Ecological Footprints in Plural: A Meta-analytic Comparison of Empirical Results

Regional Studies, Vol. 38.7, pp. 747–765, October 2004

Ecological Footprints in Plural: A Meta-analytic Comparison of Empirical Results PETER NIJKAMP*, EMILIA ROSSI† and GABRIELLA VINDIGNI‡ *Free University, De Boelelaan 1105, NL-1081 HV Amsterdam, the Netherlands. Email: [email protected] †Universita` di Bologna, Piazza Scaravilli, 2, I-40126 Bologna, Italy. Email: [email protected] ‡Universita` di Catania, Via Santa Sofia, 98, I-95123 Catania, Italy. Email: [email protected] (Received March 2003: in revised form December 2003) N P., R E. and V G. (2004) Ecological footprints in plural: a meta-analytic comparison of empirical results, Regional Studies 38, 747–765. The concept of an ecological footprint is intriguing and has prompted an avalanche of theoretical and applied research. In recent publications both the scientific basis and the policy relevance of this concept have been given ample attention, while also much empirical work has been undertaken to assess the value of the ecological footprint in different regions or countries of the world. The paper starts with a concise critical overview of the current discussion on ecological footprints. Its main aim, however, is to provide a meta-analytic assessment and interpretation of the various empirical findings in the recent literature that offer estimated values or ranges of the ecological footprint indicator. The sensitivity of ecological footprints for the stringent assumptions made in the calculation schemes is investigated using frequency analyses, cross-tabulation methods and decision-tree induction methods (a recent technique based on pattern recognition methods). The results show that methodological choice, geographical scales and year of data collection offer a significant explanation for variations in results. The paper concludes with some suggestions for further research. Biodiversity Ecological footprint Frequency analysis Cross-tabulation methods Decision-tree induction methods Sustainable development N P., R E. et V G. (2004) Des traces e´cologiques au pluriel: une comparaison me´ta-analytique des re´sultats empiriques, Regional Studies 38, 747–765. La notion de traces e´cologiques est fascinante et a provoque´ un flot de recherches fondamentale et applique´e. Dans des revues re´centes on a preˆte´ une attention particulie`re a` la base scientifique de cette notion et a` son importance pour la politique, alors que beaucoup de travail empirique a e´te´ fait aussi afin d’e´valuer les traces e´cologiques dans diverses re´gions et dans de diffe´rents pays du monde. Dans un premier temps, cet article cherche a` donner une vue d’ensemble critique et concise du de´bat actuel sur la notion de traces e´cologiques. Cependant, son but principal est de fournir une e´valuation et une interpre´tation me´ta-analytiques des diffe´rentes preuves empiriques que l’on peut trouver dans la documentation re´cente et qui fournissent des estimations ou des approximations de l’indice des traces e´cologiques. La sensibilite´ des valeurs des traces e´cologiques pour ce qui est des suppositions rigoureuses faites dans les calculs sont examine´es a` partir des analyses de fre´quences, des donne´es croise´es, et des me´thodes d’induction par arbres de de´cisions (une technique re´cente fonde´e sur des me´thodes d’identification de configuration). Les re´sultats laissent voir que le choix me´thodologique, l’e´chelle ge´ographique, et la collecte des donne´es en fonction de l’an, fournissent une explication riche des variations des re´sultats. L’article conclut par proposer des recherches futures. Diversite´ biologique Traces e´cologiques Me´thodes d’induction par arbres de de´cisions

Analyse de fre´quences Donne´es croise´es De´veloppement durable

¨ kologische Fußspuren u¨berall: ein meta-analytischer Vergleich empirischer N P., R E. und V G. (2004) O Ergebnisse, Regional Studies 38, 747–765. Der Begriff o¨kologischer Fußspuren ist faszinierend, und hat eine wahre Lavine theoretischer und angewandter Forschung veranlaßt. In ku¨rzlich vero¨ffentlichten Arbeiten hat man der wissenschaftlichen Grundlage und der politischen Relevanz dieser Vorstellung reichlich Aufmerksamkeit gezollt, wa¨hrend gleichzeitig viel empirische Arbeit unternommen wurde, um den Wert o¨kologischer Fußspuren in verschiedenen Regionen oder La¨ndern der ¨ berblick u¨ber die gegenwa¨rtige Debatte u¨ber Welt zu beurteilen. Dieser Aufsatz beginnt mit einem knappen, kritischen U o¨kologische Fußspuren. Sein Hauptziel ist jedoch, eine meta-analytische Beurteilung und Interpretation der verschiedenen empirischen Befunde der ku¨rzlich vero¨ffentlichten Literatur vorzulegen, die Scha¨tzwerte oder Reihen o¨kologischer Fußspurenwerte anbieten. Die Sensitivita¨t o¨kologischer Fußspurenwerte fu¨r strenge Kriterien, die in Berechnungsschemen verlangt werden, werden untersucht, und zwar mit Hilfe von Frequenzanalysen, Kreuztabulierungsmethoden und Entscheidungsbaumverfahren (einer ju¨ngst entwickelten Technik, die sich auf Methoden der Musterverfahren stu¨tzt). Die Ergebnisse der Autoren zeigen, daß Wahl der Methodik, geographische Skalen und Jahr der Datensammlung signifikante Erkla¨rungen fu¨r Abweichungen bei Ergebnissen liefern. Der Aufsatz schließt mit Vorschla¨gen fu¨r weitere Forschung. 0034-3404 print/1360-0591 online/04/070747-19 ©2004 Regional Studies Association http://www.regional-studies-assoc.ac.uk

DOI: 10.1080/0034340042000265241

748

Peter Nijkamp et al.

¨ kologische Fußspuren Biodiversita¨ t O Frequenzanalyse Entscheidungsbaumverfahren nachhaltige Entwicklung

Kreuztabulierungsmethoden

N P., R E. y V G. (2004) Huellas ecolo´ gicas en plural: una comparacio´ n meta-analı´tica de resultados empı´ricos, Regional Studies 38, 747–765. El concepto de huella ecolo´ gica es intrigante y ha suscitado una avalancha de investigaciones teo´ ricas y aplicadas. En recientes publicaciones se ha prestado una amplia atencio´ n tanto a la base cientı´fica como a la relevancia polı´tica de este concepto, mientras que tambie´ n se ha llevado a cabo una gran cantidad de trabajo empı´rico con el fin de evaluar el valor de la huella ecolo´ gica en diferentes regiones y paı´ses del mundo. El presente artı´culo comienza con una breve revisio´ n breve y concisa sobre el actual debate en torno a las huellas ecolo´ gicas. El objetivo principal de este artı´culo, sin embargo, es ofrecer una interpretacio´ n y evaluacio´ n meta-analı´tica de los varios resultados empı´ricos presentes en la literatura reciente que ofrecen valores o rangos estimados del indicador de huella ecolo´ gica. La sensibilidad de los valores atribuı´dos a la huella ecolo´ gica para las rigurosas suposiciones hechas en los esquemas de ca´ lculo se investiga utilizando ana´ lisis de frecuencia, me´ todos de tablas cruzadas y me´ todos de induccio´ n de a´ rboles de decisio´ n (una te´ cnica reciente basada en me´ todos de reconocimiento de patrones). Nuestros resultados muestran que la eleccio´ n metodolo´ gica, las escalas geogra´ ficas y el an˜ o de recoleccio´ n de datos ofrecen una explicacio´ n significativa de las variaciones en los resultados. Este artı´culo concluye con algunas recomendaciones para futuras investigaciones. Biodiversidad Huella ecolo´ gica Ana´ lisis de frecuencia Me´ todos de tablas cruzadas Me´ todos de induccio´ n de a´ rboles de decisio´ n Desarrollo sostenible JEL classifications: R1, R5, Q2, Q4

The inhabitants of a city, it is true, must always ultimately derive their subsistence, and the whole materials and means of their industry from the country. But those of a city, situated near either the seacoast or the banks of a navigable river are not necessarily confined to derive them from the country in the neighbourhood. They have a much wider range, and may draw them from the most remote corners of the world, either in exchange for the manufactured produce of their own industry, or by performing the office of carriers between distinct countries, and exchanging the produce of one for that of another. A city might in this manner grow up to great wealth and splendour, while not only the country in its neighbourhood, but all those to which it traded, were in poverty and wretchedness. (S, 1776, p. 406)

ENVIRONMENTAL SUSTAINABILITY IN A SPATIAL PERSPECTIVE The twentieth century has witnessed an unprecedented growth of the world economy accompanied by an unmatched decline in environmental quality. In the past two decades, the size of the economy has tripled and the world population has grown by 30% reaching more than 6 billion people. This increase in the scale of human pressure on the ecosphere has exacerbated the so-called ‘new scarcity’. For example, soil degradation is now a concern for 65% of agricultural land; water quality is degraded directly through chemical and nutrient pollution and indirectly due to the insufficient capacity of ecosystems to filter water properly; emissions of carbon dioxide (CO2) have brought about a real threat of global climate change; major losses of biodiversity have occurred in all ecosystems (W R I, 2001–02).

The need for action has been recognized by the World Commission on Environment and Development (WCED), which in 1987 identified sustainable development as the criterion against which human-made changes of the environment must be assessed (WCED, 1987). Unfortunately, the situation has not improved since and sustainability is far from being achieved. The problem might also descend from the fact that the public does not perceive the severe risks related to overexploitation of natural resources. Moreover, there seems to be a widespread – and sometimes unwarranted – opinion that timely improved technology will make available proper substitutes or that it will allow us to use the resources increasingly more efficiently. However, certain ecosystem functions, like air and water quality control, climate control or nutrient cycling, cannot be replaced and are fundamental for humankind’s survival. The ‘new scarcity’ has prompted serious doubts about the compatibility of material growth with environmental sustainability. Agenda 21, formulated at the Rio Conference in 1992 (UNCED, 1992, p. 2) admitted that ‘there has been some progress through conventional economic policy applied in parallel to environmental economics. It is now clear that this is not enough . . .’. Finding a solution to the widespread environmental degradation may require a change in one’s views on environmental externalities. Economic processes cannot be interpreted in an isolated, closed system; the economy is a dependent subsystem of the ecosphere, characterized not only by flows between both households and firms, but also by flows of matter-energy from and to the environment. As G-R (1976) has put it, what goes into the economic process represents valuable natural resources and what is thrown out of it is valueless waste.

Ecological Footprints in Plural It is clear that sustainable development is not a given endpoint-state of a compound economic–social– ecological system. It is essentially based on complex tradeoffs, related to priorities attached to inter- and intra-generational equity, weak and strong sustainability, absolute and relative (de)linking, and local versus global sustainable development (F and N, 2001a). The latter issue prompted the formulation of L A 21 (1997), which points out that local (or regional) authorities should achieve consensus on a mission statement for sustainable development action at the local level and acknowledges that local actions are needed to combine a reduction of environmental decay with an improvement of local socioeconomic conditions in both industrialized and Third World countries. Basically, the global and Local Agenda 21 call for attention and policy action regarding our current lifestyle with its high resource depletion, decay of environmental quality and increasing socioeconomic disparities, at both global and local levels. Given the high urbanization rate of the modern world (about 70%), it certainly makes sense to consider cities – or, in general, urban regions – as focal points of sustainability policies (F and N, 2001b). Most production, consumption and transportation activities in a country take place in cities or urban regions, and hence focusing policy attention on urban quality of life can be extremely effective. These areas are anchor points for administrative and policy action and at the same time they may be in a better position to induce citizen’s participation and public support through direct local involvement. In addition, from a research perspective they are clearly demarcated statistical units that might offer the necessary empirical data for policy analysis. Clearly, urban areas are in a unique position to develop proper sustainability strategies and exploit the economies of density and scale that are the decisive factors for their formation. Obviously, there also are dis-economies, like congestion, or environmental decay or disintegration of ecosystems (R and W, 1996; M and W, 2003), but the continuous growth of medium-size and large cities suggests that positive features of the city, namely the economies in terms of scale and scope advantages and the efficient use of scarce space and environmental resources tend to outweigh the negative externalities (O, 1982; N and P, 1994; S and S, 1996; C et al., 1998; G, 1998; V and N, 2002). Clearly, cities are from an economic perspective an efficient spatial organization form, but this does not necessarily hold from social or global ecological perspectives. An important issue related to spatial externalities is the interaction between urban and rural space (or between urban regions and their hinterlands). By definition, an urban area is an open system that depends on its hinterland for life support;

749

even hunting tribes needed an area much larger than that of the village to survive. Hence, cities cannot be autarchic; there are flows of (traded) goods and pollution entering and leaving it. On the one hand, food and manufactured goods flow into the city from agricultural areas and other parts of the world to satisfy citizens’ needs and eventually to become waste. Also flows of pollution enter the city as spatial externalities; they originate in other areas, close by or far away, and reach the city through water streams or the atmosphere, for instance. On the other hand, a city is usually a centre of transformation and of trade; thus, through these activities, it can create flows of pollution that may end up as spatial externalities in other regions, or it may be the source of waste material in other areas through its exported goods. Consequently, the notion of local sustainability prompts a discussion on spatial externalities. Local sustainability comprises the effects of citizens’ lifestyles and of production systems inside the urban area, as well as the derived effects of these activities for surrounding areas. The strategic role of cities and regions in spurring trade should be recognized, while the influence of the related transport sector on environmental problems should also be underlined. Clearly, if trade and transport continue to increase, they will speed up depletion of resources, given that environmental costs are not internalized, ecosystems not valued properly valued and property rights not clearly assigned. Even though it is not a priori certain that the scale effects of globalization in relation to trade, communication and transport are by definition deleterious for environmental quality, the past patterns of environmental decay and resource depletion suggest that ‘in practice current trends show an adverse pattern’ (V V-G and N, 1999, p. 332). It is, therefore, fundamental to design a mechanism to quantify the externalities connected to the economic process – and particularly to trade and transport activities – in order to address especially questions about sustainable cities and regions, where flows of goods and pollution show a permanent interaction with their hinterland (R, 2002). The ecological footprint (EF) is one of the few indicators that makes such an attempt. It offers a new way to look at sustainable development in terms of the land derived imputed use implications of human activity; it measures the claims on environmental resource stocks in land area consequences and makes it possible to look at sustainability issues from both intra(or intraregional) and supra-urban (or supraregional) perspectives. As will be argued below, the flows of goods and pollution are only partially taken into consideration because the EF is built up using only data on consumption of domestic and traded goods and on emissions of CO2. The ‘exchange’ of other types of pollutants between cities or regions and their hinterland is usually not quantified due to the great complexity

750

Peter Nijkamp et al.

that such a task involves; many spatial externalities by necessity are thus omitted. Despite its incompleteness, the EF makes a first step in the right direction by connecting the issues of globalization, trade and environmental degradation while creating a scientific debate on the merits of local or regional self-supporting systems (cf. H, 2000). The present paper will first review concisely the concept of EF while it also offers a critical analysis of this indicator by addressing inter alia the hidden relations between actual EFs and other site-specific characteristics. In many empirical studies carried out in recent years, a great variation in EFs is observed, which prompts a question on the causes of this variety. It is, therefore, an intriguing research question whether empirical EF studies show a significant commonality in results or whether differences in results can be ascribed to site-specific conditions, to differences in methodology or to geographical scale variations. Thus, the issue is not whether the empirical EF studies are wrong or good, but whether they exhibit systematic variation. The present study aims to develop a metaanalytic framework for comparing EFs in a large set of empirical studies undertaken worldwide. The main goal is then to identify the critical factors determining variations in the value of the EF in these studies. The paper is as follows. The second section presents a general framework to address local or regional sustainable development; the EF will be interpreted as one of its components, while its evolution will then be traced. Next, the EF calculation procedure will be described in more detail in the third section and some measurement issues will be addressed. Section 4 presents a clear outline of the strengths and weaknesses of the EF concept. Subsequently, in the fourth and fifth sections, some EF applications will be reviewed and they will be classified in an information base according to characteristic attributes in order to allow a comparative study of EFs. After some exploratory investigation of the database (by means of frequency analysis and crosstabulation methods), a specific meta-analytic technique, coined decision-tree induction and based on pattern recognition methods from artificial intelligence, will be applied to identify explanatory factors in our qualitative database. Finally, some research perspectives will be offered. ECOLOGICAL FOOTPRINT: CONCEPT AND EVOLUTION The EF is a recently developed analytical tool to decompose spatial sustainability; sustainable development refers to a complex system of tradeoffs involving weak and strong sustainability, absolute and relative (de)linking and local versus supralocal sustainability (F and N, 2001a). Sustainability aims at offering future generations the same socio-economic opportunities as are available at

present. If one concentrates on the multidimensional set of opportunities and not on the idea of a unidimensional welfare concept, one can deploy the notions of strong and weak sustainability. Strong sustainability implies that no category of natural capital declines over time, while weak sustainability allows for substitution between different classes of capital and also involves distributional and equity issues; it takes for granted that overall capital, both human made and natural, is constant over time. The EF concept, introduced in a pioneering article by R (1992) and subsequently elaborated by W and R (1996), stands on the concept of strong sustainability of a given area, but it extends the condition of constancy of natural capital towards other areas on which a given area may live. Clearly, an alternative economic approach might be based on a monetary indicator that might refer to a constant economic value of the stock concerned. This approach, however, has been widely criticized, in particular because it only produces incomplete results, since for some environmental services and several critical natural capital stocks there are no markets. Does economic growth aggravate environmental externalities assessed via the EF? This second issue refers to the long-lasting debate about the relationship between economic growth and environmental degradation, i.e. the issue of (de)coupling (or delinking). Many authors argue that this relation has an inverted Ushape, implying that the two phenomena concerned can be delinked (i.e. have an opposite growth pattern) once a certain level of income or welfare has been reached (for an extensive discussion, see D B, 1999). The advocates of the so-called environmental Kuznets curve (EKC) hold that once a country is sufficiently wealthy, the structure of preferences changes in favour of environmental quality, while more environmentally benign technologies are also made available; as a result, growth should improve the quality of ecosystems. This has proven true only for a few pollutants, like nitrogen dioxide (NO2) and sulphur oxides (SOx), and it seems that above a certain average income, economic improvement and environmental degradation are linked again making the previous relation representable by an N-shaped curve (cf. O, 1992). Relative decoupling means that this relationship is less than proportional, while absolute decoupling means that the idea of the EKC is accepted (growth does not spur environmental degradation). The consequences of the EKC debate for the EF concept are not unambiguous, as rising incomes in an area may lead to an export of polluting activities, so that EFs may increase. This issue deserves certainly more empirical research, which is beyond the scope of the present study. A third research angle from which spatial sustainable development can be analysed is local versus supralocal sustainability, in particular the issue of footprints with

Ecological Footprints in Plural their quantification of human pressure on the environment. The idea underlying the EF is that human impact on the world is proportional to the population, the existing technology and affluence. It revisits the concept of human-carrying capacity by inverting it; the latter can be defined as the human load, namely a population with its level of consumption and waste production, which a given area can support indefinitely, whereas the EF refers to the area necessary to sustain a given human load. The EF was first defined as: the area of ecologically productive land (and water) in various classes that would be required on a continuous basis to provide all the energy/material resources consumed, and to absorb all the wastes discharged by a population with prevailing technology, wherever on earth the land and water is located. (W and R, 1996, p. 9)

The main assumptions it rests on are that one can keep track of one’s levels of consumption and waste and that these externalities can be converted into a measurement system corresponding to the land area needed by means of a single, easily understandable figure. In addition, the required land is assumed to be used sustainably. By taking into account trade, the EF shows how much land the population uses to maintain its lifestyle, whether or not this area coincides with the population’s home region. The per capita EF is easily calculated by dividing the total EF of the spatial area considered by its population. It can be compared with the region’s available bio-productive space in order to check whether it is running a so-called ecological deficit. If it were running such a deficit, it would be importing biocapacity from other countries that on the contrary are running a surplus or depleting their natural capital stock. The ecological deficit gives an idea of the extent to which a region or city is dependent on extraterritorial productive capacity through trade or appropriated natural flows; it is a clear indicator of dependence on out-of-boundary ecosystems. The per capita EF can also be compared with the planet’s average bio-productive space per capita, although for the time being no relevant policy conclusions can be derived from this comparison. The concept of EF has prompted much scientific debate worldwide. Its meaning, though, has not changed substantially over the past few years, but the calculation methods have been refined. In a recent report by W et al. (W W F  N (WWF), 2000, p. 3), an EF is described as follows: the biologically productive area required to produce the food and wood people consume, to give room for infrastructure, and to absorb the CO2 emitted from burning fossil fuels.

751

This definition gives a better idea of the essence the EF calculation really represents, since many ecosystem services and several forms of pollution are analytically intractable or too difficult to take account of, and therefore the picture offered by the EF tends to be rather incomplete (or to offer an underestimation, as W and R, 1996, put it). Although the concept of EF has not changed over the years – it measures estimated demand for ecological space – its empirical interpretation has changed. W and S (2000, p. 352) state that a minimum requirement for ecological sustainability is that ‘footprints must be smaller than the available ecological capacity’; this has generated much debate because of the often implicitly assumed assimilation of a sustainable system to a self-sufficient one. This assumption is not necessarily implied by the EF concept. At a global level, this might be the case, as the limits of the planet cannot be denied, but a city or region is in general not by definition self-reliant because one of its features is interacting with rural areas (or in general, its hinterland) and similarly nations do not always have the characteristics necessary to produce certain goods and hence need to trade. Moreover, trade can clearly create economic advantages (gains of trade) for both trading partners, as trade theory convincingly teaches us. As a result of this criticism in W et al. (WWF, 2000), a report that contains the EFs of all countries with more than 1 million inhabitants, the authors admit that the meaning of an ecological deficit is somewhat limited. It can be useful as an indicator of dependence and especially at the urban or regional level, it quantifies effectively the nexus with the rural areas. Another feature of the EF should also be pointed out to gain some further understanding of this indicator; in the debate on growth and environmental decay, it tends to support the side of growth pessimists. At the global level, it can be thought of as based on technological scepticism, one that assumes technology will not be able to develop enough to overcome the limits imposed by natural conditions; at a national and urban level, it is based on the assumption that beneficial trade is very hard. In addition, looking at the empirical values of some calculated EFs, it is quite clear that they follow the distribution of richness of areas; basically, wealthier countries or cities seem to put more pressure on the environment. This conclusion may probably be valid, but the EF reaches this by means of a rather simplified argumentation; the bottom line is that populations that consume a lot of food and wood and use a lot of fossil energy are pressuring the ecosystems too much. Clearly, many ecosystems and landscapes are not managed sustainably, in particular because trade dependency distracts ecological attention from specific input sources. However, at no stage in the EF analysis is the sustainability of the extractive processes explicitly addressed. Some more criticism will be offered in the

752

Peter Nijkamp et al.

fifth section, after a discussion of some measurement issues. The EF calculation procedure is fundamental to the understanding of the essence of the EF; moreover, it has undergone several shifts since it was first developed. This problem will be described in somewhat more detail below. ASSESSMENT OF THE ECOLOGICAL FOOTPRINT Prefatory remarks The method used to translate the level of consumption and waste production of a given population into land needed to support it was illustrated in detail in W and R’s Our Ecological Footprint (1996). Since then, it has slightly changed and improved, but the following basic assumptions are still valid: Ω Ω

Ω Ω

Industrial harvest practices, in particular in agriculture and forestry, are supposed to be sustainable. Not all ecosystem services that humans deploy are accounted for; for instance, human appropriation of land through freshwater withdrawal and soil contamination is not considered. Each area of land is regarded as providing one function to avoid double counting. Land is divided into various use categories according to the way humans exploit it.

Wackernagel and Rees initially specified eight land categories in total: energy land, built environment, gardens, crop land, pasture, managed forest, untouched forest and non-productive areas. This classification changed through time by the authors: the categories were further aggregated and a marine EF for the sea was introduced. Likewise, the latest version of the EF calculation procedure (WWF, 2000) distinguishes between cropland, grazing land, forest, built-up land, fishing ground and energy land. To convert the lifestyle of the population considered into ‘consumed’ areas of these different land categories, disaggregated data on consumption and on agricultural yields are needed.

always a risk of double counting in EF assessment. For example, a country might import milk to produce cheese; the former may be counted twice: once it is added to milk production and next when it is considered in the domestic production of cheese into which it is transformed. However, the embodied milk in imported cheese is not considered. Therefore, as argued by Wackernagel and Rees, one needs trade-corrected data. This issue will be discussed below. Apparent consumption can be converted into land area, the so-called appropriated capacity, using global average productivity or yield for a given area. For a given good ‘c’, the footprint component in terms of hectares per capita would be as follows: Productionc òImport c ñExportc Yield c îPopulation ó

ó Appropriated capacity c At this point, the figures relative to each land category can be added to obtain the specific land-type footprints. The total EF of the analysed population is reached through a (weighted or non-weighted) sum of the footprint components. The method may appear simple, and indeed it is, but the land requirement of each considered footprint item result is a proxy as it stems from an assessment procedure that requires many simplifications and assumptions. It is therefore necessary to discuss in greater detail the matrix produced when relating the resource and energy flows of the economy to the different land categories that these flows rely on. To do so, it is appropriate to define the land categories consistently with the latest version of the calculation procedure illustrated in the Living Planet Report (WWF, 2000). Land categories Ω

Data characteristics and appropriated area To make the collection of data simple, Wackernagel and Rees divide consumption into five major categories: food, services, transportation, consumer goods and housing. Clearly, as the number of items considered increases, the EF gains accuracy. The net consumption of the selected resources can be attained by incorporating net imports in domestic production; this figure is called ‘apparent consumption’ because it differs from real household consumption. The former, in fact, includes intermediate imported goods that might be used to produce manufactured goods, but it does not account for embodied goods in imports. There is

Apparent consumption c Yield c îPopulation

Ω

Ω

Ω

Cropland: accounts for all the land area needed to grow the crops consumed by the analysed population or regional economy. It also includes crops fed to poultry and pigs (these plants are basically converted to meat and consumed in the form of chicken or pork), in addition to cereals, fruit, vegetables, coffee, tobacco, etc. Grazing land: the grazing land used by a population refers to the area required to produce animal products such as meat, dairy products from cattle, sheep and goats, and hides and wool. Forest: includes both managed and untouched forest and accounts for the land used to produce fuel wood and charcoal, round wood, paper and paperboard. Built-up land: refers to paved-over, built upon, badly eroded or otherwise degraded land. It is no longer bio-productive land.

Ecological Footprints in Plural Ω

Ω

Fishing ground: refers to that part of the ocean area that provides the greatest part of marine production. It is the area required to produce the marine fish and seafood products that the specified population consumes, including sea fish, crustaceans, cephalopods, and fish meal and oils, which are fed to animals. Energy land: takes into consideration the implications of consuming fossil fuel, hydroelectricity and other renewable energy sources. To account for fossil energy use, Wackernagel and Rees initially proposed three methods: one calculates land consumed by fossil energy use as the area necessary to produce an equal amount of a biologically produced substitute. In later works, the energy EF concerns the land area that would be required to absorb all CO2 emissions resulting from burning the fossil fuel consumed. This includes the direct use of coal, oil, gas and the indirect use from the consumption of electricity, public transport, manufactured goods or other services (WWF, 2000). Clearly, this is a crucial assumption that will be addressed below.

Consumption–land-use matrix The consumption–land-use matrix is made up of three different parts and connects the flows of resources and energy to the land areas needed to sustain them. In the upper part, the main biotic resources are dealt with. The apparent consumption of each item is translated into the appropriated area dividing it by the corresponding global average yield. The latter is preferred to local productivity figures, not only to make international comparison possible, but also because nowadays every region strongly relies on trade and because goods consumed in one country create environmental pressure elsewhere, since production chains are interregional and international. To try and solve the aforementioned problem connected to apparent consumption and to get a better estimate of household consumption, the net trade of manufactured goods can be converted into their raw material equivalent. For instance, net imports of cheese can be converted into milk and added to the apparent consumption of the latter; domestic production of cheese is already considered in milk’s account. The central part of the matrix uses detailed national energy consumption statistics to obtain the corresponding footprint component. The consumption of fossil energy in its different states is divided by the corresponding estimated assimilation capacity of forests. The sequestration rate can greatly influence the results and usually refers to young successional forests. The Living Planet Report (WWF, 2000) also considers for the first time the ocean’s role as a carbon sink, assuming it sequesters 35% of the globally emitted carbon. Previous versions of the EF calculation procedure did not account for it. The area corresponding to hydroelectricity use is estimated by dividing total production

753

by the typical space use of hydro dams and their relative corridor spaces for transmission rows, whereas nuclear power is calculated as if it were fossil fuel. The net import of embodied energy in manufactured products (e.g. the net energy used in other countries to produce manufactured goods designed for the population studied) is also accounted for in this central part of the matrix. Trade flows are transformed into energy using intensity figures; the latter is then transformed into land using emission factors and the above-mentioned sequestration rate. In the final part of the consumption–land-use matrix, the EF and the available bio-productive space per capita are assessed in two separate boxes, described as demand and supply of land area. To aggregate the EF components relative to each type of land, equivalence factors are often used. These factors did not appear in W and R (1996), who essentially built up their model by equating different land-use forms, so that arable land, for instance, is regarded as productive as grazing land or forest. The equivalence factors function as weights in the aggregation translating ‘the specific land-use (such as world average cropland) into a generic biologically productive area (global average space) by adjusting for biomass productivity’ (WWF, 2000, p. 32). The EF obtained this way is expressed in area units (e.g. 1 hectare of biologically productive space with world average productivity) per capita and can be compared with that of other populations and with domestic supply of bio-productive space. The existing area in each land category is also multiplied by the equivalence factors, to consider varying productivity among land types and yield factors, in order to take account the extent to which a local landuse category is more productive than the world average in the same category. The possibly resulting areal ecological deficit does not indicate whether an area is sustainable, but it ought to be recognized that at a global level the minimum requirement for sustainability is that humanity’s footprint be smaller than the biosphere’s biological capacity. Even though the procedure outlined here has been improved since it was first developed, it still suffers from serious shortcomings and has been severely criticized by various scientists. Some of the unattractive features of this indicator will be described below, as they may be accountable for the variation in the empirical EF estimates. ECOLOGICAL FOOTPRINT: STRENGTHS AND WEAKNESSES The EF, as like any other aggregate indicator, is constructed using a great deal of detailed information and its appeal lies in synthesis. To avoid misleading interpretations from an aggregate environmental indicator and to use it profitably in the decision-making process, one must be informed about the data source,

754

Peter Nijkamp et al.

the aggregation process, the assumptions and the (implicit) weights used. To get a clear picture of the scope of EF, its main strengths and weaknesses will now be described as a summary of various observations made in the literature. The main goal is to explore whether factors may create uncertainty and hence variation in empirical EFs. Strengths The advantages of the aggregate EF indicator are due to its single measurement scale, which makes it quite straightforward for decision-making (C, 2000). Moreover, the EF is expressed in easily and widely understandable terms, namely land area per capita. Other empirical advantages of this indicator are its relatively easy calculation method and the frequent availability of the data at different spatial scales. Turning to the conceptual level, two characteristics of the EF are often pointed out, First, it recognizes the relevance of the second law of thermodynamics, as it can be thought of as the fictitious photosynthetic surface needed to replace the free energy dissipated by humans and their industrial metabolism. Second, it quantifies the dependence of regions and cities on their host ecosystems capturing Eugene Odum’s lament that ‘great cities are planned and grow without any regard for the fact that they are parasites of the country side . . .’ (R, 2000, p. 372). Put another way, the EF measures virtually the ‘geographical distribution of environmental stress of a given area relative to other areas’ (F and N, 2002, p. 248). To summarize, the EF is derived from ecological economics principles and it is an original tool to attract the interest of the general public, helping, therefore, to create a public debate on human pressures on the environment. Various scientists agree this is the EF’s greatest strength (C, 2000; M, 2000), even though they put forward many critiques. These will be discussed below. Weaknesses Land-use measurement: the conversion and the aggregation processes. A first critical comment often expressed in the literature refers to the use of global average yields in the – somewhat artificial – conversion of consumption data into land area. This may allow for consistent comparisons between different regions’ footprints, but at the same time it may obscure the comparative ecological advantage of a specific region, as the EF represents the hypothetical land – or imputed – area that would be needed to support a living standard, given certain – sometimes less plausible – conditions, and not the area actually used for the same purpose. Clearly, lack of data may be a bottleneck, so that in an applied context the researcher may have to resort to global average productivities in each category to convert all land uses to world average bio-productivity. In the

second place, using global average yields and assuming that the current harvest practices are sustainable, conventional agriculture is regarded as environmentally benign, which is far from the truth. Therefore, even though using land area as the EF’s measurement unit contributes to its effectiveness in gaining attention, the implicit assumptions need proper justification in an empirical setting. Such a justification may likely be found in the general purpose of the EF concept to demonstrate the decoupling of human activity from nature, to explore the spatial ecological demands of different lifestyles/incomes and to find a rough estimate of the human aggregate load relative to global carrying capacity. In the view of many scientists (e.g. F, 1999, 2000; V  B and V, 1999a; V K and B, 2000) another element that does nothing but add opacity to the result is the obscure use of equivalence factors in the aggregation of the EF components. Recognizing that land area used for different purposes cannot be aggregated as if it had the same biological productivity and the same pressure on the environment led the authors to the adoption of equivalence factors, although attributing the same factor to arable land and builtupon land (because the latter has taken the former’s place, supposedly) may create some confusion about the validity or relevance of the outcomes. For example, the current popularity of three-dimensional land use options offers a new possibility for multifunctional land use through which alternative and efficient environmental strategies can be implemented (V  B and V, 1999a; F and N, 2002). Despite possibly feeble and often criticized elements in the EF approach, it is an appealing approach in that it demonstrates clearly the strong dependency of humans on the ecosystems services as well as the disproportionally high claim of wealthier regions on the ecosystems, to the extent that global ecologicalcarrying capacity has been exceeded. Clearly, the limited data availability allows only for approximate calculations under rather strict assumptions. Energy land. An interesting feature of the EF regards the way energy consumption is accounted for. The reason for adopting the CO2 emission method, i.e. converting fossil fuel energy into the land needed to absorb the corresponding CO2 emissions, is that the land area estimates are rather similar to those resulting from an alternative method proposed, i.e. fossil energy land area needed to produce an equal amount of a biologically produced substitute. While the second option may seem more in line with the idea of maintaining a constant stock of natural capital (i.e. strong sustainability), the first method refers to the sink capacity of the environment. This method has been criticized by several authors, however, since CO2 assimilation by forests is only one of many options to

Ecological Footprints in Plural compensate for CO2 emissions. The main question at stake is, of course, whether a common denominator can be found that is consistent, measurable and widely applicable. Some authors have argued that this compensation scheme may be a rather land-intensive option and that in case of a high CO2 emission, there might not be enough available land (forest) to absorb all the carbon emissions. This criticism is only correct in that a lack of a land-based carbon sink simply means that the global-carrying capacity for fossil fuel use is insufficient. Other authors have claimed that maintaining forests carrying out this function may be rather expensive and that alternative solutions like shifting to other fuels, less fuel use or other ways to prevent CO2 build up in the atmosphere may be more convenient (F, 1999, 2000; V  B and V, 1999a). This is certainly a valid observation, but it calls for a search for alternatives without disqualifying the usefulness of the EF concept. If these alternatives are neglected, the sequestration rate used in calculations should be critically analysed. The standard EF literature refers to the net CO2 sequestering potential of an immature successional forest, but this may lead one to overestimate the net CO2 uptake because it saturates to zero when succession is completed, or to underestimate the net CO2 uptake because forest growth may be enhanced by higher concentrations of CO2 in the atmosphere (H, 2000). It should be admitted, however, that if a maturing forest loses (part of ) its net sink capacity, new forests (or other alternatives) would have to be established, which may be costly. It should be added that emissions of SO2 and NOx, which are related to fossil energy use and that contribute to acidification, are not accounted for in EF calculations. The same holds for other environmental externalities such as noise, toxic substances, landscape segmentation, spatial decay, etc. In conclusion, the EF tends to offer a conservative estimate of a global EF.

Ω

Ω

Demand and supply of land. An EF can be calculated at various levels: urban, regional, national and global. Its usefulness is supposed to descend from comparing it either to the biologically productive space available in the chosen spatial unit or to the fair earth share, e.g. global average biocapacity per capita. Many argue that these comparisons cannot produce significant information in terms of sustainability (V  B and V, 1999a, b; C, 2000) and that the EF is meaningful only at a global level. There may be many reasons for this, as follows: Ω

There may be a danger of misinterpretation of the EF when associating an ecological deficit with an unsustainable situation. This would suggest that the EF is an anti-trade biased indicator and implies that some form of autarchy is desirable (V  B and V, 1999a, b; F and N, 2002). Clearly, no one would deny that

755

the demand of rich countries tends to crowd out the production of sustenance in poor countries or that the current system of international trade, which ignores environmental externalities and differences in labour laws and conditions, is probably neither sustainable nor necessarily fair (C, 2000). Of course, economists would argue that there is no doubt on the trade’s theoretical capacity to improve the allocative efficiency of the regions involved, so that economists might have problems accepting an indicator for sustainable development that equals sustainable to self-sufficient, but this criticism concerns more the interpretation of the EF than its factual basis. Another caveat on the use of information on the ecological deficit concerns the choice of the areal units analysed. National boundaries, for instance, are of a political and cultural nature; they may be related to climate conditions and specialization patterns, but they do not have a clear or explicit environmental meaning. For instance, comparing a large, scarcely populated country with a high level of consumption and an ecological surplus with a small, densely populated country with, say, an equal level of consumption but a significant ecological deficit, does not immediately offer relevant information about which is on the right track to achieve sustainability. The lifestyle that people living in these regions enjoy may be exactly the same, but the dense area seems to put more pressure on the ecosphere because of the areal unit and standardization used (land area per capita). Even though densely populated areas (e.g. urban regions) typically confer high ecological benefits – as a result of environmental and energy technology use (e.g. waste management, district heating, etc.) – they often suffer from an ecological deficit. This observation suggests that EFs have to be used with great care. Similarly, it is difficult to draw meaningful conclusions from the comparison between the EF and the fair earth share. Actually, the EF advocates admit in their most recent application of the EF (WWF, 2000) that no relevant conclusion on sustainability can stem from these comparisons; this recognition reduces somewhat the strength of the EF as a sustainability indicator. At any level apart from the global, the EF should be seen as the net input of virtual land from outside the analysed system. The EF is essentially more a meaningful ecological dependency indicator for a given area, which is by its very nature scale dependent. The geographical scale inherent in the EF calculations is, therefore, the Achilles’ heel in its methodology.

Policy directions. Another often mentioned and less attractive feature of the EF is that it does not offer straightforward and guiding policy suggestions and does not refer to instruments for environmental policy

756

Peter Nijkamp et al. Table 1. Codification of attributes

A B C D

Type of documentation Year of data collection Geographical scale Methodology

1

2

3

journal 1991–93 national techniques based on Wackernagel and Rees

research paper 1994–96 regional input–output model

commissioned report 1997–99 urban component-based approach

mechanisms (V  B and V, 1999a; M, 2000; F and N, 2002). Given that there are uncertainties connected to the results of any policy implementation, a fundamental characteristic for any sustainable development indicator is its capacity to monitor the situation in order to give guidance to the decision-makers. It seems meaningful to state that an EF serves to create awareness, as it firmly indicates the need for action and points to the distribution of environmental pressure around the world. In conclusion, from an analytical perspective, the EF is a fascinating research issue; from a policy action point of view, its relevance in terms of feasible instruments remains to be seen. Nevertheless, the EF demonstrates clearly that it is a sound precautionary principle for all countries and regions to think more critically about their ecological deficit, as a confirmation of current trends will affect the sustainability of the planet. META-ANALYTICAL FRAMEWORK FOR A COMPARATIVE ECOLOGICAL FOOTPRINT STUDY Despite clear weaknesses, the EF concept has attracted much attention and has generated a wide range of applied studies in many regions and countries. The methodological shortcomings are related to the EF calculation method and to the geographical scale bias. Empirical problems may emerge, as the calculation method has changed over the years. Furthermore, there may be flaws in empirical estimates depending on the publication channel; as argued by B and N (2000), empirical results may exhibit a publication bias. In the years following the invention of the EF, many efforts were made to quantify this concept in order to assess the pressure of different populations on the environment. This was done at different scale levels ranging from the urban or regional to the national or global level. These efforts resulted in a wide range of published and unpublished articles as well as reports by governmental and environmental organizations. A systematic overview and analysis of these EF applications can offer useful insight into the factors that determine footprints and can lead to a deeper understanding of the indicator’s meaning in practice. To

4

techniques based on Wackernagel and Rees using local yields

trace systematically commonalities and differences, the present paper will use ideas and approaches derived from meta-analysis. Meta-analysis aims to develop and deploy a framework for quantitative research synthesis based on previously undertaken studies (for reviews, see V  B et al., 1997; and F, 2002). There is a wide variety of meta-analytical techniques, but in all cases they aim to distinguish between different background variables that might explain variations in outcomes, including research methods and spatial and time data. Some features that may help to classify the great diversity in EF studies and their results and that can be found in all empirical EF studies are as follows (Table 1): Ω

Ω

Ω

Ω

Type of documentation: may influence the outcome of the application in its reliability and scientific soundness. An article published in an academic journal should be more complete, and probably less biased, than a report commissioned by a public authority, also because of strict refereeing procedures of professional scientific journals. Year of data collection: the EF calculation relies on data on production, trade and productivities, aggregates that typically change over time; for instance, consumption data from the beginning of the 1990s might be very different from the same type of information referred to a decade later, e.g. if there has been a financial crisis in the meanwhile. Geographical scale: spatial unit analysed can have peculiarities that affect the footprint. Cities, for instance, typically have higher densities that can lead to environmental advantages thanks to economies of scale or scope; these could be reflected by lower environmental pressure and, hence, EFs. However, cities are compact areas that have to import many foods, which might lead to a higher EF. Methodology: the calculation procedure adopted in empirical EF studies can vary widely. In the past, both alternative and complementary methodologies to the initial approach of W and R (1996) have been developed. On the one hand, techniques relying on input–output models have been designed as a response to the complaint that the EF does not account for interindustrial uses, assuming therefore that household consumption is on average equal to production plus net imports;

Ecological Footprints in Plural W et al. (1999a, b) try to solve this problem by converting some manufactured traded products into raw material. To express consumption in terms of land, these techniques rely on actual domestic land use (per unit of monetary outcome) and on Leontief ’s inverse matrix, which captures all backward linkages in production. Therefore, the input–output technique may be useful when calculating the EF of those countries that, for instance, are big importers of raw materials that are subsequently processed in exported goods. Clearly, the input– output model deployed should have the same geographical scale as the EF under consideration. Anyhow, a drawback of the methodologies based on the input–output model is formed by the assumptions it rests on and that limit its use, above all in forecasting. On the other hand, the so-called component-based approach can be seen as a valid complement to the original synthetic EF method when dealing with data at a subnational level and when trade statistics are not available. It considers several categories of environmental pressure (transport, waste, water, etc.) instead of looking at the consumption of raw materials and energy. It is a ‘bottom-up’ analytical approach in the sense that the EFs for certain activities are precalculated using data appropriate to the region under study. For instance, when the embodied energy of a kilogram of waste of a certain city is determined, then, through data on the amount of waste produced, the loss of energy related to the production of waste in the given urban area may be calculated. This methodology involves a sensitivity analysis of a range of data sources to determine the most representative footprint conversion factors. At the urban level, it may be feasible to extrapolate footprints through comparison of local and national consumption patterns. In addition, some researchers have chosen to adopt the EF calculation procedure by using local instead of global yields. Global average yields are usually lower than those in most developed countries. Here productivities have risen considerably in the last century thanks to the use of pesticides, fertilizers, genetically modified technology and more, in general because of the industrializing process that farming has gone through. Therefore, if a country produces efficiently (economically and ecologically) a biotic resource, but if average yields are used in the conversion process, the actual comparative advantage disappears. Then the resulting EF may be misguided and has to be interpreted cautiously. Adopting local yields may make land use less hypothetical, but it still says nothing about how sustainable are these harvest rates. Thus, the variety in the adopted methodology could be one of the explaining factors for varying EF estimates. These attributes can be used to classify the EF applications and to understand why countries with different

757

features or areas of different sizes have EFs in the same range or why areas with similar characteristics have different footprints. In this way, one can also test whether the weak elements implicit in the EF calculations have implications for the empirical results.

INFORMATION BASE OF ECOLOGICAL FOOTPRINT APPLICATIONS There is obviously a need for a comparative study on EFs. The present analysis has tried to collect a wide range of empirical EF studies in order to extract some commonalities and contrasts from them. The selected case studies cover a wide range of cities, regions and countries. All results are expressed in terms of hectares/ capita and can be compared effectively. Unfortunately, not all applications in the sample appeared to compare directly demand (EF) and supply (available biocapacity) to check for an ecological deficit or a remainder, so that it was useless to include this information in the database. Moreover, regarding countries with an ecological deficit as unsustainable has been severely criticized (see below) to the point that Wackernagel himself (WWF, 2000) has admitted that one cannot draw sheer conclusions on sustainability from the existence of observed gaps between the artificial demand and supply of land. Table 2 lists the case studies that will be analysed in the present meta-analytical experiment. Some of them analyse more than one country at a time or use more than one methodology in the same study. Therefore, the footprints that will be looked at do not entirely correspond to the number of surveyed documents. In particular, Wackernagel et al. prepared a report for the Rioò5 summit in which they presented the footprints of many countries; however, in their articles about the EF in official journals (i.e. Ecological Economics and Ambio), they analysed, respectively, Italy and Sweden. Since a great deal of information on these two mentioned countries was disposed of, it was decided to include them. The greatest part of the areas analysed appears to belong to the industrialized world; clearly, this is a consequence of the higher interest shown in developed countries towards environmental problems that, in turn, have been created mainly by their own economic activity. Some developing areas were also examined, but they constitute a much smaller fraction of the sample. In a few articles, EF time series referring to periods in the last century are calculated. They are a first step towards using this indicator as a forecasting tool. However, the vast majority of studies yield as an outcome a figure for a single year and, as suggested by R (2000), they take a snapshot of the situation, i.e. they give a static indication. Nearly all the data are from governmental sources or United Nations statistical

758

Peter Nijkamp et al. Table 2. Overview of the selected studies

Reference a

W et al. (1999a) W et al. (1999b) V V and S (2000)b

B et al. (1998) F (2001) B and S (2001) B and P (1999)c W (1999) WWF (2000)d O et al. (1998) L and M (2001) F and N (2002) H et al. (1999) P (1998) P (2000) B (2001) W-R and K (2001) S et al. (2000) R G and P N (1999) H et al. (2001)b

Analysed unit

Ecological footprint (ha/capita)

Italy Sweden Benin (l) Bhutan (l) Costa Rica (l) the Netherlands (l) Benin (g) Bhutan (g) Costa Rica (g) the Netherlands (g) New Zealand Taiwan Liverpool, UK

4.2 7.2 0.7 0.9 1.7 2.4 0.7 0.6 2.1 4.6 3.49 1.07 4.15

Orvieto, Italy Santiago de Chile India Toronto, Canada Australia Regione Marche, Italy Scotland, UK Japan Regione Liguria, Italy Guernsey Island, UK

2.25 2.64 1.061 7.6 13.6 2.93

Hong Kong, China Australia

6.3 5.96

Barcelona, Spain Austria (l) Austria (g)

3.23 4.67 3.24

2.4 6.3 3.64 8.55

Notes: aCase study extracted from the Rio ò5 Report in which the EFs of 52 different nations are calculated. b l, use of local yields; g, use of global yields. c In this application, the footprints of three medium-sized Italian towns were calculated; since Italy is already well represented, only one value for Orvieto has been taken. d India was taken to diversify the survey; in the Living Planet Report (WWF, 2000) (the source of this case study), all countries with populations over 1 million are analysed.

databases and refer to the last decade, as could be expected since the EF was invented in 1996. The values that the EF assumes in the survey range from a minimum of 0.6 ha/capita for Bhutan to a maximum of 13.6 ha/capita for the average Australian; they suggest a direct relationship between a country’s richness and its footprint. Such a correlation may seem obvious, since the EF calculation is nothing but the translation into land of a population’s standard of life or – better – of its level of consumption of biotic resources and energy. However, the so-called apparent consumption (i.e. production ò imports – exports) is highly correlated to gross national product, since it is constituted in great part by the same aggregates. The

Table 3. Coded information base on ecological footprint studies Case study 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Year of data 1 2 2 2 2 2 2 2 2 2 2 1 3 2 2 2 3 2 2 1 2 3 3 3 1 2 2 2

Type of documenta- Geograph- Methodtion ical scale ology 1 1 1 1 1 1 1 1 1 1 1 1 3 3 3 1 3 1 2 1 1 3 1 1 1 3 1 1

1 1 1 1 1 1 1 1 1 1 1 1 3 3 3 1 3 1 2 2 1 2 2 3 1 3 1 1

1 1 4 4 4 4 1 1 1 1 2 2 3 1 1 1 1 2 1 1 1 1 3 1 1 1 4 1

Ecological footprint High Very high Low Low Low Medium Low Low Medium High Medium Low High Medium Medium Low Very high Very high Medium Medium Very high Medium Very high Very high High Medium High Medium

EFs in the present sample comprise both point estimates and ranges. The calculated EFs can be classified as follows: a ‘very high’ footprint (score 1) is [6 ha/capita; ‘high’ is between 4 and 6 ha/capita (score 2); ‘medium’ is between 2 and 4 ha/capita (score 3); and ‘low’ is \2 ha/capita (score 4). For the classification of all the present empirical EF cases according to the results, see Table 3. It also contains information about the studies; to this scope, the qualitative information table has been deployed (Table 1). Table 3 forms the empirical input for the meta-analytical comparison to be undertaken in the next section. COMPARATIVE ANALYSIS OF THE ECOLOGICAL FOOTPRINT DATABASE The 20 documents surveyed gave 28 areal units that finally constitute the EF database. Each unit was classified according to its footprint and to the qualitative attributes listed in Table 1. The outcomes can be found in Table 3; the latter can also be used to present some diagrams on the frequency of different qualities or properties in the sample. Regarding the attribute ‘type of documentation’, 75% of the studies that make up

Ecological Footprints in Plural

759

Fig. 1. Frequency of type of documentation

Fig. 3. Frequency of the methodology used

Fig. 2. Areal units by geographical scale

Fig. 4. Frequency of ecological footprint value classes

the database were obtained from articles published in academic journals; 21% come from reports, while only 1/28 (3.5%) areas were from unpublished research papers. Fig. 1 shows a description of the latter. The distribution of the areal units by geographical scale is also quite unbalanced; 64% of the sample is constituted by nations and 22% of the areal units are urban areas; the remaining ones are regions (Fig. 2). Given the characteristics of the data, i.e. trade statistics, necessary to calculate an EF – at least in the original method – it is conceivable that mainly countries are analysed since often they are the only administrative regions that make available this sort of data (W and S, 2000). Fig. 3 refers to the methodology adopted to calculate the EFs of the studies belonging to the sample. Many of the footprints, 64%, were calculated relying on the methodology proposed by Wackernagel and Rees (Tech_W&R); however, at a subnational level, this can be quite imprecise since it consists of an extrapolation of the region’s or city’s EF through comparison of the consumption patterns. To obtain the EF of some areal units (18% of the sample), the authors have apparently used local yields, whereas only 11% of the sample is constituted by units whose EF is calculated through

the input–output approach. Finally, 7% are urban areas to which the component-based approach was applied. The classification of the areas according to their footprints can be seen in Fig. 4. The majority of surveyed units, 36%, have a footprint between 2 and 4 ha/capita, which has been defined a medium EF; 21% of the analysed areas have a very high EF, while 18 and 25%, respectively, have a medium and a low footprint. The paper will now seek to identify pairwise correlations among the EF attributes by adopting a crosstabulation method. Table 4 classifies the analysed spatial units according to the EF class they fall into and to the type of documentation to which they refer. Nearly all Table 4. Cross-classification of areas by type of documentation and ecological footprint

Journal Reports Research paper Total

Very high

High

Medium

Low

Total

5 1

4 1

5 4

7 0

21 6

0 6

0 5

1 10

0 7

1 28

760

Peter Nijkamp et al.

Table 5. Cross-classification of areas by geographical scale and ecological footprint Very high

High

Medium

Low

Total

3 1 2 6

4 0 1 5

4 3 3 10

7 0 0 7

18 4 6 28

National Regional Urban Total

Table 6. Cross-classification of areas by methodology and ecological footprint

Tech_W&R Input–output Component approach Tech_W&R_ly Total

Very high

High

Medium

Low

Total

4 1

3 0

8 1

3 1

18 3

1 0 6

1 1 5

0 1 10

0 3 7

2 5 28

Notes: W&R, Wackernagel and Rees. ly, Local yields.

of them, 21/28, are analysed in articles taken from scientific journals. There seems to be no connection between this attribute and the value (or code) of the EF classes. The studies extracted from commissioned reports most often yield medium results; this could be a consequence of the less strict procedures and lower completeness that tend to characterize reports commissioned by city councils or local or regional administrations. In fact, all these units are in industrialized areas that typically have high or very high EFs. Looking at the distribution of the present EF studies according to EF classes and geographical scale, some additional conclusions may be drawn (Table 5). Regions and especially cities seem to have bigger footprints than nations. This result makes sense, since the smaller is an open region, the higher will be the net input of resources. The EFs of nations seem to be distributed quite equally in the various EF classes, even though there is a predominance of countries with low EFs. Classifying the spatial units according to the methodology used to calculate their EF (Table 6), note that the component-based approach, which is adopted uniquely for urban areas, yields very high or high results. On the other hand, applying techniques that rely on the original methodology invented by Wackernagel and Rees but modifying it by using local yields (Tech_W&R), the results are somewhat lower. Once again, this is an expected outcome, because global average yields are lower than the average yields in industrialized areas that constitute the greatest part of the sample. In conclusion, the exploratory analysis of EF results has confirmed most of the authors’ prior expectations. The next section will try to identify a structure in the complex data set of EFs.

PATTERN RECOGNITION IN THE ECOLOGICAL FOOTPRINT DATABASE: DECISION-TREE INDUCTION Decision-tree results The EF information base is qualitative in nature, but despite its linguistic meaning, it may have a hidden structure. To this end, the present paper will use a recently developed pattern recognition technique that may be used for meta-analytical investigation. The present section will use a machine-learning approach based on decision-tree induction to gather meta-data and summarize findings regarding the EFs. From our set of studies on sustainable development, four classes of main attributes characteristic of these studies were extracted (Table 1). Each study also offers an estimate for the EF. From this value and the four classes of attributes, one can now derive a treelike form hierarchical classification model that uncovers the implicit structure of the present database. Finally, from this classification, a set of rules will be extracted aimed at describing the most relevant features of the database by means of conditional explanatory statements. The ability of the system as a classifier was tested using the classification accuracy criterion and the so-called confusion matrix. For more details, see the Appendix. A number of systems exist for inducing classification trees from empirical experiments with categorical information (e.g. B et al., 1984; C et al., 1986; Q, 1993). In the present EF experiment, the C5/See5 algorithm was implemented. It is one of the best known and mostly used decisiontree systems in the literature on artificial intelligence (Q, 1986). Tree pruning is an important mechanism used in See5 to prevent trees from over-fitting data. It can be employed during the process of tree construction. Another important feature of See5 is its ability to take into account misclassification cases. Decision-trees were next induced for the present set of studies to predict whether it is possible to identify a relationship among the characteristics of the studies and the variation in EFs. Table 7 and Fig. 5 show the main branches of the tree. The statistic in brackets summarizes the performance of the classification, i.e. the number of cases in each leaf, and the cases incorrectly classified by the leaf. A non-natural number of cases can arise because in a case when the value of an attribute in the tree is not known, the system splits the case and sends a fraction down each branch. These are the branches that contain most of the positive examples for each class of values of the EF. Branches correspond to classification rules, in the decision-tree literature also called ‘if–then’ rules. The model summarizes the common features of the four single classes of values. Some patterns of their interconnections, which have plausible explanations, can

Ecological Footprints in Plural Table 7. Decision tree for the data set of the ecological footprint studies (nó28) Geographical scaleóRegional: Medium (4/1) Geographical scaleóUrban: :. . . Year data collectionó1991–93: Medium (0) :. . . Year data collectionó1994–96: Medium (3) :. . . Year data collectionó1997–99: Very high (3/1) Geographical scaleó1: :. . . Year data collectionó1997–99: Low (0) Year data collectionó1991–93: High (3/1) Year data collectionó1994–96: :. . . MethodologyóComponent appr.: Low (0) MethodologyóTech_W&R: Low (13/7) MethodologyóInput–output: Medium (2/1)

now be observed. The model apparently can be described by three attributes: geographical scale, year of data collection and methodology. The most important result is the relevance of the temporal and spatial dimensions of environmental sustainability analysed in the set of our EF studies. The geographical scale of the case studies is the attribute that can best discriminate among the others. One can explain the importance of this attribute based on the large discrepancies in accessibility, environmental and resource endowments, soil characteristics, and climate conditions at a spatial level. Next, the difference in the inter-temporal dimension at the urban level appears to show relatively higher EFs. This difference can be viewed as a reflection of the difference in the spatial distribution of wealth. It can also be understood not only as a sign of unsustainability, but also as the spatial agglomeration effect due to the economies of scale and scope (V  B and V, 1999a). Differences at spatial level 1 are more difficult to explain in terms of land area, but here the type of methodology adopted in the EF studies can partly explain different results in EFs. After the construction of the decision-tree, a set of six conditional deterministic (i.e. ‘if–then’) decision

761

Table 8. Set of conditional rules Rule 1: (13/7, lift 1.9) If (Year of data collectionó1994–96) & (Geographical scaleó1) & (MethodologyóTech_W&R), then EF class Low [0.467] Rule 2: (4/1, lift 1.9) If (Geographical scaleóRegional), then EF class Medium [0.667] Rule 3: (3, lift 2.2) If (Year of data collectionó1994–96) & (Geographical scaleóUrban), then EF class Medium [0.800] Rule 4: (3/1, lift 2.8) If (Year of data collectionó1997–99) & (Geographical scaleóUrban), then EF class Very high [0.600] Rule 5: (3/1, lift 3.4) If (Year of data collectionó1991–93) & (Geographical scaleó1), then EF class High [0.600] Rule 6: (2/1, lift 1.4) If (Year of data collectionó1994–96) & (MethodologyóInput– output), then EF class Medium [0.500]

rules was induced (Table 8). For the present sample of 28 studies, these rules appear to have 11 misclassified items or a 39.3% error rate. Rules are ordered by class and subordered by confidence. The rule that most reduces the error rate appears first and the rule that contributes least appears last. The rule’s accuracy is estimated by the Laplace ratio (nñmò1)/(nò2). The indicator lift x is the result of dividing the rule’s estimated accuracy by the relative frequency of the predicted class in the training set. The results from the decision-tree induction method appear largely to confirm the results from the frequency analysis and the cross-tabulation methods, especially as far as the impact of the spatial scale, the period and the methodology deployed is concerned. Evaluation of the classification algorithm Classification accuracy is the most commonly used criterion to evaluate the success or failure of a classification system. Accuracy is measured as the percentage of events correctly classified. In the present system, this is estimated as 60.7%. In the case of balanced data sets and

Fig. 5. Results of the decision-tree induction method. EF, ecological footprint

762

Peter Nijkamp et al.

Table 9. Confusion matrix of ecological footprint studies Classification by model High 2 1

Low 2 6 3 2

Medium

7 2

Very high

Actual class

1

High Low Medium Very high

2

highlighted patterns in EF estimates among studies and have tried to isolate the relationship between particular characteristics of studies and the resulting estimates of EFs in a multivariate analysis, these are only statistical associations. It would be a significant step forward to go beyond association and to develop causal patterns, applying the above methodology to an extensive narrative review. RETROSPECT AND PROSPECT

if all errors are of equal importance, this measure summarizes the overall performance of the classifier. In the present case study, distinctions among different types of errors turn out to be important. Therefore, a so-called confusion matrix was constructed (K and P, 1988). Table 9 shows the confusion matrix for the four classes considered. It lists the correct classification against the predicted classification for each class. The number of correct predictions can be found on the diagonal of the matrix. All other numbers represent the number of errors for a particular type of misclassification error. In the context of the present study, the entries of the confusion matrix have the following meaning for each successive column: Ω

Ω

Ω

Ω

Two is the number of correct predictions that instances are classified in class ‘high’; 1 is the number of incorrect predictions that an instance of the class ‘low’ is classified in class ‘high’. Six is the number of correct predictions that instances are classified in the class ‘low’; 2 is the number of incorrect predictions that an instance of class ‘high’ was classified in the class ‘low’; 3 is the number of incorrect predictions that an instance of the class ‘medium’ was classified in class ‘low’; 2 is the number of incorrect predictions that an instance of the class ‘very high’ was classified in the class low. Seven is the number of the correct predictions that instances are classified in class ‘medium’; 2 is the number of incorrect predictions that an instance of the class ‘very high’ was classified in the class medium. Two is the number of the correct predictions that instances are classified in the class ‘very high’; 1 is the number of incorrect predictions that an instance of the class ‘high’ was classified in the class ‘very high’.

Collating the results of the present analysis and subjecting them to systematic, quantitative scrutiny, this approach can help to answer a range of questions of primary importance for decision-makers concerned with the use of the EF as an indicator for sustainability. The use of the decision-tree method has helped in the determination of what evidence is available about the variation of EFs and to identify how and why the EF estimate differs among studies. Although we have

This contribution has served as a critical but constructive assessment of the EF concept. It has not only highlighted its importance from a spatial sustainability perspective, but also questioned the validity of various assumptions made. The empirical part has arrived to test the sensitivity of the EF results for various moderator variables, in particular the role of the geographical scale, the possibility of a publication bias, the impact of the period considered and – last but not least – the influence of the methodology used. The present study has thus tried to identify the background factors explaining the variation in EFs in a range of empirical studies. The variability of EFs appears to be rather high, but the present analysis has demonstrated that the spatial scale, the year of undertaking the data collection and the specific technique adopted offer a significant explanatory contribution for this phenomenon. The relatively high EFs at the urban level prompt an intriguing question in the current localization–globalization debate. The various empirical studies demonstrate that more emphasis on localized modes of consumption and production seem to increase the range of spatial externalities. This might suggest a contradiction with the antitrade bias inherent in the EF paradigm. On the other hand, it ought to be recognized that cities are by definition based on labour specialization and hence on trade. It would be a mistake to think that localized production and consumption patterns are feasible with low EFs within strict geographical boundaries. This observation prompts also another issue, namely that of sectoral specialization in a given area. Geographical conditions are historically decisive for the emerging pattern of production opportunities and they tend to force people to adopt the economically most efficient and long-term sustainable production modes. A change toward lower EFs may likely be achieved to the detriment of economic efficiency. This also provokes a scientific reflection on local and sectoral metabolism and on the economic and ecological gains from industrial transformation. Therefore, due attention for the spatial scale in the EF calculations for a certain area is a sine qua non, as is also evident from the present results. Finally, another research issue is in order. In the context of the EF literature, the spatial component has received a dominant position. Despite much talk on multigenerational sustainability, hardly any empirical

Ecological Footprints in Plural work has been undertaken to assess the EFs of the current generation on the next cohorts or the subsequent generations. Such an exercise might perhaps sound somewhat speculative, but might bring to light interesting findings. The use of scenario experiments might perhaps be helpful in this framework for future thinking also using causal spatial–ecological patterns. Acknowledgements – The authors thank Jeroen van den Bergh and two anonymous referees for extremely helpful and constructive comments on a previous draft of the paper.

APPENDIX: LEARNING DECISION TREES WITH C5/SEE5 Through classification, it is possible to uncover information contained in a multivariate database, discovering structural relationships between class characteristics and relevant attributes of the phenomena to be classified. The method of decision-tree induction, which belongs to the class of multidimensional classification methods (such as neural network analysis, fuzzy set analysis, rough set analysis or decision-tree analysis), is used widely and increasingly for classification purposes. This method aims at analysing and predicting a class membership by a recursive partition of a multidimensional data set into more homogeneous subsets (for details, see B et al., 1984). This leads to a hierarchical decision-tree structure where instances are classified by sorting them down the tree from the root node to a leaf node, which provides the classification of the instance. Each node in the decision-tree specifies a test for the attribute of the instance concerned, and each branch descending from the node corresponds to one of the possible values for this attribute. An instance is classified by starting at the root node of the decisiontree, by testing the attribute specified by this node and by the next node down the tree branch that corresponds to the value of the attribute. This process is then repeated at the node on this branch and so forth until a leaf node is reached. In a decision-tree algorithm, the critical step method is the one used to assess splits at each internal node of the tree. Often the information-theoretic approach, which examines the entropy in relation to the information contained in a probability distribution, is deployed (S, 1948). The aim is then to select the

763

attribute that is most useful for classifying instances, based on the so-called information gain (a measure for the goodness-of-separation of a given attribute for the training examples according to their classification (for details, see DF and C, 2000). Thus, entropy is then used as a measure of the reduction of disorder when ordering a set of variables in a data set with respect to different classes. By interpreting information gain as a measure of the expected reduction in entropy, one can – by considering the next node down – define a measure of the effectiveness of an attribute in classifying the training data, caused by positioning the instances according to this attribute. The process of selecting a new attribute and positioning the training examples is then repeated for each non-terminal descendent node, this time using only the training examples associated with the node concerned. Attributes that have been incorporated higher in the tree are excluded, so that any given attribute can appear at most once along any path in the tree. Formally, the information gain of an attribute is computed by means of the corresponding entropy expression. Given a training dataset T, composed of observations belonging to one of k classes {C1, C2 , . . ., Ck}, the amount of information required to identify the class for an observation in T is:




k



freq(Cj , T) freq(Cj , T) *log2 DTD DTD jó1

Info(T)óñ ;

where freq(Cj , T) is the number of cases in T belonging to class Cj , and DTD is the total number of observations in T. It is the average amount of information required to define the class of a sample from the set T. In terms of information theory, it is called entropy of the set T. The same estimate but after separation of the set T with X is provided by: n

InfoX (T)ó ; ió1



Ti *Info(Ti ) T

The following formula is then the criterion of the attribute choice: Gain(X)óInfo(T)ñInfoX (T) This criterion is calculated for all the attributes and the one that maximizes the expression is then selected. This attribute is the test used in the current tree node, while it will be used for further tree derivation.

REFERENCES B S. and N P. (2000) Meta-analytic methods for comparative and exploratory policy research, Journal of Policy Modelling 21, 821–858. B J. (2001) Component ecological footprint: developing sustainable scenarios, Impact Assessment and Project Appraisal 19, 107–118. B J. and S A. (2001) The ecological footprint: a metric for corporate sustainability, Corporate Environmental Strategy 8, 1–10. B K. B., B R. J., C R. and B H. R. (1998) New methodology for the ecological footprint with a application to the New Zealand economy, Ecological Economics 27, 149–160.

764

Peter Nijkamp et al.

B G. and P M. (1999) Un applicazione sperimentale dell’impronta ecolica in tre piccole citta del centro e del sud d’Italia, in B G. and P A. (Eds) L’Impronta Ecolica, pp. 18–31 (available at: http://www.wwf.it/ ambiente/attenzion/13.pdf ). B L., F J., O R. and S C. (1984) Classification and Regression Trees. Wadsworth, Belmont. C R., C R. and N P. (1998) Towards sustainable city policy: an economy–environmental–technology nexus, Ecological Economics 24, 103–118. C B., K I. and B I. (1986) ASSISTANT 86: a knowledge elicitation tool for sophisticated users, in B I. and L N. (Eds) Progress in Machine Learning, pp. 31–45. Sigma, Wilmslow. C R. (2000) The dynamics of the ecological footprint concept, Ecological Economics 32, 341–345. D B S. M. (1999) Economic growth and the environment. PhD thesis, Vrije Universiteit, Amsterdam. DF R. S. and C J. C. W. (2000) Multiple criteria for evaluating machine learning algorithms for land cover classification from satellite data, Remote Sensing of the Environment 74, 503–515. F A. R. B. (1999) The essence of ecological footprints, Ecological Economics 31, 318–319. F A. R. B. (2000) Academic perspectives on eco-footprinting. Optimum Population Trust Working Paper. Draft. F J. J. (2001) Using composition of land multiplier to estimate ecological footprints associated with production activities, Ecological Economics 37, 159–172. F A. and N P. (2001a) Pathways to urban sustainability, Journal of Environmental Policy and Planning 3, 289–302. F A. and N P. (2001b) Towards a sustainable future of cities in Europe: an evaluation of sustainable city initiatives using multicriteria decision support methods, in S J. and S H. (Eds) Land-use Simulation for Europe, pp. 173–192. Kluwer, Dordrecht. F A. and N P. (2002) Spatial fingerprints and ecological footprints, in H Y. Y., N P., P J. and V W J. (Eds) The Region in the New Economy, pp. 241–262. Ashgate, Aldershot. F R. J. G. M. (2002) Accounting for dependence among study results in meta-analysis: methodology and the applications to the valuation and use of natural resources. Paper Masterpoint. Free University, Amsterdam. G-R N. (1976) Energy and Economic Myths: Institutional and Analytical Economic Essays. Pergamon, New York. G E. (1998) Are cities dying?, Journal of Economic Perspectives 12, 139–160. H H., E K. H. and K F. (2001) How to calculate and interpret ecological footprints for long periods of time: the case of Austria 1926–1995, Ecological Economics 38, 25–45. H N., M I., F R. and W M. (1999) Measuring sustainability: a time series of alternative indicators for Scotland, Ecological Economics 28, 55–73. H R. A. (2000) Ecological footprint is a vivid indicator of indirect effects, Ecological Economics 32, 357–358. K R. and P K. (1988) Editorial: Glossary of terms, Machine Learning [Special Issue: Application of Machine Learning and Knowledge Discovery Process] 30, 271–274. L M. and M S. A. (2001) A modified ecological footprint method and its application to Australia, Ecological Economics 37, 229–255. M I. (2000) Ecological footprints and sustainable development, Ecological Economics 32, 359–362. M R. and W B. (2003) Rapid worldwide depletion of predatory fish communities, Nature 423, 280–283. N P. and P A. (1994) Sustainable Cities in Europe. Earthscan, London. O L. J., K E. and W M. (1998) How big is Toronto’s ecological footprint? (available at: http:// www.city.toronto.on.ca/eia/efreport.pdf ). O J. B. (1992) Economic Incentives and Environmental Policies. Kluwer, Dordrecht. O I. (1982) Urbanisation and Environmental Quality. Kluwer, Dordrecht. P A. (2000) Valutazione dell’impronta ecologica della region Liguria (available at: http://www.ecozero.liguriainrete.it/ ecozero/impronta.htm). P P. (1998) An environmental measure of Japan’s economic development: the ecological footprint, Geographische Zeitschrift 86, 106–119. Q J. R. (1993) C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo. Q J. R. (1986) Induction of decision trees, Machine Learning no. 1, 81–106. R W. E. (1992) Ecological footprints and appropriated carrying capacity: what urban economics leaves out, Environment and Urbanization 4, 121–130. R W. E. (2000) Eco-footprint analysis: merits and brickbats, Ecological Economics 32, 371–374. R W. E. (2002) Globalization and sustainability: conflict or convergence?, Bulletin of Science, Technology and Society 2, 249–268. R W. E. and W M. (1996) Urban ecological footprints: why cities cannot be sustainable and why they are a key to sustainability, EIA Review 16, 223–248. R G G. and P N F. (1999) Application of the ecological footprint to Barcelona: summary of the calculations and thoughts on the results, Medi Ambient i Serveis Urbans 6, 67–76. S C. E. (1948) A mathematical theory of communication, Bell System Technical Journal 27, 379–423, 623–656. S R. W., P A. and L I. (2000) An ecological footprint analysis for Australia, Australian Journal of Environmental Management 7, 11–18. S A. (1776) Inquiry into the Nature and Causes of the Wealth of Nations. London. S S. and S G. (1996) Politiche di riqualificazione delle aree metropolitane: domanda di valutazione e contributo delle discipline economico-estimative, in Proceedings of the XXVI CESET Conference, Milan, Italy, 17–18 October.

Ecological Footprints in Plural

765

UNCED (1992) Agenda 21: Programme of Action for Sustainable Development. Rio Declaration on Environment and Development. Rio de Janeiro. V  B J. C. J. M., B K., N P. and P G. (1997) Meta-analysis in Environmental Economics. Kluwer, Dordrecht. V  B J. C. J. M. and V H. (1999a) Spatial sustainability, trade and indicators: an evaluation of the ecological footprint, Ecological Economics 29, 61–72. V  B J. C. J. M. and V H. (1999b) An evaluation of the ‘ecological footprint’: reply to Wackernagel and Ferguson, Ecological Economics 31, 319–321. V K G.-C. and B E. H. (2000) The ecological footprint: useful science or politics?, Ecological Economics 32, 357–358. V V-G D. B. and N P. (1999) Globalisation, transport and the environment: new perspectives for ecological economics, Ecological Economics 31, 331–346. V V D. P. and S E. M. W. (2000) Ecological footprints of Benin, Bhutan, Costa Rica and the Netherlands, Ecological Economics 34, 115–130. V E. T. and N P. (2002) Externalities in urban sustainability. Environmental versus localization-type agglomeration externalities in a general spatial equilibrium model of a single-sector monocentric industrial city, Ecological Economics 40, 157–179. W M. (1999) L’impronta ecolica di Santiago del Cile, in B G. and P A. (Eds) L’Impronta Ecolica, pp. 141–152 (available at: http://www.wwf.it/ambiente/attenzione/13.pdf ). W M., L L. and B¨  H. (1999b) Evaluating the use of natural capital with the ecological footprint, Ambio 28, 604–612. W M., O L., B P., C A. L., L´  I. S. F., M´  J. G., G´  A. I. G., S´  A. I. G. and S´  M. G. G. (1999a) National natural capital accounting with the Ecological Footprint concept, Ecological Economics 29, 375–390. W M. and R W. (1996) Our Ecological Footprint. Reducing Human Impact on the Earth. New Society, Gabriola Island. W M. and S J. (2000) Big things first: focusing on the scale imperative with the ecological footprint, Ecological Economics 32, 341–394. W-R K. and K A. (2001) Ecosystem appropriation by Hong Kong and its implications for sustainable development, Ecological Economics 39, 347–359. W C  E  D (WCED) (1987) Our Common Future. Oxford University Press, Oxford. W R I (2001–02) People and Ecosystems: The Fraying Web of Life. WRI. W W F  N (2000) Living Planet Report. WWF, Washington, DC (available at: http://www.panda.org/ livingplanet/lpr00/downloads/lpr_2000_full.pdf ); for the statistics compiled by M. Wackernagel for the report, see http:// www.rprogress.org; for an update, see http://www.panda.com