institutional factors affecting biophysical outcomes in forest management

Journal of Policy Analysis and Management, Vol. 28, No. 1, 122–146 (2009)© 2009 by the Association for Public Policy Analysis and Management Published by Wiley Periodicals, Inc. Published online in Wiley InterScience (www.interscience.wiley.com)DOI: 10.1002/pam.20405

Abstract

Although there is considerable interest in the impact of diverse policies affecting the biophysical outcomes in forests, gaining a substantial sample over time offorests under different institutional arrangements has been difficult. This articleanalyzes data from 46 forests located in six countries over time. In forests wherepolicies have been adopted for conservation, active monitoring and sanctioning bylocals is associated with positive forest conditions. Forests that allow user groupharvesting, perhaps counterintuitively, are also shown to be associated with positive forest conditions. However, conditions in community-managed forests arenot statistically different from government- or privately managed forests. Thisimplies that local communities can play an important role in achieving positiveforest conditions but that full management responsibilities need not be given toachieve these results. © 2009 by the Association for Public Policy Analysis andManagement.

INTRODUCTION

Policy scholars have long examined common pool resource (CPR) systems and thegreat variety of institutions used to govern them (National Research Council, 1986,2002; Ostrom, 1990, 2005). A CPR is a resource for which it is relatively costly toexclude users but the use of the resource is subtractable or rivalrous in consump-tion (McKean, 2000; Ostrom, Gardner, & Walker, 1994). Early scholars demon-strated that in open access, commons users will deplete the resource (Hardin, 1968).Hardin proposed that this tragedy could be avoided by either privatization or exter-nal government regulation.

Ostrom (1990) has argued that many CPRs are not open access and that commu-nities often, although not always, create institutions to prevent tragedy. There is nowmuch policy debate on how to incorporate communities into resource governance.Some argue that policies should enable local communities to manage suchresources (Uphoff, 1998; Wade, 1994). A whole literature on community-based nat-ural resource management (CBNRM) has emerged from this paradigm (see Kellertet al., 2000). Others are quite skeptical of this approach (Agrawal, 2001b; Blaikie,2006). Some argue that national preserves are the only way to save forests and otherbiologically diverse ecosystems (Terborgh, 1999). Still others advocate for someform of privatization (Anderson & Leal, 2001). In recent work, Ostrom, Janssen, andAnderies (2007) argue that there is no panacea for such problems and that localinstitutions need to be adapted that fit the particularities of a situation.

Eric A. ColemanInstitutional Factors AffectingBiophysical Outcomes in ForestManagement

Institutional Factors Affecting Biophysical Outcomes in Forest Management / 123

Journal of Policy Analysis and Management DOI: 10.1002/pamPublished on behalf of the Association for Public Policy Analysis and Management

Many types of policies have been implemented to incorporate local communitiesinto forest management activities. This paper sheds light on the effects of some ofthese policies on the biophysical outcomes in forests. It has been difficult to gain asubstantial sample over time of CPR outcomes under different institutionalarrangements to test the relative efficacy of different policies. The data setemployed in this paper, consisting of over-time observations in 46 forests, offers aunique opportunity to analyze the effects of different institutions on forest out-comes. I find that the forest management type (private, government, or local) doesnot significantly affect biophysical outcomes in the models analyzed. Instead, I findthat monitoring regularity, the existence of rights to local users to harvest from theforest, and socioeconomic conditions are associated with positive forest conditions.

For the remainder of this section, I review the literature on forest policy and mon-itoring and sanctioning. In the next section, I outline a strategy to analyze forestoutcomes across the sites and over time. In the third section, I propose and estimatean empirical model of forest conditions dependent on a number of social, eco-nomic, and biophysical variables. In the final sections, I report and discuss theestimation results and conclude with policy recommendations and a discussion ofthe shortcomings of the paper.

Forest Policy

A large amount of research examines the types of rules that might be used to effec-tively manage long-enduring resource systems. Much empirical evidence existsexamining rules and comparing them to management outcomes (Agrawal, 2001a;Dietz, Ostrom, & Stern, 2003; Ostrom, 1990). Recently, Pagdee, Kim, and Daugherty(2006) conducted a review of the existing literature in forestry and concluded thatamong the institutional variables identified as important, “effective enforcement”has one of the strongest associations with success in forest management. Thisconfirms work by Gibson, Williams, and Ostrom (2005) that argues regular sanc-tioning and monitoring of rules is a necessary condition for successfulmanagement.

Many empirical studies link institutional variations with variation in forest out-comes. For example, Gibson, Lehoucq, and Williams (2002) assess a number of biophysical criteria in Guatemalan forests to show that privatized forests do notoutperform community forests. Agrawal and Chhatre (2006) look at forests in Nepaland relate attributes of communities and their forest institutions to explain varia-tions in forest outcomes. A book edited by Gibson, McKean, and Ostrom (2000)contains many empirical studies. Varughese (2000) assesses the role of populationgrowth and many other variables on the willingness to participate in collectiveaction and the subsequent effects on forest conditions. Agrawal and Goyal (2001)examine institutions in India and find that large forest councils are more success-ful than smaller councils.

Recent work by Gibson, Williams, and Ostrom (2005) and Hayes and Ostrom(2005) attempted to compare forest conditions across diverse ecologies and coun-tries. These papers do not use physical measures of forest conditions such as basalarea or species diversity, but instead focus on (subjective) assessments of forest con-ditions by expert foresters and local community members. Ostrom and Nagendra(2006) recently extended this analysis by assessing changes in physical measuresover time. They then compare physical changes across diverse ecologies and coun-tries. However, the authors looked only at differences in mean characteristics acrossinstitutional types. In this paper, I examine the effects of institutional variation on biophysical outcomes in diverse settings, while controlling for potentially con-founding variables.

124 / Institutional Factors Affecting Biophysical Outcomes in Forest Management


User Group Monitoring and Sanctioning

Monitoring and sanctioning has been shown to be a key element in explaining suc-cessful CPR management (Gibson, Williams, & Ostrom, 2005; Hayes & Ostrom,2005; Ostrom & Nagendra, 2006; Pagdee, Kim, & Daugherty, 2006). There are twoimportant issues, however, in assessing the role of monitoring and sanctioning inforestry. First, what type of monitoring and sanctioning are most likely to be themost effective? Second, how effective can monitoring and sanctioning be?

Some research has shown that the source of monitoring and sanctioning isimportant. Externally imposed rules and monitoring institutions have often failedeven if those rules are shown to be mathematically efficient in a general CPR model(Cardenas, Stranlund, & Willis, 2000; Cardenas, 2004; Ostrom, 1990; Ostrom, Gardner,& Walker, 1994). For example, Cardenas, Stranlund, and Willis (2000) ran a seriesof field experiments in rural Colombia. Subjects were allowed to allocate resources(that is, how many months to harvest in a forest) and then received payoffs accord-ing to their share of the commons production. In five designs, subjects were allowedto make some rule about their investment in the commons. The subjects were thentold there was a probability of 1/16 that a subject would be monitored.1 When sub-jects were allowed to make their own agreements, the outcomes were at least asgood as in the rule-free environment, and often much better. Cardenas, Stranlund,and Willis (2000) then ran a series of five more experiments where the subjects weregiven an optimal quota rule for the resource system with the same probability ofmonitoring. Subjects actually increased their harvesting levels in these experi-ments.2 Cardenas, Stranlund, and Willis (2000) conclude that they “have presentedevidence that the fundamental reason for the poor performance of external controlis that it crowded out group-regarding behavior in favor of greater self-interest” (p. 1731).

In this paper, I examine the role of user group monitoring and sanctioningbecause, following Cardenas, Stranlund, and Willis (2000), it is expected to have the greatest potential effect. This augments the existing literature by reporting theeffects of user group monitoring across a number of forest CPRs in field settings toconfirm what Cardenas, Stranlund, and Willis (2000) found in field experiments. Ifuser group monitoring can improve forest conditions, then external governmentsmay, to some extent, rely on local efforts to devise and monitor their own rulesadapted to local conditions. This allows governments to allocate resources moreeffectively while at the same time preserving user group efforts of collective actionthat might otherwise be crowded out. On the other hand, government monitoringand sanctioning might still play an important role in “monitoring the monitors” andbacking up local efforts (Lejano & Ingram, 2007; Ostrom, 1990).

The other important issue is measuring the effects of monitoring and sanctioning,and its relative efficacy vis-à-vis other institutional factors. As discussed, there is afair amount of literature testing the effects of monitoring and sanctioning in com-mons problems in laboratory settings.3 Many case studies also conclude that mon-itoring and sanctioning are important.4 Why do monitoring and sanctioning playsuch a prominent role in common pool resource management? Part of the problemof analyzing monitoring and sanctioning is that it presents a second-order collec-tive action problem (Ostrom, 2005; Panchanathan & Boyd, 2004). Monitoring and

1 The experimenter would roll a die to determine whether there would be monitoring. If the die was even,then the experimenter randomly selected one of the eight subjects to monitor. Thus, the probability ofbeing monitored was 1/16.2 This confirms the hypothesis of Frey and Oberholzer-Gee (1997) and Frey (1997) that externallyimposed rules can crowd out intrinsic or community-based norms and rules and lead to worse outcomes(Ostrom, 2005).3 See Ostrom, Gardner, and Walker (1994), Cardenas, Stranlund, and Willis (2000), and Cardenas (2004).4 See Pagdee, Kim, and Daugherty (2006).


sanctioning are public goods, benefiting the group at personal costs to the individ-ual. To suggest that common pool resource problems can be solved with monitor-ing and sanctioning merely introduces a second-order dilemma. Still, many groupseffectively engage in monitoring and sanctioning activities, so these efforts need tobe taken seriously.

The decision to engage in monitoring and sanctioning is often presented in termsof one-shot decisions. If instead we think of the willingness to engage in monitor-ing and sanctioning in terms of the reputations that can be built among groupmembers over time in repeated games, then it may be possible for humans to suc-cessfully commit to such actions. There are now cogent theoretical arguments sug-gesting that users can credibly engage in costly monitoring and punishment in suchrepeated game situations (Milinski, Semmann, & Krambeck, 2002; Boyd et al.,2003; Panchanathan & Boyd, 2004; Hauert et al., 2007). In addition, there is empir-ical evidence from experimental laboratory work indicating that subjects do, in fact,engage in costly punishment (Fehr & Gachter, 2000, 2002; Ostrom, Walker, & Gardner,1992; Sefton, Shupp, & Walker, 2007) and also field experiments in developingcountries to confirm this (Henrich et al., 2006).

This literature also shows that monitoring and sanctioning are effective. In arecent paper, Sefton, Shupp, and Walker (2007) conduct a public goods game wherecooperation is measured each round by the number of tokens that subjects allocateto a group fund. The authors had three treatments where, in the latter rounds of the game, subjects could punish others in their group at a cost to themselves,reward others at a cost to themselves, or reward and punish. The authors found thatall of the treatments increased cooperation above the baseline treatment where nosanctions or rewards were allowed. The CPR experimental literature also indicatesthat both pecuniary and nonpecuniary measures can induce cooperation on thecommons (Ostrom, Walker, & Gardner, 1992). There is also much fieldwork docu-menting instances and types of social shaming that induce community members tocomply with institutional requirements (Ostrom, 1990). Monitoring exposes rulebreakers to these social consequences. Where there is no monitoring, even if sanc-tioning rules exist, they are not enforced because rule breakers cannot be caught. It is also likely that where there is no monitoring, sanctioning mechanisms may beweak (either small fines or little ability for a community to effectively shame amember).

COMPARING OUTCOMES ACROSS FORESTS

As mentioned, researchers have recently begun to compare community forest con-ditions under different management schemes and institutional types (Gibson,Williams, & Ostrom, 2005; Hayes & Ostrom, 2005; Ostrom & Nagendra, 2006). Iextend this research in three important ways. First, this paper does not solely relyon subjective measures of forest outcomes. Gibson, Williams, and Ostrom (2005)and Hayes and Ostrom (2005) relied on expert foresters’ evaluation of the forest. I use biological outcome measures taken from sampled forest plots at a number ofcommunity forests throughout the world. The outcomes measured are binary vari-ables indicating significant change over time in Basal Area, which is a strong indi-cator of biomass (Husch, Beers, & Kershaw, 2003), and the Shannon DiversityIndex (a measure of species diversity).

Second, I control for a number of potentially intervening variables that priorresearch does not. Gibson, Williams, and Ostrom (2005), Hayes and Ostrom (2005),and Ostrom and Nagendra (2006) all assess differences in evaluation criteria by look-ing at the distribution of outcomes by management type and user group monitoring.The differences these authors find may be due to the variables they are examining ordue to a host of other intervening factors. In this paper, I control for a number ofbiophysical and socioeconomic variables, as described in the next section.




Third, I assess biophysical outcomes according to conservation policies of the for-est. Andersson and Gibson (2007) have convincingly argued that the outcome mea-sures used to assess institutional arrangements have often not taken into accountpolicy goals. It may make no sense to declare a community-managed forest as a fail-ure when a forest is commercially logged, if the goal of the manager is to generaterevenue. Thus, if we are to assess forest conditions, we must be sure that the publicpolicies aim to achieve the outcome measures. Of course, in practice, this is difficultto do. Few public policies explicitly set goals to increase biomass or species diver-sity. For the purposes of this paper, I assume that forests that restrict harvesting doso implicitly with the goal in mind to increase Basal Area and the Shannon DiversityIndex. The important comparison for this paper, then, is not whether community-managed forests have more basal area than government-managed forests. A community-managed forest may have less basal area because the policy goal is toharvest from the forest. This can hardly be called a management failure. Theimportant question is whether community-managed forests have more basal areathan government-managed forests conditional on policies to restrict harvesting.

EMPIRICAL ANALYSIS

Data

I use data gathered from the International Forestry Resources and Institutions(IFRI) program. IFRI is an effort by a worldwide network of colleagues to system-atically assess social-biological forest systems that affect local communities.5Research teams are deployed to sites where a forest is used by a local community.They assess local forestry management, society, and biology of the forest.Researchers at each site follow the same set of data-gathering protocol for both bio-logical and social data. This ensures a relatively consistent way to compare dataacross forests. Over 200 site visits have been conducted by IFRI colleagues.

In this paper, I use data from 46 of these forests. Only data from forests that havebeen visited twice were used. These data come from six countries. The number ofsites from each country is reported in Table 1. These forests do not represent a ran-dom sample of all forests in the world (it is hard to imagine a process for such selec-tion). However, IFRI sites are not selected on the basis of the dependent variable—biomass, species diversity, or any comparable outcome. The criteria for selection ofIFRI sites vary across countries, time, and IFRI network colleagues’ research inter-ests; however, sites are nonetheless selected based on values of independent vari-ables. For example, often sites are selected to include one community-managed

Table 1. Distribution of sites by country.

Country Frequency Percent

GUA 1 2.17IND 5 10.87KEN 3 6.52NEP 11 23.91UGA 20 43.48USA 6 13.04

Total 46 100

5 IFRI was originally started at Indiana University, Bloomington, and has partnered with research cen-ters in countries around the world. The coordination of the IFRI program was, until recently, located atthe Workshop in Political Theory and Policy Analysis at Indiana University but is now shifting to theSchool of Natural Resources at the University of Michigan.



forest, one government-managed forest, and one privately managed forest in a localecology. None of the sites were sampled because they stood out as being a particu-larly successfully managed forest.

The data do not restrict the range of possible values for the dependent variablebut may constrain the values of the independent variables upon which the caseshave been selected. The popular inference from such a design is that the parameterestimates hold for the sample, but not the population. However, it is more correctto say that the results are generalizable to cases with similar variation in the inde-pendent variables under the sampling procedure (King, Keohane, & Verba, 1994,ch. 4). Thus, care should be taken when generalizing these results to forests thathave values of independent variables outside of the range of this study. Summarystatistics for all variables are found in Table 2.

Outcome Variables

Biological data are obtained by randomly selecting forest plots in each forest. Men-suration is then performed to determine speciation, density, tree height and diam-eter, as well as other indicators. About 30 plots are selected for each forest; however,the number of plots in each forest varies considerably due to the variability foundin diverse forests. The specific number of plots selected for each site depends on aperformance curve. A sufficient number of plots are selected so that the variationin mean estimates of a key variable (such as basal area) becomes very small if addi-tional plots are added. Determining the number of plots depends on the biophysi-cal variability and size of the forest as well as the comparisons to be made with thedata.

The most frequent number of plots selected for the forests used in this study is 30.Figure 1 shows a histogram of the number of plots selected in each forest for eachvisit. Because there are 46 forests and there are two visits per forest, the total num-ber of visits is 92. The figure clearly shows that the most frequent number of plotschosen for comparison is 30. However, many forests have fewer than 30 plots andfive forests have fewer than 10 plots. This is because there is virtually no biophysi-cal variation at forests in these sites, making more plots unnecessary. For example,two of the five forests are from a first and second visit to a site in India where theforest has been completely denuded. Thus, selecting more forest plots for mensu-ration would not improve the estimate of basal area for the forest, which is 0.

Table 2. Summary statistics.

Variable Mean Std. Dev. Min. Max.

Basal Area 0.543 0.504 0 1 46Shannon Index 0.630 0.488 0 1 46Forbid Harvest 0.326 0.474 0 1 46Monitor and Sanction 0.391 0.493 0 1 46Private 0.174 0.383 0 1 46Government 0.543 0.504 0 1 46Community 0.283 0.455 0 1 46HH Per Forest Hectare 10.587 55.446 0 377 46Forest Value 5.163 1.376 2 8 46

Socioeconomic Conditions 0 1.838 �1.183 4.664 46Prop Rights 5.017 1.664 3.54 9.04 46GDP Per Capita PPP 5483.678 11,815.81 576.3 36,837.26 46Wage 5.904 13.361 0 40 46Fuelwood Dependency 60.02 38.025 0 100 46



The outcome variables used in this paper, Basal Area and the Shannon DiversityIndex, are constructed for each forest in the following manner: First, estimates foreach of these measures is calculated for each plot during each visit (adjusting for size of the plot if there has been a change); second, a comparison of means t-testis performed to assess whether each variable has changed during the intervalbetween site visits (about five years); and finally, each measure is coded as a binaryvariable, taking the value of 1 if the measure has stayed the same or increased fromthe previous visit, or a value of 0 if the measure has decreased since the previousvisit. In coding the variables, I use a one-sided t-test at a� 0.10 level of significance.

Basal Area

Basal Area is calculated from measurements of stem diameter, d. A researcherenters a randomly selected plot in the forest and measures all trees above 10 cen-timeters. Stem diameter then is calculated at breast height. Generally, researchersuse diameter tapes for this measure.6 Basal area for tree t is then calculated asBasalAreat � p . (d2/4). The basal area for the entire plot i is then calculated as

where Ti is the total number of trees in the plot. Basal area is often used to indicatebiomass, which plays a “role in biodiversity, global climate change, carbon seques-tration, wildlife habitat evaluation, and forest fuel assessment” (Husch, Beers, &Kershaw, 2003, 236–238).

t � 1

Ti

BasalArea � i BasalArea t

Figure 1. Histogram of the number of plots randomly selected for each site and foreach visit. IFRI researchers most frequently selected 30 plots for each visit. Five siteshad fewer than 10 forest plots. This is because there was little variability in biophys-ical characteristics of these forests (they were completely denuded), and selectingmore plots would not substantially improve the estimates of biophysical variables.

6 See Husch, Beers, and Kershaw (2003, pp. 85–99) for a discussion on calculating stem diameter.


Shannon Diversity Index

The Shannon Diversity Index is a measure of the diversity of tree species in the plot.In order to derive the index, I first define two characteristics of the forest plot. Firstis the total count of all tree individuals from all species, Ti. Second, let us index the species found in each plot i by si � 1, . . . , Si, where Si is the total number ofunique species. Species abundance is calculated as the total number of individualsin the plot for each species, denoted Ais. The Shannon Diversity Index is thencalculated as

(1)

The Shannon Diversity Index measures tree diversity by weighting the total num-ber of unique species by their relative abundance in the plot. The more uniquespecies there are, the higher will be the index measure. More equal distribution ofspecies within the plot also increases the index measure.

Institutional and Policy Variables

The primary interest of this paper is to assess differences among community andgovernment forests, the effects of socioeconomic conditions, and the effects of usergroups monitoring and sanctioning. To assess these effects, it is imperative to con-trol for the conservation policies in each forest. The outcome measures mentionedin the previous section have little meaning if not attached to some policy goal.

Management Type

Three dichotomous variables are created indicating the type of forest management.Community-managed forests are those for which a group of local users are the pri-mary decision makers. Government-managed forests are those for which a centralgovernment agency has the responsibility for making management decisions.Examples include national parks and national forests. Privately managed forestsare those for which an individual or small group of individuals make managementdecisions in a formal partnership. About 54 percent of the forests in the sample aregovernment-managed, 17 percent are privately managed, and 28 percent are com-munity managed.

Monitoring and Sanctioning

User group monitoring and sanctioning is a binary variable indicating the fre-quency of user group monitoring and sanctioning in the forest, taking a value of 1if there is regular user group monitoring and sanctioning (seasonal or year round)and a value of 0 if there is occasional or no activity. There may be many differentuser groups that use a particular forest. There may be timber harvesters, mushroomcollectors, and wildlife viewers. The monitoring variable indicates the maximumlevel of monitoring and sanctioning that any user group conducts.

In addition, sanctioning activities may be pecuniary or nonpecuniary. Nonpecu-niary sanctions include verbal chastisement, forced public apology, or other socialshaming. Pecuniary sanctions may include fines, in-kind contributions of materialsor labor, or restrictions on future harvesting. Further, in many forests the sanctionvaries by the severity of the violation and the number of times a violator has bro-ken a rule. This variable is hypothesized to be positively related to the outcome vari-ables, as discussed in the introductory section.

S � 1

S

i

i

Shannon � � iisAiT

isAiTln ( )




Forbid Harvesting

Approximately one-third of the forests have policies such that no user group has aright to harvest trees. The dummy variable for forbid harvesting takes the value of1 if harvesting of trees is forbidden to all user groups and a value of 0 if harvestingis permitted to at least one user group. Note that all private forests take a value of0 because harvesting is at least permitted by the owner of the forest, even if theowner does not exercise that right. Therefore, the efficacy of privately managedforests cannot be measured with this variable.

This variable represents a de facto rule. If a forest is managed to conserve treesyet harvesting is still permitted to at least one user group, then that forest is notconsidered to have “forbid harvesting” policies. The outcome measures have themost relevance in these forests, so in the following discussion I discuss the resultsin terms of conditioning on the fact that a forest does not allow harvesting. I do notsimply eliminate forests that do not forbid harvesting, because of the efficiencygains from the additional data supplied by those observations.

Socioeconomic Conditions

To control for differences in socioeconomic conditions across sites, I calculate a sin-gle index of socioeconomic conditions from four variables: property rights, grossdomestic product adjusted for purchasing power parity, the average daily wage rateat the site, and dependency on fuelwood. Analysis of these variables suggests theyare highly correlated. To create the index, I perform principal components analysis(PCA) on these variables.7 PCA is a statistical technique used for data reduction. Itdecomposes multivariate relationships to a smaller number of core underlying fac-tors by solving for data eigenvectors (that is, principal components) that accountfor as much of the variation as possible. In this case, I solve for the first principalcomponent of the four socioeconomic variables.

PCA results indicate that the first principal component accounts for more than 84percent of the underlying variation in the four variables. The index for socioeco-nomic conditions is formed by weighting each variable according to its estimatedloading on the first principal component. Property rights, per capita gross domes-tic product adjusting for purchasing power parity, and wage rates are positivelyrelated to the socioeconomic conditions index, whereas the variable for dependenceon fuelwood is negatively related.8 Each of the socioeconomic variables is discussedbelow.

Property Rights

Strength of property rights is measured on a scale of 0 to 10 and is taken from the Gwartney and Lawson (2006) economic freedom indicators; it is weighted to theyear nearest to the time of visit.9 It is positively associated with socioeconomic con-ditions. The variable comes from the “Legal System and Property Rights” classifi-cation of that survey, which is compiled from five indexes taken from the WorldEconomic Forum Global Competitiveness Report, the World Bank GovernanceIndicators Project, and the PRS Group International Country Risk Guide. These

7 See Kaplan (2000, ch. 3) for a review of this methodology.8 Results from the principal components analysis are available upon request from the author.9 The value is included for each year where the survey has data available for the year of the visit. How-ever, in the early years of the property right index, data was only collected every five years. In these cases,property rights are assumed to be a weighted average of the year nearest to the time of visit. For exam-ple, if the visit took place in 1999 and property rights data were only available for 1995 and 2000, thenthe data would be entered as 0:8 _ PropRight2000 � 0:2 _ PropRight1995.



indices are (1) judiciary independence, (2) impartial courts, (3) protection of prop-erty rights, (4) military in politics, and (5) law and order.10

Does the enforcement of national property rights reflect the situation at IFRI sites?There is undoubtedly variation in property rights enforcement across geographiclocalities, socioeconomic status, and ethnicity within most countries. Property rightsas measured at the national government level, however, are particularly importantgiven that they represent a higher level conflict resolution mechanism to resolve dis-putes at the local level, an important design principle developed by Ostrom (1990).If these indices are high, it represents the fact that higher-level conflict resolution isless likely to be arbitrary or capricious. For example, Culas (2007, p. 435) usesstrength of contract enforcement, as measured with a country-level indicator, toshow that “improvement in institutions that empower people through secure prop-erty rights for forest resources and better environmental policies will ultimatelyreduce the pressure on resources and lead to better conservation of forestland.”

Furthermore, De Soto (2000) argues that if formal property rights are welldefined and enforced, owners are both accountable for their use of assets and arebetter able to realize the benefits from their investments. State policies that bolsterformal property rights help ensure avenues to track and defend property rightswhen they are violated. According to De Soto (2000, p. 54), strong formal propertyrights shift legitimacy of rights to the resource from “local communities to theimpersonal context of law.” Thus, state enforcement of property rights has a dis-tinctively important role in providing incentives for resource users.

GDP Per Capita Purchasing Power Parity (PPP)

Per capita Gross Domestic Product is measured adjusting for purchasing powerparity. It is taken from the World Bank World Development Indicators and isreported in nominal USD and measured at time of visit. It is positively related tosocioeconomic conditions.

Wage

Average daily wage rates are measured in nominal USD. Nominal exchange ratesare calculated at the time of visit to the forest as reported by IFRI teams. Wage rates arepositively related to socioeconomic conditions.

Fuelwood Dependency

Fuelwood dependency represents the fraction of forest user groups’ needs for fuel-wood supplied by the forest. Andersson and Gibson (2007) use fuelwood depen-dency as a proxy for socioeconomic conditions in their analysis of 29 forests inBolivia. The authors argue that this variable gives an indication both of the generallevel of development and information about the importance of forest products tousers groups. This variable was also used by Gibson, Williams, and Ostrom (2000)in their analysis of community forestry in Ecuador. This variable is negativelyrelated to socioeconomic conditions.

Control Variables

Households Per Forest Hectare

The number of households that use the forest is expected to be negatively associatedwith the outcome measures. This is for three reasons: First, more households imply

10 See Gwartney and Lawson (2006) for more details regarding the precise composition of this variable.



a greater demand for forest services; second, more households generate more com-petition for forest resources; and third, as the number of households increases, theprobability of successfully acting collectively to restrain use diminishes (Olson, 1965).There is a long history, and much empirical evidence, linking human population pres-sure to environmental conditions (Cropper & Griffiths, 1994). What constitutes ahousehold is defined by the local culture of the community of the forest being stud-ied. The absolute number of households is less important than the number of house-holds per forest hectare, which represents a potential strain on the resource.

Forest Value

The forest value is taken by combining two measures in the IFRI data set. At theconclusion of each site visit an expert forester is asked to rate the forest on a num-ber of criteria. Two of those criteria are the subsistence value of the forest and thecommercial value of the forest. Each of these measured is on a scale of 1 to 5. Thequestion is, In your best judgment, given the topography and biophysical zone inwhich this forest is located, how would you judge the (commercial or subsistence)value of the forest: 5—substantially above normal, 4—above normal, 3—normal,2—below normal, or 1—substantially below normal? Forest value is taken as thesum of commercial and subsistence value and thus has a maximum value of 10 anda minimum value of 2.11

Agrawal (2001a) argues that the value of the resource and market pressure facedby resource managers needs to be integrated into models of forest extraction. Regevet al. (1998) show that impacts of market forces can endanger the sustainability ofresources by inducing resources users to harvest above biophysically sustainablelevels. The value of the forest is an important alternative explanation to the effectsof monitoring and sanctioning in particular. The value of the forest might bothincrease the level of forest harvesting and increase the incentive to engage in mon-itoring and sanctioning; thus, it is important to control for the value of the resourceto ensure that monitoring and sanctioning are causing changes in the outcome vari-ables and not differences in forest value across sites.

The Model

I estimate a binary logit model of forest conditions, where the probability of success(equivalent or increased measures on the outcome variables) depends on the inde-pendent variables (the institutional, policy, socioeconomic conditions, and controlvariables) described in the previous sections. For notational convenience, let usdefine y as the binary outcome variable, indicating significant change between vis-its, and x as a vector of independent variables.

To estimate the model, I take measures of each independent variable at the timeof the first site visit and use this to explain any significant change, y, so that

Pr(y � 1) � �(x1b) (2)

where x1 represents the logistic cumulative distribution function (see Cameron &Trivedi, 2005) and each x is evaluated at the first visit. This mitigates any endo-geneity concerns that monitoring, for example, may be an effect rather than a causeof forest conditions. It is unlikely that a change in future forest conditions willcause forest monitoring retroactively.

11 It would be preferable to use data on price of forest products; however, these forests produce multipleforest products, some of which have market prices and some of which do not. In addition, the method-ology of weighting the prices for each product is uncertain. The value of the forest is taken from the firsttime period, as described in the next section, thus mitigating endogeneity problems of looking at totalvalue rather than just price.



RESULTS

The results of six different logit regression estimations for each outcome variableand their model diagnostics are reported in Tables 3 and 4. Because each of thesocioeconomic variables is highly correlated, separate models using different mea-sures of socioeconomic conditions are estimated: Model 1 uses the property rightsindex; Model 2 uses GDP per capita adjusted for purchasing power parity; Model 3uses average daily wage rates; Model 4 uses fuelwood dependency; Model 5 uses thecomposite index of socioeconomic conditions as described in the previous section;and Model 6 does not control for socioeconomic conditions in any way. The esti-mated standard errors are reported in parentheses and are clustered by country.

Each continuous variable in the models (each socioeconomic variable, house-holds per forest hectare, and forest value) are standardized to have a mean of 0 anda standard deviation of 1. Such a transformation helps mitigate any potential prob-lems of collinearity among the variables. The largest Variance Inflation Factor (VIF)

Table 3. Logit regression results for Basal Area.

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Monitor and Sanction 2.008*** 2.027*** 1.559* 1.933*** 1.636** 2.113***(0.78) (0.66) (0.84) (0.68) (0.73) (0.82)

Private 1.354 1.975* 0.984 1.251 1.177 1.360(0.86) (1.02) (0.86) (1.01) (0.87) (0.87)

Government 1.497 2.410*** 1.407 1.689 1.645 1.424(1.13) (0.81) (1.01) (1.32) (1.07) (1.12)

Forbid Harvest 1.033 1.099 0.863* 1.317** 1.092** 1.015(0.65) (0.70) (0.52) (0.62) (0.54) (0.68)

HH Per Forest 2.852 3.361 3.282 4.118 3.791 2.621Hectare

(3.23) (3.59) (3.33) (3.60) (3.56) (3.30)Forest Value 0.240 0.108 0.111 0.062 0.094 0.265

(0.22) (0.20) (0.23) (0.31) (0.25) (0.23)Prop Rights 0.089

(0.28)GDP Per Capita PPP 0.001

(� 0.00)Wage 0.101**

(0.05)Fuelwood Dependency 0.018***

(0.01)Socioeconomic Conditions 0.535**

(0.24)Constant 1.330 2.424** 0.884 0.456 0.455 0.937

(1.59) (1.23) (1.07) (1.05) (1.10) (1.05)

N 46 46 46 46 46 46

McFadden’s R2 0.184 0.264 0.224 0.225 0.219 0.183McKelvey & Zavoina’s R2 0.740 0.974 0.808 0.847 0.833 0.710

Log-Likelihood 25.875 23.331 24.595 24.567 24.768 25.921AIC 61.750 54.662 59.190 59.135 59.537 61.842BIC 70.893 61.977 68.334 68.278 68.680 70.985Mean VIF 2.364 1.685 1.668 1.841 1.567 1.420

Coefficients standardized for continuous variables. Standard errors, clustered on countries, in parentheses.Two-tailed hypothesis tests: *p � 0.10, **p � 0.05, ***p � 0.01.



of any variable in any of the models is 5.21 (for the property rights variable in Model1 for Basal Area); this is well below the accepted threshold of 20 suggested byGreene (2003, p. 58). The mean VIF of all variables is reported at the bottom ofTables 3 and 4.

Model Specification

A number of scalar model statistics are reported to compare the six models esti-mated for each outcome variable. These are found at the bottom of Tables 3 and 4.The McFadden R2 is calculated as 1 minus the ratio of the model log–likelihood tothe log-likelihood of an intercept-only model. When comparing results across mod-els for the same dependent variable, a higher measure indicates that the model hasa greater likelihood value. McKelvey and Zavoina’s R2 reports the fraction ofpredicted model variance to total variance given the assumed variance of the logistic

Table 4. Logit regression results for Shannon Diversity Index.

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Monitor and Sanction 1.981*** 1.614*** 1.538*** 1.946*** 1.638*** 2.040***

(0.49) (0.33) (0.36) (0.72) (0.41) (0.56)Private 0.195 0.185 0.203 0.239 0.041 0.205

(1.01) (0.75) (0.84) (1.11) (0.94) (0.98)Government 0.173 0.136 0.230 0.171 0.026 0.203

(0.71) (0.71) (0.77) (0.81) (0.68) (0.82)Forbid Harvest 0.528 0.579 0.626 0.327 0.463 0.536

(0.54) (0.53) (0.49) (0.57) (0.54) (0.51)HH Per Forest 0.496*** 0.475*** 0.466*** 0.408*** 0.450*** 0.500***Hectare

(0.15) (0.18) (0.17) (0.12) (0.14) (0.16)Forest Value 0.099 0.181 0.169 0.214 0.175 0.092

(0.29) (0.28) (0.30) (0.33) (0.31) (0.29)Prop Rights 0.045

(0.31)GDP Per Capita PPP � 0.000

(� 0.00)Wage 0.095**

(0.04)Fuelwood Dependency 0.018

(0.01)Socioeconomic Conditions 0.503

(0.45)Constant 0.414 0.848 0.238 0.839 0.046 0.202

(1.25) (0.89) (0.81) (1.65) (1.06) (0.81)N 46 46 46 46 46 46McFadden’s R2 0.181 0.224 0.209 0.220 0.207 0.181McKelvey & Zavoina’s R2 0.334 0.901 0.517 0.401 0.445 0.334

Log-Likelihood 24.812 23.517 23.964 23.645 24.026 24.822AIC 59.625 55.034 57.928 57.290 58.053 59.644BIC 68.768 62.349 67.071 66.433 67.196 68.787Mean VIF 2.364 1.685 1.668 1.841 1.567 1.420

Coefficients standardized for continuous variables. Standard errors, clustered on countries, in paren-theses. Two tailed hypothesis tests: *p � 0.10, **p � 0.05, ***p � 0.01.



cumulative distribution function. Higher measures indicate a higher fraction ofexplained variance. Although these variables should not be used for model selection(Long, 1997), they do provide some information on the ability of the model toexplain the data. Both pseudo-R2 measures indicate more explanatory power forModel 2 (the model using GDP per capita PPP), for both outcome variables.

The log-likelihood, Akaike Information Criteria (AIC), and Bayesian Informa-tion Criteria (BIC) indicate the relative likelihood of each model for a givendependent variable (see Greene, 2003, pp. 159–160). The log-likelihood repre-sents the relative likelihood of the data given the estimated model parameters.AIC and BIC measures penalize models for including nonrelevant variables.Models are often selected based on the following guidelines: a significantlygreater log-likelihood value, smaller AIC values, and smaller BIC values. Again,however, it must be cautioned that these provide indications of model relevancybut are not hard and fast rules for model selection. For both outcome variablesModel 2 has the largest log-likelihood value, the smallest AIC value, and smallestBIC measure.12

The actual coefficient estimates appear to be fairly robust across model specifi-cations. In Model 6, which omits control for any socioeconomic variable, the esti-mated coefficient for monitoring and sanctioning is larger than in all other models.It appears that socioeconomic conditions are positively correlated both with moni-toring and sanctioning and the outcome measures. Omission of socioeconomic con-ditions in Model 6 overstates the effect of monitoring and sanctioning; part of thereason that forests with monitoring and sanctioning are able to sustain Basal Areaand the Shannon Diversity Index is because they have higher socioeconomic con-ditions on average. For the remainder of the paper, I focus on the results fromModel 5 because the socioeconomic conditions index has components of eachsocioeconomic variable. When assessing the role of particular socioeconomic fac-tors, however, I will refer to the other models for ease of interpretation.

Effects of Independent Variables

Table 5 reports each independent variables’ estimated marginal effect, taken fromModel 5 for each outcome measure, while holding all other variables at their meansand conditional on being in a forest that forbids harvesting. Standard errors forthese estimates are reported in parentheses and are calculated via the delta method(Greene, 2003, p. 70). For continuous variables (households per forest hectare, for-est value, and socioeconomic conditions) the marginal effects represent the changein probability of sustaining Basal Area or the Shannon Diversity Index from a mar-ginal change in the continuous variable. For a binary variable (monitoring andsanctioning, government-management, and private-management), they represent adiscrete change in probability. For example, the effect of government managementis calculated by subtracting the predicted probability of sustaining the outcomevariable given a forest is community managed from the predicted probability givena forest is government managed, while holding all other variables at their meansand forbid harvesting at 1.13

12 In addition to the scalar model indicators, I also analyzed model residuals to check for outliers andinfluence. I plotted Pearson standardized residuals for Model 5 for both outcomes. For Basal Area, eachstandardized residual was within a range of �2:5. For the Shannon Diversity Index, however, two stan-dardized residuals laid slightly outside of that range but still within �3:3, the conventional threshold(Andersson & Gibson, 2007, p. 114).13 As discussed previously, it is important to consider the impacts of these variables in forests that for-bid harvesting. In nonlinear models, such as the logit model, the effects of each explanatory variableimplicitly depend on the level of the other model variables (Long, 1997). Thus, to assess the effects inforests that forbid harvesting, the marginal effects are assessed based on this condition.



Table 5 indicates that forests with active monitoring and sanctioning have a 0.377higher probability of sustaining Basal Area (p � 0.05) and a 0.268 higher probabil-ity of sustaining the Shannon Diversity Index (p � 0.01) than average forests thatforbid harvesting and do not engage in regular monitoring and sanctioning. Tables3 and 4 show that this effect is positive and consistent across model specificationsand outcome variables. Monitoring and sanctioning is positive, of consistent mag-nitude, and statistically significant in every model estimated.

Figures 2 and 3 show how the effects of monitoring and sanctioning change associoeconomic conditions change as measured from Model 5. Monitoring and sanc-tioning has the strongest impact in forests with low socioeconomic conditions, butas socioeconomic conditions increase, the role of monitoring and sanctioningdiminishes. As previously detailed, socioeconomic conditions are a weighting of aproperty rights index measure, per capita GDP adjusted for purchasing power par-ity, average local wage rates, and fuelwood dependency. As fuelwood dependencydecreases, the measure of socioeconomic conditions increases; therefore, the largepositive effect of socioeconomic conditions is in part due to less pressure that usergroups place on forest resources. This can be seen in the separately estimated coef-ficient for fuelwood dependency in Model 4.

As average local wage rates increase, which is the opportunity cost of harvestingfrom the forest, socioeconomic conditions increase. This also helps explain why, associoeconomic conditions increase, the probability of sustaining Basal Area and theShannon Diversity Index increases. The positive effect of wage rates can be seenseparately estimated in Model 3.

The role of formal property rights at the national level are surprisingly less impor-tant. A standard deviation increase in property rights score (1.644) increases thevalue of socioeconomic conditions by a value of 0.852 (0.5184 � 1.644, where0.1584 is the estimated eigenvalue from the principal components analysis). Socio-economic conditions, in turn, positively affect the probability of attaining the out-come measures. Although property rights effects measured separately in Model 1are positive, they are small and insignificant for both outcome measures. GDP percapita adjusted for purchasing power has a negligible effect. Thus, any role thatproperty rights or GDP play is contingent upon other socioeconomic factors.

Table 5. Marginal effects of model 5.1

Basal Area Shannon Index

Monitor and Sanction2 0.377** 0.268***(0.18) (0.08)

Private2 0.265 0.007(0.19) (0.17)

Government2 0.388 0.005(0.22) (0.12)

HH Per Forest Hectare 0.939 0.081***(0.86) (0.05)

Forest Value 0.023 0.032(0.06) (0.05)

Socioeconomic Conditions 0.133** 0.091(0.05) (0.08)

Notes: Standard errors in parentheses, calculated via the delta method. Two-tailedhypothesis tests: *p � 0.10, **p � 0.05, ***p � 0.01.1Marginal effects are conditional on Forbid Harvest � 1 and holding all other variablesat their means.2Marginal effects of binary variables are discrete changes; they are calculated by sub-tracting the predicted probability given the variable is 0 from the predicted probabilitygiven the variable is 1.



A fairly consistent finding across model specifications is the lack of effect of man-agement type on the outcome variables. The estimated coefficients for government-managed forests on Basal Area are quite large, as reported in Table 3; however, thiscoefficient estimate generally has a large variance and is not statistically distin-guishable from 0 at the a � 0.10 level. In Model 2, using GDP per capita adjustedfor purchasing power parity as a control, the effect of being a government-managedforest is statistically different from 0 at the a � 0.01 level. A similar significanteffect, however, is not found in the remaining 11 models. Also, for the ShannonDiversity Index outcome, the effect of management type is quite small and negativein three of the six models. The predicted probabilities for government- and community-managed forests, as reported in Figure 3, are nearly identical in foreststhat forbid harvesting. It appears that management type does not have a significantimpact on the outcome variables overall, although a larger sample might moreaccurately estimate standard errors to confirm this finding.

An important predictor of the Shannon Diversity Index outcome is the pressureon forest resources measured by the number of households per forest hectare. Asreported in Table 5, a marginal change in this variable decreases the probability ofsustaining the Shannon Diversity Index by 0.081, holding all other variables at theirmean and in a forest that forbids harvesting. Table 4 shows that the estimated effect

Figure 2. The effects of management type, monitoring, and socioeconomic condi-tions on Basal Area, given that no harvesting of trees is permitted in the forest, andholding all other variables from Model 5 at their mean. Government-managedforests have a higher probability (not statistically significant) of sustaining orincreasing Basal Area than community-managed forests. Forests with some moni-toring have a higher probability of sustaining or increasing Basal Area than thosethat do not. As the socioeconomic conditions improve, the probability of sustainingor increasing Basal Area increases.



is remarkably consistent across model specifications and significant in each model.However, there is not a statistically significant effect of this variable as measured inthe Basal Area models. Thus, it appears that forests are more homogeneous yethave more basal area the greater the pressure on forest resources. This may bebecause some timber (perhaps low valued) is left to grow in such forests as usersare able to identify and harvest different high-valued species at younger ages.

Perhaps surprisingly, the effect of forbidding harvesting has the opposite antici-pated effect in three of the six models estimated for Basal Area. Again, this variablerepresents a de facto rule for local users. It is interesting that those forests that seekto eliminate harvesting do a poorer job of sustaining Basal Area. There is not a sta-tistically significant effect of this variable as measured in the Shannon DiversityIndex models. This result is discussed in some detail in the following section.

DISCUSSION

The empirical results suggest that management type probably plays little role indetermining whether forests sustain Basal Area and the Shannon Diversity Index.This confirms previous work by Hayes and Ostrom (2005), Gibson, Williams, andOstrom (2005), and Ostrom and Nagendra (2006). Still, there is much ongoing

Figure 3. The effects of management type, monitoring, and socioeconomic condi-tions on the Shannon Diversity Index, given that no harvesting of trees is permittedin the forest and holding all other variables from Model 5 at their mean. Community-managed forests have a nearly identical probability of sustaining or increasing theShannon Index to government-managed forests. Forests with some monitoringhave a higher probability of sustaining or increasing the Shannon Diversity Indexthan those that do not. As the measure of socioeconomic conditions increases, theprobability of sustaining or increasing the Shannon Diversity Index increases.



debate about the efficacy of centralized resource management. Table 6 lists some ofthe most often cited advantages of central and local resource management. Theresults presented in the previous section indicate that on average these advantagesdo not distinguish one management arrangement as being more effective thananother. The results do not suggest, however, that management type is never impor-tant. In some particular instances, a particular advantage of central resourcegovernment may be more important in achieving forest conservation objectives.For example, some locally managed forests may lack the scientific and technologi-cal resources to effectively manage a forest, and governments from a central forestoffice might have such resources. On the other hand, indigenous knowledge insome cases may be more important for management purposes; in these cases, localforest management might be preferred. The point is to move beyond blueprintthinking such as believing that central government is always best or that local man-agement is always best. Instead, we need to focus on the specific managementneeds of specific forests (Ostrom, Janssen, & Anderies, 2007).

More research needs to be devoted to exploring instances of when local manage-ment might be more appropriate than central management and when it is not.Interesting hypotheses in this vein might be that where local social capital is low,central management might be more effective (to guard against the possibility oflocal tyranny). When enforcement of rules is the primary obstacle to managementsuccess, local management may be more efficient. Or when there are serious localfinancial constraints for management and investment, central management mightbe more effective. I do not endorse any of these hypotheses but simply submit themas interesting questions that need further empirical investigation.

Important policy variables do appear to have broad impacts, however. In thispaper, I find that socioeconomic conditions, forest policies to forbid harvesting,

Table 6. Strengths of central and local resource management.

Central Local

“Transboundary Spillovers” (Sigman, 2005): Effective at solving local collectiveLocal management boundaries may not action problems and building social capital coincide with ecosystem boundaries; central (McGinnis & Ostrom, 1992; Cardenas, management can be more inclusive (Blaikie, 2004; Cardenas, Stranlund, & Willis, 2000; 2006; Butler & Macey, 1996). Frey & Oberholzer-Gee, 1997; Ostrom,

1990).

Avoid empowering local tyrannies (Ostrom, Diffusion of power and rulemaking,1997). greater accountability, and increased

opportunities for participation (V. Ostrom,1987, 1991, 1997; Ostrom, Bish, & Ostrom,1988; Andersson & Gibson, 2007).

Scientific and technical capacity (Ostrom, Indigenous knowledge (Berkes, Folke, &1998; Blaikie, 2006). Colding, 1998; Berkes, Colding, & Folke,

2003; Folke et al., 1998; Blaikie, 2006).

Financial, administrative, or technological Relatively low information costs for resources (Andersson, 2003; Boone, 2003; monitoring and enforcement (McGinnis &Falleti, 2005; O’Neill, 2003). Ostrom, 1992).

Goals of conservation, as opposed to local Experimentation and isolation of policygoals of development (Taylor, 2002). failure (E. Ostrom, 1991, 1990; Low et al.,

2003; Simon, 1981).

Longer time horizons for planning (McCarthy, Shorter time horizons for planning2004; Andersson & Gibson, 2007). (McCarthy, 2004; Andersson & Gibson,

2007).



and local institutions of monitoring and sanctioning appear to play an importantrole in sustaining the outcome measures used in this paper. Each is discussed below.

Monitoring and Sanctioning

The results show a remarkably consistent positive effect of monitoring and sanction-ing, confirming the findings of a recent meta-study of community forestry by Pagdee,Kim, and Daugherty (2006). The monitoring and sanctioning variable representswhether regular local monitoring occurs and whether sanctions, broadly defined, areimposed on rule breakers. Further, this variable only indicates monitoring and sanc-tioning activities carried out by local user groups, not any activities that are carriedout by external authorities. These results add an important element to the literatureon forestry, common pool resource management, and human cooperation in general.The lesson from this paper is not that local efforts to engage in monitoring and sanc-tioning are more effective than external efforts, but that local efforts are effective.This finding draws from a number of cases and empirically measures how effectivelocal monitoring and sanctions can be. Average forests with local users that monitorand sanction are much more likely to sustain Basal Area and the Shannon DiversityIndex than forests without such users. Specifically, local efforts of monitoring andsanctioning increase the probability of sustaining Basal Area by 0.377 and increasethe probability of sustaining the Shannon Diversity Index by 0.268.

An important finding is that field data confirm the consistently positive effect of usergroup monitoring and sanctioning found in the experimental laboratory (Ostrom,Walker, & Gardner, 1992; Ostrom, Gardner, & Walker, 1994; Cardenas, Stranlund, &Willis, 2000; Cardenas, 2004; Sefton, Shupp, & Walker, 2007). Laboratory experimentsof monitoring and sanctioning have looked at both external monitoring and sanc-tioning (that is, as imposed by the experimenter) and internal monitoring and sanctioning (that is, as imposed by subjects participating in the experiment). Manyof these studies assess pecuniary incentives to comply with harvesting rules; how-ever, other research has shown that “cheap talk,” or nonbinding communication,also decreases harvesting levels in CPR situations (Ostrom, Gardner, & Walker,1992; Ostrom, Walker, & Gardner, 1994).

Two broad areas of research are needed to further investigate the effects of mon-itoring and sanctioning on forest conditions. First, more analysis of the relative effi-cacy from different sources of monitoring and sanctioning needs to be done (Tucker &Ostrom, 2005, pp. 100–101). For example, one of Ostrom’s (1990, pp. 101–102)design principles is that “monitoring . . . [is] organized in multiple layers of nestedenterprizes.” In this sense, external government monitoring and sanctioning can beseen as a complement to local monitoring and sanctioning efforts. However, thereis also evidence that government monitoring and sanctioning might crowd out localefforts and be counterproductive (Cardenas, Stranlund, & Willis, 2000; Cardenas,2004). Furthermore, local and external monitoring and sanctioning activities mightsubstitute for each other if, for example, locals feel like government officials canhandle the problem. Further empirical evidence, then, needs to be directed to inves-tigate what configurations of local and external monitoring and sanctioning arebest able to meet forest management objectives.

Second, we need a better understanding of what types of monitoring and sanc-tioning are effective. Social shaming, for example, may be a powerful deterrent insome cultures or in small groups but be less effective in large, anonymous groups.Pecuniary sanctions, however, may crowd out the efficacy of local shaming. Levittand Dubner (2005, p. 23) warn that monetary fines can “substitute an economicincentive for a moral incentive,” allowing people to buy off their guilt. Pecuniaryincentives in local communities may crowd out moral incentives to abstain fromharvesting from the forest. Further research is needed to disaggregate the effects ofpecuniary and nonpecuniary sanctions and to assess their effects.



Socioeconomic Conditions

Some authors have argued that improvements in socioeconomic conditions lead tobetter environmental conditions. The relationship is often cited as being nonlinear;that is, deforestation rates would increase in early stages of economic growth butwould later improve as economic growth ensued. Culas (2007), among others, hasconfirmed the existence of an environmental Kuznets curve for deforestation. Cropperand Griffiths (1994) argue that as incomes increase, people generally substituteother energy sources for fuelwood and shift productive assets away from agricul-ture, which can be a major driver of deforestation.

Other authors are more skeptical of this relationship (Lawn, 2007, ch. 12). Stillother authors have concluded that there is inconclusive evidence for such a rela-tionship (Arrow et al., 1995). Recent work, however, has shown the important roleof institutional arrangements mitigating the effects of economic growth on envi-ronmental impacts (Aubourg, Good, & Krutilla, 2008; Culas, 2007). Looking atdeforestation rates at the macro level, across countries and across time, for exam-ple, Culas (2007) found that macro-institutional variables, such as secure propertyrights, decrease the rate of deforestation at all income levels.

My results suggest that increases in socioeconomic conditions positively affectthe sustainability of the outcome variables between visits. As shown in Figures 2and 3 at low levels of socioeconomic conditions, the effects of increases in socio-economic conditions are quite strong, but the benefits diminish as socioeconomicconditions continue to increase. This is manifest in a steeper slope at points oflower income levels and a flatter slope at points of higher income levels. Specifi-cally, I found that as user groups rely less on the forest for fuelwood and have higherwage rates, the forest is more likely to sustain the outcome variables. This providesmicrolevel evidence to the notion that as communities develop, local forests aremore likely to sustain the outcome measures used in this study.

Forbid Harvesting

In the previous section, I presented results indicating that forests that forbid har-vesting to all user groups have significantly less Basal Area than those that do not.This might be due to the “CAMPFIRE effect.” CAMPFIRE is a program instituted inZimbabwe with the goal to empower local communities to manage wildlife(Sterner, 2003). Colonial wildlife management policy banned hunting and set uplarge game reserves. There was great resentment of this policy from both fromwhite colonialists and the native population. In 1975, Zimbabwe legalized owner-ship of animals on private land and an incredible boom occurred for sport hunting.In 1982, Zimbabwe enabled smallholders on community land to similarly profitfrom hunting. Owners now had a greater incentive to protect wildlife because theybenefited financially from them. Sterner presents evidence that hunting of wildlifedeclined at the same time that income from the CAMPFIRE program grew.14 Priorto the CAMPFIRE program, locals resented the ban on wildlife hunting, seeing it asa symbol of repression. Poachers were heroes. However, under different institu-tional arrangements where locals now incurred the costs of poaching, there was agreater incentive to prohibit others and abstain oneself from poaching.

Similarly, in the forests studied in this paper, if harvesting rights are not given tolocals, then forest conditions tend to be poor, holding all else constant. This mightbe for two reasons. First, if no harvest rights are given, there may be incentives toillegally harvest from the forest because community members do not see any bene-fit when abstaining from harvesting. Local norms may not deter illegal harvesting,as there is no local residual claimant on the resource. Illegal harvesting may hurt

14 However, there is still debate about the broad success of these programs (Jones, 2006).



“the government” but not the locals. Second, there is little incentive for others toinvest in the forest. Much research from property rights theory argues that individ-uals are more likely to invest in resources for which they have property rights (DeSoto, 2000). User groups are unlikely to invest resources in sustaining forests if theyhave no use rights and therefore no foreseeable benefit.

CONCLUSION

There is still much academic and policy debate about how to best incorporate localcommunities into natural resource management, ranging from no involvement, tojoint management sharing, to full management responsibilities given to local com-munities. The empirical results suggest that, broadly speaking, forest managementresponsibilities need not completely reside with local communities, but that com-munities still play an important role in affecting forest conditions.

The most consistent and strongest finding from this paper is that communities thatengage in monitoring and sanctioning activities are more likely to sustain basal areaand species diversity in local forests than those that do not. Full management of theresource, however, does not appear important to sustaining these outcomes. Thisimplies that whether or not forests are managed by central managers or local com-munities, management objectives would be well served by enlisting local monitors.

Local efforts of active monitoring and sanctioning have the potential to achievethe benefits often outlined in efforts of “stakeholder participation.”15 Unfortunately,efforts of stakeholder participation are often relegated to token consultation ordemands that local resource users participate in already established governmentprograms. In fact, real democratic participation implies power sharing that allowselaboration, design, and/or implementation of policies and projects (Wester, Merrey,& De Lange, 2003). The empirical results presented here suggest that when stake-holder participation is real (that is, when user groups are allowed to harvest fromthe forest and actively engage in monitoring and sanctioning activities), forest con-ditions are significantly better. These types of activities reflect the spirit of stake-holder participation and are here shown to be effective management tools.

ERIC A. COLEMAN is a doctoral student in public policy at the Indiana UniversitySchool of Public and Environmental Affairs and Department of Political Science andresearch assistant at the Workshop in Political Theory and Policy Analysis.

ACKNOWLEDGMENTS

Earlier versions of this paper were presented at the Midwest Political Science AssociationAnnual Meetings, the Workshop in Political Theory and Policy Analysis colloquium series,and the Property and Environment Research Center graduate seminars. I thank the threeanonymous reviewers, Elinor Ostrom, Wally Thurman, Burnell Fischer, David Good, DanBenjamin, and various seminar participants for extensive comments and suggestions on earlier drafts. Any remaining errors are my own. Financial support from the John T. andCatherine T. MacArthur Foundation from the Workshop in Political Theory and Policy Analysis,and from the Property and Environment Research Center is gratefully acknowledged.

REFERENCES

Agrawal, A. (2001a). Common property institutions and sustainable governance of resources.World Development, 29, 1649–1672.

Agrawal, A. (2001b). Small is beautiful, but is larger better? Forest-management institutionsin the Kumaon Himalaya, India. In C. C. Gibson, M. A. McKean, & E. Ostrom (Eds.),

15 For a recent review of stakeholder participation in the sustainable development literature, see Backstrand (2006).


People and forests: Communities, institutions, and governance (pp. 27–55). Cambridge:MIT Press.

Agrawal, A., & Chhatre, A. (2006). Explaining success on the commons: Community forestgovernance in the Indian Himalaya. World Development, 34, 149–166.

Agrawal, A., & Goyal, S. (2001). Group size and collective action—Third-party monitoring incommon-pool resources. Comparative Political Studies, 34, 63–93.

Anderson, T. L., & Leal, D. R. (2001). Free market environmentalism (rev. ed.) New York: Palgrave MacMillan.

Andersson, K. (2003). What motivates municipal governments? Uncovering the institutionalincentives for municipal governance of forest resources in Bolivia. Journal of Environmentand Development, 12, 5–27.

Andersson, K., & Gibson, C. (2007). Decentralized governance and environmental change:Local institutional moderation of deforestation in Bolivia. Journal of Policy Analysis andManagement, 26, 99–123.

Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C. S., Jansson, B.-O.,Levin, S., Maler, K.-G., Perrings, C., & Pimentel, D. (1995). Economic growth, carryingcapacity, and the environment. Science, 268, 520–521.

Aubourg, R. W., Good, D. H., & Krutilla, K. (2008). Debt, democratization, and developmentin Latin America: How policy can affect global warming. Journal of Policy Analysis andManagement, 27, 7–19.

Backstrand, K. (2006). Democratizing global environmental governance? Stakeholderdemocracy after the World Summit on Sustainable Development. European Journal ofInternational Relations, 12, 467–498.

Berkes, F., Colding, J., & Folke, C. (2003). Navigating social-biophysical systems: Buildingresilience for complexity and change. Cambridge, UK: Cambridge University Press.

Berkes, F., Folke, C., & Colding, J. (1998). Linking social and biophysical systems: Manage-ment practices and social mechanisms for building resilience. Cambridge, UK: CambridgeUniversity Press.

Blaikie, P. (2006). Is small really beautiful? Community-based natural resource managementin Malawi and Botswana. World Development, 34, 1942–1957.

Boone, C. (2003). Decentralization as political strategy in West Africa. Comparative PoliticalStudies, 36, 355–380.

Boyd, R., Gintis, H., Bowles, S., & Richerson, P. J. (2003). The evolution of altruistic punish-ment. Proceedings of the National Academy of Sciences, 100, 3531–3535.

Butler, H. N., & Macey, J. R. (1996). Pollution, externalities, and the matching principle. YaleLaw and Policy Review/Yale Journal on Regulation Symposium: Constructing a NewFederalism, 14, 23–66.

Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: Methods and applications. NewYork: Cambridge University Press.

Cardenas, J. C. (2004). Norms from outside and from inside: An experimental analysis on thegovernance of local ecosystems. Forest Policy and Economics, 6, 229–41.

Cardenas, J. C., Stranlund, J., & Willis, C. (2000). Local environmental control and institu-tional crowding-out. World Development, 28, 1719–1733.

Cropper, M., & Griffiths, C. (1994). The interaction of population growth and environmentalquality. American Economic Review, 84, 250–254.

Culas, R. J. (2007). Deforestation and the environmental Kuznets curve: An institutional per-spective. Biophysical Economics, 61, 429–437.

De Soto, H. (2000). The mystery of capital: Why capitalism triumphs in the West and failseverywhere else. New York: Basic Books.

Dietz, T., Ostrom, E., & Stern, P. C. (2003). The struggle to govern the commons. Science,302, 1907–1912.

Falleti, T. G. (2005). A sequential theory of decentralization: Latin American cases in com-parative perspective. The American Political Science Review, 99, 327–346.



Fehr, E., & Gachter, S. (2000). Cooperation and punishment in public goods experiments.American Economic Review, 90, 980–994.

Fehr, E., & Gachter, S. (2002). Altruistic punishment in humans. Nature, 415, 137–140.

Folke, C., Pritchard, L., Berkes, F., Colding, J., & Svedin, U. (1998). The problem of fitbetween ecosystems and institutions. Working Paper 2.

Frey, B. S. (1997). On the relationship between intrinsic and extrinsic work motivation. Inter-national Journal of Industrial Organization, 15, 427–439.

Frey, B. S., & Oberholzer-Gee, F. (1997). The cost of price incentives: An empirical analysisof motivation crowding-out. American Economic Review, 87, 746–755.

Gibson, C., Lehoucq, F. E., & Williams, J. T. (2002). Does privatization protect naturalresources? Property rights and forests in Guatemala. Social Science Quarterly, 83, 206–225.

Gibson, C., McKean, M. A., & Ostrom, E. (2000). People and forests: Communities, institu-tions, and governance. Cambridge, MA: MIT Press.

Gibson, C., Williams, J., & Ostrom, E. (2005). Local enforcement and better forests. WorldDevelopment, 33, 273–284.

Greene, W. H. (2003). Econometric analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

Gwartney, J., & Lawson, R. (2006). Economic freedom of the world: 2006 annual report. Vancouver, BC: The Fraser Institute. Data retrieved from www.freetheworld.com.

Hardin, G. (1968). The tragedy of the commons. Science, 162, 1243–1248.

Hauert, C., Traulsen, A., Brandt, H., Nowak, M. A., & Sigmund, K. (2007). Via freedom tocoercion: The emergence of costly punishment. Science, 316, 1905–1907.

Hayes, T. & Ostrom, E. (2005). Conserving the world’s forests: Are protected areas the onlyway? Indiana Law Review, 38, 595–617.

Henrich, J., McElreath, R., Barr, A., Barrett, J. E. C., Bolyanatz, A., Cardenas, J. C., et al.,(2006). Costly punishment across human societies. Science, 312, 1767–1770.

Husch, B., Beers, T. W., & Kershaw, J. A. (2003). Forest mensuration. Hoboken, NJ: JohnWiley & Sons.

Jones, S. (2006). A political ecology of wildlife conservation in Africa. Review of AfricanPolitical Economy, 33, 483–495.

Kaplan, D. (2000). Structural equation modeling: Foundations and extensions. Advancedquantitative techniques in the social sciences 10. Thousand Oaks, CA: Sage.

Kellert, S. R., Mehta, J. N., Ebbin, S. A., & Lichtenfeld, L. L. (2000). Community naturalresource management: Promise, rhetoric, and reality. Society & Natural Resources, 13,705–715.

King, G., Keohane, R. O., & Verba, S. (1994). Designing social inquiry: Scientific inference inqualitative research. Princeton, NJ: Princeton University Press.

Lawn, P. (2007). Frontier issues in biophysical economics. Cheltenham, UK: Edward ElgarPublishing.

Lejano, R. P., & Ingram, H. (2007). Place based conservation: Lessons from the TurtleIslands. Environment, 49, 18–26.

Levitt, S. & Dubner, S. J. (2005). Freakonomics: A rogue economist explores the hidden sideof everything. New York: Harper Collins.

Long, J. S. (1997). Regression models for categorical and limited dependent variables.Advanced quantitative techniques in the social sciences 7. Thousand Oaks, CA: Sage.

Low, B. S., Ostrom, E., Simon, C. P., & Wilson, J. (2003). Redundancy and diversity: Do theyinfluence optimal management? In J. Colding & C. Folke (Eds.), Navigating nature’sdynamics (pp. 83–114). Cambridge, UK: Cambridge University Press.

McCarthy, J. F. (2004). Changing to gray: Decentralization and the emergence of volatilesocio-legal configurations in central Kalimantan, Indonesia. World Development, 32,1199–1223.

McGinnis, M. D., & Ostrom, E. (1992). Institutional analysis and global climate change Designprinciples for robust international regimes. In M. Rice, J. Snow, & H. Jacobson (Eds.),



Proceedings of a Conference Global Climate Change: Social and Economic Research Issues,45–85.

McKean, M. (2000). Common property: What is it, what is it good for, and what makes itwork? In C. Gibson, M. McKean, & E. Ostrom (Eds.), People and forests: Communities,institutions, and governance. Cambridge, MA: MIT Press.

Milinski, M., Semmann, D., & Krambeck, H.-J. (2002). Reputation helps solve the tragedy ofthe commons. Nature, 415, 424–426.

National Research Council. (1986). Proceedings of the Conference on Common PropertyResource Management. Washington, DC: National Academy Press.

National Research Council. (2002). The drama of the commons. Washington, DC: NationalAcademy Press.

Olson, M. (1965). The logic of collective action: Public goods and the theory of groups. Cam-bridge, MA: Harvard University Press.

O’Neill, K. (2003). Decentralization as an electoral strategy. Comparative Political Studies,36, 1068–1091.

Ostrom, E. (1990). Governing the commons: The evolution of institutions for collectiveaction. Cambridge, UK: Cambridge University Press.

Ostrom, E. (1991). Rational choice theory and institutional analysis: Toward complementar-ity. American Political Science Review, 85, 237–243.

Ostrom, E. (1998). Scales, polycentricity, and incentives: Designing complexity to governcomplexity. In L. D. Guruswamy & J. A. McNeely (Eds.), Protection of global biodiversity:Converging strategies (pp. 149–67). Durham, NC: Duke University Press.

Ostrom, E. (2005). Understanding institutional diversity. Princeton, NJ: Princeton UniversityPress.

Ostrom, E., Gardner, R., & Walker, J. (1994). Rules, games, and common-pool resources. AnnArbor: University of Michigan Press.

Ostrom, E., Janssen, M. A., & Anderies, J. M. (2007). Going beyond panaceas. Proceedings ofthe National Academy of Sciences, 104, 15176–15178.

Ostrom, E., & Nagendra, H. (2006). Insights on linking forests, trees, and people from the air,on the ground, and in the laboratory. Proceedings of the National Academy of Science, 103,19224–19231.

Ostrom, E., Walker, J., & Gardner, R. (1992). Covenants with and without a sword: Self-governance is possible. American Political Science Review, 89, 404–417.

Ostrom, V. (1987). The political theory of a compound republic: Designing the Americanexperiment (2nd ed.). Lincoln: University of Nebraska Press.

Ostrom, V. (1991). The meaning of American federalism: Constituting a self-governing soci-ety. San Francisco: ICS Press.

Ostrom, V. (1997). The meaning of democracy and the vulnerability of democracies: Aresponse to Tocqueville’s challenge. Ann Arbor: University of Michigan Press.

Ostrom, V., Bish, R. L., & Ostrom, E. (1988). Local government in the United States. SanFrancisco: ICS Press.

Pagdee, A., Kim, Y.-S. K., & Daugherty, P. D. (2006). What makes community forest manage-ment successful: A meta-study from community forests throughout the world. Society andNatural Resources, 19, 33–52.

Panchanathan, K., & Boyd, R. B. (2004). Indirect reciprocity can stabilize cooperation with-out the second-order free rider problem. Nature, 432, 499–502.

Regev, U., Gutierrez, A. P., Schreiber S. J., & Zilberman, D. (1998). Biological and economicfoundations of renewable resource exploitation. Biophysical Economics, 26, 277–242.

Sefton, M., Shupp, R., & Walker, J. (2007). The effect of rewards and sanctions in provisionof public goods. Economic Inquiry, 45, 671–690.

Sigman, H. (2005). Transboundary spillovers and decentralization of environmental policies.Journal of Environmental Economics and Management, 50, 82–101.

Simon, H. A. (1981). The sciences of the artificial (2nd ed.) Cambridge, MA: MIT Press.



Sterner, T. (2003). Policy instruments for environmental and natural resource management.Washington, DC: RFF Press.

Taylor, M. (2002). Whose agenda? Reassessing the role community-based natural resourcemanagement in Botswana. Mimeo.

Terborgh, J. (1999). Requiem for nature. Washington, DC: Island Press.

Tucker, C. M., & Ostrom, E. (2005). Multidisciplinary research relating institutions and for-est transformations. In E. F. Moran & E. Ostrom (Eds.), Seeing the forests and the trees:Human-environment interactions in forest ecosystems (pp. 81–103). Cambridge, MA: MITPress.

Uphoff, N. T. (1998). Community-based natural resource management: Connecting micro andmacro processes and people with their environments. Keynote for international workshopon Community-Based Natural Resource Management, Economic Development Institute,World Bank, Washington, DC.

Varughese, G. (2000). Population and forest dynamics in the hills of Nepal: Institutionalremedies by rural communities. In C. Gibson, M. A. McKean, & E. Ostrom (Eds.), Peopleand forests: Communities, institutions, and governance (pp. 193–226). Cambridge, MA:MIT Press,

Wade, R. (1994). Village republics: Economic conditions for collective action in South India.Princeton, NJ: Princeton University Press.

Wester, P., Merrey, D. J., & De Lange, M. (2003). Boundaries of consent: Stakeholder repre-sentation in river basin management in Mexico and South Africa. World Development, 31,797–812.


institutional factors affecting biophysical outcomes in forest management

Documents