an empirical survey of testing the...

39
I E F J, Vol. 7, No. 1, January-June (2012) : 169-207 * Department of Economics, Tunghai University, Taichung, Taiwan, E-mail: [email protected] AN EMPIRICAL SURVEY OF TESTING THE ENVIRONMENTAL KUZNETS CURVE HYPOTHESIS Yi-Chia Wang * Abstract: The environmental Kuznets curve (EKC) is a long-run inverted-U-shaped relationship between a country’s environmental degradation and material living standard. As thorough as possible, this paper collects most of the pre-2007 representative studies that tested the EKC hypothesis, summarizes their similarities and categorizes their differences. With a complete review of the EKC literature, this paper concludes whether the EKC hypothesis can be supported depends critically on data sources, sample countries, length of time period, the adoption of a market exchange rate or purchasing power parity to adjust GDP, regression functional forms and various estimation techniques. Keywords:Environmental Kuznets Curve; EKC; Survey J.E.L. Codes: Q50; Q53 1. INTRODUCTION Analysisof the environment–income relationship can betraced back to the late 1960s. At that time the Club of Romesuggested in its report of ‘Limits to Growth’ that the depletion ofraw materials, energy and (non-renewable) natural resources hadshared a similar upward trend with economic development over severaldecades (Meadows et al. (1972)). This started serious concern withregard to environmental protection. After people gained satisfactionfrom a better material life, their willingness to pay for a cleanerenvironment and eagerness to reduce the intensity of energy use roseaccordingly. Especially in developed economies, pollution control,energy renewal and afforestation have become major items ofexpenditure in both public and private sectors in recent decades.Wildavsky (1988)remarked on this growing consciousness andattempted to protect the environment by concluding that ‘richer issafer and cleaner’. However, convincing empirical evidence tosupport the causality and the relationship between the evolution ofpollution and growing income was scant at the time when the aboveauthors proposed their observational arguments. Due to the constraint of data availability and quality, empiricalanalysis of the relationship between environmental degradation and acountry’s wealth did not commence until the beginning of the 1990s. 2 A pioneering working paper written by Grossman and Krueger (1991) (later published in 1995) investigatedthe relationship between GDP per capita and some air pollutantsusing data from the archive of Global Environmental Monitoring System (GEMS) for 42 low- and high-income countries from 1977 to1984.

Upload: lyngoc

Post on 19-Apr-2018

222 views

Category:

Documents


5 download

TRANSCRIPT

Page 1: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

I E F J, Vol. 7, No. 1, January-June (2012) : 169-207

* Department of Economics, Tunghai University, Taichung, Taiwan, E-mail: [email protected]

AN EMPIRICAL SURVEY OF TESTINGTHE ENVIRONMENTAL KUZNETS CURVE HYPOTHESIS

Yi-Chia Wang*

Abstract: The environmental Kuznets curve (EKC) is a long-run inverted-U-shaped relationshipbetween a country’s environmental degradation and material living standard. As thorough aspossible, this paper collects most of the pre-2007 representative studies that tested the EKChypothesis, summarizes their similarities and categorizes their differences. With a complete reviewof the EKC literature, this paper concludes whether the EKC hypothesis can be supported dependscritically on data sources, sample countries, length of time period, the adoption of a marketexchange rate or purchasing power parity to adjust GDP, regression functional forms and variousestimation techniques.

Keywords:Environmental Kuznets Curve; EKC; Survey

J.E.L. Codes: Q50; Q53

1. INTRODUCTION

Analysisof the environment–income relationship can betraced back to the late 1960s. Atthat time the Club of Romesuggested in its report of ‘Limits to Growth’ that the depletionofraw materials, energy and (non-renewable) natural resources hadshared a similar upwardtrend with economic development over severaldecades (Meadows et al. (1972)). Thisstarted serious concern withregard to environmental protection. After people gainedsatisfactionfrom a better material life, their willingness to pay for a cleanerenvironmentand eagerness to reduce the intensity of energy use roseaccordingly. Especially indeveloped economies, pollution control,energy renewal and afforestation have becomemajor items ofexpenditure in both public and private sectors in recent decades.Wildavsky(1988)remarked on this growing consciousness andattempted to protect the environmentby concluding that ‘richer issafer and cleaner’. However, convincing empirical evidencetosupport the causality and the relationship between the evolution ofpollution and growingincome was scant at the time when the aboveauthors proposed their observationalarguments.

Due to the constraint of data availability and quality, empiricalanalysis of therelationship between environmental degradation and acountry’s wealth did not commenceuntil the beginning of the 1990s.2 A pioneering working paper written by Grossman andKrueger (1991) (later published in 1995) investigatedthe relationship between GDP percapita and some air pollutantsusing data from the archive of Global EnvironmentalMonitoring System (GEMS) for 42 low- and high-income countries from 1977 to1984.

Page 2: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

170 Yi-Chia Wang

Their selected air pollutants include ambient concentrationsof sulfur dioxides (SO2),smoke/dark matters, and suspendedparticulate matters (SPM). Their random effectestimation revealedthat both ambient concentrations of SO2 and smoke/darkmattersfollowed an N-shaped trajectory against GDP per capita. That is, theconcentrationlevels of these two pollutants accumulate before GDPper capita reaches the vicinity of5,000 US$ and then declines asthe economy keeps developing until second turning points(around15,000 US$ for SO2 and 12,000 US$ for smoke/dark matters,respectively) aresurpassed. These two pollutants eventually growwithout bound as long as people continueto become wealthy.

Instead of using the measure of ambient concentrations of airpollutants,Panayotou(1993) chose airborne SO2 emissionsper capita on a national basis and partly echoed theresults from Grossman and Krueger (1991). He concluded that the relationshipbetweenSO2 emissions and income per capita reveals abell-shaped pattern, which is similar to therelationship betweenincome inequality and income per capita proposed in the mid-1950sbyKuznets (1955). Therefore, Panayotou (1993) labeled thisbell-shaped pollution–incomerelationship as the ‘environmentalKuznets curve’ (EKC), which has become a commonly-cited term in theenvironmental literature.

Most of the EKC researchers employ a panel of countries with avariety of incomelevels over a number of time points. They are infavor of estimating the following reduced-form regression:

2 3, 0 1 , 2 , 2 , , , ,i t i t i t i t i t i t i tEP Y Y Y Z= β + β + β + β + β + η + γ + (1)

where EPi,t and Yi,t respectively represent a certainindex of environmental pressure (airpollution, for example) and thelevel of income per capita in country i at time t.3 β denotes arow vector ofcoefficients of other non-income explanatory variables, zi,t,

4 and theregression leftoverincludes country-specific effects (ηi), time-specificfactors (γt) and apure white noise (i,t). Different combinations of the estimated β1, β2, and β3 can lead todistinct shapes of environmentalpressure–income relationship. Figure1demonstratesseveral distinguished forms of this relationship forthe following cases:

(1) β1 and β2 = β3 = 0 suggest that environmental pressure tends to monotonicallyincrease with economic development. Thus, the shape of their relationship yields astraight line with a positive slope β1.

(2) β1 < 0 and β2 = β3 = 0 suggest thatthere exists a decreasing trend of environmentalpressure as income per capitagrows. That is, a straight line with a negative slope β1

can bedrawn for environmental pressure against income percapita.

(3) β1 > 0, β2 < 0 and β3 = 0 (where |β1| > |β2| > 0) suggest aninverted-U quadraticrelationship between EP and Y which represents the EKC pattern. The peak of this

quadratic curveis at the turning point where1

2

.2

Yβ= −β

Page 3: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 171

(4) β1 < 0, β2 > 0 and β3 = 0 (where |β1| > |β2| > 0) suggest a U-shaped curve of the

relationship between EP and Y witha turning point1

2

.2

Yβ= −β

(5) β1 > 0, β2 < 0 and β3 > 0 (where |β1| > |β2| > |β3| > 0) suggest a cubic-polynomial N-shaped relationship in whichenvironmental pressure tends to decline after theeconomyreaches a certain level of income per capita, but climbsupwards afteranother higher level of income per capita is achieved. The two turning points are

22 2 1 3 5

3

3.

3

−β β − β ββ

(6) β1 < 0, β2 > 0 and β3 < 0 (where |β1| > |β2| > |β3| > 0) also suggest a cubic-polynomial but is an inverted-N relationship between EP and Y with two turning

points2

2 2 1 3 6

3

3.

3

−β β − β ββ

(7) β1 = β2 = β3 = 0 suggestsno correlation between EP and Y.

When estimating the reduced-form regression (1),pooled OLS and panel estimationare the commonly used econometrictechniques. Panayotou (1993), for example, appliedpooledcross-section OLS estimation using three air pollutants (emissionsper capita) asdependent variables: sulfur dioxide (SO2), suspended particles and nitrogen oxides (NOX).Hisfindings revealed that the above pollutants support the EKChypothesis with thethresholds of income per capita equaling 3,000,4,500 and 5,500 US$ (market exchangerate adjusted), respectively.However, these results may not be reliable because hisregressionsfailed to correct possible econometric problems of omitted-variablebias andheteroskedasticity in residuals.

Pooled OLS estimation was nearly abandoned after the late 1990s andgraduallyreplaced by panel estimation techniques to avoid theabove-mentioned econometricproblems. Stern and Common (2001), forexample, used the data of sulfur emissions fromA.S.L. and Associates’ yearly report for 73 countries and concluded that fixedeffectestimation for the EKC regression was preferred to randomeffect on the basis of theHausman test (Hausman (1978)). Withthe elimination of country-specific factors byapplying fixed effectestimation, their results seem to provide some support to theexistenceof an inverted-U pattern between (logarithmic) per capitasulfur emissions and(logarithmic) income per capita in the selectedfull sample and the two subsamples (OECDand non-OECD countries). Based on the significant coefficients estimated, thisconcavefunction has a turning point for the non-OECD subsample equaling 908,178 in real1990 US$ per capita (purchasing power parityadjusted), which is far beyond the samplerange of per capita incomelevels in any existing economy. This means that we can have

Page 4: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

172 Yi-Chia Wang

littleconfidence in predictions about a turning point for these countries.By contrast, theresult in the OECD group suggests that thesedeveloped countries will be on the decliningpath after achieving 9,239 US$ per capita income level, while this result from their OECDsubsample was claimed to be generated by a spurious regression.

Given the acceptance of the EKC relationship between degradingenvironment andupgrading wealth, growth optimists tend to believethat economic development is the mostuseful and surest way to greenour nature (Beckerman (1972), Simon (1981) andWildavsky (1988)). As long as an economy achieves a thresholdliving standard,environmental quality will become a luxury goodwith income elasticity greater than unity,which in turn increasesthe value of environmental amenities (Pezzey (1989), Selden andSong (1994), Baldwin (1995), and Day and Grafton (2003)).7 Thesearguments seeminglysuggest that environmental degradation may be ashort-term phenomenon at the beginningof economic development andwill be mitigated in the long run after income per capitasurpassesthe level where the turning point of the EKC occurs.

However, there exists another pattern with an N-shaped curvaturebetweenenvironmental degradation and income per capita. It suggeststhat the reduction in pollutionis also a short-run achievement. After an economy approaches another wealth level(second turningpoint) higher than the first one, the damage to the environmentwilleventually rise again. As previously noted, this has been proposedby Grossman andKrueger (1991) and then discussed further by, forexample, Grossman and Krueger (1995)and Torras and Boyce (1998). In these papers, city-level ambient concentrations of SO2

anddark matters experience only a short-run reduction as income percapita rises, andreturn to increasing trends in the long run.

Following the earliest EKC work produced byGrossman and Krueger (1991), in thesetwo decades a large number ofresearchers devoted themselves to investigating theexistence of the EKC hypothesis by implementing different measures of environmentaldegradation, choosing various data sets, employing diversepollution-related explanatoryvariables and applying advancedeconometric techniques. As thorough as possible, thispaper collectsmost of the pre-2007 representative EKC studies, summarizestheirsimilarities and categorizes their differences. To the best of myknowledge, this paperis the first attempt to categorize most ofmajor EKC studies according to differentmeasurements ofenvironmentally degradation.

2. AIR POLLUTION

Airborne pollutants are chosen as major proxies to measureenvironmental degradationthroughout the whole of EKC history. Apartfrom the accessibility of data, they areregarded as influentialfactors in deteriorating many aspects of community well-being.Forexample, rising concentration levels of suspended particles and darkmatters can affecthuman health by inducing respiratory andcardiovascular diseases. The issue of greenhousegas-driven climatechange that could be responsible for stronger hurricanes and risingsea

Page 5: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

levels is also blamed for its consequence of economic andinfrastructure loss. Thecontribution to the reduction in airbornepollutants, however, is made mainly by developedcountries withhigher income levels, compared to those developing economieswithrelatively weak progress in abating pollution. Therefore, whetherthe EKC hypothesiscan be considered valid for air pollution ismotivating the interest of both environmentalistsand economists.

There are several air pollutants that have been widely investigatedeconometrically inthe EKC literature. These include the sources ofacid precipitation (sulfur dioxide (SO2)and nitrogen oxides (NOX)), the main components of greenhouse gases (carbonmonoxide(CO) and carbon dioxide (CO2)) and the city cloudingveil of suspended particulate matters(SPM) and dark smoke. Althoughthere are other anthropogenic chemical compounds inthe atmosphere,such as chlorofluorocarbons (CFCs), halons, nitrous oxide (N2O)andmethane (CH4), emitted by human activity, they are hardlydiscussed in EKC studiesprobably because of data availability andquality. Comparison can hardly be made whenpapers are scant. As aresult, the focus of this section is mainly on these air pollutantsthathave been explored in a substantial number of studies.

It is noticeable that the use of emissions or concentrations ofairborne pollutants cangenerate different implications as well asestimation results. Before starting to categorizeand summarize EKCfindings according to various air pollutants, it is worthwhile togain abasic understanding of the differences between ambientconcentrations and emissions.

2.1. Ambient Concentrations versus Emissions

The measures of ambient concentrations and emissions providedifferent information andrepresent different aspects ofenvironmental degradation. Ambient concentrations ofcertainairborne pollutants are the proportion of the mass in a given volumeof air or area ona city basis. Therefore, the common concentrationunits are ppm (part per million,

610massof component

massof solution× ),8 ppb(part per billion,

910massof component

massof solution× ) and

(microgram percubic meter). The measure of anthropogenic emissions, on the otherhand,is based on the amount of pollutants converted from thecontents of different fuels on anational basis. Thus, the units ofemissions are generally in carbon equivalent tonnes andkilograms.From an economist’s perspective, anthropogenic emissions of airpollutants aremore directly related to economic activity such asfossil fuel combustion from powerplants, than ambient concentrationlevels that may be long-lived and generated by naturaldecayprocess.

However, the same quantity of emissions may exacerbate theenvironment moreseriously in a smaller economy than a bigger one (Kaufmann et al. (1998)). For example,one million tons of SO2 emitted by Hong Kong has more significant impact on thelocalenvironment than an equivalent amount emitted by China. Theuse of ambient

Page 6: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

174 Yi-Chia Wang

concentrations may avoid this problem of area sizewith respect to environmentaldegradation and is closely related tohuman health as well as environmental quality.

Another drawback of using emissions as the measurement ofenvironmentaldegradation is its inappropriate transformation (Kaufmann et al. (1998)). Emission data isoften calculatedusing a constant conversion rate for various ranks of fuels withdifferentpollutant contents. This problem is especially common indeveloping countries and oftenleads to biased distortion of thetrue quantity of emissions.

The weaknesses of adopting ambient concentrations to measure theenvironmentalquality are also multifold and sometimes moreproblematic than the emission data. First,the level of ambientconcentrations is not dominantly produced by economic activityandcan alter over space and time. Even in the same country, theconcentration levels of airpollutants can vary greatly from ruralareas to urban cities, from winter to summer. Inaddition, theconcentration intensity is also partly determined by natural processandgeographical properties in different areas, as, for example, itis difficult to reduce theconcentration levels with a relativelypersistent decaying process in a ‘stuffy basin’. Theseinherentweaknesses in geography and natural property reveal that ambientlevels are noteasily improved.

Although the level of ambient concentrations is measured directlyfrom reliablemonitoring equipment on a city basis, which increasesthe accuracy compared to emissiondata, it is always affected bynatural disturbances over a very short time. For example, aheavyrainfall can reduce the concentration levels of suspended particlesand dark mattersduring an afternoon. Therefore, temporarilydeclining ambient concentrations of pollutantsdoes not necessarilyimply that the overall environmental burden is mitigated.

In sum, both ambient concentrations and emissions of airbornepollutants have theirmeasuring strengths as well as weaknesses compared in Table 1 pollutants provide city-level information with regard to human health, while national-level emissions representwider environmentalissues, such as climate change and acid deposition. The correlationbetween these two proxies may be weak and no objective judgmentexists in terms of whichone is the better measure of environmentalquality. Cole and Elliott (2003) proposed thatconcentration datatend to be ‘noisier’ than emission data in the sense that differentobservation sites contain varying properties between urban and ruralareas, such astemperatures as well as precipitation rates. Thisrequires a number of dummy variables tocapture site-specificnatures. Thus, emission data are relatively ‘cleaner’ such that theyarepopularly used in recent EKC studies when measuring air-qualitydegradation. Moreover,the main advantage of emissions series is that they have longer and more complete timeseries. Urban concentrations cannot reflect much long-run impacts on the naturalenvironment. This is the main reason that most EKC studies preferemissions data.

2.2. S u l f u r D i o x i d e ( S O 2) Kuznets Curve

Although more than one-half of SO2 emissions are generated bynatural processes (source:the United Nations Environment Programme (UNEP)), such as the explosion of volcanoes,

Page 7: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 175

decaying organic matterand sea spray, human activities are more and more responsibleforthis noxious gas production. The anthropogenic sources of SO2are mainly fossil fuelcombustion from power plants, heating andcooling demands in various weather patterns,smelting of nonferrousores, automobile exhaust and certain chemical manufactures.

Early EKC studies used data of SO2 concentrations from the Global EnvironmentalMonitoring System (GEMS) initiated by the United Nations, based on the city stationslocated in both developedand developing economies spanning around 20 years. Grossmanand Krueger (1991) (and their later publication in 1995), Selden and Song (1994) andPanayotou (1997) are representativestudies using this data archive. Later, sulfur emissiondatagradually replaced concentration data in the EKC estimation becauseof their betterproperties (as discussed in the previous section)and longer time span. A.S.L. andassociates’ yearly data, forexample, measure global sulfur emissions from burning hardcoal, brown coal, and petroleum for the period 1850–1990 and this dataset was laterexpanded by Stern (2005) to 2000 (and keptupdated on his website). His extension wasbased on available sulfuremission data in publication from around 70 countries andtheextrapolation using both the econometric emission frontier model andthe EKC methodfor missing records. This long time-series data setwas employed, for example, by Stern andCommon (2001) and Bertinelli and Strobl (2005) for testing the validity of SO2EKChypothesis.

A summary of studies which estimate the relationship between ambientconcentrationsand income per capita is shown in Table 2. As can be seen, most authors prefer to usefixedeffect and random effect estimation to avoid generaleconometric problems from pooledOLS estimation. However, even usingthe same estimation technique, different authors can

Table 1Comparison of Ambient Concentrations and Emissions as theMeasure of

Environmental Degradation

Ambient Concentrations Emissions

Measurement Quantity in a given volume of Conversion from burning fuels toair or area physical units

Strengths (1) Directly related to human health (1) Directly related to economicand environmental quality activity and wider environmental

problem(2) Avoid the problem of area size (2) No natural or site-specific

with respect to environmental degradationdisturbance

(3) Directly measured by reliable (3) Data has relatively longer timemonitoring equipment span

Weaknesses (1) Partly determined by city’s (1) Emissions may be moregeographical and natural exacerbated in smaller economiesproperties

(2) Vary from time to time, space (2) Indirect and inappropriateto space transformation from fuels with

(3) Influenced by site-specific effects different pollutant content

Page 8: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

176 Yi-Chia WangT

able

2O

verv

iew

of E

KC

Stu

dies

in th

e R

elat

ions

hip

betw

een

Sulf

ur D

ioxi

de (S

O2)

Con

cent

rati

ons

and

Inco

me

Per

Cap

ita

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

GK

9142

cou

ntri

esR

ando

m E

ffec

tC

ity d

umm

y; G

eogr

aphi

c du

mm

y; C

omm

unis

t dum

my;

N-s

hape

d5,

000;

15,

000

(198

5 U

S$, P

PP)

1977

–198

8T

ime

tren

d; P

opul

atio

n de

nsit

y; T

rade

inte

nsity

SB92

149

coun

trie

sFi

xed

Eff

ect

Dem

ocra

tic d

umm

y; T

ime

tren

d; T

rade

inte

nsity

EK

C3,

700

(198

5 U

S$, P

PP)

1960

–199

0S9

414

9 co

untr

ies

Fixe

d E

ffec

tT

ime

tren

dE

KC

3,67

0 (1

985

US$

, PPP

)19

60–1

990

GK

9558

cou

ntri

esR

ando

m E

ffec

tT

he s

ame

as G

K91

; Lag

ged

inco

me

per

capi

ta;

Insi

gnif

ican

tN

/A19

70–1

990

Popu

latio

n de

nsity

; Ind

ustr

y sh

are;

Pol

icy

vari

able

sP9

730

cou

ntri

esR

ando

m E

ffec

tN

-sha

ped

5,00

0; 1

5,00

0 (1

985

US$

, PPP

)19

82–1

994

TB

9819

–42

Pool

ed O

LS

City

dum

my;

Geo

grap

hic

dum

my;

Tim

e tr

end;

N-s

hape

d3,

360;

14,

034

(198

5 U

S$, P

PP)

coun

trie

sPo

pula

tion

dens

ity; M

ean

wat

er te

mpe

ratu

re; L

itera

cy;

1977

–199

4G

INI

coef

fici

ent;

Polit

ical

rig

htK

DG

P98

23 c

ount

ries

Fixe

d E

ffec

tE

cono

mic

act

ivity

per

uni

t of

area

;U

-sha

ped

12,5

00 (

1985

US$

, PPP

)19

74–1

989

(Iro

n an

d st

eel e

xpor

ts)/

GD

P; T

ime

tren

dD

CP0

033

cou

ntri

esPo

oled

OL

ST

ime

dum

my;

Cap

ital-

labo

r ra

tioIn

sign

ific

ant

N/A

1979

–199

0A

CT

0143

cou

ntri

esFi

xed

Eff

ect

Cap

ital-

labo

r ra

tio;

U-s

hape

d10

9,00

0 (1

995

US$

, PPP

)19

71–1

996

Tra

de in

tens

ity;

Tim

e du

mm

yD

G03

Can

ada

OL

ST

ime

tren

dN

-sha

ped

19,7

84; 2

3157

(19

86 U

S$)

1974

–199

7† P

leas

e re

fer

to th

e ab

brev

iati

on o

f au

thor

s an

d as

soci

ated

dat

a so

urce

in T

able

4.

‡ The

cho

ice

of m

etho

dolo

gy is

bas

ed o

n th

e H

ausm

an te

st if

mor

e th

an o

ne e

stim

atio

n te

chni

que

is r

epor

ted.

§ My

calc

ulat

ions

of

turn

ing

poin

ts a

re b

ased

on

the

resu

lts f

rom

ori

gina

l pap

ers

if n

o tu

rnin

g po

int r

epor

ted.

§ Tur

ning

poi

nts

are

repo

rted

seq

uent

iall

y w

hen

the

rela

tions

hip

is c

ubic

.* P

PP: P

urch

asin

g Po

wer

Par

ity a

djus

ted.

* OL

S: O

rdin

ary

Lea

st S

quar

es.

Page 9: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 177

produce results with huge differences. For example, Shafik (1994) and Antweiler et al.(2001) both applied fixed effect estimation, but used a different sample and non-incomeexplanatory variables. Shafik (1994)found the inverted U-shaped relationship betweenSO2 concentrations and income per capita with an achievable turning point of 3,670 US$.In contrast, Antweiler et al. (2001) showed a U-shaped relationship between the twovariables with an extremely high turning point, 109,000US$, which implies that SO2

concentrations are monotonicallydecreasing before an economy’s wealth level reaches thisturningpoint. Experienced EKC authors such as Grossman and Krueger mayadopt thesedifferent results by saying that the relationshipbetween SO2 concentrations and income percapita exhibits an N-shaped pattern with two sequential turning points, as demonstratedintheir paper in 1991. However, the problem of serial correlatederrors in Grossman andKrueger (1991) makes the significance of estimated coefficients unreliable. Afteremploying lagged income percapita in their later publication in 1995, the previous N-shaped pattern was rejected and no significant relationship between SO2 concentrationsand contemporaneous income per capita can befound.

Contrary to the fragile results from studies that analyze SO2concentrations, the EKCrelationship between SO2 emissions andincome per capita can seemingly be supportedamong variousestimation techniques categorized in Table 3. Nevertheless, even though theEKC phenomenon seems to be valid in SO2 emissions, the estimated threshold of GDP percapita canrange from 3,000 (1985 US$) (Panayotou (1993)) to 101,166 (1990 US$) (Sternand Common (2001), full sample). From theresults reported by Stern and Common (2001),different turning points can be attributed to sample selection. Their results shows the EKCturning point in developing countries (non-OECD sub sample) is roughly 100 times higherthan that in developed nations (OECD sub sample). In addition, employing differentexplanatory variablescan also lead to other monotonic or even insignificant patterns of SO2

emission–income relationship (Gangadharan and Valenzuela (2001) and Bertinelli andStrobl (2005), for example), Table 3 provides comparable examples for a number ofstudies.

On average, given the existence of an SO2 EKC phenomenon, theestimated turningpoints in sulfur concentrations are generally lower than those in sulfur emissions, assummarized in Tables 2 and 3. This echoes Selden and Song’s (1994) propositionsuggesting that the estimated turning points for city-level concentrations tend to be atlower income levels than those for national emissions since the reduction (or dispersion) ofpollution concentrations within a city is reasonably easy and less costly to achieve and isoften an early environmental target.

2.3. Nitrogen Oxides (NOX) Kuznets Curve

NOX is considered to be a major source of acid rain. NOX includes nitric oxide (NO) andnitrogen dioxide (NO2), both of which are primarily generated from nitrogenoxidizing inthe air and the burning process of nitride. Whenburning nitride, NO, a colorless and odor-less gas that isa major proportion of NOX emissions, can be transformed into NO2 by photo

Page 10: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

178 Yi-Chia WangT

able

3O

verv

iew

of E

KC

Stu

dies

in th

e R

elat

ions

hip

betw

een

Sulf

ur D

ioxi

de (S

O2)

Em

issi

ons

and

Inco

me

Per

Cap

ita

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

P93

55 c

ount

ries

Pool

ed O

LS

Tro

pica

l dum

my;

EK

C3,

000

(198

5 U

S$, M

ER

)la

te 1

980s

Popu

latio

n de

nsity

SS94

30 c

ount

ries

Ran

dom

Eff

ect

Tim

e du

mm

y; P

opul

atio

n de

nsity

EK

C10

,292

(19

85 U

S$, P

PP)

1973

–198

4C

RB

9710

OE

CD

Ran

dom

Eff

ect

Non

eE

KC

5,70

0–6,

900

(198

5 U

S$, P

PP)

1970

–199

2C

JM97

50 U

S st

ates

Pool

ed O

LS

Non

eM

onot

onic

ally

decr

easi

ngN

/A19

88–1

994

LG

9948

US

stat

esPo

oled

OL

ST

ime

tren

dE

KC

20,1

38 (

1987

US$

)19

29–1

994

SC01

73 c

ount

ries

Fixe

d E

ffec

tN

one

EK

Cfu

ll sa

mpl

e: 1

01,1

66 (

1990

US$

, PPP

)19

60–1

990

OE

CD

: 9,2

39 (

1990

US$

, PPP

)no

n-O

EC

D: 9

08,1

78 (

1990

US$

, PPP

)G

V01

51 c

ount

ries

2SL

SG

INI

coef

fici

ent;

Enr

ollm

ent r

atio

;In

sign

ific

ant

N/A

1998

Urb

an p

opul

atio

n;Po

pula

tion

dens

ityM

LS0

348

US

stat

esSe

mi-

para

met

ric

Non

eE

KC

10,0

00 (

1987

US$

)19

29–1

994

Pool

ed O

LS

Non

eM

onot

onic

ally

incr

easi

ngN

/AC

E03

26 c

ount

ries

Ran

dom

Eff

ect

Cap

ital-

labo

r ra

tio; T

ime

tren

d;M

onot

onic

ally

dec

reas

ing

N/A

1975

–199

0T

rade

inte

nsity

PS03

74 c

ount

ries

Err

or C

orre

ctio

nN

one

EK

C10

,975

(19

90 U

S$, P

PP)

1960

–199

0B

S05

108

coun

trie

sSe

mi-

para

met

ric

Non

eM

onot

onic

ally

incr

easi

ngN

/A19

50–1

990

AX

0648

US

stat

esFi

xed

Eff

ect

Non

eE

KC

15,4

12 (

1987

US$

)19

29–1

994

STR

Mod

el¥

† Ple

ase

refe

r to

the

abbr

evia

tion

of

auth

ors

and

asso

ciat

ed d

ata

sour

ce in

Tab

le 4

.‡ T

he c

hoic

e of

met

hodo

logy

is b

ased

on

the

Hau

sman

test

if m

ore

than

one

est

imat

ion

tech

niqu

e is

rep

orte

d.§ M

y ca

lcul

atio

ns o

f tu

rnin

g po

ints

are

bas

ed o

n th

e re

sults

fro

m o

rigi

nal p

aper

s if

no

turn

ing

poin

t rep

orte

d.§ T

urni

ng p

oint

s ar

e re

port

ed s

eque

ntia

lly

whe

n th

e re

latio

nshi

p is

cub

ic.

¥ The

Sm

ooth

Tra

nsiti

on R

egre

ssio

nMod

el is

est

imat

ed u

sing

non

-lin

ear

leas

t squ

ares

.* P

PP: P

urch

asin

g Po

wer

Par

ity a

djus

ted;

ME

R: M

arke

t Exc

hang

e R

atea

djus

ted.

* OL

S: O

rdin

ary

Lea

st S

quar

es; 2

SLS:

Tw

o-st

age

Lea

st S

quar

es.

Page 11: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 179

Table 4List of Sulfur EKC Authors and Associated Pollution Data Sources

Abbreviation References Source of SO2 data

ACT01 Antweiler et al. (2001) United Nations Environment ProgrammeAX06 Aslanidis and Xepapadeas (2006) Environmental Protection Agency, The United StatesBS05 Bertinelli and Strobl (2005) A.S.L. and Associates’ yearly reportCE03 Cole and Elliott (2003) United Nations Environment ProgrammeCJM97 Carson et al. (1997) Environmental Protection Agency, The United StatesCRB97 Cole et al. (1997) Environmental Data Compendium 1993, Paris: OECDDCP00 Dinda et al. (2000) World Development Report, World BankDG03 Day and Grafton (2003) Pollution Data Branch, Environment CanadaGK91 Grossman and Krueger (1991) Global Environmental Monitoring SystemGK95 Grossman and Krueger (1995) Global Environmental Monitoring SystemGV01 Gangadharan and Valenzuela (2001) World Development IndicatorsKDGP98 Kaufmann et al. (1998) United Nations Statistical YearbookLG99 List and Gallet (1999) Environmental Protection Agency, The United StatesMLS03 Millimet et al. (2003) Environmental Protection Agency, The United StatesP93 Panayotou (1993) Global Emissions on a National BasisP97 Panayotou (1997) Global Environmental Monitoring SystemPS03 Perman and Stern (2003) A.S.L. and Associates’ yearly reportS94 Shafik (1994) Monitoring and Assessment Research Centre, LondonSB92 Shafik and Bandyopadhyay (1992) World Development Report, World BankSC01 Stern and Common (2001) A.S.L. and Associates’ yearly reportSS94 Selden and Song (1994) Global Environmental Monitoring SystemTB98 Torras and Boyce (1998) Global Environmental Monitoring System

chemical reactions. NO2, a red-brown gasthat stimulates smell and the respiratory system,can easily dissolve in water and then decompose to nitrous and nitric acid by further photochemical reactions with vapor after absorbing sunlight. It can also oxidize to nitrate andlead to rain acidification, eroding land fertility and gradually deteriorating economicinfrastructure.

NOX concentration data can be accessed in several countries, but due to variousestimation standards and equipment, pooled country estimation may not be reliable.Therefore, NOX emission data are employed in a majority of NOX EKC studies. Asummary of these studies is providedin Table 5. In the studies that support the EKCrelationship between NOX emissions and GDP per capita, the estimated turning points areroughly around 15,000 US$ withsmall variation. Panayotou (1993), however, is anobvious deviation with the estimated bell-shaped turning point equaling 5,500 US$. Thisdifference may result from the use of a market exchange rate and the pooled OLSestimation technique. de Bruyn and Heintz (1999) and Stern and Common (2001)remarked that the adoption of a market exchange rate could result in adownward biasedturning point because of the fact that it cannot adequately reflect the true ‘wealth of

Page 12: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

180 Yi-Chia WangT

able

5O

verv

iew

of E

KC

Stu

dies

in th

e R

elat

ions

hip

betw

een

Nit

roge

n O

xide

s (N

OX) E

mis

sion

s an

d In

com

e P

er C

apit

a

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

P93

55 c

ount

ries

Pool

ed O

LS

Tro

pica

l dum

my;

EK

C5,

500

(198

5 U

S$, M

ER

)la

te 1

980s

Popu

latio

n de

nsity

SS94

30 c

ount

ries

Ran

dom

Eff

ect

Tro

pica

l dum

my;

EK

C17

,843

(19

85 U

S$, P

PP)

1973

–198

4Po

pula

tion

dens

ityC

RB

979

OE

CD

Ran

dom

Eff

ect

Non

eE

KC

14,7

00–1

5,10

0 (1

985

US$

, PPP

)19

70–1

992

CJM

9750

US

stat

esPo

oled

OL

SN

one

Mon

oton

ical

ly d

ecre

asin

gN

/A19

88–1

994

LG

9948

US

stat

esPo

oled

OL

ST

ime

tren

dN

-sha

ped

8,33

3; 2

5,00

0 (1

987

US$

)19

29–1

994

GV

0151

cou

ntri

es2S

LS

GIN

I co

effi

cien

t; E

nrol

lmen

t rat

io;

Insi

gnif

ican

tN

/A19

98U

rban

pop

ulat

ion;

Popu

latio

n de

nsity

ML

S03

48 U

S st

ates

Pool

ed O

LS

Non

eN

-sha

ped

8,33

3; 2

5,00

0 (1

987

US$

)19

29–1

994

Sem

i-pa

ram

etri

cN

one

EK

C12

,000

(19

87 U

S$)

CE

0326

cou

ntri

esR

ando

m E

ffec

tC

apita

l-la

bor

ratio

; Tim

e tr

end;

Mon

oton

ical

ly d

ecre

asin

gN

/A19

75–1

990

Tra

de in

tens

ityA

X06

48 U

S st

ates

Fixe

d E

ffec

tN

one

Mon

oton

ical

ly in

crea

sing

N/A

1929

–199

4ST

R M

odel

¥

† Ple

ase

refe

r to

the

abbr

evia

tion

of

auth

ors

and

asso

ciat

ed d

ata

sour

ce in

Tab

le 6

.‡ T

he c

hoic

e of

met

hodo

logy

is b

ased

on

the

Hau

sman

test

if m

ore

than

one

est

imat

ion

tech

niqu

e is

rep

orte

d.§ M

y ca

lcul

atio

ns o

f tu

rnin

g po

ints

are

bas

ed o

n th

e re

sults

fro

m o

rigi

nal p

aper

s if

no

turn

ing

poin

t rep

orte

d.§ T

urni

ng p

oint

s ar

e re

port

ed s

eque

ntia

lly

whe

n th

e re

latio

nshi

p is

cub

ic.

¥ The

Sm

ooth

Tra

nsiti

on R

egre

ssio

nMod

el is

est

imat

ed u

sing

non

-lin

ear

leas

t squ

ares

.* P

PP: P

urch

asin

g Po

wer

Par

ity a

djus

ted;

ME

R: M

arke

t Exc

hang

e R

atea

djus

ted.

* OL

S: O

rdin

ary

Lea

st S

quar

es; 2

SLS:

Tw

o-st

age

Lea

st S

quar

es.

Page 13: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 181

Table 6List of NOX EKC Authors and Associated Pollution Data Sources

Abbreviation References Source of NOX data

AX06 Aslanidis and Xepapadeas (2006) Environmental Protection Agency, The United StatesCE03 Cole and Elliott (2003) United Nations Environment ProgrammeCJM97 Carson et al. (1997) Environmental Protection Agency, The United StatesCRB97 Cole et al. (1997) Environmental Data Compendium 1993, Paris: OECDGV01 Gangadharan and Valenzuela (2001) World Development IndicatorsLG99 List and Gallet (1999) Environmental Protection Agency, The United StatesMLS03 Millimet et al. (2003) Environmental Protection Agency, The United StatesP93 Panayotou (1993) Global Emissions on a National BasisSS94 Selden and Song (1994) Global Environmental Monitoring System

nations’. In addition, theomitted-variable problem is especially serious when using pooledOLS estimation without controlling for time- and country-specific effects, which mayfurther result in the bias of EKC turning points.

Single country EKC estimation for NOX emissions was undertaken mainly in theUnited States using data from the Environmental Protection Agency (EPA) for the periodfrom the late1920s to the present. Employing a sample period of less than 20 years for 50states, Carson et al. (1997), for example, came to the conclusion that NOX emissions aremonotonically decreasing as per capita income rises. However, applying the same pooledOLS estimation technique, List and Gallet (1999) chose alonger time span from 1929 to1994 for 48 states, and estimated an N-shaped NOX emission–income relationship. List andGallet (1999) further rejected the ‘one-size-fits-all’ reduced-form regression, saying thatthe F-test cannot accept thenull hypothesis that the intercept and slope coefficients are thesame across states. However, the above pooled OLS estimations forthe United States seemto be unreliable when using the fixed-effectsmooth transition regression (STR) model toaccount for thestate-specific effects that may lead to biased OLS coefficients (As lanidisand Xepapadeas (2006)). Although the NOXemission–income relationship was found to bemonotonically increasing byAslanidis and Xepapadeas (2006), they estimated a percapitaincome level equaling 15,658 US$ at which the increasing speed of NOX emissions tend toslow down as income percapita continuously rises.

To sum up, the existence of the EKC hypothesis for NOX emissions depends criticallyon different estimation techniques. Earlier EKC studies tend to support the inverted-Urelationship between NOX emissions and income per capita, but thishas been mostlydisproved in recent studies with newly-developed econometric methods as well as otherexplanatory variables employed.

2.4. Particulates, Dark Matters and Smoke Kuznets Curve

In addition to SO2 and NOX, suspended particulate matter (SPM), or aerosol, is consideredas another typical local airpollutant generated from both natural and anthropogenic

Page 14: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

182 Yi-Chia Wang

activities. The former includes dust, sea spray, forest fires, and volcanic explosions, whilethe latter mainly consists of incomplete fuel combustion and industrial chemical processes.Most naturally produced particulates are relatively large, causing rather minorimpacts onhuman health but graying the sky and reducing visibility. Conversely, particulatesproduced by human activities, including raised dust from streets, exhaust fume from cars,outdoor burning, construction, and arable land farming, are smaller than 10 microns, andnot only lead to eye and lung damage but also aggravaterespiratory illness (e.g. coughs andasthma). They also result inincreasing mortality rates among children and the elderlybycarrying carcinogenic or poisonous heavy metals into their lungs. Therefore,anthropogenic SPM is a serious city-level environmental problem.

The reduction in anthropogenic SPM emissions can generally beachieved byincreasing polluters’ costs in buying advanced equipment, controlling burning intensityand combustion quality of coals and seeking alternative cleaner fuels. Definitely, tofinance these costs, an economy’s priority is nothing more than rising overall wealth level.

Most of the existing SPM EKC studies are in favor of investigating SPMconcentrations rather than emissions, though their anthropogenic sources are mutuallycorrelated within a city. In the three SPM emission–income EKC studies shown in Table 7,Selden and Song (1994) and Cole et al. (1997) supported the existence of the SPM EKChypothesis using fixed effect and random effect estimation, respectively. It is noticeablethat the seven OECD countries chosen by Cole et al. (1997) have a relatively lower turningpoint (7,300–8,100 US$) compared to the pooled 30 countries selected by Selden and Song(1994) with the estimatedturning point equaling 10,289 US$. In the case of the 50 states inAmerica during the period 1988–1994, Carson et al. (1997) founda monotonicallydecreasing pattern for SPM emissions as per capita income increases, though this mightnot be reliable when consideringthe inherent weaknesses of their pooled OLS technique.

With regard to SPM concentrations, the existence of the SPM EK Chypothesis is alsofragile, as illustrated in the second part of Table 7. Although Grossman and Krueger (1991)found monotonically decreasing SPM concentration levels as income percapita rises, theirlater work in 1995 replaced previous results byan insignificant relationship after enlargingthe number of nationsas well as incorporating lagged income variables in their model.Shafik (1994) also extended her earlier work of SPM EKC regressions from Shafik andBandyopadhyay (1992) using the same sample and methodology but different data sourcesand explanatory variables. An identical inverted U-shaped relationship between SPMconcentrations and income per capita was generated in these two papers with a turningpoint of 3,280 US$. Panayotou (1993) also provided a support to Shafik andBandyopadhyay (1992) and Shafik (1994) with a slightly higher turning point, but it wasmeasured by a market exchange rate rather than adjusted by purchasing power parity.

Income per capita might not be correlated with SPM concentrations when employingtoo many pollution- and income-related explanatory variables. Torras and Boyce (1998),for example, used the concentration levels of heavy particles as the dependent variable

Page 15: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 183T

able

7O

verv

iew

of E

KC

Stu

dies

in th

e R

elat

ions

hip

betw

een

Susp

ende

d P

arti

cula

te M

atte

rs (S

PM

) and

Inc

ome

Per

Cap

ita

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

Em

issi

ons

SS94

30 c

ount

ries

Fixe

d E

ffec

tT

ime

dum

my;

Pop

ulat

ion

dens

ityE

KC

10,2

89 (

1985

US$

, PPP

)19

73–1

984

CR

B97

7 O

EC

D c

ount

ries

Ran

dom

Eff

ect

Non

eE

KC

7,30

0–8,

100

(198

5 U

S$, P

PP)

1970

–199

2C

JM97

50 U

S st

ates

Pool

ed O

LS

Non

eM

onot

onic

ally

dec

reas

ing

N/A

1988

–199

4

Con

cent

rati

ons

GK

9129

cou

ntri

esR

ando

m E

ffec

tC

ity d

umm

y; G

eogr

aphi

c du

mm

y;M

onot

onic

ally

dec

reas

ing

N/A

1977

–198

8C

omm

unis

t dum

my;

Tim

e tr

end;

Popu

latio

n de

nsity

; Tra

de in

tens

itySB

9214

9 co

untr

ies

Fixe

d E

ffec

tD

emoc

ratic

dum

my;

Tim

e tr

end;

EK

C3,

280

(198

5 U

S$, P

PP)

1960

–199

0T

rade

inte

nsity

P93

55 c

ount

ries

Pool

ed O

LS

Tro

pica

l dum

my;

Pop

ulat

ion

dens

ityE

KC

4,50

0 (1

985

US$

, ME

R)

late

198

0sS9

414

9 co

untr

ies

Fixe

d E

ffec

tT

ime

tren

dE

KC

3,28

0 (1

985

US$

, PPP

)19

60–1

990

GK

9558

cou

ntri

esR

ando

m E

ffec

tT

he s

ame

as G

K91

; Lag

ged

inco

me

Insi

gnif

ican

tN

/A19

70–1

990

per

capi

taV

97C

ross

sec

tion,

Fixe

d E

ffec

tPo

pula

tion

dens

ity; T

ime

tren

dM

onot

onic

ally

incr

easi

ngN

/AM

alay

sia

late

1970

s–ea

rly

1990

sT

B98

19–4

2Po

oled

OL

SC

ity d

umm

y; G

eogr

aphi

c du

mm

y;In

sign

ific

ant

N/A

coun

trie

sT

ime

tren

d; P

opul

atio

n de

nsity

;19

77–1

994

Lite

racy

; GIN

I co

effi

cien

t;Po

litic

al r

ight

cont

d. ta

ble

7

Page 16: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

184 Yi-Chia WangA

utho

rs†

Sam

ple

Met

hodo

logy

‡O

ther

exp

lana

tory

var

iabl

esR

elat

ions

hip

Tur

ning

poi

nts§

DC

P00

33 c

ount

ries

Pool

ed O

LS

Tim

e du

mm

y; C

apita

l-la

bor

ratio

U-s

hape

d12

,998

(19

85 U

S$, P

PP)

1979

–199

0G

V01

51 c

ount

ries

2SL

SG

INI

coef

fici

ent;

Enr

ollm

ent r

atio

;In

sign

ific

ant

N/A

1998

Urb

an p

opul

atio

n; P

opul

atio

n de

nsity

DG

03C

anad

aO

LS

Tim

e tr

end

Mon

oton

ical

ly in

crea

sing

N/A

1974

–199

7

† Ple

ase

refe

r to

the

abbr

evia

tion

of

auth

ors

and

asso

ciat

ed d

ata

sour

ce in

Tab

le 8

.‡ T

he c

hoic

e of

met

hodo

logy

is b

ased

on

the

Hau

sman

test

if m

ore

than

one

est

imat

ion

tech

niqu

e is

rep

orte

d.§ M

y ca

lcul

atio

ns o

f tu

rnin

g po

ints

are

bas

ed o

n th

e re

sults

fro

m o

rigi

nal p

aper

s if

no

turn

ing

poin

t rep

orte

d.* T

he S

PM O

LS

regr

essi

on e

stim

ated

by

DG

03 (

Day

and

Gra

fton

(20

03))

may

be

mis

-spe

cifi

ed s

ince

it f

ails

som

e ke

yF

-tes

ts.

* PPP

: Pur

chas

ing

Pow

er P

arity

adj

uste

d; M

ER

: Mar

ket E

xcha

nge

Rat

eadj

uste

d.* O

LS:

Ord

inar

y L

east

Squ

ares

; 2SL

S: T

wo-

stag

e L

east

Squ

ares

.

Page 17: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 185

against several income-related explanatory variables listed in Table 7 and found aninsignificant pollution–income relationship. Gangadharan and Valenzuela (2001) echoedTorras and Boyce (1998) by using a similar set of explanatory variables. As theindependent variables adopted by these two studiesmay be highly correlated with incomeper capita (GINI coefficientsand population density, for instance), their insignificantresultsmay be wrongly produced due to multi-collinearity problems.

Compared to SPM, dark matters and smoke are relatively finer andmore hazardous tohuman health in the short term. They result fromsimilar anthropogenic sources to SPM.Although Beckerman (1992) treated SPM and smoke as the same thing, empirical EKCstudies usually employ the data that combines dark matter and smoke together (forexample, Grossman and Krueger (1991, 1995)).

Data for dark matters and smoke in previous studies was usually extracted from theGEMS archive. Recent EKC studies for these two pollutants are relatively rare probablybecause Grossman and Krueger (1995) and Torras and Boyce (1998) have rejected thesignificant relationship between them and economic development. In addition, it isdifficult to have alternative data sources other than GEMS to compare with existingstudies. On the other hand, because of the similar properties to SPM, with relatively well-recorded data sources, EKC researchers are generally in favor of using SPM as themeasurement of local environmental degradation. Table 9 summarizes an overview of theEKC findings for dark matters and smoke.

2.5. Carbon Kuznets Curve

Carbon pollutants include carbon monoxide (CO) and carbon dioxide (CO2). In addition tothose generated from natural events, suchas forest fires, methane oxidization, and creature

Table 8List of SPM EKC Authors and Associated Pollution Data Sources

Abbreviation References Source of SPM data

CJM97 Carson et al. (1997) Environmental Protection Agency, The United StatesCRB97 Cole et al. (1997) Environmental Data Compendium 1993, Paris: OECDDCP00 Dinda et al. (2000) World Development Report, World BankDG03 Day and Grafton (2003) Pollution Data Branch, Environment CanadaGK91 Grossman and Krueger (1991) Global Environmental Monitoring SystemGK95 Grossman and Krueger (1995) Global Environmental Monitoring SystemGV01 Gangadharan and Valenzuela (2001) World Development IndicatorsP93 Panayotou (1993) Global Environmental Monitoring SystemS94 Shafik (1994) Monitoring and Assessment Research Centre, LondonSB92 Shafik and Bandyopadhyay (1992) World Development Report, World BankSS94 Selden and Song (1994) Global Environmental Monitoring SystemTB98 Torras and Boyce (1998) Global Environmental Monitoring SystemV97 Vincent (1997) Department of Environment, Malaysia

Page 18: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

186 Yi-Chia Wang

activities, CO emissions are mainly produced by the incomplete burning process of thepetro chemical fuels for economic purposes. CO is a colorless andodor-less gas lighter thanair but heavier than oxygen. Thus, over-inhalation of this gas may cause low oxygenconcentration in the blood and tissues of a human body, leading to stupor and even death.In addition to the effect on human health, CO also acts as aminor component of greenhouse gases that warm the Earth.

Green house gases are the chemical compounds found in the Earth’s atmosphere thatcan absorb the infrared radiation (heat) reflected from the Earth’s surface. Theaccumulation of these gases can trap the reflection of sunlight and gradually raise globalaverage temperatures. Thus, green house gas emissions, of which CO2 isthe majorcomponent, are regarded as the most serious global air pollutants and difficult to bereduced due to the needs of economic development.

During the past 20 years, about three-quarters of human-made CO2 was emitted byburning fossil fuels (source: the Oak Ridge National Laboratory (ORNL), 2000). Energy-related CO2 emissions, resulting from burning coal, petroleum and natural gas, consist ofmore than 80% of anthropogenic green house gas emissionsin the United States (source:Energy Information Administration, Washington, D. C., 2002), the biggest CO2 emitter inthe world. Around 1.5 billion metric tons of CO2 emissions are emittedannually by theUnited States, which constitutes nearly 50% ofcurrent global annual emissions (3.2 billiontons).

Unlike local pollutants, the reduction in carbon emissions is aninternational or jointtarget in most of the developed countries, rather than a domestic issue. In 1997, 35industrial nations were obliged to cut CO2 emissions by an overall 5.2% below 1990 levelsby 2008–12 in an agreement known as the Kyoto Protocol. Through this international pact,several high-income economies, suchas Germany and France, have successfully eased thegrowth of carbon emissions in recent years. In addition, international assistance forpollution controlling techniques in developing countries also playsa successful role incurbing global carbon emissions.

Whether this achievement requires economies to become rich is also aworth-investigating topic in the EKC literature. In Canada, CO concentrations were found to bedeclining only when income percapita lies between 20,240 and 24,321 US$ (Day andGrafton (2003)). However, this finding might not be robustin the time-series frameworkbecause of its short sample period (24 time points) as well as possible non-stationary OLSresiduals fromtheir reduced-form regression. In addition, it is uncommon to useambientconcentrations in estimating carbon EKC because international data are incomplete,unstandardized, and most importantly meaning less. CO2 is well-mixed and long-livedinthe atmosphere and regional variations are minor. CO emission data, on the contrary, arerelatively accessible in a variety of data sources.

Table10 lists several EKC studies of the CO emission–income relationship.Unsurprisingly, seven OECD economiesselected by Cole et al. (1997) were found to have

Page 19: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 187T

able

9O

verv

iew

of E

KC

Stu

dies

in th

e R

elat

ions

hip

betw

een

Con

cent

rati

ons

of S

mok

e/D

ark

Mat

ters

and

Inc

ome

Per

Cap

ita

Aut

hors

Sam

ple

Met

hodo

logy

‡O

ther

exp

lana

tory

var

iabl

esR

elat

ions

hip

Tur

ning

poi

nts§

GK

9119

cou

ntri

esR

ando

m E

ffec

tC

ity d

umm

y; G

eogr

aphi

c du

mm

y;N

-sha

ped

5,00

0; 1

2,00

0 (1

985

US$

, PPP

)C

omm

unis

t dum

my;

1977

–198

8T

ime

tren

d; P

opul

atio

n de

nsit

y; T

rade

inte

nsity

GK

9558

cou

ntri

esR

ando

m E

ffec

tT

he s

ame

as G

K91

; Lag

ged

inco

me

per

capi

taIn

sign

ific

ant

N/A

1970

–199

0

TB

9819

–42

Pool

ed O

LS

City

dum

my;

Geo

grap

hic

dum

my;

Tim

e tr

end;

Insi

gnif

ican

tN

/Aco

untr

ies

Popu

latio

n de

nsity

; Lite

racy

;19

77–1

994

GIN

I co

effi

cien

t; Po

litic

al r

ight

Abb

revi

atio

nR

efer

ence

sSo

urce

of R

elev

ant d

ata

GK

91G

ross

man

and

Kru

eger

(19

91)

Glo

bal E

nvir

onm

enta

l Mon

itori

ng S

yste

mG

K95

Gro

ssm

an a

nd K

rueg

er (

1995

)G

loba

l Env

iron

men

tal M

onito

ring

Sys

tem

TB

98T

orra

s an

d B

oyce

(19

98)

Glo

bal E

nvir

onm

enta

l Mon

itori

ng S

yste

m

‡ The

cho

ice

of m

etho

dolo

gy is

bas

ed o

n th

e H

ausm

an te

st if

mor

e th

an o

ne e

stim

atio

n te

chni

que

is r

epor

ted.

§ My

calc

ulat

ions

of

turn

ing

poin

ts a

re b

ased

on

the

resu

lts f

rom

ori

gina

l pap

ers

if n

o tu

rnin

g po

int r

epor

ted.

§ Tur

ning

poi

nts

are

repo

rted

seq

uent

iall

y w

hen

the

rela

tions

hip

is c

ubic

.* P

PP: P

urch

asin

g Po

wer

Par

ity a

djus

ted.

* OL

S: O

rdin

ary

Lea

st S

quar

es.

Page 20: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

188 Yi-Chia WangT

able

10

Ove

rvie

w o

f EK

C S

tudi

es in

the

Rel

atio

nshi

p be

twee

n C

arbo

nM

onox

ide

(CO

) and

Inc

ome

Per

Cap

ita

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

Con

cent

rati

ons

DG

03C

anad

aO

LS

Tim

e tr

end

Inve

rted

N-s

hape

d20

,240

; 24,

321

(198

6 U

S$)

1974

–199

7E

mis

sion

sSB

9214

9 co

untr

ies

Fixe

d E

ffec

tD

emoc

ratic

dum

my;

Tim

e tr

end;

Tra

de in

tens

ityE

KC

6,24

1 (1

985

US$

,PPP

)19

60–1

990

SS94

30 c

ount

ries

Fixe

d E

ffec

tT

ime

dum

my;

Pop

ulat

ion

dens

ityIn

sign

ific

ant

N/A

1973

–198

4C

RB

977

OE

CD

Ran

dom

Eff

ect

Non

eE

KC

9,90

0–10

,100

(19

85 U

S$, P

PP)

1970

–199

2C

JM97

50 U

S st

ates

Pool

ed O

LS

Non

eM

onot

onic

ally

dec

reas

ing

N/A

1988

–199

4† P

leas

e re

fer

to th

e ab

brev

iati

on o

f au

thor

s an

d as

soci

ated

dat

a so

urce

in T

able

11.

‡ The

cho

ice

of m

etho

dolo

gy is

bas

ed o

n th

e H

ausm

an te

st if

mor

e th

an o

ne e

stim

atio

n te

chni

que

is r

epor

ted.

§ My

calc

ulat

ions

of

turn

ing

poin

ts a

re b

ased

on

the

resu

lts f

rom

ori

gina

l pap

ers

if n

o tu

rnin

g po

int r

epor

ted.

§ Tur

ning

poi

nts

are

repo

rted

seq

uent

iall

y w

hen

the

rela

tions

hip

is c

ubic

.* P

PP: P

urch

asin

g Po

wer

Par

ity a

djus

ted.

* OL

S: O

rdin

ary

Lea

st S

quar

es.

Page 21: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 189

Table 11List of CO EKC Authors and Associated Pollution Data Sources

Abbreviation References Source of CO data

SB92 Shafik and Bandyopadhyay (1992) World Development Report, World BankSS94 Selden and Song (1994) Global Environmental Monitoring SystemCJM97 Carson et al. (1997) Environmental Protection Agency, The United StatesCRB97 Cole et al. (1997) Environmental Data Compendium 1993, Paris: OECDDG03 Day and Grafton (2003) Pollution Data Branch, Environment Canada

decreasing levels of CO emissions after a per capita income ranging from 9,900 to 10,100US$ was achieved. In addition, using pooled OL Sestimation, Carson et al. (1997) came tothe conclusion that CO emissions in the United States were monotonically decreasingduring the period 1988–1994. However, the estimated relationship between CO emissionsand income per capita is not consistent among the studies selected in Table 10.

The importance of green house gases, or their representative CO2 emissions, drives theproduction of historical data record for both environmental and economic research. ORNL,a laboratory organized bythe Carbon Dioxide Information Analysis Center (CDIAC), hasrecords for more than 100 years of CO2 emissions for a majority of advanced countries (forthe United Kingdom, ORNL even traces back to 1750), and at least 50 years for manydeveloping nations. ORNL’sestimation of CO2 emissions includes fossil fuels (gas,liquidand solid fuel) consumption, cement production and gas flaring. Other data sources,such as the International Energy Agency (IEA), operated by OECD, also chroniclesseveral decades of CO2 emissions by advanced measure, which is arguably better than thatof ORNL (Galeotti et al. (2006)).

It seems to be a general result that the EKC relationship between CO2 emissions andincome per capita can be produced in thestudies before 2000, but with a wide range ofturning points assummarized in Table 12. This problem may result fromthe differentsources of data and estimation techniques chosen. Using data from the World Bank andORNL, respectively, Shafik and Bandyopadhyay (1992) and Holtz-Eakin and Selden(1995) found two distinct EKC turning points equaling 4,000 and 35,428US$,respectively. Schmalen see et al. (1998) used a similar sample as Holtz-Eakin and Selden(1995) but divided their sample income range by 10 segments in which OLS estimation isindividually applied (spline model). They found that in the 10thsegment where per capitaincome is higher than 10,000 US$, CO2 emissions will be negatively correlated withincome per capita.

In the past five years, most of the early EKC findings between CO2 emissions andincome per capita were rejected and were infavor of a monotonically increasing pattern,under a variety of newly-developed econometric techniques. Bertinelli and Strobl (2005),for example, emphasized that the irsemi-parametric regression places no restriction on thefunctional form, which avoids the general weakness in the traditional reduced-form EKCregression. Their results indicated a monotonically increasing relationship between CO2

Page 22: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

190 Yi-Chia WangT

able

12

Ove

rvie

w o

f EK

C S

tudi

es in

the

Rel

atio

nshi

p be

twee

n C

arbo

n D

ioxi

de (C

O2)

Em

issi

ons

and

Inco

me

Per

Cap

ita

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

Em

issi

ons

SB92

149

coun

trie

sFi

xed

Eff

ect

Dem

ocra

tic d

umm

y; T

ime

tren

d;E

KC

4,00

0 (1

985

US$

,PPP

)19

60–1

990

Tra

de in

tens

ityS9

414

9 co

untr

ies

Fixe

d E

ffec

tT

ime

tren

dM

onot

onic

ally

incr

easi

ngN

/A19

60–1

990

HE

S95

130

coun

trie

sFi

xed

Eff

ect

Non

eE

KC

35,4

28 (

1985

US$

,PPP

)19

51–1

986

RG

9713

5 co

untr

ies

Yea

rly

OL

SN

one

Mon

oton

ical

ly in

crea

sing

N/A

1962

–199

1C

RB

977

regi

ons

Ran

dom

Eff

ect

Non

eE

KC

25,1

00–6

2,70

0 (1

985

US$

, PPP

)19

60–1

991

CJM

9750

US

stat

esPo

oled

OL

SN

one

Mon

oton

ical

ly d

ecre

asin

gN

/A19

88–1

994

MU

9716

cou

ntri

esFi

xed

Eff

ect

Non

eN

-sha

ped

11,6

36; 2

4,55

5 (1

985

US$

, PPP

)19

50–1

992

UM

9816

cou

ntri

esC

haos

stu

dyN

one

EK

C11

,426

(19

85 U

S$, P

PP)

1950

–199

2SS

J98

141

coun

trie

sSp

line

mod

elN

one

EK

C10

,000

(19

85 U

S$, P

PP)

1950

–199

0A

C99

34 e

cono

mie

sA

utor

egre

ssiv

eL

agge

d de

pend

ent v

aria

ble;

EK

C13

,630

(19

85 U

S$, P

PP)

1971

–198

9di

stri

bute

d la

gT

rade

inte

nsity

S99

Chi

naD

escr

iptio

nN

one

EK

C50

0 (1

987

Yua

n)19

72–1

995

GL

9911

0 co

untr

ies

Gam

ma

func

tion

Non

eE

KC

16,6

46 (

1990

US$

, PPP

)19

71–1

996

Wei

bull

func

tion

Non

eE

KC

15,0

73 (

1990

US$

, PPP

)G

V01

51 c

ount

ries

2SL

SG

INI

coef

fici

ent;

Enr

ollm

ent r

atio

;N

-sha

ped

5,84

7; 1

8,87

7 (1

988

US$

, PPP

)19

98U

rban

pop

ulat

ion;

Popu

latio

nde

nsity

DG

03C

anad

aO

LS

Tim

e tr

end

Mon

oton

ical

ly in

crea

sing

N/A

1958

–199

5C

E03

32 c

ount

ries

Fixe

d E

ffec

tC

apita

l-la

bor

ratio

; Tim

e tr

end;

Mon

oton

ical

ly in

crea

sing

N/A

1975

–199

5T

rade

inte

nsity

(Tab

le 1

2 co

ntin

ued…

)

Page 23: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 191A

utho

rs†

Sam

ple

Met

hodo

logy

‡O

ther

exp

lana

tory

var

iabl

esR

elat

ions

hip

Tur

ning

poi

nts§

MB

0422

OE

CD

PMG

in E

CM

£L

agge

d de

pend

ent a

ndN

-sha

ped

5,24

4; 2

0,55

7 (1

997

US$

, PPP

)19

75–1

998

inde

pend

ent v

aria

ble

BS0

511

2 co

untr

ies

Sem

i-pa

ram

etri

cN

one

Mon

oton

ical

ly in

crea

sing

N/A

1950

–199

0G

L05

108

coun

trie

sPo

oled

OL

SN

one

Mon

oton

ical

ly in

crea

sing

N/A

1971

–199

5L

0524

OE

CD

3SL

SC

omm

erci

al e

nerg

y us

e (p

er c

apita

)M

onot

onic

ally

dec

reas

ing

N/A

1975

–199

0G

LP0

6O

EC

D P

oole

dO

LS

Non

eIn

vert

ed N

-sha

ped

932;

16,

882

(199

0 U

S$, P

PP)¢

1960

–199

72,

206;

8,3

85 (

1990

US$

, PPP

non-

OE

CD

Pool

ed O

LS

Non

eM

onot

onic

ally

incr

easi

ngN

/A¢

1971

–199

7N

/A£

OE

CD

Wei

bull

func

tion

Non

eE

KC

16,5

87 (

1990

US$

, PPP

)19

60–1

997

non-

OE

CD

Wei

bull

func

tion

Non

eM

onot

onic

ally

incr

easi

ngN

/A¢

1971

–199

7(T

able

12

cont

inue

d…)

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

Rat

e of

Cha

nge

T95

137

coun

trie

sPo

oled

OL

SPe

r ca

pita

inco

me

grow

th r

ate

Posi

tive

ly c

orre

late

dN

/A19

71–1

991

DV

O98

Net

herl

and,

OL

SL

agge

d in

com

e pe

r ca

pita

Neg

ativ

ely

corr

elat

edN

/AU

K, U

SA, a

ndR

ate

of c

hang

e in

ene

rgy

pric

es;

Wes

t Ger

man

yPe

r ca

pita

inco

me

grow

th r

ate

1961

–199

3† P

leas

e re

fer

to th

e ab

brev

iati

on o

f au

thor

s an

d as

soci

ated

dat

a so

urce

in T

able

13.

‡ The

cho

ice

of m

etho

dolo

gy is

bas

ed o

n th

e H

ausm

an te

st if

mor

e th

an o

ne e

stim

atio

n te

chni

que

is r

epor

ted.

§ My

calc

ulat

ions

of

turn

ing

poin

ts a

re b

ased

on

the

resu

lts f

rom

ori

gina

l pap

ers

if n

o tu

rnin

g po

int r

epor

ted.

§ Tur

ning

poi

nts

are

repo

rted

seq

uent

iall

y w

hen

the

rela

tions

hip

is c

ubic

.£ P

oole

d M

ean

Gro

up E

stim

atio

n of

an

Err

or C

orre

ctio

n M

odel

wit

h la

g le

ngth

1.

¢ The

res

ult i

s ba

sed

on th

e da

ta s

ourc

e fr

om I

nter

nati

onal

Ene

rgy

Age

ncy,

OE

CD

.£ T

he r

esul

t is

base

d on

the

data

sou

rce

from

Oak

Rid

ge N

atio

nal L

abor

ator

y.* T

he C

O2

OL

S re

gres

sion

est

imat

ed b

y D

G03

(D

ay a

nd G

raft

on (

2003

)) m

ay b

e m

is-s

peci

fied

sin

ce it

fai

ls s

ome

key

F-t

ests

.* P

PP: P

urch

asin

g Po

wer

Par

ity a

djus

ted.

* OL

S: O

rdin

ary

Lea

st S

quar

es; 2

SLS:

Tw

o-st

age

Lea

st S

quar

es; 3

SLS:

Thr

ee-s

tage

Lea

st S

quar

es.

Page 24: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

192 Yi-Chia Wang

emissions and income percapita. Using a non-linear three-parameter Weibull function andthe generalized method of moments (GMM) estimation, Galeotti et al. (2006) found a CO2

EKC pattern in OEC Dcountries with a turning point 16,587 US$, while carbon emissionsare monotonically increasing in selected non-OECD countries.

A rough picture of the CO2 emission–income relationship canalso be drawn byinvestigating the correlation between the grow thrates of carbon emissions and GDP percapita. Tucker (1995), for example, found a positive correlation between the growthratesof CO2 emissions and GDP for 137 economies. On the contrary, four developedcountries, Netherland, the United Kingdom, the United States and West Germany, werefound to have a negative correlation between the two growth rates from 1961 to 1993.

In summary, carbon emissions have drawn great attention from EKC researchers.Recent EKC studies tend to argue that green house gasemissions are global pollutants withthe highest abatement costsamong all sources of environmental degradation. Thus,carbone missions are very likely to grow continuously so long as the economy keepsdeveloping, unless vintage technology of carbon emission control is introduced andcompliance with the international treaty is enforced. In addition, the process of afforestation may soak up a large amount of CO2 emissions, but this effect is notyet included inany EKC studies because of the difficulty in its quantitative measurement.

3. WATER POLLUTION

Water pollution is typically a local degradation of the environment when an economy isdeveloping. Undoubtedly, access to clean water resources is directly related to humanhealth and reflects acommunity’s living standard. In general, the abatement of waterpollution is relatively easy to achieve because the (opportunity) costs of developing newtechnology and regulation enforcement are lower than most air pollutants.

Levels of fecal coliform, nitrates, dissolved oxygen (DO), biochemical oxygendemand (BOD) and chemical oxygen demand (COD) arecommon indicators representingcity water quality. In fact, they areclosely related to each other. Naturally, there exists aconstantamount of oxygen produced by plants’ photo synthesis as well as water churningin rivers. The sources of oxygen consumption mainly includethe respiration of aquaticanimals, decomposition of bacteria, andvarious chemical reactions. If oxygenconsumption in a stream system exceeds its production, DO levels fall and some sensitiveanimals may move away, weaken, or even die. Overall, the natural processes of oxygenproduction and consumption are balanced in the ecological system without considering theexternal damage resulting from human activities.

So why and how does oxygen consumption increase in rivers? Wastewater fromsewage treatment plants, on-site septic systems, domestic and wild animal manure, andstorm runoff often contain organic materials and fecal bacteria (coliform and fecalstreptococci). These by products are decomposed by micro organisms that exhaust oxygenin the process. The amount of oxygen consumed by the seorganisms for decomposing the

Page 25: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 193

waste is called biochemical oxygen demand (BOD). Similarly, chemical oxygen demand(COD) is the amount of oxygen consumed by chemical oxidization for converting organicwater constituents to inorganic end products. This chemical process also consumes oxygenin rivers. Other sources of oxygen-exhausting waste include storm water run off from farmland and urban streets, feedlots, and failing septic systems.

Fecal coliforms are a group of bacteria. Although they are notharmful themselves, theycould carry pathogenic viruses and protozoans that can live in human and animal’sdigestive systems. Therefore, their presence in streams reveals that pathogenic microorganisms might also be carried by shellfish so that eating aquatic animals from fecalcoliform contaminated streams might be ahealth risk. In addition, the decomposition offecal coliforms requires the demand for oxygen, which in turn reduces DO levels andharms rivers.

Nitrate contamination is also causing serious damage to the aquatic system. Theevolution of nitrates comes from nitrogen, including ammonia (NH3), nitrates (NO3), and

Table 13List of CO EKC Authors and Associated Pollution Data Sources

Abbreviation References Source of CO2 data

AC99 Agras and Chapman (1999) Oak Ridge National LaboratoryBS05 Bertinelli and Strobl (2005) World Development IndicatorsCE03 Cole and Elliott (2003) Oak Ridge National LaboratoryCJM97 Carson et al. (1997) Environmental Protection Agency, The United StatesCRB97 Cole et al. (1997) Oak Ridge National LaboratoryDG03 Day and Grafton (2003) Pollution Data Branch, Environment CanadaDVO98 de Bruyn et al. (1998) Oak Ridge National LaboratoryGL99 Galeotti and Lanza (1999) International Energy Agency, OECDGL05 Galeotti and Lanza (2005) International Energy Agency, OECDGLP06 Galeotti et al. (2006) Oak Ridge National Laboratory

International Energy Agency, OECDGV01 Gangadharan and Valenzuela (2001) World Development IndicatorsHES95 Holtz-Eakin and Selden (1995) Oak Ridge National LaboratoryL05 Liu (2005) no reportMB04 Martínez-Zarzoso and no report

Bengochea-Morancho (2004)MU97 Moomaw and Unruh (1997) Oak Ridge National LaboratoryRG97 Roberts and Grimes (1997) Oak Ridge National LaboratoryUM98 Unruh and Moomaw (1998) Oak Ridge National LaboratoryS94 Shafik (1994) Oak Ridge National LaboratoryS99 Sun (1999) World Development Report, World BankSB92 Shafik and Bandyopadhyay (1992) World Development Report, World BankSSJ98 Schmalensee et al. (1998) Oak Ridge National LaboratoryT95 Tucker (1995) World Development Indicators

Page 26: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

194 Yi-Chia Wang

nitrites (NO2). Anthropogenic sources of nitrates mainly include waste water treatmentplants, run off from fertilized lawns, cropl and and animal manure storage areas, failing on-site septic systems, and industrial discharges that contain corrosion inhibitors. Nitrates areaquatic plants’ essential nutrients, but excess amounts of nitrates, together withphosphorus, can cause significant water problems inthe form of eutrophication, theconsequence of which will lead todramatic growth in aquatic plants (e.g. algae). Over-accumulation ofalgae absorbs the existing amount of oxygen and prevents sunlight fromreaching the deep layers of a river, which in turn expels andeven kills aquatic lives due tothe reduction in oxygen levels and temperatures.

The above-mentioned indicators of water quality are broadly recordedin the GEMS/water data archive, containing observations from up to287 stations in 58 countries from1991 onwards. Thus, it is the most preferred data source in recent water EKC studies.Other records ofwater quality indicators, such as the World Bank’s World DevelopmentReport, monitor the DO and fecal coliforms levels in major streamsfor more than 100countries, and were also used in early water EKC research.

A summary of several major water EKC studies is categorized in Table 14 according todifferent measurements of water quality. As can be seen, DO levels fall continuously asincome percapita rises in the early studies carried out by Shafik and Bandyopadhyay(1992) and Shafik (1994). Although Grossman and Krueger (1995) found that the DOEKC pattern does not significantly exist when taking into account mean temperature andlagged income variables. Torras and Boyce (1998) later proposed that DO levels mightincrease when income per capita reaches the level between 5,085 and 19,865 US$.

Choosing a similar time span, BOD and COD levels, on the other hand,were found tohave no EKC phenomenon by Grossman and Krueger (1995) for 58 countries and Vincent(1997) in Malaysia. However, Hettige et al. (2000) and Cole and Elliott (2003) bothapplied fixed effect estimation and generated a U-shaped relationship between BODemissions and income per capita but with distinct turning points, 5,000 (1990 US$) and39,698 (1987 US$), respectively.

In terms of fecal coliform, two successive studies, Shafik and Bandyopadhyay(1992) and Shafik (1994), used different data sources (World Development Report andCanada Centerfor Inland Water, respectively) but generated a similar N-shapedrelationship between fecal coliform levels and income per capita. However, some laterfindings from Grossman and Krueger (1995) and Torras and Boyce (1998) indicated thatincreasing income per capitahas no significant correlation with the levels of fecalcoliform inrivers. A similar insignificant relationship between nitrates and income wasalso found by Grossman and Krueger (1995) for 58 countries, most of which were intheir developing stage. In the 15OECD countries selected by Cole et al. (1997),however, a30-river sample produced an EKC pattern between nitrates levels and incomeper capita with two turning points, 25,000 US$ from logarithm specification of income,and 15,600 US$ from level specification.

Page 27: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 195T

able

14

Ove

rvie

w o

f EK

C S

tudi

es in

the

Rel

atio

nshi

p be

twee

n W

ater

Pol

luti

on I

ndic

ator

s an

d In

com

e P

er C

apit

a

Aut

hors

†Sa

mpl

eM

etho

dolo

gy‡

Oth

er e

xpla

nato

ry v

aria

bles

Rel

atio

nshi

pT

urni

ng p

oint

Dis

solv

ed O

xyge

nSB

9214

9 co

untr

ies

Fixe

d E

ffec

tD

emoc

ratic

dum

my;

Tim

e tr

end;

Mon

oton

ical

lyN

/A19

60–1

990

Tra

de in

tens

ity d

ecre

asin

gS9

414

9 co

untr

ies

Fixe

d E

ffec

tT

ime

tren

dM

onot

onic

ally

dec

reas

ing

N/A

¢

1960

–199

0G

K95

58 c

ount

ries

Ran

dom

Eff

ect

Mea

n te

mpe

ratu

re; T

ime

tren

d;In

sign

ific

ant

N/A

1970

–199

0L

agge

d in

com

e pe

r ca

pita

TB

9858

cou

ntri

esPo

oled

OL

ST

ime

tren

d; G

INI

coef

fici

ent;

Inve

rted

N-s

hape

d5,

085;

19,

865

(198

5 U

S$, P

PP)

1977

–199

4M

ean

wat

er te

mpe

ratu

re; L

itera

cy;

Polit

ical

rig

htB

ioch

emic

al O

xyge

n D

eman

d (B

OD

)G

K95

58 c

ount

ries

Ran

dom

Eff

ect

Mea

n te

mpe

ratu

re; T

ime

tren

d;In

sign

ific

ant

N/A

1970

–199

0L

agge

d in

com

e pe

r ca

pita

V97

Cro

ss s

ectio

n,R

ando

m E

ffec

tPo

pula

tion

dens

ity; T

ime

tren

dIn

sign

ific

ant

N/A

Mal

aysi

a la

te19

70s–

earl

y19

90s

HM

W00

13 c

ount

ries

Fixe

d E

ffec

tT

ime

tren

dU

-sha

ped

5,00

0 (1

990

US$

, PPP

)19

75–1

994

GV

0151

cou

ntri

es2S

LS

GIN

I co

effi

cien

t; E

nrol

lmen

t rat

io;

Insi

gnif

ican

tN

/A19

98U

rban

pop

ulat

ion;

Pop

ulat

ion

dens

ityC

E03

32 c

ount

ries

Fixe

d E

ffec

tC

apita

l-la

bor

ratio

; Tim

e tr

end;

U-s

hape

d39

,698

(19

87 U

S$, P

PP)

1975

–199

5T

rade

inte

nsity

Che

mic

al O

xyge

n D

eman

d (C

OD

)G

K95

58 c

ount

ries

Ran

dom

Eff

ect

Mea

n te

mpe

ratu

re; T

ime

tren

d;In

sign

ific

ant

N/A

1970

–199

0L

agge

d in

com

e pe

r ca

pita

V97

Cro

ss s

ectio

n,R

ando

m E

ffec

tPo

pula

tion

dens

ity; T

ime

tren

d;In

sign

ific

ant

N/A

Mal

aysi

a la

te19

70s–

earl

y19

90s

(Tab

le 1

4 co

ntin

ued…

)

Page 28: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

196 Yi-Chia WangA

utho

rs†

Sam

ple

Met

hodo

logy

‡O

ther

exp

lana

tory

var

iabl

esR

elat

ions

hip

Tur

ning

poi

nts§

Fec

al C

olif

orm

SB92

149

coun

trie

sFi

xed

Eff

ect

Dem

ocra

tic d

umm

y; T

ime

tren

d;N

-sha

ped

1,20

0; 1

1,40

0 (1

985

US$

, PPP

)19

60–1

990

Tra

de in

tens

ityS9

414

9 co

untr

ies

Fixe

d E

ffec

tT

ime

tren

dN

-sha

ped

1,37

5; 1

1,50

0 (1

985

US$

, PPP

) £

1960

–199

0G

K95

58 c

ount

ries

Ran

dom

Eff

ect

Mea

n te

mpe

ratu

re; T

ime

tren

d;In

sign

ific

ant

N/A

1970

–199

0L

agge

d in

com

e pe

r ca

pita

TB

9852

cou

ntri

esPo

oled

OL

ST

ime

tren

d; G

INI

coef

fici

ent;

Insi

gnif

ican

tN

/A19

77–1

994

Mea

n w

ater

tem

pera

ture

; Lite

racy

;Po

litic

al r

ight

Nit

rate

sG

K95

58 c

ount

ries

,R

ando

m E

ffec

tM

ean

tem

pera

ture

; Tim

e tr

end;

Insi

gnif

ican

tN

/A19

70–1

990

Lag

ged

inco

me

per

capi

taC

RB

9715

OE

CD

Ran

dom

Eff

ect

Non

eE

KC

15,6

00–2

5,00

0 (1

985

US$

, PPP

)19

75–1

990

† Ple

ase

refe

r to

the

abbr

evia

tion

of

auth

ors

and

asso

ciat

ed d

ata

sour

ce in

Tab

le 1

5.‡ T

he c

hoic

e of

met

hodo

logy

is b

ased

on

the

Hau

sman

test

if m

ore

than

one

est

imat

ion

tech

niqu

e is

rep

orte

d.§ M

y ca

lcul

atio

ns o

f tu

rnin

g po

ints

are

bas

ed o

n th

e re

sults

fro

m o

rigi

nal p

aper

s if

no

turn

ing

poin

t rep

orte

d.§ T

urni

ng p

oint

s ar

e re

port

ed s

eque

ntia

lly

whe

n th

e re

latio

nshi

p is

cub

ic.

£ The

res

ult i

s ba

sed

on th

e da

ta s

ourc

e fr

om C

anad

a C

ente

r fo

r In

land

Wat

er.

¢ The

res

ult i

s ba

sed

on th

e da

ta s

ourc

e fr

om W

orld

Dev

elop

men

t Rep

ort,

Wor

ld B

ank.

* PPP

: Pur

chas

ing

Pow

er P

arity

adj

uste

d.* O

LS:

Ord

inar

y L

east

Squ

ares

; 2SL

S: T

wo-

stag

e L

east

Squ

ares

.

Page 29: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 197

All in all, the EKC relationship between various water pollutants and income per capitais inconsistent. The quality of a river may presumably be affected not only by humanactivities (contaminationand abatement) but also by its characteristics, such as watertemperature and mobility. One can also infer that the selection of sample rivers and periodmay also determine the existence of water EKC hypothesis.

4. TIMBER HARVESTING

Unlike air and water pollution, deforestation does not directly affect human health, butavoiding it, or affore station, brings many benefits to the environment. First, worldwidevegetation is the lung of the Earth that absorbs an estimated 6.1 billion metric tons ofanthropogenic CO2 emissions per annum from the atmosphere in aso-called ‘CarbonCycle’.9 Thus, over-logging of trees worsens the problem of global warming. Second, theforest protects various kinds of end angered species, maintains a food chain among forestinhabitants and keeps ecological systems balanced. Third, treescreate a natural barrieragainst intense storms and rising tides. Last but not least, trees trap water in their roots sopeople canavoid disasters such as landslides. Deforestation therefore causesnothing butlong-run catastrophes to the whole ecosystem.

There are three underlying sources of deforestation: the desire to convert forest andtimber land areas to pasture and cropland forlive stock farming and agriculture; timberharvesting for architecture; and the extraction of underground resources and woodfuels. Ingeneral, population pressure is always emphasized as acrucial factor of the above sourcesof deforestation in a country’sdeveloping stage (Panayotou (1993) andCorpper andGriffiths (1994)). As an economy starts developing, itsforest areas gradually shrink butmay be recovered when people’smarginal utility of environmental amenity exceeds theconsumption of material goods. That is, growing income levels lead to a higher demand fortrees, national parks and the forest to beautify theenvironment, and therefore, it isreasonably in the field of EKC research.

Early forest-related EKC studies are in favor of adopting an annual rate ofdeforestation as a measure of environmental degradation.10 Panayotou (1993) and Shafik(1994), for example, used the records of forest areafrom World Bank’s WorldDevelopment Report for a number of countriesand generated an EKC pattern between therate of deforestation andincome per capita, though Panayotou’s estimated turning pointseemsto be downward biased due to the use of a market exchange rate andpooled OLStechniques (de Bruyn and Heintz (1999) and Stern and Common (2001)). The overview ofthis series of EKC studies is summarized in Table 16. In addition to the forest data from theWorld Bank, the United Nations’ Food and Agriculture Organization (FAO) also recordsthe forest area for many tropical countries located in Africa, Asia and Latin America.Corpper and Griffiths (1994), for example, used this data set and found a deforestationEKC pattern for selected countries in Africa and Latin America with sensible turningpoints, but this pattern is insignificant in Asia due to insufficient sample size andvariationin their fixed effect estimation. Koop and Tole (1999) also workedon the FAO’s

Page 30: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

198 Yi-Chia Wang

data set, pooled 76 tropical countries together, butfound an insignificant deforestation–income relationship. This insignificant result was also supported by Gangadharan andValenzuela (2001) using 2SLS estimation. Similarly, using annual rates of affore station(negative rates ofdeforestation), Antle and Heidebrink (1995) also claimed that there existsno significant relationship between growing trees and acountry’s wealth.

Although the problem of deforestation is part of the cost of economic development, itis difficult to take the effect of natural disasters into account. Torrential rains, for example,are commonand regular features in tropical areas and usually causecatastrophic landslidesand floods washing away soils and timbers. Therefore, the measure of deforestation seemsto be a problematic proxy of environmental degradation, especially in tropical regions.

5. PROGRESS OF THE ENVIRONMENTAL KUZNETS CURVE

Early environmental economists embarked on investigating the patterns of variouspollutants against growing income levels by asking three general questions: will theenvironmental pressure resulting from economic development also be reduced byeconomic development itself? If so, at what level of income per capita will any pollutant becurbed? Even when pollution levels are decreasing, will they rise again? At first, theyestimated a reduced-formregression for certain pollutants with levels, second- and third-order polynomial of income per capita to capture the curvature between the variables andgenerate plausible turning points. Later, they incorporate a number of non-incomeregressors on the extended topics of their interests. The inclusion of non-income regressorswas suggested by the fact that, in addition to increasing income percapita, there are otherpossible mechanisms in reducing pollution levels. These pollution-reducing channels inthe EKC literature arelisted as follows:

(1) Trade Effect: by importing pollution-intensive products, or by moving heavyfactories to adjacent countries with less strict environmental regulations, high-income economies can reduce their pollution levels through these channels known

Table 15List of Water EKC Authors and Associated Data Sources

Abbreviation References Source of Water-quality Indicator

CE03 Cole and Elliott (2003) World Development IndicatorsCRB97 Cole et al. (1997) Environmental Data Compendium 1993, Paris: OECDGK95 Grossman and Krueger (1995) Global Environmental Monitoring SystemGV01 Gangadharan and Valenzuela (2001) World Development IndicatorsHMW00 Hettige et al. (2000) Individual Country’s Environmental Protection AgencySB92 Shafik and Bandyopadhyay (1992) World Development Report, World BankS94 Shafik (1994) Canada Center for Inland Water

World Development Report, World BankTB98 Torras and Boyce (1998) Global Environmental Monitoring SystemV97 Vincent (1997) Department of Environment, Malaysia

Page 31: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 199T

able

16

Ove

rvie

w o

f EK

C S

tudi

es in

the

Rel

atio

nshi

p be

twee

n A

nnua

l Rat

e of

Def

ores

tati

on a

nd I

ncom

e P

er C

apit

a

Aut

hors

Sam

ple

Met

hodo

logy

‡O

ther

exp

lana

tory

var

iabl

esR

elat

ions

hip

Tur

ning

poi

nts§

SB92

149

coun

trie

sFi

xed

Eff

ect

Dem

ocra

tic d

umm

y; T

ime

tren

d;E

KC

2,00

0 (1

985

US$

, PPP

)19

60–1

990

Tra

de in

tens

ity

P93

55 c

ount

ries

Pool

ed O

LS

Tro

pica

l dum

my;

Pop

ulat

ion

dens

ityE

KC

1,20

0 (1

985

US$

, ME

R)

late

198

0s

CG

9462

cou

ntri

esFi

xed

Eff

ect

Popu

latio

n gr

owth

rat

e;A

fric

a: E

KC

4,76

0 (1

991

US$

, PPP

)19

61–1

991

Popu

latio

n de

nsity

;T

imbe

r pr

ice;

Tim

e tr

end

Lat

in A

mer

ica:

EK

C5,

420

(199

1 U

S$, P

PP)

Asi

a: I

nsig

nifi

cant

N/A

S94

149

coun

trie

sFi

xed

Eff

ect

Tim

e tr

end

EK

C3,

380

(198

5 U

S$, P

PP)

1960

–199

0

AH

95£

93 c

ount

ries

Pool

ed O

LS

Cou

ntry

’s to

tal a

rea;

Pop

ulat

ion;

Insi

gnif

ican

tN

/A19

87–1

991

Fore

st a

rea

KT

9976

cou

ntri

esFi

xed

Eff

ect

Popu

latio

n gr

owth

rat

e; P

opul

atio

n de

nsity

;In

sign

ific

ant

N/A

1961

–199

2G

DP

grow

th r

ate

GV

0151

cou

ntri

es2S

LS

GIN

I co

effi

cien

t; E

nrol

lmen

t rat

io;

Insi

gnif

ican

tN

/A19

98U

rban

pop

ulat

ion;

Pop

ulat

ion

dens

ity

Abb

revi

atio

nR

efer

ence

sSo

urce

of R

elev

ant d

ata

AH

95A

ntle

and

Hei

debr

ink

(199

5)W

orld

Res

ourc

es I

nstit

ute

CG

94C

orpp

er a

nd G

riff

iths

(199

4)Fo

od a

nd A

gric

ultu

re O

rgan

izat

ion,

Uni

ted

Nat

ions

GV

01G

anga

dhar

an a

nd V

alen

zuel

a (2

001)

Wor

ld D

evel

opm

ent I

ndic

ator

sK

T99

Koo

p an

d T

ole

(199

9)Fo

od a

nd A

gric

ultu

re O

rgan

izat

ion,

Uni

ted

Nat

ions

P93

Pana

yoto

u (1

993)

Wor

ld D

evel

opm

ent R

epor

t, W

orld

Ban

kS9

4Sh

afik

(19

94)

Wor

ld D

evel

opm

ent R

epor

t, W

orld

Ban

kSB

92Sh

afik

and

Ban

dyop

adhy

ay (

1992

)W

orld

Dev

elop

men

t Rep

ort,

Wor

ld B

ank

‡ The

cho

ice

of m

etho

dolo

gy is

bas

ed o

n th

e H

ausm

an te

st if

mor

e th

an o

ne e

stim

atio

n te

chni

que

is r

epor

ted.

§ My

calc

ulat

ions

of

turn

ing

poin

ts a

re b

ased

on

the

resu

lts f

rom

ori

gina

l pap

ers

if n

o tu

rnin

g po

int r

epor

ted.

§ Tur

ning

poi

nts

are

repo

rted

seq

uent

iall

y w

hen

the

rela

tions

hip

is c

ubic

.£A

H95

(A

ntle

and

Hei

debr

ink

(199

5))

used

ann

ual r

ate

of a

ffor

esta

tion

(neg

ativ

e of

def

ores

tati

on)

as th

e de

pend

ent v

aria

ble.

* PPP

: Pur

chas

ing

Pow

er P

arity

adj

uste

d; M

ER

: Mar

ket E

xcha

nge

Rat

e ad

just

ed.

* OL

S: O

rdin

ary

Lea

st S

quar

es; 2

SLS:

Tw

o-st

age

Lea

st S

quar

es.

Page 32: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

200 Yi-Chia Wang

Figure1: Different Patterns between Environmental Pressure and Income Per Capita

(1) A monotonically increasing pattern where β1 > 0 and β

2 = β

3 = 0,

(2) A monotonically decreasing pattern where β1 < 0 and β

2 = β

3 = 0,

(3) An inverted U-shaped EKC relationship where β1 > 0, β

2 < 0 and β

3 = 0,

(4) A U-shaped curve represents β1 < 0, β

2 > 0, and β

3 = 0,

(5) An N-shaped pattern reveals β1 > 0, β

2 < 0 and β

3 > 0, and

(6) An inverted N-shaped curve represents β1 < 0, β

2 > 0 and β

3 < 0.

Note: Given the significance of all βi′s, two conditions —|β

1| > |β

2| > |β

3| and β2

2> 3β

3 — must hold to

have distinct turning points.

as ‘displacement hypothesis’. Grossman and Krueger (1991) and Shafik andBandyopadhyay (1992) both concluded that higher physical trade intensity willleadto lower concentrations of air pollutants. Rather than using physical trademeasures such as export and import, Hettige et al. (1992) adopted the Dollar indexas the proxy to the openness of trade and reached similar conclusions in the issueof toxic intensity.

(2) Stringent Environmental Regulation: de Bruyn (1997) and Torras and Boyce(1998) proposed that the environmental policy, fostered by internationalagreements such as the Kyoto Protocol and emissions-trading scheme in theEuropean Union, explains why pollution can be successfully curbed in high-income nations. It is especially noticed that some local pollutants can alsobemitigated by the environmental enforcement of rule of law (Panayotou (1997)and Aslanidis and Xepapadeas (2006)).

Page 33: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 201

(3) Population Effect: a city with higher population density creates higher demandfor energy and causes unavoidable pollution and deforestation. Therefore, thereduction in population intensity has been found to have significant contributionsto curb SO2 concentrations (Grossman and Krueger (1991)) andairborne emissionsof CO, NOX, SPM and SO2 (Selden and Song (1994)), as well as discouragedeforestation (Corpper and Griffiths (1994)).

(4) Technology Progress: most of the decomposition analyses (discussed later) tendto believe that technological innovation is the most crucial factor that curbspollution. Hilton and Levinson (1998) found that the EKC phenomenon exists inautomotive lead emissions because the technological reductionin lead componentsper gallon of gasoline (pollution intensity) gradually offsets the increase of totalgasoline consumption (pollution activity).Stern (2004)also reached the conclusionthat emissions-related technology progress leads to the reduction in sulfuremissions in his selected OECD countries.

(5) Input Mix and Output Mix Effects: rapid progress in technology may be helpfulto use less pollution-intensive input, or reduce the production of environmentallytaxing output. The former was upheld by Stern (2002) with the conclusion that thereduction of energy intensity (energy consumption) contributes to 10.2%decreases of global sulfur emissions during 1973–90, while the latter was hardlydiscussed with empirical evidence.

(6) Energy Price Effect: the reduction in energy consumption can also be attributedto increasing energy prices. Unruh and Moomaw (1998) observed that the oil crisisin 1973 had triggered the reduction in CO2 emissions per capita in selected oil-dependent high-income countries. Similarly, de Bruyn et al. (1998) indicated thatenergy prices and CO2 emissions are negatively correlated in growth rates in theUnited States during the period 1961–1993.

(7) Income Inequality Effect: the absolute income effect suggests that rising incomeper capita will increase the capability to pay for environmental amenity. However,the relative income effect will reduce this capability since rising income inequalityresulting from increased income per capita may reduce people’s willingness toprotectthe environment (Magnani (2000)). Thus, the improvement of a country’sincome inequality may create median voters’ incentive towards environmentalprotection.

After the EKC history passes the stage in which people adopted miscellaneousexplanatory variables listed above to capture different aspects of pollution-reducingmechanisms, studies at thenext stage started to criticize existing results and some of themalso proposed possible corrections for conventional EKC regressions. In general,econometric problems were mainly criticized at this stage. First, the possibility of jointdetermination of income and environmental degradation will result in inconsistentestimation when suitable instrumental variables are notconsidered.11 That is, notonly can

Page 34: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

202 Yi-Chia Wang

income per capita affect the level of pollutants but the irreverse causality may exist as well.Second, without the correction of heteroskedasticity and serial correlation in regressionerrors, the estimated coefficients may be accompanied by small standarderrors anddistorted t-statistics such that the significance of coefficients may be unreliable.

Pooled OLS estimation has also been criticized for its omitted variable problem.Therefore, most EKC researchers are in favor ofusing fixed effect estimation to eliminateomitted variable bias. Holtz-Eakin and Selden (1995), for example, found an EKCrelationship between CO2 emissions and income per capita byusing pooled OLS and fixedeffect estimations, but there were huge differences in the turning points from these twotechniques, which were eight million US$ and 35,428 US$, respectively. Day and Grafton(2003) also revealed that their results of SPM and CO2 EKC regressions in the case studyof Canada failed to perform some key misspecification tests.

The recent generation of models and estimation techniques has been developed toavoid various problems when using conventional reduced-form EKC regression. Galeottiand Lanza (1999),12 forexample, estimated a non-linear Weibull function using maximumlikelihood estimation. Although this function is sometimes used inapplied environmentaland ecological economics to capture thenon-linear curvature between pollution andincome per capita, it isnot practical to include more than one explanatory variableforfurther discussion of other pollution-reducing mechanisms. Similarproblems exist inthe semi-parametric regression withoutrestrictions on the EKC functional form (Taskinand Zaim (2000) and Bertinelli and Strobl (2005), for example).

In order to distinguish different channels in which certainpollutants are growing,decomposition analysis has been used frequently in recent years. The idea ofdecomposition analysis canbe demonstrated as follows:

1 ( 2) ( 1),

1 2 ( 1)t t t t

t tt t t t

EP factor factor N factor NEP factor N

factor factor factor N factor N

− −= × × × × ×−

(2)

where the environmental pressure at time t (EPt) can bedecomposed by the product of $N$factor ratios. In general, most of the decomposition studies use factor (N–1)t and factor Nt

as GDP and total population, respectively. Therefore,( 1)t

t

factor N

factor N

− represents GDP per

capita (scale effect) and factor Nt denotes a population effect. Decompositionanalysischooses two distinct years for dividing the rate of change in EP into the sum of the rates ofchange in N factors. Takinglogarithm transformation and linearizing equation (2), wehave:

( 1)( ),

1t t

t tt t

EP factor NIn EP In In In factor N

factor factor N

−∆ = ∆ + + ∆ + ∆

(3)

Page 35: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 203

where ∆ denotes the difference of variables at between thetwo time points. The main meritof decomposition analysis is toexplicitly understand the sign and magnitude of thecontribution of every factor ratio to the growth rates of pollutants. Stern (2002), forexample, decomposed total change in global sulfur emissions into (1) the impact of shifts

in the mixed fossilfuel type ,global sulfur emissions

fossil fuel consumpon

(2) the shift between fossil and

non-fossil fuels ,fossil fuel consumption

total fuel consumption

(3) energy intensity ,total fuel consumption

totalGDP

(4) a scale effecttotalGDP

total population

and (5) apopulation effect (total population). His

calculation shows that thegrowth rate of global sulfur emissions is 28.77% from 1973 to1990,among which the contribution of scale effect consists of 53.78%,while the mainsources of the reduction in global sulfur emissionscome from the change of emissions-related technical progress (–19.86%) and the change of energy intensity (–10.2%).

Most decomposition analysts reach similar conclusions that growingincome per capitaleads to positive and significant change in sulfurand carbon emissions,13although thefactors chosen as decomposition components are fairly subjective. In addition, a linearizeddecomposition approach cannot capture a possible non-linear relationship betweenemissions and scale effects from increasing income per capita. This is the main reason whythis approach is not applicable in the EKC literature.

6. CONCLUSIONS

The relationship between environmental protection and economic development has longbeen theoretical and empirical issues. However, the measurement of environmental qualityis further complicated dueto its multi-dimensional nature (Antle and Heidebrink (1995)). Itis admitted that most of the ecological damage is difficult tomeasure and anticipate. Ofparticular importance are irreversible changes, including desertification, loss ofbiodiversity, soilerosion and the depletion of ground water reservoirs. Arrow et al. (1995)provided a short but elaborate discussion of this issue. Their two major conclusions are:

(1) If human activities are to be sustainable, we need to ensurethat the ecologicalsystems on which our economies depend areresilient.

(2) When a country has attained a sufficient high standard of living, people givegreater attention to environmental amenities. However, economic growth is not apanacea for environmental quality.

Even though measuring environmental quality is difficult, somedevelopment-relatedpollutants and deforestation can representvarious aspects of environmental degradation.This paper collectsand discusses numerous EKC studies according to different measuresof

Page 36: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

204 Yi-Chia Wang

environmental deterioration. Broadly speaking, the bell-shapedrelationship betweenincome per capita and the proxy toenvironmental pressure can be found in earlier EKCstudies, butrecent researchers tend to question their validity and applicabilityfrom manyperspectives. This paper concludes whether the EKChypothesis can be supported dependscritically on data sources,sample countries, length of time period, the adoption of amarketexchange rate or purchasing power parity to adjust GDP, regressionfunctionalforms and various estimation techniques.

Notes1. To date, the multi-country data of various air pollutants can be found in a joint project of the World

Health Organization (WHO) collaborated with the United Nations Environment Programme (UNEP) inoperating Global Environmental Monitoring System(GEMS) on air and water quality, Toxic ReleaseInventory (TRI) on toxic emissions from the US manufacturers, the long time-series dataof CO2emissions from the Oak Ridge National Laboratory (ORNL), the Environmental Data Compendiumcompiled by the OECD, WorldResources Institute (WRI) from Washington D.C., A.S.L. andAssociates’ yearly report for sulfur emissions from Washington D.C.,the emission of greenhouse gasesfrom the Department for Environment Food and Rural Affairs (DEFRA), various environmentalmeasures fromWorld Development Indicators (WDI) established by World Bank and a number ofenvironmental monitoring institutes in individual country.

2. Logarithm transformation of variables is also commonlyused.

3. To be precise, these are the variables excluding‘contemporaneous’ income per capita.

4. If 22 1 33 0,β − β β < there exists noturning point and environmental pressure tends to ‘increase’ as income

grows with curvature.

5. If 22 1 33 0,β − β β < there exists noturning point and environmental pressure tends to ‘decrease’ as income

grows with curvature.

6. Day and Grafton (2003) suggestedthat such an argument is one of many explanations of theEKCphenomenon in literature. They pointed out that this argument isinadequate, given the fact thatmany low- to middle-income countriesare at the per capita income levels lower than the reportedturningpoints in many EKC studies (collected by Ekins (2000)). Even inCanada, environmentalproblems are still an on-going issue.

7. Mass ofsolution is a homogeneous mixture of two or more substances.

8. National Energy InformationCenter (NEIC): http://www.eia.doe.gov/environment.html.Accessed: 1Apr. 2010.

9. The annual rate of deforestation is defined as, 1 ,

, 1

,i t i t

i t

F F

F−

− where F

i,t denotes thetotal forest area in

country i at time t.

10. This is the problem of simultaneity andirreversibility discussed byStern (1998). When thedependentvariable and regressors are endogenously and systematicallydetermined, both of them will becorrelated with regressionresiduals. In this situation, the estimated OLS coefficients are notconsistent.However if variables cointegrate in the long run, due tosuperconsistency, simultaneity issues aremitigated.

11. Also, seeGaleotti and Lanza (2005) andGaleotti et al. (2006).

12. These studies include Zheng (2000), Viguier (1999), Stern (2002), Hamilton and Turton (2002), Bruvolland Medin (2003) and Stern (2004).

Page 37: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 205

ReferencesAgras, J. and D. Chapman (1999), “A Dynamic Approach to the Environmental Kuznets Curve Hypothesis.”

Ecological Economics, 28(2), 267–277.

Antle, J. M. and G. Heidebrink (1995), “Environment and Development: Theory and InternationalEvidence.” Economic Development and Culture Change, 43(3), 603–625.

Antweiler, W., B. R. Copeland and M. S. Taylor (2001), “Is Free Trade Good for the Environment?”American Economic Review, 91(4), 877–908.

Arrow, K., B. Bolin, R. Costanza, P. Dasgupta, C. Folke, C. S. Holling, B. O. Jansson, S. Levin, K. G. Maler,C. Perrings and D. Pimental (1995), “Economic Growth, Carrying Capacity and the Environment.”Science, 268 (5210), 520–521.

Aslanidis, N. and A. Xepapadeas (2006), “Smooth Transition Pollution-Income Paths.” EcologicalEconomics, 57(2), 182–189.

Baldwin, R. 1995. Does Sustainability Require Growth? Cambridge, UK: Cambridge University Press: I.Goldin and L. A. Winters (eds.), The Economics of Sustainable Development.

Beckerman, W. B. (1972), “Economic Development and the Environment: a False Dilemma.”InternationalConciliation, 586, 57–71.

Beckerman, W. B. (1992), “Economic Growth and the Environment: Whose Growth? Whose Environment?”World Development, 20(4), 481–496.

Bertinelli, L. and E. Strobl (2005), “The Environmental Kuznets Curve Semi-parametrically Revisited.”Economics Letters, 88(3), 350–357.

Bruvoll, A. and H. Medin (2003), “Factors Behind the Environmental Kuznets Curve. A Decomposition ofthe Changes in Air Pollution.”Environmental and Resource Economics, 24(1), 27–48.

Carson, R. T., Y. Jeon and D. R. McCubbin (1997), “The Relationship between air Pollution Emissions andIncome: US Data.”Environment and Development Economics, 2(4), 433–450.

Cole, M. A. and R. J. R. Elliott (2003), “Determining the Trade-environment Composition Effect: the Role ofCapital, Labor and Environmental Regulations.” Journal of Environmental Economics andManagement, 46(3), 363–383.

Cole, M. A., A. J. Rayner and J. M. Bates (1997), “The Environmental Kuznets Curve: An EmpiricalAnalysis.” Environment and Development Economics, 2(4), 401–416.

Corpper, M. and C. Griffiths (1994), “The Interaction of Population Growth and Environmental Quality.”American Economic Review, 84(2), 250–254.

Day, K. M. and R. Q. Grafton (2003), “Growth and the Environment in Canada: An Empirical Analysis.”Canadian Journal of Agricultural Economics, 51, 197–216.

de Bruyn, S. M. (1997), “Explaining the Environmental Kuznets Curve: Structural Change and InternationalAgreements in Reducing Sulphur Emissions.” Environment and Development Economics, 2(4), 485–503.

de Bruyn, S. M. and R. J. Heintz (1999), “The Environmental Kuznets Curve Hypothesis,”in J. C. J. M. vanden Bergh (eds.) Handbook of Environmental and Resource Economics, Cheltenham: Edward Elgar.656–677.

de Bruyn, S. M., J. C. J. M. van den Bergh and J. B. Opschoor (1998), “Economic Growth and Emissions:Reconsidering the Empirical Basis of Environmental Kuznets Curves.” Ecological Economics, 25(2),161–175.

Dinda, S., D. Coondoo and M. Pal (2000), “Air Quality and Economic Growth: an Empirical Study.”Ecological Economics, 34(3), 409–423.

Ekins, P. (2000), Economic Growth and Environmental Sustainability: The Prospects for Growth. London,UK: Routledge.

Page 38: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

206 Yi-Chia Wang

Galeotti, M. and A. Lanza (2005), “Desperately Seeking Environmental Kuznets.”Environmental Modellingand Software, 20(11), 1379–1388.

Galeotti, M. and A. Lanza (1999), “Richer and Cleaner? A Study on Carbon Dioxide Emissions inDeveloping Countries.” Energy Policy, 27(10), 565–573.

Galeotti, M., A. Lanza and F. Pauli (2006), “Reassessing the Environmental Kuznets Curve for CO2

Emissions: A Robustness Exercise.” Ecological Economics, 57(1), 152–163.

Gangadharan, L. and M. R. Valenzuela (2001), “Interrelationships Between Income, Health and theEnvironment: Extending the Environmental Kuznets Curve Hypothesis.” Ecological Economics, 36(3),513–531.

Grossman, G. M. and A. B. Krueger (1995), “Economic Growth and the Environment.” Quarterly Journal ofEconomics, 110(2), 353–377.

Grossman, G. M. and A. B. Krueger (1991), “Environmental impacts of a North American Free TradeAgreement,” NBER Working Paper 3914, National Bureau of Economic Research (NBER), Cambridge.

Hamilton, C. and H. Turton (2002), “Determinants of Emissions Growth in OECD cCountries.” EnergyPolicy, 30(1), 63–71.

Hausman, J. A. (1978), “Specification Tests in Econometrics.”Econometrica, 46(6), 1251–1272.

Hettige, H., R. E. B. Lucas and D. Wheeler (1992), “The Toxic Intensity of Industrial Production: GlobalPatterns, Trends, and Trade Policy.”American Economic Review, 82(2), 478–481.

Hettige, H., M. Mani and D. Wheeler (2000), “Industrial Pollution in Economic Development: theEnvironmental Kuznets Curve Revisited.” Journal of Development Economics, 62(2), 445–476.

Hilton, F. G. H. and A. Levinson (1998), “Factoring the Environmental Kuznets Curve: Evidence fromAutomotive Lead Emissions.” Journal of Environmental Economics and Management, 35(2), 126–141.

Holtz-Eakin, D. and T. M. Selden (1995), “Stoking the Fires? CO2 Emissions and Economic Growth.”

Journal of Public Economics, 57(1), 85–101.

Kaufmann, R. K., B. Davidsdottir, S. Garnham and P. Pauly (1998), “The Determinants of Atmospheric SO2

Concentrations: Reconsidering the Environmental Kuznets Curve.” Ecological Economics, 25(2), 209–220.

Koop, G. and L. Tole (1999), “Is There an Environmental Kuznets Curve for Deforestation?” Journal ofDevelopment Economics, 58(1), 231–244.

Kuznets, S. (1955), “Economic Growth and Income Inequality.”American Economic Review, 49, 1–28.

List, J. A. and C. A. Gallet (1999), “The Environmental Kuznets Curve: Does One Size Fit All?” EcologicalEconomics, 31(3), 409–423.

Liu, X. (2005), “Explaining the Relationship between CO2 Emissions and National Income — The Role of

Energy Consumption.” Economics Letters, 87(3), 325–328.

Magnani, E. (2000), “The Environmental Kuznets Curve, Environmental Protection Policy and IncomeDistribution.” Ecological Economics, 32(3), 431–443.

Martínez-Zarzoso, I. and A. Bengochea-Morancho (2004), “Pooled mean Group Estimation of anEnvironmental Kuznets Curve for CO

2.” Economics Letters, 82(1), 121–126.

Meadows, D. H., J. R. Meadows and W. I. Behrens (1972), The Limits to Growth. New York: Universe Books.

Millimet, D. L., J. A. List and T. Stengos (2003), “The Environmental Kuznets Curve: Real Progress orMisspecified Models?” Review of Economics and Statistics, 85(4), 1038–1047.

Moomaw, W. R. and G. C. Unruh (1997), “Are Environmental Kuznets Curves Misleading Us? The Case ofCO

2 Emissions.” Environment and Development Economics, 2(4), 451–463.

Panayotou, T. (1997), “Demystifying the Environmental Kuznets Curve: Turning a Black box into a PolicyTool.” Environment and Development Economics, 2(4), 465–484.

Page 39: AN EMPIRICAL SURVEY OF TESTING THE …serialsjournals.com/serialjournalmanager/pdf/1343214889.pdf · An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 173

An Empirical Survey of Testing the Environmental Kuznets Curve Hypothesis 207

Panayotou, T. (1993), “Empirical Tests and Policy Analysis of Environmental Degradation at DifferentStages of Economic Development,” Discussion paper 1, Geneva: International Labour Office.

Perman, R. and D. I. Stern (2003), “Evidence from Panel Unit Root and Cointegration Tests that theEnvironmental Kuznets Curve does not Exist.” Australian Journal of Agricultural and ResourceEconomics, 47(3), 325–347.

Pezzey, J. C. V. (1989), “Economic Analysis of Sustainable Growth and Sustainable Development,”Environment Department Working Paper no. 15, World Bank.

Roberts, J. T. and P. E. Grimes (1997), “Carbon Intensity and Economic Development 1962–1991: A BriefExploration of the Environmental Kuznets Curve.” World Development, 25(2), 191–198.

Schmalensee, R., T. M. Stoker and R. A. Judson (1998), “World Carbon Dioxide Emissions: 1950–2050.”Review of Economics and Statistics, 80(1), 15–27.

Selden, T. M. and D. Song (1994), “Environmental Quality and Development: Is There a Kuznets Curve forAir Pollution Emissions?” Journal of Environmental Economics and Managements, 27(2), 147–162.

Shafik, N. (1994), “Economic Development and Environmental Quality: An Econometric Analysis.” OxfordEconomic Papers, 46, 757–773.

Shafik, N. and S. Bandyopadhyay (1992), “Economic Growth and Environmental Quality: Time Series andCross-country Evidence,” World Bank Background Papers, no. 904, Washington, DC.

Simon, J. L. (1981), The Ultimate Resource. Princetion, NJ: Princetion University Press.

Stern, D. I. (2002), “Explaining Changes in Global Sulfur Emissions: An Econometric DecompositionApproach.” Ecological Economics, 42(1-2), 201–220.

Stern, D. I. (2005), “Global Sulfur Emissions from 1850 to 2000.” Chemosphere, 58(2), 163–175.

Stern, D. I. (1998), “Progress on the Environmental Kuznets Curve?” Environment and DevelopmentEconomics, 3(2), 173–196.

Stern, D. I. (2004), “The Rise and Fall of the Environmental Kuznets Curve.” World Development, 32(8),1419–1439.

Stern, D. I. and M. S. Common (2001), “Is There an Environmental Kuznets Curve for Sulfur?” Journal ofEnvironmental Economics and Management, 41, 162–178.

Sun, J. W. (1999), “The Nature of CO2 Emission Kuznets Curve.”Energy Policy, 27(12), 691–694.

Taskin, F. and O. Zaim (2000), “Searching for a Kuznets Curve in Environmental Efficiency Using KernelEstimation.”Economics Letters, 68(2), 217–223.

Torras, M. and J. K. Boyce (1998), “Income, Inequality, and Pollution: a Reassessment of the EnvironmentalKuznets Curve.” Ecological Economics, 25(2), 147–160.

Tucker, M. (1995), “Carbon Dioxide Emissions and Global GDP.” Ecological Economics, 15(3), 215–223.

Unruh, G. C. and W. R. Moomaw (1998), “An Alternative Analysis of Apparent EKC-type Transitions.”Ecological Economics, 25(2), 221–229.

Viguier, L. (1999), “Emissions of SO2, NO

X and CO

2 in Transition Economies: Emission Inventories and

Divisia Index Analysis.” Energy Journal, 20(2), 59–87.

Vincent, J. R. (1997), “Testing for Environmental Kuznets Curves within a Developing Country.”Environment and Development Economics, 2(4), 417–431.

Wildavsky, A. (1988), Searching for Safty. New Brunswick, N.J.: Transaction Books.

Zheng, Z. (2000), “Decoupling China’s Carbon Emissions Increase from Economic Growth: An EconomicAnalysis and Policy Implications.” World Development, 28(4), 739–752.