Democratisation, Environmental and Income

Inequality.

Laura PolicardoUniversity of Siena

Abstract

Empirical economists commonly agree upon the fact that the de-mand for environmental goods is increasing with income, so as long asdemocratisation shifts the decisive power from a rich individual (auto-crat or dictator) to a poorer one (decisive voter), such a regime changeshould be associated to worse environmental conditions. Assumingthat citizens’ wealth does not depend on inherited endowments, butonly on each individual’s share on total production, I will show with atheoretical model that contrary to the expectations, democratisationmay have mixed effects on the level of environmental quality, dependingon the size of the price and income effects on the demand for environ-mental quality associated to a decrease in the decisive political actor’swealth. If indeed a poorer individual desires less environmental qual-ity than a richer individual (income effect), he also desires less of thegoods responsible for degradation (price effect) so the overall result onthe environment is ambiguous. Assuming instead that society is com-posed by two classes of individuals, one supplying an embodied factorof production and one supplying capital, and assuming moreover thatthe decisive voter belongs to the first class of individuals while theautocrat does not, democratisation is shown to be beneficial for theenvironment, the better the effect on the environment, the bigger thedifference in wealth between the two decisive political actors.

JEL Classification: D02, D31, O13, Q56

Keywords: Democracy; Environment; Environmental Inequality; In-come Inequality.

0

1 Introduction

Theoretical and empirical literature about the effects of democracy on thelevel of environmental quality is still inconclusive and does not provide clearand unilateral answers to the effect of different regimes on the environment.In the seventies, the dominant thinking was that democracy and its asso-ciated liberties to consume and procreate would have generated ecologicalcatastrophes (Desai, [10], Hardin, [22]); recently, on the contrary, scholarshave found positive effect of democracy on the environmental quality, themain reason being the necessity for democratic governments to be electedor re-elected, which makes them more prone to respect human life, and thenmore responsive to life threatening environmental degradation (Dasguptaand Maler [8], Schultz and Crockett [41], Payne [39], Gleditsh and Sverdrup[18]).

This paper is aimed at contributing on the debate about whether democ-racy may be good or not for the environment. The questions I want toanswer are: what is the role of a political regime in the determination ofa country’s environmental policy? Are democracies more environmentalistthan autocracies? In order to answer these questions, it is important firstof all to understand the difference between those two regimes. In a (direct)democracy, decisions are taken by majority voting, while in a dictatorshipthey are taken by one or a group of few individuals. In practice, therefore,the decisive political actor is different, and what makes them differ is oftentheir wealth.

It is well known that wherever in the world, at any time in history,kings, queens and dictators have been the richest and powerful people. Fromnorth to south, in developed and underdeveloped countries, in communistand non-communist countries, the only characteristic that this variegatedgroup of people has in common is a lofty position and vast fortune. Forbes,in an article appeared few years ago1, estimated the Saudi Arabia KingAbdullah Bin Abdulaziz’s wealth in around 45% of the country’s $340 billionGDP. Brunei’s Sultan Haji Hassanal Bdkiah has an estimated fortune of $20billion, and benefits from petroleum and natural gas fields. The Presidentof the United Arab Emirates, Sheikh Khalifa bin Zayed Al Nahyan ownthe 90% of the 2.5 million barrels a day exported from the UAE and healone represents two third of the GDP. Mutubu Sese Seko in the DemocraticRepublic of Congo (Zaire) during the 70’s benefitted of around 15-20% ofCongo’s operating budget and in the 80’s his fortune was estimated in $5billion (Leslie [28], p. 72). Rafael Trujillo in the Dominican Republic hada fortune which amounted to 100% of the national GDP at current pricesand his family controlled almost 80% fo the country’s industrial production

1The article is available online at http://www.forbes.com/2006/05/04/rich-kings-dictators cz lk 0504royals.html

1

(Moya Pons [34], p. 398). These are only few examples, but the list ofvery rich rulers is not over and probably it would be a manuscript on itsown. From this perspective, a regime change from dictatorship to democracyshifts the decision power from the dictator to the median voter, implyingan impoverishment of the decisive political actor and therefore a differentwillingness to pay for environmental protection.

If the demand for environmental quality was an increasing function ofincome, dictatorship would be a panacea for the environment. Numer-ous empirical studies suggest that willingness to pay for environmental im-provement is an increasing function of wealth (Hokby and Soderqvist [24],Kristrom and Riera [26], Miles, Pereyra and Rossi [33]). Richer peopletend to express higher willingness to pay for environmental improvementthan poorer people, that is, environmental quality is typically found to bea normal good. Because of this, an institutional shift from democracy todictatorship should result in an improvement of a country’s environmentalquality. As pointed out by Boyce, [5], however, this beneficial effect maynot necessarily bring to the expected result, because other factors come intoplay to mitigate or even nullify it.

First, because an increase in the decisive political actor’s income raiseshis “price” of the environmental protection. Increasing income means in-creasing the share the individual gets out of production, if production iskept constant. Since production requires some environmental depletion, itis easy to realise that environmental protection is more costly for the richbecause a decrease in global GDP results in a greater decrease in his ownincome than on the poor’s, who gets a lower share. Moreover, even thoughricher individuals may desire more environmental quality (income effect),they also desire more of the goods and services responsible for environmen-tal degradation (price effect), so the expected environmental improvementwill occur only if the negative “price effect” does not outweigh the positive“income effect”.

Second, because richer people have more resources and capabilities tosubstitute environmental quality with private consumption, giving raise tothe so-called environmental inequality. If indeed the willingness to pay(WTP) for environmental protection is an increasing function of income,different is the elasticity of such willingness to pay. Martini and Tiezzi [32]indeed find, in a panel of Italian households from 1999 to 2006, that de-spite the willingness to pay for environmental protection is higher for richerpeople, the elasticity of WTP is less than one, and therefore decreases as in-come increases, suggesting a lower capability for poorer people to substituteenvironment with private consumption.

Substantial evidence shows indeed that the less wealthy and powerfulmembers of societies may be the ones exposed to much heavier environ-mental degradation than the more well off, being able to avoid it, simply

2

relocating to cleaner living areas or using their political power to drive outpolluting industries from their neighborhoods. Gray and Shadbegian [19],for instance, obtain some evidence from US data on the paper industry forthe period 1985-1997 that polluting emissions are significantly lower in areaswith more children, older people and fewer poor people than in areas withyoung poor people without children, and Bina Agarwal [2], for instance,documented how the degradation of forest resources in rural India had par-ticularly severe effects on poor women, via impacts of their time, incomeand nutrition. Pastor, Sadd and Hipp [38], for instance, provide substan-tial evidence that minorities residential areas have an higher likelihood tohost various environmental hazards, and Sheila Foster [15] reports that poorafrican american neighbourhoods in Chester (Pennsylvania) often experiencea clustering of waste facilities in their areas.

In this paper, I provide a theoretical model through which I analyse theeffects of democratisation on the environment, meaning by democratisationa shift of the decisive power from a rich individual (dictator) to a poorerone (decisive voter).

I will assume that citizens differ only by their wealth, meaning by wealthonly income (or share they get out of production) and excluding any inher-ited endowment, to keep the model as simple as possible.

In the first instance, I will analyse the result of democratisation on thelevel of environmental quality assuming that there is no environmental in-equality, so all the citizens experience the same level of pollution.

I will later remove this assumption by introducing a class model of expe-rienced pollution, and assuming that the society is composed by two typesof individuals - employers and employees. Employers supply capital in theproduction process, while employee supply their physical labor and there-fore cannot avoid their own exposure to pollution. Employers, on the otherhands, supplying capital, can relocate far away (at a cost) from their pollut-ing factories, being not necessary their physical activity in the productionprocess.

As it is often unexpected, income inequality per se between the twodecisive political actors may not be responsible of different environmentalpolicies undertaken by the two different regimes, but environmental inequal-ity is the crucial variable affecting them. Income inequality, therefore, hasan indirect role as long as it induces environmental inequality. In countrieswhere the majority of the population is represented by employees whose ser-vices require physical presence in the firm, democratisation is expected to bebeneficial for the environment, and this is in line with the recent empiricalevidence.

The next section reviews the literature on the topic of environmentalquality and regime, and section 3 introduces the model. Section 4 presents

3

some empirical evidence and discusses some cases of regime transition, hav-ing regards to the effect on environmental policy of democratisation. Section5 concludes.

2 Literature review

Literature on the effect of political regimes on pollution or pollution controldates back to the seventies and the main view of that time was that democ-racies and their associated liberties to consume, pollute and procreate wouldhave generated environmental catastrophes (Desai [10], Hardin [22], Ophuls[37] and Heilbroner [23]).

Desai (1998, [10]) has argued that “as democracy is dependent on eco-nomic development, and since economic growth and prosperity generally re-sult in environmental pollution and ecological destruction, democracy wouldnot necessarily be protective of the environment.” This argument, however,says little about the independent effects of democracy, separate and apartfrom income.

Same criticism can be extended to Hardin (1969, [22]) and Heilbroner[23] too. In Hardin’s article on the tragedy of the commons, he does notexplicitly refer to democracy as a cause - ceteris paribus - of pollution, but herefers to it indirectly, saying that pollution is a direct cause of the increase inpopulation, and one characteristic of democracy is its freedom to procreate.

Heilbroner [23] bases its arguments on the projections made by Jay For-rester and a team of scientists of MIT about the population growth, andtherefore an always shorter carrying capacity of the earth. He moreoverestimates a total “self destruction” of the human race by the year 2100 ifthe rate of growth of the economy is kept constant.

In more recent years, however, empirical evidence suggests the contrary,that is to say, that democracies tend to respect the environment more thanautocracies, and the reason often provided is the necessity of governments torespect human life (and therefore to be more responsive to life threateningenvironmental degradation) because they want to be re-elected (Dasguptaand Maler [8], Schultz and Crockett [41], Payne [39], and Gleditsh andSverdrup [18]).

In support of this view, Congleton [6] develops a theoretical model show-ing that authoritarian regimes (i.e. dictators) often face a higher relativeprice for pollution abatement than a median voter does, due to the fact thatthe dictator has a shorter time horizon and gets a bigger fraction of income(i.e. he is richer). If indeed this is the case, autocracies will adopt less strin-gent environmental standards than democracies. In his model, however, hedeals with “global” pollution problems, and assumes that the level of expe-rienced environmental quality does not differ between the median voter andthe dictator.

4

In the empirical literature, Bhattarai and Hamming [4], for example,use a measure of institutional quality (measured by an index of politicalrights and civil liberties) to account for the role of different policy regimesin the causes of deforestation in 66 Latin American, African and Asiancountries during the period 1972-1991. Using a fixed effect panel data modelestimated by weighted least squares (thus correcting for some of the outliersand measurement errors by giving less weight to such outliers), they claimthat democratic institutions significantly reduce deforestation.

Torras and Boyce [45] use a similar technique for a panel data survey ofseven air and water pollution indicators. They basically test whether powerequality (which is higher in democracies) may affect the level of pollution inmore than 1000 locations worldwide, and find that the inclusion of variableslike Gini index of income inequality, literacy and political rights - which areproportionally correlated to power equality and therefore democracy - allhave the effect of lowering the peak of the Environmental Kuznets curve, ormaking GDP not significant in the regression.

Gallagher and Thacker [17] introduce a concept of “stock of democracy”to study its implications through time, and they also find positive effect ofdemocracy on the environment.

Barrett and Graddy [3] find that countries that are more democratictend to have lower levels of sulphur dioxide emissions, smoke, particulates,arsenic and lead. Harbaugh et al. [21] found that democracies significantlyand constantly reduce sulphur dioxide, smoke and particulates in a panel ofcities worldwide (the number of cities varies according to availability of datafor the indexes of pollution analysed).

Neumayer [35] tests Payne’s hypothesis that democratic governments aremore prone to collaborate with each other on environmental issues. He findsthat democracies sign and ratify more multilateral environmental agree-ments, participate in more environmental intergovernmental organisations,comply better with reporting requirements under international agreements,put a greater percentage of their land area under protections status using amultivariate regression in a cross-country setting.

Farzin and Bond [13] find democracy associated with lower levels ofcarbon dioxide, nitrogen oxides, volatile organic compounds, and sulphurdioxide. They also find that other covariates like income inequality, agedistribution, education and urbanization may mitigate or exacerbate thenet effect of the political regime on pollution.

Fredriksson and Wollscheid [16], use the method of propensity scorematching with a sample of 163 countries from the late 1990’s to test forthe hypothesis that democracy tend to be more environmentalist than au-tocracies. They also observe differences in the level of stringency of envi-ronmental policies among different types of democracies. Controlling forthe price of super and diesel gasoline in 1998 and 2000, age distribution,population density, urbanization, percentage of land forested, percentage of

5

Table 1: Summary of the literature review

Authors Theoretical/Empirical Effect of democracy onenvironmental quality

Desai, 1998 Theoretical -Hardin, 1969 Theoretical -Ophuls 1974 Theoretical -Heilbroner 1972 Theoretical -Dasgupta and Maler, 1995 Theoretical +Schultz and Crockett 1990 Theoretical +Payne 1995 Theoretical +Gleditsh and Sverdrup 2003 Empirical +Congleton 1992 Theoretical +Bhattarai and Hamming 2001 Empirical +Torras and Boyce 1998 Empirical +Gallagher and Thacker 2008 Empirical +Barrett and Graddy 2000 Empirical +Neumayer 2002 Empirical +Farzin and Bond 2006 Empirical +Fredriksson and Wollscheid 2007 Empirical +

land used in agriculture as well as GDP percapita and the ratio of exportplus import to GDP, they consistently find that democracies (and in partic-ular, parliamentary democracies) tend to set more stringent environmentalpolicies than autocracies.

They all find positive evidence that civil liberties and political rights areassociated with more pro-environmental behaviour.

Table 1 summarizes all the contributions cited in this section. As it ispossible to observe, older (and mainly theoretical) contributions assumethat liberties (characteristics typical of democratic countries) to procreate,consume and produce would have damaged the enviroment, even though themain critics to this view is that they do not analyze the effect of the regimeseparate and apart from all the other characteristics of a country (levelof production, population, etc...). Recent (and mainly) empirical articlesinstead find a positive association between democracy and environmentalquality.

3 The model

This section presents the model, with the aim of analysing the effect ofa regime change (in particular, democratisation) on the optimal level ofenvironmental quality. In the first instance, I will assume that the decisivepolitical actor in autocracy is richer than in democracy, so democratisationresults in a shift of the decisive power from a rich individual (the dictator)to a poor one (the median voter). Poorer people have a tendency to value

6

consumption more than environmental quality, but richer people consumemore of the good which are responsible for pollution. The first part of themodel will deal with this issue, assuming that the environment is a purepublic good whose exposure cannot avoided by anyone at any cost.

The second part instead will introduce a model of class differences inexperienced environmental quality to take into account of the effect of envi-ronmental inequality. Assuming that the decisive political actors in democ-racy and autocracy belong to different classes, and in particular, assumingthat the dictator is a capitalist employer who supply only capital in the pro-duction process and therefore is able to relocate his home far away from thepolluting sources, while the median voter is a employee who supply physicallabor and therefore is exposed to emission, I will show that democracy isbeneficial to the environment, the more beneficial the higher the differencein wealth between the dictator and the median voter.

3.1 Democratisation and income inequality

Throughout this section, I will assume that:

A. 1. The decisive political actor under an autocratic regime receives a largershare of GDP than under a democratic one.

Denote then by g the level of production, or GDP, and by e the levelof environmental quality, with both g and e assured to be positive. En-vironment is an essential factor of production, so there cannot be positiveproduction without some environmental depletion. Equation 1 representsthe relation between production and environmental quality, i.e., the trans-formation locus between environment and income:

g = f − 1

2e2 (1)

with f a positive constant. The maximum achievable level of production isbounded above due to technological constraints and this boundary is rep-resented by f . A level of production equal to f can be achieved only bycompletely depleting the environment. Conversely, the highest environmen-tal quality requires absence of production.

Suppose than that the society is composed by N individuals, with allhaving identical preference function. Each generic individual i gets a share oftotal production si, with

∑Ni=1 s

i = 1 (notice that, in general, in general si 6=sj for i 6= j) as a “personal wage” available for consumption, and therefore,the amount each individual can consume in each period is represented by

yi = sig (2)

7

and therefore, the “personal” marginal rate of transformation between in-come and environmental quality is represented by

∂yi

∂e= −esi. (3)

Each individual’s utility is a nonseparable function of income and envi-ronmental quality and it is denoted by

ui = u(yi, e) i = 1, ..., N (4)

with ui representing the level of utility for individual i, yi denotes income(or consumption) of the same individual, and e the level of environmen-tal quality. Environmental quality is a public good which is experienceduniformly across the population, hence the absence of the superscript (thisassumption, however, will be removed later). The utility function ui is as-sumed increasing and concave with respect to both arguments, and havepositive cross derivates. In other words, I have ue > 0, uy > 0, uee < 0,uyy < 0, uey > 0 and uye > 0, where ux is the first derivative of the utilityfunction with respect to the generic variable x and uxz is the first derivativeof ux with respect to another variable z. These assumptions imply thatthe utility generated by increasing income or environmental quality is pos-itive, but it increases at a decreasing rate, and that the pleasure derivedby income (cleaner environment) is higher, the higher is the environmentalquality (income).

Under assumption A.1., if the level of experienced environmental qualitydoes not differ between different people, the effect of democratisation (au-tocratisation) can be read as the effect of an impoverishment (enrichment)of the decisive political actor2. Denote with v the decisive political actorunder democracy, and with a the decisive individual under autocracy. As-sumption A.1. has the only implication that sv < sa, so the problem thedecisive citizen faces is

maxe

ui = u(yi, e) i = v, a (5)

s.t. yi = si · g (6)

g = f − 1

2e2 (7)

so they decide their optimal levels of environmental quality so as to maximisetheir utility subject to their wealth and the available technology.

2The first part of the model assumes that exposure to pollution does not vary betweenthe two political actors. A regime change therefore modifies - ceteris paribus - only thefraction of income (i.e. wealth) perceived by the two distinct individuals. In the nextsubsection, as we will see, a regime change not only modifies the wealth of the decisivepolitical actor, but also its exposure to emissions.

8

The first order condition for utility maximisation requires

−sie · uyi(yi, e) = −ue(yi, e) or (8)

−esi = − ue(yi, e)

uyi(yi, e)

(9)

where uyi represents the derivative of the utility function for individual iwith respect to the level of income yi, and ue represents the derivative ofthe utility for individual i with respect to the environmental quality. Equa-tion 9 simply says that as a necessary condition for maximising utility is theequality between the “personal” marginal rate of transformation between in-come and environment and the marginal rate of substitution between incomeand environment. The level of environmental quality e chosen by individuali, therefore, will be

e∗ =1

si· ue(y

i, e)

uyi(yi, e)

. (10)

A first question that comes into mind is whether the effect of a marginalincrease in the share of income of this citizen increases the level of environ-mental quality or not. Taking the first derivative of e∗ with respect to si,we get that

∂e∗

∂si=

1

si

[− 1

si

(ue(y

i, e)

uyi(yi, e)

)+

+ g

(ueyi(y

i, e)uyi(yi, e)− ue(y

i, e)uyiyi(yi, e)

(uyi(yi, e))2

)](11)

which is positive (so the effect of an increase in si is good for the environ-ment) when

1

sig< −

uyiyi(yi, e)

uyi(yi, e)

+ueyi(y

i, e)

ue(yi, e). (12)

Defining

ue(yi, e)

uyi(yi, e)

= MRSie,y (13)

−esi = MRT ie,y (14)

it is easy to see that equation 12 can be rewritten as3

∂MRT ie,y/∂y

MRT ie,y/y

<∂MRSi

e,y/∂y

MRSie,y/y

(15)

3For the computation of the derivative with respect to y of the marginal rate of trans-formation, we make use of equation 6 according to which si = yi/g. Substituting si into14 and taking the derivative with respect to yi, we get −e/g, which, once divided by −esi

gives 1/sig.

9

so, a positive variation in the share of income is associated to a positive vari-ation in the level of environmental quality only if the “price effect” due to anhigher share of income si (which makes also the marginal rate of transfor-mation between environment and income steeper) is lower than the “incomeeffect” due to the relative variation in the marginal rate of substitution be-tween the two goods. In other words, becoming richer is beneficial for theenvironment only if the elasticity of MRT i

e,y is lower than the elasticity ofMRSi

e,y, both with respect to income.Denoting by ev and ea the optimal level of environmental quality chosen,

respectively, by the decisive citizen in democracy and by the dictator, wehave that

ev =uev(yv, ev)

sv · uyv(yv, ev)(16)

ea =uea(ya, ea)

sa · uya(ya, ea)(17)

The optimal level of environmental quality selected in democracy will belower than the level selected in dictatorship (ev < ea) if

MRSve,y

svg<

MRSae,y

sag(18)

or, rearranging,MRSv

e,y

sv<

MRSae,y −MRSv

e,y

sa − sv(19)

Equation 18 highlights the ambiguous effect (depicted in figure 1) of aregime shift. Since the level of environmental quality experienced by boththe dictator and the decisive citizen in democracy is actually the same, whatmakes the difference in determining which regime is more pro-environmentdepends basically on the difference between “relative” marginal willingnessto pay for environmental protection. An example may clarify the interpre-tation of equation 18. Suppose the decisive political actor in democracy iswilling to pay 2 unit of income to increase by one unit environmental qual-ity. Suppose instead that the dictator is willing to pay 5, because he’s richerso he values environment more. However, the difference in wealth betweenthose two decisive citizens is huge. Suppose that the dictator gets 10% of theglobal production, so sa = 0.1. Assume that the median voter gets only 1%,so sv = 0.01. Assume that the country’s GDP (g) is, for simplicity, 1000.It follows that, in percentage, the decisive political actor in democracy iswilling to give up 0.2% of his income to protect the environment while thesame decisive citizen under dictatorship is willing to pay 5%. In this case, ofcourse, democracy is bad for the environment. If inequality between thosetwo citizens instead was not too high, and the marginal rate of substitutionfor the median voter was 6 against 8 for the dictator, sv = 0.05, sa = 0.1,

10

and keeping constant g at 1000, we have that proportionally to income, themedian voter is willing to give up 12% of to increase environmental qualityagainst 8% for the dictator. This case, of course, implies that democracy isgood for the environment.

Figure 1 provides a graphical explanation of the argument above. Sup-pose that the decisive citizen’s income increase from svg to sag. Whetherthis shift is good for the environment or not, depends on the variation of theratio between the two marginal rates of substitution between y and e for thedictator and the decisive citizen in democracy. Of course, the dictator has aMRSe,y higher than the median voter’s MRSe,y, but if this increase in thewillingness to pay for environmental quality do not compensate the increasein the share of income the dictator gets, dictatorship is bad for the environ-ment (figure 1A). Figure 1B depicts the opposite case, so when democracy isbad for the environment, because the marginal utility from consumption ofthe decisive voter in democracy is much higher than the dictator’s one, so ashift from democracy to dictatorship is good for the environment. Equation18 mans exactly what is depicted in figure 1.

Figure 1.

From another point of view, equation 19 shows that we can see democrati-sation as a bad regime for the environment if the median voter’s relativemarginal willingness to pay for environmental quality does not compensatewhat would be the difference of the marginal willingness to pay under thetwo different regimes, weighted by the difference in income between the twodecisive citizens.

3.2 A model of class differences in experienced environmen-tal quality

In order to introduce environmental inequality, i.e., the fact that environ-ment is experienced in different ways by different people, I will make theassumption that the population is divided into two classes, capitalists and

11

workers. Capitalists supply capital in the production process and receive therents their capital generates. Workers can only sell their labour in the firmsowned by the capitalists and therefore cannot avoid their exposure to thepollution produced by the firms they are employed in, either because duringthe job they are exposed to emissions, or because they cannot relocate toofar away from the firms since they have to show up at work every morning.Of course, this is not the case for the capitalist, who can choose the bestlocation to live because he does not need to physically sell his own labour4.Under the assumption that the decisive political actor in autocracy is acapitalist and in democracy is a worker, the optimal level of environmentalquality chosen by those two decisive citizens will be different than in theprevious case. The worker is assumed to have an experienced environmentalquality which reflects the real status of the environment, therefore his max-imisation problem remains unchanged. The capitalist, on the contrary, maydecide to increase his own perception5 of environmental quality, for example,by relocating in a cleaner area, but this can be done at a cost. The pricemay be related, for instance, to the transportation costs the capitalist mustpay in order to reach the city where his firms are located, so the price maybe a function of the capitalist’s home distance to the industrial district.

Denote by eax the capitalist’s (or dictator) level of experience environ-mental quality. eax is no more equal to the actual, or “real” level, since itcan be increased by relocating far away from the pollution source. In detail,the equation for eax is given by

eax = e + hc (20)

where e is the real level of environmental quality, c is the cost of relocatingin a cleaner area and h is a positive constant denoting the level of “produc-tivity” of the cost of living far from the pollution sources. In this model, wewill assume that

eax ≥ e (21)

4This assumption clearly implies that the pollution problem considered is mainly local.Consequences of pollution can be indeed local or global: global warming - a clear globalconsequence of a global pollution problem may hit both rich and poor individuals (thinkfor example to the rise of the sea level, or fires originated by the high and anomaloustemperatures). The pollution problem I am considering here can be represented, forexample, by a firm’s discharges in a lake, which affect fishering of the population livingnear the lake (and firm), or fog coming from the chimny stack of a firm, who affects mainlywho lives around it, or pollution affecting the water channel of a city, which hits mainlywho cannot afford bottled water.

5“Perceived” environmental quality may differ to “experienced” environmental qualityin the sense that the perception of the environment may be subjective and mistaken. Inthis paper however, I consider the words “perceived” and “experienced” as a synonymousbecause I exclude the possibility that a person may perceive pollution differently fromother people in the same class or even wrongly.

12

so the level of experienced environmental quality cannot be lower than thereal level. This assumption guarantees that c is an effective cost, so it cannotbe negative (the employer basically cannot accept bribes in order to live closeto the polluting factory).

Equation 20 can be rephrased as follow: if one defines c the monetaryexpense the employer pays to experience a better environment, c can be readas

c = p(eax − e) (22)

that is to say, the price for an experienced environmental improvement (p)multiplied by the difference in the experienced environmental quality chosenby the employer and the real environmental quality (eax − e). If this isthe case, it is possible to notice, by equation 20, that, taking e from theright hand side of the equation to the left, the price for an experiencedenvironmental improvement p is defined as 1/h.

p ≡ 1

h(23)

Since the capitalist - or employer - pays a cost c > 0, his level of incomeavailable for consumption will be reduced of that amount, so

ya = sag − c (24)

and technology remains identified by equation 1.The new problem the decisive political actor in dictatorship faces is then

maxe,c

ua = u(ya, eax) (25)

subject to equation 20 and 24. The first order conditions for utility max-imisation are:

∂ua

∂e= 0 ⇒ ea∗ =

ueax(ya, eax)

sauya(ya, eax)(26)

∂ua

∂c= 0 ⇒ 1

h=

ueax(ya, eax)

uya(ya, eax)(27)

Equation 27 represents the equilibrium level of expenditure, which mustequate the productivity of the “insurance” against pollution and the amountof the environmental quality the dictator is willing to give up to become oneunit richer. Combining equations 26 and 27 I get

ea∗ =1

sah(28)

so the optimal level of environmental quality chosen by the dictator dependsnegatively only on the share he can get out of production and the productiv-ity of the expenditure in increasing the perception of environmental quality.

13

The reason for this result is simple: a richer citizen enjoys environmentalquality more than a poor one, but his opportunity cost to pay for environ-mental protection is higher than for a poor citizen. If he can increase hisown perception of environmental quality of an amount that is greater thanthe amount of environment he is willing to give up to become one unit richer,then it is worthwhile to pay.

In general, taking the equilibrium relations of the two decisive politicalactors, represented by equation 28 and equation 16, I get that the optimallevel of environmental quality will be lower in autocracy if

h >sv

sa· uy

v(yv, ev)

uev(yv, ev)(29)

that is to say, if the level of productivity of the cost devoted to preventexposure to pollution is greater than the amount of environmental qualitythe median voter is willing to give up to become a unit richer, times theratio of the income share of production of the two decisive actors. So, in-come inequality between the dictator and the median voter is not a goodpredictor of the effect of a regime change on the environmental policy un-dertaken by a country, it is environmental inequality that matters, as longas the expenditure for substituting environment with private consumptionis productive (i.e. increase the perception of environmental quality) enough.This result is exacerbated when income inequality between the two politicalactors is big, making autocracy more prone to choose bad environmentalpolicies even for lower levels of h.

Figure 2 represents two possible equilibria that a country may end-upin, depending on the type of regime. Assume that a dictator gets a largershare of output than a potential median voter in a democracy; if the dic-tator can pay to avoid his own exposure to pollution and the median votercannot, the dictator will choose a lower optimal level of environmental qual-ity. Denote by svy − emax and say − emax the private transformation locusbetween income of the potential median voter and the dictator, respectively.At point c the median voter solves his own maximisation problem, equatinghis MRSe,y (that is to say, the amount of money he is willing to spendto improve the environment) with the amount of money he would get byproducing more (this is expressed in terms of e, and is equal to −se). De-note this equilibrium level of environmental quality by e∗v. The dictatoroptimises his utility function with respect to two controls, the level of per-ceived environmental quality and the cost of insurance against exposure topollution. The equilibrium condition is represented by equation 28.

Figure 2

14

The question is: if the dictator can protect himself against pollution bypaying an insurance, will his optimal level of e be lower than the optimallevel of the median voter? The answer is yes, provided that at point dthe amount of income the dictator has to give up to increase environmentalquality by one unit (−∂y/∂e), is greater than his willingness to pay (in termsof income) to increase quality by one unit (MRSe,y).

If this is so, the dictator has the interest to reduce his optimal level ofenvironmental quality such that he can buy the insurance and increase isperceived level of environmental quality. In figure 2, e∗a represents the “real”level of environmental quality chosen by the dictator, while e′ is the “expe-rienced” level of environmental quality. If this is the case, then, democracyis definitely better for the environment.

4 Motivations - Empirical evidence and some casestudies

As the previous section shows, in the presence of environmental inequalitybetween the two decisive political actors in dictatorship and democracy,democracy is likely to be beneficial for the environment, the better theeffect on the environment, the bigger the income inequality between thosetwo citizens.

Provided that this paper is mainly theoretical and this section does notpretend to be a formal test of the results implied by the theoretical model,but only a mere description of what happens in the data, I will show withsome row statistics and simple OLS regressions that democracy is beneficialfor the environment.

15

Looking at some simple raw statistics6, using a sample of 47 transitioncountries, I observe that during the period 1990-2002 the average concen-tration of PM10 recorded during spells of dictatorship is about 1.36 timesbigger than during democratic periods, despite the average level of percapitaGDP is 2.36 times bigger during democracy than during dictatorship spells7.

Figure 4.1

Across the period 1950-2002, the average intensity of CO2 emissionsproduced per unit of GDP is 1.17 times larger in periods of dictatorshipthan in periods of democracy8. Several countries show clearly a decrease inthe intensity of emission per unit of GDP in proximity to the regime shift(see fig. 4.2):a simple fixed-effect regression of variations of percapita CO2 emissions overa constant, variations of percapita income and a dummy for democracyrepresented in equation 30, shows that percapita emissions are consistentlyreduced during democratic periods9:

6Data for CO2 emissions are from Marland, G., T.A. Boden, and R. J. Andres [31], forPM10 concentrations are from World Bank - World Development indicators’ Database,data for percapita GDP are from Maddison [30], and data for democracy are from Prze-worski’s dataset, available at http://politics.as.nyu.edu/object/ przeworskilinks.html

7This average is computed over all 13 years and all over the countries, conditioned toperiods of democracy or dictatorship. The average concentration of PM10 during democ-racy is 69.67906 and during dictatorship is 89.14407. Percapita GDP during democraticperiods is, on average, 4,281.742 against 1,813.18 during dictatorship.

8This average is computed over all the 53 years and over all the countries, conditionedto periods of democracy or dictatorship. The data for periods of democracy show anintensity of CO2 emissions (in Kg of carbon) per unit of income of 0.1269 against 0.1485during periods of dictatorship.

9All the results are significant at 1% level, standard errors in parenthesis

16

Table 2: Regime changes (years)

Przeworski (1950-2002)Country No. Switch to dem. Switch to dict.Albania 1 1992 -Bangladesh 1 1991 -Bolivia 3 1979, 1982 1980Brazil 2 1979 1964Bulgaria 1 1990 -Burundi 2 1993 1996Central Afr. Rep. 1 1993 -Chile 2 1990 1973Colombia 1 1958 -Congo (Brazzav-ille)

2 1992 1997

Cote d’Ivoire 1 2000 -Czechoslovakia∗ 1 1990 -Ecuador 3 1979 1963, 2000El Salvador 1 1984 -Ghana 5 1969, 1979, 1993 1972, 1981Greece∗ 2 1974 1967Guinea-Bissau 1 2000 -Haiti 1 1994 -Hungary 1 1990 -Indonesia 1 1999 -Kenya 1 1998 -Korea Rep.∗ 3 1960, 1988 1961Laos 1 - 1959Lesotho 1 1993 -Madagascar 1 1993 -Malawi 1 1994 -Mali 1 1992 -Mexico 1 2000 -Moldova 1 1996 -Nepal 2 1991 2002Nicaragua 1 1984 -Niger 3 1993, 2000 1996Nigeria 4 1979, 1999 1966, 1983Pakistan 5 1972, 1988 1956, 1977, 1999Panama 3 1852, 1989 1968Peru 7 1956, 1963, 1980, 2001 1962, 1968, 1990Philippines 2 1986 1965Poland 1 1989 -Portugal∗ 1 1976 -Romania 1 1990 -Senegal 1 2000 -Sierra Leone 4 1996, 1998 1967, 1997South Africa 1 1994 -Spain∗ 1 1977 -Sri Lanka 2 1989 1977Venezuela 1 1959 -Zambia 1 1991 -

17

∆CO2it =8.229862

(1.89233)+

0.1170303

(0.0079892)∆GDPit +

−17.3097

(3.356346)Demit (30)

with ∆CO2it denoting variations in the level of percapita CO2 emissionsexpressed in Kg of carbon occurred between t−1 and t for country i, ∆GDPit

variations in the level of percapita income occurred between t − 1 and tfor country i, and Demit is a dummy variable coded 1 during periods ofdemocracy, and 0 otherwise.

The estimated coefficients of equation 30 indicate that emissions are pos-itively correlated to income, and that one dollar increase of GDP requires0.117 Kg more carbon. However, for each year a country has been demo-cratic, we assist to an average decrease in percapita carbon utilisation of17.3 Kg. Graphically, the estimated coefficient related to the dummy fordemocracy represents the average “kink” in the two fitted lines in figure 3.2,before and after the regime shift, for the whole sample of countries.

Figure 4.2 shows, for Colombia, South Africa, Spain and Bulgaria, theactual (dotted) and fitted (line) levels of intensity of emissions (expressed intons of carbon per unit of GDP). The vertical line in each subfigure repre-sents the date of the regime shift. All the four countries have experienced atransition to democracy after long periods of dictatorship. After the regimeshift we observe a reverse pattern for emissions; while before democratisa-tion the tendency is to increase the intensity of CO2 emissions in production,later we observe a decline, which is persistent through time.

18

Figure 4.2

In early democracies, environmental policies may be often out of thepublic agenda. The needs of protecting human rights are instead one ofthe main reason for democratisation. The protection of human rights oftenmanifests as the need of a job for everyone, social reforms aimed at reducingpoverty and pressures for meeting the basic needs for the whole population.In this last category falls the right of everyone to have an healthy life forthe present and future generations, and therefore environmental manage-ment becomes of crucial importance. How a country chooses to achieve thegoal of a better environment however varies between different nations, andsometimes it is a direct consequence of other policies undertaken and maynot be explicitly regulated. In Colombia, for example, the sharp reductionin the intensity of emission is associated with the land reform which wasaimed both at creating jobs and reducing poverty. Under the PresidentCarlos Lleras Restrepo (1966-1970) the Colombian Instutute of AgrarianReform (INCORA) promoted the redistribution of usable land to the peas-ants and unemployed workers in the country, issuing more than 60.000 titlesin 1968 and 1969 alone. As a result of that, after few years the economy wasmore diversified than before, labor productivity was higher, and inequal-ity lower. He also implemented an aggressive and broad program of socialand economic reforms, creating, among others, a national saving fund, aninstitute for the family wellbeing, the institute to protect non-renewableresources, an agency to promote export etc. His successor Misael Pastrana

19

Borrero (1970-1974) carried on the development increasing economic growththrough encouraging housing construction and giving financial incentives tocommercial agriculture. As a result jobs increased as well as wages, and healso promoted the first national environmental legislation in Latin America.

In South Africa the picture is similar. The election of the new govern-ment in 1994 was meant to put in place new policies to promote development,directed mainly at alleviating poverty, creating jobs and meeting the basicneeds for the majority of South Africans. In this last context, it was neces-sary to define a clear policy objectives in the area of environmental qualityand the use of natural resources. The Constitution provided (and provides)a powerful safeguard in shaping future economic and social development inan environmentally sustainable way: it lays down among the fundamentalrights of every citizen that “every person shall have the right to an envi-ronment that is not detrimental to his or her health and wellbeing” andthat “every person shall have the right to access to all information held bythe state or any of its organs at any level of government insofar as suchinformation is required for the protection or exercise of any of his or herrights”.

In Spain the decrease in the intensity of emission coincides with an in-crease in the investments in environmental areas. During the sixties and theearly seventies Spain have had an uncontrolled process of industrialisation(Font and Morata [14]). The institutionalisation of the environmental pol-icy started in the seventies, with the creation in 1971 of the InterministerialCommission for the Environment, but investments in the environmental ar-eas were not significant until 1978 (De Esteban and Lopez Lopez [9]), whenthe new Constitution introduced an article that ascribes as duty to ensurerational use of natural resources, to protect and improve the quality of lifeand to defend or restore the environment.

Lastly, Bulgaria, contrary to other countries, had very bad environmentalconditions and democratisation was the result of (mainly) environmental-ist protests. Bulgaria’s dramatic environmental conditions were inherited bythe inefficient and obsolete technologies used during the communist era. Anyregulation for environmental protection was missing and it was estimatedthat two thirds of Bulgarian suffered of health problems due to pollution,in 1988 the top three causes of death in Bulgaria were cardiovascular ill-nesses, cancer, and respiratory illnesses10. The first demonstration againstthe Communist party in Bulgaria took place in Rousse at the Romanianborder in the north of the country. On february 1988 pram-pushing moth-ers marched through the main street of Rousse protesting that governmentswas doing nothing on behalf of the “international proletariat” about thechemical plant in Girgiu on the other side of the border in Romania, whichfor decades had been belching chlorine pollution in the air in Rousse. A

10U.S. Library of Congress, http://countrystudies.us/bulgaria/29.htm

20

committee to save Rousse was founded and its activities were directed notagainst the regime of Nicolae Ceausescu in Romania but against Bulgariaand Todor Zhivkov. Here, environmental problems were the driving forcetoward democratisation. For what concerns this work, however, the issue isnot whether democracy causes an environmental improvement of whetherenvironmental issues cause democracy. What is important here is that en-vironmental goals can be obtained only through democratic institutions.What comes first is not crucial, it is only that environmental policies wereundertaken after the regime transition to democracy, necessary to achievethose objectives.

This paper contributes with a theoretical model to address the questionof whether democratisation is good for the environment or not, taking intoconsideration the issues of price effects for environmental protection andsubstitutability of the environment with private consumption. This model,moreover, provides a possible explanation of the recent empirical evidence,which sees democracy associated to a more pro-environmental policies.

5 Conclusion

This theoretical model sheds some lights on the possible mechanisms un-derlying the observed empirical evidence of an environmental improvementfollowed by democratisation. In the absence of environmental inequality, ashift of the decision power from the dictator to the median voter may leadto an environmental improvement only if the “relative” (with respect to in-come) marginal willingness to pay for an additional unit of environmentalquality is greater for the median voter than the dictator. This results issimply due to the fact that the dictator, appropriating of a larger share ofglobal production, faces the tradeoff between taking possession of an addi-tional fraction of income or giving up an amount of income and having abetter environment. As long as the fraction of income the dictator can takepossession increases, the opportunity cost to pay for environmental protec-tion increases as well. On the contrary, since the median voter is poorerthan the dictator, his demand for environmental quality will be lower, butalso its opportunity cost to pay for additional environmental protection islower due to the fact that his own share of income is low. The global ef-fect of a regime change from dictatorship to democracy is than ambiguous,and depends primarily on the interaction of the differences of wealth be-tween the two decisive political actors and their marginal willingness to payfor environmental protection. If instead one assumes that the dictator canenjoy a better environment because of his ability to pay in order to avoidexposure to pollution, the basic result of this model is that environmentalquality chosen by the dictator does not depend on utility anymore, but only

21

on his own share of global production he can take possession of, and theproductivity of the expenditure done for reducing the exposure to pollu-tion. Basically, what I called “insurance for exposure to pollution” reducesthe marginal utility from environmental quality and therefore reduces themarginal willingness to pay for environmental protection. As long as thisinsurance’s productivity is higher than the amount of environmental qual-ity the dictator is willing to give up to become one unit richer, then it isworth for the dictator paying. The total amount the dictator will spend insuch an insurance depends on its productivity, stopping when the amountof environmental amenities he can renounce to become marginally richer isexactly equal to the level of productivity of the insurance. The direct effectof this insurance on the level of environmental quality is detrimental: thedictator will find it worthwhile to increase production because doing so hecan maximise his own consumption (or wealth) and pay for reducing his ownexposure to pollution generated by the productive activity. Global produc-tion will be greater (and correspondently, the global level of environmentalamenities will be lower), the greater the productivity of the insurance policy,because the dictator’s perceived environmental quality will be higher.

Concluding, income inequality per se is not a valid indicator to predictwhether a change of the decision power from a dictator to the median voterwill result in a more environmental protection or not. Wealth inequalityinstead plays an important role in conjunction with environmental inequal-ities to explain the recent evidence of beneficial effects of democracy on theenvironmental quality. The explanation for this effect is simple and relieson the assumption that in authoritarian regimes the person who holds thedecision power is likely to be rich and so he can afford - contrary to thedecisive citizen in a democracy which is supposed to be poorer - this in-surance against exposure to pollution. If this is the case, as the empiricalevidence on the distribution of the environmental burden suggests, we caneasily explain why democratisations are associated to reductions in humaninduced emissions.

Some scholar may wonder, however, if this effect may be induced bythe environmental Kuznet Curve, assuming that democratisation occurs atthe turning point. Looking at the statistics of equation 30, and to the rowstatistics related to concentration of PM10, it can be reasonably excludedthat this sharp reduction in the intensity of CO2 emission or concentrationof PM10 was the result of generic development, like technological progressor changes in the composition of GDP or any reason that has been adductedto justify the Environmental Kuznet Curve (EKC). Even assuming that anEKC exists for such a kind of pollution indicators, the average percapitaincome during periods of democracy is far below the Selden and Song’s[43] minimum estimate of the “turning point” for CO2 emissions, whichwas supposed to be around $6,241.00, and for PM10 this turning point hasbeen estimated by Grossman and Krueger [20] in less than $5,000.00 (These

22

estimates are computed using real percapita GDP at constant prices, withreference to 1985 US$. My data for GDP are expressed in 1990 US$, buthaving them expressed in 1985 US$ the average percapita income duringdemocratic spells between 1990 and 2002 is 4,162,79, and 1,914.93 duringdictatorship. Increasing the time of observation from 1950 to 2002, I get thatpercapita GDP in 1985 US$ during democracy is, on average, 4,286.08, andduring dictatorship, 2,405.68, in any case far below the estimated turningpoints). Moreover, as Stern [44] pointed out in one of his recent surveys,any estimation of the EKC’s turning point tend to be larger, the greateris the number of developing and underdeveloped countries included in thesample, which makes the EKC of dubious existence. The fact that intensityof emissions have declined simultaneously to an increase in income, therefore,is not an indicator that the turning point of the EKC has reached, since thereare good reasons to believe to stay in the increasing part of the EKC (so, tothe left of the turning point).

The evidence is therefore supporting the basics of the model. Democrati-sation, in a society where the majority of the population is not self-employedand supply physical labour to the production process, is likely to be bene-ficial for the environment, the bigger the effect, the larger the difference inthe share of income the two decisive citizens can take possession of.

Appendix

This appendix is introduced in support of section 4, and motivates the rea-sons for selecting the econometric model represented by equation 30.

As I already stressed in the text, section 4 does not pretend to be a formaltest of the results implied by the theoretical model, but just a mere descrip-tion of the main evidences that data show. However, a formal motivationsfor the choice of the simple econometric model used in section 4 is due.

First of all, I checked stationarity of the main variables used (percapitaGDP and CO2 emissions). Tests of unit root of percapita GDP and CO2emissions using Levin, Lin and Chu (LLC) test, Im, Pesaran and Shin (IPS)test, and finally ADF and PP-Fisher tests show that the two variables areintergrated of order one. In the absence of cointegration, therefore, a regres-sion of a I(1) variable over another I(1) variable would produce inconsistentestimates of the coefficients (see appendix table A for results of the unit roottests).

In the Engle-Granger approach, cointegration is tested by verifying thatthe residual series generated by the regression of one I(1) variable over an-other I(1) variable is stationary. To verify whether percapita CO2 emissionsand percapita GDP are cointegrated, I use the approach suggested by Kao[25]. The global ADF statistic for Kao residual cointegration test with the

23

Table 3: Summary statistics of the primary variables used for descriptivestatistics

Variable Description Obs mean Std.Dev. min max

CO2 Kg of carbonpercapita (1950-2002)

2255 563.663 834.6266 0 4470

PM10 Concentrations ofparticulate mat-ters, microgramsper cubic meter(1990-2002)

611 72.10575 48.19759 11.9218 274.45

GDP Percapita GDP,1990$

2452 2778.26 2677.37 289.15 16572.83

Dem Dummy forDemocracy(Przeworski)

2234 .3531782 .4780645 0 1

Ineq Household’s in-come inequality(UTIP)

1129 42.00001 6.666231 19.81 62.32

Y ear Year (from 1950to 2002)

2491 1976 15.30013 1950 2002

N 2491

24

null of no cointegration shows a t-stat of 3.499820 with a p-value of 0.0002,so this test strongly suggests that those two series are cointegrated. Mad-dala and Wu [29] combined tests from individual cross-sections (trace andmaximum eigenvalue tests) reject the hypothesis of absence of cointegrationat 1% level, and, as expected, accept the hypothesis that there exists onecointegrating relation.

So, despite the regression of percapita GDP over percapita CO2 emis-sions produces “superconsistent” estimates and models like VAR and VECMcan be implemented, I prefer to regress those variables in differences becausethe coefficient attached to the dummy variable Demit (which denotes peri-ods of democracy) can be directly associated to figure 4.2 because it repre-sents the “average kink” of the panel, that is to say, how the variation ofCO2 emissions decreases as a consequence of democracy, given the growthof income.

Environmental issues became important after the 70’s (Zurn [47]), andconsidering that the period covered by my dataset for CO2 emissions be-gins in 1950 and ends in 2002, one may think of a possible structural break.However, enviromental problems were raised mainly by developed and richeconomies, and considering that the sample I am using is composed mainlyby developing and underdeveloped countries, with few exceptions, I thinkthat the structural break problems can be overlooked without affecting themain result of the estimation. Underdeveloped and developing countriesindeed did not participate until recently (often after the end of my observa-tional period) to international agreements for pollution abatement becausethey claim(ed) a right to grow like the rich and developed countries did in thepast without any external imposition. Moreover, they also raised the prob-lem of “environmental justice”, according to which the burden of emissions’reduction should be born by rich countries, because they are responsible forthe main global pollution problems like greenhouse gases and actually theykeep polluting not to satisfy their needs for survival, but their luxuriouslifestyles (United Nations Framework Convention on Climate Change).

In equation 30, I regressed GDP over CO2 emissions using a panel fixed-effect model, which allows to control for time-invariant differences amongcountries. Due to the fact that differenced variables are stationary (resultsof panel unit root tests on the differenced variables not shown), results ofthis estimation are consistent.

I also check whether the inclusion of potentially important variables inthe regression (i.e. income inequality) would have affected (or nullified)the effect of democracy. Including in the regression variations in incomeinequality11, not only the sign of the coefficients attached to democracy and

11Data for inequality are from EHII dataset of the University of Texas Inequality Project

25

GDP do not vary, but variations in income inequality do not appear to bestatistically significant, although negative. This is probably due to the factthat inequality takes more time to give its effect on pollution abatement.

Considering the sample I am using - mainly composed by underdevelopedand developing countries - data on possible important variables that maybias the estimates because they are not included in the regression, are notalways available. Education is, for instance, one of these variables. It isreasonable however to believe that - similarly to what happens for incomeinequality - its effect can be seen more in the long than in the short time,and that the inclusion of such a variable would produce a coefficient that isstatistically not significant once introduced differentiated in equation 30.

Moreover, I check whether the results are robust with respect to differentsubsets of countries. I found that results are less stable for countries thatexperience more than one regime switch. Bolivia, Brazil, Chile, Burundi,Congo, Ecuador, Ghana, Greece, South Korea, Laos, Nepal, Niger, Nigeria,Pakistan, Panama, Peru, Philippines, Sierra Leone and Sri Lanka are thecountries which - during the period 1950 through 2002, had at least tworegime changes, with Peru having 7 regime switch, Pakistan and Ghana 5,Sierra Leone and Nigeria 4 etc. A high political instability is then a causefor the implementation of less stringent environmental policies. Results ofthe estimate for those “highly unstable” countries show indeed that emis-sions depends exclusively on production, because the coefficient attached todemocracy becomes statistically not significant.

Appendix Table A - Panel Unit root test summaryExogenous variables: individual effect and individual trendLags included: 1

Series : CO2Method Stat. Prob** Cross Section Obs

Null: Common unit root processLevin, Lin & Chu 1.58923 0.9440 46 2164

Null: Individual unit root processIm, Pesaran and Shin W-Stat 3.77347 0.9999 46 2164ADF-Fisher Chi-Square 55.8579 0.9989 46 2164PP-Fisher Chi-Square 106.053 0.1501 46 2210

Series : GDPMethod Stat. Prob** Cross Section Obs

Null: Common unit root processLevin, Lin & Chu -0.20043 0.4206 47 2357

Null: Individual unit root processIm, Pesaran and Shin W-Stat 3.46602 0.9997 47 2357ADF-Fisher Chi-Square 66.3949 0.9862 47 2357PP-Fisher Chi-Square 58.8221 0.9983 47 2404**Probabilities for Fisher tests are computed using an asymptotic Chi-Square distribution.All other tests assume asymptotic normality

(UTIP), available at http://utip.gov.utexas.edu/data.html

26

References

[1] Acemoglu D., Robinson J.A. and Verdier T., 2004. Kleptocracy andDivide-and-Rule: A Model of Personal Rule. Journal of the EuropeanEconomic Association, Vol. 2, No. 2/3, Papers and Proceedings of theEighteenth Annual Congress of the European Economic Association,pp. 162-192.

[2] Agrawal, B., 1992. The Gender and Environment Debate: Lessons fromIndia, Feminist Studies 18, pp. 119-158.

[3] Barrett S. and Graddy K., 2000. Freedom, Growth and the Environ-ment. Environment and Development Economics 5, pp. 433-456.

[4] Bhattarai M., and Hammig, M., 2001. Institutions and the Environ-mental Kuznets Curve for Deforestation: a Cross Country Analysis forLatin America, Africa and Asia. World Development 29, pp. 995-1010.

[5] Boyce J., 2007. Inequality and Environmental Protection, in Inequal-ity, Cooperation and Environmental Sustainability, Princeton Univer-sity Press.

[6] Congleton, Roger D., 1992. Political institutions and pollution con-trol,The Review of Economics and Statistics 74 (3), pp. 412-421

[7] Costa D., 1997. Less of a luxury: the rise of recreation since 1888.National Bureau of Economics Research, Working Paper no. 6054.

[8] Dasgupta, P. and Maler, K. G., 1995. Poverty, Institutions, and theEnvironmental Resource-Base, in Behrman, J., Srinivaan, T.N. (eds.),Handbook of Development Economics, vol. 3A. Elsevier Science, Ams-terdam. (Chapter 39).

[9] De Esteban Alonso and A. Lopez Lopez, 1993. Environmental Policyin: “Spain and EC Evaluated”, London.

[10] Desai, Uday, 1998. Environment, Economic Growth, and Government,Ecological Policy and Politics in Developing Countries. Albany, NY.State University Press of New York Press, 1-45.

[11] Dinan T., and Lim D., 2000. Who Gains and Who Pays Under CarbonAllowance Trading? Washington, D.C.: Congressional Budget Office.

[12] Downs A., 1957. An Economic Theory of Political Action in a Democ-racy. Journal of Political Economy 65, pp. 135-150.

[13] Farzin, Y. Hossein, and Craig A. Bond., 2006. Democracy and environ-mental quality. Journal of Development Economics 81, pp. 213-235.

27

[14] Font N. and F. Morata, 1998. Spain: Environmental Policy and PublicAdministration. A Marriage of Convenience Officiated by the EU? in:Hanf K. and A.I. Jansen, “Governance and the Environment in WesternEurope; Politics, Policy and Administration”, Longman.

[15] Foster, Sheila, 1998. Justice from the Ground-Up: Distributive Inequal-ities, Grassroots Resistance, and the Transformative Politics of the En-vironmental Justice Movement. California Law Review 86, pp. 775-841.

[16] Fredriksson and Wollscheid, 2007. Democratic institutions versus auto-cratic regimes: the case of enviromental policy Public Choice no. 130,pp. 381-393.

[17] Gallagher K. P., and Thacker S. C., 2008. Democracy, Income andEnvironmental Quality, Political Economy Research Institute (PERI)Working Paper No. 164, University of Massachusetts Amhrest.

[18] Gleditsch, Nils P., and Bjorn Sverdrup, 2003. Democracy and the Envi-ronment. In Human Security and the Environment: International Com-parisons, edited by Edward Paper and Michael Redclift. London: Elgar.

[19] Gray and Shadbegian, 2004. Optimal Pollution Abatement - WhoseBenefits Matter, and How Much?, Journal of Environmental Economicsand Management 47, pp. 518-534.

[20] Grossman G.M. and A.B. Krueger, 1995. Economic Growth and theEnvironment, The Quarterly Journal of Economics 110, pp. 353-377.

[21] Harbaugh W.T., Arik Levinson, and David Molloy Wilson, 2002.Re-examining the Empirical Evidence for an Environmental KuznetsCurve. The Review of Economics and Statistics 84, pp. 541-551.

[22] Hardin, Garret, 1968. The Tragedy of the Commons. Science 162, pp.1243-1248.

[23] Heilbroner, Robert, 1972. Growth and Survival. Foreign Affairs 51(1),pp. 139-153.

[24] Hokby S., Soderqvist T., 2001. Elasticities of demand and willingnessto pay for environmental services in Sweden. Beijer discussion paperseries No. 137, Beijer International Institute of Ecological Economics,The Royal Swedish Academy of Sciences, Stockholm.

[25] Kao Chihwa, 1999. Spurious Regression and Residual-Based Tests forCointegration in Panel Data, Journal of Econometrics 90, pp. 1-44.

[26] Kristrom B., Riera P., 1996. Is The income elasticities of environmentalimprovements less than one?. Environmental and Resource Economics7, pp. 45-55.

28

[27] Larsen R.E., 2001. Revealing demand for nature experiences using pur-chase data on equipment and lodging. Discussion paper No. 305, Statis-tic Norway, Research Department.

[28] Leslie, Winsome J., 1987. The World Bank and Structural Transforma-tionin Developing Countries:The Case of Zaire. Boulder, CO: LynneRienner Publishers.

[29] Maddala G.S. and Shaowen Wu, 1999. A Comparative Study of Unitroot test with Panel Data and a New Simple Test, Oxford Bulletin ofEconomics and Statistics 61, 631-652.

[30] Maddison Angus, 2009. Statistics on World Population, GDP andPer Capita GDP, 1-2006 AD, updated March 2009, available athttp://www.ggdc.net/maddison/

[31] Marland, G., T.A. Boden, and R. J. Andres, 2008. Global, Regional,and National CO2 Emissions. In “Trends: A Compendium of Dataon Global Change”. Carbon Dioxide Information Analysis Center, OakRidge National Laboratory, U.S. Department of Energy, Oak Ridge,Tenn.

[32] Martini C., and Tiezzi S., 2010. Is the Environment a Luxury? AnEmpirical Investigation using Revealed Preferences and Household Pro-duction. Quaderni del Dipartimento di Economia Politica, n. 599.

[33] Miles D., Pereyra A., and Rossi M., 2002. The Consistent estimation ofincome elasticity of environmental amenities in Uruguay. Stud. Econ.no. 17.

[34] Moya Pons, Frank, 1995. The Dominican Republic: A National History.Princeton, NJ: MarkusWeiner.

[35] Neumayer E., 2002. Do democracies exhibit stronger international en-vironmental commitment? A cross-country analysis. Journal of PeaceResearch no. 39, pp. 139-164.

[36] Olson M., 1965. The Logic of Collective Action, Public Goods and theTheory of Groups, Harvard University Press.

[37] Ophuls W., 1974. The Scarcity Society. Harper’s 248(1487), pp. 47-52.

[38] Pastor M. Jr, Jim Sadd and John Hipp, 2001. Which Came First?Toxic Facilities, Minority Move-in, and Environmental Justice. Journalof Urban Affairs 23, pp. 1-21.

[39] Payne R.A., 1995. Freedom and the Environment, Journal of Democracy6, pp. 41-55.

29

[40] Pereyra A., and Rossi M., 1998. Are environmental goods a luxury?Working paper no. 13, University of Uruguay.

[41] Schultz, C. B., and T. R. Crockett, 1990 Economic Development,Democratization, and Environmental Protection in Eastern Europe.Boston College Environmental Affairs Law Review 18, pp.5384.

[42] Scruggs L., 1998. Political and Economic Inequality and the Environ-ment, Ecological Economics 26, pp. 259-275.

[43] Selden T.M., and D. Song, 1994. Environmental Quality and Develop-ment: Is there a Kuznets Curve for Air Pollution Emissions? Journalof Environmental Economics and Management 27, pp. 147-162.

[44] Stern, David, 2004. The Rise and Fall of the Environmental KuznetsCurve. World Development 32, pp. 1419-1439.

[45] Torras M., and Boyce J. K., 1998. Income, Inequality and Pollution:a Reassessment of the Environmental Kuznets Curve. Ecological Eco-nomics 25, pp. 147-160.

[46] Veblen T., 1899/1934. The Theory of the Leisure Class. New York,Modern Library.

[47] Zurn M., 1998. The Rise of International Environmental Politics: aReview of Current Research. World Politics 50, pp. 617-649.

30