Ekc stern

Environmental Kuznets Curve

DAVID I. STERNRensselaer Polytechnic Institute

Troy, New York, United States

1. Introduction

2. Theoretical Background

3. Econometric Framework

4. Results of EKC Studies

5. Theoretical Critique of the EKC

6. Econometric Critique of the EKC

7. Other Evidence

8. Decomposing Emissions

9. Conclusions

Glossary

cointegration Classical regression analysis assumes thatthe error term in a regression is normally andindependently distributed. An extreme violation of thiscondition is when the error term is a random walk. Sucha regression is termed a spurious regression and theresults are not reliable. It usually indicates that arequired variable with random walk behavior has beenomitted from the model or that the variables in themodel follow random walks and are not related to eachother.

consistent estimation May show bias and inefficiency,but both are reduced toward zero as the sample sizeincreases.

developed economies Countries with high income levels;typically have an income per capita greater than U.S.$10,000, life expectancy greater than 70 years, andliteracy rates greater than 95%.

developing economies Countries with low- and middle-income levels; equivalent to the term ‘‘Third World.’’

econometrics The specialized branch of mathematicalstatistics applied to the analysis of economic data.

economic growth An increase in economy-wide economicproduction usually measured by an increase in grossdomestic product; also, the process of the economygrowing over time.

environmental degradation Any human-caused deteriora-tion in environmental quality.

externality When a production or consumption activityhas unintended damaging effects, such as pollution, on

other firms or individuals and no compensation is paidby the producer or consumer to the affected parties(negative externality); when activities have beneficialeffects for others, such as freely available researchresults, and the recipients do not pay the generator ofthe effect (positive externality).

heteroskedasticity When the variance of the error term in aregression model varies systematically across observa-tions or cases.

income elasticity The percentage increase in some variablewhen income increases by 1%.

income per capita Usually computed as the gross nationalproduct of a country for one year divided by the totalpopulation.

integrated variable A random walk is an integratedvariable—the current value is equal to the previousperiod’s value plus a random shock. This means that theeffect of these shocks on the future does not decline overtime. Rather, the current value of the variable is anintegration of all past shocks.

omitted variables bias Occurs when variables that shouldbe in a regression model are omitted and they arecorrelated with those variables that are included.

panel data A data set that includes observations of anumber of individuals, countries, firms, and the likeover a number of time periods such as months or years.

simultaneity bias When there is mutual feedback betweentwo variables and ordinary regression analysis cannotprovide consistent estimates of model parameters.

sustainable development If the current average well-beingof people could be maintained into the indefinite future.

technological change The invention and introduction ofnew methods of production as well as new products.

The environmental Kuznets curve is a hypothesizedrelationship among various indicators of environ-mental degradation and income per capita. Duringthe early stages of economic growth, degradationand pollution increase, but beyond some level ofincome per capita (which will vary for differentindicators) the trend reverses, so that at high incomelevels economic growth leads to environmentalimprovement. This implies that the environmental

Encyclopedia of Energy, Volume 2. r 2004 Elsevier Inc. All rights reserved. 517

impact indicator is an inverted U-shaped function ofincome per capita. Typically, the logarithm of theindicator is modeled as a quadratic function of thelogarithm of income.

1. INTRODUCTION

An example of an estimated environmental Kuznetscurve (EKC) is shown in Fig. 1. The EKC is named forSimon Kuznets, who hypothesized that income in-equality first rises and then falls as economic develop-ment proceeds. Emissions of various pollutants,such as carbon dioxide, sulfur, and nitrogen oxides,are tightly coupled to the use of energy. Hence, theEKC is a model of the relationship among energy use,economic growth, and the environment.

The EKC is an essentially empirical phenomenon,but most of the EKC literature is statistically weak. Itis very easy to do bad econometrics, and the historyof the EKC exemplifies what can go wrong. The EKCidea rose to prominence because few paid sufficientattention to the relevant diagnostic statistics. Little orno attention has been paid to the statistical proper-ties of the data used such as serial dependence andrandom walk trends in time series, and few tests ofmodel adequacy have been carried out or presented.However, one of the main purposes of doingeconometrics is to test which apparent relationshipsare valid and which are spurious correlations.

When we do take such statistics into account anduse appropriate techniques, we find that the EKC doesnot exist. Instead, we get a more realistic view of theeffect of economic growth and technological changeson environmental quality. It seems that most indica-

tors of environmental degradation are monotonicallyrising in income, although the ‘‘income elasticity’’ isless than 1.0 and is not a simple function of incomealone. Time-related effects, intended to model tech-nological change common to all countries, reduceenvironmental impacts in countries at all levels ofincome. However, in rapidly growing middle-incomecountries, the scale effect, which increases pollutionand other degradation, overwhelms the time effect. Inwealthy countries, growth is slower and pollutionreduction efforts can overcome the scale effect. This isthe origin of the apparent EKC effect.

The econometric results are supported by recentevidence that, in fact, pollution problems are beingaddressed and remedied in developing economies.

This article follows the development of the EKCconcept in approximately chronological order. Sec-tions II and III review in more detail the theorybehind the EKC and the econometric methods usedin EKC studies. Sections IV to VI review some EKCanalyses and their critiques. Section VII discusses therecent evidence from Dasgupta and colleagues andothers that has changed the way in which we viewthe EKC. The final two sections discuss an alternativeapproach, decomposition of emissions, and summar-ize the findings.

2. THEORETICAL BACKGROUND

The EKC concept emerged during the early 1990swith Grossman and Krueger’s pathbreaking study ofthe potential impacts of the North American FreeTrade Agreement (NAFTA) and Shafik and Bandyo-padhyay’s background study for the World Develop-ment Report 1992. However, the idea that economicgrowth is necessary for environmental quality to bemaintained or improved is an essential part of thesustainable development argument promulgated bythe World Commission on Environment and Devel-opment in Our Common Future.

The EKC theme was popularized by the Interna-tional Bank for Reconstruction and Development’s(IBRD) World Development Report 1992, whichargued that ‘‘the view that greater economic activityinevitably hurts the environment is based on staticassumptions about technology, tastes, and environ-mental investments’’ and that ‘‘as incomes rise, thedemand for improvements in environmental qualitywill increase, as will the resources available forinvestment.’’ Others have expounded this positioneven more forcefully, with Beckerman claiming that‘‘there is clear evidence that, although economic

0

50

100

150

200

250

300

0 5000 10000 15000 20000 25000 30000

GNP per capita ($)

Kilo

gram

s S

O2

per

capi

ta

FIGURE 1 Environmental Kuznets curve for sulfur emissions.

Data from Panayotou (1993) and Stern et al. (1996).

518 Environmental Kuznets Curve

growth usually leads to environmental degradationin the early stages of the process, in the end thebest—and probably the only—way to attain a decentenvironment in most countries is to become rich.’’However, the EKC has never been shown to apply toall pollutants or environmental impacts, and recentevidence challenges the notion of the EKC in general.The remainder of this section discusses the economicfactors that drive changes in environmental impactsand that may be responsible for rising or decliningenvironmental degradation over the course of eco-nomic development.

If there were no change in the structure ortechnology of the economy, pure growth in the scaleof the economy would result in a proportionalgrowth in pollution and other environmental im-pacts. This is called the scale effect. The traditionalview that economic development and environmentalquality are conflicting goals reflects the scale effectalone. Panayotou, a proponent of the EKC hypoth-esis, argued,

At higher levels of development, structural change towardsinformation-intensive industries and services, coupled withincreased environmental awareness, enforcement of envir-onmental regulations, better technology, and higher envir-onmental expenditures, result in leveling off and gradualdecline of environmental degradation.

Therefore, at one level, the EKC is explained bythe following ‘‘proximate factors’’:

1. Scale of production implies expanding produc-tion, with the mix of products produced, the mix ofproduction inputs used, and the state of technologyall held constant.

2. Different industries have different pollutionintensities, and the output mix typically changesover the course of economic development.

3. Changes in input mix involve the substitutionof less environmentally damaging inputs to produc-tion for more damaging inputs and vice versa.

4. Improvements in the state of technology in-volve changes in both (a) production efficiency interms of using less, ceteris paribus, of the pollutinginputs per unit of output and (b) emissions-specificchanges in process that result in less pollutant beingemitted per unit of input.

These proximate variables may, in turn, be drivenby changes in underlying variables such as envi-ronmental regulation, awareness, and education. Anumber of articles have developed theoretical modelsabout how preferences and technology might interactto result in different time paths of environmental

quality. The various studies make different simplify-ing assumptions about the economy. Most of thesestudies can generate an inverted U-shape curve ofpollution intensity, but there is no inevitability aboutthis. The result depends on the assumptions madeand the value of particular parameters. It seems fairlyeasy to develop models that generate EKCs underappropriate assumptions, but none of these theore-tical models has been empirically tested. Further-more, if in fact the EKC for emissions is monotonicas recent evidence suggests, the ability of a model toproduce an inverted U-shaped curve is not necessa-rily a desirable property.

3. ECONOMETRIC FRAMEWORK

The earliest EKCs were simple quadratic functions ofthe levels of income. However, economic activityinevitably implies the use of resources, and by thelaws of thermodynamics, use of resources inevitablyimplies the production of waste. Regressions thatallow levels of indicators to become zero or negativeare inappropriate except in the case of deforestationwhere afforestation can occur. This restriction can beapplied by using a logarithmic dependent variable.The standard EKC regression model is

lnðE=PÞit ¼ ai þ gt þ b1 lnðGDP=PÞit

þ b2½lnðGDP=PÞ�2it þ eit; ð1Þ

where E is emissions, P is population, GDP is grossdomestic product, e is a random error term, and lnindicates natural logarithms. The first two terms onthe right-hand side are intercept parameters that varyacross countries or regions i and years t. Theassumption is that although the level of emissionsper capita may differ over countries at any particularincome level, the income elasticity is the same in allcountries at a given income level. The time-specificintercepts are intended to account for time-varyingomitted variables and stochastic shocks that arecommon to all countries.

The ‘‘turning point’’ level of income, whereemissions or concentrations are at a maximum, canbe found using the following formula:

t ¼ exp½�b1=ð2b2Þ�: ð2Þ

The model is usually estimated with panel data.Most studies attempt to estimate both the fixedeffects and random effects models. The fixed effectsmodel treats the ai and gt as regression parameters,whereas the random effects model treats them ascomponents of the random disturbance. If the effects

Environmental Kuznets Curve 519

ai and gt and the explanatory variables are corre-lated, the random effects model cannot be estimatedconsistently. Only the fixed effects model can beestimated consistently. A Hausman test can be usedto test for inconsistency in the random effectsestimate by comparing the fixed effects and randomeffects slope parameters. A significant differenceindicates that the random effects model is estimatedinconsistently due to correlation between the ex-planatory variables and the error components.Assuming that there are no other statistical pro-blems, the fixed effects model can be estimatedconsistently, but the estimated parameters areconditional on the country and time effects in theselected sample of data. Therefore, they cannot beused to extrapolate to other samples of data. Thismeans that an EKC estimated with fixed effectsusing only developed country data might say littleabout the future behavior of developing countries.Many studies compute the Hausman statistic and,finding that the random effects model cannot beestimated consistently, estimate the fixed effectsmodel. But few have pondered the deeper implica-tions of the failure of this test.

Tests for integrated variables designed for use withpanel data find that sulfur and carbon emissions andGDP per capita are integrated variables. This meansthat we can rely only on regression results thatexhibit the cointegration property. If EKC regressionsdo not cointegrate, the estimates will be spurious.Very few studies have reported any diagnosticstatistics for integration of the variables or cointe-gration of the regressions, and so it is unclear whatwe can infer from the majority of EKC studies.

4. RESULTS OF EKC STUDIES

The key features differentiating the broad variety ofEKC studies can be displayed by reviewing a few ofthe early studies and examining a single indicator(sulfur) in more detail. This section also discussesEKCs for total energy use, seen by some as a proxyindicator for all environmental impacts.

The early EKC studies appeared to indicate thatlocal pollutants, such as smoke and sulfur emissions,were more likely to display an inverted U-shapedrelation with income than were global impacts suchas carbon dioxide. This picture fits environmentaleconomics theory; local impacts are internalized in asingle economy or region and are likely to give rise toenvironmental policies to correct the externalities on

pollutees before such policies are applied to globallyexternalized problems.

Grossman and Krueger produced the first EKCstudy as part of a study of the potential environmentalimpacts of NAFTA. They estimated EKCs for SO2,dark matter (fine smoke), and suspended particles(SPM) using the GEMS data set. This data set is apanel of ambient measurements from a number oflocations in cities around the world. Each regressioninvolved a cubic function in levels (not logarithms) ofpurchasing power parity adjusted (PPP) per capitaGDP and various site-related variables, a time trend,and a trade intensity variable. The turning points forSO2 and dark matter are at approximately $4000 to$5000, whereas the concentration of suspendedparticles appeared to decline even at low income levels.

Shafik and Bandyopadhyay’s study was particularlyinfluential in that the results were used in the WorldDevelopment Report 1992. They estimated EKCs for10 different indicators using three different functionalforms. Lack of clean water and lack of urbansanitation were found to decline uniformly withincreasing income and over time. Deforestationregressions showed no relation between income anddeforestation. River quality tended to worsen withincreasing income. However, local air pollutant con-centrations conformed to the EKC hypothesis, withturning points between $3000 and $4000. Finally,both municipal waste and carbon emissions per capitaincreased unambiguously with rising income.

Selden and Song estimated EKCs for four emis-sions series—SO2, NOx, SPM, and CO—using long-itudinal data from World Resources. The data wereprimarily from developed countries. The estimatedturning points all were very high compared withthose in the two earlier studies. For the fixed effectsversion of their model, the estimated turning pointswere as follows (converted to 1990 U.S. dollars usingthe U.S. GDP implicit price deflator): SO2, $10,391;NOx, $13,383; SPM, $12,275; and CO, $7,114.This study showed that the turning point foremissions was likely to be higher than that forambient concentrations. During the initial stages ofeconomic development, urban and industrial devel-opment tends to become more concentrated in asmaller number of cities that also have rising centralpopulation densities, with the reverse happeningduring the later stages of development. So, it ispossible for peak ambient pollution concentrationsto fall as income rises, even if total nationalemissions are rising.

Table I summarizes several studies of sulfur emis-sions and concentrations, listed in order of estimated


income turning points. On the whole, the emissions-based studies have higher turning points than do theconcentrations-based studies. Among the emissions-based estimates, both Selden and Song and Cole andcolleagues used databases that are dominated by, orconsist solely of, emissions from Organization forEconomic and Cooperative Development (OECD)countries. Their estimated turning points were$10,391 and $8,232, respectively. List and Gallet

used data for 1929 to 1994 for the 50 U.S. states.Their estimated turning point is the second highest inTable I. Income per capita in their sample rangedfrom $1,162 to $22,462 (in 1987 U.S. dollars). Thisis a wider range of income levels than was found inthe OECD-based panels for recent decades. Thissuggests that including more low-income data pointsin the sample might yield a higher turning point. Sternand Common estimated the turning point at more

TABLE I

Sulfur EKC Studies

Author(s)

Turning point

(1990 U.S.

dollars)

Emissions or

concentrations

Purchasing

power parity

Additional

variables

Data source

for sulfur Time period

Countries/

cities

Panayotou (1993) 3,137 Emissions No — Own estimates 1987–1988 55 developed

and

developingcountries

Shafik and

Bandyopadhyay

(1992)

4,379 Concentrations Yes Time trend,

locational

dummies

GEMS 1972–1988 47 cities in 31

countries

Torras and Boyce(1998)

4,641 Concentrations Yes Incomeinequality,

literacy,

political andcivil rights,

urbanization,

locational

dummies

GEMS 1977–1991 Unknownnumber of

cities in 42

countries

Grossman andKrueger (1994)

4,772–5,965 Concentrations No Locationaldummies,

population

density, trend

GEMS 1977, 1982,1988

Up to 52 citiesin up to 32

countries

Panayotou (1997) 5,965 Concentrations No Populationdensity,

policy

variables

GEMS 1982–1984 Cities in 30developed

and

developing

countries

Cole et al. (1997) 8,232 Emissions Yes Country

dummy,

technology

level

OECD 1970–1992 11 OECD

countries

Selden and Song

(1994)

10,391–10,620 Emissions Yes Population

density

WRI: primarily

OECD

source

1979–1987 22 OECD and 8

developing

countries

Kaufmann et al.(1997)

14,730 Concentrations Yes GDP/Area, steel

exports/GDP

UN 1974–1989 13 developed

and 10developing

countries

List and Gallet

(1999)

22,675 Emissions N/A — U.S. EPA 1929–1994 U.S. states

Stern and

Common (2001)

101,166 Emissions Yes Time and

country

effects

ASL 1960–1990 73 developed

and

developing

countries


than $100,000. They used an emissions databaseproduced for the U.S. Department of Energy by ASLthat covers a greater range of income levels andincludes more data points than do any of the othersulfur EKC studies.

The recent studies that use more representativesamples of data found that there is a monotonicrelation between sulfur emissions and income, just asthere is between carbon dioxide and income.Interestingly, an estimate of a carbon EKC for apanel data set of OECD countries found an invertedU-shaped EKC in the sample as a whole. The turningpoint was at only 54% of maximal GDP in thesample. A study using a spline regression implied awithin-sample turning point for carbon for high-income countries. All of these studies suggest that thedifferences in turning points that have been found forvarious pollutants may be due, at least in part, to thedifferent samples used.

In an attempt to capture all environmentalimpacts of whatever type, a number of researchersestimated EKCs for total energy use. In each case,they found that energy use per capita increasesmonotonically with income per capita. This resultdoes not preclude the possibility that energy intensity(i.e., energy used per dollar of GDP produced)declines with rising income or even follows aninverted U-shaped path.

A number of studies built on the basic EKC modelby introducing additional explanatory variablesintended to model underlying or proximate factorssuch as ‘‘political freedom,’’ output structure, andtrade. On the whole, the included variables turnedout to be significant at traditional significance levels.However, testing different variables individually issubject to the problem of potential omitted variablesbias. Furthermore, these studies did not reportcointegration statistics that might indicate whetheromitted variables bias was likely to be a problem ornot. Therefore, it is not really clear what can beinferred from this body of work.

The only robust conclusions from the EKCliterature appear to be that concentrations ofpollutants may decline from middle-income levelsand that emissions tend to be monotonic in income.Emissions may decline over time in countries at manydifferent levels of development. Given the likely poorstatistical properties of most EKC models, it is hardto come to any conclusions about the roles of otheradditional variables such as trade. Too few qualitystudies of other indicators apart from air pollutionhave been conducted to come to any firm conclusionsabout those impacts as well.

5. THEORETICAL CRITIQUE OFTHE EKC

A number of critical surveys of the EKC literaturehave been published. This section discusses thecriticisms that were made of the EKC on theoretical,rather than methodological, grounds.

The key criticism of Arrow and colleagues andothers was that the EKC model, as presented in theWorld Development Report 1992 and elsewhere,assumes that there is no feedback from environ-mental damage to economic production given thatincome is assumed to be an exogenous variable. Theassumption is that environmental damage does notreduce economic activity sufficiently to stop thegrowth process and that any irreversibility is notsevere enough to reduce the level of income in thefuture. In other words, there is an assumption thatthe economy is sustainable. However, if higher levelsof economic activity are not sustainable, attemptingto grow fast during the early stages of development,when environmental degradation is rising, may proveto be counterproductive.

It is clear that the levels of many pollutants perunit of output in specific processes have declined indeveloped countries over time with technologicalinnovations and increasingly stringent environmentalregulations. However, the mix of effluent has shiftedfrom sulfur and nitrogen oxides to carbon dioxideand solid waste, so that aggregate waste is still highand per capita waste might not have declined.Economic activity is inevitably environmentallydisruptive in some way. Satisfying the material needsof people requires the use and disturbance of energyflows and materials. Therefore, an effort to reducesome environmental impacts might just aggravateother problems. Estimation of EKCs for total energyuse is an attempt to capture environmental impactregardless of its nature.

Various critics argued that if there was an EKC-type relationship, it might be partly or largely a resultof the effects of trade on the distribution of pollutingindustries. The Hecksher–Ohlin trade theory, thecentral theory of trade in modern economics,suggests that, under free trade, developing countrieswould specialize in the production of goods that areintensive in the production inputs they are endowedwith in relative abundance: labor and naturalresources. The developed countries would specializein human capital and manufactured capital-intensiveactivities. Part of the reductions in environmentaldegradation levels in developed countries and part ofthe increases in environmental degradation levels in


middle-income countries may reflect this specializa-tion. Environmental regulation in developed coun-tries might further encourage polluting activities togravitate toward developing countries.

These effects would exaggerate any apparentdecline in pollution intensity with rising incomealong the EKC. In our finite world, the poorcountries of today would be unable to find othercountries from which to import resource-intensiveproducts as they themselves become wealthy. Whenthe poorer countries apply similar levels of environ-mental regulation, they would face the more difficulttask of abating these activities rather than out-sourcing them to other countries. There are no clearsolutions to the impact of trade on pollution from theempirical EKC literature.

Some early EKC studies showed that a number ofindicators—SO2 emissions, NOx, and deforesta-tion—peak at income levels around the currentworld mean per capita income. A cursory glance atthe available econometric estimates might have ledone to believe that, given likely future levels of meanincome per capita, environmental degradationshould decline from this point onward. However,income is not normally distributed but rather veryskewed, with much larger numbers of people belowmean income per capita than above it. Therefore, it ismedian income, rather than mean income, that is therelevant variable. Both Selden and Song and Sternand colleagues performed simulations that, assumingthat the EKC relationship is valid, showed thatglobal environmental degradation was set to rise fora long time to come. More recent estimates showedthat the turning point is higher; therefore, thereshould be no room for confusion on this issue.

6. ECONOMETRIC CRITIQUE OFTHE EKC

Econometric criticisms of the EKC concern fourmain issues: heteroskedasticity, simultaneity, omittedvariables bias, and cointegration issues.

Heteroskedasticity may be important in thecontext of cross-sectional regressions of groupeddata. Data for a country are a sum or mean ofobservations for all of the regions, firms, or house-holds within that country. Smaller residuals may beassociated with countries with higher total GDPs andpopulations. Taking heteroskedasticity into accountin the estimation stage seems to significantly improvethe goodness of fit of globally aggregated fittedemissions to actual emissions.

Simultaneity may arise because increasing energyuse drives increasing income, but increasing incomemay also increase the demand for energy in consump-tion activities. In addition, if environmental degrada-tion is sufficiently severe, it may reduce nationalincome, whereas the scale effect implies that increasedincome causes increased environmental degradation.A Hausman-type statistical test for regressor exogene-ity can be used to directly address the simultaneityissue. No evidence of simultaneity has been foundusing such a test. In any case, simultaneity bias is lessserious in models involving integrated variables thanin the traditional stationary econometric model. TheGranger causality test can be used to test the directionof causality between two variables. In tests ofcausality between CO2 emissions and income invarious individual countries and regions, the overallpattern that emerges is that causality runs fromincome to emissions or that there is no significantrelationship in developing countries, whereas caus-ality runs from emissions to income in developedcountries. However, in each case, the relationship ispositive, so that there is no EKC-type effect.

Three lines of evidence suggest that the EKC is anincomplete model and that estimates of the EKC inlevels can suffer from significant omitted variablesbias: (1) differences between the parameters of therandom effects and fixed effects models tested usingthe Hausman test, (2) differences between theestimated coefficients in different subsamples, and(3) tests for serial correlation. Hausman-type teststatistics show a significant difference in the para-meter estimates for estimated random effects andfixed effects models in the majority of EKC studies.This indicates that the regressors—the level andsquare of the logarithm of income per capita—arecorrelated with the country effects and time effects,indicating that the regressors are likely correlatedwith omitted variables and that the regressioncoefficients are biased. As expected, given the Haus-man test results, parameter estimates turn out to bedependent on the sample used, with the global anddeveloping country estimates of emissions EKCs forsulfur and carbon showing a turning point atextremely high income levels and the developedcountry estimates showing a within-sample turningpoint. In the cases where serial correlation in theresiduals was tested, a very high level was found,indicating misspecification in terms of either omittedvariables or missing dynamics.

Tests for integrated variables designed for use withpanel data indicate that sulfur emissions per capita,carbon emissions per capita, and GDP per capita are


integrated. This means that we can rely only onregression results that exhibit the cointegrationproperty. Results for cointegration are less clear-cutthan those for integration in the individual timeseries. For sulfur emissions in 74 countries from 1960to 1990, approximately half of the individual countryEKC regressions cointegrate, but many of these haveparameters with ‘‘incorrect signs.’’ Some panelcointegration tests indicate cointegration in allcountries, and some accept the non-cointegrationhypothesis. But even when cointegration is found, theforms of the EKC relationship vary radically acrosscountries, with many countries having U-shapedEKCs. A common cointegrating vector in all coun-tries is strongly rejected.

7. OTHER EVIDENCE

Dasgupta and colleagues present evidence thatenvironmental improvements are possible in devel-oping countries and that peak levels of environmentaldegradation will be lower than those in countries thatdeveloped earlier. They present data that showdeclines in various pollutants in developing countriesover time. They show that although regulation ofpollution increases with income, the greatest in-creases occur from low- to middle-income levels. Inaddition, diminishing returns to increased regulationwould be expected, although better enforcement athigher income levels would also be expected. There isalso informal or decentralized regulation in develop-ing countries. Furthermore, liberalization of devel-oping economies over the past two decades hasencouraged more efficient use of inputs and lesssubsidization of environmentally damaging activities.Multinational companies respond to investor andconsumer pressure in their home countries and raisestandards in the countries in which they invest.Furthermore, better methods of regulating pollution,such as market instruments, are having an impacteven in developing countries. Better information onpollution is available, encouraging government toregulate and empowering local communities. Thosewho argue that there is no regulatory capacity indeveloping countries would seem to be wrong.

Much of Dasgupta and colleagues’ evidence isfrom China. Other researchers of environmental andeconomic developments in China came to similarconclusions. For example, China is adopting Eur-opean Union standards for pollution emissions fromcars with an approximately 8- to 10-year lag.Clearly, China’s income per capita is far more than

10 years behind that of Western Europe. Further-more, China has reduced sulfur emissions and evencarbon emissions during recent years. Ambient airpollution has been reduced in several major cities,and with the exception of some encouragement ofroad transport, the government is making a sustainedeffort in the direction of environmentally friendlypolicies and sustainable development.

8. DECOMPOSING EMISSIONS

As an alternative to the EKC, an increasing numberof studies carry out decompositions of emissions intothe proximate sources of emissions changes describedin section II. The usual approach is to use indexnumbers and detailed sectoral information on fueluse, production, emissions, and so on that, unfortu-nately, is unavailable for most countries. Econo-metric decomposition models may require lessdetailed data. Decomposition models represent therelations between energy use and the environmentmore explicitly than do EKC models.

The conclusion from the studies conducted so faris that the main means by which emissions ofpollutants can be reduced is by time-related techni-que effects, particularly those directed specifically atemissions reduction, although productivity growthor declining energy intensity has a role to play,particularly in the case of carbon emissions wherespecific emissions reduction technologies do not yetexist. Estimates of emissions-specific reductions forsulfur range from approximately 20% globally overa 20-year period to 50 to 60% in West Germany andThe Netherlands during the 1980s alone.

Although the contributions of structural change inthe output mix of the economy and shifts in fuelcomposition may be important in some countries andat some times, their average effect seems lessimportant quantitatively. Those studies that includedeveloping countries find that technological changesare occurring in both developing and developedcountries. Innovations may first be adopted prefer-entially in higher income countries but seem to beadopted in developing countries with relatively shortlags. This is seen, for example, regarding lead ingasoline, where most developed countries had sub-stantially reduced the average lead content of gaso-line by the early 1990s but many poorer countriesalso had low lead contents. Lead content was muchmore variable at low income levels than at highincome levels.


9. CONCLUSIONS

The evidence presented in this article shows that thestatistical analysis on which the EKC is based is notrobust. There is little evidence for a common invertedU-shaped pathway that countries follow as theirincomes rise. There may be an inverted U-shapedrelation between urban ambient concentrations ofsome pollutants and income, although this should betested with more rigorous time series or panel datamethods. It seems unlikely that the EKC is acomplete model of emissions or concentrations.

The true form of the emissions–income relationshipis likely to be monotonic, but the curve shifts downover time. Some evidence shows that a particularinnovation is likely to be adopted preferentially inhigh-income countries first with a short lag before it isadopted in the majority of poorer countries. However,emissions may be declining simultaneously in low-and high-income countries over time, ceteris paribus,although the particular innovations typically adoptedat any one time could be different in variouscountries.

It seems that structural factors on both the inputand output sides do play a role in modifying the grossscale effect, although on the whole they are lessinfluential than time-related effects. The incomeelasticity of emissions is likely to be less than 1.0but not negative in wealthy countries, as proposed bythe EKC hypothesis.

In slower growing economies, emissions-reducingtechnological change can overcome the scale effect ofrising income per capita on emissions. As a result,substantial reductions in sulfur emissions per capitawere observed in many developed countries duringthe past few decades. In faster growing middle-income economies, the effects of rising income over-whelmed the contribution of technological change inreducing emissions.

SEE ALSO THEFOLLOWING ARTICLES

Development and Energy, Overview � EconomicGrowth and Energy � Economic Growth, Liberal-ization, and the Environment � Energy Ladder inDeveloping Nations � Environmental Change andEnergy � Sustainable Development: Basic Conceptsand Application to Energy

Further Reading

Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C.,

Holling, C. S., Jansson, B-O., Levin, S., Maler, K-G., Perrings,

C., and Pimentel, D. (1995). Economic growth, carryingcapacity, and the environment. Science 268, 520–521.

Beckerman, W. (1992). Economic growth and the environment:

Whose growth? Whose environment? World Dev. 20,

481–496.

Cole, M. A., Rayner, A. J., and Bates, J. M. (1997). The

environmental kuznets curve: An empirical analysis. Environ.Dev. Econ. 2, 401–416.

Dasgupta, S., Laplante, B., Wang, H., and Wheeler, D. (2002).

Confronting the environmental kuznets curve. J. Econ. Per-spect. 16, 147–168.

Grossman, G. M., and Krueger, A. B. (1994). Environmentalimpacts of a North American Free Trade Agreement. In ‘‘The

U.S.–Mexico Free Trade Agreement’’ (P. Garber, Ed.). MIT

Press, Cambridge, MA.International Bank for Reconstruction and Development. (1992).

World Development Report 1992: Development and the

Environment. Oxford University Press, New York.

Kaufmann, R. K., Davidsdottir, B., Garnham, S., and Pauly, P.(1997). The determinants of atmospheric SO2 concentrations:

Reconsidering the environmental Kuznets curve. Ecol. Econ.25, 209–220.

List, J. A., and Gallet, C. A. (1999). The environmental Kuznetscurve: Does one size fit all? Ecol. Econ. 31, 409–424.

Panayotou, T. (1993). ‘‘Empirical Tests and Policy Analysis of

Environmental Degradation at Different Stages of Economic

Development, working paper WP238.’’ International LaborOffice, Geneva, Switzerland.

Panayotou, T. (1997). Demystifying the environmental Kuznets

curve: Turning a black box into a policy tool. Environ. Dev.Econ. 2, 465–484.

Perman, R., and Stern, D. I. (2003). Evidence from panel unit root

and cointegration tests that the environmental Kuznets curve

does not exist. Austr. J. Agric. Resource Econ. 47, 325–347.Selden, T. M., and Song, D. (1994). Environmental quality and

development: Is there a Kuznets curve for air pollution?

J. Environ. Econ. Environ. Mgmt. 27, 147–162.

Shafik, N., and Bandyopadhyay, S. (1992). ‘‘Economic Growthand Environmental Quality: Time Series and Cross-Country

Evidence,’’ background paper for World Development Report

1992. The World Bank, Washington, DC.Stern, D. I. (1998). Progress on the environmental Kuznets curve?

Environ. Dev. Econ. 3, 173–196.

Stern, D. I. (2002). Explaining changes in global sulfur emissions:

An econometric decomposition approach. Ecol. Econ. 42,

201–220.

Stern, D. I., and Common, M. S. (2001). Is there an environmental

Kuznets curve for sulfur? J. Environ. Econ. Environ. Mgmt. 41,

162–178.Stern, D. I., Common, M. S., and Barbier, E. B. (1996). Economic

growth and environmental degradation: The environmental

Kuznets curve and sustainable development. World Dev. 24,

1151–1160.Torras, M., and Boyce, J. K. (1998). Income, inequality, and

pollution: A reassessment of the environmental Kuznets curve.

Ecol. Econ. 25, 147–160.World Commission on Environment and Development. (1987).

‘‘Our Common Future.’’ Oxford University Press, Oxford, UK.


Technology

Ekc stern