O

Auv

Aa

b

c

a

ARRA

KCLS

1

as(Medatctpte

i

1h

Ecological Indicators 34 (2013) 103– 112

Contents lists available at SciVerse ScienceDirect

Ecological Indicators

jou rn al hom epage: www.elsev ier .com/ locate /eco l ind

riginal article

n exploratory spatial analysis of illegal coca cultivation in Colombiasing local indicators of spatial association and socioecologicalariables

lexander Rincón Ruiza,∗, Unai Pascualb, Milton Romeroc

ICTA, Autonomous University of Barcelona, SpainDepartment of Land Economy, University of Cambridge, UKBasque Centre for Climate Change (BC3) and IKERBASQUE, Basque Foundation for Science, Bilbao, Spain

r t i c l e i n f o

rticle history:eceived 21 July 2012eceived in revised form 19 March 2013ccepted 2 April 2013

eywords:oca cropsocal indicators

a b s t r a c t

The cultivation of coca (Erythroxylon coca) by smallholders represents the first step in the world’s largestillegal agribusiness: cocaine. The literature on illicit crops generally considers the global and nationalfactors that influence its expansion but little attention is paid to the local conditions where cultivationtakes place. This study analyzes the economic, environmental, institutional and social factors associatedwith the existence of coca crops within municipalities in two different years (2001 and 2008). We dothis from a regional perspective, by taking into account the particular factors that characterize the areasassociated with coca municipalities and estimating the impacts of coca cultivation on natural ecosystems.

patial analysis For the proposed analysis we use Exploratory Spatial Data Analysis (ESDA) to visualize and describe thelocal spatial distributions and clusters, implementing the Local Indicators of Spatial Association (LISA)for the year 2001 and 2008. Our results show that no individual factor is exclusively associated with thecoca areas; in fact, a group of common factors characterizes these regions. The analysis indicates that theexpansion of coca cultivation has produced a process of deforestation mainly affecting humid tropicalforested ecosystems.

© 2013 Elsevier Ltd. All rights reserved.

. Introduction

Tropical deforestation is a complex social-ecological problemnd has been explored from many disciplinary perspectives empha-izing economic, social, institutional and environmental factorsAngelsen, 1999; Deininger and Minten, 2002; Evans et al., 2008;

artinez-Alier et al., 2010; Norgaard, 1994; Scott, 1998; Solomont al., 2005; UNODC, 2010). While most research on tropicaleforestation has focused on agricultural frontier expansion andssociated activities such as timber extraction and land conversiono monocultures, little attention has been paid to the deforestationaused by cultivation of illegal crops such as coca for the produc-ion of cocaine. This is surprising as the literature has identified thisroduction as a driver of deforestation in tropical regions and par-icularly in countries like Colombia (Armenteras et al., 2006; Etter

t al., 2006; Farrell and Thorne, 2005).

Although coca cultivation is not the main cause of deforestationn Colombia, it is affecting ecosystems through its negative impact

∗ Corresponding author.E-mail address: alexander.risvid@gmail.com (A. Rincón Ruiz).

470-160X/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.ecolind.2013.04.008

on biodiversity, especially in forests with suitable environmentalconditions for its cultivation (Álvarez, 2007; Dávalos et al., 2011).The displacement of illicit crops to other areas has created a defor-estation problem in Colombia, as discussed by (Dávalos et al., 2011),using land use coverage maps. There are various underlying factorsregarding the location of coca-growing areas in Colombia. Somesuch factors that have been mentioned in the literature include landtenure insecurity (Fajardo, 2002), the existence of political insta-bility and armed conflicts (Díaz and Sánchez, 2004; Garcés, 2005;Vargas, 2005), low social and institutional development (Thoumi,2005a,b,c), high levels of rural poverty (Dion and Russler, 2008),violence and inaccessibility (Álvarez, 2001; Álvarez, 2003; Rinconet al., 2006), forced displacement (CODHES, 2009; Ibánez and Vélez,2008), income and other economic aspects (Angrist and Kugler,2008; Ibanez and Carlsson, 2009; Ibanez, 2010; Rocha, 1997, 2000),and environmental aspects (Álvarez, 2003; Dávalos et al., 2011).

Most of the socio-economic and institutional information avail-able for analysis that includes the factors mentioned above is

found at the level of the administrative political units in the coun-try: national, departmental and municipal. However, no study hasincluded an in-depth analysis using disaggregated spatial level data(with data on a municipal level for the whole country) to better

1 ical In

ueglaptcaSlac

f(ttCpaaam

si4fitpveeupet

2

ifiT(cPt12ct

w1zMpirldg2

we have made an appropriate declaration regarding its limitationsin Section 5.

04 A. Rincón Ruiz et al. / Ecolog

nderstand the regional factors associated with the existence andxpansion of coca crops in Colombia. Most statistical analyses arelobal (and do not take into account regional and local particu-arities and realities); for example, Dion and Russler (2008) havettempted to explain the permanence of coca crops in Colombia,roviding evidence that poverty has a non-linear effect. However,he authors were not able to decipher the local particularities ofoca crop areas since their analysis was based on the nationalnd departmental scale rather than a municipal scale. Díaz andánchez (2004) used a smaller scale (municipal level) but only ana-yzed the relationship between land use and armed conflict, which,lthough important, only offers a partial view of the illegal coca-rop-growing problem in Colombia.

This paper applies a spatial analysis at the municipality levelor the period from 2001 to 2008 with the following objectives:i) determine the relationship between illicit crops and the fac-ors generally mentioned in literature and by experts, by assessinghe presence of coca cultivation for illegal use in 2001 and 2008 inolombia; (ii) identify the factors that may be associated with theresence and expansion of coca cultivation between 2001 and 2008t a local scale and identify the differences with global analysis. Tochieve these objectives we carried out an exploratory spatial datanalysis (ESDA) using information about social, economic, environ-ental and institutional factors at the municipal level.The paper is organized as follows: Section 2 describes the case

tudy area; Section 3 describes the methodological issues relat-ng to data and methods used during the research; and Section

presents the results obtained, divided into two sections. Therst shows a change in the geographical distribution of coca cul-ivation and estimates the relative proportion of land that wasreviously forest and woody ecosystems being used for coca culti-ation in 2001 and in 2008. In the second, we analyze the economic,nvironmental, institutional and social factors associated with thexistence of coca crops for two periods of time (2001 and 2008)sing ESDA. Section 5 discusses the results and identifies someolicy implications regarding the efforts needed to inhibit thexpansion of coca cultivation in Colombia. Finally, Section 6 con-ains the main conclusions.

. Study area

Colombia is located in the northwestern corner of South Amer-ca and bordered by the Atlantic and Pacific oceans as well as byve other countries (Venezuela, Brazil, Peru, Ecuador and Panama).he climate is tropical: hot and humid with two rainy seasonsMarch–May and September–December). Colombia consists of fiveontinental natural regions: the Andean (AN), Caribbean (CA),acific (PA), Orinoquian (OR) and Amazon (AM) regions. The coun-ry is divided into 32 departments and 1101 municipalities (Map). According to our estimates, based on the coca censuses (UNODC,006, 2008a,b, 2009), in 23 departments and 274 municipalities,oca was present for at least one year during the period from 2001o 2008.

Colombia produces more than 50% of the coca used in world-ide cocaine production, and has an average planting density of

10,000 plants per hectare (UNODC, 2006, 2008b, 2009). The Ama-on is the region with the highest coca yield, principally in theeta and Guaviare departments, which average 9900 kg per ha

er year. In 2001 most of the coca production was concentratedn the Amazon region. The lowest coca yield occurs in the Pacificegion (2600 kg per ha per year), but despite this comparatively

ow amount, a large part of the country’s coca cultivation has beenisplaced to the Pacific region in recent years, partly due to theovernment’s eradication policies in the Amazon region (UNODC,006, 2008b, 2009).

dicators 34 (2013) 103– 112

3. Data and methods

3.1. Data for the identification of factors associated with cocacultivation

The identification of possible factors associated with coca cropsareas (CCAs) included in the analysis was carried out in three ways.Firstly, the relevant literature was reviewed. Secondly, expertswere consulted regarding social, economic, environmental andinstitutional issues related to illicit crops in Colombia, mainlyresearchers and staff from the Colombian Integrated Illicit CropsMonitoring System (SIMCI, from the Spanish acronym); the Colom-bian National Office of Narcotics (DNE, from the Spanish acronym);the Ministry of Defense; and various NGOs including the Transna-tional Institute (TNI), the Arcoiris Foundation and Acción Andina;as well as researchers from Colombian Universities. Thirdly, field-work was carried out in two municipalities, Rosario and Leiva(located in Narino, which experienced the biggest increase incoca cultivation between 2001 and 2008). Interviews with pub-lic institutions (health department, local government and nationalpolice), and civil society (local population and indigenous com-munities such as the Awá community1) were also carried out.These varied sources of information were instrumental in creat-ing a diagnosis of the regional factors associated with coca crops inColombia. This diagnosis identified five key factors: (1) poverty,(2) weakness and low presence of the state, (3) violence andarmed conflict, (4) inaccessibility and (5) favorable biophysicalconditions.

After identifying the possible factors behind the cultivation ofcoca, spatial and alphanumeric information was collected, takingtwo key criteria into account. The information was required to be:(1) available at the municipal level for the whole country and (2)collected from an official source. To organize the data, informa-tion was classified according to social, economic, institutional andenvironmental factors. Table 1 shows a set of selected variableswith information at the municipal level covering the whole country.Some of the key variables identified could not be included becauseno information was available at the municipal level for the wholecountry; this was the case for the following variables: income distri-bution, land tenure, social capital and state weakness. The numberof violent acts and murders by illegal armed groups was used asa proxy for the weakness of the state. The spatial and alphanu-meric information on coca crops was obtained from the IntegratedSystem of Monitoring of Illicit crops of the United Nations on Drugsand Crime (SIMCI, from the Spanish acronym). We selected the per-centage of coca area per municipality as variable “X” and the othervariables as “Y” in order to analyze their spatial association (in ourresearch the names “X” and “Y” do not imply causality).

The best information available for important spatial andalphanumeric variables on coca crops at the municipality level inColombia was identified for the years 2001 and 2008. In cases inwhich complete information was not available for a specific year,the existing information from the closest year was used. Addition-ally, in cases where data were available for one year only, thisinformation was used as a reference for both years. Here we intro-duce the assumption that although the variables may have changedover time, they have maintained their spatial distribution betweenmunicipalities throughout the examined period. We recognize thatthis may introduce inaccuracies in our analysis and for this reason

1 The Awá Community is an indigenous community that inhabits the southwestof Colombia, in the Departments of Narino and Putumayo (also found in Ecuador).

A.

Rincón

Ruiz

et al.

/ Ecological

Indicators 34 (2013) 103– 112

105

Table 1Selected variables at the municipal level.

Tipo devariable

Sigla Category Variable name Year Source of data Unit Additional explanation

Var X PCA Analysis variable Percentage of themunicipal territory incoca area

2001–2008 Coca maps (shape) – Integrated Illicit CropsMonitoring System (SIMCI: Spanish acronymfor “Sistema Integrado de Cultivos Ilicitos”)

Percentage Percentage of hectares of cultivated coca incomparison to the total area of themunicipality. Scale 1:10.000

Var Y GDPC Economic Growth DomesticProduct per capita

2002–2007 National Administrative Department ofStatistics – DANE, estimation by the authors

Billions Annual gross domestic product per person

PRD Economic Primary roads density 2005 Road map (shape) – Geographic InstituteAgustín Codazzi – Estimation by the authors

Mts/roads/ha Meters of primary roads per hectare. Scale:1:500.000

FDP Social Forced displacement ofpopulation rate -expulsion

2001–2008 Presidency of the Republic of Colombia –Presidential Agency for Social Action andInternational Cooperation

Number of displacedpeople per 1000inhabitants

Number of forced displaced people byviolende and conflict. This information istaken from the National System of Integralattention of Displaced People (“SistemaNacional de Atención Integral a laPoblación Desplazada”)

RUBN Social Rural Unsatisfied BasicNeeds Index

2005 National Administrative Department ofStatistics - (DANE Spanish acronym for“Departamento Administrativo Nacional deEstadística”)

Index from 0 to 100(From ‘no basic needsatisfied’ to completelysatisfied, respectively)

The index is determined through 5indicators to define if the basic needs of therural population are covered. The selectedindicators are: Inadequate housing,Housing with critical overcrowding,housing with inadequate services, housingwith high economic dependence, housingwith children at school-age but that arenot attenting school.

RDP Social Rural densitypopulation

2005 National Administrative Department ofStatistics - (DANE Spanish acronym for“Departamento Administrativo Nacional deEstadística”)

Number of persons Number of persons living in the rural zonesof each municipality

IMD Institutional Index of municipaldevelopment* (2001 –2008)

2001–2008 National Planning Department of Colombia(DNP – Spanish acronymun for “DepartamentoNacional de Planeación”) – Dirección deDesarrollo Territorial Sostenible DDTS

Index from 0 to 100(where 0 means lowmunicipaldevelopment)

Synthetically measuring the performanceof municipalities in social and financialvariables. The variables taken are: % ofpopulation in header,% of households withwater supply,% of households withsewage,% households with energyservices,% of people without NBI 2005 perhead, % of people without NBI 2005 (rest),%alphabetics, % school attendance, taxrevenues per capita (current $), municipalpublic investment per capita (current $), %of non-reliance on transfers

VAIA Institutional Number of violentaction by Illegal armedgroups (used as proxyto presence of illegalarmed groups),

2001–2007 Data Base – Andes University and defenseministery of Colombia

Number of violentaction

Number of violent action by Illegal armedgroups (FARC, AUC, ELN) por municipio,Incluye actos terrotistas, asaltos, atentados,Bloqueos de vias, Emboscadas,hostigamientos, ataques a población

MR Institutional Murder Rate by Illegalarmed groups (used asproxy to presence ofillegal armed groups),

2001 –2008 Colombian National Police - Estimation by theAuthors

Number of homocidesper 100.000inhabitants

Number of homicides per 100,000inhabitants committed by illegal armedgroups(FARC, AUC, ELN). Homicidescommitted by common crime are nottaken into account.

PPF Environmental Percentage of temunicipal territory inprimary forest area(2000)

2000 Colombia Ecosystem map (shape) - Institute ofHydrology, Meteorology and EnvironmentalStudies of Colombia (IDEAM - Spanishacronymun for “Instituto de Hidrología,Meteorología y Estudios Ambientales deColombia) - Estimation by the authors

Percentage Percentage of hectares of primary forest incomparison to the total area of themunicipality. Scale 1:500.000

106 A. Rincón Ruiz et al. / Ecological Indicators 34 (2013) 103– 112

gions

a

3

pytmboett2a

Map 1. Continental Natural Re

Table 1 indicates the year for which information could beccessed for each of the variables used in the analysis.

.2. Coca crops in forest and woody ecosystems

The analysis of coca crops in forest and woody ecosystems waserformed based on information from the ecosystem map for theear 2000, at a scale of 1:500,000, obtained from manual interpre-ation of Landsat satellite data (IDEAM et al., 2007). The ecosystem

ap from 2000 was overlaid with the coca census data (providedy SIMCI for this study) for two years, 2001 and 2008, and the areaf coca-cultivating land in both years that had been forest or woodycosystems in 2000 was calculated. The purpose of this analysis was

o give an estimation of the degree to which the different ecosys-em types had been affected by coca cultivation in both 2001 and008. These estimations were made from the natural cover (forestnd other woody formations) that existed in 2000 (IDEAM et al.,

of Colombia and Departments.

2007). These forests and formations were grouped into 16 differentecosystem types according to the climate and geological positionin the landscape.

3.3. Exploratory spatial analysis of coca cultivation

Two types of exploratory spatial data analysis (ESDA) werecarried out to shed light on the key factors associated with cocacultivation in Colombia, following the methods in Anselin (1995)and Patacchini and Rice (2007).

First a global multivariate spatial correlation analysis wasapplied in order to calculate, at the municipal level for all ofColombia, a single measurement of spatial correlation between the

percentage of land area under coca cultivation and the selectedeconomic, social, institutional and environmental variables. Theestimation was based on the multivariate Moran’s I coefficient(Anselin et al., 2002), and is referred to as the “global analysis” as it

ical Indicators 34 (2013) 103– 112 107

ita

cb(Itcaai2

s2ica

fgTwItftimivHhtLaYot

wMlsonaswe

4

4

bcqA

k

Table 2Relative proportions of coca crops in 2001 and 2008 in different ecosystems.

Natural Cover Area (Ha)

2001 2008

Tropical humid forest of the Orinoco and Amazon 67,600 20,555

A. Rincón Ruiz et al. / Ecolog

s based on a country-wide analysis without taking regional charac-eristics into account. This method shows how the analyzed factorsre associated with existence of illicit crops at the aggregated level.

The second ESDA approach involved a local multivariate spatialorrelation analysis to identify local patterns of spatial associationsased on the Local Indicator of Spatial Analysis – LISA approachAnselin, 1995). This approach was used to decompose the Moran’s

coefficients by identifying the contribution made by each observa-ion at the local level. We used a LISA extension for the multivariatease developed by Anselin et al. (2002). The multivariate LISA offersn indication of the degree of association between the percent-ge of coca area in each municipality and the variables of interestn neighboring municipalities (Anselin, 1995; Anselin et al., 1996,002).

Both global and local multivariate spatial analyses require con-tructing a ‘weight matrix’ (Anselin, 2005; Anselin et al., 2007,006), to define a local neighborhood value around each munic-

pality as a geographical unit. The weight matrix by contiguity washosen as the best option due to the heterogeneity of the polygonsnd the wide range of neighborhoods per municipality.2

Both global and local multivariate spatial correlation analysesollow Moran’s I statistic visualized in the four quadrants of theeneralized Moran scatter plot (Anselin, 1993; Anselin et al., 2002).his plot represents the analyzed variables in a standard form inhich the slope of the regression line represents the global Moran’s

statistic. In the present analysis it indicates the spatial associa-ion at the global (i.e. national) level. The scatter plot also reflectsour types of local spatial correlations between municipalities andheir neighboring areas. The upper right quadrant contains munic-palities with high values in the X variable (i.e., percentage of the

unicipal territory in coca area, or PCA) surrounded by municipal-ties with high values in the chosen Y variable (any of the other keyariables selected); we refer to these municipalities as having anigh–High (H–H) association. The lower right quadrant consists ofigh values of the X variable associated with low or zero values forhe Y variable, i.e. High–Low (H–L) association. On the lower left,–L consists of municipalities with low or zero values for the X vari-ble surrounded by municipalities with low or zero values for the

variable. Finally, the L–H quadrant on the upper left contains lowr zero values for the X variable with high neighboring values forhe Y variable.

A LISA cluster map (Anselin, 1995; Anselin et al., 2006, 2002)as used to show the location of observations with significant localoran statistics, highlighting significant spatial clusters and out-

iers, as well as showing the centers of the clusters. Hence, thepatial range of the clusters should be seen in the broader contextf the region; that is, including the neighboring regions which areot highlighted in the map. We represent the municipalities associ-ted with H–H, H–L, L–H and L–L values in the LISA cluster map. Thetatistical significance level of the various correlation coefficientsas estimated using 10,000 permutations (Anselin, 2005; Anselin

t al., 2002).

. Results

.1. Forest, woody ecosystems and coca crops (2001–2008)

The results show a marked change in the geographical distri-ution of coca cultivation from 2001 to 2008. Specifically, coca

ultivation decreased in the region of the Amazon and the Orino-uia and it increased in the Pacific, Caribbean and the north of thendean Region (Map 2). Based on the natural cover and ecosystems

2 There are three kinds of weight matrices: contiguity (queen, rook), distance, and-nearest neighbour.

Tropical rainforest in the Pacific 1982 8166Foothills of the Amazon and Pacific 3442 5155Forest in Catatumbo, Caribiam and Andes regions 17,181 9043

in the year 2000, it was found that in 2001 the most affected ecosys-tems by the expansion of coca cultivation were the humid tropicalforests, especially those in the departments of Orinoco and Amazon.In 2001, a total of 67,700 ha (40%) of the coca area was located inthe humid tropical forest and this had decreased to approximately20,500 ha (33%) by 2008. In addition, during the 2000 s, a displace-ment of coca crops to other regions took place, mainly towardsthe humid tropical forests of the Pacific Region. For example, thedata show that while in 2001, 1900 ha (1%) of the entire coca areawas located in the humid tropical forest of the Pacific region, thishad increased to 8100 ha (11%) by 2008. Similarly, while in 2001,around 3400 ha (2%) of the total coca area was located in the humidtropical forests found in the foothills of the Amazon and the Pacificregions, this had increased to 5200 ha (7%) by 2008.

Our results indicate that forest is the habitat type which ismost affected by coca cultivation, including those regions that hadnot been previously associated with coca cultivation, such as theecosystems of the Pacific region (see Map 2, Table 2).

4.2. Factors associated with coca cultivation (2001–2008)

The existence of coca crops in certain regions in Colombia is notonly due to favorable environmental conditions for its cultivation,but also depends on certain key social, economic and institutionalfactors. There follow the results of the local and global spatial mul-tivariate analyses, in which these factors are spatially analyzed atthe local scale.

4.2.1. Multivariate global spatial analysisFor both years (2001 and 2008), a significant global spatial asso-

ciation was identified between the percentage area under cocacultivation at the municipality level and all of the selected vari-ables (see Table 3). An inverse relationship was found between cocaarea and the following variables: Rural Population Density (RPD),Gross Domestic Product per Capita (GDPC), Primary Road Density(PRD) and the Index of Municipal Development (IMD). By contrast,a positive relationship was found between coca area and the RuralUnsatisfied Basic Needs index (RUBN), Murder Rate by illegal armedgroups (MR), Forced Displacement of Population (FDP), Violent Actscommitted by Illegal Armed groups (VAIA) and Percentage of Pri-mary Forest area (PPF).

4.2.2. Multivariate local spatial analysisFour key results were obtained from the multivariate local

spatial analysis. First, in 2001 the municipalities with coca culti-vation showed a significant spatial correlation with the variablesanalyzed in the global analysis, obtaining similar results in bothsign and significance of the correlations, except for Gross Domes-tic Product per Capita (GDPC), which was not significant at thelocal level. Second, none of the analyzed variables was exclusivelyassociated with the CCAs; that is, significant associations werealso identified in regions where coca crops were absent. Third,

between 2001 and 2008 a displacement of coca-growing areastowards the Pacific, the Caribbean and northern area of the Andeanregion was found. In the new CCAs significant spatial correla-tions were found, similar to the CCAs in 2001. Between 2001 and

108 A. Rincón Ruiz et al. / Ecological Indicators 34 (2013) 103– 112

nd coc

2eawb(m(U

TMa

a

Map 2. Annual average growth of coca a

008 the coca crops were displaced to areas with similar social,conomic, institutional and environmental characteristics. Fourth,lthough there were no individual factors exclusively associatedith CCAs, a distinct set of common factors between them could

e identified, including a low Index of Municipal DevelopmentIMD), low Primary Road Density (PRD), high Percentage of Pri-

ary Forest (PPF), high level of Forced Displacement of Population

FDP), significant presence of illegal armed groups and a high Ruralnsatisfied Basic Needs index (RUBN).

able 3ultivariate global spatial analysis (significant associations with percentage of coca

rea 2001 and 2008).

Category Variable association coefficient

2001a 2008a

Social Rural Unsatisfied BasicNeeds Index (RUBN)

0.035 0.179

Social Rural Densitypopulation (RDP)

−0.008 −0.018

Social Forced displacement ofpopulation rate -expulsion (FDP)

0.125 0.206

Economic Growth DomesticProduct per capita(GDPC)

−0.044 −0.073

Economic Primary Road Density(PDR)

−0.086 −0.238

Institutional Index of Municipaldevelopment (IMD)

−0.099 −0.253

Institutional Murder Rate by Illegalarmed groups (MR)(Proxy for illegalgroups presence)

0.017 0.147

Institutional Number of Violentaction (VAIA)(Proxy for illegalgroups presence)

0.170 0.180

Environmental Percentage of primaryforest (PPF)

0.125 0.276

Significative association (5%).

a crops over natural forest (2001–2008).

The results of the LISA analyses are presented below withaccompanying maps. Although we generated LISA maps for all thevariables, we have only included those which are significant in mostof the municipalities, as we did not wish to overload the documentwith maps. We found that the variables that showed significancein most of the territory were the Rural Unsatisfied Basic NeedsIndex (RUBN), the Index of Municipal Development (IMD) and thepercentage of the municipal territory in primary forest area (PPF).The variable designated Forced Displacement of Population rate -expulsion (FDP), despite its high significance in the global correla-tions, was not very representative in the LISA analyses carried outof the territory in the year 2001.

4.3. Environmental factors

While a positive association between CCAs and areas with a largepercentage of the municipal territory in primary forest area (PPF)was found for 2001 (H–H clusters), areas with a high PPF were alsoassociated with municipalities that had a low presence of coca crops(L–H clusters). However, in 2008, an expansion of the H–H clusterswas found towards the Pacific and the Caribbean regions, indicatingthat coca crops were displaced to areas with a relatively high PPF(Map 3).

4.4. Social factors

In 2001, the coca-growing municipalities were associated withhigh levels of Rural Unsatisfied Basic Needs (RUBN) indices – (H–Hclusters). This type of cluster was mainly found in the Amazon andOrinoquia regions, where the coca crops were mainly concentrated(Map 4). Nevertheless L–H clusters could also be identified, whichindicates municipalities with low or null coca levels were associ-ated with zones of high RUBN (i.e., the Pacific and the Caribbeanregions). By 2008, an expansion of the H–H clusters could be seen

in the Pacific, Caribbean and northern areas of the Andean region,while a decrease in the H–H clusters was observed in the Ama-zon region. L–L clusters could also be identified in the Andeanregion. In 2008, the variable Forced Displacement of Population

A. Rincón Ruiz et al. / Ecological Indicators 34 (2013) 103– 112 109

2008)

rac

tRtnasP

4

rbc

Map 3. LISA map for PPF (2001–200) and PCA (2001 –

ate – expulsion (FDP) showed similar behavior to the RUBN vari-ble. However, for the year 2001 there was a difference: H–Hlusters were only found in some municipalities in Amazonas.

Additionally, in 2001 there was a significant spatial associa-ion (H–L clusters) between coca-growing municipalities and a lowural Population Density (RDP). However, clusters showing statis-ically significant spatial relationships between areas with low orull coca levels and low RPD (L–L clusters) in other regions couldlso be found. In the year 2008, the existence of CCAs associatedignificantly with low RDP (H-L clusters) was found towards theacific and the north of the Andean region.

.5. Economic factors

Compared to the global multivariate analysis (Table 3), differentesults were found at the local level regarding the local associationetween levels of Gross Domestic Product per Capita (GDPC) andoca cultivation. While there was an inverse significant association

Map 4. LISA map for RUBN (2005) and PCA (2001–2008) b

by municipality (significant clusters at 0.05 sig. level).

at the global level in 2001 and 2008, there was no statistically sig-nificant association between them at the local level. In addition,Primary Roads Density (PRD), which is a factor that facilitates mar-ket development, showed a significant negative spatial associationwith the relatively more intense coca municipalities (H–L clusters),implying that CCAs that are isolated by road transport have a higherpresence of coca cultivation. The H–L clusters in the traditional cocaregions in 2001 and the expansion of the H–L clusters, by 2008,towards the Pacific, Caribbean and the northern region of the Andes,confirms this finding.

4.6. Institutional factors

Using the Index of Municipal Development (IMD), the local spa-

tial association analysis showed the presence of H-L clusters inthe CCAs (Map 5), implying that coca-growing municipalities wereassociated with municipalities which had relatively low levels ofIMD in 2001. However, areas with low levels of IMD could also

y municipality (significant clusters at 0.05 sig. level).

110 A. Rincón Ruiz et al. / Ecological Indicators 34 (2013) 103– 112

2008

bwttaVirlaacoTd

5

aybtiwtPrtf(hAmouiwbb

Map 5. LISA map for IMD (2001–2008) and PCA (2001 –

e found to be associated with municipalities where the coca areaas low or absent. By 2008, the H–L clusters had shifted towards

he Pacific and Caribbean regions from the Amazon region. Addi-ionally, variables linked with violent acts and murders by illegalrmed groups (number of Violent Acts by Illegal Armed groups –AIA, and Murder Rate by illegal armed groups – MR), both result-

ng from institutional weakness of the state, showed the followingesults. In 2001, H–H clusters (municipalities with high percentageevels of coca area associated with zones with high levels of VAIAnd MR) were identified in the CCAs. However, L-H clusters werelso found, indicating that in areas with a low to null presence ofoca, illegal armed groups were also present. Similar results werebtained in 2001 and 2008, in relation to the significant variables.he only difference was that the significant H–H clusters had beenisplaced to the southwest and the north of the country.

. Discussion

One of the points for discussion is the limitations caused by using single year for four of the variables considered in the LISA anal-sis. Information for both these years (2001 and 2008) could note found; only one datum was available in each case. With regardo the variables Rural Population Density (RPD) and Rural Unsat-sfied Basic Needs (RUBN), based on the data from previous years

e found that although there had been temporal changes, the spa-ial distribution was broadly maintained. In the case of the variablerimary Road Density (PRD), according to dialogue with experts onoad development in Colombia, there have not been big changes tohe proportions of roads between municipalities; we have there-ore made the assumption that although the Primary Road DensityPRD) has indeed changed, the distribution between municipalitiesas stayed constant. With respect to Percentage of Primary Forestrea (PPF), given that Colombia has only produced one ecosystemap (in the year 2000) and that we consider it fundamental to

ur analysis to include one variable of this type, we decided tose it despite the inaccuracies that would be introduced, apply-

ng the same assumption with respect to the spatial distribution ase introduced above. We believe that the analytical value gained

y including the variables outweighs any inaccuracies that coulde introduced in the results by doing so.

) by municipality (significant clusters at 0.05 sig. level).

Regarding the more general results found in the analysis, weconcur with (Thoumi, 2005a,b,c) that particular variables, suchas the economic profitability generated by the illegality and thehigh global demand for cocaine, cannot on their own explainwhy coca cultivation and cocaine production is concentrated incountries like Colombia. In fact, the analysis shows that thereis no single regional factor associated exclusively with the pres-ence of coca in certain areas in Colombia. For example, factorssuch as armed violence and conflict, the high levels of unsatisfiedbasic needs in rural areas, and the prevalence of poverty, oftencited as variables associated with the expansion of the cultiva-tion of coca (Angrist and Kugler, 2008; Departamento Nacionalde Planeación DNP, 2003; Dion and Russler, 2008), are not exclu-sively associated with the coca-growing areas in Colombia. Thisfinding also concurs with the view that illegal groups have existedand developed in Colombia independently of coca cultivation (Díazand Sánchez, 2004; Klein, 2007) and with the view that illicitcrops do not seem to be linked directly to the initiation of armedconflicts and hence the presence of illegal armed groups, eventhough they seem to prolong pre-existing armed conflicts (Ross,2004).

Regarding the comparison between coca cultivation in 2001 and2008 with respect to their location in what had been forest orwoody ecosystems in 2000, our results might represent an overes-timation because part of the natural forest in 2000 may have beentransformed first to agriculture or another type of land use and onlyafter that been used for coca cultivation, rather than having been adirect result of transformation of natural forest to coca crops. Nev-ertheless, the results clearly show an increase in coca cultivationin areas of humid tropical forest, which is of great environmentalimportance.

When more specific results are taken into account, we wouldlike to point out that forest cover in isolated areas provides suitableconditions for concealing coca cultivation. But while the litera-ture tends to find a positive relationship between road density anddeforestation (Ali et al., 2005; Perez-Verdin et al., 2009; Vance andGeoghegan, 2002), in the case of CCAs, a low road density, in turn

associated with low accessibility, is linked to the maintenance ofcoca cultivation. This implies that coca cultivation can remain in thearea and refrain from expanding to other forested areas. Neverthe-less, once land under coca cultivation expands, it is a major driver of

ical In

de

ttcftoorw

ttl(2acelhaGocc

pttC

6

tatiaerbaPi

atltfkbawtsiptit

A. Rincón Ruiz et al. / Ecolog

eforestation due to the establishment of roads and tracks (Rincont al., 2006).

While the global spatial analysis indicates an inverse correla-ion between Gross Domestic Product per Capita (GDPC) and CCAs,he association is not statistically significant when the analysis isarried out at the local level. In the latter case, coca crops can beound in both high and low GDPC areas, which is explained byhe fact that some coca municipalities are also important in termsf extractive natural resources such as gold mining, oil and palmil plantations. In future research it is necessary to analyze theelationship between coca municipalities and the distribution ofealth, as currently no data are available at the municipal level.

It is often suggested in the policy arena that in order to solvehe problem of expansion of illicit coca in Colombia, CCAs needo be transformed into economically productive zones based onegal commodity productive activities, such as palm oil plantationsDepartamento Nacional de Planeación DNP, 2007; FEDEPALMA,006, 2007). We posit, based on the data analyzed, that this isn over-simplification of the problem because the expansion ofoca crops is not exclusively associated with poverty and low lev-ls of GDPC. If CCAs continue to exhibit characteristics such asow state presence and low levels of municipal development, it isighly probable that even with crops like palm oil, deforestationnd the associated social conflicts will remain (Aristizabal, 2009;oebertus, 2008). New productive projects such as oil palm andther crops might replace illicit crops, though it is more likely thatoca will be displaced to other municipalities that share the rightonditions, as identified in the analysis.

We would also like to stress that some of the current mitigationolicies such as aerial aspersion of illicit crops might just create aemporary effect (Walsh et al., 2008), since favorable characteris-ics for coca expansion would still be present in other regions ofolombia.

. Conclusion

Based on exploratory spatial data analysis (ESDA), we have iden-ified correlation patterns of coca-growing regions associated with

diverse set of characteristics (social, economic and environmen-al) between 2001 and 2008. The analysis has also proved usefuln showing that illicit crops are primarily grown in areas whichre characterized by forest and woody ecosystems and that thesecosystems seem to be the most affected by illicit crops. Also, ouresults indicate that forest is the habitat type which is most affectedy coca cultivation, including those that had not been previouslyssociated with coca cultivation, such as the ecosystems of theacific region (see Map 2). This has caused several significant andn many ways irreversible impacts on forest ecosystems.

The analysis has allowed the finding of a significant global spatialssociation in the two years of analysis (2001 and 2008) betweenhe percentage of area under coca cultivation at the municipalityevel and each of the chosen variables depicting the environmen-al, social and economic context in Colombia. When the analysisocuses at a local level the results appear in a new light and variousey messages need to be flagged: First, none of the variables cane considered to be exclusively associated with the coca-growingreas since significant associations were also identified in regionsithout coca. Second, although there are no individual regional fac-

ors exclusively associated with the coca-growing areas, there is aet of common factors that are characteristic of these areas, includ-ng low levels of municipal development, low road density, high

resence of primary forest, high presence of forced displacement,he presence of illegal armed groups and a high prevalence of unsat-sfied basic needs. Third, a spatial displacement of the coca cropsowards the Pacific, the Caribbean and northern area of the Andean

dicators 34 (2013) 103– 112 111

region can be observed between 2001 and 2008. In the new coca-growing areas a similar set of significant correlations was foundregarding the variables analyzed in 2001. This implies that between2001 and 2008 the coca crops were displaced to areas with similarsocial, economic, institutional and environmental characteristics.

The analysis shows that coca-growing areas in Colombia haveparticular regional characteristics that create favorable conditionsfor loss of forest and woody ecosystem. Despite CCAs being mostlythose areas that maintain a large proportion of primary forest andhaving favorable environmental conditions for coca production,these areas remain not only physically isolated (low road density),but also socially deprived (as reflected by their high index of unsat-isfied basic needs index), as well as institutionally impaired (asreflected by their low municipal development indices). These char-acteristics together create a favorable context for the existence ofillegal armed groups to enforce control over the territory and itspopulation through violence and forced displacement. Under thesecircumstances the forests and the local population become easilyexploitable for illegal activities. Through violence, a new structureof rules is created in these regions, in which the forest is consid-ered an easily and freely utilizable resource. This, in combinationwith the rural population remaining in deprived social conditions,creates the favorable scenario for deforestation by coca crops forillegal use.

If the root causes for the existence of such deprived social, insti-tutional and economic factors are not tackled in a structured way,large areas in Colombia will continue to provide the necessaryconditions for the continuation of illegal activities such as coca cul-tivation for cocaine production as well as other illegal extractiveactivities such as mining (e.g. coltan and gold) or palm cultivation.Whether such illegal ventures will flourish or perish in Colombiais thus a matter of reframing past development paradigms andembracing sustainable development (Ballvé, 2009).

Acknowledgments

We thank the Programme Alßan for financial support and theconstructive comments made by Susanne Menzel, Suzette Flantuaand Joan Martinez Alier. We would like to thank Francisco Thoumifor his guidance at the beginning of the research, and the IntegratedIllicit Crops Monitoring System (SIMCI in Spanish) for their supportand supply of spatial information on coca crops. Thanks to BlancaEdilma for your support all these years.

References

Ali, J., Benjaminsen, T.A., Hammad, A.A., Dick, Ø.B., 2005. The road to deforestation:an assessment of forest loss and its causes in Basho Valley, Northern Pakistan.Global Environ. Change A 15, 370–380.

Álvarez, M., 2001. Could peace be worse than war for colombia’s forests? In: TheEnvironmentalist. Kluwer Academic Publishers, The Netherlands, pp. 305–315.

Álvarez, M., 2007. Environmental damage from illicit drug crops in Colombia. In: DeJong, W.D., Deanna, Abe, Ken-ichi (Eds.), Extreme Conflict and Tropical Forests.Springer, Dordrecht, The Netherlands, p. 184.

Álvarez, M.D., 2003. Forests in the time of violence – conservation implications ofthe Colombian war. J. Sustainable For. 16, 47–68.

Angelsen, A., 1999. Agricultural expansion and deforestation: modelling the impactof population, market forces and property rights. Journal of Development Eco-nomics 58, 185–218.

Angrist, J.D., Kugler, A.D., 2008. Rural windfall or a new resource curse? Coca,income, and civil conflict in Colombia. Rev. Econ. Stat. 90, 191–215.

Anselin, L., 1993. The Moran scatterplot as an ESDA tool to assess local instabil-ity in spatial association. In: Fischer, M., Scholten, H., Unwin, D. (Eds.), SpatialAnalytical Perspectives on GIS. Taylor and Francis, London, pp. 93–115.

Anselin, L., 1995. Local indicators of spatial association—LISA. Geogr. Anal. 27,93–115.

Anselin, L., 2005. Exploring Spatial Data with GeoDa: A Workbook. Center forSpatially Integrated Social Science/Spatial Analysis Laboratory, Department ofGeography – University of Illinois, Urbana, Champaign.

Anselin, L., Bera, A.K., Florax, R., Yoon, M.J., 1996. Simple diagnostic tests for spatialdependence. Reg. Sci. Urban Econ. 26, 77–104.

1 ical In

A

A

A

A

A

BC

D

D

D

D

D

D

E

E

F

F

FF

G

G

I

I

12 A. Rincón Ruiz et al. / Ecolog

nselin, L., Sridharan, S., Gholston, S., 2007. Using exploratory spatial data analysisto leverage social indicator databases: the discovery of interesting patterns. Soc.Indic. Res. 82, 287–309.

nselin, L., Syabri, I., Kho, Y., 2006. GeoDa: an introduction to spatial data analysis.Geogr. Anal. 38, 5–22.

nselin, L., Syabri, I., Smirnov, O., 2002. Visualizing Multivariate Spatial Correlationwith Dynamically Linked Windows. University of Illinois, Urbana-Champaign,Urbana, IL.

ristizabal, A., 2009. Efectos del monocultivo de la palma de aceite en los mediosde vida de las comunidades campesinas: el caso de Simití–sur de Bolívar, Mas-ter of rural development. Pontificia Universidad Javeriana de Colombia, Bogotá,Master thesis.

rmenteras, D., Rudas, G., Rodriguez, N., Sua, S., Romero, M., 2006. Patterns andcauses of deforestation in the Colombian Amazon. Ecological Indicators 6,353–368.

allvé, T., 2009. The dark side of plan Colombia. Nation 15 (June), 9.ODHES, 2009. Víctimas emergentes - Desplazamiento, derechos humanos y

conflicto armado en 2008. In: Boletín informativo de la Consultoría para losDerechos Humanos y el Desplazamiento,. Codhes - Consultoría para los DerechosHumanos y el Desplazamiento, Bogotá, Colombia, pp. 15.

ávalos, L., Bejarano, A., Hall, M., Correa, H.L., Corthals, A., Espejo, O.J., 2011. Forestsand drugs: coca-driven deforestation in tropical biodiversity hotspots. Environ.Sci. Technol. 45, 1219–1227.

eininger, K., Minten, B., 2002. Determinants of deforestation and the eco-nomics of protection: an application to Mexico. Am. J. Agric. Econ. 84,943–960.

epartamento Nacional de Planeación (DNP), 2003. Plan Nacional de Desar-rollo 2002–2006/Hacia un estado comunitario. Departamento Nacional dePlaneación, Bogotá, pp. 284.

epartamento Nacional de Planeación (DNP), 2007. Documento CONPES No 3477.Estrategia para el desarrollo competitivo del sector palmero colombiano. Depar-tamento Nacional de Planeación (DNP), Bogotá, pp. 30.

íaz, A.M., Sánchez, F., 2004. Geography of Illicit Crops (Coca Leaf) and Armed Con-flict in Colombia. Center of Studies for Economic Development, Department ofEconomics. Universidad de los Andes.

ion, M.L., Russler, C., 2008. Eradication efforts, the state, displacement and poverty:explaining coca cultivation in Colombia during Plan Colombia. J. Latin Am. Stud.40, 399–421.

tter, A., McAlpine, C., Wilson, K., Phinn, S., Possingham, H., 2006. Regional patternsof agricultural land use and deforestation in Colombia. Agric. Ecosyst. Environ.114, 369–386.

vans, T., York, A., Ostrom, O., 2008. Institutional dynamics, spatial orga-nization, and landscape change. In: Wescoat, J.L., Johnston, Douglas M.(Eds.), Political Economies of Landscape Change. Springer, Netherlands,pp. 111–129.

ajardo, D., 2002. Para sembrar la paz hay que aflojar la tierra. Instituto de EstudiosAmbientales (IDEA). Universidad Nacional de Colombia, Bogotá, Colombia.

arrell, G., Thorne, J., 2005. Where have all the flowers gone? Evaluation of theTaliban crackdown against opium poppy cultivation in Afghanistan. Int. J. DrugPolicy 16, 81–91.

EDEPALMA, 2006. La palma social – Video. FEDEPALMA, Colombia.EDEPALMA, 2007. The Faces of the Palm Oil, the Relevance of the Oil Palm Agro-

Industry in Colombia. FEDEPALMA, Bogotá.arcés, L., 2005. Colombia: the link between drugs and terror. Journal of Drug Issues

35 (1), 83–105.oebertus, J., 2008. Palma de aceite y desplazamiento forzado en Zona

Bananera: trayectorias entre recursos naturales y conflicto. Colombiainternacional/Departamento de Ciencia Política y Centro de Estudios Interna-cionales. Facultad de Ciencias Sociales, vol. 67. Universidad de los Andes, Bogotá,

pp. 152–175.

banez, M., Carlsson, F., 2009. A survey-based choice experiment on coca cultivation.J. Dev. Econ. (In Press).

bánez, A.M., Vélez, C.E., 2008. Civil conflict and forced migration: the micro determi-nants and welfare losses of displacement in Colombia. World Dev. 36, 659–676.

dicators 34 (2013) 103– 112

Ibanez, M., 2010. Who crops coca and why? The case of Colombian farmers, Discus-sion Papers, Courant Research Centre PEG.

IDEAM, IGAC, IAvH, Invemar, Sinchi, IIAP., 2007. Ecosistemas continentales, cos-teros y marinos de Colombia. Instituto de Hidrología, Meteorología y EstudiosAmbientales (IDEAM), Instituto Geográfico Agustín Codazzi (IGAC), Instituto deInvestigación de Recursos Biológicos Alexander von Humboldt (IAvH), Institutode Investigaciones Ambientales del Pacífico Jhon von Neumann (IIAP), Institutode Investigaciones Marinas y Costeras José Benito Vives De Andréis (INVEMAR)e Instituto Amazónico de Investigaciones Científicas (Sinchi), Bogotá, Colombia.

Klein, N., 2007. Shock Doctrine: The Rise of Disaster Capitalism. Metropolitan Books,New York.

Martinez-Alier, J., Kallis, G., Veuthey, S., Walter, M., Temper, L., 2010. SocialMetabolism, Ecological Distribution Conflicts, and Valuation Languages. ECOL.ECON 70, 153–158.

Norgaard, R., 1994. Development Betrayed: The End of Progress and a Coevolutio-nary Revisioning of the Future. Routledge.

Patacchini, E., Rice, P., 2007. Geography and economic performance: exploratoryspatial data analysis for Great Britain. Region. Stud. 41, 489–508.

Perez-Verdin, G., Kim, Y.-S., Hospodarsky, D., Tecle, A., 2009. Factors driving defor-estation in common-pool resources in northern Mexico. J. Environ. Manag. 90,331–340.

Rincon, A., Romero, M., Bernal, N., Rodriguez, N., Rodriguez, J., 2006. Modelamientode presiones sobre la biodiversidad en la Guayana. Revista Internacional desostenibilidad. Tecnología y Humanismo 1, 34.

Rocha, R., 1997. Aspectos económicos de las drogas ilegales, Editorial Planeta.Rocha, R., 2000. La economía colombiana tras 25 anos de narcotráfico. In: Siglo del

Hombre. UNDCP, Bogotá, Colombia.Ross, M.L., 2004. What do we know about natural resources and civil war? J. Peace

Res. 41, 337–356.Scott, J., 1998. Seeing Like a State How Certain Schemes to Improve the Human

Condition Have Failed. Yale University Press, New Haven,London.Solomon, K., Anadón, A., Luiz, C.A., Marshall, J., Sanin, L.H., 2005. Estudio de los efectos

del Programa de Erradicación de Cultivos Ilícitos mediante la aspersión aérea conel herbicida Glifosato (PECIG) y de los cultivos ilícitos en la salud humana y en elmedio ambiente. Comisión Interamericana para el Control del Abuso de Drogas(CICAD) – División de la Organización de los Estados Americanos.(OEA).

Thoumi, F., 2005a. The Colombian competitive advantage in illegal drugs: the roleof policies and institutional changes. J. Drug Issues 35 (1), 7–25.

Thoumi, F., 2005b. The causes of illegal drug industry growth in the Andes, anti-drug policies and their effectiveness, Centro Editorial Universidad del Rosarioed. Centro de Estudios y Observatorio de Drogas y Delito – Facultad de Economía.Universidad del Rosario, Bogotá, Colombia, pp. 57.

Thoumi, F., 2005c. Why a country produces drugs and how this detennines policyeffectiveness: a general model and some applications to Colombia. In: Rojas, C.,Meltzer, J. (Eds.), Elusive Peace: International, National and Local. Dimensionsof Conflict in Colombia. Palgrave Macmillan, New York.

UNODC, 2006. Características Agroculturales de los Cultivos de Coca en Colombia.United Nations Office on Drugs and Crime – UNODC, Colombia, Bogotá.

UNODC, 2008a. Colombia coca cultivation survey. United Nations Office on Drugsand Crime - UNODC, Bogota, Colombia, pp. 100.

UNODC, 2008b. Análisis Multitemporal de cultivos de coca, período 2001–2006.United Nations Office on Drugs and Crime, Bogotá.

UNODC, 2009. Análisis Multitemporal de cultivos de coca, período 2007–2008.United Nations Office on Drugs and Crime – UNODC, Bogotá.

UNODC, 2010. The Globalization of Crime A Transnational Organized Crime ThreatAssessment. United Nations Office on Drugs and Crime – UNODC.

Vance, C., Geoghegan, J., 2002. Temporal and spatial modelling of tropical defor-estation: a survival analysis linking satellite and household survey data. Agric.Econ. 27, 317–332.

Vargas, R., 2005. Narcotráfico, guerra y políticas antidrogas. Acción Andina, Bogotá,Colombia.

Walsh, J., Sánchez, J., Salinas, Y., 2008. Chemical Reactions Fumigation: SpreadingCoca and Threatening Colombia’s Ecological and Cultural Diversity. WashingtonOffice on Latin America (WOLA), Washington.