Interactions between Democracy and Environmental Quality: Toward a More Nuanced Understanding

Abstract

This paper seeks to contribute to existing debates on the relationship between democracy and environmental quality. More specifically, we aim to provide nuance and insight into the question as to whether democratic regimes are better equipped to protect the environment. After critically reviewing theoretical arguments and providing an overview of existing empirical studies, the paper proposes an approach which consists of the use of non-parametric correlations between democracy and environmental quality, and a consideration of the interactions between democracy, government effectiveness, economic prosperity, and perceptions of corruption. Crucially, we show that, although a positive correlation can be found between levels of democracy and environmental quality, the picture is somewhat blurred if data are stratified using criteria such as government effectiveness and corruption perceptions. Consequently, the main argument the paper pursues is that, to assess the relationship between democracy and environmental quality, intervening factors and their effects need to be acknowledged and taken into account.

1. Democracy and the Environment

The association between environmental quality and democracy is a hotly debated issue in the scholarly literature [1,2,3]. Literature tends to assume a positive relationship between democracy and environmental quality [4]. Sometimes, an extra layer of sophistication is added to the previous claim, where the democratic capital stock is argued to be more important than the current level of democracy [5]. For example, Battig and Bernauer [6] present democracies as political states/institutions which have a slow capacity to move from political and legal commitments (policy output) to emission reductions (policy outcomes). However, empirical findings are often conflicting and, therefore, it is rather difficult to conclude that there is an absolute and robust link between environmental quality and democracy [7,8,9,10]. The picture becomes more complex if the concept of democracy itself is opened up. Whilst it is not the aim of this paper to thoroughly debate the modern concept of democracy, it is worth recounting that the condition and the effectiveness of democracy as a form of governance and of the capacity and willingness of citizens to engage with environmental issues are evolutionary and need to be seen as such. A few points are relevant in this context. Though liberal democracy is on the ascendance since the end of the Cold War as an aspirational form of governance, democracy in the 21st century proved to be both dynamic and fragile and not just in transitioning states. Firstly, as Carter and Stokes [11] argue, democratic states are increasingly viewed by their citizens as unable to deliver on basic public goods and services, and, with prevailing economic globalization, there are perceived limits to the traditional role played by state intervention in socio-economic issues. Secondly, liberal democracy is not seen to take account of often profound social and economic disparities, posing restrictions on individual and community opportunities to participate in politics, which in turn generates apathy, disengagement, and ultimately the breakdown of the kind of political community implied by Western liberal democracy [11]. There is also the compelling claim that the ascendance of neoliberalism and rise of international corporations which are able to operate within weak regulatory frameworks effectively killed off the effectiveness of collective pro-environment actions, giving rise instead to weak and ineffective individualized approaches [12].

1.1. Is There a Positive Relationship between Democracy and the Quality of State’s Environmental Policies?

The case for there being a positive relationship between environmental quality and democracy received considerable theoretical support. A positive link is often described as “green democracy” or “green state” [13]. Following this line of thinking, the arguments by Payne [14] and Midlarsky [15] in support of there being a positive link between environmental quality and democracy point to the essential characteristics of democracy, namely (i) the existence and protection of individual rights, (ii) citizenship, (iii) the existence of free scientific communities and environmental lobbying groups, (iv) respect for the rule of law, and (v) existence of green consumerism. In essence, in democracies, citizens are at liberty to exercise their rights to gather and disseminate environmental information, advocate environmental values, and exercise pressure in favor of environmentally friendly projects and policies [16,17]. In other words, in democracies, the possibility of green lobbying is taken for granted [18], and this may influence and shape legislation. However, it should be stressed that, whilst participation rights are generally conceived as necessary preconditions for effective environmental governance, they are not, on their own, sufficient for such a goal [19].

Democracies are responsive toward their citizenry because elections are the legitimizing basis of the power of decision-makers. Therefore, the necessity to be elected introduces a minimum level of accountability to society’s preferences, i.e., environmental quality preferences are important especially to (free and wealthy) societies that pay increasing attention to environmental and climate change issues [6,20,21]. The latter is a postulate compatible with the Environmental Kuznets Curve hypothesis [22]. Also, under democratic regimes, scientific communities, as well as environmental lobby groups, may have influential voices and can try to influence the policy-making process and national debates [8,14]. Payne [14] also shows the democratic countries’ ability to satisfy people’s preference for the environment and acknowledges various aspects of the markets when he assesses the “green” dimensions of democracies. These include green consumerism [23], the use of economic incentives and disincentives for environmental management [24], and corporate responsibility [25].

Another claim that links democracy and environmental performance is that democratic countries are more likely to participate in international environmental agreements and treaties. Weiss and Jacobson [26], as well as Bättig and Bernauer [6], claim that democracies tend to act in accordance with international laws and also to act multilaterally, rather than unilaterally. Participation in international treaties, inter alia, might also facilitate citizens’ participation in the public sphere as it offers them opportunities to elevate their domestic (environmental) priorities to the international level of impact by working with like-minded transnational movements and groups. Furthermore, developing countries may gain financial assistance by participating in international environmental agreements [27].

In contrast to the arguments mentioned above, Ball [28] argues that there is no fundamental association between environmental protection and democracy. Someone can perfectly embrace a pro-environmental political stance, while at the same time reject democracy on philosophical or strategic grounds. Such profound ambivalence from some environmentalists to democracy is an example of the means/ends debate. There are two branches of green political theory which are hostile to the idea of democracy, namely eco-authoritarianism and eco-radicalism [29]. The proponents of eco-authoritarianism argue that a democratic regime is not equipped for addressing major ecological imperatives, such as resource scarcity and over-population. The early version of eco-authoritarianism was often referred to as “survivalism” [30], which belongs to the intellectual tradition known as “neo-Malthusianism” [31]. Such a tradition frequently blames population growth for the pressure to consume scarce (natural) resources and, accordingly, overpopulation is conceived of as the main reason behind environmental quality degradation [32,33,34,35]. Although the (neo)-Malthusian pessimism dominates policy discourse, it is not immune to criticisms [36]. For example, the proponents of political ecology severely criticized (neo)-Malthusianism as an oversimplifying and misleading narrative [37]. Furthermore, while the fundamental notion of ecological limits to growth remains convincing, advocates of “de-growth” start redeeming population growth and vigorously stress the importance of less growth and down-scaling when discussing ecological limits and environmental quality [38,39].

Eco-radicals, on the other hand, preoccupied with expanding the community of moral rights holders to animals, trees, and plants, do not pay attention to democratic procedures and are primarily concerned with outcomes [40]. According to Vanderheiden [41] what dichotomizes radical environmental groups from mainstream ones are the “ecotages”, which are the extra-legal tactics of vigilante violence against those who are responsible for serious ecological damages. Therefore, while a democratic regime may foster the proliferation of green ideas, it does not necessarily mean that an eco-authoritarian or an eco-radical stance manifests democratic principles.

There are also studies that find only limited conditional effects of democracy on environmental policy. Povitkina [20] explains the connection between democracy and CO₂ emissions. She claims that, if corruption is low, more democracy is associated with lower CO₂; if it is high, democracies appear to be doing no better than authoritarian regimes. The conditional effects are taken from different perspective by Barrett and Graddy [42], who state that an increase in political and civic freedoms reduce a number of pollution variables, whilst Torras and Boyce [43] claim that political rights and civil liberties have particularly strong impacts on the quality of the environment in low-income countries.

Bearing in mind the discussion presented above, the aim of the rest of this paper is to present findings and analysis that shed light on the relationship between democracy and the quality of environmental policies. Firstly, we summarize the main empirical studies which examine the environment–democracy debate.

The paper is based on relevant data from the Environmental Performance Index (EPI) to capture environmental quality and the Democracy Index (DI) for the democracy proxy. Drawing from Fredrikson and Wollscheid [24], this paper explicitly examines “interactions”, which, as noted at the start of this paper, refer to interplay between factors that produce an effect on environmental quality. Such effects may have decisive role in explaining the nature and the strength of the association between environmental quality and democracy.

2. The State of Play: Democracy and the Quality of Environmental Policy

While most previous studies indicate a positive, or mostly positive, link between democracy and environmental quality, there are quite a few studies which reveal that such a link is neither clear-cut nor always consistent. For example, Neumayer [44], Esty and Porter [45], and Winslow [3] prove mostly positive impacts, while Li and Reuveny [8] show how democracy minimizes environmental degradation in terms of carbon dioxide emissions, nitrogen oxide, land degradation, deforestation, and organic pollution in water. There are many analyses which present various variables with inconclusive results [46,47,48]; the relevant literature is summarized in

Table 1. Summary of the empirical studies exploring the link between democracy and environmental quality.

Authors	Environmental Quality (EQ) Indicators	UN SDGs	Democracy Index	Number of Countries	Methodology	Time Span of EQ Proxy	Verdict
Barrett and Graddy (2000) [42]	Many pollutants	(11) sustainable cities and communities	PR and CL	60–160	Panel data techniques	1977–1987	Mostly Positive
Neumayer (2002) [44]	Many indicators of environmental commitment	(15) life on land and (13) climate action	PR and CL, VAN, Polity project	153	OLS cross-section data	1998	Positive
York et al. (2003) [46]	Ecological footprint	(15) life on land	PR and CR	142	OLS cross-section data	1996	Inconclusive
Esty and Porter (2005) [45]	UPC, SO2, and energy use	(7) affordable and clean energy	PR and CL	40–70	Bilateral regression	2001	Positive
Winslow (2005) [3]	SO2, smoke and SPM	(11) sustainable cities and communities	Polity III Project index, and PR and CR	46	Panel data techniques	1971–1992	Positive
Li and Reuveny (2006) [8]	Many indicators	(6) clean water and sanitation and (7) affordable and clean energy	Polity IV Project index	143	Panel data techniques	1961–1997	Non-monotonic and differentiated
Pellegrini and Gerlagh (2006) [48]	ESI	(15) life on land, (13) climate action, and (6) clean water and sanitation	Polity IV Project index	54	OLS cross-section data	2005	Inconclusive
Buitenzorgy and Mol (2011) [1]	Deforestation	(15) life on land	Polity IV Project index	177	OLS cross-section data	1990–2000	Inverted U-shape
Mukherjee and Chakraborty (2013) [49]	EPI	(15) life on land, (13) climate action, and (6) clean water and sanitation	DI from Economist Intelligence Unit	146	OLS cross-section data	2008	Positive
You et al. (2015) [47]	CO2	(13) climate action	PR and CR (Freedom House)	98	Quantile regression	2013	Inconclusive
Policardo (2016) [2]	CO2 and PM10	(13) climate action and (11) sustainable cities and communities	Custom-made index of democratic transition	47	Interrupted time series cointegration	1950–2002	Positive
Halkos and Paizanos (2017) [50]	Many air pollutants	(13) climate action and (11) sustainable cities and communities	Polity IV Project index	94	Dynamic panel data techniques	1970–2008	Mostly positive

PR: political rights, CL: civil liberties, VAN: Vanhanen’s index, ESI: Environmental Sustainability Index, EPI: Environmental Performance Index, DI: Democracy Index, SO₂: sulfur dioxide, UPC: urban particulate concentration, SPM: suspended particulate matter, CO₂: carbon dioxide. PM₁₀: concentration of particulate matter 10 micrometers or less in diameter. The numbers in parentheses indicate the classification of Sustainable Development Goals (SDGs) provided by the United Nations (UN); see https://www.un.org/sustainabledevelopment/sustainable-development-goals/.

.

The ambiguity which emerges from Table 1 may stem either from the complexity of the issue or from the incompatibility between various approaches encountered in the literature. In our view, the incompatibility refers mainly to the lack of consensus concerning the choice of data and indicators and/or the diverse methodologies. In particular, there are only two studies which use composite environmental indicators such as the Environmental Performance Index (EPI) [49] or the Environmental Sustainability Index (ESI) [48], which was the forerunner of the EPI, to capture the concept of environmental quality.

Composite environmental indicators offer a means of aggregating multiple indicators to track and communicate complex (environmental) systems. By definition, a composite environmental indicator outperforms a single one, such as the one which is expressed by the concentration of specific pollutants. By contrast, however, the majority of the relevant literature attempts to capture environmental quality by using single or multiple pollutant indices [2,3,8,42]. At the same time, there are cases where environmental quality is captured through indicators of ecological footprint [46], deforestation [1], or even various proxies of countries’ environmental commitment [44]. Presumably, while all these environmental proxies used in the scholarly literature are adequate in capturing elements of environmental quality, it is rather hard to compare due to the differences in scales, units, and dimensions.

Furthermore, as Munck [51] reminds us, the conceptualization of democracy and its various postulates or attributes are difficult to precisely hone because of its dynamic and nebulous nature. Firstly, there is the Freedom House index, which is used by some researchers [45,47,51]. Freedom House determines the democratic status of a country on the basis of its ratings on two institutional dimensions, namely political rights and civil liberties [52]. Secondly, there is the Polity IV Project used by Li and Reuveny [8] and others [48,50]. The Polity Project is a multidimensional index which captures various aspects of democratic regimes such as executive recruitment, constraints on executive authority, and political competition and participation. Winslow [3] combines these two sources, while Neumayer [44] adds a third source of data, the Vanhanen Index (V index) This focuses on the electoral system and in particular on competition and degrees of participation. However, such definitions of democracy can miss the point as to how collective decisions are reached [53].

Very often, a literal reading of democracy associates its meaning with majoritarian decision rule (see References [54,55]). Majority rule and popular accountability are captured by the Democracy Index (DI) from the Economist Intelligence Unit (EIU) [56]. Mukherjee and Chakraborty [49] use the DI, while Policardo [2] considers a custom-made index to capture democratic transition.

As Table 1 shows, the applied methodologies primarily comprise ordinary least squares (OLS) and panel data techniques (fixed and random effects). Occasionally, less frequent techniques are also encountered in the relevant literature. For example, Esty and Porter [45] use bilateral regression, and You et al. [47] apply quantile regression, while Halkos and Paizanos [50] employ a partial adjustment model.

3. Materials and Methods: Capturing the Interaction between Democracy and Environmental Quality

This paper utilizes the Environmental Performance Index (EPI) as an indicator for environmental quality [57]. The EPI is a composite index which takes into account both human health and the protection of ecosystems. Within these two pillars, the EPI scores a country’s performance on the basis of nine issues comprising 20 specific variables. Arguably, EPI is the most comprehensive composite indicator used to reflect environmental quality [58,59,60]. The EPI is published every two years, beginning in 2006. In the last available data of 2016, the EPI contains an extra dimension which was missing from the previous reports, namely climate change. As a result, the 2016 dataset is not directly analogous to the previous sets of data; thus, the analysis of this paper was restricted from 2006 to 2014.

Likewise, the Democracy Index (DI) from the Economist Intelligence Unit was chosen as an indicator to classify the democratic regimes of selected countries. The reason for such a choice is that the DI is the only continuous measure of democracy with a wide coverage available for the time span of this paper. The DI is based on 60 indicators grouped into five different categories measuring pluralism, civil liberties, and political culture. The DI was first published in 2006 and covers almost the entire world [61,62,63,64].

To recapitulate at this stage, the aim of this paper is to make a contribution to the discussion as to whether democracy is good for the environment by looking at interactions between different democracy indicators. In regression models, typically, the interaction effects are taken into account by adding the cross-product of two explanatory variables, a technique known as moderated regression analysis [65]. However, modeling can inflate the problem of multicollinearity [66], which is likely to occur in similar models due to endogeneity issues [67]. Multicollinearity is usually considered as a problem since it very often results in unstable coefficients (i.e., having wrong signs or implausible magnitude) and, therefore, the effect of independent variables on dependent variables is difficult to assess [68].

To avoid these modeling problems, the paper abstains from using regressions analysis. Instead, the correlation coefficient between the EPI and the DI was estimated as a means of assessing the statistically significant relationship between these two variables. It should be stressed here that a significant correlation between two variables does not imply direct causality but, instead, only shows whether two variables are synchronized, i.e., vary in a similar way [69]. In order to avoid the usual criticism of coincidental results, the correlation coefficients were estimated for all available periods, i.e., from 2006 to 2014.

We used non-parametric correlation, otherwise known as Spearman rank order correlation, since it does not rely on the assumption of normally distributed data [70]. The non-normality of the data usually inflates the estimates derived by the Pearson correlation [71,72,73]. The choice of non-parametric correlation over the parametric equivalent was justified based on the normality tests applied to the data; we used the most efficient so-called Shapiro–Wilk and Anderson–Darling tests [74].

As Shpitser [75] argues, the common truism “correlation does not imply causation” is common precisely because co-dependence of potential cause and potential effect can be fully explained away by, for example, a third non-observed variable. A straightforward and standard technique to examine the role of such missing variables is to stratify the dataset in question [76]. By stratifying the entire sample into several sub-samples and examining the heterogeneity in the correlation coefficient estimates, it is possible to determine whether interaction effects exist or not. Specifically, if the (correlation coefficient) estimates systematically vary in a monotonic way within sub-samples, then there is evidence that the variable used for data stratification affects the link between environmental quality and democracy. Likewise, if there is no significant heterogeneity of these estimates within sub-samples, then it is unlikely that interaction effects exist.

Although the stratified samples provide a sound conceptual approach for exploring interaction effects, it poses an interesting interpretation problem. Since the sample size affects both the estimates and the critical values of correlation coefficients, then the direct comparison or interpretation of these coefficients (estimated from samples with different sizes) is not straightforward. Although various methods were proposed for comparing correlation coefficients [70,77], to our best knowledge, there is nothing for the interpretation problem. The latter means that inference intervals for describing the strength of the likely association between two correlated variables (e.g., the interval (0.5, 0.75) indicates a strong relationship, etc.) should be adjusted for the sample size. To overcome such a troublesome situation, we propose the following procedure that considers sample size by incorporating the critical values into the inference, which computes the statistic labeled “test”:

$$tes{t}_i=\frac{rh{o}_i^e-rh{o}_i^{cr}}{rh{o}_i^{cr}},$$

(1)

where $rh{o}_i^e$ stands for the Spearman coefficient estimate and $rh{o}_i^{cr}$ is the critical value of Spearman coefficient for a given sample size. A straightforward interpretation of Equation (1) is that the “test” can be seen as the proportional change of $rh{o}_i^e$ compared to its critical value. The maximum value of “test” is the one where $rh{o}_i^e\to 1$. Therefore, we constructed another index which can facilitate the interpretation of comparing different correlation coefficients. We termed it “Ratio” and defined it as $Rati{o}_i=tes{t}_i/\max \left(tes{t}_i\right)$. In turn, on the basis of Ratio’s magnitude, we may have a robust inference of the relative strength of Spearman estimates from different samples. In particular, we propose the following quartiles:

Ratio Values	Inference
Ratio >0.75	Very strong association
0.5 < Ratio ≤ 0.75	Strong association
0.25 < Ratio ≤ 0.5	Moderate association
Ratio ≤ 0.25	Weak association

4. Results

The first step here is to examine whether the dataset follows a normal distribution, the results of which are given in

Table 2. Normality tests for the Environmental Performance Index (EPI).

		Shapiro–Wilk test	Reject Normality	Anderson–Darling test	Reject Normality
2006	statistic	0.968	Yes	1.201	Yes
	p-value	0.003		<0.01
2008	statistic	0.959	Yes	1.913	Yes
	p-value	0.000		<0.01
2010	statistic	0.986	No	0.666	No
	p-value	0.125		0.080
2012	statistic	0.990	No	0.311	No
	p-value	0.473		>0.15
2014	statistic	0.981	Yes	0.776	Yes
	p-value	0.020		0.043

for the EPI and in

Table 3. Normality tests for the Democracy Index (DI).

		Shapiro-Wilk Test	Reject Normality	Anderson-Darling Test	Reject Normality
2006	statistic	0.959	Yes	1.985	Yes
	p-value	0.00		< 0.01
2008	statistic	0.963	Yes	1.691	Yes
	p-value	0.000		< 0.01
2010	statistic	0.959	Yes	2.049	Yes
	p-value	0.000		< 0.01
2012	statistic	0.971	Yes	1.410	Yes
	p-value	0.002		< 0.01
2014	statistic	0.967	Yes	1.634	Yes
	p-value	0.001		< 0.01

for the DI. These tables present the estimates obtained and the estimated probability (denoted as a p-value) concerning the null hypothesis. Typically, a small value of the estimated probability (typically ≤0.05) indicates strong evidence against the null hypothesis [69]. Consequently, in such a case, we can reject the normality assumption and argue that there is evidence that the data tested are not from a normally distributed population. Specifically, the normality hypothesis, concerning EPI data, is rejected for 2006, 2008, and 2014, but cannot be ruled out for 2010 and 2012.

Table 3 gives the respective normality tests for the DI. It is clear that the normality assumption is rejected in all cases.

Consequently, by jointly taking into account Table 2 and Table 3, the assumption of normal distributed data can be rejected. Hence, the Spearman rank order correlation is the appropriate statistic for assessing the likely association between the EPI and the DI.

In turn, the tables below give the main results of the paper.

Table 4. Non-parametric correlations between EPI and DI; * for N = 100 and significance level α = 0.5.

	2006	2008	2010	2012	2014
Spearman rho	0.726	0.697	0.628	0.697	0.629
N	132	144	153	126	160
Critical value *	0.197	0.197	0.197	0.197	0.197
Reject H0	Yes	Yes	Yes	Yes	Yes

presents the correlation coefficients for all countries in the period under examination.

In order to assess the possibility of synchronized variables, two steps were followed sequentially. Firstly, it was considered whether a statistically significant relationship between these variables exists. This can be done by comparing the value of an estimate with the critical value of the specific test. If the (correlation coefficient) estimate exceeds the critical value, the relationship is statistically significant, and vice versa. A typical source of critical values for Spearman’s test is Reference [78] (p. 962). Secondly, the strength of the likely association between the examined variables was determined by comparing the absolute value of the estimate with the pre-specified standards. In social sciences, these standards which can be used to assess the evidence of data synchronization are the following: a very strong correlation (0.7–1), a strong correlation (0.5–0.7), a moderate correlation (0.3–0.5), a weak correlation (0.1–0.3), and no correlation (0–0.1).

The results presented in Table 4 seem to suggest a clear and strong association between environmental quality and democracy. In other words, Table 4 confirms a clear and sound conclusion that democracy is good for the environment.

4.1. Gaining a More Nuanced Picture by Breaking down Democracy into More Factors

The paper holds that, when it comes to exploring the relationship between democracy and the environment, the devil is in the detail. As Acemoglu et al. [79] argue, it may be the role of omitted variables that inflate the association between environment and democracy. In order to shed some light on this controversial issue, we stratify the dataset on the basis of some institutional and economic criteria, as follows:

a)

Government effectiveness (GE) drawn from the Worldwide Governance Indicators (WGI);

b)

Countries’ classification by income proposed by the World Bank;

c)

Corruption Perception Index (CPI) taken from Transparency International (TI).

In doing this, it is possible to examine whether these variables (government effectiveness, economic prosperity, and corruption perceptions) affect the association between democracy and environmental quality. The choice of the previous variables was based on compelling arguments in the scholarly literature [7,80].

The GE index represents one of the six aspects of governance monitored by the World Bank in order to release the WGI The effectiveness of governance primarily refers to the quality of policy formulation and implementation, and the credibility of the government’s commitment to these policies. The WGI indicators are based on several hundred variables collected from 31 different data sources, capturing people’s perceptions of governance as reported by survey respondents, non-governmental organizations (NGOs), commercial business information providers, and public sector organizations globally [81]. The values of GE were normalized and the whole sample was stratified into four groups. These were (a) low government effectiveness (score 0–0.25), (b) moderate government effectiveness (score 0.26–0.5), (c) high government effectiveness (score 0.51–0.75), and (d) very high government effectiveness (score 0.76–1).

Table 5. Non-parametric correlations between EPI and DI by government effectiveness.

		2006	2008	2010	2012	2014
Government Effectiveness
Low	Spearman rho	−0.020	−0045	−0.194	−0.083	−0.538
	N	15	11	18	9	12
	Critical value	0.521	0.618	0.472	0.7	0.587
	Reject H0	No	No	No	No	No
Moderate	Spearman rho	0.290	0.375	0.240	0.222	0.160
	N	55	65	61	48	61
	Critical value	0.267	0.245	0.237	0.284	0.237
	Reject H0	Yes	Yes	Yes	No	No
	Test	0.09	0.53	0.01
	Max value	2.75	3.08	3.22
	Ratio	3.1%	17.2%	0.4%
	Inference	Weak	Weak	Weak
High	Spearman rho	0.782	0.474	0.622	0.456	0.401
	N	20	30	35	31	46
	Critical value	0.447	0.362	0.335	0.340	0.281
	Reject H0	Yes	Yes	Yes	Yes	Yes
	Test	0.75	0.31	0.86	0.34	0.43
	Max value	1.24	1.76	1.99	1.94	2.56
	Ratio	60.6%	17.6%	43.2%	17.6%	16.7%
	Inference	Strong	Weak	Moderate	Weak	Weak
Very high	Spearman rho	0.281	0.477	0.416	0.549	0.485
	N	20	21	23	25	27
	Critical value	0.447	0.435	0.415	0.398	0.382
	Reject H0	No	Yes	Yes	Yes	Yes
	Test		0.097	0.002	0.379	0.270
	Max value		1.30	1.41	1.51	1.62
	Ratio		7.4%	0.2%	25.1%	16.7%
	Inference		Weak	Weak	Moderate	Weak

presents the correlation coefficients disaggregated by Government effectiveness groups.

Table 5 reveals an interesting set of results. Firstly, when government effectiveness is low, then there is no statistically significant relationship between democracy and environmental quality. Some examples of countries which belong to this group are Afghanistan, Togo, Sudan, and Zimbabwe. Secondly, as government effectiveness increases, there is evidence of a weak association between democracy and environmental quality, albeit not always statistically significant. Examples of countries that fall within this group are Zambia, Nepal, Morocco, and Ecuador. Under high government effectiveness, the evidence reflects a statistically significant association between environmental quality and democracy. However, the strength of this association is not consistent through time. It is mostly weak for 2008, 2012, and 2014, moderate for 2010, and strong for 2006. Some examples from this group are Greece, Poland, Spain, Uruguay, and Jamaica. Finally, when government effectiveness is very high, there is evidence of a mostly weak link between democracy and environmental quality, although not always statistically significant. Examples from this group are Australia, Germany, Finland, and Canada.

Similarly, mixed results concerning the role of government effectiveness on environmental quality were previously reported in the scholarly literature [82]. Some researchers argue that government effectiveness, broadly defined as the capacity of state institutions to implement legitimate objectives effectively, seems to decline in young democracies [83]. Likewise, Bäck and Hadenius [84] argue that the relationship between democracy and government effectiveness is not monotonic and depends upon the level of democracy. Furthermore, in authoritarian states, there is a negative relationship between government effectiveness and environmental quality, while at the same time environmental protection increases with higher government effectiveness [85].

The results presented in Table 5 suggest a complex association between democracy, environmental quality, and government effectiveness. Only when there is high government effectiveness (GE score 0.5–0.75) can we affirm a clear and mostly weak association between democracy and environmental quality. Such an observation may imply that high government effectiveness in designing and enforcing policies for environmental protection seems to be among the necessary prerequisites for a democratic regime to mobilize these social processes that are beneficial for the environment. In all other cases, there are either non-existent links (statistically insignificant) or mixed evidence (co-existence of significant and non-significant results). By reversing the previous argument, one may try to explain the cases of low and moderate government effectiveness. In these cases where governments are ineffective at applying policies for environmental protection, democracy’s role may be restricted. In stark contrast, when governments are very highly effective, probably the role of democracy is already accomplished. Nevertheless, Mol [86] is rather skeptical with the notion that associates state capacity and power with environmental protection. Above and beyond that, another plausible explanation may concern the quality of the data used. In particular, WGI data were severely criticized as being the product of “a complex atheoretical and poorly articulated hypothesis” [87].

Table 6. Non-parametric correlations between EPI and DI by income groups.

		2006	2008	2010	2012	2014
Income Group
High	Spearman rho	0.666	0.564	0.690	0.637	0.668
	N	32	36	41	39	41
	Critical value	0.351	0.330	0.309	0.317	0.309
	Reject H0	Yes	Yes	Yes	Yes	Yes
	Test	0.90	0.71	1.23	1.01	1.16
	Max value	1.849	2.030	2.236	2.155	2.236
	Ratio	48.5%	34.9%	55.1%	46.9%	52.0%
	Inference	Moderate	Moderate	Strong	Moderate	Strong
Middle	Spearman rho	0.447	0.424	0.352	0.487	0.293
	N	62	71	74	65	80
	Critical value	0.251	0.233	0.229	0.245	0.220
	Reject H0	Yes	Yes	Yes	Yes	Yes
	Test	0.78	0.82	0.54	0.99	0.33
	Max value	2.984	3.292	3.367	3.082	3.545
	Ratio	26.17%	24.90%	15.95%	32.05%	9.36%
	Inference	Moderate	Weak	Weak	Moderate	Weak
Low	Spearman rho	0.054	−0.021	−0.011	0.346	0.021
	N	38	37	36	22	40
	Critical value	0.321	0.326	0.330	0.425	0.313
	Reject H0	No	No	No	No	No

depicts the association between the EPI and the DI split by the World Bank classification by income for brevity and simplicity, the lower–middle-income and the upper–middle-income groups were joined together under the label of middle income. This is a standard aggregation in the literature [88].

The conclusion to be drawn from this analysis is that the role of democracy is nonexistent for low-income countries since no statistically significant link exists between the EPI and the DI (Table 6). Such a link is statistically significant for middle-income countries (weak to moderate) and for high-income countries (moderate to strong). Judging from the strength of this link, it can be argued that democracy is probably good for the environment for rich countries, while there is evidence of a weak link for middle-income countries. Such a result supports the assertion made in Reference [42] where it is impossible to assess the implication of freedoms on environmental quality independent of incomes. A plausible explanation may be that a rich country, compared to a poor one, is more likely to acknowledge and embrace post-materialist values. The presence of these values can predict pro-environmental attitudes [88]. In turn, it is possible that these attitudes could be mobilized toward societal goals that include environmental protection.

Li and Reuveny [8] found that democracies have a positive effect on income distribution; however, Ravallion et al. [89] claim that the impact of income inequality on environmental degradation depends on the marginal propensity to emit (MPE). If the poor have a higher MPE than the rich, reducing income inequality will increase emissions of pollution; if they have lower MPE, then the reduction of inequality reduces the pollution. It means that more variables broaden the view on the democracy/income and environmental concerns.

Finally,

Table 7. Non-parametric correlations between EPI and DI by Corruption Perception Index (CPI) groups.

		2006	2008	2010	2012	2014
CPI group
Seriously corrupted	Spearman rho	−0.082	0.311	0.418	0.393	0.065
	N	16	17	18	14	16
	Critical value	0.503	0.485	0.472	0.538	0.503
	Reject H0	No	No	No	No	No
Moderate corrupted	Spearman rho	0.630	0.091	0.390	0.351	0.212
	N	16	24	21	24	30
	Critical value	0.503	0.406	0.435	0.406	0.362
	Reject H0	Yes	No	No	No	No
	Test	0.25
	Max value	0.99
	Ratio	25.6%
	Inference	Moderate
Clean	Spearman rho	0.476	0.463	0.374	0.432	0.228
	N	68	65	70	75	91
	Critical value	0.239	0.245	0.235	0.237	0.206
	Reject H0	Yes	Yes	Yes	Yes	Yes
	Test	0.99	0.89	0.59	0.82	0.11
	Max value	3.184	3.082	3.255	3.219	3.854
	Ratio	31.1%	28.9%	18.2%	25.6%	2.8%
	Inference	Moderate	Moderate	weak	Moderate	Weak
Very Clean	Spearman rho	0.237	0.284	0.262	0.421	−0.113
	N	35	46	39	19	22
	Critical value	0.335	0.291	0.315	0.460	0.425
	Reject H0	No	No	No	No	No

displays the association between the EPI and the DI split by the CPI scores. The CPI ranges from 0 (high corruption) to 100 (absence of corruption). As in the case of government effectiveness, four groups were created: (a) seriously corrupted (score 0–25); (b) moderately corrupted (score 26–50); (c) clean (score 51–75); and (d) very clean (score 76–100).

Table 7 shows that the positive association between EPI and DI is statistically significant only for the third group, i.e., the clean countries classified on the CPI score. The exception of the second group in 2006 is ignored on the basis that there is no statistically significant relationship in all other years; it must be a data problem. The third group is the most populated one and comprises all the relatively clean countries. Some examples are the United States of America (USA), France, Cyprus, and Malta. However, judging by the strength of such an association between EPI and DI, it can be argued that, although this is a statistically valid relationship, it is quite weak. While it is generally accepted that the control of corruption is a clear and direct determinant of environmental quality [90], there is no consensus about its indirect impact through a democratization process. For example, Morse [91] provides evidence of a positive link between corruption and environmental sustainability; however, at the same time, he also gives a clear warning about aggregate measures of preferences and perceptions as the CPI. By contrast, Povitkina [20] argues that corruption undermines the function of democracy and, consequently, corrupted countries cannot exploit the well-known advantage of democratic regimes to improve environmental quality. Welsh [92] argues that the effect of corruption on environmental quality is not monotonic. Only for low-income countries may controlling corruption improve environmental quality [92].

A similar line of reasoning as in the case of government effectiveness can be put forward here. Specifically, if a country is corrupted, or perceived as such, the democratic institutions are likely to be thin and ineffective to contribute toward environmental quality (see countries such as Yemen, Kenya, or Mexico). On the contrary, a very clean government, such as that in Switzerland, Sweden, Finland, or New Zealand, is most likely to be in a full democratic state where the regime’s nature already capitalized on the quality of the environment.

5. Conclusions

This paper revisits the environment/democracy debate by providing a concise review of some of the major elements of the existing discussion. The paper provides another look at the issue by using appropriate aggregate measures of environmental quality and democracy and then examining the most important interaction effects.

The results are twofold. Firstly, after taking into account the whole dataset, we show a positive and statistically significant association between environmental quality and democratic regime. However, such a clear result vanishes if one examines the interaction effects. Specifically, the positive role of government effectiveness should be expected only for those countries that are effective enough in designing and implementing policies (GE score 0.51–0.75). Nothing can be argued for the remaining countries. In addition, only weak evidence exists that the control of corruption may affect the role of democracy on the environmental quality, and this evidence only refers to countries that are relatively free from corruption (CPI score 51–75). By contrast, a country’s income level clearly affects the positive association between democracy and environmental quality. In other words, the anticipated positive link between democracy and environmental quality should be assessed in the light of possible interaction effects.

The policy implications of these results are profound, in the sense that they support the rhetoric of the Environmental Kuznets Curve. Although it is tempting to use linear thinking and to advise a country to first get richer and then to introduce democratic improvements and control corruption in order to significantly improve environmental quality, the latter is a simple consequence of the modernization hypothesis under which economic growth typically consolidates democratic institutions. Against the conventional wisdom of the modernization hypothesis which still dominates the debate, the empirical evidence does not provide unanimous support that a significant causal relationship exists between income and democracy [93,94,95].

Additionally, we would like to highlight three interrelated points that reinforce the relevance of the article and attempt to answer the important question of “so what?”. One of the crucial findings is that future research on the subject should take a more nuanced approach to try to understand the relationship between democracy and the standard of environmental policy in any given state or region. The research revealed that the assumption “more democracy is equal to a higher level of environmental policy” needs to be qualified by looking at other factors and variables, as identified in this article. Moreover, there is a policy-related argument. The findings of the paper suggest that donors and International Organizations, when supporting environmental campaigns in developing and transitional states, need to make sure that their strategies are grounded in a sufficiently sensitive understanding of democracy and democratization, including awareness of the effects of corruption and levels of political participation. The future research needs also to be able to take account of the possibility of democracy back-sliding and its effects on environmental policy, since the rise of populism in Europe and the USA already, in some cases, brought with it a disavowal for environmentalism.

Finally, a few words about the main limitations of the paper are necessary. Specifically, the approach adopted does not provide any causality analysis between democratic regimes and environmental quality. It is our contention that the results from this paper can provide valuable guidance toward a more thorough investigation of the causality issue via more sophisticated methods. Secondly, our paper was constrained by the quantity and quality of the available data. Therefore, further research is needed to unfold the likely causality between income, democracy, and the environment.