ArticlePDF Available

Explaining Success on the Commons: Community Forest Governance in the Indian Himalaya

Authors:

Abstract and Figures

In the past two decades, scholarship on resource use and management has emphasized the key role of institutions, communities, and socio-economic factors. Although much of this writing acknowledges the importance of a large number of different causal variables and processes, knowledge about the magnitude, relative contribution, and even direction of influence of different causal processes on resource management outcomes is still poor at best. This paper addresses existing gaps in theory and knowledge by conducting a context-sensitive statistical analysis of 95 cases of decentralized, community-based, forest governance in Himachal Pradesh, and showing how a range of causal influences shape forest conditions in diverse ecological and institutional settings in the Indian Himalaya. In focusing attention on a large number of cases, but drawing on findings from case studies to motivate our analysis and choice of causal influences, our study seeks to combine the strengths of single case-oriented approaches and larger-N studies, and thereby contributes to a more thorough understanding of effective resource governance.
Content may be subject to copyright.
Explaining Success on the Commons: Community
Forest Governance in the Indian Himalaya
ARUN AGRAWAL
University of Michigan, Ann Arbor, USA
and
ASHWINI CHHATRE
*
Duke University, Durham, USA
Summary. In the past two decades, scholarship on resource use and management has empha-
sized the key role of institutions, communities, and socio-economic factors. Although much of this
writing acknowledges the importance of a large number of different causal variables and processes,
knowledge about the magnitude, relative contribution, and even direction of influence of different
causal processes on resource management outcomes is still poor at best. This paper addresses exist-
ing gaps in theory and knowledge by conducting a context-sensitive statistical analysis of 95 cases
of decentralized, community-based, forest governance in Himachal Pradesh, and showing how a
range of causal influences shape forest conditions in diverse ecological and institutional settings
in the Indian Himalaya. In focusing attention on a large number of cases, but drawing on findings
from case studies to motivate our analysis and choice of causal influences, our study seeks to com-
bine the strengths of single case-oriented approaches and larger-Nstudies, and thereby contributes
to a more thorough understanding of effective resource governance.
Ó2005 Published by Elsevier Ltd.
Key words — resource governance, community-based conservation, local institutions, monitoring
and enforcement, gender and environment, South Asia, Himalayas
1. INTRODUCTION
In the past two decades, scholarship on re-
source use and management has emphasized
the key role of institutions, communities, and
socio-economic factors. Literatures on com-
mon property (Ostrom, 1990), political ecology
(Neumann, 1998), rural sociology (Goldman &
Schurman, 2000), resource economics (Lise,
2000), and environmental politics (Gibson &
Marks, 1995) examine what explains variations
in resource governance and conditions. Even
those who downplay certain kinds of insti-
tutions are willing to acknowledge the im-
portance of communal institutional forms
(Campbell et al., 2001; Virtanen, 2002; cf.
Rangan, 1997; Ribot & Peluso, 2003). Much
of this writing emphasizes the importance of
many different causal variables and processes.
*We would like to thank our research team for the
immense effort and time they invested in the data col-
lection for the research project on which the paper is
based. Special thanks are due to Satya Prasanna,
Akshay Jasrotia, and Vishal Sharma. High levels of
cooperation in the villages we visited made the field
research by the team so successful. We appreciate the
difficulties and inconvenience to which our respondents
were put many times by our visits. Comments, sugges-
tions, and constructive criticisms from Krister Anders-
son, Mark Baker, Daniel Brown, Graham Marshall,
Lisa Naughton-Treves, and three anonymous reviewers
are gratefully acknowledged. We especially thank Elinor
Ostrom and Kent Redford for their careful point-by-
point engagement with the arguments in the paper.
Finally, we are grateful for the research support pro-
vided by the National Science Foundation under its
grant# SBR 9905443, and to the Ford Foundation for
support to the International Forestry Resources and
Institutions Program through its grant #950-1160-2.
Final revision accepted: July 22, 2005.
World Development Vol. 34, No. 1, pp. 149–166, 2006
Ó2005 Published by Elsevier Ltd.
Printed in Great Britain
0305-750X/$ - see front matter
doi:10.1016/j.worlddev.2005.07.013
www.elsevier.com/locate/worlddev
149
But existing knowledge about the magnitude,
relative contribution, and even the direction
of influence of different causal processes re-
mains poor at best (Agrawal, 2001).
Relevant knowledge about the magnitude
and relative importance of different causal vari-
ables is still relatively poor in part because of
the analytical approach that dominates studies
of resource management and conservation.
Much of this literature relies on a case study
or comparative case study approach (Baland
& Platteau, 1996; Berkes, 1989, 2004; NRC,
1986). Statistical analyses are far more rare,
especially those based on data from the local
level. Case studies can be remarkably effective
in providing in-depth knowledge of specific
conjunctures and highlighting the importance
of causal processes significant in those conjunc-
tures. They can potentially also be invaluable
tools to identify the direction of causal forces
and specify the contextual features that lend a
particular cause its leverage over outcomes
(Ragin, 1997). But they are less effective when
the objective is to assess the magnitude or rela-
tive importance of different causal factors
(Goldthorpe, 1997). This paper adds to existing
theory and knowledge by examining how a
range of causal influences shape forest con-
ditions in diverse ecological and institutional
settings in Himachal Pradesh in the Indian
Himalaya. Our study is based on a context-
sensitive statistical analysis of 95 instances of
decentralized, community-based, institutional-
ized governance of forest resources. We use sta-
tistical techniques to probe potential causal
mechanisms, but also draw on findings from
case studies and intensive fieldwork to motivate
our analysis, choice of causal influences, and
interpretation of regression results. Our study
thus combines the strengths of single case-
oriented approaches and larger-Nstudies, and
contributes to a more thorough understanding
of effective resource governance.
The ensuing discussion considers five classes
of causal influences: biophysical, demographic,
economic, institutional, and socio-political.
Existing studies have recognized each of these
as being instrumental in influencing resource
governance outcomes (Alvarez & Naughton-
Treves, 2003, p. 269; Brown & Pearce, 1994;
Rocheleau & Edmunds, 1997). For each cate-
gory of influence, we examine variables that
have been highlighted in the literature and
which appear relevant to the Himachal Pradesh
context as discussed below. In general terms,
our paper engages discussions related to com-
mon property and analyses of local governance
of forest resources. More specifically, it high-
lights two important issues. The first relates di-
rectly to the many different causal influences
emphasized in the case study literature on for-
est governance. Our analysis shows that within
each class of causal influence, there are at least
some variables that exercise statistically signifi-
cant effects on outcomes. Thus, much case
study literature on the commons is on target
in the specific types of causal influences that it
has highlighted. However, it may also be off
target in the causal influences that particular
case studies have ignored. Empirical measures
of the theoretical constructs that serve as inde-
pendent variables in causal analyses are often
correlated, even if only to some degree. It is rea-
sonable to infer, therefore, that empirical anal-
yses that focus only on a restricted set of causal
influences likely inflate the significance of the
variables they consider even as they ignore the
relevance of excluded variables. Studies that
focus on single cases are especially prone to
such biases. Our position does not imply that
systematic studies must always analyze the
effects of scores of independent variables.
Rather, we make a plea for the inclusion of
variables known to be theoretically among the
most important in the social and ecological
contexts being studied.
The second finding, a direct corollary of the
methodological point above, concerns the
important causal role of biophysical variables
in explaining local resource condition. In the
analysis that follows, biophysical factors prove
to be critically important in explaining varia-
tion in the dependent variable: forest condition.
The importance of this finding for the literature
on common property lies in the fact that the
general context of our study shares many fea-
tures with the general context of studies of
common property: high ecological and social
variability. Indeed, a number of major scholars
have explicitly identified high levels of variation
in biophysical factors, and therefore resource
flows, as the source of pressures for self-organi-
zation and local cooperation (Scott, 1976;
Wade, 1994; see also Baland & Platteau, 1996;
Berkes, 1989; NRC, 1986; Ostrom, 1990). Our
results suggest that when scholars of commons
seek to explain resource governance-related
outcomes in ecologically and socio-culturally
variable contexts, they need to attend more
carefully to the influence of biophysical factors
on socio-cultural conditions and resource gov-
ernance outcomes.
150 WORLD DEVELOPMENT
In the next section of the paper, we provide a
brief description of the Himachali context,
especially highlighting aspects relevant to the
ensuing analysis. We then describe our meth-
ods, explain the variables used in the statistical
analysis, and outline the model to be tested.
After discussing the results of the analysis and
the theoretical issues concerning findings re-
lated to the different variables, we conclude
with a discussion of the scope of our paper,
and implications for future studies of the com-
mons and local resource governance.
2. ECOLOGICAL AND INSTITUTIONAL
ASPECTS OF FOREST GOVERNANCE IN
HIMACHAL PRADESH
The state of Himachal Pradesh is ecologically
highly diverse owing to distinct climatic and
physiographic factors. Nestled in the western
Himalaya, its elevation varies from 350 to
7,000 m above mean sea level. With nearly
3,200 identified plant species, forests in Hima-
chal constitute its most important biological
wealth. More than two-thirds of the state is
administratively classified as forest (37,600 sq.
km). However, only about 25% of this area is
dense forest, with about a third being perma-
nent snow, and approximately 10% classified
as protected areas.
Steep altitudinal gradients shape variations in
forest characteristics. Beginning with sub-trop-
ical scrub forests in the foothills, species com-
position and biological diversity change with
elevation (see Figure 1). The lower hills (below
900 m) are a mosaic of dry scrub, and dry and
moist deciduous forests. They are interspersed
with stands of Himalayan Pine (Pinus rox-
burghii), dominant between 1,000 and
2,000 m. Oak (Quercus spp.) and deodar
(Cedrus deodara) make their appearance at
1,800 m, and are present in single or mixed
stands with Rhododendron (Rhododendron
spp.) up to a height of 3,000 m. Beyond
3,000 m, Silver Fir (Abies pindrow) and Spruce
(Picea smithiana) are dominant in single or
mixed stands. Mixed stands of Silver Fir and
Birch (Betula alnoides) typically characterize
the transition from mixed forests to alpine
meadows. The tree line ends at approximately
3,700 m. Studies of high levels of ecological
variability in factors such as temperature and
moisture have pointed to their relationship with
levels of plant biodiversity elsewhere (Brock-
way, 1998). Himachal Pradesh is no exception.
Altitudinal effects on vegetation played an
important role in our research design and sam-
ple selection.
The population of Himachal Pradesh was
6.01 million in 2001. The average population
density (108 persons per sq. km) varies from 2
persons in the cold desert district of Lahaul
and Spiti to 330 persons in Hamirpur district.
The urban population is concentrated in 59
towns and smaller settlements. The overwhelm-
ing majority of the population is rural (more
than 90%), living in more than 16,000 villages
(DOP, 1997). Although the net sown area is less
than 15%, agriculture is the occupation of
nearly two thirds of the population (DOP,
1997). Forests are critical to hill agriculture,
and also contribute directly to livelihoods. Over
the last century, they have provided land for
distribution to the landless, been the sites of
extensive road construction and infrastructure
development, and served as important commer-
cial resources by providing resin for turpentine
(Pinus roxburghii), raw materials for paper and
pulp (bamboo (Bambusa bambos) and bhabhar
grass (Euloliopsis binata)), and wood packing
cases for Himachal Pradesh’s important fruit
industry.
In this context of high population density
and competing uses, a number of different insti-
tutional mechanisms to secure the formal par-
ticipation of local residents are in evidence
across the forested landscape in Himachal Pra-
desh. Such institutional arrangements include
self-initiated systems, cooperatives, corporate
clan-owned forests, sacred forests, and coman-
aged forests. Through these arrangements,
communities in Himachal Pradesh govern the
full range of different forest types found in the
state.
3. FORESTS AND INSTITUTIONS: NOTES
FOR AN ANALYTICAL FRAMEWORK
Classifications such as open access, private,
community, and state ownership are too coarse
grained according to scholars of the commons
and property rights to convey variations in
institutional arrangements at the community
level (McKean, 1992). Not only are these clas-
sifications coarse, but they are also inadequate
as causal explanations of outcomes: it is insuf-
ficient to say that resources are better governed
simply because they are under common or pri-
vate property regimes, or in a bad condition be-
cause they are managed by governments (Dietz,
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 151
Ostrom, & Stern, 2003). Rather than replace a
coarse-grained set of categorical variables with
finer-grained categorical variables such as
cooperative, comanagement, or sacred, we base
our analysis on underlying institutional vari-
ables that are common to the different commu-
nity-level governance regimes, and which are
more directly related to causal processes.
Thus, for our analytical framework, we draw
upon property rights and new-institutionalist
theories (McKean, 1992; Ostrom, 1990, 2003).
Since the theoretical literature on the subject
is well developed, we do no more than summa-
rize the most important elements of the argu-
ment, and focus primarily on the empirical
analysis. According to institutionalists inter-
ested in resource governance, forest conditions
are a function of a large number of factors of
which institutional variables are an extremely
important set. Resource characteristics and
biophysical variables form the context within
which socio-political and economic characteris-
tics of users and institutional variables shape
resource management outcomes. Institutional
variables include those related to existence, rep-
resentation of users, enforcement of rules, and
relationship with external authorities (see Agra-
wal, 2001 for a review).
The specific causal mechanisms at work in
the different categories of institutions are a re-
sult of the varying configurations of underlying
institutional factors. To identify the effects of
institutional variables on resource conditions,
however, it is necessary to take into account
Figure 1. Vegetation cover and districts in Himachal Pradesh.
152 WORLD DEVELOPMENT
the role of the biophysical, economic, social,
political, and demographic context as well.
Although institutional theories recognize the
importance of these contextual factors (Os-
trom, 2005), it is a rare study that explicitly
incorporates variables representing all these
classes of influences into the analysis. It is pre-
cisely this gap in the literature that our paper
seeks to fill.
4. METHODS, STUDIED VARIABLES,
DATA, AND STATISTICAL MODEL
We collected data in 2001 in 205 forests
across Himachal Pradesh. The present article
is based on the subset of 95 cases where village
communities and groups of individuals in a vil-
lage community jointly manage forests. The
remaining 110 cases of forests are managed by
the forest or the revenue department. We se-
lected villages for data collection across the alti-
tudinal gradient in the state, sampling equally
from the lower hills (<900 m above mean sea
level), middle hills (between 900 and 1,800 m),
and high hills (>1,800 m). Within each eleva-
tion class, we selected cases to represent differ-
ent institutional regimes, in proportion to
their distribution across the three elevation
classes. This strategy of case selection, while
not random, ensures that all major types of for-
ests and institutional regimes are represented in
our sample. Further, this sample selection strat-
egy also ensures that the cases are not picked
on the basis of variations in the value of the
dependent variable. We should note that it is
near-impossible to identify a fully random sam-
ple for local institutional or forest types owing
to the non-existence of any comprehensive lists
that contain the relevant information.
In collecting data, we were especially atten-
tive to a suite of biophysical, economic, demo-
graphic, institutional, and socio-political
variables as described below. The questions
we used during our field work come from the
set of data collection instruments developed
by the International Forestry Resources and
Institutions (IFRI) Program at Indiana Univer-
sity (Poteete & Ostrom, 2004; the full set of
instruments can be obtained from the Work-
shop in Political Theory and Policy Analysis
at Indiana University). We trained our field
investigators regarding the meaning of different
questions, our analytical framework, and in
conducting individual and group interviews in
villages. Our unit of analysis is the management
unit at the community level. Responses to ques-
tions were triangulated by multiple interviews
with different individuals and groups within a
community. These groups and individuals in-
cluded upper and lower caste men and women,
decision makers, and where relevant, guards
and other office-holding individuals in the local
community institutions.
The dependent variable in our analysis is
‘‘Forest Condition.’’ It is measured by an
index, based on group responses for the condi-
tion of the forest from (1) upper caste men, (2)
upper caste women, (3) lower caste men, (4)
lower caste women, (5) forest department
guard, and (6) the investigators for each village.
All responses for a village forest were averaged
into a single measure to yield the Forest Condi-
tion Index, a continuous variable whose value
varies between 1 for very bad forest condition
and 5 for very good forest condition. The index
correlates highly with all the component re-
sponses. We also collected forest plot data on
number of trees and their diameter at breast
height for a subset of 30 forests. The correla-
tion between forest condition index and aver-
age stem density is 0.68. We therefore use the
index based on community group responses
for the condition of forests as a reasonable
proxy variable to represent forest condition.
We realize that using community perceptions
as a proxy for forest condition introduces a
measure of subjective error in our analysis.
However, the index also includes the assess-
ment of the investigators who had experience
in collecting forest plot data and therefore were
conversant with biological measures. The
Forest Condition index correlates highly with
the assessment of the investigators (r= 0.80),
increasing our confidence in the use of the index
as our dependent variable. Needless to say, if
resources had permitted, the collection of bio-
logical data for every forest would have im-
proved both our analysis and the confidence
in our findings.
To explain the observed variations in Forest
Condition, we operationalized the following or-
dinary least squares (OLS) regression model
according to which Forest Condition is a linear
function of a suite of causal variables and a
randomly distributed error term:
Forest Condition
¼aþbðbiophysicalÞþcðeconomicÞ
þdðdemographicÞþgðinstitutionalÞ
þkðsocio-politicalÞþe;eNð0;r2Þ
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 153
where biophysical = (elevation, aspect, rain,
conifer); Economic = (treecrop, distance to
market, fodder supply, utility, adverse effect,
utility *conifer); Demographic = (size, popu-
lation change, cattle months, migratory sheep,
cattle number); Institutional = (duration,
guard, comanage, competition, fines, fines *
size); and Socio-political = (landless, village
conflict, gender relations, and gender conflict).
Our choice of an OLS model was influenced
by the nature of our dependent variable: it is
continuous and symmetric around the mean.
Table 1 provides the basic summary statistics
for our independent variables.
‘‘Elevation’’ is an important influence on for-
est ecology and condition in mountain regions
because it affects a host of other variables,
including temperature, energy demands, ac-
cessibility, and agricultural possibilities. We
measured it for each sampled forest as an
important biophysical variable that should be
positively related to forest condition. ‘‘Aspect’’
and ‘‘Conifer,’’ two other biophysical variables
on which we collected data are dichotomous,
and represent independent features of biophys-
ical processes. Conifer is coded as ‘‘1’’ when
coniferous species constitute more than 80%
of the woody vegetation in a forest. We hypoth-
Table 1. Summary statistics for the variables (n = 95)
Variable Mean
a
Standard deviation Minimum Maximum
Forest Condition 3.23 0.85 1 5
Biophysical variables
Elevation (m) 1125.55 666.5 400 3,450
Log of elevation 6.86 0.57 5.99 8.15
Aspect 41 NA 0 1
Conifer 24 NA 0 1
Rain (mm) 1588.5 505.87 866 2297
Economic variables
Treecrop 6.14 16.46 0 98
Distance to market 3.06 1.28 1 5
Fodder supply 59.31 21.15 10 100
Utility 3.06 1.06 1 4.75
Adverse effect 1.33 0.57 1 3
Utility *Conifer 0.79 1.44 0 4.75
Demographic variables
Size (no. of households) 117.15 177.44 1 1,200
Population change 0.26 0.14 0 0.84
Cattle-Months 5.57 5.00 0 12
Grazing of migratory sheep 0.74 1.35 0 6
Cattle-Number 159.41 327.16 0 2,000
Institutional variables
Duration 2.82 1.46 1 5
Guard 15 NA 0 1
Comanage 31 NA 0 1
Competition 0.22 0.55 0 2
Fines 18.26 64.52 0 500
Fines *Size 10679.88 64979.77 0 600,000
Socio-political variables
Landless 0.20 0.21 0 0.88
Village conflict 0.77 0.87 0 3
Gender relations 1.99 0.57 1 3
Gender conflict 9 NA 0 1
a
For dichotomous variables, the value is the number of positive responses.
154 WORLD DEVELOPMENT
esize that it is positively related to forest condi-
tion, because subsistence utility of coniferous
species to villagers is likely low, and there are
significant government restrictions on commer-
cial harvesting. Aspect is coded as ‘‘1’’ if the
forest is located on slopes facing north, north-
east, or northwest. In Himachal Pradesh,
north-facing slopes are generally moister and
assist in vegetative growth compared to south-
facing slopes that receive direct sunlight and
are more arid. Aspect should also bear a posi-
tive relationship with forest condition. Finally,
we also use data on ‘‘Rainfall’’ to take into ac-
count precipitation effects on forest condition,
and hypothesize that the relationship will be
positive. However, our rainfall data are drawn
from official sources and exists only at the
district level. It is relatively coarse data whose
limitations we could not overcome owing to
resource constraints during data collection.
To measure different aspects of local eco-
nomic conditions, economic relationships of
households to forests, and the articulation be-
tween localities and markets, we use a set of
variables broadly categorized as ‘‘Economic
Variables.’’ ‘‘Treecrop’’ measures the amount
of private land devoted to horticulture and is
one of our measures of market pressure. The
high incidence of fruit production in the state
prompts the use of this variable. We expect
‘‘Treecrop’’ to be negatively correlated to forest
condition. In one of our models, we used
‘‘Cashcrop’’ as an alternative measure of mar-
ket influence—this second variable indicates
the amount of private land dedicated to non-
tree cash crops in the village, and we expect it
also to have a negative relationship with forest
condition. ‘‘Market’’ measures the road dis-
tance to the nearest market, a commonly used
indicator of market pressure (Angelsen &
Kaimowitz, 1999; DeVelice & Martin, 2001).
We rescaled distance in kilometers so that this
variable ranges from 1 (for less than 1 km) to
5 (for more than 20 km). We expect it to be
positively related to forest condition: greater
the distance from market, better the forest con-
dition. ‘‘Fodder Supply’’ indicates the propor-
tion of fodder that comes from non-forest
sources for the village as a whole. We include
it as an indicator of the extent to which villag-
ers rely on forests for their livestock’s fodder
needs, and anticipate that the relationship of
forest condition with fodder supply should be
negative. Greater supply of fodder from non-
forest sources reduces dependence on forests,
and therefore incentives to manage them. Aver-
age benefits that villagers receive from the for-
est for their household subsistence activities
are indicated by ‘‘Utility,’’ an index of group
responses ranging from 1 for very low to 5 for
very high. Higher the utility, the better should
be the condition of the forest, because villagers
are more likely to be interested in protecting
forests with greater utility. ‘‘Adverse Effect’’
indicates the extent to which villagers would
be adversely affected where the forests are no
longer available. The variable is coded from 1
to 3 where a coding of ‘‘1’’ indicates highly
adverse effects, and ‘‘3’’ indicates no adverse
effects. We expect that villagers will be more
interested in protecting forests whose absence
will affect them more adversely, and therefore
this variable should have a negative relation-
ship with forest condition (because it is reverse
coded). Coniferous forests are generally not as
useful for subsistence purposes as mixed or
broad-leaved forests. To examine whether there
are any interaction effects between perceived
utility and the type of vegetation, we use an
interaction term, ‘‘Utility *Conifer.’’ We ex-
pect this term to have a negative sign because
high utility coniferous forests are valued for
their harvestable timber, and therefore likely
to be in worse condition.
A large amount of literature has argued for
the importance of population and related
demographic processes in influencing forest
condition. Because of the significant impor-
tance placed on these processes in the existing
literature, we use demographic indicators that
would take into account both human and
livestock population pressures. We include
measures of animal population because con-
sumption pressures on forests can be high even
with a low human population if per household
animal holdings are high. ‘‘Size’’ corresponds
to the number of households as the basic mea-
sure of population effects in the largely subsis-
tence economy of rural Himachal Pradesh.
We expect, on the basis of much of the litera-
ture on population and resources, that it will
have a negative relationship with forest condi-
tion. ‘‘Population Change’’ in percentage terms
over the previous decade signifies the rate of
change of population. We anticipate that the
higher the rate of change, the worse the con-
dition of the forest. For animal population,
we use three variables. The first ‘‘Grazing of
Migratory Sheep’’ indicates the number of
months for which migrant shepherds use forest
resources. Higher number of months for which
migratory sheep browse in the forest should
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 155
lead to a worse forest condition. ‘‘Cattle-
Months’’ indicates the number of months
for which village animals graze in the forest.
Finally, ‘‘Cattle-Number’’ represents the num-
ber of village cattle that graze in the community
forest. Both these variables should be nega-
tively related to forest condition.
Institutional factors, especially in the litera-
ture on common property but also more gener-
ally, have come to be viewed as extremely
important in shaping access to forest resources
and the nature of activities that occur in forests.
They are therefore also very important in influ-
encing forest conditions. In our analysis, we
use several variables to take different aspects
of institutions into account. ‘‘Duration’’ is the
number of years for which community-based
conservation has been in force in a given case.
It ranges from 1 to 5, with 1 denoting very re-
cent initiatives (post-1990) and 5 standing for
older community-based management systems
(pre-1930). We anticipate that older institutions
will have a positive impact on forest condition
because over time, the local residents who are
subject to these institutions would have ad-
dressed operational and functional problems.
‘‘Guard’’ and ‘‘Comanage’’ are dichotomous
variables. If the local community hires a guard
for enforcement the variable is coded 1, other-
wise it is coded 0. Similarly, if government or
forest department officials are involved in facil-
itating decision making within local community
institutions, the variable is coded 1, otherwise
0. We expect both of these variables to have
a positive sign. ‘‘Competition’’ represents
whether more than one individual competes
for positions in the executive body of the com-
munity organization. It varies between 0 and 2,
where 0 indicates no competition, and 2 indi-
cates that there is almost always some competi-
tion. Greater competition indicates that
positions on the executive body of commu-
nity-level forestry institutions are more mean-
ingful, and therefore, it is likely to be
positively related to forest condition. ‘‘Fines’’
measures the number of individuals fined by
the community institution in the past two years.
We view this variable as an indicator of the ex-
tent to which institutional enforcement is actu-
ally present in a village, and expect it to have a
positive relationship with forest condition.
Finally, we include an interaction term
‘‘Fines *Size’’ because we expect rule enforce-
ment to become more critical as group size in-
creases (see Poteete & Ostrom). Collectively,
the different institutional variables allow us to
examine whether, and the extent to which the
longevity of community institutions, differences
in their functioning, and the strictness of
enforcement are related to the condition of
the forest.
Socio-economic and socio-political variables
at the village level affect the ability of users to
cooperate with each other in conserving forests,
as also the dynamics of forest governance
regimes (Kant, 2000). While we have already
discussed a range of socio-economic variables,
the more political variables constitute the final
set of causal influences our model examines.
‘‘Landless’’ is the proportion of households
with less than 0.4 ha of agricultural land. This
variable potentially affects forest condition in
two ways. It likely increases harvesting pres-
sures on the forest, and therefore exerts a neg-
ative effect on forest condition. But a higher
proportion of landless in the village is also
likely to increase overall dependence of villag-
ers on the forest, potentially prompting them
to try to protect it better. ‘‘Village Conflict’’
measures the level of reported conflict in the vil-
lage, ranging from 0 for no conflict to 4 for vio-
lent conflict. Higher levels of conflict should
bear a negative relationship with forest condi-
tion. In addition, we also include two gender-
related variables in the analysis. Indeed, several
authors have highlighted the important role of
gender and meaningful women’s participation
in the success of community-based conserva-
tion (Agarwal, 2001; Rocheleau & Edmunds,
1997). ‘‘Gender Relations’’ measures whether
women hold positions of power in village orga-
nizations, ranging from 1 for equitable distribu-
tion between men and women in these positions
to 3 for distribution that greatly favors men.
We expect women to be more active in protect-
ing forests in areas where forests are not in
a good condition, and therefore, lower levels
of this variable should be related to higher
involvement of women in decision-making
positions. ‘‘Gender Conflict’’ is a categorical
variable for the presence or absence of conflict
in the village along gender lines (coded 1 if
gender conflicts are present). The presence of
gender-related conflicts, we expect, will be
negatively related to forest condition.
Our statistical analysis thus takes into account
a broad array of variables. By examining their
effects on forest condition simultaneously, we
hope to identify the extent to which each
variable exerts independent causal effects on
outcomes even when the impact of other theo-
retically relevant variables is taken into account.
156 WORLD DEVELOPMENT
5. RESULTS AND DISCUSSION
The results of our analysis are presented in
Table 2 and described below. The table lists
the regression coefficients of variables we used
in three different models. We include the results
for three models to examine the sensitivity of
the estimates in the first model to changes in
model specification. It is clear that the coeffi-
cient estimates and standard errors are similar
across the models: for no variable do the coef-
ficient estimates change by more than one stan-
dard error across the three models. The first
column in the table presents the results of the
model that most closely conform to the vari-
ables discussed above. The overall similarity
of the coefficients, their standard errors, and
statistical significance, and the different indica-
tors of model fitness such as the R
2
, the ad-
justed-R
2
, and the F-statistic suggest that the
statistical relationships in the initially specified
model are robust against perturbations. The
Table 2. Dependent variable—Forest Condition
Variable Model 1 Model 2 Model 3
Biophysical variables
Log of elevation 0.417 (0.145)*** 0.29 (0.145)** 0.414 (0.147)***
Aspect 0.683 (0.128)*** 0.683 (0.128)*** 0.682 (0.129)***
Rain (mm) 0.0004 (0.00013)*** 0.0004 (0.00013)*** 0.0004 (0.0001)***
Conifer 1.536 (0.457)*** 1.598 (0.461)*** 1.52 (0.463)***
Economic variables
Treecrop 0.007 (0.004)* 0.007 (0.004)**
Cashcrop 0.004 (0.003)
Distance to market 0.269 (0.052)*** 0.259 (0.052)*** 0.27 (0.052)***
Fodder supply 0.008 (0.003)** 0.009 (0.003)*** 0.008 (0.003)**
Dependence 0.35 (0.117)*** 0.378 (0.118)*** 0.348 (0.118)***
Utility 0.18 (0.068)** 0.171 (0.068)** 0.18 (0.069)**
Utility *Conifer 0.406 (0.143)*** 0.435 (0.144)*** 0.403 (0.145)***
Demographic variables
Population change 1.639 (0.494)*** 1.462 (0.495)*** 1.65 (0.506)***
Size 0.001 (0.0006)** 0.001 (0.0006)** 0.0013 (0.0007)*
Grazing of migratory sheep 0.221 (0.068)*** 0.218 (0.069)*** 0.223 (0.070)***
Grazing of cattle (months) 0.038 (0.015)** 0.04 (0.015)** 0.038 (0.015)**
Grazing of cattle (number) 0.0005 (0.0002)** 0.0005 (0.0002)** 0.0005 (0.0002)**
Area of forest 0.00002 (0.0001)
Institutional variables
Guard 1.22 (0.200)*** 1.161 (0.203)*** 1.1214 (0.202)***
Comanage 0.528 (0.156)*** 0.510 (0.156)*** 0.526 (0.157)***
Duration 0.130 (0.053)** 0.102 (0.052)* 0.131 (0.054)**
Competition 0.433 (0.152)*** 0.458 (0.152)*** 0.428 (0.156)***
Fines *Size 5.60
e06
(1.58e06)*** 5.77
e06
(1.59e06)*** 5.61
e06
(1.59e06)***
Socio-political variables
Gender relations 0.235 (0.108)** 0.267 (0.109)** 0.236 (0.109)**
Gender conflict 0.418 (0.226)* 0.452 (0.226)** 0.41 (0.231)*
Landless 0.657 (0.314)** 0.834 (0.311)*** 0.656 (0.316)**
Village conflict 0.131 (0.074)* 0.181 (0.076)** 0.132 (0.075)*
No. of observations 95 95 95
R
2
70.28% 70.07% 70.30%
Adjusted-R
2
60.09% 59.8% 59.54%
F-statistic F
(24,70)
= 6.90 F
(24,70)
= 6.83 F
(25,69)
= 6.53
Root MSE 0.537 0.539 0.540
Prob > F0.0000 0.0000 0.0000
*
,
**
, and
***
signify statistical significance at 0.1, 0.05 and 0.01 levels, respectively.
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 157
R
2
for the three models is 70.28%, 70.07%, and
70.30%; and the adjusted-R
2
is 60.09%, 59.8%,
and 59.54%, respectively. These numbers indi-
cate that the basic model explains a high
proportion of the observed variance. The F-
statistic for the three models is 6.90, 6.83, and
6.53, with p< 0.00001 for each model. The
residuals for all three models are normally dis-
tributed. To check for heteroskedasticity, we
used the Breusch–Pagan Test, and for the pos-
sibility of non-linearities the Ramsay Regres-
sion Specification Error Test (RESET). We
also examined whether individual observations
have a strong influence by using the Cook’s
Distance and Leverage statistics. These diag-
nostics indicate that the models are well speci-
fied, and add to our confidence in the scope
of our findings. Finally, we calculated the Var-
iance Inflation Factor for each variable to
check for multicollinearity. The scores for the
different variables indicate that collinearity is
an issue only with Conifer and Utility *Coni-
fer which are strongly correlated. However,
we did not omit either of these variables for
theoretical reasons as discussed above.
In assessing the results of the analysis, the
first important point to note is that all the dif-
ferent categories of causal influences include
at least some statistically significant variables
even if there are other variables that are not
highly significant. Some of the observed rela-
tionships do not conform to the stated hypoth-
eses in the previous section. Exploring these
variables and relationships in greater detail
helps identify connections among the hypothe-
sized causal and outcome variables that may be
instructive in other contexts as well.
For biophysical factors, all the variables in-
cluded in the analysis are statistically signifi-
cant. The coefficient for the logged value of
‘‘Elevation’’ confirms that forests in higher ele-
vations are likely to be in better condition. The
result for ‘‘Aspect’’ indicates that north-facing
slopes, as hypothesized, are more likely to have
forests in good condition. Coniferous forests
tend to be in better condition than mixed or
broad-leaved forests. However, rainfall has a
negative sign, indicating that forests in districts
with higher levels of rainfall are in worse condi-
tion. We believe that the measure of total rain-
fall at the district level that we used is likely
inadequate to capture the complexities in the
relationship between rainfall patterns and for-
est condition, especially since the administra-
tive districts in Himachal Pradesh do not
follow ecological boundaries. But overall, the
results suggest that biophysical variables play
a very important role in shaping forest condi-
tions, and studies of forest governance need
to take these variables into account so as not
to overemphasize the role of socio-economic
or institutional influences over forest condition.
Economic and demographic variables have
received perhaps the most careful and general
attention in studies of conservation. The well-
known equation, ‘‘I=P*A*T’’ summarizes
the conviction of many that the conjunction
of Population, Affluence, and Technology
defines Impact on environmental outcomes
(Ehrlich & Ehrlich, 1990). Consider the six
economic variables in our analysis: Treecrop,
Distance to Market, Fodder Supply, Utility,
Dependence, and the interaction term Uti-
lity *Conifer. The first two are related to the
impact of market forces, the remaining four
concern the role of forests in the household
and village economy.
All the variables are statistically significant,
although not always in the hypothesized direc-
tion. Treecrop has a negative relationship with
forest condition (as hypothesized), indicating
that the greater the prevalence of fruit trees in
village agriculture, the less likely are forests to
be in a good condition. Our data also indicate
that forests are likely to be in a better condition
the closer the markets are. Although our results
contrast with those in a significant literature
on the effects of roads and markets (Angelsen
& Kaimowitz, 1999; Southworth & Tucker,
2001), a number of recent studies have reported
similar findings as ours (Gautam, Shivakoti, &
Webb, 2004; Rudel, Bates, & Machinguiashi,
2002). We account for the observed relation-
ship by suggesting that distance from roads
can also be a proxy for distance from govern-
ment offices, and that official presence acts as
a disincentive to deforestation. Thus, in con-
texts where state officials are effectively present
in local contexts, distance from roads or mar-
kets is not an efficient measure of economic
pressures because its effects are confounded
by those of government influence. Indeed, other
studies that explain greater forest density in
locations closer to roads essentially use a simi-
lar explanatory mechanism (Agrawal & Yad-
ama, 1997; Alvarez & Naughton-Treves, 2003;
Nagendra, Southworth, & Tucker, 2003). In
contrast, the Treecrops variable is a more direct
representation of articulation with external
market forces given the social and political
economy of Himachal Pradesh. It indicates
whether village forests face pressure for logging
158 WORLD DEVELOPMENT
as a result of demands for wood from owners of
apple orchards and similar fruit tree crops.
Cashcrop also represents the articulation of vil-
lage communities with markets. But in the spe-
cific context of Himachal Pradesh where forest
land is not being converted for cultivation, this
variable does not capture the influence of mar-
kets on forest conditions to the same degree as
Treecrop.
The third economic variable we use, Fodder
Supply, indicates the proportion of fodder real-
ized from non-forest sources, and has a nega-
tive relationship with forest condition as
hypothesized. In a similar fashion, the variable
Utility, which indicates the overall subsistence
benefits from forests to villagers, is positively
related with forest condition. The results for
both variables can be interpreted to mean that
when villagers assess their community forest
to be more useful for subsistence and liveli-
hoods, they make greater efforts to protect for-
ests. Other scholars have remarked upon the
extent to which livelihoods of the poor are
dependent on forests, and the complexity of
the concept of dependence (Fisher, 2004). Our
conclusion that communities are likely to try
to protect and maintain forests when they rely
on them for subsistence is supported at least
in part by Lise’s (2000) finding for three Indian
states (Haryana, Uttar Pradesh, and Bihar)
that high levels of forest dependence encourage
greater participation in forest governance.
Adverse effect—how villagers perceive the
loss of the forest to affect them—is positively
related to forest condition. This finding is con-
trary to our initial hypothesis, and underscores
the complexity of the concept of dependence
(Beckley, 1998; McSweeney, 2002). We believe
that two different factors may be used to ex-
plain the tension between the results for Fodder
Supply and Utility, and Adverse Effect. The
joint result for these three variables suggests
that it is subsistence rather than general benefits
from forests that prompt villagers to express
the need to conserve forests. Further, it is likely
that when subsistence benefits of forests are
highlighted in a question, villagers are more
likely to express the need to protect forests than
when a more abstract question about how they
will be affected by loss of forest. It is also pos-
sible that when villagers do not view a forest
as important to them, its condition is better
because villagers are not extracting too much
from the forest.
The final economic variable is the interaction
term, Utility *Conifer, and it has a negative
relationship with forest condition. This result
indicates that for coniferous forests, higher
the utility, lower the condition of the forest.
Coniferous forests are typically used for timber
and the negative coefficient of the interaction
terms—coupled with positive coefficients for
both Utility and Conifer—suggests that if vil-
lagers view such forests as having high utility
(in terms of the supply of timber), the forests
are likely not to be in a good condition. A cor-
ollary implication may be that mixed and
broadleaved forests that are seen as having high
utility are more likely to be in a better condi-
tion. This interpretation of the coefficient is
consistent with the results for the two variables,
Fodder Supply and Utility.
All the five demographic variables in our
analysis (Size, Population Change, Grazing of
Migratory Sheep, Cattle-Months, and Cattle-
Number) are statistically significant, but the
unexpected signs for some of them indicate that
some theoretically interesting causal processes
may be at play. Consider the first two variables
related to population size and the rate at which
population changes. Size has a positive rela-
tionship with Forest Condition and Population
Change has a negative relationship. The results
plausibly indicate that the greater the rate of
population change, the more likely are commu-
nity forests to be in a bad condition. A possible
causal mechanism that may explain this finding
is that when population changes are rapid, it is
more difficult for users to change existing forest
use and management practices and for existing
institutions to accommodate the impact of
change. The unexpected positive relationship
of Size (number of households) holds for all
three models. It would appear that for the
range within which household numbers vary
in our dataset, more people leads to better for-
ests. This result is similar to what Tiffen, Mor-
timore, and Gichuki (1994) point out in their
work on soil erosion in Kenya. Larger numbers
of people may adopt more labor-intensive agri-
cultural technologies and provide greater labor
to invest in land improvement, thereby reduc-
ing soil erosion. In a similar fashion a larger
number of people, as long as this number is
not excessive, can improve forest condition by
increasing aggregate household contributions
for conservation and strengthening the institu-
tions that facilitate conservation.
The signs of coefficients for livestock related
variables also do not always accord with our
hypothesized relationships. For Cattle-Months,
the association with forest condition is negative
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 159
as hypothesized. For Grazing of Migratory
Sheep, however, the forest condition is posi-
tively related to the number of months sheep
are present in the community forest. We believe
that both for this and the Cattle-Numbers var-
iable, reverse causality may be at work. That is,
migrant shepherds may be taking their sheep to
forests that are in better condition, and villag-
ers may be sending more cattle to better forests,
rather than it being the case that more grazing
by cattle and more months of the presence of
migrant sheep produces better forests.
In model 3, we examine whether the area of
forest is important in influencing its condition.
Our hypothesis was that larger forests should
be in better condition. The results do not
support the hypothesized relationship. We
conclude that like other demographic variables,
there is no direct relationship between forest
size and condition. Just as the effects of popula-
tion size are mediated by a large number of
institutional, economic, and socio-political
variables, so are those of greater availability
of forested land. Once these other factors are
taken into account, forest area does not have
a statistically significant effect on forest condi-
tion.
A large number of existing studies have as-
serted the critical importance of institutional
and socio-political factors in shaping resource
use, governance, and outcomes (Armitage,
2002; McKean, 1992; Ostrom, 1990). If some
have asserted the importance of local control
in reducing resource extraction from forests
(Edmonds, 2002), others have pointed to the
significant differences that de facto institutional
arrangements produce in forest governance-
related outcomes (Agrawal & Yadama, 1997;
Wily, 2001).
In our analysis, we used only five of the six
institutional variables that we had initially
chosen: Duration, Guard, Comanage, Compe-
tition, and Fines *Size. The variable, Fines,
turned out to be highly correlated with Size,
the demographic variable we use to represent
population of the village (r= 0.75), and with
the interaction term, Fines *Size (r= 0.92),
leading to collinearity in the regression models.
Both because we have other institutional vari-
ables (Guard) that tap into the issue of enforce-
ment in a roughly similar way as does Fines,
and because the variables Size and Fines *Size
appear to us to be theoretically more impor-
tant, we dropped Fines from our analysis. Size
is theoretically more important because it rep-
resents the demographic argument that number
of people who use a resource affects the condi-
tion of the resource, and its conservation.
Fines *Size is theoretically important because
it tests for the possibility that in villages with
larger number of households, there may be
scale effects associated with enforcement.
The five institutional variables are all statisti-
cally significant. The time for which a local com-
munity-level institution has been in existence
(Duration) is positively related to forest condi-
tion. The coefficient for Guard is also statisti-
cally significant, but the sign is negative,
indicating that in forests that are in good condi-
tion, there is less enforcement. The negative sign
for guard is in tension with the findings of a
number of studies in which the presence of lo-
cally appointed and paid guards is associated
with forests that are in good condition (Gibson,
Williams, & Ostrom, 2005; McKean, 1992). Our
interpretation of this finding is that in the stud-
ied cases, villagers are more likely to hire guards
and impose fines more frequently if their forests
are not in a good condition in an effort to im-
prove their forests. Thus, the causal arrow sug-
gested by our data runs in the reverse direction
from what we had hypothesized. Two different
factors may be at work. Villagers hire guards
either when they see adverse changes occurring
in forests that were earlier in good condition,
or they hire guards for forests whose future con-
ditions they believe they can improve. Thus, the
current condition of forests is not a good proxy
for the causal relationship at work. The variable
Comanage, representing government officials’
involvement in community decision making, is
negatively related with forest condition and
prospects for conservation. We pose three likely
explanations based on qualitative evidence:
powerful external figures hinder the adoption
of governance rules that are best suited to local
conditions; comanagement processes introduce
new and sometimes large amounts of external
funds into the local context, a development that
may exacerbate the ill effects of powerful exter-
nal figures; and finally, comanaging government
officials often devolve control over forests that
are not in a good condition to begin with. In-
deed, in Himachal Pradesh, the forest depart-
ment devolved control over forests beginning
in the 1980s, when there was significant bureau-
cratic reluctance against participating in
comanagement. The forests that local commu-
nities came to comanage were often therefore
in poor shape.
Greater competition in the selection of office
bearers in community institutions is associated,
160 WORLD DEVELOPMENT
as hypothesized, with better forest condition.
Finally, Fines *Size has a negative sign. Espe-
cially in combination with the positive sign for
Size, this finding suggests that in larger groups,
a high incidence of fines is negatively associated
with Forest Condition. The reason may be that
as group size increases, more fines are particu-
larly detrimental to features of community—
trust, reciprocity, conservationist-norm-ori-
ented activities—that confer positive valence
upon community-based conservation. It may
also be the case that smaller communities
address the problem of enforcement without
relying on monetary fines.
The coefficients of the socio-political vari-
ables (Landless, Village Conflict, Gender Rela-
tions, and Gender Conflict) are also statistically
significant. These variables represent also the
specific effects of group heterogeneity on forest
condition (Poteete & Ostrom, 2004; Varughese
& Ostrom, 2001). We find that greater landless-
ness is positively related with forest condition,
which may indicate that villagers attempt
to protect forests better when landlessness is
higher. This result reflects the situation in
Himachal Pradesh: low levels of social and eco-
nomic inequality at the village level. In such a
situation, it is likely that Landless captures sub-
sistence dependence rather than heterogeneity
within the community. Higher levels of Village
Conflict are related to forests in worse condi-
tion. More conflict-ridden social relationships
in the village likely make decision making
around forest protection difficult, as pointed
out by some other scholars as well (Johnson
& Forsyth, 2002).
The relationship of the two gender vari-
ables—Gender Relations and Gender Con-
flict—with Forest Condition is quite
interesting. The positive sign of the first one
of these indicates that one of the major argu-
ments advanced by gender theorists is likely
on target: that involvement of women in insti-
tutionalized decision making improves the
prospects for better resource conservation
(Agarwal, 1998, 2001; Sarin, 1995). As ethno-
graphic and sociological studies of household
division of labor in hill areas have noted,
women are typically charged with collection
of forest products such as fodder and firewood.
If they are in leadership positions, they can cre-
ate regulatory mechanisms that are more suited
to the local context as well as their needs for
forest products. The analysis also indicates that
the presence of gender conflicts is positively
associated with better forest condition. The
causal implication of these two variables taken
together is that women are likely involved in
decision making and positions of power only
after there is some conflict related to gender
issues. Clear examples illustrating these gender
dynamics occurred in the community forests
of Shanag in Kullu, and Thalli and Kuthah in
Mandi. In all three villages, women gained
decision-making positions only after local
forests were viewed as deteriorating rapidly.
Women as a group were most affected by this
decline, and there were substantial disagree-
ments within the village community around
how to address these issues. Gender-related
conflicts thus do not necessarily produce nega-
tive effects as far as women’s participation in
decision making is concerned.
Table 3 presents information about how and
to what extent the causal variables included in
our analysis affect the predicted value of the
outcome variable: forest condition. The second
column of the table is an array of the coeffi-
cients of the variables, as derived from model
1. The third and fourth columns indicate the
value of the independent variable at the 25th
and the 75th percentile levels. The fifth column
indicates how the predicted value of Forest
Condition changes in response to a change in
the causal variable from the 25th to the 75th
percentile. The final column lists the mean effect
on the predicted value of Forest Condition. In
simple terms, we may say that Table 2 indicates
how well the data fit the regression line in an n-
dimensional space where ‘‘n’’ is the number of
variables, and Table 3 indicates how much the
regression line (or the dependent variable) is af-
fected by changes in the value of the causal
variables.
Even a cursory look at the table is sufficient
to show the large contribution of biophysical
variables to the predicted value of forest condi-
tion. Elevation and Conifer generate the great-
est change in forest condition with changes in
their value. Further, elevation is correlated with
several other independent variables. The impli-
cation is obvious—studies of local resource
governance that set aside the treatment of bio-
physical variables and focus mainly on socio-
economic, demographic, and/or institutional
variables to explain forest condition likely over-
estimate the impact of the variables included in
the analysis. One may defend the omission of
biophysical variables from an analysis on the
ground that although changes in these variables
may produce substantial changes in forest con-
dition, there is little that can be done about
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 161
them. Therefore, from a policy standpoint, it is
unnecessary to be concerned about variables
such as elevation, aspect, or rainfall. We dis-
agree. Inclusion of relevant variables in an
analysis helps generate a more precise and
therefore valuable sense of the power and limits
of the causal factors that policy interventions
can influence.
The important influence of biophysical vari-
ables on outcomes is quite probably related to
the high variability of ecological and biophysi-
cal conditions in Himachal Pradesh. Our in-
ferences about the importance of including
biophysical variables in analyses of common
pool resource related outcomes are especially
germane to other contexts characterized by sig-
nificant variability—typically ones with which
scholars of the commons are often concerned.
Much of the work on fisheries, in mountainous
regions, and in semi-arid environments will fit
the description of a context that has high and
localized variability in biophysical conditions.
A second point to note is that changes in val-
ues of the demographic variables included in
our analysis have a smaller overall effect on
the predicted value of Forest Condition in com-
parison to Institutional Variables. Among the
different demographic variables, the one that
Table 3. Mean predicted value of Forest Condition
Variable Model
coefficients
25th
percentile
75th
percentile
Effect on
Forest Condition
Mean effect
Biophysical variables
Log of elevation 0.417 6.31 7.378 0.446 2.866
Aspect
a
0.683 0 1 0.683 0.683
Rain 0.0004 1,069 2,297 0.528 0.684
Conifer
a
1.536 0 1 1.536 1.536
Economic variables
Treecrop 0.007 0 5 0.035 0.043
Distance to market 0.269 2 4 0.538 0.825
Fodder supply 0.008 40 75 0.2998 0.508
Utility 0.18 2 4 0.36 0.552
Adverse effect 0.35 1 2 0.35 0.468
Utility *Conifer 0.406 0 1 0.406 0.32
Demographic variables
Size (no. of households) 0.001 20 152 0.181 0.161
Population change 1.639 0.166 0.333 0.273 0.426
Cattle-Months 0.038 0 12 0.457 0.213
Grazing of migratory
sheep (months)
0.221 0 1 0.221 0.164
Cattle-Number 0.0005 0 150 0.082 0.088
Institutional variables
Duration 0.130 2 4 0.26 0.367
Guard
a
1.22 0 1 1.22 1.22
Comanage
a
0.528 0 1 0.528 0.528
Competition 0.433 0 1 0.433 0.096
Fines *Size 5.60
e06
0 400 0.0022 0.0598
Socio-political variables
Landless 0.657 0.027 0.277 0.164 0.133
Village conflict 0.131 0 1 0.131 0.101
Gender relations
b
0.235 1 3 0.472 0.469
Gender conflict
a
0.418 0 1 0.418 0.418
Note: The figures in the fifth column denote the absolute contribution to the value of Yover the inter-quartile range
of X(for dichotomous variables, the difference is between the effect for 0 and 1).
a
Denotes a dichotomous variable.
b
The range for ‘‘Gender Relations’’ is between 15th and 85th percentile, respectively, because the inter-quartile range
is constant at the value of 2.
162 WORLD DEVELOPMENT
affects outcomes most is population change
rather than the absolute number of user house-
holds. We can say that in contexts such as that
of Himachal Pradesh, and for the range of var-
iation in demographic variables that our data
represent, the arguments of many social scien-
tists that there is no straightforward link
between population levels and resource degra-
dation are well directed (Leach & Mearns,
1996; Tiffen et al., 1994; Varughese & Ostrom,
2001). This link is mediated through institu-
tional form. The strength and resilience of insti-
tutional enforcement and the durability of
institutions are highly significant factors. Our
analysis, thus, confirms much scholarship on
the importance of locally negotiated institu-
tional arrangements.
A third important finding worth highlighting
is the role of gender-related variables in our
analysis. The variables concerning gender that
we include in our model are each statistically
significant. Changes in the values of both pro-
duce substantial impact on the predicted value
of forest condition. This impact exists above
and beyond the effects of a large number of
other causal variables included in the analysis.
Our study also indicates that women’s partici-
pation in forestry-related community institu-
tions is not an easy or smooth process in
rural contexts where they are typically assigned
more burdensome parts of household and agri-
cultural labor. But the positive association of
gender-related conflicts and forest condition
suggests that such obstacles ultimately have a
positive effect on forestry outcomes.
6. DISCUSSION AND CONCLUSION
In conclusion, it is useful to highlight two is-
sues fundamental to our study—one relates to
methods, the other concerns how contextual
factors imbue specific variables with their cau-
sal significance. In this study, we have opted
for an approach that can harness the strengths
of a multivariate, large-Ndriven analysis in
detecting patterns in large amounts of data—
something that is not easy to accomplish by a
case-oriented approach. But it should be obvi-
ous that our statistical analysis of data, choice
of variables, manner of operationalizing the
variables, and interpretation of regression coef-
ficients have relied on the findings of many case
studies on local resource governance, and our
own intensive field-based empirical work. We
have highlighted the importance of conducting
statistical work on local resource use and gov-
ernance because so much of the literature on
the subject is driven by single case-oriented
analytical lens. Undoubtedly, it is easier to
identify causal mechanisms in intensive studies
of specific cases—perhaps a factor that explains
why so much existing work on the subject has
relied on fieldwork and case analysis.
Our data analysis suggests however that the
generalizability of causal mechanisms identified
in case studies can be ascertained only by
undertaking statistical–analytical work to de-
tect broad patterns in the data: something at
which fieldwork and case analysis are weak.
But an intimate knowledge of field conditions
and the context of the data—what we call
‘‘thick reading’’ and ‘‘close analysis’’—are crit-
ical to interpret statistical patterns. To illus-
trate, during fieldwork in the villages of
Shanag, Thalli, and Kuthah, we learned of
the involvement of women in decision-making
positions as local forests deteriorated. We also
learned about the tense social interactions sur-
rounding the decline of community forests
and the incorporation of women in decision-
making positions. However, whether this set
of gender-and-forest interactions was relevant
for the full sample of representative cases on
which we were collecting data could not
become obvious until we had carried out our
statistical analysis. On the other hand, the
knowledge of what was happening in the field
was essential to give meaning and significance
to the correlation we observed between forest
conditions, and gender relations and gender
conflict.
The same process of interpreting the data and
statistical results helped us make sense of the
relationships we identified between forest con-
dition and the presence of a guard. The impor-
tance of aspect and its relationship to forest
condition, similarly, became apparent as a re-
sult of what our interviewees told us about
how it helps tree growth. We interpret the vari-
ables, Utility and Fodder Supply, to indicate
that residents protect forests better when they
consider them as providing more subsistence
benefits. This interpretation is again based on
what we learned in the field. A rigorous test
of competing hypotheses about causal direction
would require longitudinal data for the sam-
pled villages. The cited examples of the rela-
tionships between statistical correlations and
their interpretation should, however, convey
our sense that the conventional tension between
case study work and statistical analysis is an
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 163
artifact of particular methodological commit-
ments. Careful, meaningful, and systematic
knowledge about how resource governance un-
folds and is shaped requires an inescapable con-
versation between statistics and contextual
expertise.
These observations about method are related
to what our study tells us about the intercon-
nections between context and the relevance of
specific variables. Consider two variables we
use in our analysis: Elevation and Distance to
Market. Variation in Elevation is one of the
central factors in our analysis that explains dif-
ferences in Forest Condition, our dependent
variable. It is in significant measure this vari-
able that prompts us to call for greater atten-
tion to biophysical variables in analyses of the
governance of the commons. It is related to
many of the other independent variables we
use: Treecrop, Cashcrop, Size, Grazing by
migratory sheep, Months of grazing by cattle,
Duration, and Guard. Many of these relation-
ships seem evident—for example, size of vil-
lages should decline as one moves to a greater
elevation in mountain regions. However, if in
a particular study, data exist only on village
size and the information on elevation is ig-
nored, the reported relationship between forest
condition and village size will likely be an over-
or under-estimate. Elevation produces an im-
pact on forest condition through the way it
affects these other variables to which it is re-
lated rather than a direct raw impact.
One can interpret the importance of elevation
on forest condition in our analysis in another
way as well. It is quite possibly a proxy for tem-
perature (a variable on which we could not
gather local data, and which is therefore not in-
cluded in our analysis). Elevation, therefore,
likely also represents the effects of temperature
on forest growth, demand for firewood and
cooking materials, agricultural growth possibil-
ities and thus Treecrop and Cashcrop, and so
forth. It is no surprise that changes in this var-
iable have large effects on the state of the
dependent variable.
The idea can also be illustrated in a different
manner with reference to the important role of
distance to market in our analysis. In our data,
closeness to the market is associated with for-
ests that are in better condition. The finding
runs counter to that in many existing studies,
as noted. Our reinterpretation of the statistical
relationship is that distance to market con-
founds the effects of market pressures and offi-
cial constraints on harvesting of forest product.
This reinterpretation is driven especially by an
awareness of the political–economic geography
of Himachal Pradesh where the presence of the
state is near ubiquitous, and its influence runs
along channels created by roads. Similarly,
the scale at which variations in the influence
of roads on forest condition become discernible
is far finer than it would be in a context where
population density is low.
Our study, thus, underscores the importance
of contextual variations and awareness of such
variations in knowing how to specify and oper-
ationalize the variables of interest. This need
for intimate familiarity with data, informed
by knowledge of field conditions, has a
prime implication for the study of the com-
mons. Variations in how the same factors oper-
ate and should be operationalized in different
macro-contexts should make us pessimistic
about the possibility of a universal theory of
the commons. What Jon Elster suggests about
the study of local justice—‘‘it is a very messy
business, and... it may be impossible to iden-
tify a set of necessary and sufficient conditions
that constitute a theory of local justice’’
(1992, p. 14)—is likely also true for the
study of the commons and local resource gover-
nance.
REFERENCES
Agrawal, A. (2001). Common property institutions and
sustainable governance of resources. World Develop-
ment, 29(10), 1649–1672.
Agrawal, A., & Yadama, G. (1997). How do local
institutions mediate the impact of market and
population pressures on resource use. Development
and Change, 28(3), 435–465.
Agarwal, B. (1998). Environmental management, equity
and ecofeminism: Debating India’s experience. Jour-
nal of Peasant Studies, 25(4, July), 55–95.
Agarwal, B. (2001). Participatory exclusions, community
forestry, and gender: An analysis for South Asia and
a conceptual framework. World Development, 29(10),
1623–1648.
Alvarez, N. L., & Naughton-Treves, L. (2003). Linking
national agrarian Policy to deforestation in the
Peruvian Amazon: A case study of Tambopata,
1986–1997. Ambio, 32(4), 269–274.
Angelsen, A., & Kaimowitz, D. (1999). Rethinking the
causes of deforestation: Lessons from economic
164 WORLD DEVELOPMENT
models. World Bank Research Observer, 14(1),
73–98.
Armitage, D. (2002). Socio-institutional dynamics and
the political ecology of mangrove forest conservation
in Central Sulawesi, Indonesia. Global Environmental
Change, 12(3), 203–217.
Baland, J. M., & Platteau, J. P. (1996). Halting degra-
dation of natural resources: Is there a role for rural
communities? Oxford: Clarendon Press.
Beckley, T. M. (1998). The nestedness of forest depen-
dence: A conceptual framework and empirical explo-
ration. Society and Natural Resources, 11(2, March),
101–120.
Berkes, F. (Ed.) (1989). Common property resources:
Ecology and community-based sustainable develop-
ment. London: Belhaven Press.
Berkes, F. (2004). Rethinking community-based conser-
vation. Conservation Biology, 18(3), 621–630.
Brockway, D. G. (1998). Forest plant diversity at local
and landscape scales in the Cascade Mountains of
southwestern Washington. Forest Ecology and Man-
agement, 109(1–3), 323–341.
Brown, K., & Pearce, D. (Eds.) (1994). The causes of
tropical deforestation: The economic and statistical
analysis of factors giving rise to the loss of the tropical
forests. London: UCL Press.
Campbell, B., Mandondo, A., Nemarundwe, N., Sithole,
B., De Jong, W., Luckert, M., & Matose, F. (2001).
Challenges to proponents of common property
resource systems: Despairing voices from the social
forests of Zimbabwe. World Development, 29(4,
April), 589–600.
DeVelice, R. L., & Martin, J. R. (2001). Assessing the
extent to which roadless areas complement the
conservation of biological diversity. Ecological Appli-
cations, 11(4, August), 1008–1018.
Dietz, T., Ostrom, E., & Stern, P. (2003). The struggle to
govern the commons. Science, 302(special issue,
December 12), 1907–1912.
DOP (1997). Department of Planning Ninth Five Year
Plan 1997–2002. Shimla: Department of Planning,
Government of Himachal Pradesh.
Edmonds, E. V. (2002). Government-initiated commu-
nity resource management and local resource extrac-
tion from Nepal’s forests. Journal of Development
Economics, 68(1), 89–115.
Ehrlich, P., & Ehrlich, A. (1990). The Population
Explosion. New York: Simon and Schuster.
Elster, J. (1992). Local justice: How institutions allocate
scarce goods and necessary burdens. New York:
Russell Sage Foundation.
Fisher, M. (2004). Household welfare and forest depen-
dence in Southern Malawi. Environment and Devel-
opment Economics, 9, 135–154, Part 2.
Gautam, A., Shivakoti, G., & Webb, E. (2004). Forest
cover change, physiography, local economy, and
institutions in a mountain watershed in Nepal.
Environmental Management, 33(1), 48–61.
Gibson, C. C., & Marks, S. A. (1995). Transforming
rural hunters into conservationists: An assessment of
community-based wildlife management programs in
Africa. World Development, 23(6), 941–957.
Gibson, C. C., Williams, J., & Ostrom, E. (2005). Local
enforcement and better forests. World Development,
33(2), 273–284.
Goldman, M., & Schurman, R. A. (2000). Closing the
’Great Divide’: New social theory on society and
nature. Annual Review of Sociology, 26, 563–584.
Goldthorpe, J. H. (1997). Current issues in comparative
macrosociology: A debate on methodological issues.
Comparative Social Research, 16, 1–26.
Johnson, C., & Forsyth, T. (2002). In the eyes of the
state: Negotiating a ‘‘rights-based approach’’ to
forest conservation in Thailand. World Development,
30(9, September), 1591–1605.
Kant, S. (2000). A dynamic approach to forest regimes
in developing economies. Ecological Economics, 32(2,
February), 287–300.
Leach, M., & Mearns, R. (Eds.) (1969). The lie of the
land: Challenging received wisdom on the African
environment. Oxford, UK, and Portsmouth, NH:
James Currey and Heinemann.
Lise, W. (2000). Factors influencing people’s participa-
tion in forest management in India. Ecological
Economics, 34(3, September), 379–392.
McKean, M. (1992). Success on the commons: A
comparative examination of institutions for common
property resource management. Journal of Theoret-
ical Politics, 4(3), 247–281.
McSweeney, K. (2002). Who is ‘‘forest-dependent’’?
Capturing local variation in forest-product sale,
Eastern Honduras. Professional Geographer, 54(2,
May), 158–174.
Nagendra, H., Southworth, J., & Tucker, C. (2003).
Accessibility as a determinant of landscape transfor-
mation in western Honduras: Linking pattern and
process. Landscape Ecology, 18(2), 141–158.
NRC (National Research Council) (1986). In Proceed-
ings of the conference on common property resource
management. Washington DC: National Academy
Press.
Neumann, R. P. (1998). Imposing wilderness: Struggles
over livelihood and nature preservation in Africa.
Berkeley: University of California Press.
Ostrom, E. (1990). Governing the commons: The evolu-
tion of institutions for collective action. New York:
Cambridge University Press.
Ostrom, E. (2003). How types of goods and property
rights jointly affect collective action. Journal of
Theoretical Politics, 15(3), 239–270.
Ostrom, E. (2005). Explaining institutional diversity.
Princeton, NJ: Princeton University Press.
Poteete, A. R., & Ostrom, E. (2004). Heterogeneity,
group size and collective action: The role of institu-
tions in forest management. Development and
Change, 35(3), 437–461.
Ragin, C. C. (1997). Turning the tables: How case-
oriented research challenges variable-oriented re-
search. Comparative Social Research, 16, 27–42.
Rangan, H. (1997). Property vs. control: The state and
forest management in the Indian Himalaya. Devel-
opment and Change, 28(1), 71–94.
Ribot, J. C., & Peluso, N. L. (2003). A theory of access.
Rural Sociology, 68(2), 153–181.
COMMUNITY FOREST GOVERNANCE IN THE INDIAN HIMALAYA 165
Rocheleau, D., & Edmunds, D. (1997). Women, men,
and trees: Gender, power, and property in forest and
agrarian landscapes. World Development, 25(8),
1351–1371.
Rudel, T. K., Bates, D., & Machinguiashi, R. (2002). A
tropical forest transition: Agricultural change, out-
migration, and secondary forests in the Ecuadorian
Amazon. Annals of the Association of American
Geographers, 92(1), 87–102.
Sarin, M. (1995). Regenerating India’s forests: Recon-
ciling gender equity with joint forest management.
IDS Bulletin: Institute of Development Studies, 26(1),
83–91.
Scott, J. C. (1976). The moral economy of the peasant.
New Haven: Yale University Press.
Southworth, J., & Tucker, C. (2001). The influence of
accessibility, local institutions, and socioeconomic
factors on forest cover change in the mountains of
western Honduras. Mountain Research and Develop-
ment, 21(3), 276–283.
Tiffen, M., Mortimore, M., & Gichuki, F. (1994). More
people–less erosion: Environmental recovery in Kenya.
London: John Wiley.
Varughese, G., & Ostrom, E. (2001). The contested role
of heterogeneity in collective action: Some evidence
from community forestry in Nepal. World Develop-
ment, 29(5), 747–765.
Virtanen, P. (2002). The role of customary institutions
in the conservation of biodiversity: Sacred forests
in Mozambique. Environmental Values, 11(2),
227–241.
Wade, R. (1994). Village republics: Economic conditions
for collective action in South India. Oakland: ICS
Press.
Wily, L. A. (2001). Reconstructing the African com-
mons. Africa Today, 48(1, Spring), 77–99.
166 WORLD DEVELOPMENT
... As already noted, these forests are formally owned by the Government of Nepal but often are used by communities and may effectively be open access. The outcome variables are indices of technical and governance aspects of forest collective action and a forest outcome index, which are composite measures that summarize underlying aspects of forest management and outcomes, and are common in analysis of forest collective action (e.g., Agrawal & Chhatre, 2006;Anderson & Agrawal, 2011;Beyene et al., 2016). These indices, which are constructed based on a total of eight questions posed to members of both CF and Non-CF groups, also allow us to combine disparate aspects of collective action, communicate the forest management situation in a holistic way, and perform statistical tests. ...
... A significant portion of groups also restrict the tools and methods that households can use to cut fuelwood (e.g., axes are not allowed and/or only a large knife called a khukuri can be used). 6 Based on the literature, 15 focus group discussions, and two expert consultations, we identified 14 observable, potential confounders, which could affect whether communities opt to participate in the CF Program. These are shown in the first column of the first table in the Appendix. ...
Article
Forest tenure reform and devolution are taking place around the world, especially in lower-income countries, primarily to support environmental and livelihood goals. However, whether forest devolution to communities results in improved forest collective action is not fully understood. In this paper, we assess the effect of Nepal’s formal Community Forestry (CF) Program on forest collective action and outcomes. We utilize data from a nationally representative random sample of 1300 households who manage and use 130 forests. By using covariate matching, we demonstrate a strong positive effect of the CF program on forest outcomes and more active forest collective action. Our results are generally robust to the presence of unobservable confounders and are consistent at the national level and across geographic and topographic regions, forest sizes, forest-managing population structures, and forest management durations.
... Cette approche renforce la cohésion sociale et facilite la coopération. Des études suggèrent que lorsque les populations locales sont habilitées à gérer les ressources de leur territoire de manière autonome, elles étaient capables de le faire de manière durable(Agrawal, 2001;Agrawal & Chhatre, 2006). De plus, Berkes & Adhikari, (2006) ont montré que la décentralisation des pouvoirs de gestion des ressources naturelles contribuait à l'autonomisation des femmes et des jeunes. ...
Thesis
Full-text available
Global biodiversity is under extreme pressure, marked by a significant increase in species extinctions over the last 300 years and a decline in most vertebrates over the last five decades, mainly due to human activities. Crocodilians are also concerned, with 50% of their species categorized as threatened. Consequently, it is essential to improve the effectiveness of conservation programs. This thesis contributes to broadening and deepening knowledge of conservation approaches and population inventory methods, with a particular focus on crocodilians. Through community-based conservation, I emphasize the importance of involving indigenous peoples and local communities in conservation projects and considering their social, economic and environmental viewpoints. Reconciling conservation and development objectives increases the chances of success and sustainability. My work has also led to the development of a standardized method for monitoring crocodilians using drones. This efficient, non-invasive methodology is suitable for crocodilian species found in open environments. This technology, accessible to a wide range of users, including indigenous peoples and local communities, promotes their empowerment and the protection of ecosystems. This work offers new perspectives for conservation by combining community involvement and technological advances, for a more effective, inclusive and sustainable approach.
... While local communities can be effective at monitoring biodiversity, community management may be associated with strong species preference, resulting in declined biodiversity over time (Shahabuddin and Rao, 2010). However, critics of mainstream approaches to CBC claim that well-intentioned, top-down efforts aimed at reinforcing preexisting CBC unintentionally disempower local communities by conceiving them as passive subjects rather than as active participants of conservation projects (Agrawal and Chhatre, 2006;Agrawal, 2001;Agrawal, 2005;Goldman, 2003;Pimbert and Pretty, 1995). CBC offers no singular "one-size-fits-all" approach. ...
Article
Full-text available
As dominant approaches to biodiversity loss and climate change continue to fail in mitigating current socio‐environmental crises, scholars and activists are exploring novel conceptual frameworks to drive transformative change in conservation. Among these, more‐than‐human care has emerged as a concept at the intersection of feminist debates about care and post‐humanist discussions around the more‐than‐human. Although more‐than‐human care is increasingly referenced in the literature, it remains sparsely situated in the Global South. This article investigates the concept's potential to articulate care relations that can foster more plural, respectful, and transformative biodiversity conservation practices in local communities. Our research is situated in an artisanal fishing village in Brazil, where we engaged with both human and non‐human members of the community to engage with their perspectives on more‐than‐human care and its relevance to their local concerns. From our analysis, four key dimensions emerged: the vital and everyday nature of caring; the relationality between humans and non‐humans; reciprocity; and the fostering of flourishing for as many as possible. Our core argument is that the interaction between community practices and care theory enables new perspectives that center on daily and reciprocal care relations often overlooked in mainstream conservation approaches. By situating the conceptual potential of more‐than‐human care within the Global South, we underscore its ability to make the invisible visible and to inspire action for transformative change.
Article
• Nationwide evaluation of community forest conservation and its formalization. • Spatio-temporal analysis of forest losses revealed deforestation pressure. • Deforestation lower in community forests even in deforestation hotspots. • Formalization was not associated with a significant change in deforestation. • High regional differences in effectiveness of community-based conservation. Governments are legally formalizing an increasing number of community forests by sharing and transferring tenure rights over state-owned forestland in an effort to reduce deforestation. However, there has been little evidence on whether their conservation effectiveness could be further strengthened through formalization. In Thailand, the Royal Forest Department began to register community forests in 2000. The objective of this study was to assess the effectiveness of community forests in moderating the impact of deforestation pressures, highlighting the intrinsic ability of communities to protect their forests, and the effects of a legal formalization. In a spatial evaluation approach, statistical matching and fixed-effects models were used to analyze the effect of community-based forest conservation and its formalization on deforestation rates. Each analysis was conducted in provincial areas sampled from northern, north-eastern, and southern Thailand (680 community forests) to compare the impact of varying levels of deforestation pressure over a 14-year period from 2000 to 2014. The large majority of sampled communities protected their forests against substantial deforestation during the entire observation period, with 82 % experiencing less than 1 ha of deforestation and 60.15 % experiencing no deforestation at all. The median relative deforestation rate over this period was 0.21 %, with an interquartile range of 1.82 %. Their efforts reduced the likelihood of forest loss in regions of high deforestation pressure from approximately 30 % to almost zero. In contrast, the threat of deforestation did not significantly change after a formal registration. These findings were similar across different regions despite their biophysical and socio-economic differences. These findings suggest that while community efforts are central to forest conservation, the benefits of formalization in enhancing communal efforts appear to be subtle and thus remain inconclusive in the current context. Registered community forests are still affected by forest encroachment despite their successful conservation efforts. Thus, communities require stronger support from forest officials and local law enforcement agencies in both legal and technical capacities.