The great stagnation and environmental sustainability: A multidimensional perspective

Since the 2008/09 Great Financial Crisis, we have witnessed a prolonged period of persistent global economic slowdown termed the “ Great Stagnation ” . This study examines how this “ new normal ” is associated with critical environmental dynamics (i.e., biodiversity, water, forest, agriculture, emissions) in areas and groups with different socio-environmental characteristics (i.e., income groups, continents, forest cover, biome, environmental performance index). Mixed results are shown. For instance, we find a deterioration in terrestrial and marine biodiversity, especially in middle- and high-income countries in Africa and Europe. This includes a reduction in the global fish stock, driven by countries in Africa. In contrast, the Great Stagnation is associated with reductions in PM 2.5 (lower- and upper mid-income countries), CH 4 emissions (upper mid-income countries and Europe), forest loss (upper mid-income countries and Asia), and increases in species habitat index (across most groupings). Our evidence indicates that periods of economic slowdown, such as the great stagnation, on their own cannot ensure a transition to a sustainable socio-environmental system and may be associated with significant negative environmental effects. Managing our transition to sustainability will require concerted policy efforts across multiple environmental domains, not only on carbon emissions, and during periods of both strong and weak economic growth rates.

Yet, there is quantitative as well as qualitative evidence that slower economic growth, on its own, cannot ensure transition to a sustainable socio-environmental system. A recent study examined how episodes of economic slowdown (e.g., financial crises), over the last four decades, impact air quality (Pacca, Antonarakis, Schröder, & Antoniades, 2020). The authors found that although the reduction in growth rates has a positive impact on the environment, this impact is short-lived, rather heterogeneous across different groups of countries, and disappears or turns negative 1-2 years after the beginning of these episodes (regardless of their duration). Qualitative evidence from case studies support these findings. For instance, studying the effect of the 1997 and 2008 financial crises in East Asia, Elliott (2011) finds that any positive environmental consequences were short-lived, while negative impacts endured. The latter include pressures for "further deforestation, agricultural expansion at the expense of water and soil quality, and lax enforcement of pollution regulations" (ibid. 179). Moreover, the priority for both government and the private sector in the post-crisis environment was investments that would generate "quick returns to compensate for losses rather than pursuing longer-term environmental and financial sustainability" (ibid. 180) (for a recent literature review see Pacca et al., 2020).
The above evidence is derived from episodes related to shock events, for instance abrupt economic slowdowns due to financial crises. Therefore, these findings are characterized by a "shock-bias", i.e. they are derived from conditions that signify temporary, short-term diversions from a "normal" economic trend. This "shock-bias" is overcome by a different set of studies that are based on statistical modelling and computer-based simulation techniques. These studies attempt to assess how a prolonged period of growth slowdown or no growth will impact on environmental sustainability (e.g., Barrett, 2018;Hardt & O'Neill, 2017;Jackson & Victor, 2015). Yet, although these studies overcome the shock-bias, they bear the weaknesses associated with the attempt to model and project social reality.
Thus, existing empirical findings on the interplay between economic slowdown and environmental sustainability are coming either from past shock events or from future-oriented simulation studies. In this paper, we attempt to advance the state of the art in the existing literature by adopting a rather different approach. The period after the 2008/09 Great Financial Crisis (GFC) has been characterized by a slow and fragile economic recovery. Despite very low interest rates and unprecedented liquidity support by Central Banks, growth rates have remained below their historical trend. This below-trend growth dynamics grew beyond the shadow of the GFC, acquiring characteristics of a "new normal" (El-Erian, 2009), defined by King (2019) as the period of "great stagnation." This period thus provides a solid ground to examine the potential impact of a systemic economic slowdown on environmental sustainability at a global level.
In this context, this paper sets out to examine how this period of Great Stagnation is associated with key aspects of environmental sustainability at a global level. To do so, we examine 15 environmental indicators in 217 countries. We simultaneously consider six environmental categories: biodiversity, forest, water, agriculture, and atmospheric emissions. To explore potential determinants of the relationship, we adopt a five-dimensional clustering of countries that accounts for income-level, geographical position, environmental performance, forest cover, and dominant biome (see Table 1). To the best of our knowledge, this is the first paper that adopts such a comprehensive approach and reports non-simulated findings on how a prolonged period of economic slowdown has impacted on different aspects of environmental sustainability. For this reason, our approach is more exploratory rather than confirmatory. Our aim is to examine the relationship between our variables, allowing for a range of potential interlinking mechanisms that impact on environmental dynamics in periods of slower growth. Our results contribute to the post-growth literature (e.g., Jackson & Senker, 2011;Victor, 2012), by offering new evidence and insights that go beyond existing findings coming from "shock events" and simulation-based projections.
The paper is structured as follows. Section 2 reviews the literature on the link between the great stagnation and the natural environment. Section 3 describes our quantitative approach, the sources of the data, and the 15 environmental indicators considered. Section 4 presents the results and robustness checks. Sections 5 and 6 comprise a discussion of the main results, the empirical limitations, and the concluding remarks. The global economic shock from the Great Financial Crisis proved more consequential than a temporary and reversible V-shaped disruption (i.e., featuring a sharp downturn and a rapid recovery). Advanced economies have been locked into a long-term low-growth trajectory, referred to as a "new normal" (El-Erian, 2009), "new mediocre" (Lagarde, 2014), "secular stagnation" (Summers, 2015) or the "great stagnation" (King, 2019). This trend does not only apply to advanced economies. As Figure 1 demonstrates a significant rupture in GDP trend is observable at global, high-income and middle-income countries levels. According to King (2019), "[t]he world economy is stuck in a low growth trap."

T A B L E 1 Number of countries per subgroup
Prior to the GFC advanced economies grew by about 2.27% per year, whereas in the period since, it has averaged just 1.39% (IMF, 2020). The US, for example, experienced a growth of only around 1% yearly, significantly lower than the pre-crisis period of 3% GDP per capita between 1950 and 2000 (Haldane, 2015). Similar low growth rates occurred in several other advanced economies, which saw their average growth falling from 3.5% in the 1990s to 1.86% during (IMF, 2020. Also, after the GFC, China, the second largest economy in the world, entered into a "new normal" of significant lower growth rates (see Appendix A in Table A1). The respective rupture in the global GDP trend is captured in Figure 2.
There is a growing consensus that economic growth at the current rate of depletion and degradation of environmental assets cannot continue indefinitely. Several scholars suggest that the economy needs to slow down (e.g., Jackson, 2016;Kallis, 2018 (Barrett, 2018;Hardt & O'Neill, 2017).
The uncertainty around the environmental outcomes of the great stagnation comes partly from the fact that we lack solid, robust evidence on the impact of economic slowdown on the environment.
Existing models often overlook the multiple aspects of ecosystem or are based on a reduced representation of the socio-economic system (Hardt & O'Neill, 2017;Spash & Schandl, 2009 understanding of the interplay between an economic slowdown and the natural environment, this study proposes an approach, which avoids the use of macroeconomic model assumptions by using a real example of stagnation period as our empirical test. The literature surrounding the relationship between environmental quality and economic growth is extensive. The popular environmental Kuznets Curve, which argued that economic growth has an inverted U-shape relationship with environmental quality (Kuznets, 1955), no longer provides a relevant framework for this relationship (Stern, 2004) as it has been challenged by several empirical analyses (see Pacca et al., 2020). Although some data may suggest that wealthy countries seem to decrease their environmental impact over time, several potential externalities may hinder their behavior.
For example, wealthy countries may reduce their domestic portion of materials extraction through international trade, while the overall mass of material consumption significantly increases (Wiedmann et al., 2015). Similarly, data on improving management of public waste disposal in wealthy countries do not take into account international waste trade (Cotta, 2020;Kellenberg, 2015).
On the other hand, there is a growing literature surrounding the effect of economic slowdown on environmental sustainability. Existing evidence suggests that environmental degradation tends to decrease straight after economic shocks occurs, but negative impacts endure (Elliott, 2011;Lekakis & Kousis, 2013;Pacca et al., 2020;Siddiqi, 2000).
Endurance may be due to reinforced industrial activity and/or a shift toward weaker environmental protection and conservation policies. The latter may lead to a lax enforcement of pollution regulations, further deforestation, and agricultural expansion at the expense of water and soil quality. In contrast, Monteiro, Russo, Gama, Lopes, and Borrego (2018) suggest that recessions may lead to long-term changes in consumer behavior with a potentially positive impact on the environment (i.e., shift to lower energy consumption). Overall, the literature finds mixed evidence on the long-term effect of economic shocks on environmental quality. Our analysis adds to the existing studies by going beyond a shock event and examining multiple aspects of environmental quality (i.e., biodiversity, forest, emissions, water, and agriculture) and potentially critical factors for sustainable transitions.

| Approach
This study uses the period of "great stagnation" as a test case to examine the potential impact of a slowing down in economic growth on the environment. We use aggregate data at country level on environmental indicators for the level of biodiversity, agricultural and forestry activity, water resources, and atmospheric emissions. We use a model that comprises a dynamic panel data model using a GMM specification. Our variable of interest comprises a dummy as a proxy for the period of slow growth, the great stagnation. This variable is coded as zero from the beginning of our dataset up to 2009 and one from financial crisis (see Figure 2). This evident fall in global GDP in constant terms is translated in a significant global output loss, and it is followed by a constant period of slow growth.
The environmental indicator of interest of country i in time t, as dependent variable can be denoted as Y it (i = 1,…, n; t = 1,…, T), thus the model can be written as: where controls is a vector of control variables, Y it − 1 is the lagged dependent variable to attenuate for potential omitted variable bias which might arise, as well as capturing the dynamic and temporal dependence of the independent variable. Finally, ε is the error term. We control for several indicators based on existing studies (Antoniades, Widiarto, & Antonarakis, 2019;Pacca et al., 2020). Appendix B, Table B1 provides a list with the controlling variables adopted for each regression.
We use the Arellano-Bond specification (Arellano & Bond, 1991) which includes second and deeper lags as instruments for the first lag of the dependent variable. The GMM models allow the correction of the potential bias resulting from the endogenous relationship between economic crisis and our environmental variables. Specifically, we adopt the two-step system GMM, or Arellano-Bover/Blundell-Bond estimator, which augments Arellano-Bond by adding the level equation in addition to the difference one and drastically improves its efficiency (Roodman, 2009). Furthermore, we compare our main specification (two lags) to a higher number of lags to make sure that results are robust across lag choices.
For the purpose of the analysis, the sample is divided into five groupings to enable a better understanding of the potential drivers in the relationship between great stagnation and environmental sustainability (Table 1) Table C1). Fifth, we divide countries into four subgroups based on their average score in the Environmental Performance Index over 16 years (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). The first two groupings account for heterogeneity across income levels and geographical locations. The remaining three groupings account for country-specific environmental characteristics as an intervening viable on the impact of great stagnation. In particular, we are interested to examine whether "forest cover," biomes and environmental performance make a difference in the degree of socio-environmental resilience in the context of a slowing down economy.   Tables 3 and 4 show the great stagnation's effect on the 15 environmental performance indicators. Specifically, Table 3 shows the regression results for the income-and continent groups, whereas Table 4 shows the results for the forest, biome, and environmental performance groups. The regressions comprise control variables as well as the lagged dependent variable.

| Biodiversity
At global level, the great stagnation is negatively associated with five out of six biodiversity variables examined. Terrestrial Protected Areas decreased by 1.6% (national weights) and 3.7% (global weights), Lastly, the economic stagnation is associated with a negative impact on marine protected areas for high-income (−33.1%) countries, mainly in temperate biomes (−30.4%) and in Europe (−37.2%). The effect is negative for upper mid-(6.7%) and high-EPI (33%) countries, whilst this is positive for low-EPI (10.9%) countries.

| Agriculture
The great stagnation is linked to an increase in agricultural land at a global level by 0.8%, according to the World Development Indicators

| Forest
The great stagnation is associated with a reduction in forest loss for upper mid-income countries and in the Asian continent, with decreases of 6.7% and 13%, respectively. There is a significant impact of stagnation on forest loss at the global level although the regressions do not pass either of the two statistical assumptions (Hansen, AR2 tests) thus these results are not considered. Regarding treecovered areas from the CCI, the effect of stagnation is small, negative, and significant for low-(−0.3%) and high-income (−0.2%) countries, and for Europe (0.1%). Tree covered areas are also negative and statistically significant for mid-forest area (−0.2%), and low-(−0.2%), upper mid-(−0.1%), and high-EPI (−0.2%) countries.

| Water
The great stagnation is associated with a reduction in unsafe drinking water for high-income countries (−1.2%) and for Europe (−1.6%); however, we observe an increase in unsafe water for Americas (−2%) and for mid-forest area countries (−2.3% Note: All indicators have been transformed into logarithmic form (see Table D1 in Appendix D).

T A B L E 3 Regression results on the effect of global stagnation on distinct environmental indicators for the income and continents subgroups
Income group
T A B L E 4 Regression results on the effect of global stagnation on distinct environmental indicators for the forest, biome, and environmental performance index subgroups (3) (8) *p < 10% is the significance level. **p < 5% is the significance level. ***p < 1% is the significance level.

| Emissions
Looking at atmospheric emissions, the great stagnation is linked to an increase in CO 2 emissions in low-income countries (6.9%), high-forest area (2%), and a dominant tropical biome (3%), but a decrease in high-

| Robustness of the results
A series of robustness checks have been carried out to assess the consistency of the estimates across different specifications (see Table F1, Appendix F). We used an alternative GMM specification, with the lagged dependent variable and stagnation treated as endogenous, as some unobserved factors with a potential effect on the stagnation period could contribute to determining the slow growth period as well as simultaneously change environmental policies (Pacca et al., 2020). All other controlling variables have been treated as strictly exogenous. The results from the modified GMM from We provide a visual representation of our results in Figure 3.

| Macro-level findings
The period of great stagnation offers a fertile ground to test the implications of a sustained slowdown in economic activity on the environment. We expect that a reduction in economic growth rates would in principle have a positive impact on the environment (Krausmann et al., 2009;UNEP, 2011), Yet, we also expect that a reduction in growth rates will have an adverse impact on important socioeconomic indicators, such as on livelihoods, employment and investments, which may negatively affect the impact of the economic slowdown on the environment. Understanding the complex relationship among these dynamics is critical to manage the needed transition to a socio-environmentally sustainable model. Overall, despite the historically exceptional and persistent slowdown of growth rates that we have seen over the last decade, this period is not associated with exclusively positive environmental dynamics. The relationship between great stagnation and the environment is mixed. Our results indicate a deterioration in biodiversity, and some increases in CO 2 and N 2 O emissions, along with improvements in PM 2.5 and CH 4 emissions, forest loss, and species habitat index.
Thus, although economic growth has negative environmental consequences (OECD, 2008;Vadén et al., 2020;Ward et al., 2016;Wiedmann et al., 2015), the environmental consequences of slower economic growth appear to be mixed, diverse across different regions, and sensitive to policy responses (Bowen & Stern, 2010). Therefore, degrowth as a strategy to transitioning to sustainability cannot be thought of as a quantitative target or threshold. Rather it should be thought in qualitative terms as a strategy, a set of policies that aim to rebase our economic model on a more sustainable footing. report refers to as "tapering off" effect during this period (available data up to 2017) (see also Lewis et al., 2019). Yet, our TPA results in national weights demonstrate that the negative impact is driven by developments in the American continent (reduction of 2.4%). This is a more worrying finding as it may point to a geographically concentrated scaling back of protected areas, a process called Protected Area Downgrading, Downsizing and Degazettement (or PADDD) (ibid:7), with primary causes being industrial level activities, energy projects and local land pressures (Mascia et al., 2014). These results come to support recent findings that demonstrate for instance that PADDD has impacted Brazil's protected area network (Pack et al., 2016). North America is also affected. According to Lewis et al. (2019:577), the USA has the highest negative footprint in this area between 2006 and 2016, although the reason for this is not mostly PADDD but changes in the IUCN definition of a protected area. Another potential driver here, especially in central and southern America, Africa, and Asia is the increase in cattle farming and oil seed production, which has been a major factor in biodiversity loss (Marques et al., 2019). In the case of high-income countries, biodiversity loss may have also been driven by a shift toward austerity policies that are associated with a worsening of environmental standards and protection (Botetzagias, Tsagkari, & Malesios, 2018;Lekakis & Kousis, 2013).  (e.g., BBC, 2018;Hilborn et al., 2020;McClanahan, 2019;Mcclanahan et al., 2019). Furthermore, marine protected areas do not necessarily reduce fishing pressures (Agardy, di Sciara, & Christie, 2011;Bates et al., 2019), especially if only a small percentage of oceans is protected (Dasgupta, 2018). Yet, the difference that the great stagnation makes to fisheries is notable, although data are available only until 2014, and therefore the duration of great stagnation here is only 5 years.
The Great Stagnation had differing impact on the Species Protection Index (SPI) and the Species Habitat Index (SHI), but it is notable that in both cases this impact was rather diffused (across most income groups and several continents). The reduction in SPI at global level (−2.6%) is driven by reductions in the African (−1.3%) and European (−1.5%) continents. On the other hand, there is moderate positive impact on SHI at a global level (0.055%) that spreads across most income groups (except low-income) and continents (except Oceania).
These results suggest that although both terrestrial protected areas and species protection are decreasing, the existing habitat ecosystem could be improving, or it is likely that the decrease in the trend of habitat destruction could be levelling off for a number of species and ecosystems. This levelling off since 2010 has been seen in the similar Living Planet Index (Grooten & Almond, 2018). Another reason could be due to the natural growth of vegetation seen from remote sensing (involved in calculating the SHI) as well as effective conservation of the protected areas (note data for SPI and SHI are available only until 2014).

| Agriculture
To examine changes in agricultural land during the great stagnation,

| Forests
The relationship between great stagnation and forest cover is also mixed depending on the dataset used. During the stagnation period, forest loss (obtained from the Global Forest Watch) did not change at the global level but was significantly lowered in upper mid-income countries (−6.7%), and in Asia (−13%). Globally, deforestation was coming up against international efforts to decrease, halt, or restore and PROBA-V) meaning many pixels will be a mosaic of forest and non-forest types. Also, it would be expected that a decrease in forest cover would be met with an increase in agricultural land from the CCI, but this is only matched in high-income countries (see Table 3).

| Water
Dynamics regarding unsafe drinking water (UDW) during the great stagnation differ across income and continent levels. High-income countries, driven by the European continent, experience a decrease in UDW (−1.16% and −1.57% respectively). Yet in the Americas, we observe an increase in UDW by 1.98% (we also observe an increase of 3.37% in Asia, but the regression does not pass the AR2 test, thus we cannot guarantee a robust result). The increase in the Americas may be driven by developments both in central and south America, due to economic hardship, but also in the United States, which experienced significant challenges in this area recently (Suh, 2019).

| Emissions
Finally, the results on GHG emissions are mixed, pointing that a period of degrowth is not necessarily associated with a reduction in greenhouse gas emissions, even if we have seen initiatives and agreements related to these emissions in the last 10-15 years including the 2009 Copenhagen summit, the 2015 Paris Agreement, the 2008 EU Ambient Air Quality Directive, and the WHO Global Platform on Air Quality and health since 2014. In terms of air quality, we observe a reduction in PM 2.5 emissions by 1.85% in lower-mid-income countries and 4.6% in upper-mid-income countries. For CO 2 emissions, we find no impact at global level, a decrease in high-income countries (−2.5%), driven by Europe (−1.8%) and temperate biome areas (−1.35%), but a significant increase in low-income countries (6.9%). CO 2 has been increasing also in countries with tropical biomes, and with high-forest area. A potential explanation of our emission results is the decrease in green government spending in East Asia after the Great Financial Crisis (i.e., spending for renewable energy use, waste reduction and recycling, emission control programmes, and green energy efficient technologies) (Elliott, 2011;Pacca et al., 2020). Furthermore, the rise in CO 2 , CH 4, and N 2 O emissions could be linked to Chinese investments in low-income countries in southeast Asia (Brown, 2016;Frost, 2004;Frost & Ho, 2005;Pheakdey, 2013;Yeh, 2016) as well as in central America (Sanborn & Chonn, 2015) and central Africa (Shen, 2013). We also observe increases in N 2 O emissions in low-income countries (2.2%). The main driver for the CO 2 and N 2 O increases in low-income countries should be the historically high growth rates experienced in these countries during this period (Steinbach, 2019) (Hurley, Storrie, & Peruffo, 2016). Lastly, in contrast to prior research (Santamouris et al., 2013;Saffari et al., 2013), our PM 2.5 findings do not point to a clear shift in energy consumption habits by people in developing countries, such as the use of alternative energy sources (i.e., wood burning). To this extent, we suggest future research to consider for example the role played by environmental related (green) technologies as well as the circular economy. Constraints in global data availability did not allow us to integrate such factors in our analysis. We further recommend using a narrower approach for later research that would potentially focus on a few specific environmental quality indicators, to improve the empirical analysis with more control variables and robust indicators (e.g., using longer time periods and/or different indicators).

| CONCLUSIONS
Periods of economic slowdown have been linked to an amelioration of environmental quality due to a slowdown in energy use, resource extraction (water, timber, and minerals), and greenhouse gas emissions. This further emphasizes the direct link of the economy to the environment (Bowen & Stern, 2010;Dietz & Adger, 2003;Mills & Waite, 2009;Stern, 2006), which has been demonstrated once again during the Coronavirus pandemic (Antonarakis, 2020). Recent evidence has shown that improvement in environmental quality during economic crises is short-lived, and the environment deteriorates again 1 or 2 years after the break-out of the crisis (Elliott, 2011;Pacca et al., 2020).
The decade after the Great Financial Crisis is one of the longest periods of persistent global economic slowdown after WWII. We used this setting to examine the impact of a systemic economic slowdown on environmental sustainability at the global level. We developed a novel research approach in which we examine the relationship between the great stagnation and 15 environmental indicators comprising of five distinct environmental categories (i.e., biodiversity, water, forest, agriculture, and emissions) across five groupings (i.e., income groups, continents, forest cover, biome, and environmental performance index). In this way, we account for a large number of potential drivers of environmental change in periods of slow growth.
We find that the period of great stagnation is associated with mixed environmental dynamics, both in terms of direction and location. For instance, we observe that the great stagnation is associated with a deterioration in biodiversity in terrestrial as well as marine ecosystems at the global level, and with improvements in some air pollutant emissions (PM 2.5 , CO 2 , CH 4 ) in middle-and high-income countries. These mixed effects are differently distributed across continents. For instance, the deterioration in "terrestrial protected areas" is driven by the Americas, in "marine protected areas" by Europe, of the "species protection index" by Europe and Africa, and of "global fish stocks" by Africa. Respectively, improvements in CO 2 and CH 4 emissions are concentrated in Europe, while, at the same time, we observe increases in CH 4 emissions in Asia.
The level of income seems to be a key determinant for environmental outcomes during the period of great stagnation, followed by the continental grouping that most times come to add specificity in relation to income groups. Groupings related to forest area, biome, and environmental performance index do not have a major impact on explaining environmental effects of the stagnation period, save some evidence of worsening environment quality in medium forest cover and temperate countries.
The Sustainable Development Goals (SDGs) have strongly advocated for the synergy between economic growth, poverty alleviation, and environmental improvement. The importance of this synergy has been widely advocated during the Coronavirus pandemic too (Dasgupta & Andersen, 2020;Florizone, 2020;Nature Editorial, 2020). Yet, for GDP growth to be sustainable it would have to be decoupled from energy and material use and thus environmental impacts, and there is limited evidence that this is possible for all countries across many environmental domains (Vadén et al., 2020;Ward et al., 2016;Wiedmann et al., 2015). APPENDIX D.
APPENDIX E.

Biodiversity indicators
The environmental indicators adopted are listed in Table 2 country's total catch that come from taxa that are classified as either over-exploited or collapsed. This value is then averaged for all species occurring in a country, with all species weighted equally. Marine protected areas measure the percent of a country's Economic Exclusion Zone set aside as a marine protected area (Wendling et al., 2018).

Agriculture indicators
Agricultural land, in hectares, from the The World Bank (2020b) refers to the share of land area that is arable, under permanent crops, and under permanent pastures. Arable land includes land defined by the FAO as land under temporary crops (double-cropped areas are counted once), temporary meadows for mowing or for pasture, land under market or kitchen gardens, and land temporarily fallow. Land abandoned as a result of shifting cultivation is excluded.

Forest indicators
Tree cover loss from the CCI, is defined as "stand replacement disturbance," or the complete removal of tree cover canopy at the Landsat pixel scale, presented in hectares. In the Global Forest Watch (2020) database, tree cover is defined, in hectares, as all vegetation greater than 5 m in height and may take the form of natural forests or plantations across a range of canopy densities.

Water indicators
Unsafe drinking water measures the actual outcomes from lack of access or use of improved sources of drinking water. It measures unsafe drinking water using the number of age-standardized disability-adjusted life-years lost per 100,000 persons (DALY rate) due Note: Unsafe drinking water and PM 2.5 air pollution have been interpolated up to four consecutive years due to missing observations. All variables have been transformed into logarithmic form.
to exposure to unsafe drinking water. Data for this indicator come from the Institute for Health Metrics & Evaluation's (IHME) Global Burden of Disease (GBD) study (EPI-Yale, 2020). Wastewater treatment comes from the United Nations (2020) and measures the percentage of population connected to a wastewater treatment plant through a public sewage network. This indicator does not take into account independent private facilities, used where public systems are not economic.

Atmospheric emissions indicators
The dataset comprises three greenhouse gas emissions: carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O). CO 2 emissions comprise emissions mostly from sources such as the consumption and production of fossil fuels, including coal, peat, petroleum, and natural gas and the production of cement. CH 4 emissions are a major part of the global greenhouse gas emissions, encompassing emissions from agriculture, produced mostly by humans and livestock animals as well as natural sources such as wetlands. Also, to a minor extent, methane is produced from rice production and waste and from industrial activity. N 2 O emissions are produced mostly from the agricultural sector, especially the use of manure and nitrogen fertilizers (Davidson, 2009). Terrestrial protected areas (%EEZ-National Weights)