deforestation distribution and development

Global Environmental Change 11 (2001) 193}202

Deforestation, distribution and development

Gary Koop*, Lise ToleDepartment of Economics, Adam Smith Building, University of Glasgow, Glasgow, UK

Centre for Development Studies, Adam Smith Building, University of Glasgow, Glasgow, UK

Received 28 August 1998

Abstract

This paper investigates the role played by distributional factors in mediating the e!ects of growth and development on forestdepletion in tropical developing countries. A key "nding of the paper is that the distributional pro"le of a country signi"cantlydetermines whether economic development will have either a positive or a negative e!ect on the rate of forest loss. In countries wherelevels of inequality are high, development will tend to exacerbate deforestation rates while in countries where distributional pro"lesare more egalitarian, the negative e!ects of growth and development on forest cover will be ameliorated. 2001 Elsevier ScienceLtd. All rights reserved.

Keywords: Deforestation; Poverty; Inequality; Development strategies; Economic growth; Environmental Kuznets curve

1. Introduction

The promotion of economic growth has formed thecenterpiece of development planning and policymakingin the post-war period (IFAD, 1993; World Bank, 1990).In recent years, economists have complemented this fo-cus on growth with attention to basic needs, investmentin human capital formation, and environmental protec-tion. Although not included in formal measures of eco-nomic output or GDP, the quality of a country's naturaland human resource base is now recognized to be animportant determinant a!ecting economic growth andthe prospects for improved social welfare (Leonard, 1989;World Bank, 1990).

This shift in emphasis from strictly economic growthconsiderations toward issues of sustainability and welfarehas emerged in response to increasing recognition of theproblems now facing many developing countries. The"rst is the persistence of acute poverty. Despite the con-siderable economic progress made by developing socie-ties in the post-war period, the absolute numbers ofpeople living in poverty remains unacceptably high(Steer, 1992; World Bank, 1990; van der Gaag, 1991;

*Corresponding author. Tel.: 0141-330-6399.E-mail addresses: [email protected] (G. Koop), ltole@

socsci.gla.ac.uk (L. Tole).

IFAD, 1993). According to the 1990World Bank'sDevel-opment Report an estimated 1 billion people in the devel-oping world live in poverty, nearly half of them insoutheast Asia. International poverty pro"les suggestthat these people are likely to live in rural areas, to be infemale-headed households, to be farmers or agriculturalworkers, to be landless or near-landless, and to bea member of an ethnic minority (World Bank, 1990;IFAD, 1993; Fields, 1980). In addition, wide intra-coun-try disparities in income and wealth persist, not onlywithin regions and countries of the developing world butbetween countries as well (World Bank, 1990; Fields,1980; van der Gaag, 1991).

A second problem is the growing environmentalcrisis many developing countries now face. As Leonard(1989) notes, an increasing number of these poor aresituated in marginal, resource-poor and ecologically de-graded areas. Highly exploitative patterns of subsis-tence behavior in these areas lead to decreasing agricul-tural productivity due to environmental decay and re-source depletion. Similarly, declining resource stocksand environmental quality undermine economic growthand development, exacerbating already deep socialinequalities, and hindering the onset of the demographictransition.

A substantial economic literature exists devoted tounderstanding the role of poverty and inequality ineconomic growth. Similarly, a smaller and more recentbody of literature exists on the relationship between

0959-3780/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.PII: S 0 9 5 9 - 3 7 8 0 ( 0 0 ) 0 0 0 5 7 - 1

the environment and economic growth. (Sections 2 and 3below provide a very brief introduction to some of theseworks). However, most of the existing literature oneconomic development and the environment restrictsitself to qualitative assessments of these relationships ormerely speculates on the probable consequences for de-velopment arising from environmental degradation andexcessive population growth (Myers, 1993). Despite thefact that distributional issues have been observed to havean important impact on economic growth and growthitself to have important consequences for the environ-ment, there has been surprisingly little attempt to analyzethe links between redistribution and the environment.

This paper will attempt to explore the nature of therelationship between the environment, poverty, inequal-ity and economic growth. An important issue addressedis whether inequality and poverty have any signi"cantmediating e!ect on observed relationships betweengrowth and a key indicator of the decline in environ-mental quality: the rate of forest loss. In other words,does a country's distributional pro"le in#uence the rela-tionship between economic development and a country'sdeforestation rate?

After a brief discussion outlining the dimensions ofpoverty and inequality in developing countries, the paperdescribes recent empirical evidence on the relationshipbetween development and distribution and the environ-ment. Following this discussion, the subsequent sectionwill empirically analyze relationships using panel datafrom 48 developing countries. The estimated equationsform the basis of a number of broad generalizations or`stylized factsa about the linkages between these keyfactors. Variables chosen for analysis in the study arepredicted on the basis of the literature to in#uence de-forestation outcomes, and as such, re#ect key aspects ofthe development/deforestation/distribution nexus. Itshould be stressed at the outset that these relationshipsare not the only ones of importance in#uencing defores-tation rates. Neither should it be assumed that thevariables chosen for analysis are the only ones compris-ing this nexus. These relationships form part of alarger web of factors a!ecting forest loss, the rela-tions and interrelations of which are highly complex.Despite this complexity (which makes it impossible tomodel these relationships adequately), the study aims toprovide the basis for the development of more informa-tive models and country-level investigations of these rela-tionships in the future.

2. Patterns of poverty and inequality in developingcountries

Empirical research on the distributional aspects ofgrowth provides a number of implications for the presentstudy:

(i) Much evidence points to a positive role forgrowth in reducing poverty; the incomes of the poor,in other words, have for the most part tended to risewith growth (Fields, 1989; World Bank, 1990). Countriesthat have made substantial progress in reducing theincidence of poverty have also been those in whichincome per capita has risen sharply (Fields, 1980,1989). Although inter-temporal surveys of changesin distributions are lacking, this progress has beenmost pronounced in East Asia (recent "nancial crisesnotwithstanding). There is also some indication thatincome distributions have worsened in the past decadeand a half in Latin America and sub-Saharan Africa(World Bank, 1990; Cardoso and Helwege, 1992; Fields,1988b).

(ii) However, rapid economic growth is neither itselfa necessary nor a su$cient condition for the reduction ofpoverty since slowly growing countries have achievedconsiderable progress in this sphere (van der Gaag,1991; World Bank, 1990; Fields, 1980, 1989). Countriesthat have grown very slowly, such as Sri Lanka andmuch of sub-Saharan Africa, have only been able toraise incomes to any appreciable degree through strongredistributive e!orts by government (World Bank,1990). However, to paraphrase one expert, such policye!orts have been increasingly di$cult for governments toimplement in the absence of strong growth (Fields, 1980,1988a).

(iii) Economic growth does not necessarily lead toeither greater inequality or greater equality of distribu-tion, although countries with initially egalitarian incomedistributions tend to grow faster than those with in-egalitarian systems (Fields, 1989; Alesina and Perotti,1994; Alesina and Rodrik, 1994; Persson and Tabellini,1994; Birdsall et al., 1995). The initial level of incomedistribution appears to be a key factor in whether thepoor will bene"t from rapid growth. Further, growth hastended to reduce income inequality in the long run, butthe e!ects of growth on income inequality have been lessdramatic than on the alleviation of poverty (Fields, 1980,1989).

(iv) None of the above trends should obscure the factthat rapid population growth has increased the absolutenumbers of poor in many countries despite overall de-clines in the relative incidence of poverty (Fields, 1980,1988a, 1989).

(v) International growth and development compari-sons suggest that which distributional outcome occursdepends critically on the kind of public policy environ-ment that governments choose to foster. The pattern ofdevelopment appears to be as or more important, inother words, than the actual rate of growth in e!ectingpositive distributional changes (Fields, 1980, 1988a,1994; Ahluwalia, 1976). Summarizing the "ndings ofa large number of international case studies, the WorldBank (1990) concluded that countries that have exhibited

194 G. Koop, L. Tole / Global Environmental Change 11 (2001) 193}202

`balanced growtha * that is, growth where all sectorsand strata of society have bene"tted * are thosewhich have both grown rapidly and achieved sharpreductions in poverty in a short period of time.According to these studies, a key factor in the success ofsuch countries has been their capacity to create labor-intensive employment opportunities for the mass ofthe poor while simultaneously fostering an appropriateenvironment in which new capital investment andskills development continually complement and reinforceone another (Barro, 1991; Persson and Tabellini, 1994;van der Gaag, 1991; World Bank, 1990; Birdsall et al.,1995).

The available evidence also suggests that distributivemeasures are most e!ective in alleviating poverty whenthey encourage rapid expansion in non-agricultural sec-tors. Raising the purchasing power of rural regionsthrough land reform, agricultural extension, and im-provements in crop yields can raise e!ective demand inthe countryside and prompt the expansion of non-farmpursuits. However, there appear to be limits on the extentto which raising productivity in the agricultural sectoralone can create both the quantity and the qualityof employment opportunities required for sustainedeconomic growth (Fields, 1980).

3. Economic growth & patterns of environmental change

Recent empirical research suggests that the relation-ship between economic growth and environmentalchange exhibits a range of patterns, with many environ-mental indicators improving as incomes rise, someworsening, and others worsening and then improving(Steer, 1992). The growth of prosperity is also associatedwith improved living standards with respect to factorsthat bear directly on the quality of the environment (e.g.access to safe drinking water and sanitation) (Beckerman,1992).

Several studies, for example, have consistently founda strong correlation between per capita income levels andimproving environmental quality. This relationship is

The East Asian economies of Thailand, Korea, and Taiwan, Sin-gapore and Hong Kong exemplify these patterns well. These countriesare notable for their strong export-oriented labor-intensive growthstrategies and high levels of investment in human capital development.These policies have been credited with creating rapid growth in theeconomy as a whole, raising wages and employment prospects for themass of the population. In contrast, comparatively low levels of invest-ment in human resources, coupled with the persistence of an import-substituting, industrial-urban bias in development, have been blamed incountries like Brazil, India and the Philippines, for the marginal anduneven participation of the poor across sectors and their relatively lowskills level and productivity (Fields, 1980; Birdsall et al., 1995; WDR,1990).

commonly referred to in the literature as the environ-mental Kuznets curve (EKC). Concentrations of sulfurdioxide and smoke, for instance, have been found toincrease with per capita GDP at low levels of nationalincome, decrease with per capita GDP between US$4000andUS$8000 (1985 dollars), and then level o! (Grossmanand Krueger, 1995; Selden and Song, 1994).

Similarly, Hettige et al. (1992) "nd an inverted U-shaped pattern holds for GDP and the toxic intensity (i.e.toxic releases per unit of production) of manufacturingindustries. However, they also "nd that toxic emissionsfor this indicator depend heavily on the growth rate ofincome and the policy regime adopted by the country.For instance, fast-growing closed developing economiesbecame signi"cantly more toxic intensive in their manu-facturing sectors during the 1970s, a trend that acceler-ated in the 1980s. In contrast, fast-growing, openeconomies experienced essentially toxic-neutral struc-tural change in the 1970s and underwent a signi"cantshift to less pollution-intensive structures in the 1980s.This same pattern appears to hold for slow and medium-growing open economies. Thus, while dirty industriesexpanded faster than cleaner industries in the developingworld and faster than comparable industries in indus-trialized economies, not all developing countries exhib-ited growth in the toxic intensity of their emissions. Poor,closed but growing economies had the highest toxicintensity of their manufacturing mix, whereas the rapidlygrowing open higher income countries actually sawnegative growth in the toxic intensity of their emissions.

Birdsall and Wheeler (1992) "nd a similar patternholds in respect to intensity change in pollution by indus-try for Latin America. Results indicate pollution intensityto be lower (although still positive) at higher levels of percapita income between the 1960s and 1980s for all LatinAmerica countries. Growth rates in the toxic intensity ofemissions for closed economies that grew rapidly duringthis period were much higher than for fast-growing openeconomies.

Similarly, Wheeler and Martin (1992) "nd that theadoption of clean pulping technology is determined mostsigni"cantly by a country's trade regime, with more openeconomies adopting clean technologies at a faster ratethan closed economies. de Bruyn et al. (1998) use time-series data for four countries (USA, West Germany,Netherlands) on GDP and intensity of emissions ofSO

, NO

and CO

a nd "nd a positive relationship

(with the exception of SOemissions in the Netherlands)

between these indicators and economic growth. How-ever, they also note that this positive e!ect can becounteracted by `the intensity of use e!ecta, which be-comes more important as income grows: in other words,technological and structural factors may counteract thepositive impact of economic growth on emissions byreducing the intensity of pollution-generating materialsuse.

G. Koop, L. Tole / Global Environmental Change 11 (2001) 193}202 195

Kaufmann et al. (1998) "nd a U-shaped relationshipbetween income and SO

concentrations, one they at-

tribute to changes in energy use from fuels with a highsulfur content in the downward portion of the curve, toincreases in energy consumption in the upward portionof the curve that outweigh the e!ects of shifting towardlower-sulfur content fuels. Conversely, when the spatialintensity of economic activity is substituted for income,they "nd an inverted U-shaped relationship holds. Thelatter, they argue, may be a more important factor inobserved reductions in SO

concentrations than income

alone. However, they caution that the eventual positiveimpact of spatial intensity of economic activity on SO

concentrations is vulnerable to increases in population,which in turn can o!set any positive impacts associatedwith income gains.

Ambient levels of suspended particles and COemis-

sions have been found to increase monotonically withGDP (Holtz-Eakin and Selden, 1995; Sha"k, 1994. Un-ruh and Moomaw (1998) "nd this relationship to benon-linear and and punctuated by shocks that canchange the trajectory of the relationship. They note thatwhile CO

emissions have followed regular incremental

paths as GDP grows, this relationship has also beensubject to abrupt transitions, which appear to be corre-lated with exogenous factors that are independent ofincome growth (e.g. oil shocks).

Cropper and Gri$ths (1994) "nd an inverted U-shaped relationship between per capita income and ratesof deforestation for Africa and Latin America but not forAsia. In contrast, Koop and Tole (1999) "nd no suchevidence, and conclude that the variability in governmentpolicies and other factors a!ecting land use patternsdi!er too signi"cantly across countries for there to becommonalities in the structure of their forest qualityindicators. Eakins (1997) examines the relationship be-tween income and an aggregate indicator developed bythe OECD. This index not only includes CO

, SO

, and

NOemissions, energy intensity, sewage treatment, and

solid waste generation, but other key environmental indi-cators, such as imports of timber and cork, threatenedspecies of mammals and birds, kilometers of private road,number of protected areas, water use, and nitrate ferti-lizer application. The author "nds no evidence to supportthe presence of an EKC.

In contrast to these studies, Rothman (1998) argues fora consumption-based approach to measuring environ-mental indicators. Production-based measures (e.g. pol-lution intensity of use) which dominate current researchobscure: (a) the extent to which improvements in envir-onmental quality are the result of both the spatial dislo-cation of dirty industries and the unsustainable demandfor resources elsewhere; (b) the fact that most of theproductive activities in higher-income countries, whereenvironmental quality indicators have improved, areoriented primarily toward the satisfaction of consumer

demand; (c) the full monetary costs and lifestyle changesneeded to reverse environmental degradation when thefocus is largely on the production of pollutants per eco-nomic unit. Accordingly, the author provides evidencebased on a consumption indicator measuring ecologicalfootprints/appropriated carrying capacity for 52 nationsagainst GDP per capita. He "nds no evidence of thepresence of an EKC hypothesis for any functional form(e.g. linear, log}log and quadratic).

Similarly, Mason (1997) and Low and Yeats (1992)present evidence to suggest that with respect to (a) inparticular, global gas reductions in some countries havebeen achieved largely through the displacement of indus-tries to less-developed countries. With respect to (c) Suriand Chapman (1998) make the additional observationthat, the issue of the appropriate measure aside, studies"nding an EKC are misleading since the production ofincreasingly non-energy use intensive goods which can beobserved as incomes rise obscures the fact that many ofthe activities fueling economic growth require largeamounts of energy-intensive inputs. For this reason, theauthors argue that a better measure of environmentalstress is total commercial energy consumption. Usingthis measure, the authors "nd that the nature of the EKC(i.e. its inverted-U shape) can be accounted for by exam-ining energy use at its source: In industrializing countries,exports of manufactured goods account for the upwardsloping portion of the EKC, and in industrialized coun-tries, imports of manufactured goods, the downwardsloping portion. However, they caution that the esti-mated turning point for energy consumption lies welloutside the sample range, particularly when trade e!ectsare accounted for.

Finally, Torras and Boyce (1998) empirically analyzethe relationship between air and water quality indicatorsknown to improve with income. They "nd that in manycases it is not economic growth per se behind observedrelationships between income and the environment.Changes in the distribution of income, and in particular,increases in income which enhance the power of theaverage citizen to demand environmental protection, ap-pear to be key. The authors "nd that literacy, politicalrights and civil liberties measures are strong predictors ofpollution levels in low-income countries. Improvementsin these environmental indicators are associated with lesspollution at low income levels where the generally strongrelationship between income and environmental qualityas measured by these indicators weakens greatly. Incomeinequality as measured by the Gini ratio, was found tohave less consistent e!ects on environmental quality indi-cators. Greater income inequality was associated withmore pollution and less access to safe water, heavy par-ticles and dissolved oxygen, and appeared to increasewith greater income inequality in the low-income coun-tries. Results were weaker for higher-income countriesbut similar patterns were found, suggesting that power is


Note also that research suggesting the presence of a positive rela-tionship between some environmental indicators and economic growthhas tended to ignore issues of carrying capacity and ecological resil-iancy. Arrow et al. (1995) and Kaufmann and Cleveland (1995) observe,for example, that current studies have utilized data related to resourcestocks or to pollution emissions but do not consider possible synergisticenvironmental e!ects or consequences for the renewability of the re-source base arising from such degradation. Thus, it is quite conceivablethat the environment may actually be deteriorating as some environ-mental quality indicators are showing steady improvement. Long-runsustainability, in other words, depends fundamentally on the capacityof the environmental resource to renew itself and absorb and processwaste.Similarly, Stern et al. (1996) argue that the actual indicator used to

measure environmental quality may or may not reveal a state ofimprovement. For example, ambient levels of SO

, and CO

, NO

do

not tell us anything about the level of acid deposition agriculturalecosystems are facing.

a particularly important factor in#uencing environ-mental quality when income is very low. The authors alsoreport evidence to suggest that for somemeasures greaterinequality exacerbates observed relationships betweenincome and the environment.

4. Distribution, development and deforestation: sources oftestable hypotheses

Di!erences in the nature of the relationships betweeneconomic growth and indicators of environmental qual-ity may re#ect di!erences in policies toward certainindicators, the costs of environmental degradation andits abatement, and/or variations in the production struc-tures of countries at di!erent levels of economic develop-ment. As economic activities shift away from agricultureto industry and then again to services and high-techindustries, `the intensity of environmental services usedup in the production of a unit of GDP declinesa(Radetzki, 1992). Other processes accompanying thisshift can also a!ect the level of pollution and intensity ofnatural resource use: the extension of property rights toopen access reserves; the di!usion of new technologies;greater e$ciency in the use of resources prompted bygreater trade openness; the inclusion of environmentalprotection objectives in the political process; growingpublic valuation of and awareness of the environment;and so on (Radetzki, 1992; Grossman and Krueger, 1995;Birdsall and Wheeler, 1992; Behgin et al., 1994; Wheeler1996). With respect to the latter factor, it has been consis-tently noted in the literature (e.g. see Rothman, 1998) thatthe kind of environmental problems that appear to im-prove with income (generally, at the middle income level)tend to be those which can best be described as beingnon-separable in space and time. These tend to havelocalized impacts (e.g. sewage pollution) with observablehealth e!ects and do not require trans-national coopera-tion to address (e.g. global warming).

However, as Steer (1992) and Grossman and Krueger(1995) note, there is nothing inevitable about suchprocesses. Economic growth may or may not lead toimproved environmental quality and sustainability in theuse of natural resources. The kind of policy regime thatgovernments choose to foster* the land use and envir-onmental protection policies they implement and thesectors and strata they choose to target for development* will also play a critical role in which environmentaloutcome will dominate, irrespective of the technologicaland structural e!ects induced by economic growth (Steer,1992).

An important factor in#uencing a country's environ-ment-development trajectory is the extent to whicheconomic expansion alleviates poverty pressures on theenvironment and promotes environmental awareness.Whether this happens will depend greatly on whethergovernments pursue development policies that: enablethe poor to bene"t from economic growth, lead tobetter provision of government services, improve agricul-tural and industrial productivity, and encourage techno-logical innovation. As discussed in Section 1, it is quitepossible to raise aggregate income without appreciablychanging the living standards of the mass of the popula-tion. (However, as noted, achieving or sustainingprogress in redistribution is di$cult in the absence ofrobust growth.)

Thus, environmental degradation may worsen ifdistributive policies marginalize the poor despite the factthat the economy may be growing quickly (Steer, 1992;IFAD, 1993). In addition, environmental quality mayworsen despite other positive spin-o!s for the environ-ment arising from structural changes induced by eco-nomic growth in consumption, production and socialwelfare.

An important question is whether social welfare policymay have an important mediating e!ect on environ-mental outcomes of economic growth. In other words, towhat degree, if any, are the overall e!ects of economicexpansion on the environment in#uenced by a country'sdistributive policies? The next section will attempt toanswer this question by determining the role played bypoverty and inequality in mediating deforestation ratesin developing countries. It is hypothesized that whileeconomic growth may lead to deforestation as outputexpands, the scale of this exploitation may be exacer-bated by social conditions. For example, large numbersof impoverished people denied access to resources, parti-cularly land, can exacerbate already strong pressures onforests arising from the overall growth in the scale ofeconomic activity. High levels of poverty and inequalitymay also accelerate forest decline by hindering thetransition to demographic stability. Worsening povertyand inequality may also perpetuate reliance on the re-source base, thereby exacerbating resource shortages andlessening environmental quality.


A drawback of the data is that it relies heavily on individualgovernment surveys to supply information. Another drawback is thatthe data does not distinguish between forest types. Thus, even landwhich has lost its primary forest cover but has been replanted withmonocultural tree plantations is included in the de"nition of forestcover. So, too, are heavily degraded forests, which to all intents andpurposes have ceased to resemble viable ecosystems. Finally, studiesmeasuring the relationship between growth and forest cover depletionmay underestimate the contribution of development to the deforesta-tion process, since the amount of forest cover at any given period oftime is always a by-product of previous land-clearance activities (seePearce and Brown, 1994, for a discussion of the methodological issuessurrounding the measurement of tropical forests).

In this paper we adopt the FAO/UNEP 1994 de"nition of tropical.Tropical forests encompass a wide variety of ecosystems (e.g. evergreen,semi-evergreen, semi-deciduous, premontane, woodlands, alpine, sa-vannah, and shrub) situated in countries receiving between 200 to20,000 mm of rainfall annually (see WRI, 1994}1995).

A poor welfare pro"le may also a!ect forest coverin other, less direct ways, by undermining any positivebene"ts accruing to the environment arising fromeconomic or institutional changes elsewhere. Similarly,by contributing to political instability, poverty andinequality can undermine the security of propertyrights associated with sustainable resource use * andultimately* economic growth itself (Keefer and Knack,1995).

In short, a country in which growth is not accom-panied by the alleviation of poverty or inequality wouldbe expected to have higher deforestation rates than onecharacterized by more balanced and equitable growththat improves the lives of the mass of the poor. Deforesta-tion, in other words, would be expected to be higher incountries like Brazil and the Philippines, where growthhas not been accompanied by the eradication of povertyand/or inequality to any signi"cant degree. Countries inwhich growth has either been slow or has stagnated butin which governments have made deliberate e!orts todistribute public resources (e.g. Sri Lanka, Tanzania)should have lower deforestation rates than those coun-tries where poor growth performance has also beenmatched by a poor distributive pro"le (e.g. India,Haiti). Those countries that have managed to grow fairlyquickly (e.g. Jamaica, Bangladesh) but have failed never-theless to achieve appreciable reductions in levels ofpoverty or inequality, would be expected to have higherrates of deforestation than those countries where growthhas been more balanced (e.g. Malaysia, Singapore,Korea).

5. Empirical analysis

This section empirically investigates the relative im-pact of distributional factors on the relationship betweeneconomic growth, development and deforestation. Webegin with a standard panel data model for deforestation(e.g. Cropper and Gri$ths, 1994; Koop and Tole,1999). Deforestation is de"ned in this study as the per-centage annual decrease in forest area. Data on forestcover loss for 48 tropical developing countries are de-rived from the FAO Production Yearbook for variousyears between the period, 1961}1992. Although theFAO Production data is not the only existing source ofglobal forest data, it is the most comprehensive,covering more countries and spanning a longer periodthan any other source (Allen and Barnes, 1985). Thisdeforestation measure is based on a de"nition of forestcover that includes all woody vegetation. More speci"-cally, forests and woodland refer to all land under naturalor planted stands of trees, regardless of productivity, inaddition to land that has been cleared of forests but willbe reforested in the near future. This broad de"nitionlargely avoids problems associated with inconsistencies

in de"nition arising from the use of more restrictive forestde"nitions.

Although all types of forest cover are represented bythe data, we include in this study only developing coun-tries whose predominant ecosystems are tropical. Tropi-cal forest ecosystems dominate the developing world. It istheir progressive destruction more than that of any otherecosystem, which is causing the greatest internationalalarm in view of the rapid rate at which they are beingdepleted, their unique biodiversity, and the importantecological and material contributions they make tohuman welfare.

The level of economic development and rate ofeconomic growth are measured by GDP per capita andthe % change in GDP per capita. Both are from thecommonly used Penn World Table, which adjusts toa common set of international prices. Two demographicmeasures * population density and change in popula-tion * important for their impacts on the consumptionand production processes that fuel economic growth,poverty and environmental degradation * are also in-cluded in the analysis. Population density is measured asthe number of people per hectare of land area, andpopulation growth, as the percentage annual increase intotal population. Both indicators are derived from FAOProductionData and the PennWord Table, respectively.

The data above (i.e. deforestation, GDP, GDP growth,population growth) are panel data (i.e. are both acrosscountries and over time). Since we also wish to includedistributional variables, for which data are not usuallyavailable in time-series form (income or land inequalitymeasures are typically based on government surveyswhich are carried out at most once a decade), we cannotuse a standard panel data framework, but rather choosea model of the form

y"

#

x#

dx


Note also that statistical tests indicate that the use of a "xed e!ectsspeci"cation is required (i.e. the intercept of the regression relationship,, varies across countries). However, such a regression speci"cation

excludes the use of das a separate explanatory variable, which can only

be included by allowing it to interact with the x, s. (see Greene, 1997,

chapter 14, for a discussion of estimation and testing with panel datamodels). Formally, we use "xed e!ects estimation for unbalancedpanels.

A mathematical way of conceptualizing these results is to note thatall statistically signi"cant coe$cients which involve any of the distribu-tion variables are positive.

where yis deforestation in country i at time t, x

is

explanatory variable j in country i at time t (i.e. j"1indicates GDP per capita, j"2 GDP growth, j"3population density and j"4 population growth) anddis a distributional measure for country i. Note that

dhas only an i subscript, indicating that we have only

one data point per country. We consider for our distribu-tional variables (i.e. d

) one measure of income inequality,

the Gini coe$cient (GINI) of income, and two of landinequality, the Gini coe$cient for land (LANDGINI)and the percentage of land held by the top 20% oflandowners (LANDT20), respectively. This last measureis included since large landholdings are typically moreaccurately measured than are smallholdings. All thesevariables are taken from IFAD (1993). Higher values forthese variables indicate a greater degree of inequality.A list of countries, years, and data de"nitions are given inthe appendices. Summary statistics for all variables aregiven in Table 3 in Appendix B.

In practice, distributional issues likely re#ect underly-ing economic/social/political structures that change onlyslowly over time so that the information loss involved innot having time-series data for distributional variables isprobably quite small.

Note also that the marginal e!ects (i.e. the e!ect ondeforestation of a one unit change in an explanatoryvariable) of explanatory variable j is

#

d. This

implies, for instance, that the e!ect of increasing GDPper capita on deforestation can vary across countriesdepending on the value of the distributional variable.In short, this model explicitly addresses the questionof interest in this study, namely, `How do distributionalissues a!ect the deforestation/development relation-ship?a.

Table 4 in Appendix B provides results from "xede!ects estimation of the model with each of the threedi!erent distributional measures. To aid in interpreta-tion, we subtract the mean from each of the distributionalvariables so that a country, for instance, with an averagelevel of income inequality, for instance, will haveGINI"0.

One implication of this approach is that the "rst fourrows of estimates (i.e. those labelled GDP, GDP growth,Pop. density and Pop. Growth) can be interpreted asmeasuring the e!ects of these variables on deforestationfor a country with an average amount of inequality.Looking at these four rows, we "nd a reasonably consis-tent picture for all three of our distributional measures.

That is, GDP and population growth have little e!ect ondeforestation for countries with an average amount ofinequality. However, GDP growth is negatively andpopulation density positively, associated with deforesta-tion for these average countries. This latter result is notsurprising: higher population density is associated withmore deforestation. The former result is perhaps lessexpected: faster GDP growth is associated with lessdeforestation.

The numbers in the bottom half of Table 4 are themost important ones in that they illustrate how distribu-tional considerations a!ect deforestation/developmentor deforestation/demographic relationships. Althoughthere are some con#icts across distributional measures(i.e. the income and land inequality measures yield some-what di!erent results), the general pattern of the resultswhich are statistically signi"cant is consistent with thehypothesis of this paper. That is, more inequality causesthe deforestation/development relationship to worsen.

If we look, for instance, at the results for GDP for theland inequality measures, we "nd that for countries withaverage levels of inequality, GDP seems to have no e!ecton deforestation (i.e. the marginal e!ect,

#

d, re-

duces to since d

"0 for these average countries; and

the coe$cient is statistically insigni"cant). However,

for countries with above average levels of inequalityd'0 and the marginal e!ect for GDP now becomes

signi"cantly positive. In other words, if we considercountries with a high degree of land inequality, GDP anddeforestation are positively related (i.e. more GDP isassociated with more deforestation). In contrast, in thecase of those countries characterized by greater equalityin their distribution of land, d

is negative and hence the

marginal e!ect of GDP on deforestation is signi"cantlynegative. In other words, in more equal countries, moreGDP is associated with less deforestation.

A similar, but statistically weaker, story can be told forthe income inequality measure. That is, higher levels ofinequality tend to worsen the deforestation/GDP growthand deforestation/population growth relationships.Conversely, lower levels of inequality improve theserelationships.

6. Conclusion

This paper has sought to establish some key stylizedfacts about the role of distribution in mediating thee!ects of development and growth on forest depletion intropical developing countries. A key "nding of the paperis that distributional pro"le is a signi"cant determinant


Table 1

Region Number(millions)

Headcountindex (%)

Life expectancy(years)

Under "vemortality(per '000)

Poverty gap Net primaryenrollmentrate

Sub-Saharan Africa 180 47 50 196 11 56East Asia 280 20 67 96 1 96South Asia 520 51 56 172 10 74Middle East & North Africa 60 31 61 148 2 75Latin America & Caribbean 70 19 66 75 1 92TOTAL 1110 25 60 98 2.5 78.6

Adopted From WDR (1990, Table 2.1).De"ned as the aggregate income shortfall of the poor as a % of aggregate consumption.

Table 2List of countries with years of availability

Bangladesh 61}92 Grenada 84}90 Niger 61}89Belize 80}92 Guatemala 61}92 Nigeria 61}92Bolivia 61}92 Guyana 61}90 Pakistan 61}92Botswana 61}89 Haiti 61}89 Panama 61}92Brazil 61}92 Honduras 61}92 Paraguay 61}92Cameroon 61}92 India 61}92 Peru 61}92Cape Verde 61}92 Indonesia 61}92 Philippines 61}92Colombia 61}92 Jamaica 61}91 Rwanda 61}92Costa Rica 61}92 Kenya 61}92 Sierra Leone 61}92Cote d'Ivoire 61}92 Liberia 61}86 Sri Lanka 61}92Dominican Rep. 61}92 Malawi 61}92 Suriname 61}89Ecuador 61}92 Malaysia 61}92 Thailand 61}92El Salvador 61}92 Mauritania 61}92 Togo 61}92Ethiopia 61}86 Mexico 61}92 Trinidad & Tob. 61}91Gambia 61}90 Myanmar 61}89 Uganda 61}92Ghana 61}92 Nepal 61}86 Zambia 61}91

Table 3

Mean SD

% decrease in forest cover 0.01 0.02GDP per capita ($000) 1.73 1.70% increase in GDP 0.01 0.07People per hectare of land 0.64 1.02% increase in population 0.03 0.02% land held by top 20% 0.64 0.17Gini coe$cient for land 0.53 0.15Gini coe$cient for income 0.46 0.84

of whether economic development will have either a pos-itive or a negative e!ect on the rate of forest depletion. Itappears that in countries where levels of inequality arehigh, development will tend to exacerbate deforestationrates. Conversely, in countries where distributionalpro"les are above the average for egalitarianism,distributional factors will tend to have a positive impact,ameliorating the negative e!ects of growth and develop-ment outcomes on forest cover. The more egalitarianthe country, in other words, the less deforestation. Therole of distributional factors appears to work indepen-dently of the strong positive e!ect observed for popula-tion pressures on deforestation rates as well as thenegative (albeit much weaker) e!ect of fast growth ondeforestation (i.e. fast growth rates are associated withless forest loss).

We can now supplement the study's "ndings withthose of the economic literature on poverty, inequalityand development discussed in Section 2. According tothis literature, more egalitarian countries also tend togrow faster. This study has demonstrated that themore egalitarian a country is the less deforestationit is likely to have; in addition, the faster a countrygrows the less deforestation. However, irrespective ofwhether countries exhibit fast or slow growth, distribu-tional factors will play a signi"cant role in determiningthe severity of the impacts of economic development onthe environment.

This strong but varying role for distribution in medi-ating development/deforestation outcomes across di!er-ent income levels and rates of economic growth may alsoexplain the contradictory "ndings of the environmentalKuznets literature. As noted, some environmental indi-cators appear to worsen and then improve with eco-nomic growth; others appear to worsen or to improvesteadily; others show no discernible pattern. It could bethat, among other factors, variations in environmentaltrajectories may relate to di!erences in distributionalfactors across countries, with some environmental qual-ity indicators, in other words, being more susceptible todistributional in#uences than others. Forests, unlike

some indicators (e.g. air quality), are closely tied to thedistribution of land. Further, being a signi"cant factor ofproduction in rural areas, land is also a major source ofincome. If, as many studies suggest, certain air and waterquality indicators may improve with income this mayre#ect the political reality that government interventionin these areas is less contentious than is distributing landor income. Indeed, as the history of developing societieshas repeatedly shown, such reforms have often proved tobe as di$cult to implement as they are rare.


Table 4

Measure of distribution

Land owned by top 20% Land gini Income gini

GDP !0.001 (0.001) !0.001 (0.001) !0.001 (0.001)GDP growth !0.014 (0.008) !0.014 (0.008) !0.010 (0.008)Pop. density 0.004 (0.002) 0.005 (0.002) 0.007 (0.003)Pop. growth 0.061 (0.057) 0.065 (0.061) !0.003 (0.061)GDPdistribution 0.023 (0.009) 0.026 (0.012) 0.012 (0.017)GDP growthdistribution !0.032 (0.045) !0.046 (0.056) 0.110 (0.087)Pop. densitydensity !0.006 (0.023) 0.003 (0.030) 0.031 (0.033)Pop. growthdistribution !0.071 (0.355) !0.059 (0.562) 0.547 (0.447)

Errors in parentheses.Indicates signi"cance at the 10% level.Indicates signi"cance at the 20% level.Indicates signi"cance at the 5% level.

Appendix A

The regional poverty pro"les are given in Table 1.

Appendix B. Data sources

The deforestation data is obtained from the FAO'sAgrostat-PC land use and input diskette and uses thestudy's forest and woodland variable. This source is alsoused to obtain total land area for calculating populationdensity. Economic variables are from the Penn WorldTable (version 5-6), a commonly used data source whichcorrects GDP "gures for cross-country price di!erencesto ensure that numbers are comparable across countries.Data for the study's distributional variables, income,land inequality and welfare are from IFAD (1993) (seeTable 2).

Appendix C. Empirical results

Summary statistics for data are shown in Table 3 andthe regression results in Table 4.

References

Ahluwalia, M., 1976. Inequality, poverty and development. Journal ofDevelopment Economics 3, 307}342.

Alesina, A., Rodrik, D., 1994. Distributive politics and economicgrowth. Quarterly Journal of Economics 109, 465}490.

Alesina, A., Perotti, R., 1994. The political economy of growth: a criticalsurvey of the recent literature. TheWorld Bank Economic Review 8,351}371.

Allen, J.C., Barnes, D.F., 1985. The causes of deforestation in develop-ing countries. Annals of the Association of American Geographers75, 163}184.

Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling,C.S., Jansson, B.-O, Levin, S., MaK ler, K.G., Perrings, C., Pimental,D., 1995. Economic growth, carrying capacity and the environment.Science 268, 520}521.

Barro, R.J., 1991. Economic growth in a cross-section of countries.Quarterly Journal of Economics CVI, 408}443.

Beckerman, W., 1992. Economic growth and the environment: whosegrowth? whose environment?. World Development 20, 481}496.

Behgin, J., Roland-Holst, D., van der Mensbrugghe, D., 1994. A surveyof the trade and environment nexus: global dimensions. OECDWorking Paper, Paris.

Birdsall, N., Wheeler, D., 1992. Trade policy and industrial pollution inLatin America: where are the pollution havens? In: Low, P. (Ed.),International Trade and the Environment. World Bank DiscussionPaper, vol. 159. The World Bank, Washington, DC.

Birdsall, N., Ross, D., Sabot, R., 1995. Inequality and growth recon-sidered: lessons from East Asia. The World Bank Economic Review9, 477}508.

Cardoso, E., Helwege, A., 1992. Below the line: poverty in Latin Amer-ica. World Development 20, 19}37.

Cropper, Gri$ths, 1994. The interaction of population growth andenvironmental quality. American Economic Review 84, 250}254.

de Bruyn, S.M., van den Bergh, J.C.J.M., Opschoor, J.B., 1998.Economic growth and emissions: reconsidering the empirical basisof environmental Kuznets curves. Ecological Economics 25,161}175.

Eakins, P., 1997. The Kuznets curve for the environment and economicgrowth: examining the evidence. Environmental Planning A 29,805}830.

Fields, G.S., 1980. Poverty, Inequality, and Development. New York,Cambridge University Press.

Fields, G.S., 1988a. Income distribution and economic growth. In:Ranis, G., Shultze, T.P. (Eds.), The State of Development Econ-omics. Oxford, Blackwell.

Fields, G., 1988b. Employment and economic growth in Costa Rica.World Development 16, 1493}1509.

Fields, G.S., 1989. Changes in poverty and inequality in developingcountries. World Bank Researcher 4, 167}185.

Fields, G.S., 1994. The Kuznets curve: a Good Idea but....Mimeo.Cornell University, Ithaca, NY.

Greene,W., 1997. EconometricMethods, (Third International Edition).Prentice-Hall International Edition, London.

Grossman, G.M., Krueger, A.B., 1995. Economic growth and the envi-ronment. Quarterly Journal of Economics 110, 3}353.

Hettige, H., Lucas, R.E.B., Wheeler, D., 1992. The toxic intensity ofindustrial production: global patterns, trends and trade policy.American Economic Review Papers and Proceedings 82, 478}481.

Holtz-Eakin, D., Selden, T.M., 1995. Stoking the "res? COemissions

and economic growth. Journal of Public Economics 57, 85}101.IFAD, 1993. The State of Rural World Poverty. FAO, Rome.


Kaufmann, R.K., Cleveland, C.J., 1995. Measuring sustainability:needed * An interdisciplinary approach to an interdisciplinaryconcept. Journal of Ecological Economics 15, 109}112.

Kaufmann, R.K., Davidsdottir, B., Garnham, S., Pauly, P., 1998. Thedeterminants of atmospheric SO

concentrations: reconsidering the

environmental Kuznets curve. Ecological Economics 25, 209}220.Keefer, P., Knack, S., 1995. Property rights, inequality, and growth.

University of Maryland. IRIS Working Paper Series 53, 1}34.Koop, G., Tole, L., 1999. Is there an environmental Kuznets curve

for deforestation? Journal of Development Economics 58, 231}244.Leonard, H.G., 1989. Environment and the Poor: Development Strat-

egies for a Common Agenda. U.S. Third World Policy Perspectives11. Washington, DC, Overseas Development Council.

Low, P., Yeats, A., 1992. Do dirty industries migrate? In: Low, P. (Ed.),International Trade & the Environment. World Bank DiscussionPapers, 159. The World Bank, Washington, DC.

Mason,M., 1997. A look behind trend data in industrialization: the roleof transnational corporations and environmental impacts. GlobalEnvironmental Change 7, 113}127.

Myers, N., 1993. Population, environment, and development. Environ-mental Conservation 20, 205}216.

Pearce, D., Brown, K. (Ed.), 1994. The Causes of Tropical Deforesta-tion: The Economic and Statistical Analysis of Factors Giving Riseto the Loss of the Tropical Forests. UBC Press, Vancouver.

Persson, T., Tabellini, G., 1994. Is inequality harmful for growth?American Economic Review 84, 600}621.

Radetzki, M., 1992. Economic growth and environment. In: Low, P.(Ed.), International Trade & the Environment. World BankDiscussion Papers, 159. The World Bank, Washington, DC.

Rothman, D.S., 1998. Environmental Kuznets curves * real progressor passing the buck? A case for consumption-based approaches.Ecological Economics 25, 177}194.

Selden, T.M., Song, D., 1994. Environmental quality and development:is there a Kuznets curve for air pollution emissions. Journal ofEnvironmental Economics and Management 27, 147}162.

Sha"k, N., 1994. Economic development and environmental quality: aneconometric analysis. Oxford Economic Papers 46, 757}773.

Steer, A., 1992. The environment for development. Finance & Develop-ment 29, 18}21.

Stern, D.I., Common, M.S., Barbier, E.B., 1996. Economic growthand environmental degradation: the environmental Kuznetscurve and sustainable development. World Development 24,1151}1160.

Suri, V., Chapman, D., 1998. Economic growth, trade and energy:implications for the environmental Kuznets curve. Ecological Econ-omics 25, 195}208.

Torras, M., Boyce, J., 1998. Income, inequality, and pollution: a reas-sessment of the environmental Kuznets curve. Ecological Econ-omics 25, 147}160.

Unruh, G.C., Moomaw,W.R., 1998. An alternative analysis of apparentECK-type transitions. Ecological Economics 25, 221}229.

van der Gaag, J., 1991. Poverty in the developing world. EuropeanEconomic Review 35, 343}349.

Wheeler, D., Martin, P., 1992. Prices, policies and the internationaldi!usion of clean technology: the case of wood pulp production,In: Low, P. (Ed.), International Trade and the Environment.World Bank Discussion Paper, 159. The World Bank, Washington,DC.

Wheeler, D., 1996. My brother's keeper: new evidence on pollutionhavens and environmental injustice. Mimeo, World Bank.

World Bank, 1990. World Development Report. Oxford UniversityPress, Oxford, New York.

World Resources Institute (WRI), 1994}1995. World Resources.Oxford University Press, Oxford, New York.


deforestation distribution and development

Documents