World Development Vol. 33, No. 9, pp.

1419–1434, 2005
Ó 2005 Elsevier Ltd. All rights reserved
Printed in Great Britain
www.elsevier.com/locate/worlddev 0305-750X/$ - see front matter
doi:10.1016/j.worlddev.2004.10.006

Village Economic Opportunity, Forest Dependence,

and Rural Livelihoods in East Kalimantan,
Indonesia
*
SONYA DEWI, BRIAN BELCHER and ATIE PUNTODEWO
Center for International Forestry Research, Indonesia
Summary. — The changing role of forests in people’s livelihoods in frontier areas is important from
the perspective of poverty alleviation and forest conservation. This study explores the link between
expanding economic opportunities, forest dependence, and welfare in 73 villages. Village economic
options, forest cover, and land suitability for agriculture and forestry are determining factors of
people’s well-being. Increased accessibility to markets and deforestation are strongly associated
with economic diversity at the village level. Increased economic diversity, larger areas of forests,
more intensive land use, higher endowments of agricultural land and forest, and higher village pop-
ulation are related to increased well-being.
Ó 2005 Elsevier Ltd. All rights reserved.

Key words — Asia, Indonesia, spatial analysis, geographical targeting, income portfolio, rural
nonfarm employment

1. INTRODUCTION livelihoods is needed to better understand their

impact and to target interventions aimed at
Increasing interest in rural poverty allevia- improving livelihoods in forest areas.
tion has resulted in a new focus on the ‘‘for- Some argue that forest resources can be put
est-dependent poor’’ and on the actual and to work to help improve the livelihoods of the
potential contributions of forests to livelihoods poor (Scherr, White, & Kaimowitz, 2002),
(Angelsen & Wunder, 2003; Arnold, 2001; while others believe that forests have a limited
Dove, 1993; World Bank, 2001; Wunder, potential to contribute to poverty reduction
2001). Forest-rich areas are frequently associ- (Wunder, 2001). Part of the discrepancy be-
ated with rural poverty. Remote locations, tween these views originates in different
rugged terrain, low population densities, limited assumptions about the scope for creating new
communications and transportation infrastruc- opportunities for rural people to take advan-
ture, infertile soils, and difficult climates act as tage of forest resources. Typically, the poor
constraints that limit forest harvesting and the
conversion of forests to agricultural land. The
same factors limit the economic opportunities * This paper was developed from a poster presented at
of people living in the area (Ashley & Maxell, the International Conference on Rural Livelihoods,
2002; Wunder, 2001). Development in frontier Forest, and Biodiversity, held in Bonn, Germany, in
areas in the form of resource extraction pro- 2003. We thank the editorial committee for this invita-
jects, infrastructure development, or industrial tion. The paper has benefited from insightful and cons-
development, for example, rapidly changes tructive comments and suggestions from Arild Angelsen,
both the physical and the economic landscapes William Sunderlin, Lini Wollenberg, Benoit Mertens,
and the corresponding opportunities for local and two anonymous reviewers. The research was sup-
people (Brookfield, Potter, & Byron, 1995; ported by grants from the Canadian International De-
Padoch & Peluso, 1996). Analysis of the rela- velopment Agency (CIDA) and CIFOR’s core donors.
tionship between such developments and rural Final revision accepted: October 25, 2004.
1419
1420 WORLD DEVELOPMENT

residents of forest-rich areas have not had a Efforts to encourage equitable and durable
voice in deciding how to use forests and have improvements in human well-being need to
lacked opportunities to transform forest re- take account of the direct and indirect liveli-
sources into wealth for themselves. They tend hood impacts of resource-extraction projects,
to lack ownership and control over the re- and the impacts of broader development ef-
sources, and do not have access to the markets forts. In this paper, we examine these issues
or the skills and contacts needed to take advan- as they have been played out in East Kaliman-
tage of opportunities. And, while there is a glo- tan, a forest-rich and fast-changing part of Bor-
bal trend toward devolving natural resource neo. Agriculture is the mainstay in the area,
management rights and responsibilities to local based primarily on a shifting cultivation system
communities, this is usually in degraded forest in which regenerating forest is used to replenish
areas and in many cases has resulted in in- soils and control weeds, and the secondary
creased state control (Edmunds & Wollenberg, forest is managed for a variety of economic
2003). Governments are reluctant to give up products (Lahjie, 1996; Sardjono & Samsoedin,
control over valuable revenue sources. There 2001). Forests are very important in local live-
is a tendency to assign rights to well-connected lihoods. Many valuable products are harvested
investors as in Sarawak (Cooke, 2002), Kali- from the forest and rivers for direct consump-
mantan (Dove, 1993, 1996; Dove & Kammen, tion and for sale. We use the term ‘‘agro-
2001), and Thailand (Vandergeest & Peluso, forestry’’ throughout the paper to define this
1995). Decisions about forest use and invest- integrated system, in which households rely
ments in forest areas are often determined by on agriculture and forestry both simultaneously
outsiders interested in short-term extraction of and sequentially for their livelihoods.
forest resources, with little concern about the The main drivers of land use change in the
long-term economic impact on local people. area are forest concessions, timber and estate
Large-scale resource extraction, forest conver- crop plantations, mining, road building by pro-
sion, and other development projects in forest jects and public investment, and migration
areas tend to provide little direct benefit to for- (both government sponsored ‘‘transmigration’’
est-dependent communities, and more com- and independent migration) (Brookfield et al.,
monly have a negative impact as local 1995; Padoch & Peluso, 1996). We investigate
peoples’ access to resources is curtailed and the impact of forest-based and infrastructure
the physical resource base is degraded (Brook- developments on economic opportunities and
field et al., 1995; Potter & Lee, 1998). welfare in the study area. We focus on the rela-
On the other hand, development in forested tionships between village-level income diversity
areas might be expected to create new opportu- (as an indicator of economic opportunity) and
nities for generating employment, income, and village welfare. We look at the relationship of
wealth. The World Bank (2002) identifies lim- those characteristics to the location of indus-
ited land and market opportunities as a major trial resource extraction projects, roads, and
constraint to poverty reduction. Road building, other economic features like forest cover,
for example, increases people’s access to facili- land-use suitability, and political and economic
ties and resources and also reduces transport centers. This allows an empirical test of the role
costs. Several studies in rural Latin America of these various factors in determining welfare.
found that factors outside the agricultural We use the village as the unit of analysis for
sector raised demand for nonfarm goods and two reasons: First, communities in the area
services and resulted in increased local incomes tend to be small (mean = 150 households, but
and the accumulation of capital for investment. many have fewer than 100) and relatively
Investment further increases rural nonfarm homogeneous, and the main drivers of change
wages and self-employment through produc- operate at a scale larger than a single commu-
tion and expenditure linkages (‘‘economic nity. The main differences in income opportuni-
transformation’’) to reduce the incidence of ties, welfare, and other factors of interest are
poverty (Reardon, Berdegué, & Escobar, found across communities rather than within
2001). Benefits arise in the form of improved them. Second, for an analysis of spatial pat-
transportation facilities, better market access, terns, we need to cover an area large enough
new markets for agricultural products, oppor- to capture a range of stimuli. This approach
tunities for petty trading, and some employ- is consistent with the geographic targeting of
ment, at least in the short term (Brookfield small administrative regions advocated by Big-
et al., 1995). man and Fofack (2000). We supplement the
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1421

village-scale analysis with household level data ences in the analysis. These include a gold
and case studies to explore how different de- mine, a coal mine, an oil-palm plantation, a
grees of forest dependence at the household le- rubber plantation, some transmigration sites
vel are reflected at the community level. in West Kutai, and oil-palm plantations in the
Paser district.

2. STUDY AREA
3. DATA
The study area includes the contiguous Kabu-
patens (districts) of Paser and Kutai Barat in (a) Spatial data
the Indonesian province of East Kalimantan,
covering 1,111,107 ha (Figure 1). The area is Geo-referenced data were collected from a
undergoing a rapid change. Previously heavily variety of sources, including (i) topographic
forested, it has been deforested by large-scale and infrastructure maps produced by the Na-
resource extraction, small-scale agricultural tional Coordinating Agency for Surveys and
expansion, and two episodes of large-scale for- Mapping in 1991; (ii) Land Systems and Suit-
est fire, one in 1982 and the other after the ability maps (Regional Physical Planning Pro-
study period, in 1997–98. Marked improvement gram for Transmigration—RePPProt, 1982);
in transportation and communication infra- and (iii) maps of company locations produced
structure in recent years, in-migration from by the Ministries of Forestry and Transmigra-
other areas of Indonesia, and new markets for tion in 1997. Land-cover maps were produced
agricultural and forest products and for labor using unsupervised classification of Landsat
have led to important socioeconomic changes. TM images for 1992 and 1996, with three clas-
There are clear gradients from villages that ses: (i) mature forest; (ii) young secondary for-
have been strongly affected by development to est; (iii) nonforest land. This classification
villages that remain remote and less affected scheme allows an accurate and clear classifica-
by modernization. tion because each of the three classes has a dis-
The most important activities driving change tinctive signature, and it fits well with our
in the study area are logging (forest conces- definition of deforestation. We calculated the
sions), 1 timber plantations, 2 coal mining 3, deforestation rate as the area of mature forest
oil-palm plantations, 4 smallholder rubber and young secondary forest in period 1 that
plantations 5, and transmigration projects, 6 as changed to nonforest in period 2, divided by
well as road building 7 (Figure 1). All except total forest in period 1 per village area. Small-
smallholder rubber planting and roads compete holder deforestation patches were distinguished
directly with local people for resources by limit- from others by using spatial characteristics
ing access and/or by depleting or degrading the including size and location. In shifting cultiva-
resource base. Some employment is generated tion systems, the length of the fallow period is
by these activities, but there has been a tendency commonly used as an indicator of land use
to hire from outside the local area. Local people intensity (Inoue, 2000). In this study, a higher
do not tend to have the skills and experience or, proportion of new agricultural plots created
according to some employers, the work ethic, to from young secondary forest (i.e., land that
compete for anything but low-paid jobs. There had been cultivated too recently to permit
are projects to also develop roads, improve mar- mature forest to regenerate) was taken as an
ket access, and create new markets for locally indicator of higher land use intensity. An index
produced supplies, and some services. Some of land use intensity was calculated as the
companies are required by law to support com- proportion of agricultural plots cleared from
munity development in their surrounding areas, young secondary forest relative to total small-
although there are few demonstrable impacts. holder deforestation patches.
Table 1 summarizes information about each A road map was developed by combining the
project in the study area. official national maps, Landsat TM 1992, and
The boundary of the study area was defined maps produced by companies. Access to mar-
based on data availability. Cloud cover in satel- kets and other features were calculated in terms
lite images is a key limiting factor in this high- of travel time. Accessibility was weighted by the
rainfall area; no ‘‘cloud-free’’ images were quality of roads. Four classes of roads were
available. Projects located outside the bound- identified: (i) the trans-Kalimantan highway,
aries of the study area were included as influ- which is a surfaced road built and maintained
1422 WORLD DEVELOPMENT

Figure 1. Map of study area showing villages, transportation network, and large-scale projects.

by the province to connect district capital cities which are unsurfaced roads but usually of a
in Kalimantan; (ii) district roads, which are sur- good quality to accommodate heavy equip-
faced roads, some of which were built in con- ments; and (iv) logging and plantation roads,
junction with transmigration projects; (iii) which are low-quality roads built by compa-
mining roads, built by mining companies, nies, and often not usable during the rainy sea-
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1423

Table 1. Summary of projects within the extent of the study area

Projects Area (ha) District Established References
Logging concessions 326,416 Paser and West Kutai Varies Integrated Forest Fire
Management (IFFM) map
Oil-palm estates 13,141 Paser Mid-1980s IFFM map
Timber plantation 29,414 Paser and West Kutai 1992 IFFM map
Coal mine 15,998 Paser 1986 Kompas, August 28, 2001

son. Class (iv) roads have a shorter life, quickly 4. METHODS

becoming unusable without regular mainte-
nance. (a) Indices of economic diversity and
Officially, districts are subdivided into sub- well-being at the village level
districts based on watershed boundaries. Vil-
lage boundaries are based on local custom, We developed two indices of village eco-
are not recognized by the state, and are not nomic diversity and peoples’ well-being at the
marked on any available map. We estimated village level.
these boundaries by drawing polygons through
midpoints between neighboring villages using (i) Village economic diversity index (EDI)
the Thiessen polygon technique. This roughly Economic diversity at the village level is de-
approximates the customary method of defin- fined in terms of heterogeneity of income
ing boundaries, 8 and, judging from field obser- sources among households in a village. In
vations and interviews, it gave a reasonable remote rural locations such as our case study
estimate. Still, we needed to be cautious in area, where agriculture and forestry are the pre-
interpreting the results as incorrect boundaries dominant economic activities and where there
influence several variables. are few alternative opportunities, economic
diversity is low. It is hypothesized that develop-
(b) Village-level socioeconomic data ment will lead to new employment opportuni-
and household survey information ties and therefore increased economic diversity.
We developed a village EDI using the
Secondary, village-level socioeconomic data Shannon–Weaver diversity index, which is a
were obtained from the Agricultural Census measure used commonly in ecology and
of the Central Board of Statistics: Village Po- information theory. 10 The index is formulated
tential, 1993 9, the latest 10-yearly census data as
X
available. These data were used to develop EDI ¼
lnðpi Þ;
indices of village-level economic diversity and pi
well-being and are described below.
Primary data were collected using structured where pi is the proportion of households in a
interviews of household heads in randomly se- village that rely primarily on each main income
lected households covering between 15% and source and i = 1, . . . , n, where n is the number
50% of households in each of 10 villages in of classes of main income sources. A household
1997–98 (Figure 1). The villages were selected is classified into one category based on its main
to represent a range of points along a gradient income source and pi’s add up to one. The in-
of distance to projects and other developments, dex ranges from 0 to ln(n). In a village where
stratified by district. Questionnaires were de- all households have the same main income
signed to capture information on demograph- source, EDI = 0. Where there is an even distri-
ics, assets, income by source, and land uses. bution of all possible main income sources
These data give us detailed village case studies among households in a village, EDI = ln(n).
to compare with the main analysis based on In this case, with four possible main income
the Agricultural Census data. The comparabil- sources, perfect heterogeneity would give an
ity may be compromised by the five-year lag be- EDI of ln(4). 11
tween the Agricultural Census (1993) and the The Agricultural Census of the Indonesian
household survey. However, we found a good Central Board of Statistics 1993 aggregates
correspondence between similar data in the main income sources into four classes: agricul-
two data sets. ture, mining, wage, and services. This usefully
1424 WORLD DEVELOPMENT

groups sources of income as farm, mining, and VDI ¼ ððh
hmin Þ=ðhmax
hmin Þ
nonfarm wages. 12 It should be noted that the þ ðe
emin Þ=ðemax
emin Þ
index is sensitive to the number and definition
of classes used. In our study area, people þ ða
amin Þ=ðamax
amin ÞÞ=3;
typically integrate agriculture and forestry where h is the infant survival rate; e is the
activities in an indigenous agroforestry system, school enrollment rate for children of age 7–
and we follow this in classifying these activities 11 years; a is the assets (motor bikes and
as one. motor boats) per household, 14 and subscripts
Because agroforestry is the most common max and min refer to the maximum and mini-
main source of income, the EDI of a village will mum values in the area. The minimum and
be increased when some households derive their maximum possible values of VDI are 0 and 1,
main income from sources other than agrofor- respectively. The index gives an equal weight
estry. The index is used as a measure of increas- to health, education, and wealth. For our data-
ing economic opportunities in an area where set, Cronbach’s alpha reliability estimate of
such opportunities are typically limited. A dif- VDI was 0.65. This is close to the common
ferent index would be needed in urban and other cut-off point of 0.7 for demonstrating internal
areas with higher levels of diversity at time zero. consistency in the index data (Nunnaly, 1978).
(ii) Well-being at the village level (VDI)
The second index is a measure of people’s (iii) Relationships between EDI and VDI
well-being at the village level, based on their Village economic diversity, measured by
health, education, and wealth relative to other EDI, is expected to increase as new opportuni-
villages in the study area. It loosely follows ties are created by improved transportation
the idea of the UNDP Human Development infrastructure, product markets, labor markets,
Index (HDI), which uses several socioeconomic and land and resource availability. (The last is
indicators relating to income per capita, health, affected by deforestation.) The VDI is hypothe-
and education to compare average national sized to be determined by economic opportuni-
levels of welfare. 13 ties, measured by the EDI, as well as by the
The Village Development Index (VDI) is cal- availability of agricultural and forest resources,
culated as follows, based on the Agricultural and agricultural practices. Population density is
Census Data: also expected to contribute to the VDI in this

Table 2. List of variables

Variables Descriptions
Economic diversity index Measure of heterogeneity of main income sources
of households in a village
Village development index Measure of development or well-being at the village level
in terms of health, education, and wealth
Forest cleared over total Proportion of area cleared from forest over
the total cleared area by smallholders
Population Village population
Time to district Distance (in hours) from the settlement to the district capital
Time to subdistrict Distance (in hours) from the settlement to the subdistrict capital
Time to logging Distance (in hours) from the settlement
to the closest active logging company
Time to transmigration Distance (in hours) from the settlement
to the closest transmigration site
Provincial road, district road, Density of provincial road, district road, mine road,
mine road, logging/plantation road and logging/plantation road (km/ha)
Forest 92 Proportion of village area covered by forest in 1992
Deforestation 92–96 Deforestation rate from 1992 to 1996 (changes from forest and
regrowth to nonforest land divided by forest cover area in 1992)
Agrosuitability Proportion of suitable area for agroforestry over total area
Land-use intensity Proportion of area of young secondary forest in 1992 changed to
nonforest land in 1996 over total smallholder deforestation
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1425

remote area, through increasing economic Table 3. Spatial error model of village EDIa
interactions. We also included proximity to Variable Coefficient z-Value
the district and subdistrict capital as this vari-
able measures access to the local center of trade Provincial road 206.31 5.17**
District road 54.13 3.41**
and services, and also to the government ser-
Mine road
8.65
0.41
vices such as health care, education, and infra-
Logging/plantation road 23.16 2.11*
structure development.
Time to logging 4.10E
03 0.43
Time to transmigration 0.02 3.35**
(b) Spatial statistical analyses of EDI and VDI
Deforestation 1992/1996 2.27 3.55**
Lambda
0.38
2.79**
We explore the empirical relationships: (i) be- a
tween village economic diversity and distances With maximum likelihood estimation, number of
from projects and developments and (ii) among observation = 73, R2 = 0.82, LIK = 5.62, AIC = 2.76.
*
Significance level of 0.05.
well-being at the village level, village economic **
Significance level of 0.01.
diversity, and other factors using regression
analysis of spatial data and village-level socio-
economic data (Table 2). We use Space Stat,
a software designed to build spatial economet- gration sites are strongly and positively corre-
ric models, in conjunction with ArcView, a lated with one another. Therefore, we included
GIS software. Space Stat provides a set of diag- only one variable—distance to transmigration
nostic tools to determine an appropriate model site—in the analysis to avoid a collinearity
to handle a particular form of spatial depen- problem. This collinearity reflects the deliberate
dence when it exists (Anselin, 1999). In order development of particular economic activities
to incorporate spatial dependence in the ana- within transmigration projects. In the study
lysis, a neighborhood matrix is generated by area, the main transmigration area was devel-
considering villages to be connected when they oped in conjunction with an oil-palm planta-
share boundaries and a transportation link tion and processing factory and, to a lesser
(either road or river). extent, with timber plantation and rubber pro-
We further explored the links between EDI/ jects. Transmigrants were provided with an oil-
VDI and the role of forests in livelihoods using palm plantation to manage, and the oil-palm
the household survey data from 10 villages. company has actively encouraged neighboring
We calculated the proportion of cash income villages to cultivate oil palm in the so-called
earned by each household from agroforestry ‘‘plasma’’ area (the area around the ‘‘nucleus
products (including rattan, honey, fruit, game, estate’’).
fish, other forest products, and rice) for each The best-fit spatial error model suggests that
sample village. Households that earn more than unexplained factors between neighboring vil-
half of their cash income from agroforestry lages are strongly correlated. Table 3 shows
activities are classified as agroforestry-depen- that better road access increases EDI. The pro-
dent households. Village-level agroforestry vincial road gives increased opportunities for
dependence is then measured as the proportion income and employment, presumably by link-
of village households that are agroforestry ing villages to the broad network of markets
dependent. The income portfolio for three se- for different products and labor, improving
lected villages is described to illustrate the access to information and facilitating in- and
breadth of income-earning activities. out-migration. Villages connected by provincial
roads show high values of economic diversity.
District roads have a similar but less pro-
5. RESULTS AND DISCUSSION nounced effect. This was expected, as accessibil-
ity, in terms of the quality of the road and the
(a) Factors associated with village EDI places it connects, is less than that provided
by the provincial road. The finding that roads
Spatial statistical analysis was run on the are associated with a higher village economic
subset of variables that gives the best-fit spatial diversity is consistent with findings from Latin
error model (Table 3), with EDI as the depen- America in which access to road infrastructure
dent variable. Figure 2 shows the map of and closeness to towns are shown to be robust
EDI. The distances to plantations (timber, rub- determinants of rural nonfarm employment
ber, and oil palm), to mines and to transmi- and incomes (Reardon et al., 2001).
1426 WORLD DEVELOPMENT

Figure 2. Village economic diversity index map.

Logging/plantation roads also lead to a higher proximity to logging companies does not affect
EDI, but the effect is smaller than that of other EDI. This implies that these projects do not
road classes. This class of road typically pro- provide sufficient economic opportunities for
vides access to smaller markets, both for prod- local people to adopt them as a main source
ucts and labor, and mainly to markets directly of income. This is consistent with reports that
related to the activities of the resource extrac- in Borneo most labor for logging companies
tion company and its employees. Mining roads was hired from outside the area (Brookfield
do not show statistically significant relation- et al., 1995). Logging roads are not perfectly
ships with EDI, which can be explained by correlated with logging companies because the
the fact that local people were not permitted roads may remain after the company (and asso-
to use them during the study period. ciated economic activities) has left the area.
While logging/plantation roads are associ- The EDI correlates positively with deforesta-
ated with a higher village economic diversity, tion (including conversion to agriculture and to
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1427

Figure 3. Village development index map.

other activities) during the 1992–96 period. (b) Well-being at the village level (VDI)
This may indicate that some activities that re-
sulted in deforestation also provided nonfarm The map of VDI is presented in Figure 3,
income opportunities for local people. and the best-fit spatial error model of VDI is
1428 WORLD DEVELOPMENT

Table 4. Spatial error model of VDIa when the village has sufficient suitable land.
Variable Coefficient z-Value As discussed earlier, opportunities for local
livelihoods to improve in the ‘‘poverty elimi-
Economic diversity index 0.19 3.70** nation’’ sense (see Sunderlin et al., this vol-
Agrosuitability 0.27 7.86**
ume) based on agriculture and forestry have
Land-use intensity 0.15 2.39*
been limited, but appropriate interventions at
Forest 92 0.53 6.79**
the project and policy level might help (e.g.,
Population 8.87E
05 5.39**
Scherr et al., 2002).
Time to district capital
0.03
1.13
Given the fact that the area has a low popu-
Time to subdistrict capital 0.01 0.38
Lambda
0.44
3.30**
lation density overall (mean = 13.95 persons/
a
sq. km), it is not surprising to find that the pop-
With maximum likelihood estimation, number of ulation has a highly significant positive coeffi-
observation = 73, R2 = 0.98, LIK = 63.65, AIC = cient. Small isolated villages are unable to

113.31.
*
Significance level of 0.05.
generate economic dynamism.
**
Significance level of 0.01.
(c) Agroforest dependencies/role of forests
in village samples
presented in Table 4. A higher village economic
diversity leads to a higher VDI. In the case For the 10 village samples (Figure 1), the
study area, increases in EDI generally mean EDI (from the census data) tends to increase
that a higher proportion of households in a vil- and then decrease as agroforest dependence
lage rely on the nonagricultural sector for their (from aggregated household survey data)
main source of income. The proportion of land decreases (R2 = 0.84). On the other hand, the
cleared by smallholders from land that was a proportion of village area initially covered by
young secondary forest (relative to the amount forests seems to be the major factor determin-
cleared from mature forests) is positively corre- ing the proportion of households that are
lated with VDI. This is mostly related to the primarily dependent on forest products and
availabilities of fertilizer and herbicide and land agroforestry with a very strong, quadratic
pressure, which are both driven by market and relationship (R2 = 0.93) (Figure 4).
economic opportunities. Agroforest dependence shows a weak qua-
A larger extent of land suitable for agrofor- dratic relationship with VDI. High agroforest
estry practices leads to a higher VDI. That is, dependence is associated with a relatively high
the traditional system provides a relatively VDI. The VDI declines as agroforest depen-
good living if there is an adequate resource dence declines, but then they increase together,
base. A large proportion of village land with with the least agroforest-dependent villages
forest cover, measured at the beginning of the having the highest VDI. The two villages with
study period, has an even larger positive coeffi- high agroforest dependence and a high well-
cient, showing that the forest has contributed to being index have large areas of good quality
higher welfare. forests, but there are obviously other variables
While, on the one hand, a higher initial involved in determining people’s well-being at
forest cover and land quality for agroforestry the village level (Figure 5).
are associated with higher welfare, on the
other hand a higher EDI (shift toward less (d) Agroforest dependence case studies
agroforestry-dependent communities) and
higher land use intensities are also associated Modang, Muser, and Legai, the three villages
with a higher welfare. The role of forest in in the lower end of the agroforest dependence
the EDI–VDI context is intriguing. While axis have a low forest cover, but a high EDI
deforestation correlates positively with EDI rank. At the other extreme, Besiq and Rantau
and with VDI, a larger proportion of village Layung villages have a high forest cover and
area initially covered by forests also increases a high dependence on agroforestry and a low
VDI. This shows two different possible devel- EDI. Sample villages with an intermediate level
opment paths. The EDI may increase as the of agroforest dependence and forest cover di-
forest cover declines, especially where suitable verge in their VDIs such as Kendisiq, Muara
land for agroforestry is limited. Or agrofor- Siram, and Muara Lambakan have a low
estry can serve as the main source of liveli- VDI while Kasungai and Muara Nilik have a
hood while maintaining a high forest cover higher VDI.
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1429

1.2

R2 = 0.93
Besiq
1.0
R Layung

Agro-forest dependencies .8

M Siram M Lambakan

.6 Kasungai
M Nilik
Kendisiq

Muser
.4
Modang Legai

.2
.4 .5 .6 .7 .8 .9 1.0
Proportion of forest cover in 1992

Figure 4. Scatterplot of proportion of forest cover in 1992 and agroforest dependencies.

.70
Kasungai
.68 Legai Besiq
Muser M Nilik
Modang
.66
R Layung
Village Development Index

.64

.62

.60

.58
Kendisiq M Siram
.56 R2=0.29

M Lambakan
.54

.52
.2 .4 .6 .8 1.0 1.2
Agro-forest dependencies

Figure 5. Scatterplot of agroforest dependencies and village development index.

We can illustrate this with the examples of district capital. We observed some intensified
the villages of Legai, Muara Lambakan, and agricultural practices here with the use of herbi-
Besiq (Figure 6). Legai represents a village with cide, pesticide, and short fallow periods. Arti-
a high EDI, low forest cover, and a high VDI. sanal gold mining is the primary source of
It is located in the Paser district, very close to income, followed by agroforestry (rattan, cof-
the trans-Kalimantan highway and to the sub- fee, rubber, and other crops), nonfarm income
1430 WORLD DEVELOPMENT

Figure 6. Mean household income portfolio in three villages.

(trading, wage labor, services), and forest prod- but with a history of several episodes of logging
ucts (wild rattan, logs). by different logging companies. Most of the vil-
Muara Lambakan, also in the Paser district, lage area is unsuitable for intensive farming due
is a remote village with no industries nearby, to thin topsoil and a rough terrain. This village
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1431

represents the group with a moderate agrofor- resource exploitation projects, but also transmi-
est dependence, moderate forest cover, low gration) created different opportunities and
EDI, and a low VDI. The largest proportion constraints for local people.
of income is from forest-based activities and The analysis shows that large-scale commer-
the largest part of that comes from game and cial logging did not benefit the local people.
fish. Income from farming is slightly lower than Mines, oil-palm plantations, and transmigra-
that from forest. Off-farm income as rattan tion projects created some new markets for lo-
workers in a few large gardens in the same vil- cal products that replace traditional activities,
lage comprises a substantial proportion of but these projects do not show trickle-down ef-
income. Nonfarm income contributes almost fects outside the agricultural sector except
20% of the total cash income. through improvement in the transportation
Besiq village, in West Kutai, has a large area network. These projects often focus on one
of forest, high agroforest dependence, low EDI, main economic activity such that in the areas
and a relatively high VDI. The area was re- around palm-oil estates and transmigration
mote, with no road access 15 and river access areas, there is a low economic diversity.
limited during the dry season. Forest products, Roads are very important to village eco-
including rattan, rubber, game, fish, and honey, nomic diversity, with a higher quality, better-
provide the main source of cash income. Rattan connected roads being associated with a higher
agroforest and domestic animals contribute a diversity. Mining roads are the exception.
small proportion of total cash income, followed They do not result in local economic diversity
by nonfarm income, which adds very little to because they are not open for public access in
the total cash income. the study area. Forest conversion to other
uses is generally associated with increased vil-
lage economic diversity, again with both
6. CONCLUSIONS AND IMPLICATIONS positive (pull) and negative (push) factors
involved.
The indices developed and the analyses Generally speaking, well-being, as measured
conducted have proven useful to explain rela- by education, health, and assets, increases with
tionships between various driving forces and more village economic diversity. However, vil-
village welfare and economic diversity. The ap- lage economic diversity is not the single most
proach could also be used as a tool for sce- important determinant of well-being. Higher
nario-based land use planning at the district levels of forest resources and suitable land for
level. Poor data availability was the main con- agroforestry are also associated with higher
straint we faced in using this approach. With welfare. Relatively remote, well endowed forest
actual village boundaries rather than approxi- villages with limited economic alternatives
mations, the analysis could have been more show a high well-being relative to other villages
accurate. Cloud-free satellite images would give in the area being studied. A good forest endow-
a more comprehensive data set to work with. ment allows people to live well at or near the
Also, more frequently collected secondary data subsistence level. But opportunities for forest-
from government statistics would allow the use based poverty elimination, in the sense of lifting
of time-series analysis in conjunction with the people safely out of poverty, have been limited.
spatial analysis to cover both temporal and The worst-off villages are those with poor re-
spatial dynamics. source endowments and limited alternative in-
Under the conditions prevailing during the come-earning opportunities.
study period, the local people were highly dis- But the future need not mirror the past. Rich
advantaged. In general, health service and edu- resource endowments could be used in other
cation quality were generally low, and the local ways, with more benefits accruing to local
people were not involved in major resource people. Forest and agricultural livelihoods can
extraction decisions. They had little opportu- potentially support relatively high levels of
nity to benefit from the rich natural resources welfare. Improving local people’s access to
in the areas where they lived. This may change resources in their vicinity and their capacity
with more decentralized governance (since to transform them is critical for enabling them
2001), but it is too early to assess the impact to attain better health, education, and other
of that. As forest resources have been depleted, well-being improvements. More investment in
people have turned to alternative activities. A infrastructure, markets, and other factors that
variety of externally controlled projects (mainly support village economic diversity should be
1432 WORLD DEVELOPMENT

encouraged. In areas where land suitable for increasing economic opportunities is even more
agriculture and forestry are less available, important.

NOTES

1. Logging roads make areas of forest/land more easily 5. The main rubber producers in Indonesia are small-
accessible and reduce transportation and marketing holders. There have been attempts to support the sector
costs for local products (Dove, 1996). Logging camps by providing planting materials, fertilizer, pesticides,
create temporary markets for products and labor, equipment, and land titles, as well as cash payments to
though it is a common practice in the study area to cover subsistence expenses during the period before the
hire workers from outside areas. Logging activities also trees mature. Two important projects in the vicinity of
compete for forest resources with local people; typically, the study area were the Tree Crop Smallholder Devel-
local people have been prohibited from entering conces- opment Project funded by the World Bank (IBRD),
sion areas. started in 1974, and by the Asian Development Bank,
started in 1992 (Budiman, 1999).
2. State policy to assign ‘‘degraded lands’’ for large-
scale timber plantations often targets (deliberately or 6. Transmigrants were typically settled in purpose-built
not) managed forest gardens and secondary forests that communities, often on lands claimed by local people.
are part of local peoples’ agroforestry systems, resulting This resulted in competition and some conflict for forest
in a direct competition for land and, often, conflict. resources and land between transmigrants and the local
These plantations were usually operated in conjunction people. As part of these projects, new roads, as well as
with logging concessions, and the plantation concessions housing, schools, and health facilities were built. Some
were used as ‘‘licenses’’ to clear-cut in the guise of transmigration projects in the study area were developed
clearing for planting (see, e.g., Manurung, 2001). Ben- in conjunction with timber plantations, oil-palm plan-
efits for local people are similar to those available from tations, and processing factories. The projects created
logging concessions. Roads are built, which leads to the new markets for various products and for labor. Also,
employment of some villagers mainly as nonskilled there is usually some technology transfer as in-migrants
workers, and temporary markets are created for local bring different ideas and approaches that may be
produce (Brookfield et al., 1995). In our study area, adopted by local people.
timber plantations were established in conjunction with
the transmigration program. 7. The Trans-Kalimantan Highway, the most signifi-
cant road development in the area, was built in the mid-
3. The study area has rich coal deposits. Mining in 1980s by the provincial government to connect major
Indonesia has been strictly controlled by the central cities in Kalimantan. Many of the transmigration roads
government, with a top-down approach that has been were built around the same time. Project-related roads
criticized for its negative impacts on the local people. have been built (and abandoned) as and when needed for
Conflicts between mining companies and local people companies’ operational purposes such that from the
are common (Down to Earth, 2001; World Bank, villagers’ point of view, they only provide a temporary
2001). Mining competes for access to forest lands, and access to resources and sometimes to the market. The
prohibitions against trespassing are strictly enforced. history of road building in the area is not well
While mining concessions may be smaller than forest documented.
concessions, the forest cannot be expected to recover
after the end of mining activities. Roads built by 8. Local village heads reported that boundaries were
mining companies are better made and better defined by agreement, using natural landmarks close to a
maintained than logging roads, and markets for midpoint that fairly divided resources between two
labor, supplies, and services are larger and more neighboring villages.
long-lasting than those created by logging and timber
plantations. 9. Village Potential data include human capital
(demography, workforce), natural captial (forest, mine,
4. As with timber plantations, development of oil-palm wetlands, etc.), social captial (institutions, networks),
estates implies a direct competition for land with small- and produced capital (health facilities, school buildings,
scale farmers (Manurung, 2000). Positive contributions infrastructure, mills, factories, etc.). These data were
to the area include road construction, development of collected by village administrative officials and then
processing facilities, and creations of markets. compiled by the Central Board of Statistics (BPS)
 VILLAGE ECONOMIC OPPORTUNITY, FOREST DEPENDENCE, AND RURAL DEPENDENCE 1433

(Suhariyanto, K. pers. com. 2004). The data were cross- frequencies in most of those eight categories. For our
checked by our enumerators with several key informants study area, ‘‘wage labor’’ includes labor in construction,
and were found to be consistent. rattan harvesting, oil-palm harvesting, rubber tapping,
sawmilling, logging, etc.
10. The Shannon–Weaver index measures diversity in a
way very similar to that of the Simpson index 13. The HDI has been criticized on several counts. In
(Baumgärtner, 2002). Using the Shannon–Weaver index, particular, Hicks (1997) has pointed out the problem of
our data are normally distributed, while computation distributional inequality within countries, which is
using the Simpson index shows a tendency to depart overlooked in the HDI. We share Hicks’s concern about
from the normal distribution, that is, with insensitivity in distribution and the need for disaggregating the popu-
the lower diversity. lation. However, as we are using much smaller units of
analysis (small villages as opposed to national popula-
11. This index is very different from the diversity index tions), inequalities are much less pronounced. Our VDI
proposed by Chang (1997), which measures diversity of and the HDI are not comparable as they are measured at
household income portfolios and its contribution to the different scales with different data for different purposes.
total, and from diversification indices presented in
Wagner and Deller (1998), which measure economic 14. We use the number of motorcycles and motor boats
diversity and industrial linkages at the state level in the per capita as a proxy for wealth. Barham, Takasaki, and
United States. This index does not capture diversifica- Coomes (1999) argue that the wealth of households in
tion at the household level, but rather at the aggregate tropical rainforest areas is best measured by assets that
level, in this case, the village. Therefore, our EDI does are related to livelihood strategies. In their study, key
not link explicitly with the household income portfolio assets are land, productive capital, and nonproductive
or household economic diversification that is often assets that together provide a basis for producing
discussed in the livelihood literature (e.g., Ellis, 2000; subsistence and cash income, a buffer against bad times,
Reardon et al., 2001). Using the index as a measure of and a better life style. In our study area, boats and
economic diversity one assumes that improved oppor- motorcycles perform all three functions. People rely on
tunities generated by external developments and the private transport to travel to forests and farms for their
capabilities of people to take advantage of those productive activities, to transport their products to the
opportunities reduce the dominance of agriculture in market and at times to generate income through
people’s livelihoods at the village level. providing transport services. People will buy an extra
motorcycle or a boat in good times and sell during bad
12. The BPS’s 10 income categories are agriculture times, using these goods, in effect, as a ‘‘savings
(fishery, crop, cattle ranching, fish farming, forestry), account.’’ Motorcycles and boats also have a role in
mining (including sand and rock), wage earning, crafts ‘‘conspicuous consumption,’’ as a common way to
and home industries, electricity and water, construction, demonstrate wealth. There are other assets that we
trading, transport, finance, services, and others. Except think should be important in measuring wealth (e.g.,
for agriculture, mining and wage earning, the last eight chainsaws), but these are not included in the data
categories were aggregated into one class called ‘‘ser- collected by BPS.
vice’’ for the purpose of our analysis. This was done
because the data of our study area showed low 15. A road was opened after the study period.

REFERENCES

Angelsen, A., & Wunder, S. (2003). Exploring the Baumgärtner, S. (2002). Measuring the diversity of what?
poverty-forest link: Key concepts, issues and research And for what purpose? A conceptual comparison
implications. Occasional Paper No. 40. Center for of ecological and economic measures of biodiversity.
International Forestry Research, Bogor. Bigman, D., & Fofack, H. (2000). Introduction and
Anselin, L. (1999). SpaceStat Version 1.80 User Guide, overview. In D. Bigman & H. Fofack (Eds.),
Urbana Champaign. Geographical targeting for poverty alleviation: Meth-
Arnold, M. (2001). Forestry, poverty and aid. Occasional odology and applications. Washington, DC: The
Paper No. 33(E). Center for International Forestry World Bank.
Research, Bogor. Brookfield, H., Potter, L., & Byron, Y. (1995). In place
Ashley, C., & Maxell, S. (2002). Rethinking rural of the forest: Environmental and socioeconomic trans-
development. Development Policy Review, 19(4). formation in Borneo and the Eastern Malay Peninsula.
Barham, B. L., Takasaki, Y., & Coomes, O. T. (1999). Tokyo: The United Nations University.
Rain forest livelihoods: income generation, house- Budiman, A. F. S. (1999). Partnership of rubber
hold wealth and forest use. Unasylva, 198. producer association with smallholders. Paper
1434 WORLD DEVELOPMENT

presented at the Agricultural Producer Organizations: Seminar Nasional: Hutan dan Lingkungan, Tema:
Their Contribution to Rural Capacity Building and ‘‘Era Menanam: Membangun Kelestarian Hutan dan
Poverty Reduction, Washington, DC. Lingkungan’’. Auditorium Rektorat IPB, Kampus
Cooke, M. F. (2002). Vulnerability, control and oil palm Darmaga, Bogor.
in Sarawak: Globalization and a new era? Develop- Nunnaly, J. C. (1978). Psychometric theory. New York:
ment and Change, 33(2), 189–211. McGraw-Hill.
Dove, M. R. (1993). A revisionist view of tropical Padoch, C., & Peluso, N. L. (Eds.). (1996). Borneo in
deforestation and development. Environment Conser- transition: People, forests, conservation, and develop-
vation, 20. ment. New York: Oxford University Press.
Dove, M. R. (1996). So far from power, so near to the Potter, L., & Lee, J. (1998). Tree planting in Indonesia:
forest: A structural analysis of gain and blame in Trends, impacts and directions. Occasional Paper,
tropical forest development. In C. Padoch & N. L. 18, 1–76.
Peluso (Eds.), Borneo in transition: People, forest, Reardon, T., Berdegué, J. A., & Escobar, G. (2001).
conservation, and development (pp. 41–58). Kuala Rural nonfarm employment and incomes in Latin
Lumpur: Oxford University Press. America: Overview and policy implications. World
Dove, M. R., & Kammen, D. M. (2001). Vernacular Development, 29(3), 395–409.
models of development: An analysis of Indonesia Sardjono, M. A., & Samsoedin, I. (2001). Traditional
under the New Order. World Development, 29(4), knowledge and practice of biodiversity conservation:
619–639. The Benuaq Dayak community of East Kalimantan
Down to Earth (2001). Communities and companies. Indonesia. In C. J. P. Colfer & Y. Byron (Eds.),
Down to Earth Newsletter, 50(August). People managing forests: The links between human
Edmunds, D., & Wollenberg, E. (Eds.). (2003). Local well-being and sustainability. Resources for the
forest management. London: Earthscan Publications. Future and Center for International Forestry Re-
Ellis, F. (2000). Rural livelihoods and diversity in devel- search (CIFOR).
oping countries. New York: Oxford University Press. Scherr, S. J., White, A., & Kaimowitz, D. (2002).
Inoue, M. (2000). Mechanism of changes in the Ken- Making markets work for forest communities. Wash-
yah’s swidden system: Explanation in terms of ington, DC: Forest Trends and CIFOR.
agricultural intensification theory. In E. Guhardja, Sunderlin, W. D., Angelsen, A., Belcher, B., Burgers, P.,
M. Fatawi, M. Sutisna, T. Mori, & S. Ohta (Eds.), Nasi, R., Santoso, L., et al. (this volume). Liveli-
Rainforest ecosystems of East Kalimantan: El Niño, hoods, forests and conservation in developing coun-
drought, fire and human impacts. Tokyo: Springer- tries: An overview. World Development.
Verlag. Vandergeest, P., & Peluso, N. (1995). Territorializa-
Lahjie, A. M. (1996). Traditional land use and Kenyah tion and state power in Thailand. Theory and
Dayak farming systems in East Kalimantan. In C. Society, 24.
Padoch & N. L. Peluso (Eds.), Borneo in transition: World Bank (2001). Indonesia: Environment and natural
People, forests, conservation, and development resource management in a time of transition. Wash-
(pp. 150–161). New York: Oxford University Press. ington, DC: The World Bank.
Manurung, T. (2000). Pembangunan perkebunan kelapa World Bank (2002). A sourcebook for poverty reduction
sawit di Indonesia: Ancaman terhadap hutan alam. strategies: vol. 1. Washington DC: World Bank.
Warta Fahutan Online, 1–6. Wunder, S. (2001). Poverty alleviation and tropical
Manurung, T. (2001). Potret pembangunan hutan tana- forests: What scope for synergies. World Develop-
man industri di Indonesia. Paper presented at the ment, 29(11), 1817–1833.