INTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVE (EKC): A LATIN AMERICAN STUDY

8/6/2019 INTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVE (EKC): A LATIN AMERICAN STUDY

1/54

END OF MASTER PROJECTEND OF MASTER PROJECTEND OF MASTER PROJECTEND OF MASTER PROJECT

INTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVEINTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVEINTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVEINTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVE

(EKC): A LATIN(EKC): A LATIN(EKC): A LATIN(EKC): A LATIN AMERICANAMERICANAMERICANAMERICAN STUDYSTUDYSTUDYSTUDY

Italo Raul A. ArbulItalo Raul A. ArbulItalo Raul A. ArbulItalo Raul A. Arbul VillanuevaVillanuevaVillanuevaVillanueva

1111

MASTER IN TOURISM AND ENVIRONMENTAL

ECONOMICS

(MTEE)

End of Master Project

INTRODUCING INSTITUTIONAL VARIABLES IN

THE ENVIRONMENTAL KUZNETS CURVE (EKC): A

LATIN AMERICAN STUDY

By

Italo Arbul Villanueva

June 2010


2/54





2222

Abstract

Several studies have examined the relationship between environmental degradation and

per capita income. However, most of them did not take into account institutional quality

and just focused on macroeconomic factors. The purpose of this paper is to fill this gap

in the literature by investigating the effects on the Environmental Kuznets Curve (EKC)

when institutional quality variables are introduced, especially those related to corruption

and rent-seeking behavior.

This study considers 18 Latin American economies and panel data for 19982005. We

employed a standard reduced-form modeling approach with pool estimation. In order to

introduce the heterogeneity of the different countries three different models were used.

The first model corresponds to the basic Environmental Kuznets Curve (Basic Model),

the second one introduced a series of additional economic variables into the previous

formulation (Extended Model N 1), and finally, the third one introduced institutional

variables into the previous formulation (Extended Model N 2). The robustness of the

estimation is checked to ensure there were no serial correlation and heteroskedasticity in

the model.

The expected results from this investigation lead us to support the EKC hypothesis while

confirming the importance of improvements in political institutions and governance for

better environmental performances in the region.


3/54





3333

Corruption is worse than prostitution.The latter might endanger the morals of anindividual, the former invariably endangers

the morals of the entire country

Karl Krauss.


4/54





4444

Acknowledgements

First of all, I would like to take this opportunity to thank my family for all their love and

support during this time. They are the greatest gift God may have ever given to me and I

want to dedicate this effort to them.

I would also like to thank my tutor, Mr. William Nilsson, for all the precious time he

devoted to this work. I feel really grateful for his invaluable guidelines and guideness

which helped me to achieve the goals of this research.

Finally, I would like to thank the same way my friends and colleagues for their valuable

comments, discussions and clarifications which improved the quality of this research.

This paper was written at the University of Balearic Islands (UIB) during the years 2009

and 2010 thank to a scholarship award from Fundacin Carolina in association with

Fundacin Barcel. I would like to mention how grateful I am to all this institutions for

their trust and hospitality.


5/54





5555

INDEX

1. INTRODUCTION.................................................................................................. 6

2. LITERATURE REVIEW ....................................................................................... 9

2.1. Theoretical Models Regarding Economic Growth and Environmental

Degradation................................................................................................................ 9

2.1.1. The IPAT Model............................................................................................... 9

2.1.2. The Environmental Kuznets Curve .............................................................. 122.2. The Importance of Institutions to Solve Environmental Problems............ 14

2.2.1. Corruption as an Institutional Problem ....................................................... 16

2.2.2. Some theoretical Approaches Regarding to Corruption and the

Environment............................................................................................................. 17

3. METHODOLOGY............................................................................................... 21

3.1. Econometric Technique ............................................................................... 23

4. DATA ................................................................................................................. 27

5. EMPIRICAL FINDING........................................................................................ 31

5.1. Statistical Contrasts: Panel Data Homogeneity.......................................... 37

5.2. Graphical Analysis of the Environmental Kuznets Curve.......................... 39

5.3. The Impact of Institutions on Environmental Quality................................. 43

6. CONCLUSIONS................................................................................................. 47

7. REFERENCES................................................................................................... 50


6/54





6666

1. INTRODUCTION

A key policy objective of international efforts to mitigate the adverse effects of global

climate change is the reduction of global CO2 emissions. The success of these efforts

depends on a large degree of commitment by major CO2 producing nations to meet

global emissions targets. This is the reason why it is so important to understand the

main variables which impact over the CO2 emissions.

Over the last years several studies used the Environmental Kuznets Curve (EKC)

hypothesis to evaluate the impact of several economic variables on environmental

indicators. The EKC is a hypothesized relationship between several indicators of

environmental degradation and per capita income. This model states that in the early

stages of economic growth, degradation and pollution increases, but after achieving a

certain level of per capita income this trend reverses (showing an inverted U shape).

The EKC is a reduced form relationship, in which the level of pollution is modeled as a

function of per capita income without specifying the links between both of them.

Grossman and Krueger (1995) characterize these missing links as environmental

regulations, technology and industrial composition1.

In this sense, as mentioned before, part of the links is related to regulations of the

economic environment. It is clear that countries with diverse political systems use their

natural resources in different manners and this fact suggests that systems of governance

have important effects on resource use (and on environmental damage).

According to Deacon and Mueller (2004) there are four ways in which a nations political

system is linked to the way its natural resources are used:

1For further references about this point see Torras, M., Boyce, J.K. (1998). Income, inequality, and pollution: A

reassessment of the Environmental Kuznets Curve.


7/54





7777

(1) When property rights to resources are weak, competition to acquire them can be

wasteful and characterized by rent-seeking and violent conflict. The link to

political systems comes about because ownership claims are most likely to be

weak or ambiguous in countries where the rule of law is not well-established.

(2) When a countrys political system is unstable or non-representative, the

individuals claim to a resource stocks future return can be rendered insecure.

This reduces the payoff to natural resource conservation, leading to faster

depletion of resource stocks.

(3) When a countrys natural resources are capable of generating significant rents,

but institutions of democratic governance and the rule of law are not well-

established, corruption by government officials responsible for resource

management can encourage rent-seeking behavior, dissipating the benefits

those resources would otherwise confer.

(4) The mix of private vs. public good outputs produced by a nations natural

resources may be affected by its political system. When a countrys government

does not represent the interests of the entire population, but rather acts on behalf

of a select group, the use of resource stocks to provide public good amenitiesmay be under-emphasized.

Creating property rights that provide incentives for efficient use of natural resources, and

adapting these rights to changes in the conditions, requires that economic agents are

able to cooperate and promote a State that is able to enforce rules in order to solve

market failures and avoid opportunistic behavior. The ability to cooperate, promote and

enforce rules, however, depends on the political institutions that determine natural

resource use.

This previous ideas will provide us the starting point for this study. In particular, the main

objective of this work is to understand and measure the possible impacts of changes in


8/54





8888

institutional quality indices and income on the environment using the EKC hypothesis for

a sample of 18 Latin American countries.

We selected this region because of the important contribution that natural resources and

environmental services have in the productive structure of these countries. Furthermore,

in this region several institutional performance indicators are considered fairly low

compared to developed countries. Both characteristics make this region one of the most

interesting locations to analyze the potential impact of institutional improvements in

environmental quality.

In this sense, new institutional indexes related to democratic behavior constructed by theWorld Bank were introduced in order to measure the impact of institutional qualities on

the environmental impact of the air. Previous studies have attempted to measure the

impact of democracies on economic behavior; however, these studies focused their

attention on measuring just a unique qualitative variable which is a democratic index2 but

few attention was given on the performance and effectiveness of these democracies (for

example, a democratic country could have high level of perceived corruption which

harms the government effectiveness even though political rights are not harmed).

Section 2 of this paper briefly summarizes the theory behind the EKC and how does

rent-seeking behavior allocates resources in a different way than the social optimum

desired level. Section 3 reviews the methodology used in our empirical analysis, and

Section 4 presents the data sources and variables required to achieve the goals of this

research. Section 5 shows the main empirical results and finally, section 6 and 7 shows

the main conclusions of the research and the references used.

2Most of these studies use the Freedom House index which measure if a country is more or less democratic following just

the perception on Civil Liberties and Political Rights inside the country. For additional information about these studies seeNeumayer, E. (2002) and Culas, R. (2007).


9/54





9999

2. LITERATURE REVIEW

The main purpose of this section is to survey the relevant theoretical approaches that will

guide our empirical analysis. In this sense, this chapter will first explore the main theory

regarding the impact of human economic activity over environmental degradation and,

later, on some principal-agent models in which the presence of corruption and rent-

seeking activities arise and determines the divergences between the social desirable

level of environmental quality and the real outcome.

2.1. Theoretical Models Regarding Economic Growth and Environmental

Degradation

The relationship between environmental degradation and economic growth has received

increasing attention in recent years due to evident impacts of climate change on human

economic activities and quality of life.

Many studies have clarified that the main impacts of economic activities on the

environment are part of the so called long-term effects, in this sense, some

macroeconomic models were developed to analyze the main impacts and to consider

alternative policies in an adjustment process. The following lines will be devoted to

explain a brief summary of the most important models which are the IPAT Model and the

Environmental Kuznets Curve (EKC).

2.1.1. The IPAT Model

There are many studies related to environmental impact of economic growth, but as

Gans and Jst (2005) mentioned, one of the first of all decomposition studies is an


10/54





10101010

identity introduced by Ehrlich and Holdren3. They describe the environmental impact of

an economic system by the following equation:

I P A T

In this expression I denotes the environmental impact; P represents the population

size, A stands for affluence, and T for the state of technology applied.

For empirical investigations, it must be indicated by which observable variables the

environmental impact, the affluence and the state of technology should be measured.

Normally emissions (E) were used as an indicator for the environmental impact, the

affluence was measured by per capita gross domestic product (Y/P), and the state of

technology by the emissions per unit of gross domestic product (E/Y).

If we take the logarithm of the previous equation and the derivatives with respect to time

the authors get that the average annual relative change of emissions is assigned to the

sum of the average annual change of emissions per unit of gross domestic product, the

average annual change of per capita gross domestic product, and the average annual

change of population size, respectively.

3Ehrlich, P., J. Holdren (1971). Impact of population growth.


11/54





11111111

This identity has been applied quite frequently in order to describe the importance of

different factors determining environmental damage. The analysis was normally carried

out on different levels of aggregation, i.e. for nations, regions or for the whole world.

However, empirical application of these approaches depends crucially on several

assumptions:

(1) In empirical applications the components of the IPAT equation are specified such

that it is an identity. In these applications the term T (technology is the residual of

an accounting identity).

(2) The development of the variables on the right hand side is independent of each

other.

(3) No other factor than affluence, technology and population determine the

environmental impact.

(4) The change of the variables during the time horizon captured by a study could be

described by an exponential function

The previous criticism of the use of decomposition analysis is in a striking contrast to the

use of this approach, in particular in ecological economic investigations. There are

numerous studies which try to identify the driving forces of the environmental impact of

economic systems on the basis of the IPAT identity. However, the use of the IPAT

identity is often criticized because of many conceptual problems, in particular when used

for empirical investigations.


12/54





12121212

2.1.2. The Environmental Kuznets Curve

Even though the IPAT model made an important contribution to the literature, its

theoretical basis still led some gaps that several years later would lead to the

development of the so called Environmental Kuznets Curve (EKC).

Grossman and Krueger (1995), among others, found that for a number of environmental

variables, the relationship between per capita income and environmental degradation

takes an inverted U shaped form, this means that environmental quality initially worsens

but ultimately improves with income. This apparent empirical relationship has been

called the Environmental Kuznets Curve because of its similarity to the relationshipbetween per capita income and income inequality first suggested by Simon Kuznets in

19554.

Source: Tisdell

4 For further references see Anand, S. and Kanbur, S. (1993). The Kuznets process and the inequality-development

relationship.


13/54





13131313

Corresponding to the early stage of economic growth, the awareness of environmental

problems is low or negligible and environment friendly technologies are not available.

Environmental degradation increases with growing income up to a threshold level

beyond which environmental quality improves with higher per capita income. This

relationship can be shown by an inverted U shaped curve. It is described as the EKC

following the observation of Grossman.

As Stern (2004) mentioned, the relationship between per capita income and

environmental quality depends on scale, composition and technology effects. If the

pollution intensity of aggregate output was fairly constant across countries, we would

expect environmental quality to worsen with income, as greater output generates morepollution.

This EKC hypothesis is intended to represent a long-term relationship between

environmental impact and income. As economic development accelerates with the

intensification of agriculture and other resource extraction, at the take-off stage, the rate

of resource depletion begins to exceed the rate of resource regeneration, and waste

generation increases in quantity and toxicity (scale effect).

At higher levels of development, structural change towards information-intensive

industries and services, coupled with increased environmental awareness, enforcement

of environmental regulations, better technology and higher environmental expenditures,

results in a gradual decline of environmental degradation. As income moves beyond the

EKC turning point, it is assumed that transition to improving environmental quality starts.

Thus, it could be a depiction of the natural process of economic development from a

clean agrarian economy to a polluting industrial economy, and, finally, to a clean service

economy.

On the other hand, environmental quality could improve with income if this scale effect

were eclipsed by the combination of the two other effects. With increasing per capita


14/54





14141414

income, the composition of output shifts among sectors which differ in their pollution

intensity of output. For instance, the service sector may grow relative to the

manufacturing sector. This effect (composition effect) can change the pollution intensity

of output. Furthermore, several sectors of the economy may adopt less polluting

technologies, either because of market driven technological advance or government

regulation (technological effect).

Grossman and Krueger (1995) likewise speculated that the strongest link between

income and pollution is, in fact, via an induced policy response and these policies are in

turn induced by popular demand: As nations or regions experience greater prosperity,

their citizens demand that more attention should be paid to the noneconomic aspects oftheir living conditions. The richer countries, which tend to have relatively cleaner urban

air and relatively cleaner river basins, also have relatively more stringent environmental

standards and stricter enforcement of their environmental laws than the middle-income

and poorer countries.

The public-good character of environmental quality means that effective demand

requires solutions to a market failure. Implicit in such policy-based explanations of the

Environmental Kuznets Curve, then, there is a simple theory of induced innovation: As

per capita income rises, societies improve their abilities to redress market failure. If this

is empirically true, we can ask what it is about higher-income societies that facilitate

solutions to market failures.

2.2. The Importance of Institutions to Solve Environmental Problems

Payne (1995) has probably provided part of the most comprehensive theoretical treatise

in favor of a positive impact of democracy on the environment.

Payne argued that in democracies citizens are better informed about environmental


15/54





15151515

problems (freedom of press) and can better express their environmental concerns and

demands (freedom of speech), which will facilitate an organization of environmental

interests (freedom of association), which will in turn increase pressure on policy

entrepreneurs operating in a competitive political system to respond positively to these

demands (freedom of vote), both domestically as well as via international cooperation.

On the other hand, in non-democratic systems, governments are likely to restrict the

access of their population to information, restrict the voicing of concerns and demands,

restrict the organization of interests and isolate them-selves from the citizens

preferences.

In other words, in democracies if citizens are concerned about environmental problems,

this will eventually require policymakers to exhibit stronger environmental commitment to

address these concerns and honors the demand for environmental protection measures.

Democracies with emphasis on private property rights and individual liberty provide the

opportunities for individuals and businesses to make full use of their potential to expand

production and consumption, which, if not sufficiently counteracted by environmental

regulation, will increase pressure on the environment.

However, while a good theoretical case can be made for a positive link between

democracy and environment, there are a number of considerations pointing in the

opposite direction5. The link between democracy and environment is therefore a

complex one and there is still debate about it.

5For further references see Leff (1964).


16/54





16161616

2.2.1. Corruption as an Institutional Problem

All political regimens have faced corruption, even the most ancient ones. In the case of

democracies this problem takes increasing importance since it harms all the positive

effects of this political system. In this sense, in the presence of corruption policymakers

and public officers6 take actions which are different from the social commitment they are

working for.

For the purposes of this discussion, corruption in natural resource and environmental

management could be defined using the definition of Robbins (2000) as the use or

overuse of community natural resources with the consent of a state agent by those notlegally entitled. It is the extension of existing non-economic relationships (family,

friendship, and other socially obligating relations) to determine access to these use

rights through normative systems of expected exchange.

It is expected that the more regulated an economy is and the larger the amount of

resources administered by the state are, the higher will be the potential rents in the

hands of public officials related to corruption. In this sense, public sector corruption

serves the private interests of bureaucrats and criminals by taking away from citizens

their rights to live in a clean environment, misallocating environmental resources, and

diverting funds from conservation and preservation.

Therefore, the main consequences of this higher incentives means that corruption leads

to ecologically unsustainable resource use and this is the reason why we should take

into account the reduction of corruption activities as an alternative to reduce the

negatives externalities produced by this inefficient resource allocation.

6Those in charge of the enforcement of the policies.


17/54





17171717

2.2.2. Some theoretical Approaches Regarding to Corruption and the

Environment

Fredriksson7 in his paper of 1997 established a model in which a producer lobby

attempts to influence the incumbent governments environmental policy by offering

prospective bribes.

In this model, the incumbent weighs bribes (established as political contributions) and

the aggregate social welfare derived from the outcomes of environmental policies8. It is

important to mention that in this model, bribes are explicitly given in order to influence

government policy and not elections.

The model examines the incentives of an incumbent government and a single producer

lobby group in a three-stage model.

First stage: The lobby group offers the incumbent government a bribe schedule.

The size of the promised bribe depends on the attractiveness of the policy

chosen by the government in the second stage. The lobby must also consider

the exogenous probability that the policy is indeed implemented.

Second stage: The government selects its optimal environmental policy and

collects from the lobby group the bribe associated with its policy choice.

Third stage: Is related to the implementation stage. It is important to mention that

with some exogenous probability the incumbent faces a political crisis or

challenge which leads to a loss of power at the beginning of this period. Note

that if the incumbent government fails to remain in power throughout all the

stages, the lobby groups bribes give no return9.

7See Fredriksson (2003). Political instability, corruption and policy formation: The case of environmental policy.

8 In this model is argued that this weighing factor serves as a useful measure of the level of corruption in the political

system9Unless the new government implements the policy promised by its predecessor without being committed to it.


18/54





18181818

The predictions that emerge are that the effect of corruption on the stringency of

environmental policy is conditional on the degree of political instability. Corruption

reduces the stringency of environmental regulations, but the effect is reduced as the

degree of political instability increase, therefore the incentive to offer a bribe is reduced

when its expected return falls.

An increase in political instability has two opposing partial effects. First, bribery becomes

less attractive for the producer lobby because the likelihood that the government remains

in office throughout the policy implementation stage is reduced, and thus the bribe

becomes less likely to pay off. This effect of instability is particularly pronounced whenthe level of corruption is high.

However, this previously mentioned effect is counterbalanced by the fact that the

government now sees bribes as relatively more attractive. The government is now less

likely to be in office during the policy implementation stage, and thus the probability is

lower that it will derive utility from its own policy choices. This effect of instability is

strongest when the degree of corruption is low. The net impact of political instability on

environmental policy thus depends on the level of corruption.

Another important contribution to the theoretical approaches regarding to the impact of

corruption (or rent-seeking behavior) in environmental outcomes was introduced by

Lpez and Mitra (2000) that proposed an extension of Downs work were it is considered

that the only goal of political parties is to repeat the rewards of holding office (win the

elections). Under the assumption of rent-seeking behavior, then it would be correct to

assume that the government maximizes a function which depends on its probability of

being re-elected as well as on rents. Therefore, the welfare function could be defined

as:


19/54





19191919

Where is the probability of being re-elected, c is the lobby payments or rents

accruing to the government, and a is a coefficient associated with the degree of

corruptibility of the government or a measure of the importance that it attaches to lobby

payments.

We assume that is linear and increasing in social welfare for the relevant range of

social welfare. We also assume that elections reveal public preferences correctly.

Therefore, we can express this idea with the following formulation:

Where is the social welfare function, F is the national product or net revenue

function, x is the amount of pollution which is a variable considered as a factor of

production, and t is a growth factor such as human or physical capital that is

increasing through time.

Social welfare is increasing in F( ) and decreasing in x. Thus, apart from being a factor

of production, x has a direct negative effect on the social welfare function 10.

As a conclusion for this model, the authors state the following proposition for a

cooperative scenario:

If corruption takes the form of cooperative government-private sector interactions, then:

(i) pollution levels will be above the socially optimal levels for any level of income; (ii) an

inverted-U-shaped relationship between income and pollution will exist; (iii) the turning

point of the inverted U-shaped curve will occur at a higher per capita income (and higher

pollution level) than in the socially optimal equilibrium.

10This model assume that the degree of corruptibility of the government affects its probability of re-election only through

its effect on welfare.


20/54





20202020

On the other hand, in the case of a non-cooperative scenarios the authors state the

following proposition:

If government corruption takes the form of a non-cooperative Stackelberg interaction

between the government and the firm, with the latter as a leader, and payment functions

are linear in output, then: (i) pollution is always above the social optimum for any level of

per capita income; (ii) a turning point in the pollution per capita income relationship

always exists as long as a turning point exists in the socially optimal pollution-income

relationship, but such a point is likely to occur at a higher per capita income (and higher

pollution level) than the socially optimal one.

It is striking that, despite the contrast between the assumed behavior of firms and

government that these two scenarios imply, the results derived are highly consistent.

This means that irrespective of the type of interaction between the firm and the

government, for any level of per capita income, pollution levels are always above the

socially optimal level when corruption exists.

Finally, the authors argued that unless economic growth process brings about a rapid

reduction of corruption in developing countries, pollution will remain much higher in these

countries than the levels reached in currently developed ones even when their per capita

incomes were comparable.


21/54





21212121

3. METHODOLOGY

As was mentioned before, the main goal of this work is to test if institutional variables

that measure the performance of those governments which are perceived as less

connected with society needs and perceived as more corrupt tend to harm more the

environment.

Throughout this section seeks to review an aspect of econometrics, specification and

estimation of models with panel data. The econometric methodology of panel data has

an important tradition and experience in economic analysis.

The article published by Balestra and Nerlove (1966) is considered by many authors as

the first point of reference in this topic. However, the fact that in recent years it has been

possible to obtain statistics that combine cross-sectional data observed for a time period

of considerable size, this is one of the reasons that justify the increase in researches and

economic studies, published in professional journals, which use this methodology.

When speaking of panel data, generally refers to a sample of observations on a set of

operators (i = 1,2,3, ..., N) over (t = 1,2, 3, ..., T) points in time. The great difference in

the application of this methodology using macroeconomic data is that the sizes of N and

T are similar, while in the application for microeconometric models, it is expected that the

number of observations around the cross-section (N) should be considerably higher than

the temporal observations (T).

Comparing the information provided by a panel with the information given by only cross-

sectional or time series sample, some differences can be found that help to identify the

main advantages that, from the econometric modeling point of view, offers the use of this

information combined.


22/54





22222222

The first advantage of panel data is that it provides more information, more variability,

less collinearity among variables, more degrees of freedom and more efficiency; all of

these elements are the main limitations of many studies which used cross-sectional or

temporal information only.

In the case of aggregate time series is obvious the problem that generates the high

correlation between independent variables; this problem can be avoided (or at least

decrease substantially) when adding new variability related to cross-section observations

which means that we use a larger sample for the estimation. Similarly, in the case of

time series analysis, if the time dimension is small, the dynamic relationships that include

delays in endogenous and exogenous variables results in an unreliable estimate, and inmany cases, results could be useless. Therefore, the information provided by the cross-

sectional observations (T x N) will allow better estimation of the parameters, even when

the time dimension is short. This previous explanation applies perfectly to our empirical

work in which is being taken measurements of CO2 emissions for a number of countries

in Latin America, however, the data collection of this emissions was made in recent

years, making clear the limitations of the lack of an extensive series for each country.

A second advantage of this methodology is related to the fact that panel data suggest

that individuals, families, businesses, states or regions analyzed are heterogeneous. The

cross-sectional data only (or just cannot) control the temporal heterogeneity, this means

that there is the risk of not taking into account all the information and the final results

could be a biased parameter.

A third advantage of panel data methodology is that it allows a more precise analysis of

the dynamics of adjustment of economic variables. The panel data provide better

information to study the duration of economic phenomena such as unemployment,

poverty, etc and, more importantly, if the time dimension is large enough, it could beused to quantify the rate of adjustment of economic policy measures taken to improve


23/54


24/54





24242424

Several previous empirical studies use cross-country data to measure, among other

factors, the relationship between income and environmental degradation. These studies

use models of the form:

Where:

: Dependent variable.

: Notation to represent a given country.

: Notation to represent a given explanatory variable.

: Constant term.

: Vector of coefficients.

: Matrix of explanatory variables.

: Random error term.

These models implicitly assumed that a common structure exists across all countries

(each country denoted by the letter i); that is, that the effect on emissions of changes in

any given explanatory variable will be the same for every country. In other words, every

country will have the same and . More intuitively, an extra unit increase in variable Xj

in Per12 will have exactly the same effect on CO2 emissions as an extra unit increase in

the same variable (Xj) in Guatemala.

Some papers attempt to avoid the weakness of cross-sectional studies by using panel

data instead and include in the set of explanatory variables those that could capture the

12For example, GDP per capita.


25/54





25252525

heterogeneity, or by employing fixed or random effects methods to estimate models. In

this sense, given that the dataset used in this research have a reduce amount of

temporal dimensions (T), long run effects will not be capture in an adequate way by fixed

and random effect models. Therefore, the use of panel data (pool estimation) with the

introduction of specific economic variables (additional explanatory variables) is the best

way to model the differences in behavior across countries13.

A general formulation of the pool model can be expressed as:

Where:

: Dependent variable.

: Notation to represent a given country.

: Notation to represent a given explanatory variable.

: Notation to represent a given period of time.

: Constant term.

: Vector of coefficients.


: Random error term.

As mentioned before, the use of panel data model allows the combination of time series

and cross-section analysis. In this sense, under the hypothesis of no correlation

13For further information about this theme, see Nilsson (2008).


26/54





26262626

between the exogenous variable and the individual effect, the model can be estimated

directly by ordinary least squares (OLS).

The main problem in this case is that the model error (which is a mix of the individual

effect and the random term), generates a high probability of autocorrelated and

heteroscedastic behavior, with the consequent impact on the efficiency property of the

estimator. However, the recent development of econometric software has introduced

alternatives to correct this problem.

Now that we have defined the main characteristics and main strengths of the

methodology, we will begin our work with a pooled regression model where all timeseries and cross-country observations are assumed to obey the same regression

relationship. We begin by regressing CO2 emission on GDP per capita and GDP per

capita squared using data for all countries. This enables us to investigate the

Environmental Kuznets Curve in its purest form; we refer to it as our Basic Model. The

second step is to consider additional explanatory variables in the model which will be

called Extended Model14 and should allow us to determine what is the effect on the

Environmental Kuznets Curve when additional explanatory variables which might

capture cross-country differences are included.

Finally, it should be addressed that complete panels of data could not be obtained for all

countries in the dataset. This is a common problem with panel data and can be corrected

by using balanced panel estimation methods. This method in panel estimation have

some losses in efficiency by using all available observations but, excluding those for

countries that are not observed in all years of the dataset, even more, there is also no

consideration of the years which do not include a complete dataset for all variables.

14This model has two versions.


27/54





27272727

4. DATA

Our data sources are several international institutions such as the World Bank (WB15),

Food and Agriculture Organization of the United Nations (FAO), Freedom House,

International Monetary Fund (IMF) and the Economic Commission for Latin-American

and the Caribbean (CEPAL16).

The dependent variable under consideration are carbon dioxide (CO2) emissions

(measured in metric tons per capita) which are those stemming from the burning of fossil

fuels and the manufacture of cement. The dataset for CO2 emissions is composed of

measurements for several Latin American countries between the years 1998 to 2005.

The following table shows the list of countries included in the estimation17 and the codes

which identify these countries in each institution from where the data was taken.

15World Development Indicators Online Database.

16We use the initials of the name in Spanish (Comisin Econmica para Amrica Latina y el Caribe - CEPAL)

17Not all Latin American countries could be included in this study due to lack of an appropriate dataset for all the relevant

variables related to the purpose of the research.


28/54





28282828

Pas / rea COD FAO COD WB COD FMI

Argentina 9 ARG 213

Bolivia 19 BOL 218

Brasil 21 BRA 223

Colombia 44 COL 233

Costa Rica 48 CRI 238

Chile 40 CHL 228

Ecuador 58 ECU 248

El Salvador 60 SLV 253

Guatemala 89 GTM 258

Honduras 95 HND 268

Mxico 138 MEX 273

Nicaragua 157 NIC 278

Panam 166 PAN 283

Paraguay 169 PRY 288

Per 170 PER 293

Repblica Dominicana 56 DOM 243

Uruguay 234 URY 298

Venezuela (Repblica Bolivariana de) 236 VEN 299

The panel includes a wide range of macro-level information on socioeconomic

characteristics of countries. The variables are shown in the following table which

contains the definition and an explanation of those variables.


29/54





29292929

CODE VARIABLE UNIT SCALE SOURCE

VAI Voice and Accountability Index IndexRanging from a bout -2.5 to 2.5, with highe r values

corresponding to better governance outcomes.World Bank

PSI Political Stability & Absence of Violence/TerrorismIndex

Index Ranging from a bout -2.5 to 2.5, with highe r valuescorresponding to better governance outcomes.

World Bank

GEI Government Effectiveness Index IndexRanging from a bout -2.5 to 2.5, with highe r values


RQI Regulatory Quality Index IndexRanging from a bout -2.5 to 2.5, with highe r values


RLI Rule of Law Index IndexRanging from a bout -2.5 to 2.5, with highe r values


CCI Control of Corruption Index IndexRanging from a bout -2.5 to 2.5, with highe r values


POP Total Population Hab. 1000 hab. FAO

GDP Gross domestic product, current prices U.S. dollars (Bil lions) Bi llion of US$ IMF

GDPPC Gross domestic product per capita, current prices U.S. dollars IMF

GDPPCPPPGross domestic product based on purchasing-power-

parity (PPP) per capita GDPCurrent international dol lar IMF

UNEMP Average annual unemployment rate Percentage CEPAL

RURP Rural Population Hab. 1000 hab. FAO

RURPR Rural Population % of total population FAO

EXDBT External Debt US$ Mil lions of US$ CEPAL

INF Inflation PercentageData for inflation are averages for the year, not end-of-

period data.IMF

CO2 CO2 Emissions per capita metric tons per capita World Bank

Per capita Gross Domestic Product (GDPPC) is obtained to act as a proxy for the per

capita income of a country. Additional country specific data is gathered in order to control

for possible influences of these national characteristics in explaining CO2 emissions.

For the purpose of this work we used a set of recent institutional indexes constructed by

the World Bank in order to evaluate the main characteristics a democracy should have

through aggregate indicators. Even though the institution estimates indicators related to

six dimensions of governance, three of them: PSI (Political Stability and Absence of

Violence/ Terrorism Index), VAI (Voice and Accountability Index) and CCI (Control of

Corruption Index) are of main interest for this study in order to test the theoretical models

proposed by Fredriksson, Payne and the one proposed by Lopez and Mitra,

respectively18

within the Environmental Kuznets Curve theory. The indicators are

18All this models have been detailed in section 2.


30/54





30303030

constructed using an unobserved components measured in units ranging from about -2.5

to 2.5, with higher values corresponding to better governance outcomes.

The governance indicators presented here reflect the statistical compilation of responses

on the quality of governance given by a large number of enterprise, citizen and expert

survey respondents in industrial and developing countries, as reported by a number of

survey institutes, think tanks, non-governmental organizations, and international

organizations.


31/54





31313131

5. EMPIRICAL FINDING

As mentioned before, heterogeneity across countries is the central focus of the empirical

analysis in this research. Therefore, the introduction of additional variables will be the

most appropriate method in the panel data model in order to focus on differences across

countries with certain characteristics. A general formulation of the pooled model can be

expressed as:

As mentioned before, pooled regression model are those where all time series and

cross-country observations are assumed to obey the same regression relationship19.

The first step was to estimate the so called Basic Model results from regressing CO2

emission (our environmental variable) on GDP per capita (GDPPC) and GDP per capita

squared (GDPPC^2) using data for all countries. This enables us to investigate the

Environmental Kuznets Curve in its purest form.

The Basic Model estimation gave the expected results (see next table); this means thatthe expected sign and statistically significance of both coefficients led to confirm the

quadratic formulation of the Environmental Kuznets Curve in its traditional theoretical

formulation.

It is important to mention also that, as noted in the description of the panel data

methodology, the property of efficiency of each specification depends on the disturbance

associated with each type of specification is homoskedastic (constant variance).

Therefore the need for a general estimate (because the variance-covariance matrix is no

longer a scalar matrix) rises.

19For further references see Greene (2002).


32/54





32323232

In this sense, White20 supplied a method to correct asymptotic variance estimator that

can be applied to the panel estimation. This correction can be obtained by the following

expression:

Where:

: Estimated coefficients.


: Estimation residuals.

Thus, for all specifications ("Basic Model" and "Extended Model"21) this correction was

applied and calculated automatically by the econometric software used (EViews)22.

Even though the countries used in the sample came from the same region (Latin

America) some heterogeneity across them should be expected. Therefore, we consider

adding additional explanatory variables in order to capture these characteristics. The

Extended Model allow us to determine what happens to the Environmental Kuznets

Curve after explanatory variables which might capture cross-country differences are

added.

20For further information about these tests see Arcarons, J. and Calonge, S. (2008)

21Both versions.

22Furthermore, we used cross-section weighting in the estimation.


33/54





33333333

In the case of the Extended Model two different versions were estimated, the first one

which will be called Extended Model N 1 will include the basic Environmental Kuznets

Curve formulation including a series of additional economic variables which aim to

capture the main characteristics of the economy and, therefore, the heterogeneity of the

countries included in the research.

The second version of the Extended Model is called Extended Model N 2 and includes

the set of institutional variables taken from the World Bank into the previous model in

order to assess the impact of government performance on the Environmental Kuznets

Curve.

The following table shows the estimation results for the three models:

Dependent Variable: CO2 Emissions

Linear estimation after one-step weighting matrix

White cross-section standard errors & covariance (d.f. corrected)

INDEPENDENT

VARIABLE

Coefficient t-Statistic Coefficient t-Statistic Coefficient t-Statistic

Constant 0.036332 0.759852 0.053741 0.836159 -0.571314 -4.08082***

GDPPC 0.000839 17.44645*** 0.000908 18.64481*** 0.0007 10.93161***

GDPPC 2 -4.57E-08 -7.154299*** -7.37E-08 -9.084384*** -3.49E-08 -3.012249***

RURP -2.89E-05 -4.003441*** -5.35E-05 -7.85784***

EXDBT 1.12E-05 9.053483*** 1.04E-05 9.940891***

INF 6.30E-06 1.031276 0.007472 6.27357***

UNEMP -0.249857 -0.900385 0.244624 0.321209

CCI -0.501422 -3.302914***

CCI 2 0.745271 9.907683***

PSI -0.104618 -1.779556***

VAI -0.238984 -15.97884***

GEI 0.607658 2.679024***

RQI -0.105945 -1.185813

RLI -0.273657 -2.378383***

WEIGHTED

STATISTICS

R-squared

Adjusted R-squared

F-statistic

UNWEIGHTED

STATISTICS

R-squared

* Significance at 10%

** Significance at 5%

*** Significance at 1%

BASIC MODEL EXTENDED MODEL N 2EXTENDED MODEL N 1

0.664879

464.2624***

0.666314

0.413158 0.6513

0.768356

0.739676

26.79089***

0.748843

0.745378

216.1635***

0.459292


34/54


35/54





35353535

which make higher efforts to control corruption in the public administration tend to

improve their environmental quality.

The results from the Extended Model N 2 on the other hand did not confirm the idea

proposed by Fredriksson23 which could be explained by the fact that even though

corruption and political instability in the region are not as good as in other developing

regions, the level of political instability should not be considered as extremely high as the

theoretical model propose by the author mentioned. Therefore, the results showed that

an increase in the political stability in the region have a positive effect over

environmental quality, this could be explained by the fact that governments with less

instability are more concerned on the efficiency of their policies (which includeenvironmental policies) and to solve the main environmental problems their residents

face.

On the other hand, the results also help to confirm the ideas of Payne which argued that

democratic governments citizens (which are supposed to be informed about

environmental quality status of their regions) can express better their environmental

concerns and demands, which will in turn increase pressure on policy entrepreneurs

(operating in a competitive political system) to respond positively to these demands. In

this sense, in order to test this hypothesis the variable VAI (Voice and Accountability

Index) was introduced as an explanatory variable and the results showed an expected

negative significant sign which means that countries which have strengthen their

democratic mechanisms of representation tend to improve their environmental quality.

In developing countries the necessity of increasing production in order to develop and

promote employment is the main concern of public authorities. This does not mean that

countries in the region are not concerned about environmental quality but t the weight of

this policies are not equal to those related to economic growth promotion. In this sense,

23 That argued that when political instability increases, bribery becomes less attractive for lobbies (because of the

likelihood that the government remains in office) that seeks to change environmental policies


36/54





36363636

the variables GEI is related to government effectiveness showed that an improvement in

this point should be related to efficiency in the allocation of resources which would lead

to an increase in factors productivity inside the country and, therefore, to increase

production which, as mentioned before, increases also pollution.

The impact of public policies on the environmental quality should not only be considered

by all the regulation given by the government but also by the way these governments

make this regulation accomplish its goals, this is what institutional economists usually

call enforcement. The variable RLI (rule of law index) tries to capture this effect and, as

we can appreciate, there is a significant negative effect of this variable on environmental

quality which means that improvement in the enforcement capacity of governments willlead to a reduction on CO2 emissions as Deacon and Mueller suggested.

Finally, it is import to highlight that due to the inclusion of these additional explanatory

variables, the statistical significance of the variable INF (inflation) changed. This could

be explained by the fact that in the Extended Model N 1 we omitted relevant

explanatory variables (institutional variables) that generated a bias in the estimation and

led to accept the null hypothesis of no significance. In this sense, once this problem is

corrected with the inclusion of institutional variables in the model, we found that

countries with higher inflation rates tend to pollute more the environment.

The positive relationship between inflation rates and CO2 emissions could be explained

by the fact that higher inflation rates in the short run does have an impact on production.

In the long run all expansive monetary policies would be expected to translate their

effects on nominal variables but not on real production, however as we mention before,

given that the dataset used in this research do not have an extensive temporal

dimension range, the results obtained for the coefficient of the INF variable does not

capture the long-run effect.


37/54





37373737

5.1. Statistical Contrasts: Panel Data Homogeneity

In order to confirm our results and test if the equations estimating "individual" (cross-

section) and "temporary" (time series) have a better goodness of fit that the joint panel

data estimation, our analysis attempted to test the so called "Panel Data Homogeneity

Contrast24. There are two ways to make the contrast: in relation to a cross-section

sample and for the temporary sample.

However, due to the nature of our work, the contrast will be more appropriate to take into

account the cross-section sample, which is detailed below25.

Panel Data Estimation (Pool):

Cross Section Estimation:

And the contrast is carried through the following expression26:

24This test consists of a general verification of whether or not it is appropriate to specify by panel data methodology.

25For further information about these tests see Arcarons, J. and Calonge, S. (2008) which is the main reference for the

purpose of this section.26In this specification it is important to mention that the VE component is equivalent to the sum of squared errors (SSE) or

sum of squared residuals (SSR) which is the notation used in other references.


38/54





38383838

Under this test, the rejection of the null hypothesis (if the statistic exceeds the value in

tables) leads to invalidate the joint specification of the panel data methodology. As

shown in the table below, statistical inference accepts the null hypothesis and, therefore,

confirmed our initial statement that the use panel data estimation is appropriate.

Basic Model Extended Model N 1 Extended Model N 2

Sum of Squared Residuals

of Individual Equations19.17 9.95 n.a.

Sum of Squared Residuals

of Pool Estimation474.23 455.27 47.75

F-Statistic 6.07 12.37 n.a.

F-Prob 1.00 1.00 n.a.

It is important to mention that this test could only be applied to the Extended Model N1

because the introduction of additional explanatory variables in the Extended Model N 2

made impossible the estimation of the model for a single year due to lack of degrees of

freedom, therefore, we will consider that the results of the Extended Model give enough

information about the desirability of a panel estimation over singles equations.


39/54





39393939

5.2. Graphical Analysis of the Environmental Kuznets Curve

With the results obtained from the three models a graphical analysis of the

Environmental Kuznets Curve was performed in order to analyze the change in the

turning point due to the inclusion of institutional variables into the model as explanatory

variables. The following graph shows the expected CO2 emissions per capita as a

function of per capita gross domestic product (GDPPC) assuming that other explanatory

variables remain at their mean level27.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

727

1,

580

2,

433

3,

285

4,

138

4,

991

5,

844

6,

697

7,

550

8,

403

9,

256

10,

109

10,

962

CO2Emission

GDPPC

Environmental Kuznets Curve

Basic Model

Extended Model N 1

Extended Model N 2

As we can appreciate, the Basic Model showed the highest level of pollution (CO2

emissions) related to the GDPPC level in middle and high income countries of the

sample; however a different outcome was accomplished by the Extended Model N 1

27As we are analyzing a panel data for Latin American countries, in order to get the mean value of each explanatory

variable, we calculate the mean value for each period and country.


40/54





40404040

which showed the highest levels of environmental degradation within the lowest values

of GDPPC28.

An important issue to analyze in the Environmental Kuznets Curve is related not only to

the functional form and variables included, but also on the estimated turning point of the

model and how does it change according to the inclusion of additional explanatory

variables. The following table shows the results of the turning points of the three

models.

Basic ModelExtended

Model N 1

Extended

Model N 2

Turning Point(US$) 9,256 6,058 10,109

The use of economic variables into the basic Environmental Kuznets Curve (Extended

Model N 1) does have an important impact on the estimation of the turning point which

reduced its value in approximately 35%29. On the other hand, a different outcome was

obtained in the Extended Model N 2 which showed a relative small difference with the

Basic Model.

The estimation of the Environmental Kuznets Curve with only economic variables

(Extended Model N 1) gave biased estimators which would explain the big difference in

the value of the turning point. In this sense, as we can appreciate from the graph and

the value of the turning point of the Extended Model N 2, the inclusion of additional

explanatory variables related to institutional p

Documents

INTRODUCING INSTITUTIONAL VARIABLES IN THE ENVIRONMENTAL KUZNETS CURVE (EKC): A LATIN AMERICAN STUDY