of years tions for the selected years. The model shows a high goodness of fit, considered. All significant variables included in the model show the as the explanatory variables used in our estimated model explain expected signs, or signs that are consistent with previous findings in above 83% of the inequalities in CO2 emissions per capita between the literature, and the VIF test showed no multicollinearity problems. developing countries. All variables included are significant for all years, Besides the variables finally included in the model, we tested other except for two variables for a couple of years (Agriculture_shi for years economic, demographic and climate variables. In short, as regards cli- 2000 and 2005 and Urbanizationi for years 1995 and 2000). However, mate variables, we also tested the climate normal for the maximum these variables also show signs and coefficient magnitudes for these daily temperature; as regards demographic variables, we also tested 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 156 PADILLA ROSA AND JADOTTE the variables population density and average household members; as data supports an N-shaped form for the cross-country relationship regards economic variables, besides the share of agriculture, we also between CO2 emissions per capita and GDP per capita. This same tested the impact of the share of industry and trade openness. These pattern was previously found by other authors for different sets of variables were not significant, so we discarded them from the model countries and periods (Alshubiri & Elheddad, 2019; Duro et al., 2017; estimated as a basis of our later decomposition in order to be able to Halicioglu, 2009; Martínez-Zarzoso & Bengochea-Morancho, 2003, apply the decomposition method consistently. We comment below 2004; Moomaw & Unruh, 1997; Shafik, 1994). The absolute value of the justification for testing the different variables and the results the coefficients of GDP in our estimation are higher than those obtained in each case. As the main concern of our analysis is the found in a previous study also employing a semi-log estimation for a decomposition of the inequality in CO2 emissions per capita, and for sample including both developed and developing countries (Duro reasons of space, we do not provide all the results with the variables et al., 2017), so that income is more relevant in determining the level that were not significant (or showed in the main text, but just provide of emissions for our sample. However, the results in the literature the results of the chosen model with significant variables that will are mixed and other papers just found a positive monotonic relation- allow the later decomposition). The results of the regressions includ- ship (see e.g., Shafik & Bandyopadhyay, 1992; or Roca et al., 2001), ing these variables are available in the Tables A1–A3 of the while many others found support for the environmental Kuznets Appendix A. curve hypothesis for CO2 emissions, that is, an inverted-U shaped The coefficients of the semi-log model are interpreted as semi- relationship (Cole et al., 1997; Holtz-Eakin & Selden, 1995) (see elasticities. That is, an increase of one unit in the value of any inde- Shahbaz & Sinha, 2019, for an extended review of the literature on pendent variable leads to a β% increase in the CO2 emissions per the issue). These mixed results may derive from different model capita. As regards GDP per capita, the results indicate a cubic rela- specifications, estimation methodologies, sample of countries and tionship with CO2 emissions per capita, as the coefficients of the periods analyzed. quadratic and cubic variable are significant, though small. That is, the Therefore, though economic affluence has a clear impact on emis- sions, being a positive correlation between them for the lower and the higher part of the income per capita distribution, our results show 1.7 that, for a segment of the income per capita distribution of countries, 1.6 the relationship changes its sign, which is a consistent result across 1.5 the different years estimated. In any case, as Cole (1999, p. 96) states, 1.4 “the impact of economic development on the environment is clearly complex in nature. It is important to note, however, that whilst eco- 1.3 nomic growth may facilitate some environmental improvements, this 1.2 is not an automatic process and will only result from investment and 1.1 policy initiatives.” Our result just shows the relationship between CO2 1 emissions per capita and GDP per capita for a cross-section of coun- 1995 2000 2005 2010 2014 tries in different moments of time and should not be interpreted as predicting the path that will be followed by the different individual F I G U R E 1 Inequality in CO2 emissions per capita between countries as their income changes over time (Piaggio & Padilla, 2012). developing countries, 1995–2014. Source: Prepared by the authors with World Bank data [Colour figure can be viewed at If the objective was to predict the behavior of countries, which is not wileyonlinelibrary.com] the case here, the research should be based on longitudinal studies on T A B L E 2 Quartiles of the Quartile 1995 2000 2005 2010 2014 Change 1995–2014 distribution of CO2 emissions per capita 0 (Min. value) 0.016 0.017 0.021 0.025 0.045 184.8% between developing countries, 0.25 (Q1) 0.120 0.148 0.158 0.187 0.220 83.3% 1995–2014 0.5 (Q2 or median) 0.377 0.335 0.284 0.406 0.531 40.8% 0.75 (Q3) 0.702 0.765 0.986 0.975 1.058 50.7% 1 (Max. value) 8.657 6.526 7.088 6.641 7.088 18.1% Note: the value in the first column denotes the share of countries with CO2 emissions (t) per capita below the value of the other columns for each year. The first row shows the minimum value of CO2 emissions per capita across developing countries for each year. The last row shows the maximum value of CO2 emissions per capita across developing countries for each year. The three rows in the middle show the values separating the quartiles of the distribution of CO2 emissions per capita between developing countries. Notice that CO2 emissions are considered here in levels, while Table 1 included them in logarithms as in the estimation. Source: Prepared by the authors with World Bank data. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 157 TABLE 3 Cross-section regressions on the determinants of CO2 emissions per capita in developing countries, selected years (1995–2014) 1995 2000 2005 2010 2014 Economic affluence GDPpci 0.0043087*** 0.0051087*** 0.0041723*** 0.002815*** 0.0023652*** (0.0011882) (0.0009592) (0.0007637) (0.0007467) (0.000612) GDPpc2i 2.25e06** 2.70E06*** 2.02E06*** 1.12E06*** 8.82E07*** (9.23e07) (7.01e07) (4.67e07) (3.80e07) (2.96e07) GDPpc3i 3.58e10* 4.53E10*** 3.06E10*** 1.42E10** 1.07E10** (2.09e10) (1.48e10) (8.51e11) (5.84e11) (4.25e11) Economic structure Agriculture_shi 0.0182897** 0.010126 0.0107789 0.017622** 0.0173629** (0.0082313) (0.0067682) (0.0069369) (0.008562) (0.0077447) Demography Urbanizationi 0.0109378 0.0049553 0.0167026*** 0.0138178** 0.0128213*** (0.0067653) (0.0062471) (0.0058792) (0.0053484) (0.0047432) Pop15–64i 0.1023719*** 0.0941024*** 0.0860454*** 0.067435*** 0.0552043*** (0.0233102) (0.0179294) (0.0153804) (0.0145942) (0.0128721) Climate tmini 0.0510659*** 0.0409441*** 0.0206693** 0.0185509* 0.0321508*** (0.010571) (0.0102079) (0.0101873) (0.0108914) (0.0095407) Intercept 0.7600811 1.097137 1.190788 0.4932817 1.674291** (1.392212) (1.011626) (0.9082756) (0.8935976) (0.80962) N 65 67 68 68 68 F 40.54*** 43.19*** 47.95*** 41.92*** 44.95*** R2 0.8327 0.8367 0.8484 0.8303 0.8398 Adjusted R2 0.8122 0.8173 0.8307 0.8105 0.8212 Note: Statistical significance is denoted by (*) at the 10% level (p < .10); (**) at the 5% level (p < .05); and (***) at the 1% level (p < .01). Standard errors within parentheses. Source: Prepared by the authors with World Bank data. individual countries (De Bruyn et al., 1998; Moomaw & Unruh, 1997; case of China (Shen, 2006) or Uruguay (Piaggio et al., 2017). There are Roca et al., 2001). various explanations for our outcome. Service and industrial sectors As regards economic structure, we first tested two variables of are to a certain extent complementary sectors that use to grow production composition, the share of agriculture value-added on total together in low income countries. The idea that the service sector is a GDP and the share of industry value-added on total GDP, while we clean sector ignores that, besides the high direct emissions of the left services as the base sector. The expected sign in the case of the transport activity, other service activities, on top of their direct energy latter variable was positive, as a greater share of industry is usually consumption, require other sectors, from which they purchase inter- expected to be associated, other things equal, to more energy inten- mediate goods, to consume energy and materials. Several studies con- sive sectors and so correlated with higher emissions, while in the case tradict the misperception of the service sector as a non-material of agriculture we expected the contrary to hold. However, in the case sector (Alcántara & Padilla, 2009; Fourcroy et al., 2012; Gadrey, 2010; of the industry share, the coefficient was not significant for any of the Nansai et al., 2009; Piaggio et al., 2015; Rosenblum et al., 2000; years of the sample. That is, for developing countries and the years Suh, 2006). According to Gadrey (2010) a larger share of the service analyzed there is not a significant impact in emissions where the share sector in the economy leads to more energy consumption. Friedl and of industry on GDP increases (across countries) reducing the share of Getzner (2003) found the share of services to be positively correlated services. Moreover, for the last years of the sample (2010 and 2014) with greater emissions for the case of Austria. Our results would be the non-significant coefficient was even negative. This may seem a consistent with those findings, as they indicate that a 1% increase in surprising result, and actually contrasts with the results obtained by the share of agriculture, meaning a decrease in the joint share of Duro et al. (2017) with a similar methodology and model for a sample industry and services, would lead to a significant decrease of emis- of countries that also included developed countries. The share of sions (between 1% and 1.8%, according to the year). Finally, we also industrial sectors was also found as significant in other studies for the tested trade openness (measured as the ratio of the sum of exports 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 158 PADILLA ROSA AND JADOTTE and imports on GDP). It only appeared significant for the year 1995, sample. This result is consistent with the previously found by Jones with a very low positive value, being far from significant for the other (1991) and Parikh and Shukla (1995) for cross-section analyses for years, so we discarded the variable from the model. When using the developing countries, and Cole and Neumayer (2004) and Duro et al. variable as just the percentage of exports on GDP the coefficient was (2017) for samples with developing and developed countries (though not significant for any year. This result would be consistent with for in our analysis on developing countries we find greater coefficients example, the results of Cole et al. (1997), which seems to be also the since 2005 that this last work). However, the results in the literature most common result for other pollutants than CO2 (Grossman & are mixed and, for example, Hossain (2011) found that this variable Krueger, 1991). It is also consistent with the result of Lamb et al. had different impacts on different countries for a set of newly indus- (2014) who found that trade openness (measured just as the percent- trialized countries. Sharma (2011) and Balsalobre-Lorente et al. (2022) age of exports on GDP) was not a significant determinant of territorial found a negative effect of urbanization on CO2 emissions per capita emissions for a set of 87 countries for the year 2008, though they for a panel of 69 countries and for BRICS countries, respectively. found it significant with a positive coefficient when consumption- Urban habits and infrastructures may be associated with more energy based emissions were employed instead. The variable was neither sig- consumption, more use of fossil fuels instead of biomass (emissions nificant in the analysis of Sharma (2011) for a panel of 69 countries. In from biomass are not considered), longer displacements for commut- contrast, Shahbaz et al. (2017) found that trade openness had a posi- ing and buying and more use of motor vehicles (Cole & tive significant impact on CO2 emissions per capita for a panel of Neumayer, 2004; Parikh & Shukla, 1995). However, there are also 107, with results that hold also for subpanels of low, middle and high other effects such as more technical innovation, energy and land use income, while for individual countries the coefficient was negative or efficiency and access to information which may in the long run help to positive depending on the country. Chang et al. (2014) also found that decrease emissions (Jiang & Hardee, 2011), which would explain the the ratio of net exports had a positive impact on carbon footprint per mixed results of the impact of urbanization on CO2 emissions. Lamb capita for a panel of 98 countries. In contrast, Atici (2009) found trade et al. (2014) found that urbanization was a significant determinant of openness to be negatively correlated with emissions in Central and emissions, though only with the usual territorial-based emissions but Eastern Europe. Some individual country studies have found trade not with consumption-based carbon emissions. In our estimation for openness to have a negative relation with CO2 emissions, such as developing countries there is a clear positive relationship between Piaggio et al. (2017) for Uruguay and Friedl and Getzner (2003) for urbanization and CO2 emissions, which, as stated above for other Austria, while others have found the opposite, such as Halicioglu determinants, does not mean that this relationship could predict the (2009) for Turkey, which would depend on the specific trade composi- relationship over time of the individual countries of the sample. tion and openness processes followed in these different countries. As We checked other demographic variables but found that they indicated above, the fact that our cross-national analysis does not find were not significant. In short, we tested for the impact of average this coefficient as significant, does not mean that the variable may not household size on CO2 emissions per capita but found no significant have an impact on the evolution of the emissions of some individual influence. The variable has been suggested to impact on resource con- countries over time, as the literature suggests that may be the case. sumption and CO2 emissions (Liu et al., 2003; MacKellar et al., 1995). With respect to the coefficients of demographic variables, the Cole and Neumayer (2004) found that this variable was significant percentage of population aged 16–64, Pop16–64i, shows a significant and negatively correlated with CO2 emissions (their dependent vari- positive value for all years, meaning that an additional 1% of working able was not in per capita terms). According to our results, however, it age population (with respect to total population) is associated with does not seem to have a significant impact on the CO2 emissions per around 10% more emissions at the beginning of the period, but just capita of developing countries. We also tested for population density, 5.5% at the end. This result is in line with the results of Duro et al. which has often been included in the analysis of the determinants of (2017) for a sample of both developing and developed countries. Cole several environmental pressures and the impact of which is ambigu- and Neumayer (2004) also found that population aged 15–64 was a ous in the literature. Onafowora and Owoye (2014) found a significant significant determinant of CO2 emissions, though only when they did and positive influence on CO2 emissions in Brazil, China, Egypt, Japan, not include in the model urbanization and household members, while and Mexico, while in the other countries they considered the variable was not significant otherwise; though in their case they (South Korea, Nigeria, and South Africa) the impact on CO2 emissions used total instead of per capita CO2 emissions, which may explain was statistically insignificant. Lamb et al. (2014) found that population some differences in their results. This population segment is the density was not a significant determinant of carbon emissions (neither potentially active population, which according to our results seems for consumption- or territorial-based carbon emission). Our results are associated with more polluting habits and consumption patterns. The consistent with this last paper, as the coefficient of population density coefficient, however, seems to decline over time, though maintaining was far from being significant for any of the years. its significance. Concerning climate variables, we included the average minimum As regards the other demographic variable included in the final daily temperature in the model, as colder temperatures are associated model, a (1%) increase in urban population with respect to total popu- with greater heating requirements (Neumayer, 2002), and so greater lation, Urbanizationi, is associated with (above 1%) greater CO2 emis- CO2 emissions from fuel combustion. We also checked the impact of sions per capita and the coefficient is significant for most years of the the average maximum daily temperature. This variable is, obviously, 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 159 highly correlated with the average minimum daily temperature tested inequality was the main driver of CO2 emissions per capita inequality. in our model; however, the later was more significant and for more The first paper applied a multiplicative decomposition of inequality years than the former, so we just left the minimum daily temperature into the factors of the Kaya identity for a world sample, finding that in the model, in line with the related literature. The variable was sig- affluence was responsible for up to 60% of emissions inequality, while nificant and negatively associated with CO2 emissions per capita for the second applied it to the European Union countries (a more all years. In short, a country with an average minimum daily tempera- income-homogeneous sample) and found that, though at the begin-  ture 1 C below others would be statistically associated with 5% more ning of the period (1990) energy intensity was more important (due to emissions per capita in 1995 but just 3.2% in 2014. York et al. (2003), the energy inefficiency of the eastern countries), at the end, affluence Neumayer (2002, 2004), Lamb et al. (2014) and Duro et al. (2017) also clearly was the main determinant of emissions. Contrary to our found this influence of colder temperatures for samples including both method, these two studies restrict the factors to those of the Kaya developing and developed countries, though we find a bit larger identity and impose the same elasticity to them, while our methodol- impact of this variable for developing countries. ogy allows for the inclusion of any influencing factor, allowing a better determination of the impact of affluence on total inequality in CO2 emissions per capita. As regards the explanation of the changes in the 3.2 | Decomposition of the inequality in CO2 relative contribution over time, taking the whole period we can see emissions per capita between developing countries that the small increase in the relative contribution of this variable is caused by the dispersion effect (i.e., income inequality), which goes in Following expression (5), we use the coefficients estimated, together the opposite direction of the coefficient effect (Table 5). Therefore, with the dispersion of each variable to determine the contribution of the persistence of significant differences in per capita income each factor to the inequality in CO2 emission per capita between between developing countries—particularly its trajectory between developing countries that are shown in Table 4. In addition, expres- 1995 and 200—would explain why per capita income continues to be sion (6) allows for the decomposition of the change experienced the main explanatory factor of inequality in CO2 emissions per capita between each period in the relative contribution of each factor into between developing countries at the end of the period. the coefficient effect (due to the change in the relationship between With reference to the sectoral composition of production, the the variable and CO2 emissions as shown by the coefficients in the share of agriculture on GDP accounts for above 10% of inequality in regressions) and the dispersion effect (due to the inequality in the var- emissions, though it is slower for 2000 and 2005. The small change in iable between countries), which is shown in Table 5. its relative importance between the beginning and the end of period The component that contributed most to the inequality in CO2 would be mainly explained by the dispersion effect (Table 5). How- emissions per capita was economic affluence (which aggregates the ever, looking at the sub-periods, we see that there has not been a contribution of the different GDP variables), with a contribution over constant tendency in the contribution of the different effects to the 30% for all the years analyzed. The factor increased its contribution changes experienced in the relative contribution of this factor significantly between 1990 and 2000, when it reached its maximum between the sample years. Actually, while the coefficient (β) esti- relative contribution (41.1%), but decreased in the next 10 years to mated at the end of the period is very similar to the one in 1995, it end the period with a similar contribution to that at the beginning. decreases significantly between 1995 and 2000, which is caused by Comparing the results with the ones of Duro et al. (2017), who the coefficient effect; that is, the lower relevance of the factor in included both developed and developing countries, we find that, for explaining CO2 emissions per capita. The contribution of agriculture developing countries there is a significantly greater relative contribu- to inequality is lower than the one found by Duro et al. (2017) for tion of affluence to emissions inequality (almost double for the years their more heterogeneous sample (the share of this sector is generally 1995 and 2005, which were also included in their sample). Duro and greater for developing countries, than for the world average). In any Padilla (2006) and Padilla and Duro (2013) also found that income case, the relevance of this variable, together with the fact that the TABLE 4 Factor contribution to the inequality in CO2 emissions per capita between developing countries (%) 1995 2000 2005 2010 2014 Affluencei 30.9% 41.1% 35.1% 31.7% 32.1% Agriculture_shi 10.4% 5.6% 6.3% 11.7% 11.1% Urbanizationi 8.2% 3.3% 12.3% 10.0% 9.6% Pop15–64i 20.7% 24.1% 27.1% 24.5% 22.3% tmini 11.5% 8.5% 3.3% 3.3% 7.6% Residual 18.3% 17.5% 15.9% 18.7% 17.3% Total 100.0% 100.0% 100.0% 100.0% 100.0% Source: Prepared by the authors with World Bank data. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 160 PADILLA ROSA AND JADOTTE TABLE 5 Changes in the relative contribution of each factor and decomposition into coefficient and dispersion effects Absolute contribution change Dispersion effect % Coefficient effect % 1995–2000 Affluencei 0.1011 0.0099 10% 0.0912 90% Agriculture_shi 0.0485 0.0035 7% 0.0449 93% Pop15–64i 0.0339 0.0551 162% 0.0211 62% Urbanizationi 0.0484 0.0080 17% 0.0404 83% tmini 0.0301 0.0092 30% 0.0209 70% Residual 0.0080 0.0080 100% 0.0000 0% 2000–2005 Affluencei 0.0600 0.0013 2% 0.0586 98% Agriculture_shi 0.0071 0.0032 46% 0.0038 54% Pop15–64i 0.0304 0.0558 183% 0.0254 83% Urbanizationi 0.0897 0.0031 3% 0.0866 97% tmini 0.0512 0.0184 36% 0.0328 64% Residual 0.0159 0.0159 100% 0.0000 0% 2005–2010 Affluencei 0.0332 0.0077 23% 0.0409 123% Agriculture_shi 0.0540 0.0086 16% 0.0453 84% Pop 15–64i 0.0256 0.0421 164% 0.0677 264% Urbanizationi 0.0232 0.0023 10% 0.0209 90% tmini 0.0004 0.0034 974% 0.0038 1074% Residual 0.0284 0.0284 100% 0.0000 0% 2010–2014 Affluencei 0.0038 0.0150 394% 0.0112 294% Agriculture_shi 0.0062 0.0046 74% 0.0016 26% Pop15–64i 0.0224 0.0270 121% 0.0494 221% Urbanizationi 0.0041 0.0033 81% 0.0074 181% tmini 0.0429 0.0108 25% 0.0322 75% Residual 0.0140 0.0140 100% 0.0000 0% 1993–2014 Affluencei 0.0117 0.0872 744% 0.0755 644% Agriculture_shi 0.0063 0.0122 193% 0.0059 93% Pop15–64i 0.0163 0.2069 1267% 0.1905 1167% Urbanizationi 0.0139 0.0002 1% 0.0141 101% tmini 0.0388 0.0060 15% 0.0447 115% Residual 0.0096 0.0096 100% 0.0000 0% Source: Prepared by the authors with World Bank data. share of industry on GDP was not significant, clearly reinforces the between developing countries than it is in explaining total inequality, notion that agriculture is the least polluting sector and that both ser- if we compare our results with Duro et al. (2017). The slight increase vice and industry are associated with more emissions than the agricul- in its relative contribution in the period is mainly explained by the ture sector for developing countries, which is reflected in the 10%– coefficient effect; that is, it becomes relatively more important in 11% contribution of agriculture to emissions inequality between explaining CO2 emissions from fossil fuels per capita over time. developing countries. The other demographic variable, population aged 15–64, is the Urbanization contributes around 10% to emissions inequality second most relevant variable in explaining inequality in CO2 emis- between developing countries, and the contribution at the end of the sions per capita between developing countries, with a contribution period is just slightly greater than at the beginning. Since 2000, this that is always above 20%. The relevance of this factor was also found factor is more relevant (about the double) in explaining inequality by Duro et al. (2017), but for their sample of countries it was found as 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 161 T A B L E 6 Contribution of the 1995–2000 2000–2005 2005–2010 2010–2014 1995–2014 different factors to the change in CO2 emissions inequality (%) Affluencei 193.9% 139.6% 106.7% 29.6% 26.1% Agriculture_shi 118.2% 26.8% 110.3% 15.2% 7.8% Pop15–64i 115.9% 273.4% 62.4% 12.3% 2.5% Urbanizationi 54.8% 115.7% 82.4% 37.2% 14.0% tmini 78.4% 145.8% 4.1% 21.1% 27.4% Residual 36.1% 30.5% 45.4% 26.7% 22.2% Total 100.0% 100.0% 100.0% 100.0% 100.0% Inequality change 4.3% 3.4% 4.4% 15.0% 19.6% Source: Prepared by the authors with World Bank data. the main factor, while for developing countries, though very impor- contribution to the inequality of CO2 emissions per capita between tant, is still clearly below the contribution of affluence. The relative developing countries, which explains that at the end of the period importance of this factor fluctuates over the period, increasing signifi- affluence explains a slightly larger share of the total inequality of CO2 catively until 2005 and decreasing thereafter, though its relative con- emissions per capita than at the beginning. In this case, the trajectory tribution is always below the contribution of affluence. The small of the inequality in affluence between developing countries has increase in its relative importance over the whole period is explained impeded an even greater reduction of the inequality in CO2 emissions by the dispersion effect; that is, the increase would be explained by per capita. the trajectory of the differences in population aged 16–64 between It is particularly noticeable the contribution to the reduction of developing countries. Meanwhile, the coefficient effect may have CO2 emissions per capita inequality of average daily temperatures, played in the opposite direction, as the coefficient estimated for this with a contribution to the reduction well above its contribution to variable decreased significantly during the period; that is, it became CO2 inequality over the period. In this case, its contribution to the relatively less important in determining emissions (Table 3). In fact, it reduction was mainly due to the lower coefficients of this variable at is noticeable the strong movement in opposing directions of both the end of the period. That is, it became less relevant in explaining effects. CO2 emissions from fuel combustion. The average daily temperature contribution to inequality As for population aged 16–64, the contribution of this factor to decreases during the period, though at the end of the period it still the decrease in the inequality in CO2 emissions per capita between explains 7.6% of the inequality. The decrease in its relative importance developing countries is relatively small (much below its share in in explaining inequality is mainly caused by the coefficient effect; that explaining this inequality). This is due to the persistence of the differ- is, the fact that temperatures are a bit less relevant (lower absolute ences in the variable across developing countries. As regards urbaniza- coefficient) in explaining CO2 emissions per capita from fossil fuels at tion, agricultural share and the residual term, their contribution to the the end of the period. reduction of inequality is not much different from their share in The residual contribution is quite stable over the period (between explaining this inequality. 15.9% and 18.7%). It may be gathering the impact of several omitted determinants, such as different technological levels, institutional dif- ferences, social factors, and others for which we do not have available 4 | CONCLU SION AND P OLICY and suitable data for all countries and years. In any case, there are not IMPLICATIONS relevant changes in the contribution of this factor during the period, so the potentially omitted variables may not change our conclusions We have explored the differences in CO2 emissions per capita as regards the evolution in the influence of the significant factors between developing countries and how these are influenced by a considered. series of affluence, structural, demographic, and climate variables. For Finally, Table 6 indicates the percentage contribution of the dif- this, we have first performed a regression analysis, contributing to the ferent variables to the change in inequality over the period, computed literature on the determinants of CO2 emissions with new evidence with expression (7). All the factors contributed to the reduction in for the case of developing countries. The results show an N-shaped inequality between developing countries experienced over the period, cubic relationship with GDP per capita, and a significant, though of a either because of the evolution of their own inequality as for the moderate magnitude, negative impact on emissions of the agriculture changes in their influence on emissions, as shown by previous tables. share of GDP and average daily minimum temperatures, while urbani- It can be highlighted the contribution of affluence to the reduc- zation and particularly the share of potentially active population tion in the inequality in CO2 emissions per capita from fuel combus- would significantly increase emissions per capita across countries. An tion. This factor was responsible for 26.1% of this decrease. In any interesting result is that, contrary to similar cross-country studies case, its contribution to the reduction is slightly below its global including developed countries, the industry share of GDP is not a 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 162 PADILLA ROSA AND JADOTTE significant determinant of emissions for developing countries, so that particularly associated with more pollution-intensive habits in devel- pursuing a direct transformation to service economies instead of oping countries and in explaining CO2 emissions per capita differences industrializing their economies would apparently not be associated between them. Developing countries are experiencing a high rate of with lower emissions across developing countries. Other variables urbanization that will continue in coming years. Those countries that such as trade openness, population density or family unit members promote a smart urbanization, avoiding sprawl through smart urban were discarded from the model used as the basis for our inequality design and, particularly, avoiding locking in carbon-intensive infra- decomposition, as they were found not significant determinants of structures will be able to attenuate the impact of urbanization on CO2 emissions per capita for developing countries. emissions, while they could gain much by encouraging the potential By using the regression-based inequality decomposition method positive aspects, such as a more efficient energy and land use and eas- (Fields, 2003), our analysis indicates the weight of each significant deter- ier access to public equipment, infrastructures, markets and informa- minant of emissions in explaining the inequality in CO2 emissions per tion. The more countries succeed on this, the less relevant this factor capita observed between developing countries. It highlighted the princi- would become in the future in explaining the inequality in CO2 emis- pal contribution of economic affluence to these differences (above sions. Finally, the search of leap-frogging into a service economy, 30%), which contrasts with the previous only other application of this without passing through industrialization processes, does not seem to method to international inequality in CO2 emissions per capita for a be a policy that guarantees lower emissions for developing countries sample including developing and developed countries (Duro per se, as services are actually associated and complementary to many et al., 2017), where, despite being quite important, this factor was not polluting activities. The processes of specialization may affect emis- the most important contributor to inequality. Actually, our results show sions differences according to the particular subsectors and regula- clear differences in the relevance of the different factors in explaining tions that are put into place in order to facilitate development. The the inequality in CO2 emissions per capita for developing countries than impact on emission inequality will also depend on the degree of inter- were found in that analysis. The potentially active population factor national specialization, that is, whether countries tend toward greater shows it to be the second main contributor to this inequality (above specialization in different activities or whether there is a tendency to 20%), though far from the weight found in the cited work. Other factors converge to similar production compositions. with different influence are agriculture share, which accounts for a lower Our analysis has provided useful insights for environmental poli- part of the inequality between developing countries, and urbanization, cies for developing countries, which are highly interested in the suc- which appears to be a more relevant factor in explaining emissions cess of global climate policies that partly depend on their inequality between developing countries than for a mixed sample. This commitments. Knowledge of both the determinants of CO2 emissions last result is a consequence of the greater impact on emissions shown per capita for developing countries as well as on the specific factors by the larger coefficient—that is, a greater semi-elasticity—found for this that explain the differences between them provide useful information variable in our study on developing countries. on the variables that lead some developing countries to pollute more Part of the inequality in CO2 emissions per capita is explained by or less than others. Moreover, the information obtained is more pre- climate variables; however, most of the differences are explained by cise and useful for developing countries than what could be derived socioeconomic and demographic variables which may be influenced from an analysis mixing countries of different development levels, as by political changes. Though affluence is the most relevant factor we have identified significant differences in the relevance of these explaining differing emissions between countries, the relationship factors for developing countries. Finally, our analysis has also provided between emissions and income or production is a complex one, as information on the variables whose changes may lead to convergence shown by the cubic relationship found in this study. Hence, it toward lower emissions levels across developing countries and their deserves further research to ascertain how, isolating it from the other relevance in achieving this desired outcome. detected significant factors, some developing countries are able to emit less with the same or even higher level of income per capita than ACKNOWLEDG MENTS others. The role of technology, which in our study is not properly cap- We are grateful for the constructive comments of the three anonymous tured and may be partly hidden in the residual term (and possibly reviewers and the editor. This research received financial support from partly in the nonlinear components of affluence), will be of great rele- the Government of the Republic of Korea (South Korea) under the vance in the future to allow the development of countries without Global Facility on Growth for Development Trust Fund. We are also passing through the most pollution-intensive stages that developed grateful for the useful comments from David Cal MacWilliam. All errors countries and emerging economies have been through (Padilla, 2017). and omissions are the authors' responsibility. The views expressed in As regards the potentially active population, it is clear that the particu- this paper are those of the authors and do not necessarily represent the lar consumption and transport habits related to this demographic views of the World Bank or the Autonomous University of Barcelona. group are determining emissions and differences between countries, so that policies should focus on influencing these patterns. Moreover, FUNDING INF ORMATI ON projections on the age structure of the population may be an informa- This research received financial support from the Government of the tive tool to predict difficulties in controlling CO2 emissions and even Republic of Korea (South Korea) under the Global Facility on Growth its influence on emissions inequality trends. Urbanization seems to be for Development Trust Fund. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 163 CONF LICT OF IN TE RE ST Friedl, B., & Getzner, M. (2003). Determinants of CO2 emissions in a small The authors declare that they have no known competing financial open economy. Ecological Economics, 45(1), 133–148. Gadrey, J. (2010). The environmental crisis and the economics of services: interests or personal relationships that could have appeared to influ- The need for revolution. In F. Gallouj & F. Djellal (Eds.), The handbook ence the work reported in this paper. of innovation and services (pp. 93–125). Edward Elgar. Grossman, G. M., & Krueger, A. B. (1991). Environmental impacts of a north ORCID American free trade agreement. NBER working paper 3914. National Bureau of economic research (NBER), Cambridge. Emilio Padilla Rosa https://orcid.org/0000-0002-9510-8262 Halicioglu, F. (2009). An econometric study of CO2 emissions, energy con- sumption, income and foreign trade in Turkey. Energy Policy, 37(3), RE FE R ENC E S 1156–1164. Alcántara, V., & Padilla, E. (2009). Input–output subsystems and pollution: Heil, M. T., & Wodon, Q. T. (1997). Inequality in CO2 emissions between An application to the service sector and CO2 emissions in Spain. Eco- poor and rich countries. The Journal of Environment and Development, logical Economics, 68(3), 905–914. 6(4), 426–452. Alshubiri, F., & Elheddad, M. (2019). Foreign finance, economic growth Holtz-Eakin, D., & Selden, T. M. (1995). Stoking the fires? CO2 emissions and CO2 emissions nexus in OECD countries. International Journal of and economic growth. Journal of Public Economics, 57(1), 85–101. Climate Change Strategies and Management, 12(2), 161–181. Hossain, M. S. (2011). Panel estimation for CO2 emissions, energy con- Atici, C. (2009). Carbon emissions in central and Eastern Europe: Environ- sumption, economic growth, trade openness and urbanization of mental Kuznets curve and implications for sustainable development. newly industrialized countries. Energy Policy, 39, 6991–6999. Sustainable Development, 17(3), 155–160. Intergovernmental Panel on Climate Change (IPCC). (2014). Climate change Balsalobre-Lorente, D., Driha, O. M., Halkos, G., & Mishra, S. (2022). Influ- 2014: Impacts, adaptation, and vulnerability. Working group II contribu- ence of growth and urbanization on CO2 emissions: The moderating tion to the fifth assessment report of the intergovernmental panel on cli- effect of foreign direct investment on energy use in BRICS. Sustainable mate change. Cambridge University Press. Development, 30(1), 227–240. Jiang, L., & Hardee, K. (2011). How do recent population trends matter to Chang, D. S., Yeh, L. T., & Chen, Y. F. (2014). The effects of economic climate change. Population Research and Policy Review, 30(2), 287–312. development, international trade, industrial structure and energy Jones, D. W. (1991). How urbanization affects energy-use in developing demands on sustainable development. Sustainable Development, 22(6), countries. Energy Policy, 19(7), 621–630. 377–390. Lamb, W. F., Steinberger, J. K., Larkin, A. B., Peters, G. P., Roberts, J. T., & Cole, M. A. (1999). Limits to growth, sustainable development and envi- Wood, F. R. (2014). Transitions in pathways of human development ronmental Kuznets curves: An examination of the environmental and carbon emissions. Environmental Research Letters, 9(1), 1–10. impact of economic development. Sustainable Development, 7(2), Liu, J., Daily, G. C., Ehrlich, P., & Luck, G. W. (2003). Effects of household 87–97. dynamics on resource consumption and biodiversity. Nature, Cole, M. A., & Neumayer, E. (2004). Examining the impact of demographic 421(6922), 530–533. factors on air pollution. Population and Environment, 26(1), 5–21. MacKellar, F. L., Lutz, W., Prinz, C., & Goujon, A. (1995). Population, Cole, M. A., Rayner, A. J., & Bates, J. M. (1997). The environmental Kuz- households, and CO2 emissions. Population and Development Review, nets curve: An empirical analysis. Environment and Development Eco- 21(4), 849–886. nomics, 2(4), 401–416. Martínez-Zarzoso, I., & Bengochea-Morancho, A. (2003). Testing for an Cowell, F. (2011). Measuring inequality. Oxford University Press. environmental Kuznets curve in Latin American countries. Revista de Das, T., & Parikh, A. (1982). Decomposition of inequality measures and a Análisis Economico, 18, 3–26. comparative analysis. Empirical Economics, 7, 23–48. Martínez-Zarzoso, I., & Bengochea-Morancho, A. (2004). Pooled mean De Bruyn, S. M., Van den Bergh, J. C. J. M., & Opschoor, J. B. (1998). group estimation of an environmental Kuznets curve for CO2. Econom- Economic growth and emissions: Reconsidering the empirical basis ics Letters, 82(1), 121–126. of environmental Kuznets curves. Ecological Economics, 25(2), Moomaw, W. R., & Unruh, G. C. (1997). Are environmental Kuznets curves 161–175. misleading us? The case of CO2 emissions. Environment and Develop- Diffenbaugh, N. S., & Burke, M. (2019). Global warming has increased ment Economics, 2(4), 451–463. global economic inequality. Proceedings of the National Academy of Sci- Nansai, K., Kagawa, S., Suh, S., Fujii, M., Inaba, R., & Hashimoto, S. (2009). ences, 116(20), 9808–9813. Material and energy dependence of services and its implications for Dinda, S. (2004). Environmental Kuznets curve hypothesis: A survey. Eco- climate change. Environmental Science and Technology, 43(12), 4241– logical Econonomics, 49(4), 431–455. 4246. Duro, J. A., & Padilla, E. (2006). International inequalities in per capita CO2 Neumayer, E. (2002). Can natural factors explain any cross-country differ- emissions: A decomposition methodology by Kaya factors. Energy Eco- ences in carbon dioxide emissions? Energy Policy, 30(1), 7–12. nomics, 28, 170–187. Neumayer, E. (2004). National carbon dioxide emissions: Geography mat- Duro, J. A., & Padilla, E. (2011). Inequality across countries in energy inten- ters. Area, 36(1), 33–40. sities: An analysis of the role of energy transformation and final energy Onafowora, O. A., & Owoye, O. (2014). Bounds testing approach to analy- consumption. Energy Economics, 33(3), 474–479. sis of the environment Kuznets curve hypothesis. Energy Economics,  -Figueras, J., & Padilla, E. (2017). The causal factors of Duro, J. A., Teixido 44(C), 47–62. international inequality in CO2 emissions per capita: A regression- Padilla, E. (2017). What can developing countries gain from a green trans- based inequality decomposition analysis. Environmental and Resource formation? In T. Altenburg & C. Assmann (Eds.), Green industrial policy: Economics, 67, 683–700. Concept, policies, country experiences (pp. 22–37). UN Environment and Fields, G. S. (2003). Accounting for income inequality and its change: A German Development Institute/Deutsches Institut für Entwicklung- new method, with application to the distribution of earnings in the spolitik (DIE). United States. Research in Labor Economics, 22, 1–38. Padilla, E., & Duro, J. A. (2009). Analysis of the international inequality in Fourcroy, C., Gallouj, F., & Decellas, F. (2012). Energy consumption in ser- CO2 emissions: The contribution of income inequality and other fac- vice industries: Challenging the myth of non-materiality. Ecological tors. In I. H. Wadell (Ed.), Income distribution: Inequalities, impacts and Economics, 81, 155–164. incentives (pp. 171–183). Nova Science Publishers. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 164 PADILLA ROSA AND JADOTTE Padilla, E., & Duro, J. A. (2013). Explanatory factors of CO2 per capita Shahbaz, M., Nasreen, S., Ahmedd, K., & Hammoudeh, S. (2017). Trade emission inequality in the European Union. Energy Policy, 62, 1320– openness–carbon emissions nexus: The importance of turning points 1328. of trade openness for country panels. Energy Economics, 61, 221–232. Padilla, E., & Serrano, A. (2006). Inequality in CO2 emissions across coun- Shahbaz, M., & Sinha, A. (2019). Environmental Kuznets curve for CO2 tries and its relationship with income inequality: A distributive emissions: A literature survey. Journal of Economic Studies, 46(1), approach. Energy Policy, 34(14), 1762–1772. 106–168. Parikh, J., & Shukla, V. (1995). Urbanization, energy use and greenhouse Sharma, S. S. (2011). Determinants of carbon dioxide emissions: Empirical effects in economic development: Results from a cross-national study evidence from 69 countries. Applied Energy, 88(1), 376–382. of developing countries. Global Environmental Change, 5(2), 87–103. Shen, J. (2006). A simultaneous estimation of environmental Kuznets Piaggio, M., Alcántara, V., & Padilla, E. (2015). The materiality of the imma- curve: Evidence from China. China Economic Review, 17(4), 383–394. terial. Services sectors and CO2 emissions in Uruguay. Ecological Eco- Shorrocks, A. F. (1982). Inequality decomposition by factor components. nomics, 110, 1–10. Econometrica, 50(1), 193–211. Piaggio, M., & Padilla, E. (2012). CO2 emissions and economic activity: Stern, N. (2006). Stern review on the economics of climate change. Cam- Heterogeneity across countries and non-stationary series. Energy Pol- bridge University Press. icy, 46, 370–381. Suh, S. (2006). Are services better for climate change. Environmental Sci- Piaggio, M., Padilla, E., & Román, C. (2017). The long-term relationship ence and Technology, 40(21), 6555–6560. between CO2 emissions and economic activity in a small open econ- York, R., Rosa, E. A., & Dietz, T. (2003). STIRPAT, IPAT and ImPACT: Ana- omy: Uruguay 1882–2010. Energy Economics, 65, 271–282. lytic tools for unpacking the driving forces of environmental impacts. Roca, J., Padilla, E., Farré, M., & Galletto, V. (2001). Economic growth and Ecological Economics, 46(3), 351–365. atmospheric pollution in Spain: Discussing the environmental Kuznets curve hypothesis. Ecological Economics, 39(1), 85–99. Rosenblum, J., Horvath, A., & Hendrickson, C. (2000). Environmental impli- cations of service industries. Environmental Science and Technology, How to cite this article: Padilla Rosa, E., & Jadotte, E. (2023). 34(22), 4669–4676. The determinants of the inequality in CO2 emissions per Shafik, N. (1994). Economic development and environmental quality: An capita between developing countries and their implications for econometric analysis. Oxford Economic Papers, 46, 757–773. environmental policy. Sustainable Development, 31(1), Shafik, N., & Bandyopadhyay, S. (1992). Economic growth and environmen- tal quality: Time-series and cross-country evidence. Working papers 151–169. https://doi.org/10.1002/sd.2380 WPS 904 (Background paper for world development report 1992). The World Bank. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 165 APPENDIX A A.1 | RESULTS OF REGRESSIONS INCLUDING NON-SIGNIFICANT VARIABLES T A B L E A 1 Cross-country regression on the determinants of CO2 emissions per capita in developing countries, including industry share of GDP and population density 1995 2000 2005 2010 2014 Economic affluence GDPpci 0.0045938*** 0.0054248*** 0.0041544*** 0.0026783*** 0.0022976*** (0.0012373) (0.0010124) (0.0008207) (0.000783) (0.0006422) GDPpc2i 2.59E06*** 2.95E06*** 2.03E06*** 1.05E06*** 8.62E07*** (9.63E07) (7.31E07) (4.90E07) (3.93E07) (3.04E07) GDPpc3i 4.42E10** 5.05E10*** 3.07E10*** 1.31E10** 1.05E10** (2.19E10) (1.54E10) (8.85E11) (5.99E11) (4.33E11) Economic structure Agriculture_shi 0.0177908* 0.0054323 0.0098024 0.0244997** 0.0205419** (0.0095648) (0.0079245) (0.0083038) (0.0101146) (0.0092152) Industry_shi 0.0053623 0.0061171 0.0029325 0.0091257 0.00364 (0.0093939) (0.0069932) (0.0072079) (0.0075115) (0.0065966) Demography Urbanizationi 0.0112925 0.0052283 0.0166314** 0.0126455** 0.0116814** (0.0073531) (0.0067927) (0.0063475) (0.0055119) (0.0048604) Pop_15-64i 0.1120865*** 0.0955881*** 0.0883899*** 0.0706301*** 0.0614523*** (0.0273812) (0.0202917) (0.0176241) (0.0164901) (0.0146145) Pop_densityi 0.0004135 0.0000769 0.0001582 0.0003454 0.0004693 (0.0006476) (0.000537) (0.0004912) (0.0004654) (0.000394) Climate tmini 0.0463598*** 0.0383259*** 0.0192611* 0.0149795 0.0276824*** (0.0118543) (0.0109674) (0.0112633) (0.0118389) (0.0103326) Intercept 1.478362 1.587155 1.396747 0.7908071 1.565547 (1.596097) (1.13882) (1.034041) (0.9964531) (0.8785616) N 63 64 66 67 67 F 30.78*** 32.14*** 35.21*** 32.22*** 34.48*** R2 0.8394 0.8427 0.8498 0.8357 0.8448 Adjusted R2 0.8122 0.8165 0.8257 0.8098 0.8203 Note: Statistical significance is denoted by (*) at the 10% level (p < .10); (**) at the 5% level (p < .05); and (***) at the 1% level (p < .01). Standard errors within parentheses. Source: Prepared by the authors with World Bank data. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 166 PADILLA ROSA AND JADOTTE TABLE A2 Cross-country regression on the determinants of CO2 emissions per capita in developing countries, including average family size 1995 2000 2005 2010 2014 Economic affluence GDPpci 0.0039704*** 0.0048065*** 0.0041457*** 0.0024568*** 0.0020315*** (0.0013673) (0.0011625) (0.0008741) (0.0008457) (0.0006807) GDPpc2i 2.03E06* 2.52E06*** 2.03E06*** 9.77E07** 7.46E07** (1.03E06) (8.47E07) (5.48E07) (4.48E07) (3.40E07) GDPpc3i 3.15E10 4.29E10** 3.16E10*** 1.30E10* 9.25E11* (2.32E10) (1.81E10) (1.03E10) (7.16E11) (5.00E11) Economic structure Agriculture_shi 0.0217167** 0.0096116 0.0111549 0.0204713** 0.0202195** (0.010371) (0.0078925) (0.0078551) (0.009471) (0.0085152) Demography Urbanizationi 0.0122971 0.0028862 0.0149189* 0.0134293** 0.0129594** (0.0082827) (0.007874) (0.0074549) (0.0061357) (0.0054115) Pop_15-64i 0.100944*** 0.1119065*** 0.0990547*** 0.0861143*** 0.0684715*** (0.0301279) (0.0245045) (0.0203961) (0.0186352) (0.0156367) Householdi 0.0222988 0.0947441 0.0558641 0.1166426 0.074668 (0.0950194) (0.0927968) (0.0784022) (0.0740466) (0.0646928) Climate tmini 0.054808*** 0.0420796*** 0.0202116 0.0162542 0.0295373*** (0.0121619) (0.0120469) (0.012081 (0.0124682) (0.0107826) Intercept 0.3212388 2.377729 2.113467 0.9066825 0.7504471 (1.962344) (1.546185) (1.376608 (1.246411) (1.077843) N 55 56 58 60 60 F 31.13*** 30.21*** 35.36*** 33.08*** 33.96 R2 0.8441 0.8372 0.8524 0.8384 0.842 Adjusted R2 0.817 0.8095 0.8282 0.8131 0.8172 Note: Statistical significance is denoted by (*) at the 10% level (p < .10); (**) at the 5% level (p < .05); and (***) at the 1% level (p < .01). Standard errors within parentheses. tmini is almost significant at 10% for year 2005 (p = .101). Source: Prepared by the authors with World Bank data. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 167 TABLE A3 Cross-country regression on the determinants of CO2 emissions per capita in developing countries, including trade openness 1995 2000 2005 2010 2014 Economic affluence GDPpci 0.0037119*** 0.0050249*** 0.0041215*** 0.0027686*** 0.0023671*** 0.001145 2.69E06 0.0007785 0.0007807 0.0006228 GDPpc2i 1.98E06** 4.57E10*** 2.02E06*** 1.11E06*** 8.93E07*** 8.95E07 1.53E10 4.76E07 3.98E07 3.04E07 GDPpc3i 3.16E10 4.57E10*** 3.09E10*** 1.43E10** 1.10E10** 2.02E10 1.53E10 8.62E11 6.11E11 4.38E11 Economic structure Agriculture_shi 0.0166239** 0.0080251 0.0080282 0.0153005 0.0136497 0.0079512 0.0069669 0.0075331 0.0092707 0.0081951 Tradei 0.0055248** 0.0004076 0.0010436 0.0005195 0.0009727 0.0026189 0.0021817 0.0017111 0.0025878 0.0021213 Demography Urbanizationi 0.0126648** 0.0072037 0.0192472*** 0.0162959* 0.015175*** 0.0061968 0.0062956 0.0061576 0.0061336 0.0054625 Pop_15-64i 0.1248726*** 0.101317*** 0.0911194*** 0.0685034*** 0.0542655*** 0.0223851 0.0180123 0.0151579 0.015126 0.0133868 Climate tmini 0.0346591*** 0.0380528*** 0.0174526* 0.0197887* 0.0344158*** 0.0102016 0.0107291 0.0100845 0.0112401 0.0097276 Intercept 2.431804* 1.671822 1.764719* 0.2868031 1.511569* 1.40475 1.038934 0.9066277 0.9343298 0.8225286 N 60 64 64 64 65 F 42.38 37.23 43.19 33.78 38.72 2 R 0.8693 0.8441 0.8627 0.8309 0.8469 Adjusted R2 0.8487 0.8214 0.8427 0.8063 0.825 Note: Statistical significance is denoted by (*) at the 10% level (p < .10); (**) at the 5% level (p < .05); and (***) at the 1% level (p < .01). Standard errors within parentheses. Agriculture_shi is almost significant at 10% for years 2010 and 2014 (p = .105 and .101, respectively). Source: Prepared by the authors with World Bank data. 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 168 PADILLA ROSA AND JADOTTE TABLE A4 Countries included in the estimations shown in the main text 1995 2000 2005 2010 2014 Angola Angola Afghanistan Afghanistan Afghanistan Burundi Burundi Angola Angola Angola Benin Benin Burundi Burundi Burundi Burkina Faso Burkina Faso Benin Benin Benin Bangladesh Bangladesh Burkina Faso Burkina Faso Burkina Faso Bolivia Bolivia Bangladesh Bangladesh Bangladesh Bhutan Bhutan Bolivia Bolivia Bolivia Cote d'Ivoire Cote d'Ivoire Bhutan Bhutan Bhutan Cameroon Cameroon Cote d'Ivoire Central African Republic Central African Republic Congo, Dem. Rep. Congo, Dem. Rep. Cameroon Cote d'Ivoire Cote d'Ivoire Congo, Rep. Congo, Rep. Congo, Dem. Rep. Cameroon Cameroon Comoros Comoros Congo, Rep. Congo, Dem. Rep. Congo, Dem. Rep. Cabo Verde Cabo Verde Comoros Congo, Rep. Congo, Rep. Egypt, Arab Rep. Egypt, Arab Rep. Cabo Verde Comoros Comoros Eritrea Eritrea Egypt, Arab Rep. Cabo Verde Cabo Verde Ethiopia Ethiopia Eritrea Egypt, Arab Rep. Egypt, Arab Rep. Ghana Ghana Ethiopia Ethiopia Ethiopia Guinea Guinea Ghana Ghana Ghana Gambia, The Gambia, The Guinea Guinea Guinea Guinea-Bissau Guinea-Bissau Gambia, The Gambia, The Gambia, The Honduras Honduras Guinea-Bissau Guinea-Bissau Guinea-Bissau Haiti Haiti Honduras Honduras Honduras Indonesia Indonesia Haiti Haiti Haiti India India Indonesia Indonesia Indonesia Kenya Kenya India India India Kyrgyz Republic Kyrgyz Republic Kenya Kenya Kenya Cambodia Cambodia Kyrgyz Republic Kyrgyz Republic Kyrgyz Republic Kiribati Kiribati Cambodia Cambodia Cambodia Lao PDR Lao PDR Kiribati Kiribati Kiribati Lesotho Liberia Lao PDR Lao PDR Lao PDR Morocco Lesotho Liberia Liberia Liberia Moldova Morocco Lesotho Lesotho Lesotho Madagascar Moldova Morocco Morocco Morocco Mali Madagascar Moldova Moldova Moldova Mongolia Mali Madagascar Madagascar Madagascar Mozambique Myanmar Mali Mali Mali Mauritania Mongolia Myanmar Myanmar Myanmar Malawi Mozambique Mongolia Mongolia Mongolia Niger Mauritania Mozambique Mozambique Mozambique Nigeria Malawi Mauritania Mauritania Mauritania Nicaragua Niger Malawi Malawi Malawi Nepal Nigeria Niger Niger Niger Pakistan Nicaragua Nigeria Nigeria Nigeria Philippines Nepal Nicaragua Nicaragua Nicaragua Papua New Guinea Pakistan Nepal Nepal Nepal Rwanda Philippines Pakistan Pakistan Pakistan 10991719, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/sd.2380 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [01/03/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License PADILLA ROSA AND JADOTTE 169 TABLE A4 (Continued) 1995 2000 2005 2010 2014 Sudan Papua New Guinea Philippines Philippines Philippines Senegal Rwanda Rwanda Papua New Guinea Papua New Guinea Solomon Islands Sudan Sudan Rwanda Rwanda Sierra Leone Senegal Senegal Sudan Sudan El Salvador Solomon Islands Solomon Islands Senegal Senegal Eswatini Sierra Leone Sierra Leone Sierra Leone Sierra Leone Chad El Salvador El Salvador El Salvador El Salvador Togo Eswatini Sao Tome and Principe Sao Tome and Principe Sao Tome and Principe Tajikistan Chad Eswatini Eswatini Eswatini Tunisia Togo Chad Chad Chad Tanzania Tajikistan Togo Togo Togo Uganda Tunisia Tajikistan Tajikistan Tajikistan Ukraine Tanzania Tunisia Tunisia Tunisia Uzbekistan Uganda Tanzania Tanzania Tanzania Vietnam Ukraine Uganda Uganda Uganda Vanuatu Uzbekistan Ukraine Ukraine Ukraine Yemen, Rep. Vietnam Uzbekistan Uzbekistan Uzbekistan Zambia Vanuatu Vietnam Vietnam Vietnam Zimbabwe Yemen, Rep. Vanuatu Vanuatu Vanuatu Zambia Yemen, Rep. Yemen, Rep. Yemen, Rep. Zimbabwe Zambia Zambia Zambia Zimbabwe Zimbabwe Zimbabwe