EC O LO G IC A L E C O N O M ICS 6 5 ( 2 00 8 ) 6 5 0 6 61

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / e c o l e c o n

ANALYSIS

A consumption-based approach to environmental Kuznets curves using the ecological footprint indicator
Marco Bagliania,d , Giangiacomo Bravob , Silvana Dalmazzonec,d,
a

IRES (Istituto di Ricerche Economico Sociali) Piemonte, Via Nizza 18, 10125, Torino, Italy Dipartimento di Studi Sociali, Universit di Brescia, Via San Faustino, 74B, 25122 Brescia, Italy c Dipartimento di Economia S. Cognetti de Martiis, Universit di Torino, Via Po, 53, 10124 Torino, Italy d IRIS Interdisciplinary Research Institute on Sustainability, Universit di Torino and Universit di Brescia, Italy
b

AR TIC LE I N FO
Article history: Received 27 February 2006 Received in revised form 14 January 2008 Accepted 14 January 2008 Available online 4 March 2008 Keywords: Environmental Kuznets curve Ecological footprint Biocapacity Consumption Environmental cost-shifting Economic development Trade JEL classification: Q00; Q01; Q20; Q50; Q56; O13; R14

ABS TR ACT
Recent research suggests that consumption-based measures offer an insightful perspective on the debate on the relationship between economic growth and the environment. In this article we deepen the consumption-based line of inquiry by investigating the empirical evidence in support of the environmental Kuznets hypothesis using 2001 ecological footprint data for 141 countries. We perform Ordinary Least Squares and Weighted Least Squares analysis on linear, quadratic and cubic functions, in standard and logarithmic specifications, as candidate models to represent the relationship between per capita income and environmental pressure. We replicate the cross country analysis also by estimating the regression function directly, through a nonparametric regression. In our analyses, with and without weighing data by population, the results do not show evidence of de-linking. 2008 Elsevier B.V. All rights reserved.

1.

Introduction

The relationship between economic growth and environmental impact has been an object of debate in environmental economics for over a decade. Several authors argue that whereas in most countries, at low income levels, an increase in national income corresponds to increased environmental pressure, in later stages of development a stronger demand

for greener goods and environmental regulation, improved technology and more abundant resources available for investment generally lead to a de-linking between economic growth and environmental degradation. The inversion in trend would give rise to an inverted-U relationship between indicators of environmental degradation and income per capita similar to that found by Kuznets (1955) in his work on inequality at different stages of development and hence

Corresponding author. Dipartimento di Economia, Universit di Torino, Via Po, 53, 10124 Torino, Italy. Tel.: +39 011 6704410; fax: +39 011 6703895. E-mail address: silvana.dalmazzone@unito.it (S. Dalmazzone). 0921-8009/$ see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolecon.2008.01.010

EC O L O G IC A L E C O N O M ICS 6 5 ( 2 0 08 ) 65 0 6 61

651

conventionally known as the environmental Kuznets curve (EKC). The EKC hypothesis and its policy implications are contested by theoretical analyses that stress the physical impossibility of an unbounded growth in a limited world. The critiques concern the assumption, implicit in the idea of a virtuous circle between economic growth and environmental quality, that there is no threshold level of environmental damage that can produce irreversible ecological consequences or have a negative impact on the level of income (e.g. Selden and Song, 1994; Arrow et al., 1995; Hettige et al., 2000). Furthermore, according to several methodological studies (Holtz-Eakin and Selden, 1995; Perman and Stern, 2003; Stern, 2004; Bradford et al., 2005; Galeotti et al., 2006), most of the EKC literature contains econometric weaknesses that crucially influence their results. As the body of research concerning the EKC pursues successive refinements, the empirical evidence remains mixed, with local air pollutant concentrations often conforming to the EKC hypothesis and most other indicators of environmental quality showing monotonic relations with income or no relation at all (e.g. Grossman, 1995; Cole et al., 1997; de-Bruyn et al., 1998; Azomahou et al., 2006). Recent research suggests that, whereas the indicators generally adopted focus, for each country, on the emissions generated by local production processes, more consumption-based measures could offer interesting new insights (Rothman, 1998; Roca, 2003). The objective of this paper is to deepen the consumptionbased line of inquiry through an investigation based on the ecological footprint a concise indicator of environmental pressure that estimates the total quantity of natural services that a population uses, by calculating the total area of land and water ecosystems necessary to sustainably provide all the resources needed for consumption and to reabsorb the residuals. We investigate the empirical evidence using 2001 ecological footprint (EF) national data for 141 countries, using both total ecological footprint and its disaggregated components. We perform Ordinary Least Square and Weighted Least Square analysis on linear, quadratic and cubic functions, in standard and in logarithmic specification, as candidate models to represent the relationship between per capita income and environmental impact. We replicate the analysis also by estimating the regression function directly, through a nonparametric regression. Finally, we investigate the relationship between a country's biocapacity and the tendency to show EKC patterns. Our cross country analyses do not show evidence of delinking: the overall best fit is offered by cubic functional forms that, rather than exhibiting the usual N-shape, depict an environmental pressure de facto monotonically increasing at all GDP levels. This result becomes all the more clear-cut if the population variable is used to weight the data, so as to consider also the total pressure that each country imposes on global ecosystems. The first part of the paper offers a brief overview of the literature on the environmental Kuznets curves most relevant for our work (Section 2), and explains our choice of the indicator used as dependent variable (Section 3). After presenting the conceptual foundations of the models being tested (Section 4), we analyze the data and discuss the results (Section 5). Section 6 concludes.

2.

Theoretical background

Early analyses of the EKC have been reviewed in detail for example in Forrest (1995), Ekins (1997), Ansuategi et al. (1998), Borghesi (1999), Dasgupta et al. (2002). Stern (2004) includes also more recent contributions and, besides describing the empirical evidence, discusses the theoretical and methodological critiques to the EKC literature. Beginning with Grossman and Krueger (1992), the literature, in order to understand the determinants of the EKC relationship, has decomposed such relationship into the different effects economic growth may have on environmental quality. Following that line, also Panayotou (1993, 1997, 2000), Grossman (1995), Kaufmann et al. (1998), Torras and Boyce (1998), Islam and Vincent (1999) and Galeotti (2003) converge in identifying three distinct structural forces: (i) a scale effect, by which a larger scale of economic activity implies the extraction of more natural resources and the creation of more residuals; (ii) a composition effect, referring to the structural changes in the economy that lead, as income rises, to an increase in the share of less polluting activities; (iii) a technology effect, describing the change in resource and emission intensity of production due to technological modernization. The environmental pressure E generated in each country then results from the following identity:
n X j1

!   n X Yj Ej Y Y C j Tj Y Yj j1

where Y is GDP, Yj is sectoral GDP, Ej is environmental pressure due to sector j, Cj is the share of the GDP in sector j over total GDP, and Tj is the jth sector's environmental pressure intensity. Although most authors use emissions as the dependent variable, we prefer a more general expression such as environmental pressure, so as to include not only endof-pipe effects but also impacts such as deforestation and other forms of resource depletion. This decomposition is production-based in that it ascribes to each country the environmental damage generated by domestic production activities. It is also possible to conceive of an alternative procedure, one that assigns to each country the environmental pressure generated anywhere in the world in order to produce the goods ad services consumed by its population. Production and consumption-based approaches actually consider, on the whole, the same environmental data, but use a different accounting principle: the former assign the environmental impact to each country on the basis of the geographical location of the sources, the latter on the basis of consumer responsibility (Proops et al., 1993, 1999; Munksgaard and Pedersen, 2001; Ferng, 2003; Bastianoni et al., 2004). Several authors stress the interest of looking at the EKC hypothesis from a consumption-based viewpoint. Ekins (1997) notes that when income increases, a change towards a greener structural composition of the economy is not necessarily due to a corresponding change in consumption patterns. An increased demand for environmental quality may not lead to a shift to a cleaner production process but rather to a movement of production outside of the country, with an unchanged

652

EC O LO G IC A L E C O N O M ICS 6 5 ( 2 00 8 ) 6 5 0 6 61

consumption pattern maintained via international trade (Rothman, 1998; Andersson and Lindroth, 2001; Bagliani and Bravo, 2005). The mere displacement of environmental pressure from one country to another, known as the Ekins' pollution-haven hypothesis, is supported by Suri and Chapman's (1998) empirical analysis of the effects of cross country movements of goods embodying pollution. When trade variables (such as import-manufacturing ratios) are introduced, the turning point of their EKC for energy-related pollutants moves from an already high $55,000 to about $224,000, far beyond an attainable level of income per capita. Aldy (2005) estimates pre-trade (production-based) and post-trade (consumption-based) CO2 EKCs on United States electricity-related emissions. He finds that consumption-based EKCs peak at significantly higher incomes than production-based EKCs, so that emissions-intensive trade appears to play an important role in shaping the incomeemissions relationship. Also Roca (2003) and Nahman and Antrobus (2005), among others, maintain that focussing on the evolution of a country's consumption structure allows us to go beyond analyses that neglect the international displacement of environmental costs. Rothman (1998) argues in favour of aggregate consumptionbased indicators such as the OECD composite indicator used by Ekins (1997)1, measures of total material requirement (Adriaanse et al., 1997), ecotoxicity indicators (Ayres and Marinas, 1995), the environmental space (Opschoor, 1995), and the ecological footprint. As an example, he proposes a concise analysis relating ecological footprint to real GDP per capita, based on early data on ecological footprint for 52 countries. He tests four model specifications linear, quadratic, log-linear, log-quadratic finding no evidence in support of an inverted-U behaviour. In the meantime, the methodology for the calculation of the ecological footprint has significantly improved. The sources of primary data have been standardized in a common accounting framework, which has increased their reliability and coverage. Spatial resolution is now an order of magnitude higher than in former accounts. More reliable ecological footprint analyses from the perspectives of trade are now feasible, since some of the new sources distinguish between changes in stocks, production, waste and secondary uses (Wackernagel et al., 2002a; Monfreda et al., 2004). Furthermore, improved sources have made it possible to extend a systematic calculation of ecological footprint to almost all countries (WWF and UNEP-WCMC, 2004). In this paper we extend Rothman's (1998) cross country investigation in terms of spatial coverage (from 52 to 141 countries) and of explanatory models considered, as well as by adding more in depth analyses based on ecological footprint components and biocapacity.

sustainability which represents the critical natural capital requirements of a defined economy or population in terms of the corresponding biologically productive areas (Wackernagel et al., 1999: 377). A country's footprint is the total area required to produce the resources it consumes and to absorb the waste it generates, using prevailing technology.2 It is a consumptionbased indicator because it includes all the natural capital directly or indirectly used for the supply of the goods and services consumed by the local population, independently of where the supplying area is located: it counts the natural capital embodied in locally consumed goods and services, whether domestically produced or imported, and subtracts the natural capital embodied in exports. After the pioneering analyses by Rees and Wackernagel, the concept of ecological footprint has been adopted in a progressively larger number of studies applied to geographical regions, nations as well as specific productive activities. The early works are extensively reviewed in the Ecological Economics special issue on the ecological footprint (vol. 32, n. 2, 2000). Both theoretical and applied studies have reached influential scientific outlets such as Nature (Rees, 2003) and PNAS (Wackernagel et al., 2002b). The most representative collections of data are the biannual WWF Living Planet Reports, that contain EF data for all countries with more than a million inhabitants up to 2001 (WWF and UNEP-WCMC, 2000, 2002, 2004) and, for the 25 European countries, the WWF 2005 Report to the European Parliament (WWF International et al., 2005). An important element in this accounting approach is represented by the calculation of biocapacity. Complementarily to the domestic demand for natural capital provided by the ecological footprint, biocapacity estimates the domestic supply of natural capital by calculating a country's total area of ecologically productive land. This joint information enables the calculation of an environmental budget by subtracting ecological footprint from biocapacity for a definite period of time. An ecological deficit (surplus) corresponds to a negative (positive) balance-sheet, meaning that the local consumption of natural resources is greater (lower) than the regeneration rate of local ecosystems.3 Both indicators take into account six different types of bioproductive areas: cropland, grazing land, forest area, fishing grounds, built-up land, and energy land.4 Different aggregations of these components can be of use depending on the purpose of the analysis. The major distinction is between
This area is measured in global hectares (gha) one hectare of ecologically productive land with world average productivity. 3 Further refinements of the EF methodology have introduced the concept of ecological footprint of production, that calculates the natural capital requirements of local production activities based on the same variables and data used for the standard (consumption-based) EF. The excess of EF of production over a country's biocapacity is called ecological overshoot' and represents the liquidation of local natural capital (Monfreda et al., 2004). If a country presents ecological deficit but not ecological overshoot it means that the excess environmental pressure over local biocapacity has been compensated by natural capital embodied in imports. 4 Energy land refers to the forest areas needed for the sequestration of the CO2 generated in energy production by fossil fuels. For detailed definitions of the various components see e.g. WWF and UNEP-WCMC (2004: 35).
2

3.

The case for ecological footprint

Introduced by Rees (1992) and developed by Rees and Wackernagel (1994) and Wackernagel and Rees (1996), the ecological footprint is a concise indicator of environmental
1 Which includes, besides emissions of key pollutants such as NOx, CO2, SO2 and fertilizers, also consumption-based measures such as water abstraction, generation of municipal solid waste, and private road transport.

EC O L O G IC A L E C O N O M ICS 6 5 ( 2 0 08 ) 65 0 6 61

653

energy (which comprises both direct consumption and energy embodied in goods and services) and non-energy footprint (which includes all the other five types). Such distinction allows us to discern the uses of natural capital that affect only the global environment by increasing the greenhouse effect, from those uses whose impact is felt on local ecosystems (e.g. generation of household waste, overgrazing, overfishing). Ecological footprint is an indicator centered on the use of renewable resources. There are therefore important components of environmental impact that this methodology cannot deal with. Emissions of heavy metals, radioactive materials, persistent synthetic compounds and any other pollutant for which nature has no significant assimilative capacity cannot be translated into areas required for their absorption, and are therefore excluded by definition from the EF account. As to emissions for which ecosystems have some assimilation capacity, current data already include CO2 whereas ongoing theoretical work is trying to increase coverage to include, for example, sulphur emissions (Vieira, 2004). Whereas resource extraction in terms of biomass (e.g. timber and agricultural harvesting, hunting and fishing) is accounted for in detail within the EF, capturing freshwater withdrawals still presents difficulties since their impact on biocapacity crucially depends on local conditions. Despite these limitations, ecological footprint represents a powerful indicator of the dynamics of renewable resource use, capturing a significant share of environmental pressure both on the input side (extraction of almost all renewable resources) and on the output side (waste absorption and CO2 sequestration). This comprehensive view is particularly important in studying the EKC whose aim is describing a general relationship between economy and the environment. Among similarly comprehensive consumption-based indicators, including also the recently proposed emergetic ecological footprint (Zhao et al., 2005; Chen and Chen, 2006) and embodied exergy footprint (Chen and Chen, 2007), the EF is the only one for which data are available, within a standardized database, for almost all countries. Alternative indicators centered on consumption such as virtual water consumption (e.g. Hoekstra and Hung, 2002; Hoekstra and Chapagain, 2007), although offering an extremely valuable perspective, are limited to one kind of environmental pressure.

Table 1 Regression models

i) Linear EF = b0 + b1GDP + where b1 N 0 ii) Quadratic EF = b0 + b1GDP + b2GDP2 + where b2 b 0 or b2 N 0 iii) Cubic EF = b0 + b1GDP + b2GDP2 + b3GDP3 + where b3 N 0 is a random error term.

(ii) A quadratic function with a negative quadratic term and a vertex at a level of income within the data range depicts the environmental Kuznets curve, which implies, as is well known, that environmental pressure increases at initial stages of growth but at a decelerating rate, up to a point when the first derivative changes sign and a de-linking of economic growth and environmental degradation takes place.5 In this scenario, braking forces must be at work that are strong enough to overwhelm the scale effect and the other factors that make environmental degradation augment as income increases. Such braking forces may include: - physical constraints to the capacity for consumption by individuals (a decreasing scale effect); - a declining natural resource intensity of individual demand associated with an increase in the size of the service sector relative to the industrial one, once basic needs have been satisfied (the demand side of the composition effect); - a stronger tendency to give priority to environmental protection in wealthier countries, both via market mechanisms (through an increased demand for green products) and through higher demand for environmental regulation (technology effect); - on the supply side, the availability of the resources necessary for the development and diffusion of cleaner technologies (technology effect), and - increasing returns to pollution abatement, when abatement efficiency rises with an increase in the scale of abatement (Managi, 2006). A quadratic function with a positive quadratic term would imply that at high levels of GDP there is a positive feedback between income and environmental pressure. This can take place as a result of several interconnected and partially overlapping factors, among which: - higher incomes may induce changes in individual consumption bundles towards goods and services with higher impact on the environment, both in terms of

4. Conceptual foundations of the tested models

The models considered in our analysis are the traditional (i) linear, (ii) quadratic and (iii) cubic forms, summarized in Table 1. (i) If the relationship between GDP and environmental impact, in per capita terms, were linearly increasing, any increase in income would impose a proportional environmental toll, in terms of resource extraction and/or generation of residuals. In the context of a decomposition analysis this implies that the scale effect is the prevailing driving factor. The corresponding scenario is one of an ecologicaleconomic system where neither changes in GDP composition by sector nor technological improvements are decisive in influencing environmental pressure, and neither are economies of scale in nature-intensity per unit of GPD of the kind observed by Panayotou (1997).

A quadratic function with a turning point for levels of GDP beyond the data range depicts a scenario equivalent to that of a monotonically increasing degradation.

654

EC O LO G IC A L E C O N O M ICS 6 5 ( 2 00 8 ) 6 5 0 6 61

Table 2 2001 cross country data: OLS regressions R2

EF Linear Quadratic Cubic 0.665 0.752 0.766 50.645 7.933 0.000 0.006

p(F)

b0
1.4590 (0.1172) 1.0680 (0.1158) 0.8772 (0.1317)

b1
0.0002 (1.16 10 05) 0.0004 (3.37 10 05) 0.0006 (6.89 10 05)

b2

b3

Turning point

7.92 10 09 (1.14 10 09) 2.26 10 08 (5.32 10 09)

26,279.64 2.94 10 13 (1.04 10 13)

Energy EF Linear Quadratic Cubic

0.548 0.639 0.647 35.284 3.103 0.000 0.080

0.6055 (0.0946) 0.3272 (0.0971) 0.2256 (0.1123)

0.0001 (9.32 10 06) 0.0003 (2.83 10 05) 0.0004 (5.88 10 05)

5.64 10 09 (9.56 10 10) 1.34 10 08 (4.53 10 09)

24,906.36 1.57 10 13 (8.90 10 14)

Non-energy EF Linear Quadratic Cubic

0.507 0.548 0.562 12.995 4.390 0.000 0.038

0.8528 (0.0594) 0.7397 (0.0653) 0.6588 (0.0752)

0.0594 (5.86 10 06) 0.0001 (1.96 10 05) 0.0002 (3.94 10 05)

2.29 10 09 (6.43 10 10) 8.51 10 09 (3.03 10 09)

29,448.89 1.25 10 13 (5.961 10 14)

Dependent variable: per capita EF 2001. Independent variable: per capita GDP 2001. Significance codes: p b 0.001, p b 0.01, p b 0.05. Standard errors are in parentheses.

environmental unfriendly ways of satisfying given needs and in terms of the creation of new needs; - pollution control technologies may have decreasing returns; - economic growth may bring about environmental unfriendly technological change, both in the sense of increased extractive capacity and of resource-intensive products; - even technological change implying resource or energy saving per unit of product may induce a rebound effect due to behavioural responses by which increases in efficiency are overcompensated by a rise in demand for the same or other commodities (e.g. Binswanger, 2001). These factors, although not falling neatly into the EKC decomposition (the first can be seen as the consumption side of the composition effect, the second and third can result from the technology effect, whilst the forth from an interaction between the technology and the scale effects) may lead to an increasing rate of environmental degradation per unit of income as countries become richer. (iii) A cubic function (with a positive cubic term) depicts a model in which high levels of income are eventually associated with increasing environmental degradation. The most common behaviour is an N-shaped function in which, after an initial EKC-like phase, environmental degradation increases again. This can be due to the same factors changes in consumption patterns, decreasing returns to pollution control technologies, environmentally unfriendly technological change, and the rebound effect that can justify the reemergence of a positive relationship between GDP and environmental impact described in the case of the U-shaped behaviour in point (ii).

5.

Data and analysis

Our models are estimated from ecological footprint (EF) and biocapacity (BC) data for the 141 countries where reliable data exist for both ecological footprint and biocapacity, i.e. all countries with more than a million inhabitants. The source for total ecological footprint and its components as well as for biocapacity is the Living Planet Report 2004 (WWF and UNEP-WCMC, 2004), which at the time of writing provided the most recent and consistent figures available. The ideal would obviously be to run an analysis on panel data, but sufficiently long time series were not available for single countries. We therefore develop a cross-sectional analysis on national data aimed at drawing insights from a comparison with the numerous cross country empirical studies of the EKC literature (for example Selden and Song, 1994; Cole et al., 1997; Rothman, 1998; Hettige et al., 2000; Bertinelli and Strobl, 2005). Gross Domestic Product (GDP) and population data come from the World Development Indicators 2001 database (World Bank, 2002). Monetary figures are expressed in 1995 US dollars. We perform Ordinary Least Squares (OLS) on the three models summarized in Table 1, both in standard and logarithmic specification. In order to evaluate potential effects of heteroskedasticity and outliers, we also perform robust6 regressions. The same analyses are then run weighting the data by each country's population. Since the outcomes of the ordinary and robust procedures are substantially equivalent,
6 As robust regression procedure we use iteratively reweighted least square with Huber (1981) weight function.

EC O L O G IC A L E C O N O M ICS 6 5 ( 2 0 08 ) 65 0 6 61

655

Fig. 1 Ecological footprint per capita vs. GDP per capita, 2001 cross country data. The chart shows linear, quadratic and cubic OLS regression curves.

we limit our discussion to OLS and WLS results. In addition we conduct a nonparametric analysis in order to rule out arbitrariness in the choice of functional forms.

5.1.

OLS analysis

We analyze 2001 per capita GDP and ecological footprint for 141 countries. Regression coefficients for the linear, quadratic and cubic models, estimated with OLS, are shown in

Table 2. At the 1% level all estimations are statistically significant. The prevailing model according to an F test for nested models is a cubic function that, however, rather than exhibiting the usual N-shape, is monotonically increasing (mathematically characterized by the absence of the maximum and minimum, substituted by an inflection point) (Fig. 1). The analysis on the logarithmic specifications selects the linear model, thus confirming a monotonically increasing trend (Table 3).

Table 3 2001 cross country data: OLS regressions on logarithmic specifications R2

EF Linear Quadratic Cubic 0.793 0.796 0.799 2.034 2.160 0.156 0.144

p(F)

b0
2.4438 (0.1361) 1.5498 (0.6439) 2.9590 (3.1342)

b1
0.4156 (0.0180) 0.1696 (0.174) 1.7232 (1.2995)

b2

b3

Turning point

0.0161 (0.0113) 0.2729 (0.1751)

5.26 0.0113 (0.0077)

Energy EF Linear Quadratic Cubic

0.728 0.728 0.738 0.000 5.053 0.997 0.026

4.7416 (0.2331) 4.737 (1.1110) 7.0428 (5.3539)

0.5941 (0.0308) 0.5929 (0.2999) 4.3525 (2.2199)

8.14 10 05 (0.0196) 0.6709 (0.2991)

3,643.43 0.0294 (0.0131)

Non-energy EF Linear Quadratic Cubic

0.621 0.628 0.628 2.730 0.024 0.101 0.878

2.0687 (0.1421) 0.9810 (0.6710) 0.4856 (3.2915)

0.2833 (0.0188) 0.0154 (0.1811) 0.2234 (1.3647)

0.0196 (0.0118) 0.0478 (0.1839)

0.39 0.0012 (0.0081)

Dependent variable: log(per capita EF 2001). Independent variable: log(per capita GDP 2001). Significance codes: p b 0.001, p b 0.01, p b 0.05. Standard errors are in parentheses.

656

EC O LO G IC A L E C O N O M ICS 6 5 ( 2 00 8 ) 6 5 0 6 61

Fig. 2 Ecological footprint per capita vs. GDP per capita, 2001 cross country data. Nonparametric regression curve.

We then replicate the cross country analysis by estimating the regression function directly, through a nonparametric regression. The weighting function is based on a logistic density function, and the predicted outcome is presented in Fig. 2. The result is a monotonically increasing function, without minima or maxima. The emerging picture is one of an environmental impact rapidly increasing with per capita GDP until around 17,000 US dollars, with subsequent slower growth.

In the analysis on the logarithmic specification, however, the linear model prevails in both aggregate EF and its energy and non-energy components (Table 3).7

5.3.

The weight of population

5.2.

Ecological footprint by components

Disaggregating the ecological footprint into its energy and non-energy components is of interest in order to test whether the monotonically increasing pattern emerging in the analysis of ecological footprint is to be ascribed predominantly to local effects on the ecosystem (non-energy component) or rather to global energy-related (i.e. greenhouse) effects. The regression relative to the energy component is the only one in which the quadratic specification, significant at the 0.1% level, fits slightly better than the other models and has a turning point corresponding to a GDP of 24,906 US dollars per capita, inside the data range (Table 2). The inversion in trend, however, corresponds to an ecological footprint of 6.54 gha per capita, that represents an unsustainable demand of natural capital given that global biocapacity per capita was estimated at 1.8 gha in 2001 (WWF and UNEP-WCMC, 2004:10). The non-energy component is best fitted by a cubic model, although the even larger data dispersion reduces the statistical significance of the regression. Energy appears to be the factor of environmental impact marginally more susceptible to the composition and technology effects generally advocated as sources of EKC behavior.

Using ecological footprint and GDP data in per capita terms permits a comparison between countries with different population sizes, but hides a crucial aspect of the problem, namely the total impact that each country imposes on global ecosystems (the product of per capita footprint times population). For instance, the per capita ecological footprint of both USA and Kuwait are equal to 9.5 gha whilst their total footprints are, respectively, 2,736 millions and 23 millions of gha. In order to investigate this aspect we run weighted least square regressions, using the population variable to weight the data. This analysis is legitimated by the fact that the correlation between population and GDP per capita is almost zero (R = 0.007). Weighted and unweighted analyses represent two different perspectives: the development path and its relationship with environmental impact in the unweighted regressions, the actual impact that countries have on global ecosystems in the weighted ones. In the WLS regression the best fitting model is a cubic specification (significant at the 0.1% level) presenting a

7 In Table 3 (energy component) the pairwise comparison between regressions presents the only case in which the quadratic model does not perform better than the linear, whereas the cubic model performs better than the quadratic, leaving open the possibility that the cubic model performs better than the linear. Compared directly with the linear, however, the cubic model does not score significantly better than the latter (F = 2.5263, p(F) = 0.08367).

EC O L O G IC A L E C O N O M ICS 6 5 ( 2 0 08 ) 65 0 6 61

657

Fig. 3 Ecological footprint per capita vs. GDP per capita, 2001 cross country data. The chart shows linear, quadratic and cubic WLS regression curves.

maximum and a minimum at virtually the same ordinate, qualitatively depicting a function that increases at low levels of income, reaches a plateau, and then increases again at higher GDPs per capita (Fig. 3, Table 4). This holds also for the disaggregated energy and non-energy components, whose

relationships with per capita income are best described by cubic models of a similar shape (Table 4). As with OLS, the analysis on the logarithmic forms sees the linear model prevailing on both aggregate EF and its energy and non-energy components (Table 5). What emerges with

Table 4 2001 cross country data: WLS regressions R2

EF Linear Quadratic Cubic 0.810 0.812 0.856 1.878 42.211 0.173 0.000

p(F)

b0
1.2240 (0.0896) 1.1670 (0.1010) 0.7565 (0.1089)

b1
0.0002 (8.07 10 06) 0.0002 (3.65 10 05) 0.0007 (7.71 10 05)

b2

b3

Turning point

1.34 10 09 (1.11 10 09) 3.93 10 08 (5.93 10 09)

89,605.36 7.34 10 13 (1.13 10 13)

Energy EF Linear Quadratic Cubic

0.801 0.801 0.836 0.122 29.010 0.727 0.000

0.4952 (0.0620) 0.4849 (0.0701) 0.2390 (0.0786)

0.0001 (5.58 10 06) 0.0001 (2.53 10 05) 0.0004 (5.56 10 05)

2.46 10 10 (7.71 10 10) 2.30 10 08 (4.28 10 09)

284,614.40 4.39 10 13 (8.15 10 14)

Non-energy EF Linear Quadratic Cubic

0.708 0.716 0.771 5.022 32.701 0.027 0.000

0.7292 (0.0393) 0.6882 (0.0438) 0.5269 (0.0485)

0.0001 (3.53 10 06) 0.0001 (1.58 10 05) 0.0003 (3.44 10 05)

9.74 10 10 (4.82 10 10) 1.59 10 08 (2.64 10 09)

49,344.45 2.88 10 13 (5.04 10 14)

Dependent variable: per capita EF 2001 (gha). Independent variable: per capita GDP 2001 (USD). Weight variable: population 2001 (millions of individuals). Significance codes: p b 0.001, p b 0.01, p b 0.05. Standard errors are in parentheses.

658

EC O LO G IC A L E C O N O M ICS 6 5 ( 2 00 8 ) 6 5 0 6 61

Table 5 2001 cross country data: WLS regressions on logarithmic specifications R2

EF Linear Quadratic Cubic 0.845 0.845 0.849 0.072 2.931 0.789 0.089

p(F)

b0
2.7808 (0.1214) 2.6099 (0.6520) 2.1120 (2.8323)

b1
0.4531 (0.0164) 0.4080 (0.1697) 1.5620 (1.1629)

b2

b3

Turning point

0.0028 (0.0107) 0.2693 (0.1560)

71.74 0.0116 (0.0068)

Energy EF Linear Quadratic Cubic

0.783 0.784 0.786 0.506 1.563 0.478 0.213

4.6792 (0.1958) 5.4117 (1.0502) 0.1685 (4.5847)

0.5945 (0.0265) 0.7876 (0.2734) 1.5410 (1.8824)

0.0122 (0.0172) 0.3028 (0.2525)

32.30 0.0138 (0.0110)

Non-energy EF Linear Quadratic Cubic

0.745 0.745 0.751 0.004 2.958 0.949 0.088

2.5460 (0.1227) 2.5048 (0.6594) 2.2906 (2.8641)

0.3352 (0.0166) 0.3244 (0.1716) 1.6767 (1.1760)

0.0007 (0.0108) 0.2714 (0.1578)

236.46 0.0118 (0.0069)

Dependent variable: log(per capita EF 2001) (gha). Independent variable: log(per capita GDP 2001) (USD). Weight variable: population 2001 (millions of individuals). Significance codes: p b 0.001, p b 0.01, p b 0.05. Standard errors are in parentheses.

statistical and economic significance is that EF tends to increase when countries get richer.

5.4.

Biocapacity as independent variable

The theoretical arguments generally used to explain an inverted-U relationship between income and environmental degradation could be summarized into wealth-induced changes

in consumer demand, both in terms of a shift towards sectors with less environmental impact and in terms of increased demand of environmental quality. The latter could in turn be potentially affected by the biocapacity locally available in any given country. One could expect to observe a heightened concern for environmental degradation, given the level of income, in contexts where pollution, congestion and resource scarcity are more serious; and conversely less sensitiveness towards

Table 6 2001 cross country data: multiple regressions R2

OLS Linear Quadratic Cubic 0.683 0.760 0.773 46.527 7.522 0.000 0.007

p(F)

b0
1.2790 (0.1318) 0.9585 (0.1246) 0.7818 (0.1377)

b1
0.0002 (1.16 10 05) 0.0004 (3.40 10 05) 0.0006 (6.88 10 05)

b2

b3

b4
0.0777 (0.0282) 0.0546 (0.0249) 0.0513 (0.0243)

Turning point

7.57 10 09 (1.14 10 09) 2.17 10 08 (5.26 10 09)

26,485.74

2.83 10 13 (1.03 10 13)

WLS Linear Quadratic Cubic

0.857 0.856 0.881 0.001 27.993 0.978 0.000

0.9626 (0.0873) 0.9635 (0.0939) 0.6899 (0.1002)

0.0002 (7.44 10 06) 0.0002 (3.32 10 05) 0.0005 (5.70 10 09)

2.52 10 11 (9.95 10 10) 2.97 10 08 (5.70 10 09)

5.70 10 13 (1.08 10 13)

0.1845 (0.0275) 0.1847 (0.0283) 0.1438 (0.0270)

3,556,595.00

Dependent variable: per capita EF 2001 (gha). Independent variables: per capita GDP 2001 (USD). per capita biocapacity 2001 (gha). Weight variable: population 2001 (millions of individuals). The b4 coefficient refers to biocapacity. Significance codes: p b 0.001, p b 0.01, p b 0.05. Standard errors are in parentheses.

EC O L O G IC A L E C O N O M ICS 6 5 ( 2 0 08 ) 65 0 6 61

659

Table 7 2001 cross country data: multiple regressions on logarithmic specifications R2

OLS Linear Quadratic Cubic 0.805 0.808 0.811 2.447 2.344 0.120 0.128

p(F)

b0
2.3714 (0.1348) 1.4166 (0.6278) 3.1490 (3.046913)

b1
0.3992 (0.0183) 0.1371 (0.1694) 1.7795 (1.2632)

b2

b3

b4
0.0866 (0.0297) 0.0881 (0.0296) 0.0884 (0.0294)

Turning point

0.0172 (0.0110) 0.2772 (0.1702)

3.99

0.0114 (0.0075)

WLS Linear Quadratic Cubic

0.889 0.890 0.890 1.048 0.171 0.308 0.679

2.2682 (0.1241) 1.7025 (0.5648) 2.7155 (2.5117)

0.3803 (0.0171) 0.2319 (0.1455) 0.6585 (1.0406)

0.0093 (0.0091) 0.0487 (0.1405)

0.0025 (0.0061)

0.1987 (0.0268) 0.2014 (0.0270) 0.2052 (0.0286)

12.47

Dependent variable: log(per capita EF 2001) (gha). Independent variables: log(per capita GDP 2001) (USD). log(per capita biocapacity) 2001 (gha). Weight variable: population 2001 (millions of individuals). The b4 coefficient refers to log(biocapacity). Significance codes: p b 0.001, p b 0.01, p b 0.05. Standard errors are in parentheses.

environmental issues in areas still perceived as lands of plenty, where the biocapacity per capita is high. Should this hypothesis hold, we would then observe a stronger tendency towards Ushaped incomeenvironment relationships in countries with low biocapacity per capita. In order to explore this we run OLS and WLS multivariate regressions to test our original models where we introduce linearly the biocapacity as a further independent variable. This means adding a further term b4BC in the models summarized in Table 1. The results, presented in Table 6, confirm the cubic as best fitting model both in the OLS and the WLS regressions. The biocapacity coefficients are significant at the 0.1% level in the WLS regressions, at the 5 or 1% levels in the OLS regressions. In the analysis on the logarithmic specification the prevailing model is linear both in the OLS and WLS case. Biocapacity is significant at the 1% level in the OLS regression and at the 0.1% in the WLS one (Table 7). When population is taken into account the role of biocapacity in determining a country's behaviour towards the environment acquires a nontrivial weight: in the nonlogarithmic case, an increase of 1 gha per capita in biocapacity results in an increase of the country's ecological footprint between 0.14 and 0.18 gha per capita.

6.

Discussion and conclusions

The analysis proposed in this paper leads to conclude that, when using the ecological footprint as dependent variable to investigate the relationship between economic growth and the environment, one does not find evidence in favour of an inverted-U behaviour. As a whole, rather than the decoupling of impact from GDP per capita, the scenario supported by our statistical evidence is one of an unbounded growth of environmental pressure as GDP per capita rises. In the OLS and WLS regressions on the non-logarithmic models, a cubic functional form is always the one with the best fit. The only exception is in the OLS analysis of the single EF components,

where one case (energy EF) shows evidence of an inverted U-shape. The analysis on the logarithmic specification, where the prevailing model is always linear, reinforces these results. The nonparametric regression as well shows a monotonically increasing behaviour of environmental degradation vs. GDP per capita. Ours adds to the large number of recent studies that do not find support for the EKC hypothesis (e.g. Azomahou et al., 2006; Deacon and Norman, 2006; Richmond and Kaufmann, 2006) in contrast with, among many, Panayotou (1993), Grossman and Krueger (1995), Paudel et al. (2005), Mazzanti et al. (2006). In our results, the absence of EKCs may derive from the fact that consumption-based indicators like the ecological footprint account for the displacement of environmental damage away from high income countries. It is a hint that the change in the composition of production often advocated as a drive behind the EKC can take place also through a change in the localization of supply not only through changes in technology and in the composition of demand. The localization of supply is modified by importing a large share of the goods whose production employs polluting technologies as well as of the biomass required as nutrition by human population, and by de-localizing dirty national production processes to low income countries by foreign direct investments. For instance, Mayer et al. (2005) argued that the forest protection policies adopted in Finland and other European countries in recent decades, without a simultaneous decrease in the domestic consumption of wood, resulted in a dramatically increased logging pressure on Russian forests. Similar trends are evidenced by Berlik et al. (2002) for USA's demand for wood, and Schtz et al. (2004) on the spatial distribution of global consumption and extraction of natural resources. National environmental policies may result in a simple export of environmental pressures with no net gain in the overall conservation of nature. Our concern is not arguing the superiority of one or the other between consumption and production-based

660

EC O LO G IC A L E C O N O M ICS 6 5 ( 2 00 8 ) 6 5 0 6 61

approaches. Production-based and consumption-based indicators, both useful depending on the aim of the analysis, simply imply a different choice on where to ascribe the responsibility for the generation of environmental impact. Rather, we aim at drawing the attention to the different conclusions stemming from the two approaches when applied to testing an EKC hypothesis. The crucial difference between them is the fact that a consumption-based approach captures the potential delocalisation effects that remain hidden in production-based analyses. In order to claim that economic growth is the road to a clean environment one would need empirical evidence on the existence of an inverted U-shaped relationship between per capita income and consumptionbased environmental indicators, condition that would guarantee an actual reduction in environmental impact. In our study, the absence of an inversion in trend in ecological footprint when GDP per capita rises appears to indicate that indefinite economic growth within a clean environment cannot be achieved simultaneously by the whole planet, since it can only work locally until there are countries whose environment is allowed to deteriorate.

Acknowledgements
We gratefully acknowledge the helpful suggestions from Sergio Giaccaria, Simone Landini, Gianni Amisano, and two anonymous referees.

REFERENCES
Adriaanse, A., Bringezu, S., Hammond, A., Moriguchi, Y., Rodenburg, E., Rogich, D., Schutz, H., 1997. Resource flows: the material basis of industrial economies. World Resources Institute, Wuppertal Institute, The Netherlands Ministry of Housing, Spatial Planning and Environment. National Institute for Environmental Studies, Washington DC. April. Aldy, J.E., 2005. An environmental Kuznets curve analysis of U.S. state-level carbon dioxide emissions. Journal of Environment and Development 14 (1), 4872. Azomahou, T., Laisney, F., Nguyen, V.P., 2006. Economic development and CO2 emissions: a nonparametric panel approach. Journal of Public Economics 90 (67), 13471363. Andersson, J., Lindroth, M., 2001. Ecologically unsustainable trade. Ecological Economics 37, 113122. Ansuategi, A., Barbier, E., Perrings, C., 1998. The environmental Kuznets curve. In: van den Berghand, J.C., Hofkes, M.W. (Eds.), Theory and Implementation of Economic Models for Sustainable Development. Kluwer Academic Publishers, Dordrecht, pp. 139164. Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C.S., Jansson, B.-O., Levin, S., Mler, K.-G., Perrings, C., Pimentel, D., 1995. Economic growth, carrying capacity, and the environment. Science 268, 520521. Ayres, R.U., Marinas, K., 1995. Waste potential energy: the ultimate ecotoxic? Economie Applique IVIII, 95120. Bagliani, M., Bravo, G., 2005. The hidden appropriation of environmental sustainability. In: Tacchi, E.M. (Ed.), Sustainability: Development and Environmental Risk. Foxwell & Davies, London, pp. 4758. Bastianoni, S., Pulselli, F.M., Tiezzi, E., 2004. The problem of assigning responsibility for greenhouse gas emissions. Ecological Economics 49, 253257.

Berlik, M.M., Kittredge, D.B., Foster, D.R., 2002. The Illusion of Preservation: A Global Environmental Argument for the Local Production of Natural Resources. Harvard Forest Paper No. 26. Harvard University, Petersham, MA. Bertinelli, L., Strobl, E., 2005. The environmental Kuznets curve semi-parametrically revisted. Economics Letters 88, 350357. Binswanger, M., 2001. Technological progress and sustainable development: what about the rebound effect? Ecological Economics 36, 119132. Borghesi, S., 1999. The environmental Kuznets curve: a survey of the literature. Nota di lavoro 85.99, Fondazione Eni Enrico Mattei, Milano. Bradford, D.F., Fender, R.A., Shore, S.H., Wagner, M., 2005. The environmental Kuznets curve: exploring a fresh specification. B.E. Journals in Economic Analysis & Policy 4 (1), 128. Chen, B., Chen, G.Q., 2006. Ecological footprint accounting based on emergy. A case study of the Chinese society. Ecological Modelling 198 (12), 101114 15. Chen, B., Chen, G.Q., 2007. Modified ecological footprint accounting and analysis based on embodied exergy: a case study of the Chinese society. Ecological Economics 61 (23), 355376. Cole, M.A., Rayner, A., Bates, J.M., 1997. The environmental Kuznets curve: an empirical analysis. Environment and Development Economics 2, 401416. Dasgupta, S., Laplante, B., Wang, H., Wheeler, D., 2002. Confronting the environmental Kuznets curve. Journal of Economic Perspectives 16 (1), 147168. Deacon, R.T., Norman, C.S., 2006. Does the environmental Kuznets curve describe how individual countries behave? Land Economics 82 (2), 291315. de-Bruyn, S.M., van den Bergh, J.C., Opschoor, J.B., 1998. Economic growth and emissions: reconsidering the empirical basis of environmental Kuznets curves. Ecological Economics 25 (2), 161175. Ekins, P., 1997. The Kuznets curve for the environment and economic growth: examining the evidence. Environmental Planning 29, 805830. Ferng, J.J., 2003. Allocating the responsibility of CO2 over-emissions from the perspective of benefit principle and ecological deficit. Ecological Economics 46, 121141. Forrest, A.S., 1995. A turning point? Environment Forum 12, 2430. Galeotti, M., 2003. Environment and economic growth: is technical change the key to decoupling? Fondazione Eni Enrico Mattei, Working Paper No. 90.2003. Galeotti, M., Manera, M., Lanza, A., 2006. On the robustness of robustness checks of the environmental Kuznets curve. Fondazione Eni Enrico Mattei, Working Papers No. 22.2006. Grossman, G.M., 1995. Pollution and growth. In: Goldin, I., Winters, L.A. (Eds.), The Economics of Sustainable Development. OECD, Paris, pp. 1946. Grossman, G.M., Krueger, A.B., 1992. Environmental impacts of a North American free trade agreement. Working Paper No. 3914. National Bureau of Economic Research, Cambridge, MA. Grossman, G.M., Krueger, A.B., 1995. Economic growth and the environment. Quarterly Journal of Economics 112, 353378. Hettige, H., Mani, M., Wheeler, D., 2000. Industrial pollution in economic development: the environmental Kuznets curve revisited. Journal of Development Economics 62, 445476. Hoekstra, A.Y., Hung, P.Q., 2002. Virtual water trade: a quantification of virtual water flows between nations in relation to international crop trade. Value of Water Research Report Series, 11. UNESCO-IHE, Delft, The Netherlands. Hoekstra, A.Y., Chapagain, A.K., 2007. The water footprints of Morocco and the Netherlands: global water use as a result of domestic consumption of agricultural commodities. Ecological Economics 64 (1), 143151. Holtz-Eakin, D., Selden, T., 1995. Stoking the fires? CO2 emissions and economic growth. Journal of Public Economics 57, 85101.

EC O L O G IC A L E C O N O M ICS 6 5 ( 2 0 08 ) 65 0 6 61

661

Huber, P.J., 1981. Robust Statistics. Wiley, New York. Islam, N., Vincent, J., 1999. Unveiling the incomeenvironment relationship: an exploration into the determinants of environmental quality. Working Paper No. 701, Institute for International Development, Harvard, MA. Kaufmann, R.K., Davidsdottir, B., Garnham, S., Pauly, P., 1998. The determinants of atmospheric SO2 concentrations: reconsidering the environmental Kuznets curve. Ecological Economics 25, 209220. Kuznets, S., 1955. Economic growth and income inequality. American Economic Review 1, 128. Managi, S., 2006. Are there increasing returns to pollution abatement? Empirical analytics of the environmental Kuznets curve in pesticides. Ecological Economics 58 (3), 617636. Mayer, A.L., Kauppi, P.E., Angelstam, P.K., Zhang, Y., Tikka, P., 2005. Importing timber, exporting ecological impact. Science 308, 359360. Mazzanti, M., Montini, A., Zoboli, R., 2006. Municipal waste production, economic drivers, and new waste policies: EKC evidence from Italian regional and provincial panel data. Fondazione Eni Enrico Mattei, Working Papers No. 155.2006. Monfreda, C., Wackernagel, M., Deumling, D., 2004. Establishing national natural capital accounts based on detailed ecological footprint and biological capacity assessements. Land Use Policy 3, 231246. Munksgaard, J., Pedersen, K.A., 2001. CO2 accounts for open economies: producer or consumer responsibility? Energy Policy 29, 327334. Nahman, A., Antrobus, G., 2005. The environmental Kuznets curve: a literature survey. South African Journal of Economics 73 (1), 105120. Opschoor, J.H.B., 1995. Ecospace and the fall and rise of throughput intensity. Ecological Economics 15, 137140. Panayotou, T., 1993. Empirical tests and policy analysis of environmental degradation at different stages of economic development, World Employment programme research, Working Paper 238, International Labor Office, Geneva. Panayotou, T., 1997. Demystifying the environmental Kuznets curve: turning a black box into a policy tool. Environment and Development Economics 2, 465484. Panayotou, T., 2000. Economic growth and the environment. In: Mler, K.-G., Vincent, J. (Eds.), Handbook of Environmental Economics. North-Holland, Amsterdam. Paudel, K.P., Zapata, H., Susanto, D., 2005. An empirical test of environmental Kuznets curve for water pollution. Environmental and Resources Economics 31, 325348. Perman, R., Stern, D., 2003. Evidence from panel unit root and cointegration tests that the environmental Kuznets curve does not exist. Australian Journal of Agricultural and Resource Economics 47 (3), 325347. Proops, J.L.R., Faber, M., Wagenhals, G., 1993. Reducing CO2 Emissions: A Comparative InputOutput Study for Germany and the UK. Springer, Berlin. Proops, J.L.R., Atkinson, G., v. Schlotheim, B.F., Simon, S., 1999. International trade and the sustainability footprint: a practical criterion for its assessment. Ecological Economics 28, 7597. Rees, W.E., 1992. Ecological footprints and appropriated carrying capacity: what urban economics leaves out. Environment and Urbanization 4, 121130. Rees, W.E., 2003. A blot on the land. Nature 421, 898 27 February. Rees, W.E., Wackernagel, M., 1994. Ecological footprints and appropriated carrying capacity: measuring the natural capital

requirements of the human economy. In: Jansson, A.-M., Hammer, M., Folke, C., Costanza, R. (Eds.), Investing in Natural Capital: The Ecological Economics Approach to Sustainability. Island Press, Washington, DC, pp. 362390. Richmond, A.K., Kaufmann, R.K., 2006. Energy prices and turning points: the relationship between income and energy use/carbon emissions. Energy Journal 27 (4), 157180. Roca, J., 2003. Do individual preferences explain the environmental Kuznets curve? Ecological Economics 45, 310. Rothman, D.S., 1998. Environmental Kuznets curves real progress or passing the buck? A case for consumption-based approaches. Ecological Economics 25, 177194. Schtz, H., Moll, S., Bringezu, S., 2004. Globalisation and the shifting environmental burden: material trade flows of the European Union. Wuppertal Papers No. 134e, Wuppertal Institute fr Klima, Wuppertal. Selden, T., Song, D., 1994. Environmental quality and development: is there a Kuznets curve for air pollution emissions? Journal of Environmental Economics and Management 27, 147162. Stern, D., 2004. The rise and fall of the environmental Kuznets curve. World Development 32, 14191439. Suri, V., Chapman, D., 1998. Economic growth, trade, and energy: implications for the environmental Kuznets curve. Ecological Economics 25 (2), 195208. Torras, M., Boyce, J.K., 1998. Income, inequality and pollution: a reassessment of the environmnental Kuznets curve. Ecological Economics 25, 147160. Vieira, R., 2004. Integrating sulphur emissions in the ecological footprint. 8th Biennial Conference of the International Society for Ecological Economics, Montral, Canada. 1114 July. Wackernagel, M., Rees, W.E., 1996. Our Ecological Footprint: Reducing Human Impact on the Earth. New Society Publishers, Gabriola Island. Wackernagel, M., Onisto, L., Bello, P., Callejas, L.A., Lpez, F.I., Garca, J., Guerriero, A., Guerrero, S., 1999. National natural capital accounting with the ecological footprint concept,. Ecological Economics 29, 375390. Wackernagel, M., Monfreda, S., Gurarie, E., 2002a. Improvements to national footprint accounts since the Living Planet Report 2000. Redifining Progress. May 23, www.rprogress.org/programs/ sustainabilityindicators/ef/index_html. Wackernagel, M., Schulz, N., Deumling, D., Linares, A., Jenkins, M., Kapos, V., Monfreda, C., Loh, J., Myers, N., Norgaard, R., Randers, J., 2002b. Tracking the ecological overshoot of the human economy. Proceedings of the National Academy of Science 99 (14), 92669271. World Bank, 2002. World Development Indicators 2002. The World Bank, Washington, DC. WWF and UNEP-WCMC, 2000. Living Planet Report 2000. World Wide Fund For Nature, Gland, Switzerland. WWF and UNEP-WCMC, 2002. Living Planet Report 2002. World Wide Fund For Nature, Gland, Switzerland. WWF and UNEP-WCMC, 2004. Living Planet Report 2004. World Wide Fund For Nature, Gland, Switzerland. WWF International, Global Footprint Network, Netherlands Committee for the World Conservation Union, 2005. Europe 2005. The Ecological Footprint. World Wide Fund For Nature, Gland, Switzerland. Zhao, S., Li, Z., Li, W., 2005. A modified method of ecological footprint calculation and its application. Ecological Modelling 185 (1), 6575.