Cross-country distribution dynamics of carbon emissions and intensity: Before and after the global financial crisis

Wei, Yigang; Wojewódzki, Michał; Cheong, Tsun Se; Shi, Xunpeng Article Cross-country distribution dynamics of carbon emissions and intensity: Before and after the global financial crisis Review of Economic Analysis (REA) Provided in Cooperation with: International Centre for Economic Analysis (ICEA), Waterloo, Ontario Suggested Citation: Wei, Yigang; Wojewódzki, Michał; Cheong, Tsun Se; Shi, Xunpeng (2024) : Crosscountry distribution dynamics of carbon emissions and intensity: Before and after the global financial crisis, Review of Economic Analysis (REA), ISSN 1973-3909, International Centre for Economic Analysis (ICEA), Waterloo (Ontario), Vol. 16, Iss. 3, pp. 343-370, https://doi.org/10.15353/rea.v16i3.5332 This Version is available at: https://hdl.handle.net/10419/328169 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/4.0/ Review of Economic Analysis 16 (2024) 343-370 1973-3909/2024343 343 www.RofEA.org Cross-country distribution dynamics of carbon emissions and intensity: Before and after the global financial crisis YIGANG WEI Beihang University and Beijing Key Laboratory of Emergency Support Simulation Technologies for City Operation, Beijing, China MICHAŁ WOJEWÓDZKI Lingnan University, Hong Kong TSUN SE CHEONG Hang Seng University of Hong Kong XUNPENG SHI University of Technology Sydney This study aims to explore the levels of regional disparity in carbon emissions and intensity among different countries. Our study employs the distribution dynamics approach to uncover transition probabilities and the long-term evolution of relative per capita carbon emissions (REPC) and relative carbon intensity (REPGDP) across 204 countries. We split the analysis period into pre-crisis (2000-2007) and post-crisis (2007-2016) and divided countries into four income groups. The results indicate the emergence of new convergence clubs post-crisis in both REPC and REPGDP. Furthermore, the majority (many) of the low- (high) income countries congregate to extremely low (above the global average) REPC levels in the long run. Finally, using mobility probability plots, we identify low-  Wojewódzki: corresponding author, Department of Finance, Faculty of Business, Lingnan University, 8 Castle Peak Rd - Lingnan, Tuen Mun, New Territories, Hong Kong; michalwojewodzk[email protected]. Wei: School of Economics and Management, Beijing, China and Beijing Key Laboratory of Emergency Support Simulation Technologies for City Operation, Beijing, [email protected]; Cheong: The Hang Seng University of Hong Kong, School of Business, Hang Shin Link, Shatin, New Territories, Hong Kong SAR, [email protected]; Shi: The Australia-China Relations Institute, University of Technology Sydney, ACRI, 15 Broadway, Ultimo, NSW, 2007, Australia, [email protected]. Competing interests and funding: none Acknowledgements: The research is supported by grants from the Beijing Social Science Fund (20GLC054). The authors also thank the anonymous reviewers for insightful comments that helped us improve the quality of the paper. © 2024 Yigang Wei, Michał Wojewódzki, Tsun Se Cheong and Xunpeng Shi. Licensed under the Creative Commons Attribution - Noncommercial 4.0 Licence (http://creativecommons.org/licenses/by-nc/4.0/. Available at http://rofea.org. Review of Economic Analysis 16 (2024) 343-370 344 www.RofEA.org (high-) income countries with REPC (REPGDP) levels of 2.3 (3.8) times the global average to have the highest probabilities of around 100 (65) per cent of diverging further above the worldwide average in the coming years. The study expands knowledge about convergence-divergence patterns in carbon emissions and intensity, which is crucial for energy management and effective climate policies. Moreover, it can aid in climate change projections and promote a fairer climate framework, encouraging high-emission countries to take greater responsibility Keywords: Carbon emissions; Global financial crisis; Distribution dynamics; Convergence clubs; Income levels JEL Classifications: C14; G01; O13; Q54 1 Introduction Greenhouse gas (hereafter GHG) emitted into the atmosphere from burning fossil fuels significantly contributes to global climate change. Thus, effective control of the relentless growth of GHG emissions has become an irreplaceable mission for global society (Shahrour et al 2023). Accordingly, 196 countries signed the Paris Agreement in 2015 with the primary target of slowing down global warming to below 2°C (if possible 1.5°C) compared to preindustrial levels (Cheng et al 2023). The signatories also pledged to define their climate actions known as Nationally Determined Contributions (hereafter NDCs). However, Climate Action Tracker (2021) forecasted that assuming all of the existing NDCs were delivered, the global temperature would still increase by 2.4°C. Furthermore, the NDCs are voluntary actions, thereby giving rise to countries under-committing, which, together with the implementation gap in numerous countries, makes reaching the global warming goals unlikely. The knowledge of convergence-divergence patterns in carbon emissions is significant in energy management and formulation of efficient climate measures and policies (e.g., Kounetas 2018, Li and Wei 2021). Knowing the dynamic trajectory of emissions also helps facilitate climate change projections. Moreover, the convergence could result in a fairer climate framework, thereby increasing the odds of countries with high emissions assuming a greater responsibility for climate commitments (Aldy 2006, Delgado 2013, Erdogan and Solarin 2021). Indeed, the Kyoto and Paris Accords highlight the substantial positive impact of emissions convergence and shared environmental policies on the probability of reaching the desired goals. Against this backdrop, the convergence in cross-country carbon emissions has attracted growing attention in the environmental economics literature (e.g., Rios and Gianmoena 2018, Li et al 2020). On the one hand, numerous studies test the presence of convergence in per capita carbon emissions measures (e.g., Westerlund and Basher 2008, Li et al 2020). On the other hand, the recent surveys by Acar et al (2018) and Payne (2020) show that the convergence of CO2 intensity (per unit of GDP emissions) remains relatively unexplored, and the empirical results are mixed at best. Such gaps in the literature are worrying, given that many countries WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 345 www.RofEA.org state carbon commitments in terms of either CO2 emissions or CO2 intensity 1 (Bhattacharya et al 2020). Furthermore, the two measures of a country’s carbon emissions can take opposite routes (Parker and Bhatti, 2020). In terms of research methods, many studies test the presence of sigma, beta, and stochastic convergence (e.g., Jobert et al 2010, Li and Lin 2013, Erdogan and Solarin, 2021). Other parametric, semi-parametric, and nonparametric methods have also been used to test absolute vis-à-vis conditional convergence (e.g., Brock and Taylor 2010, Delgado 2013), stochastic convergence (Chang and Lee 2008), spatial convergence (Rios and Gianmoena 2018), and club convergence (Parker and Bhatti 2020). However, fewer studies employ the distribution dynamics approach (hereafter DDA) developed by Quah (1993, 1997) (e.g., Aldy 2006, Criado and Grether 2011). To the best of our knowledge, Kounetas (2018), Li et al (2021) and Wojewodzki et al (2023) are the only studies examining the convergence in both measures of carbon emissions using the DDA. However, they neither investigated the transitional dynamics of carbon emissions across countries focusing on different income levels nor examined the effect of the GFC. Moreover, Kounetas (2018) uses a relatively small sample limited to 23 EU countries, and his analysis does not include the MPP tool. Considering the above-outlined gaps in environmental research, this study makes three new contributions to the environmental economics literature. To the best of our knowledge, we are first to examine the long-run, dynamic convergence-divergence pattern of (1) relative per capita carbon emissions (hereafter REPC) and (2) relative carbon intensity (hereafter REPGDP) across 204 countries 2 . Furthermore, we deliver a nascent analysis of the impacts of the 2008 global financial crisis (hereafter, GFC) and unprecedented expansionary fiscal and monetary policies on the future evolution of global REPC and REPGDP levels. For instance, we document worrying developments in transitional dynamics and long-run steady-state equilibria of both REPC and REPGDP measures post-GFC period compared with the pre-GFC period. Such findings might indicate the negative effect of GFC-related expansionary policies (e.g., massive borrowing, quantitative easing, and investment in carbon-intensive industries) on both carbon emissions measures' long-run steady-state equilibria and convergence process. Third, this paper delves into the profound influence of carbon emissions and intensity on the distribution dynamics of nations grouped by income levels. Specifically, along with the 1 For example, China and India have stated their emissions reduction goals regarding CO2 intensity (Bhattacharya et al. 2020). 2 REPC (REPGDP) is a ratio of country-specific annual per capita carbon emissions (intensity) to the average carbon emissions (intensity) of all sampled countries in a given year. Because of the global scope of this research (204 countries), we treat the sample’s annual average value as a proxy for the global averages. Therefore, a country’s REPC or REPGDP above (below) one implies that this country’s REPC or REPGDP is above (below) the global average in a given year. Review of Economic Analysis 16 (2024) 343-370 346 www.RofEA.org ergodic distribution, we also use a novel display tool of DDA introduced by Cheong and Wu (2018), the Mobility Probability Plot (hereafter, MPP). This new tool offers specific visual information regarding the probability mass distribution in the coming years. We document that the low- (high) income countries with REPC (REPGDP) values of around 2.3 (3.8) times the global average emissions have the most significant probabilities of around 100 (65) per cent of diverging further above the worldwide mean. This information, in turn, translates into a “policy priority list” of countries meriting the most urgent climate policies and actions. The rest of this study is organised as follows. Section 2 provides an extensive review of relevant literature. Section 3 introduces data and research methods. Section 4 discusses the empirical results. The last section concludes the research findings and policy implications. 2 Literature Review A burgeoning body of environmental literature has tested the convergence hypothesis of crosscountry per capita carbon emissions. For example, Ezcurra (2007) use the DDA and find strong evidence of convergence towards the mean across 140 countries from 1960 to 1999. On the contrary, based on the same method and a panel of 166 countries from 1960 to 2002, Criado and Grether (2011) document an increased divergence and higher levels of CO2 emissions in the long run. El-Montasser et al (2015) also report the results inconsistent with emissions convergence in a study of high-income G7 countries. Most recently, Lee et al (2023) employed stochastic convergence to analyse per capita emissions for 30 OECD countries from 1960 to 2018, revealing the lack of convergence. On the contrary, Chang and Lee (2008) find significant evidence of stochastic convergence in a panel of 21 high-income OECD countries using the minimum LM unit root tests. More recently, Rios and Gianmoena (2018) test the spatial convergence clubs hypothesis vis-à-vis the conditional convergence hypothesis in a sample of 141 countries and document the emergence of three clubs. However, in a worldwide sample of countries, Fallahi (2020) show that per capita carbon emissions are non-stationary and highly persistent at both the global and regional levels. Many researchers examine per capita carbon emissions, focusing on countries’ economic development and income levels. For instance, in the sample of 100 countries, Nguyen Van (2005) employs the DDA and shows a long-run convergence only across 26 developed countries. Similarly, Aldy (2006) documents converge in emissions across only 23 developed OECD members but find an opposite pattern (divergence) among 88 countries. Payne and Aspergis (2021) identify three convergence groups among 27 low-income countries and five clubs among 38 lower-middle-income countries from 1972 to 2014. In a sample spanning from 1870 to 2002, Westerlund and Basher (2008) and Delgado (2013) find evidence of the convergence across developed and developing countries alike. Similarly, Ahmed et al (2017) study 162 countries and show convergent patterns in 20, 13 and 5 high-income, middle-income, WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 347 www.RofEA.org and low-income countries, respectively. According to Li et al (2020), convergence is significantly faster across developed countries vis-à-vis developing countries. The empirical literature suggests that the episodes of financial crises significantly influence global carbon emissions, intensity, and convergence patterns. In a study of fourteen Asian countries, Parker and Bhatti (2020) document different transition paths in the convergence process of per capita carbon emissions before and after the 1997 Asian financial crisis. The authors observe the emergence of four (three) convergence clubs pre- (post) crisis. According to Wang et al (2021), the shock of GFC brought an initial U-turn in an ongoing upward (downward) trend in global carbon emissions (intensity), followed by a return to the previous pattern between 2009 and 2011. Li et al (2020) argue that the GFC and unprecedented postcrisis policies adopted by different countries have led to cross-country divergence in emissions. Recent surveys of empirical studies (Acar et al 2018, Payne 2020) agree that while the convergence hypothesis has been tested extensively for cross-country per capita carbon emissions, relatively few researchers have examined the convergence of CO2 intensity. For instance, Lindmark (2004) graphically examines the patterns of carbon intensity across 56 countries during the 1870-1992 period and reports high-income countries converging with their low-income counterparts. Camarero et al (2013) use Phillips and Sul's (2007) convergence-club approach to identify four clubs of carbon intensity across 19 OECD countries and a nonconvergent pattern in four industrialised countries. Zhu et al (2014) find evidence of convergence in a panel of 89 countries between 1980 and 2008. Zang et al (2018) examine a sample of 201 countries from 2003 to 2015 and document club convergence among all three groups of countries: low-, middle-, and high-income. According to Bhattacharya et al (2020), two convergence clubs exist in cross-country consumption-based carbon emissions. Moreover, their forecast based on a panel of 70 countries suggests that between 2014 and 2030, the number of convergence clubs in consumption-based carbon intensity will increase. Nevertheless, the dynamic aspects of per capita CO2 emissions vis-à-vis CO2 intensity remain largely unexplored. Only three cross-country studies employ the DDA approach to carbon emissions and intensity at the same time. In a sample of 23 EU countries during the 1970-2010 period, Kounetas (2018) find no evidence of convergence. Li et al (2021) investigate the REPC and REPGDP measures in 178 countries. They document significant disparities in the transitional dynamics of both measures between 71 countries that signed the Belt and Road Initiative (hereafter BRI) cooperation with China and 107 non-BRI countries. Wojewodzki et al (2023) study transitional dynamics and evolution of per capita CO2 emissions and CO2 intensity across countries with different urbanisation levels and agrarian orientations. However, none of the three studies examine (1) the dynamics and future evolution of carbon emissions and intensity concerning countries’ income levels and (2) the effect of the GFC on transitional dynamics and long-run steady-state equilibrium. Review of Economic Analysis 16 (2024) 343-370 348 www.RofEA.org The existing body of environmental literature has yielded conflicting results 3 and left some critical gaps. We aim to fill the abovementioned gaps by investigating the transitional dynamics of carbon emissions and intensity in a global sample of 204 countries from 2000 to 2016. This study adopts the DDA approach and the MPP tool that Cheong and Wu developed (2018). 3 Methodology and Data The data are compiled from the World Bank’s World Development Indicators (WDI) database. Two measurements of carbon emissions are employed: relative CO2 emissions per capita (REPC) and relative CO2 intensity (REPGDP). REPC is measured as metric tons of CO2 per capita in each country divided by the world average in a particular year. Similarly, REPGDP is measured as kg of CO2 emissions per 2010 USD of GDP in each country divided by the world average for this year. Using relative values, the disparity amongst the nations can be displayed directly as a REPC or REPGDP value of 1, which indicates that a country’s CO2 emissions or intensity is equal to the global average. In contrast, a value smaller (greater) than 1 means the level below (above) the global average. The unbalanced panel covers 204 countries, starting in 2000 and ending in 2016 4 . That is because the data for 2017 and beyond were unavailable for many countries at the time of collection. Table 1 presents annual descriptive statistics for both variables. We can observe one measure of central location (median) and one measure of dispersion (the coefficient of variation, hereafter, CV). Table 1. Annual (2000-2016) descriptive statistics for the REPC and REPGDP variables Variable Year (2000-2016) Global REPC 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 Median 0.59 0.59 0.59 0.58 0.53 0.54 0.59 0.55 0.55 0.60 0.55 0.56 0.55 0.57 0.57 0.57 0.58 CV 1.43 1.45 1.42 1.39 1.36 1.37 1.38 1.33 1.28 1.28 1.25 1.25 1.26 1.23 1.26 1.25 1.22 Global REPGDP Median 0.78 0.79 0.77 0.75 0.74 0.75 0.74 0.75 0.70 0.74 0.72 0.71 0.71 0.72 0.74 0.75 0.78 CV 1.07 1.03 1.04 1.03 1.00 0.95 0.94 0.93 0.90 0.84 0.84 0.86 0.92 0.94 0.82 0.79 0.79 Source: Authors’ calculations based on the World Bank’s World Development Indicators (WDI) 3 Mixed results in prior studies can be attributed to different sample sizes and research methods. Specifically, studies using larger samples and dynamic analyses tend to find evidence against crosscountry convergence. 4 The number of countries is the lowest (199) in 2000, while 204 were covered between 2008 and 2016. This means that each country has a minimum of 9 consecutive annual observations. WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 349 www.RofEA.org Table 1 shows similar decreasing trends regarding the variability in global carbon emissions and intensities. Specifically, the CV values for REPC (REPGDP) dropped from 1.43 or 143% (1.07 or 107%) to 1.22 (0.79). This means a presence of sigma convergence in two major proxies for emissions over time, similar to the results of Zang et al (2018). This, in turn, constitutes good news from the perspective of desirable cross-country convergence to the global mean. However, the percentage change (decrease) in CV between 2000 and 2016 was larger (85%) for REPC than that for REPGDP (74%), at odds with the changes documented by Zang et al (2018) for the 2003-2015 period. Notwithstanding, global annual variability in the REPGDP has remained substantially larger than that of REPC throughout the sampled period. As for the median values, these remained virtually unchanged between 2000 and 2016, ranging from 0.6 to 0.53 (0.79 to 0.7) in 2009 and 2004 (2001 and 2008) for the REPC (REPGDP) variable. Since the global annual mean value for both measures of relative carbon emissions equals one, we can conclude that the distribution of both variables is negatively skewed, especially for the REPC measure, due to its median values being significantly below one. This highlights relatively fewer (more) countries with REPC and REPGDP above (below) the global average during the investigation period. Time series econometric analyses are frequently used in forecasting. However, it is essential to acknowledge that econometric models are limited in their ability to forecast the dependent variable, as they do not provide insights into the shape of the underlying distribution (Liu et al 2022). Instead, such methods only offer insights into several significant distribution features since they solely focus on predicting the dependent variable. Notwithstanding, since distribution is a two-dimensional entity, it is not feasible to predict the future distribution's overall shape solely through time series econometrics. Consequently, by ignoring information regarding, e.g., multimodal distributions, the econometric analysis may lead to contaminated or misleading results (Quah 1997, Maasoumi et al 2007). Likewise, traditional (econometric) methods cannot comprehensively overview the distribution pattern and its dynamic changes (Cheong and Wu 2018). In contrast, the DDA developed by Quah (1993, 1997) focuses on examining the shape of the distribution and how it changes over time. Thus, while traditional econometric techniques can be utilised to calculate the slope parameter and assess the impact of a driving factor, the DDA allows for examining the effects of determinants on the entire distribution. This involves dividing the data into smaller datasets and applying the DDA to each subset individually. By comparing the distributions of these datasets, one can better understand the impacts of different driving factors 5 . 5 However, one limitation of the DDA is that the study of determinants can only be conducted one at a time. Review of Economic Analysis 16 (2024) 343-370 350 www.RofEA.org Furthermore, the DDA possesses a significant advantage over traditional econometric analysis regarding its resilience to outliers. This stems from the fact that the computation of the DDA relies on the probability of entities transitioning between different states, which is contingent upon the occurrence of the entities rather than their measured values. In stark contrast, traditional econometric analysis is susceptible to the influence of outliers, as it relies on calculating the slope parameter, which can be significantly affected by outliers. Moreover, one of the tools of the DDA is the ergodic distribution, which represents the future trend's steady-state distribution (Wei et al 2020). This analytical technique can comprehensively depict the underlying trend and the future evolution and intensity of, for example, carbon emissions. Quah’s DDA can be divided into two major categories: the traditional Markov transition matrix analysis and the stochastic kernel approach. One issue of the traditional Markov transition matrix analysis is the arbitrary boundary of the state associated with the selection of grid values. In contrast, demarcation can be achieved objectively in the stochastic kernel approach, which can be viewed as an improvement of the traditional Markov transition matrix approach. As a result, this study employs the stochastic kernel approach. The bivariate kernel estimator can be represented as the equation (1). 𝑓 󰆹(𝑥, 𝑦)= 1 𝑛ℎ1ℎ2 ∑𝐾(𝑥−𝑋𝑖,𝑡 ℎ1,𝑦−𝑋𝑖,𝑡+1 ℎ2) 𝑛 𝑖=1 (1) where n stands for the number of observations, and x (y) stands for the relative CO2 emissions of an entity at period t (t+1). Xi,t (Xi,t+1) represents an observed value of relative CO2 emissions at time t (t+1). Furthermore, terms h1 and h2 correspond to the bandwidths computed using the approach established by Silverman (1986), and K is the normal density function. Due to the data sparseness, we use the adaptive kernel with flexible bandwidth, which was first proposed by Silverman (1986). There are two steps in implementation. It involves the computation of a pilot estimate at the beginning, and then the bandwidth is adjusted by a factor that reflects the kernel density. Under restrictive assumptions that the studied variable’s distribution at time t +τ depends on t only, the process is of first order and doesn’t change over time; the relationship between the distributions at periods t and t +τ can be represented by the equation (2). 𝑓𝑡+𝜏(𝑧)= ∫𝑔𝜏 ∞ 0(𝑧|𝑥)𝑓𝑡(𝑥)𝑑𝑥 (2) where 𝑔𝜏(𝑧|𝑥) represents the transition probability kernel, which plots the distribution from period t to t +τ, while the term 𝑓𝑡(𝑥) stands for the kernel density function of the variable’s distribution at period t. Moreover, the term 𝑓𝑡+𝜏(𝑧) captures the τ-period-ahead density function of z conditional on x. Because annual transitions are used in the analysis, the sample size will be larger, and the estimation results will be more reliable. Given that it exists, the ergodic density function can be calculated from equation (3). WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 357 www.RofEA.org Figure 5. Time trends in annual CVs of REPC variable by different income groups Source: Authors’ calculations based on the World Bank’s World Development Indicators (WDI) The picture painted in Fig 6. largely contrasts that in Fig 5. First, we can observe that the lowest variability, with an overall decreasing trend over time, can be attributed to the low-income countries. Figure 6. Time trends in annual CVs of REPGDP variable by different income groups Source: Authors’ calculations based on the World Bank’s World Development Indicators (WDI) 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 C.V. of REPC High-income Low-income 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 C.V. of REPGDP High-income Low-income Review of Economic Analysis 16 (2024) 343-370 358 www.RofEA.org On the contrary, the annual CVs for the most affluent economies (yellow-coloured plot) followed an increasing trend (from 0.76 to 0.95), translating into a robust divergence between 2000 and 2016. While we can observe a similar reversed (increasing) pattern in the bluecoloured plot (lower-middle-income) to that in Figure 5, this group of countries made the most significant progress regarding within-group convergence in carbon emissions. While Figure 5 and 6, based on the sigma convergence tool (CV), are highly informative, they are limited in their ability to forecast the dependent variable, as they do not provide insights into the shape of the underlying distribution. Given the above backdrop, we are the first to employ the ergodic distribution and the MPP to analyse REPC and REPGDP across countries divided into four groups based on income levels 6 . Figure 7 indicates that most low-income and many lower-middle-income countries congregate around minimal REPC levels in the long-run steady-state equilibrium. On the other hand, most high-income economies converge to above the global average REPC levels. Such apparent divergent long-run trend between the poorest and wealthiest countries is grim news and at odds with the EKC. This, together with pre/post-GFC analysis (see Fig. 1), suggests that the GFC and expansionary policies introduced by many industrialised countries could result in a long-run increase in REPC and the persistent income gap between the rich and the poor. Figure 7 also highlights by far the most (least) significant convergence process for the low- (high-) income group of countries captured by the tallest (shortest) and the least (most) spread out ergodic distribution. Such findings starkly contrast with the evidence reported by Li and Lin (2013). Furthermore, we can observe that except for upper-middle-income countries, multiple convergence clubs emerge, whilst all peaks in panels A and B (D) are situated below (above) the global average REPC equal to one. Thus, assuming the transitional dynamics remain unchanged, we can expect only conditional convergence in REPC across the three groups of countries: low-, lower-middle-, and high-income. Such a finding contrasts that of Zang et al (2018), who examined a sample of 201 countries from 2003 to 2015 and found evidence of club convergence only among high-income countries. In addition, we can observe that ergodic distributions become more spread out as we move up the income ladder, i.e., from panel A to panel D. This observation, in turn, is contrary to the results of Li et al (2020) and suggests the least and the most significant convergence process across the richest and the poorest countries, respectively. The findings in Fig 7. highlight the complex relationship between income levels and carbon emissions, emphasising the need for targeted policies and international cooperation to address climate change effectively across diverse economic contexts. 6 See Table A1 in the Appendix for the list of countries by their income levels as of the end of 2016. During the 2000-2016 period, numerous (a few) countries have moved into higher (lower) income brackets in line with the World Bank’s annual classification. As of 2016, 30, 53, 55, and 66 low-, lowermiddle, upper-middle, and high-income countries are in the sample. WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 359 www.RofEA.org Figure 7. Ergodic distributions for REPC by different income groups A Low-income B Lower-middle-income C Upper-middle-income D High-income Notes: The horizontal and vertical axes show the value of the REPC variable and the proportion, respectively. Figure 8 shows that the MPP for the high (low) income countries is the least (the most) volatile, translating into the least (the most) significant aggregate net mobility probability in years to come. The weakest aggregate net mobility probability suggests a lower likelihood of substantial shifts in high-income countries' per capita carbon emissions patterns over time. Instead, we can expect a relatively stable or stagnant emissions trajectory, with gradual or incremental changes in REPC. This, in turn, could result from the interplay of established infrastructure, mature economies, and strict environmental regulations that limit rapid changes in carbon emissions of the high-income countries. On the contrary, more volatile aggregate net mobility probability signals a more dynamic and evolving REPC landscape, reflecting rapid changes in economic 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Review of Economic Analysis 16 (2024) 343-370 360 www.RofEA.org activities, rapid industrialisation, technological transitions, policy interventions, and limited regulatory oversight across many low-income countries. Figure 8. MPPs for REPC by different income groups Notes: The horizontal and vertical axes show the value of the REPC variable and the MPP, respectively. Additional important implications can be drawn from identifying (1) the sections of MPPs above the global average REPC values positioned above the horizontal axis and (2) the intersections and tangent points between the MPPs and the horizontal axis. That is because (1) it enables us to pinpoint carbon emitters above the global average with a specific range of REPC levels and exact probabilities to diverge further away/above the global average in the coming years. Similarly, (2) indicates sticky and above the global average REPC values around which the entities from different income groups would congregate in years to come. Given limited resources and environmental policy goals of convergence/reduction in CO2 emissions (Wei et al 2022, Wei et al 2023), countries with above-average REPC levels and net upward mobility probabilities ranging from 100 to zero should be placed on the policy priority list. Figure 9 indicates that for the low- (lower-middle) income countries, the alarming ranges of REPC occur from 1.5 to 2.6 (1.25 to 1.4 and 2 to 2.45), while for the upper-middle- (high) income countries around a REPC value of 1.5 and from 2.45 to 2.6 (below 1.25). However, the top spot in the policy priority list should be reserved for low-income countries with REPC values of around 2.3. This is because the MPP representing these entities reaches the maximum WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 361 www.RofEA.org positive net mobility probability of 100 at a REPC value of 2.3. This, in turn, implies that upon reaching such a REPC level, the country’s emissions have a 100 per cent probability of moving even further above the global average in years to come. Moreover, Figure 8 indicates that lowand upper-middle-income countries experience a development trap in their REPC. This is a positive piece of information because it means that whenever lowand upper-middle-income countries achieve REPC values of 2.9 and 4, they will encounter a reduction in REPC and a downward move within the distribution in the following years. However, the green and yellow plots representing more affluent economies approach the development trap at higher REPC levels of 9 and 21.5, respectively. Summing up, findings based on the first measure of CO2 emissions (REPC) are essential from the perspective of future environmental policies aiming at emissions reduction and convergence. For instance, the results imply the onus on high-income countries to reduce their per capita carbon emissions. Furthermore, more significant variability in the REPC observed across the rich countries suggests they would have more spare capacity and flexibility in reducing CO2 emissions. On the other hand, the poorer countries appear to have ample time before converging to the global average, except for the outliers with a net mobility probability of 100 at a REPC value of 2.3, identified in Figure 8. Figure 9 shows that the ergodic distribution for the low-income (lower-middle-income) group is denser (more spread out), with more entities converging around lower (higher) REPGDP values corresponding to three peaks located at 0.35, 0.6 and 1.8 (0.7, 1.1 and 2). Furthermore, the distribution in panel C for the upper-middle-income countries is the most dispersed, with many countries congregating around REPGDP values far above the mean (2.4). However, the most concentrated is the ergodic distribution for the wealthiest group of countries, with a single peak significantly below the global mean (REPGDP value of 0.32) and a long, thin right tail. Concerning the emergence of convergence clubs, these appear in panels A to C only, signifying conditional (absolute) convergence at best across lowand medium-income (high-income) countries. Such observation only partially corroborates Zang et al (2018) results of clubs across all income groups in their study of 201 countries. Overall, findings from Figure 9 suggest that the onus is on the upper-middle-income emitters to reduce carbon intensity because some countries from this income group converge toward the long-run carbon intensities significantly above the global average. Such a finding corroborates Dong et al (2020), who document that the upper-middle-income is the primary source of worldwide CO2 emissions post-2008 financial crisis. Moreover, the results suggest that middle-income economies incur relatively higher environmental externalities (emissions per USD of GDP) for their economic development (Zang et al 2018, Wang et al 2021). Nevertheless, the most dispersed ergodic distribution for upper-middle-income countries implies greater flexibility in reducing carbon intensity. Policymakers could leverage this flexibility to implement policies and strategies tailored to specific country contexts, including Review of Economic Analysis 16 (2024) 343-370 362 www.RofEA.org investment in clean technologies, renewable energy, energy efficiency, and sustainable infrastructure. Figure 9. Ergodic distributions for REPGDP by different income groups A Low-income B Lower-middle-income C Upper-middle-income D High-income Notes: The horizontal and vertical axes show the value of the REPGDP variable and the proportion, respectively. The MPPs in Figure 10 indicate that countries with REPGDP values below 0.35, irrespective of income levels, experience a positive net probability of moving upward in the distribution in future years. From the policy perspective, Figure 10 is interesting because it pinpoints the countries with specific above-the-global average REPGDP values and positive net probability of moving higher in the future distribution. Therefore, assuming the emissions reduction goals are based on CO2 intensity, such countries merit special attention, i.e., they should enter the climate policy priority list. We can observe that low-income countries with a range of REPGDP values around 1.65, 3.15, and 6.3 should be placed on the priority list due to the stickiness of their above-average relative emissions. By the same token, high-income countries with relative carbon intensities WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 363 www.RofEA.org between 3.3 and 4.5 are also problematic. In particular, industrialised economies with REPGDP values around 3.8 have the most significant (65%) probability of further diverging from and above the global average regarding their REPGDP in years to come. This, in turn, places them high on the policy priority list. On the contrary, two middle-income groups appear to be the least problematic. Figure 10. MPPs for REPGDP by different income groups Notes: The horizontal and vertical axes show the value of the REPGDP variable and the MPP, respectively. 5 Conclusion and Policy Implications The main findings of this study can be summarised in three points. First, the post-GFC period ergodic distributions are characterised by the emergence of additional convergence clubs in relative per capita carbon emissions (REPC) and relative carbon intensity (REPGDP). Besides, the post-GFC distribution is significantly more spread out, suggesting that the global long-run convergence process in REPC ex-post-GFC becomes less significant. Such results, in turn, might imply the adverse effects of GFC and post-crisis expansionary policies aimed at economic recovery undertaken by developing and developed countries (e.g., in China and the US). Overall, the results are pessimistic from the perspective of global warming and environmental policies because the more significant number of clubs makes future international climate negotiations and ambitious targets more complex and challenging. Second, regarding countries grouped by income levels, the results are very different based on the analyses for the REPC vis-à-vis the REPGDP variable. For instance, a solid divergent Review of Economic Analysis 16 (2024) 343-370 364 www.RofEA.org trend exists in the long-run distribution of REPC. With a vast majority (most) of the poorest (richest) countries congregating at extremely low (above the global average) REPC levels. From the perspective of environmental policies aiming at reduction/convergence in global CO2 emission, the results based on the REPC variable suggest that the onus should be majorly on the industrialised (high-income) countries. However, using the REPGDP variable, the convergence process is the least significant among upper-middle-income countries, with many countries clustering around REPGDP values far above the mean (2.4). Therefore, based on the REPGDP measure, upper-middle-income countries should bear a more significant share of the future carbon emissions commitments. Third, using the MPP tool, we identify countries with specific REPC and REPGDP levels that merit a place in the environmental policy priority list. Specifically, low-income countries with REPC values of around 2.3 have around 100 per cent probability of moving further up in the distribution in future years. By the same token, high-income countries with REPGDP values around 3.8 have the most significant (65 per cent) probabilities of diverging further from and above the global average. Therefore, in line with the global warming environmental policies aiming at CO2 emissions reduction/convergence, the above countries should be at the top of the priority list. This study offers several policy implications. First, the results deliver nascent evidence supporting the usefulness of the MPP display tool. For instance, an MPP-based “policy priority list” can inform policy prioritisation efforts by highlighting areas where emissions abatement is most urgently required. Policymakers, in turn, could employ this information to allocate limited resources and prioritise interventions targeting specific countries exhibiting unfavourable emissions trends, thereby maximising the effectiveness of climate mitigation efforts globally. Therefore, we advocate the periodic/annual implementation of the MPP framework as an integral part of the global reference system, helping countries update, optimise and manage their climate policies and carbon regulations. Second, we analysed the transitional dynamics and long-run trends in two variables: (1) REPC, which focuses on carbon emissions per capita, and (2) REPGDP, which accounts for emissions relative to economic output. Thus, the documented divergence in the forecast trends between REPGDP and REPC has several implications for multilateral climate negotiations and global environmental policies. (1) countries with high per capita emissions (as reflected in REPC) may argue for policies prioritising emission reductions based on individual consumption patterns. Conversely, countries with high carbon intensity relative to GDP (as indicated by REPGDP) may advocate for policies that focus on reducing emissions associated with economic activities. Balancing these competing interests will be crucial for achieving equitable and effective climate agreements. (2) the results suggest the need for sector-specific approaches to emission reduction. For instance, economies with high carbon intensity may need to decarbonise specific sectors (e.g., energy production, industry, or transportation). Tailoring WEI, WOJEWÓDZKI, CHEONG, SHI Carbon Emissions and Intensity 365 www.RofEA.org mitigation efforts to address sector-specific challenges can enhance the effectiveness of climate policies and facilitate smoother negotiations. Meanwhile, countries with high REPC may require policies targeting lifestyle changes, consumption patterns, and urban development. Third, estimating the distribution of the DDA and the MPP methods among all the countries can help commensurate intergovernmental cooperation plans by prioritising carbon tax policies across the countries. For instance, findings based on the preversus post-crisis dynamics in global CO2 emissions align with predictions that assuming a "business as usual scenario" will increase the number of clubs in cross-country carbon intensity (REPGDP) and emissions per capita (REPC) alike. We advocate that policymakers consider these projections when designing long-term environmental policies and strategies. This could include implementing proactive measures to mitigate carbon intensity growth, such as setting ambitious emission reduction targets, promoting renewable energy adoption, or implementing carbon pricing mechanisms. Furthermore, the research outcomes depict the future development trend of national carbon emissions and intensity, thereby guiding the government to allocate capital and technical resources more efficiently to promote energy transition. Climate change cooperation organisations can encourage and prioritise their investment in these countries and regions and improve knowledge diffusion, particularly in regions with outdated and imbalanced carbon emissions and intensity. The study employs the visual tools of the DDA method, which is inherently limited due to restrictive assumptions of no changes in the transitional dynamics of the study variable. This, in turn, may influence the accuracy of transition probabilities and long-term evolution projections. Moreover, due to data limitations, our sample spans between 2000 and 2016. However, the unprecedented changes in the global social and political economy, such as the COVID-19 worldwide pandemic and the outbreak of the military conflict in Ukraine, substantially impact the energy market and the trajectory of global emissions. Given the above backdrop, exploring the potential structural changes and convergence patterns in carbon emissions and intensities for the most recent period is imperative. Moreover, by focusing on carbon emissions among countries with different income levels, we possibly overlook other factors influencing emissions disparities, such as socioeconomic, political, and environmental variables. The analysis of pre-crisis and post-crisis periods may not capture all relevant economic or ecological shocks impacting emissions trajectories. Consequently, future research could expand the scope of study to incorporate a broader range of factors influencing carbon emissions convergence-divergence patterns, such as technological advancements, policy interventions, and socio-economic indicators. Review of Economic Analysis 16 (2024) 343-370 366 www.RofEA.org References Acar S, Söderholm P and Brännlund R (2018) Convergence of per capita carbon dioxide emissions: implications and meta-analysis. Climate Policy, 18, 512–525. Ahmed M, Khan AM, Bibi S and Zakaria M (2017) Convergence of per capita CO2 emissions globally: Insights via wavelet analysis. Renewable and Sustainable Energy Reviews, 75, 8697. Aldy JE (2006) Per capita carbon dioxide emissions: convergence or divergence? Environmental Resources Economics, 33, 533–555. Atasoy BS (2017) Testing the environmental Kuznets curve hypothesis across the U.S.: Evidence from panel mean group estimators. Renewable and Sustainable Energy Reviews, 77, 731–47. Bhattacharya M, Inekwe JN and Sadorsky P (2020) Consumption-based and territory-based carbon emissions intensity: Determinants and forecasting using club convergence across countries. Energy Economics, 86, 104632. Camarero M, Picazo-Tadeo AJ and Tamarit C (2013) Are the determinants of CO2 emissions converging among OECD countries? Economics Letters, 118, 159–162. Chang C and Lee C (2008) Are per capita carbon dioxide emissions converging among industrialized countries? New time series evidence with structural breaks. Environment and Development Economics, 13, 497-515. Cheng L, Shen J and Wojewodzki M (2023) A cross-country analysis of corporate carbon performance: an international investment perspective. Research in International Business and Finance, 64, 101888. https://doi.org/10.1016/j.ribaf.2023.101888 Cheong TS and Wu Y (2018) Convergence and transitional dynamics of China’s industrial output: a county-level study using a new framework of distribution dynamics analysis. China Economic Review, 48, 125-138. Cheong TS, Li VJ and Shi X (2019) Regional disparity and convergence of electricity consumption in China: A distribution dynamics approach, China Economic Review, 58, 101154. Climate Action Tracker (2021) Warming projections global update - November 2021. Available at https://climateactiontracker.org/documents/997/CAT_2021-1109_Briefing_Global-Update_Glasgow2030CredibilityGap.pdf Criado CO and Grether JM (2011) Convergence in per capita CO2 emissions: A robust distributional approach. Resource and Energy Economics, 33, 637-665. Delgado MS (2013) A smooth coefficient model of carbon emissions. Empirical Economics, 45, 1049–1071. Dong K, Hochman G and Timilsina GR (2020) Do drivers of CO2 emission growth alter overtime and by the stage of economic development? Energy Policy, 140, 111420.