Research

INTRODUCTION

Many governments in low-income countries have promoted community-based approaches to forest management and embarked on forest devolution programs to promote forest conservation and improve local livelihoods (Faguet 2014). Forests play an important role in climate change mitigation, protection of biodiversity, and improved livelihoods of nearby communities (FAO 2020). There is increasing international attention toward forests with major initiatives across the world on forest protection and reforestation. For example, the Africa Forest and Landscape Restoration Initiative (AFR100) aims to reforest more than 100 million hectares of land in 31 African countries by 2030. Given the common-pool resource nature of most forests, the eventual success of such initiatives depends on whether there are effective and inclusive collective action institutions at the local level. Devolving use rights and management responsibilities to local communities is expected to create a sense of ownership that incentivizes sustainable exploitation by motivating forest users to respect extraction rules and monitor and sanction transgressions (Ostrom and Nagendra 2006, Faguet 2014). About one-third of forests in low-income countries are now managed by local communities (Blackman et al. 2017). However, the impacts of community-based resource management are ill understood, which suggests a need for improved learning about the design, implementation, and governance of community-based forest management (CBFM) programs. Most CBFM programs are implemented by external actors, such as the government or non-governmental organizations, often with the participation of local communities in their design or implementation. Although these programs come with intensive community mobilization, training, and capacity-building components at the beginning, these support mechanisms eventually disappear as the project funding ends, leaving local communities vulnerable to governance challenges.

Developing evidence on the causal link between different governance structures and outcomes of interest is crucial to understand how to best support CBFM and to institutionalize accountability. Local governance is a key factor in addressing collective action challenges and determining community-level outcomes in the management of commons (e.g., Ostrom 1990, 2009). But what exactly constitutes “good local governance” is still a topic of debate (Agrawal 2001, Dietz et al. 2003, Ostrom et al. 2007, Saunders 2014, Baggio et al. 2016, Young et al. 2018). Institutional features typically mentioned as principles of good governance range from community participation in decision making and rule setting, to adequate monitoring and sanctioning of transgressions (Hardin 1978, Ostrom 1990, Agrawal 2003, Ostrom et al. 2007). The main challenge for sustainability science is isolating the impact of alternative governance designs on desirable outcomes. Institutional features typically arise endogenously and over a long time span in “properly functioning” communities, making the analysis of causality difficult. Without exogenous variation in community’s governance structures (i.e., more or less random occurrence of good governance across a sample of communities), one cannot judge the impact of these governance features on outcomes relevant to the community. Furthermore, desirable institutional features typically are observed in bundles, making it hard to isolate the specific effects of each one of them for analytical purposes (Slough et al. 2021a).

We used an experimental design to analyze how adding alternative versions of externally implemented monitoring to an existing CBFM affects environmental outcomes in Ethiopia. To reverse trends of deforestation and forest degradation, the Ethiopian government initiated a CBFM program in the late 1990s. So-called “forest user groups” (FUGs) received exclusive use rights for clearly demarcated forest blocks, enabling members to extract forest resources for consumption or sale. In return, they were expected to manage forest resources in a sustainable manner, restrict further settlement and agricultural expansion, and pay an annual rent (Ameha et al. 2014b). Our study area concerns the Adaba and Dodola districts in which 132 FUGs were created (maximum 30 members, blocks of 360 hectares). Many groups in charge of managing their forest block do not have properly functioning monitoring institutions (Kahsay and Bulte 2019). In our study area, all FUGs are governed by an executive committee consisting of five members: the chairperson, vice-chairperson, secretary, cashier, and representative member, typically populated by members of the local elite (Kahsay and Bulte 2021). Although the power structure within the executive committee varies across FUGs, the chairperson is often the most powerful person and leader. There is also a general assembly, consisting of all group members, in which strategic issues are discussed and agreed upon. A common problem in CBFM programs in developing countries is that group leadership may be unaccountable, and elite capture is widely perceived as a big problem (e.g., Agrawal and Gupta 2005, Kahsay and Bulte 2021). Elite capture refers to circumstances in which local elites (individuals with higher socio-political status) control decision-making processes and misappropriate or “grab” a disproportionately large share of forest benefits (Fritzen 2007, Lund and Saito-Jensen 2013). Tesfaye et al. (2015) and Kahsay and Bulte (2021) documented such cases in our study area. One mechanism may be side selling of forest resources, and another may be diversion of group income. Testing alternative designs of an externally motivated monitoring program on top of an existing CBFM to increase the accountability of community leaders is therefore of high practical and scientific relevance.

Our study focuses on the challenge of ensuring accountability of FUG leaders, a central element of local governance. We distinguish between bottom-up and top-down monitoring regimes to improve the accountability of the local leader, thereby assisting FUGs to improve their own governance. Bottom-up monitoring involves supporting local communities to enhance their capacity to monitor and track forest use and management, as well as hold their leader accountable. Top-down monitoring is done by local governments responsible for forest management (Gupta and Koontz 2019). We randomly assigned FUGs to one of the three intervention groups (top-down monitoring + punishment for worst-performing leaders, top-down monitoring + reward for best-performing leaders, bottom-up monitoring + punishment for badly performing leaders) and a control group. In the bottom-up monitoring group, FUG members were encouraged and trained to engage in (increased) forest management and to use monitoring as well in evaluating their leader’s performance. Underperforming leaders could be removed from their positions after a democratic, anonymous voting process (but user rights would remain at the community level and not revert to the state). In the top-down monitoring groups, governmental officials engaged in intensified monitoring, and either rewarded good leadership or punished a failing group leader. Rewards take the form of a material gift, namely a solar panel with market value of US$60. Punishment implies the group leader is removed from the position of authority by the regulating agency (the Oromia Forest and Wildlife Enterprise, OFWE), which is well within the terms of agreement between OFWE and FUGs. Afterward, group members elect a new group leader, and user rights remain at the community level. The three interventions are implemented in addition to an existing CBFM, in which communities manage their forest resources. The purpose of these interventions is to improve the governance of FUGs and the effectiveness of CBFM programs in promoting livelihood and conservation outcomes.

Initially, we hypothesized that both top-down and bottom-up monitoring treatments led to greater forest conservation than the control group (or business-as-usual CBFM). A broader literature on community involvement in development interventions (i.e., beyond natural resource management) suggests that increased monitoring of both types can potentially discipline leaders and result in improved development outcomes. For example, Olken (2007) found that top-down monitoring reduced missing expenditures in infrastructure interventions in Indonesia, and Björkman and Svensson (2009) documented a positive impact of bottom-up monitoring on health outcomes in Uganda (a finding that was subsequently contested by Raffler et al. 2019).

Conditional on moderating conditions, our three treatments are expected to operate through the same three mechanisms: improved leader responsiveness and performance, increased community member participation, and, ultimately, increased information to guide decision making. Figure 1 shows our theory of change. A key feature of our experiment is that differences in the treatments do not lead to consequential variation. Following Ferraro and Agrawal (2021:6), a “treatment variation is consequential if different versions of the treatment create variations in mechanism effects.” Having no consequential variation in our treatments increases the internal validity of comparing them to each other. To this end, all our treatments affect our outcome variables through the same mechanisms. Forest inventory data across the FUGs are one of the unique features of our study. As proxies for conservation success, we used an increase in the number of potential crop trees per hectare (PCT/ha: trees of marketable species with a minimum diameter at breast height), as well as reductions in the number of charcoal spots and stumped trees per ha, and also the forest experts’ rating of forest regeneration and overall forest status. We recognize that trees alone do not capture the spectrum of forest biodiversity (Spicer et al. 2020).

Monitoring interventions are likely to affect leader’s effort and behavior through the threat of sanctioning (removal from power in the interventions with punishment) or the prospect of a reward (in one of the top-down monitoring versions). This is consistent with Kahsay and Bulte (2021), and with results of Grossman and Hanlon (2014). In addition, presentation of monitoring reports to the general assembly meeting of FUGs may imply possible informal sanctions (social pressure). Thus, the FUG leader is more likely to behave in accordance with group interests, i.e., increase the effort in forest patrolling, reduce side selling of forest products, and involve group members in group decisions. Second, group members are expected to increase their involvement in group activities including forest patrolling because (1) they follow the leader’s example of increased engagement in forest patrolling and other group activities (e.g., meeting and joint works); and/or (2) group members are more likely to believe that their (collective) actions will succeed in the face of the monitoring interventions and are more likely to engage in group activities (see also Hermalin 1998, Dionne et al. 2010). The combined leaders’ and members’ forest patrolling efforts as well as the engagement of group members in other group activities (meetings and joint works) and decision making are likely to reduce illegal forest extraction by the leader, group members, and outsiders. We are fully aware that, ultimately, the final impact of monitoring on improved leader responsiveness and member participation will depend on moderating site-specific factors, for example, potential information asymmetries, the level of group empowerment vis-à-vis the leader, the possibility of collusion between leaders and monitors, the group’s ability to overcome social dilemmas, and incentives for voluntary monitoring by members. It’s hard to produce exogenous variation in these factors, and they therefore remain outside of the scope this work. Most importantly, our randomization efforts should ensure no systematic bias in the degree to which these factors are present in our communities.

Evidence of the impact of CBFM on conservation and livelihood outcomes is mixed and heterogeneous (see e.g., Porter-Bolland et al. 2012, Buntaine et al. 2015, and Blackman et al. 2017 for conservation outcomes, and Sikor and Nguyen 2007, Coleman and Fleischman 2012, and Ameha et al. 2014b for social and economic outcomes). Kahsay and Bulte (2021) compared the impact of top-down and bottom-up monitoring on livelihood outcomes among FUGs in our study area and found that only top-down monitoring improved income of FUG members. A recent Proceedings of the National Academy of Sciences (PNAS) special issue presented causal evidence on the impact of externally implemented community monitoring on common pool resources (CPR) using preregistered and harmonized randomized control trials (RCTs) across heterogeneous contexts (forest and water resources in six sites located in four continents). In a meta-analysis of these experimental studies, Slough et al. (2021a) found that the community monitoring intervention was widely adopted in five of the six sites with a modest effect on CPR extraction and status. These effects are mediated primarily through increased accountability (Ferraro and Agrawal 2021). However, results are heterogeneous across study sites. For instance, Christensen et al. (2021) and Eisenbarth et al. (2021) found no effects of community-based forest monitoring on deforestation in, respectively, Liberia and Uganda. Eisenbarth et al. (2021) found an increase in forest loss in non-monitored forest areas in treatment groups, suggesting a negative spillover effect of the monitoring intervention. In contrast, Slough et al. (2021b) found that community-based monitoring reduces forest cover loss among Indigenous communities in the Peruvian Amazon.

Our study context differs from these PNAS studies. With the exception of Bernedo Del Carpio et al. (2021), who studied community monitoring in community-based water management organizations in Costa Rica, none of these studies involved formally established resource user groups. Similar to Kahsay and Bulte (2021), our context focuses on CBFM in which local communities are formally organized into FUGs and enter into agreement with a government, detailing their rights and responsibilities. Our study expands on the previous literature in two ways. First, we compared the forest conservation impact of bottom-up monitoring to an approach involving top-down scrutiny by the government. Second, although our bottom-up intervention involved similar activities to earlier studies (e.g., community workshops, recruiting and training local monitors, disseminating the monitoring report with group members), it added a unique feature by facilitating the punishment of underperforming leaders; our intervention has teeth. Our rich dataset also allows us to shed light on the mechanisms linking interventions to forest conservation. Details about the data, context, treatments, and variables used are available in Appendix 1 (supplementary information section).

BACKGROUND AND CONTEXT

In Ethiopia, forest cover declined from 40% in the late 19th century to approximately 4% in recent years (Dessie and Christiansson 2008). To reverse this trend of deforestation and forest degradation, the Ethiopian Government initiated a large community-based forestry program in the late 1990s. The key principle was that forest user groups (FUGs) received exclusive use rights for clearly demarcated forest blocks, enabling them to extract forest resources for consumption or sale. The program is particularly widespread in the Oromia Regional State, with about 900 established FUGs overall. One of the earliest CBFM programs was implemented in the Adaba-Dodola forest (our study area) by the Ethiopian Government in collaboration with the German Society for International Cooperation (GIZ). The Adaba-Dodola forest has an estimated forest cover of 73,600 ha (WBISPP 2001) and is located on the northern slopes of the Bale Mountains, some 320 km from Addis Ababa. The program was implemented with the participation of the local community. There were a series of information and awareness meetings with village residents, who later set-up a committee responsible for identifying a list of membership criteria. Village residents subsequently approved, with a majority vote, three main membership eligibility criteria: settlement proximity to the forest area, permanent residence in a village, and traditional and customary use rights (Ameha et al. 2014b). Eligible village members voluntarily joined FUGs with a maximum group size of 30 households. With an average 12 ha per member, groups manage forest blocks up to 360 ha (Amente 2005, Ameha et al. 2014a). Each FUG draws up a bylaw stipulating rights and responsibilities of leaders and members, so some level of bottom-up scrutiny is part of all FUGs, but the extent to which ordinary members dare to stand up to their leaders varies across FUGs. Currently, there are 132 FUGs in the Adaba-Dodola CBFM program, with each village having multiple FUGs, each with access to its own forest area. The livelihood of group members includes subsistence agriculture, livestock production, and forest utilization. On average, group members derive one quarter of their income from forest resources, highlighting the importance of the forest stock for their livelihoods.

Rights and duties of groups are written in a contract signed by the FUG and the responsible branch of the government. Part of the agreement is that group members are allowed to extract forest products. In return, they should sustainably manage the forest, restrict settlement and agricultural expansion, protect the forest from incursions by others, and pay an annual rent (Kosfeld and Rustagi 2015, Ameha et al. 2014b). The agreement between FUGs and the government remains valid indefinitely, or until harvesting exceeds the allowable cut by more than 10%. In that case, the agreement is terminated and use rights revert to the government. An important part of the management responsibility concerns protecting forests from illegal harvesting by outsiders (and sometimes group members). This involves extensive patrolling of the forest, an activity planned and overseen by group leaders, and in which leaders participate themselves. When a transgression or theft is detected, the punishment is likely severe and includes confiscation of materials and payment of fines, and for group members, this may also include extra work assignment and reduced sharing in benefits from the forest. All FUG members are required to engage in forest patrolling. About 91% of the FUGs in our study have rules on forest patrolling. However, forest patrolling is a public good and there are concerns about free riding. The FUG leaders should play a role in mitigating free riding problems by enforcing forest patrolling rules.

Although leaders in our context are not directly paid by FUG members for their service, they get indirect benefits. First, leaders often receive labor contributions from members, e.g., during crop planting and harvesting seasons. Second, leaders represent group members in meetings and trainings organized by the government or NGOs, and this often involves payments for participation and opportunities for external networking. Finally, leadership positions are associated with status and networking in the local community. Nearly 70% of the FUGs in our sample do not have an internal monitoring committee to keep track of leader activities and performance. When one is in place, the scope of monitoring and their mandate is much narrower than what we implement in the context of our interventions (Kahsay and Bulte 2019). Some 50% of the groups have so far changed their leaders at least once, but leader turnover is slower than may be expected based on FUG bylaws (Kahsay and Medhin 2020).

The selection of the Adaba-Dodola CBFM program as our study site was motivated by four main reasons. First, CBFM program is widely practiced in the area with about 132 established FUGs, which guarantees us a large sample of FUGs for our study. Second, unlike in other areas, the Adaba-Dodola CBFM program kept a baseline and follow-up forest inventory data for each FUG. Without this data, it would not have been possible to estimate the impact of our monitoring interventions. Although data from satellite imagery is another option, it would have been difficult to detect significant differences between the experimental groups given the relatively short treatment period. Third, FUGs in our sample have a manageable size of 30 households unlike in some areas, where group size can reach up to several thousand households. This made it easier to understand the forest governance in detail. Finally, the area borders the important Bale Mountain National Park, which is home to various endangered and critically endangered species.

MATERIALS AND METHODS

Experimental design

We partnered with Oromia Forest and Wildlife Enterprise (OFWE), which is the branch of government responsible for forest management in the study region, and used an RCT design to randomly assign the 132 forest user groups to 4 experimental groups: a control group (with business-as-usual monitoring and governance) and 3 treatment groups. In one treatment group, we intensify monitoring by group members, and in the two others, this responsibility was assumed by the government, represented by officials from the OFWE. The interventions are described in detail below (see Figure 1 for theory of change). Additional details of the treatment groups can be found in Kahsay and Bulte (2021).

Bottom-up monitoring with punishment

We facilitated the creation of group-level monitoring committees (one per FUG). Creating such a committee is an obvious choice for intervening organizations that seek to promote engagement of group members in scrutinizing group leadership. All elected committee members accepted the responsibility, were trained, and provided with stationary materials and score cards to facilitate monitoring. A monitoring protocol was developed by OFWE experts with the participation of FUG representatives and was subsequently given to committee members. The main function of the committee was to monitor forest use and management, including inspection of forest blocks, books and records, and organization of group discussions and interviews with anonymous group members and non-members. Six months into the program, the monitoring committee presented a summary report of its findings to the group’s general assembly. This report was “benchmarked” by OFWE, providing members with information about how their leaders performed compared to other leaders in the program. Finally, FUG members anonymously voted whether the leader should be (1) complimented for his efforts, (2) reprimanded for poor performance, or (3) dismissed. In case of severe underperformance and leader dismissal, OFWE facilitated an election process of new leaders, again involving anonymous majority voting. Removal from power implies a loss of indirect benefits and possibly social stigma. It is important to emphasize that our intervention is likely different from the type of bottom-up monitoring that has grown organically within successful groups.

Top-down monitoring with punishment or reward

The study involved two types of top-down monitoring by OFWE expert, using similar assessment criteria, approaches, and tools provided to the committee members in the bottom-up monitoring group. In both top-down monitoring groups, OFWE experts visited the forest, checked the books, and organized conversations with forest users. In one group, monitoring was combined with punishment. Oromia Forest and Wildlife Enterprise would present its draft report to the FUG’s general assembly to validate its findings and collect additional information. The final monitoring report was used by OFWE management to compliment, reprimand, or dismiss group leaders. In the case of dismissal, OFWE assisted FUGs to elect a new leader. The other top-down monitoring group combined top-down scrutiny with a reward for top performing leaders. This treatment group was based on the same procedures, but now OFWE management selected the three best leaders and awarded them a solar panel (with market value of 60$US or 197$US in PPP terms). Rewards were provided in a workshop attended by all FUG leaders from this treatment group.

The monitoring experts sent by OFWE may have previous engagements with the groups. To mitigate potential collusion between the OFWE experts and group leaders, OFWE experts were randomly assigned to each group for each meeting (so an expert assigned to one group may not visit the same group in the next engagement). Furthermore, the decision of whether to punish leaders or not (and reward leaders or not) is done by OFWE management, not by the experts.

Table 1 gives the allocation of groups across the treatment groups. The experiment included the entire population of FUGs. The ex-ante power analysis suggests a minimum detectable effect below previous findings in the literature. See Appendix 1 and Table A1.1 for details of power analysis.

The timeline of the experiment was as follows. We collected baseline survey data in March and April 2017. Interventions started in May 2017 with group-level start-up meetings. Follow-up visits to FUGs took place in July 2017 and general assembly meetings were organized in December 2017 and January 2018, more than 18 months after starting the program. Finally, the current analysis is based on forest inventory data collected in the period between September and December 2019. Our forest data were therefore collected nearly 2.5 years after the start of the interventions. Although the duration of our interventions follows previous randomized controlled trials in the forest domain (e.g., Jayachandran et al. 2017, Christensen et al. 2021), this is a relatively short interval to observe significant changes in standing forests. We therefore also consider “flow variables” that capture the intensity of on-going extraction such as number of charcoal spots and stumps as well as the status of the forest and its regeneration.

Data and identification strategy

Most of our data are from OFWE’s database, which includes administrative information on group characteristics, such as year of establishment, group size, composition and location of groups, and size of the forest block they manage. Oromia Forest and Wildlife Enterprise also provides three waves of data on forest conservation, based on extensive inventories. This involves (1) systematic sampling of circular plots with a radius of 25 m at 100 m intervals between individual plots and between straight lines (approximately one plot per hectare, in total about 48,000 plots were measured), and (2) counting the number of potential crop trees (PCTs) and mature trees (MTs) for each tree species. Trees are recorded on a recording sheet. The PCTs are young and healthy trees that are over 2 m in height with a diameter at breast height (DBH) of up to 40 cm for group 1 species (Podocarpus falcatus, Juniperus procera, Hagenia abyssinica, Olea europaea ssp. cuspidata, Ekebergia capensis, Gymnosporia arbutifolia subsp. arbutifolia, and Schefflera abyssinica) and up to 25 cm for group 2 species (mostly Pittosporum abyssinicus, Hypericum lanceolatum, Erica arborea, and Rapanea melanophloeos but also other species not listed in group 1). Mature trees are trees with DBH exceeding 40 cm for group 1 tree species and exceeding 25 cm for group 2 species. Data were collected by trained forest experts.

Two waves of collecting forest inventory data took place before the start of our experiment. The first inventory was conducted in 2004, before the establishment of most FUGs. This baseline inventory is available for 117 forest blocks, managed by 117 FUGs. The second inventory was conducted in 2012, for only 49 forest blocks. Data from the first and second inventory were used in previous studies (Rustagi et al. 2010, Kosfeld and Rustagi 2015, Ameha et al. 2016, Kahsay and Bulte 2019, Kahsay and Medhin 2020). Part of the PCT/ha variable collected in 2019 (third inventory) was used in Kahsay et al. (2021) to study the association between gender balance in FUG management and forest outcomes. The first forest inventory data were included in our regression models to control for baseline differences in forest condition (see Appendix 1, Tables A1.3, A1.5, and A1.6). The second inventory data were used to show that groups assigned to the treatment groups have, on average, the same number of PCT/ha (see Appendix 1, Table A1.3).

Our dependent variables are collected during the third inventory, which took place in 2019. Forest data are available for 120 forest blocks. Following Rustagi et al. (2010), Kosfeld and Rustagi (2015), and Kahsay and Medhin (2020), we used the density of potential crop trees (PCT/ha) as our main performance indicator. Mature trees are less informative for our purposes because these large trees are unlikely to be extracted due to lack of access to suitable machinery. The main advantage of our PCT/ha variable is that it can respond quickly to changes in local forest management, which is important in light of the relatively short duration of the research intervention. Trees slowly enter the class of PCTs as they mature and graduate to the class of MTs as they age further, but PCT densities can change rapidly in response to changes in extraction. Although our study period is presumably too short to pick up the effect of increased regeneration efforts, it is sufficiently long to capture changes in thinning and extraction intensity. In this light, it is important to mention that PCTs are the type of trees that local community members use for their own consumption and sale in the form of firewood and charcoal, and for construction purposes.

Potential crop tree counts are not a perfect proxy of forest quality. For example, PCT densities are uninformative about the tree species composition and distribution, or the conservation status of specific tree species. Obviously, it is silent about biodiversity more generally, beyond trees, as well. Unfortunately, we do not have access to data to include these dimensions in our analysis. However, we do have access to four additional performance indicators collected during the third forest inventory. These are (1) the total number of charcoal burning spots per hectare; (2) the total number of stumped trees per hectare; (3) the share of the sample plots whose regeneration status was rated as “good” or “very good” by OFWE forest experts; and (5) the share of sample plots whose overall status was rated as “good” or “very good.” Selling charcoal is one of the most common sources of forest income for forest users in our study area because it is the main source of energy to urban residents in Ethiopia. Previous studies have shown charcoal production is one of the main drivers of forest degradation (Sedano et al. 2016). We do not have data on the total amount of charcoal produced, but we believe that the number of charcoal burning spots may proxy for charcoal production. Similarly, counts of tree stumps have previously been used as a measure of forest degradation (Williams-Linera 2002). We aggregate individual variables in one index of forest conservation, following the approach by Kling et al. (2007). We define the summary index as the equally weighted average of z-scores of the components. The z-scores are calculated by subtracting the control group mean and dividing by the control group standard deviation.

We also implemented two survey waves among randomly selected FUG members, before and after implementing the monitoring interventions. On average, we surveyed one-third of the members from each group, as well as all group leaders (1222 members at baseline) and collected information on socio-demographic characteristics, political affiliation, and forest governance. We aggregated these co-variates at group level, and group-level baseline averages are used as co-variates in some models. We summarize dependent variables and co-variates in Appendix 1, Table A1.2.

Because of the randomized nature of our intervention, our statistical strategy can be simple. Various factors that affect conservation outcomes of FUGs such as local settings and circumstances, geographic characteristics (e.g., distance to market), level of income and poverty, alternative income generating activities (e.g., off-farm income activities), and climatic factors such as drought are, on average, the same across the experimental groups. This enables us to focus only on one source of exogenous variation, the monitoring intervention. Table A1.3 in Appendix 1 shows that treatment and control groups have, on average, the same characteristics and forest resources at baseline. Hence, randomization “worked” in terms of balancing baseline group characteristics. For 12 groups in our sample, we did not have access to forest conservation data (an attrition rate of 9%). Appendix 1, Table A1.4 shows that groups included in the analysis have, on average, the same characteristics as groups without forest data. This suggests that attrition is not systematic. Our identification strategy rests on regressing end-line forest outcome variables for group g, F_g, on group-level treatment dummies T_jg (j = 1, 2, 3), baseline co-variates (X_g), and kebelle fixed effects. Kebelles are a low level of government administration in Ethiopia. The average kebelle population is about 6500 people in our region. Our 120 FUGs are located in 13 kebelles, and the fixed effects capture local differences in geo-physical and climatic conditions, market development, and village-level governance.

(4)

In Equation (1), ε_g captures unobserved time-varying factors causing variation in forest outcomes. Vector β_1j contains our parameters of interest, which captures the causal effect of treatment j on forest outcome F. Vector X_g includes the following baseline variables: average age of FUG members, share of literate members, average livestock holding, average land holding, distance to daily market, year of FUG establishment, group size, share of female members, and size of the forest block. These co-variates are chosen based on the collective action literature and previous studies in the study area. Expanding the set of co-variates by including political affiliation of members does not affect the estimated results (see Appendix 1, Table A1.8). Our ANCOVA model also controls for the number of potential crop trees per hectare in 2004 and kebelle fixed effects; the latter are included to capture village level characteristics that may explain the forest outcome such as geo-physical factors and quality of local administration. We will also estimate variants of (1) in which we explain variation in our measures of group governance and member participation. These auxiliary models are intended to shed light on the mechanisms linking the interventions to conservation results.

RESULTS

Compared to the mean value of PCT for the control group at endline, top-down monitoring with punishment increases PCT/ha by about 53% (about 17 PCT/ha). Figure 2 summarizes the estimated coefficients for our main outcome variable, the number of potential crop trees per hectare (PCTs/ha) for both 95% and 90% confidence intervals. These estimated coefficients are based on ANCOVA models and include the number of potential crop trees per hectare in 2004 and kebelle fixed effects as explanatory variables (see column 3 in Table A1.5 in Appendix 1). Although all three interventions increased the number of PCTs/ha, only the intervention involving top-down monitoring with punishment of underperforming leaders is statistically significant.

Our results remain the same when using our alternative outcome variable, the forest conservation index, instead of PCTs/ha. Figure 3 summarizes the estimated coefficients for our forest conservation index for both 95% and 90% confidence intervals. Again, this is based on ANCOVA models, and we again control for kebelle fixed effects (see column 3 in Table A1.6 in Appendix 1). Estimation results for individual variables of the forest conservation index are presented in Appendix 1, Table A1.7. We aggregated the variables such that lower scores are associated with greater conservation. The estimated coefficients for both top-down monitoring treatments is significantly different from zero (p < 0.1), providing robust evidence that top-down monitoring of local leaders contributes to conservation in the context of CBFM.

We used the survey data to explore the mechanisms (highlighted in Figure 1) through which forest outcomes are affected by the three variations of monitoring. All interventions are designed to incentivize leaders to allocate more effort to their job, i.e., organizing, overviewing, and participating in patrols of the forest block to prevent incursions and resource theft by outsiders (or group members). We consider the (self-reported) number of days that FUG leaders spent in the forest for patrolling and monitoring purposes, and distinguish between the group leader and the full executive committee. The variable captures the number of days that group leadership is to some extent involved in monitoring and does not necessarily measure the number of “full days.” Results are summarized in Table 2.

Leaders exposed to top-down monitoring with the risk of punishment spent more days patrolling the forest than their peers in the control group. The same is true for leaders in the top-down monitoring with reward group, but this difference is much smaller and not significant. Leaders in the bottom-up monitoring seem to spend slightly fewer days patrolling than their counterparts in the control group, but this difference is also insignificant (Table 2, column 1). We also found that executive committee members in both of the top-down monitoring groups spent more days patrolling than their peers in the control group, which implies that the group leader is able to incentivize or force his fellow group leaders to supply greater effort. In contrast, there was no significant difference in patrolling effort between executive members from the bottom-up monitoring group and the control group (Table 2, column 2).

Although all the monitoring interventions increased patrolling effort by group members compared to the control group, this increase is significantly greater in the top-down monitoring groups than in the bottom-up monitoring group (p < 0.1). If leaders spend more time in the forest, this may reduce opportunities for theft by outsiders and group members thereby promoting conservation. The presence of leaders may also “crowd in” effort by other members, who may wish to signal their commitment to the group or copy the behavior of high-stratus individuals. It is also possible that leaders can more credibly demand that other group members work harder to patrol the forest when they do so themselves. These channels are consistent with the literature on leading by example in general (e.g., Hermalin 1998, Dionne et al. 2010) and the influence of local leaders on conservation outcomes in particular (e.g., Kosfeld and Rustagi 2015, Kolinjivadi et al. 2019, Duong and De Groot 2020). Column 3 in Table 2 supports the idea that group members also spent more days patrolling in the forest.

The increase in effort by group members in the bottom-up group likely reflects a response of members to the encouragement and support of OFWE staff (instead of a response to more effort by group leaders). But, as evident from Figures 2 and 3, this extra member effort does not translate into significant conservation success, thus behavioral change of FUG leadership is essential to promote conservation. Perhaps ordinary group members are unlikely to punish transgressions, especially when encountering side-selling or theft by high-status individuals. Anecdotal evidence suggests that punishment by leaders is tougher than punishment by ordinary members. If elite capture is a problem, and elites are powerful relative to other group members, then top-down monitoring by relevant authorities is required to hold leaders accountable. Empowering ordinary group members to hold their leaders accountable fails to produce the desired results. Indeed, we do observe that fewer leaders are removed from their position in the bottom-up group than in the top-down group with punishment (respectively, six leaders and nine leaders were dismissed), even if leader performance in the bottom-up group is markedly worse in terms of livelihood outcomes (Kahsay and Bulte 2021) and conservation outcomes.

DISCUSSION

Community-based forest management has decreased deforestation in developing countries in some cases (e.g., Porter-Bolland et al. 2012, Blackman et al. 2017, Baragwanath and Bayi 2020) and fails to decrease deforestation in other cases (e.g., Buntaine et al. 2015, BenYishay et al. 2017, Kraus et al. 2021). Community-based forest management is supposed to incentivize local communities to sustainably manage forest resources, particularly compared to government-managed forests, which are often de facto open access resources. Design, implementation, and governance issues have played a role in the heterogeneous conservation outcomes of CBFM (Agrawal and Gupta 2005, Ostrom 2009).

We show that in the context of our study, externally enhancing accountability through top-down monitoring by a government organization increases the number of potential crop trees and improves the overall state of forest conservation as assessed by forestry experts. We did not find that externally initiated bottom-up monitoring affected leader behavior in the context of CBFM in our study site. Consistent with earlier evidence, from other contexts and sectors, we found that local communities may struggle to hold their own leaders accountable (Olken 2007, Tesfaye et al. 2015, Casey 2018, Raffler et al. 2019, Kahsay and Bulte 2021). Kahsay and Bulte (2021) found that ordinary members did not benefit from our bottom-up monitoring intervention in terms of increased income. The current study documents that, compared to the control group, forest conservation does not improve either. This may reflect persistent information asymmetries, i.e., if group members are less efficient and effective in monitoring than government experts, or if power asymmetries prevent group members to follow-up on their findings.

Our results suggest that forest outcomes can be improved by strengthening the capacity of FUGs to monitor their forest use and management, and hold their leaders accountable. In our context, top-down monitoring by governmental officials delivered a significant improvement in forest outcomes. We believe it does so by providing FUGs with an assurance that evidence will be heard, a watchdog against abuse of power by local elites, and a platform to discuss governance challenges without fear of repercussions. As a result, we see increased effort in forest patrolling by group leaders, members of the executive committee, and ordinary members. Our results do not rule out the possibility of combining top-down and bottom-up monitoring as recently suggested (Casey 2018, Gupta and Koontz 2019, Raffler et al. 2019). Rigorous evidence has yet to come to support this.

Several caveats to our main results should be mentioned. The first caveat is conceptual in nature. It’s important to emphasize that our interventions are implemented in addition to existing CBFM and represent an addition or extension of a pre-existing local community regime. Top-down monitoring in this study is neither a replacement for CBFM, nor a substitution for monitoring efforts by local communities. When assisting FUGs with top-down monitoring by a governmental authority, special care should be taken not to exacerbate existing power asymmetries and/or concentrate decision-making power in the hands of an “enforcers” because this will effectively undermine the whole purpose of CBFM. Similarly, the use of monitoring by governments should be carefully designed to avoid eroding the community’s motivation for engaging in natural resource management in general and monitoring in particular. This type of crowding out has been documented in previous research (e.g., Cardenas et al. 2000) and should be avoided.

Another conceptual caveat concerns the nature of our bottom-up monitoring intervention. On average, the impact of fostering bottom-up monitoring produces disappointing results, but this should not be misconstrued as an argument against bottom-up monitoring per se. Our study focuses on an externally initiated intervention, which may be different from endogenously evolving community-initiated monitoring regimes. Although our study speaks to whether outside agencies can readily promote community engagement with monitoring, it is silent on the usefulness of homegrown monitoring in successful groups. However, heterogeneity at the group level is likely important. Kahsay and Bulte (2021) documented that bottom-up monitoring fails to improve average livelihood outcomes of group members, however the intervention’s impact varied depending on the extent to which groups were able to earn an income based on their forest resources prior to the intervention. The impact of outside interventions can leverage or crowd-out pre-existing endogenous institutions. For example, auxiliary empirical analyses (not shown) suggest that the positive conservation impact of the top-down monitoring intervention studied is driven by outcomes in FUGs when prior to the intervention forest-based income of group members was low, thereby where community-based monitoring did not work well. In contrast, top-down monitoring did not meaningfully promote forest conservation for the subsample of FUGs in which group members earned high forest-based incomes at baseline. Apparently pre-existing community-level institutions worked there.

Second, future studies can improve upon our measures of forest conservation. Our main variable captures the density of potential crop trees, which leaves out many important aspects of forest conservation. Our index of forest conservation is also primarily tree based. Moreover, our intervention period was only 2.5 years, which is admittedly short to pick up many relevant changes in the domain of forest conservation. We encourage future studies that consider a longer time frame.

Third, an issue related to design choices warrants attention. To preserve statistical power, we could not implement a factorial (i.e., 2 × 2) design and had to drop the treatment arm combining bottom-up monitoring with a reward for good leadership. It is an open question about how this combination would have fared.

Finally, care should be taken to transplant findings from this case study to other contexts. Forest user groups in our study area are, on average, 12 years old, and were established and supported by the Ethiopian Government in collaboration with GIZ. Local conditions regarding FUG formation and capacity will be quite different elsewhere, as will be ecological conditions and power asymmetries between local parties.

CONCLUSION

This study documents the impact of adding bottom-up and top-down monitoring to CBFM on several measures of forest conservation using a field experiment and contributes to the debate about the design of effective and transparent institutions for community-based forest conservation. We document that adding top-down monitoring to CBFM promotes forest conservation in the context that we study. This suggests that external scrutiny by outsiders matters for leader behavior. We found that enhanced bottom-up monitoring by group members themselves does not improve forest outcomes, which is consistent with recent experimental findings by Christensen et al. (2021; in Liberia) and Eisenbarth et al. (2021; in Uganda), and contrasts with Slough et al. (2021b), who find evidence of reduced forest cover loss due to community-based monitoring in the Peruvian Amazon. Our results suggest that adding top-down, governmental monitoring to an existing CBFM improves conservation by inducing leaders to exert more effort, which crowds in effort by other group members. Whether similar effects are obtained in other contexts should be explored in future studies.

Our results are not about reducing the role of communities in forest management, reducing the involvement of community members in forest patrolling, or monitoring of their leaders. Our results are a plea for outside oversight by the relevant authority to help communities overcome local power asymmetries. Community management appears to promote conservation, and in most settings, it is hard to imagine how forests can be sustainably managed without engaging local communities. But within the sample of CBFM forest blocks, conservation success can be (further) improved by keeping the governmental organization responsible for forest management involved in the monitoring of group leadership.

LITERATURE CITED

Africa Forest and Landscape Restoration Initiative (AFR100). 2022. Investing in Africa’s local land restoration champions. Africa Forest and Landscape Restoration Initiative, Midrand, South Africa. https://afr100.org/

Agrawal, A. 2001. Common property institutions and sustainable governance of resources. World Development 29:1649-1672. https://doi.org/10.1016/S0305-750X(01)00063-8

Agrawal, A. 2003. Sustainable governance of common-pool resources: context, methods, and politics. Annual Review of Anthropology 32(1):243-262. https://doi.org/10.1146/annurev.anthro.32.061002.093112

Agrawal, A., and K. Gupta. 2005. Decentralization and participation: the governance of common pool resources in Nepal’s Terai. World Development 33:1101-1114. https://doi.org/10.1016/j.worlddev.2005.04.009

Ameha, A., H. Meilby, and G. L. Feyisa. 2016. Impacts of participatory forest management on species composition and forest structure in Ethiopia. International Journal of Biodiversity Science, Ecosystem Services and Management 12:139-153. https://doi.org/10.1080/21513732.2015.1112305

Ameha, A., O. J. Nielsen, and H. O. Larsen. 2014b. Impacts of access and benefit sharing on livelihoods and forest: case of participatory forest management in Ethiopia. Ecological Economics 97:162-171. https://doi.org/10.1016/j.ecolecon.2013.11.011

Ameha, A., H. O. Larsen, and M. Lemenih. 2014a. Participatory forest management in Ethiopia: learning from pilot projects. Environmental Management 53:838-854. 10.1007/s00267-014-0243-9

Amente, G. 2005. Rehabilitation and sustainable use of degraded community forests in the Bale Mountains of Ethiopia. Dissertation. Institute of Silviculture, Albert Ludwigs University of Freiburg, Breisgau, Germany.

Baggio, J. A., A. J. Barnett, I. Perez-Ibarra, U. Brady, E. Ratajczyk, N. Rollins, C. Rubiños, H. C. Shin, D. J. Yu, R. Aggarwal, J. M. Anderies, and M. A. Janssen. 2016. Explaining success and failure in the commons: the configural nature of Ostrom’s institutional design principles. International Journal of the Commons 10(2). https://doi.org/10.18352/ijc.634

Baragwanath, K., and E. Bayi. 2020. Collective property rights reduce deforestation in the Brazilian Amazon. Proceedings of the National Academy of Sciences 117(34):20495-20502. https://doi.org/10.1073/pnas.1917874117

Bardhan, P. 2002. Decentralization of governance and development. Journal of Economic Perspectives 16(4):185-205. https://doi.org/10.1257/089533002320951037

BenYishay, A., S. Heuser, D. Runfola, and R. Trichler. 2017. Indigenous land rights and deforestation: evidence from the Brazilian Amazon. Journal of Environmental Economics and Management 86:29-47. https://doi.org/10.1016/j.jeem.2017.07.008

Bernedo Del Carpio, M., F. Alpizar, and P. J. Ferraro. 2021. Community-based monitoring to facilitate water management by local institutions in Costa Rica. Proceedings of the National Academy of Sciences 118(29):e2015177118. https://doi.org/10.1073/pnas.2015177118

Björkman, M., and J. Svensson. 2009. Power to the people: evidence from a randomized field experiment on community-based monitoring in Uganda. Quarterly Journal of Economics 124(2):735-769.

Blackman, A., L. Corral, E. S. Lima, and G. P. Asner. 2017. Titling indigenous communities protects forests in the Peruvian Amazon. Proceedings of the National Academy of Sciences 114:4123-4128. https://doi.org/10.1073/pnas.1603290114

Buntaine, M. T., S. E. Hamilton, and M. Millones. 2015. Titling community land to prevent deforestation: an evaluation of a best-case program in Morona-Santiago, Ecuador. Global Environmental Change 33:32-43. https://doi.org/10.1016/j.gloenvcha.2015.04.001

Cardenas, J. C., J. Stranlund, and C. Willis. 2000. Local environmental control and institutional crowding-out. World Development 28(10):1719-1733. https://doi.org/10.1016/S0305-750X(00)00055-3

Casey, K. 2018. Radical decentralization: does community-driven development work? Annual Review of Economics 10:139-163. https://doi.org/10.1146/annurev-economics-080217-053339

Christensen, D., A. C. Hartman, and C. Samii. 2021. Citizen monitoring promotes informed and inclusive forest governance in Liberia. Proceedings of the National Academy of Sciences 118:e2015169118. https://doi.org/10.1073/pnas.2015169118

Coleman, E. A., and F. D. Fleischman. 2012. Comparing forest decentralization and local institutional change in Bolivia, Kenya, Mexico, and Uganda. World Development 40(4):836-849. https://doi.org/10.1016/j.worlddev.2011.09.008

Dessie, G., and C. Christiansson. 2008. Forest decline and its causes in the south-central rift valley of Ethiopia: human impact over a one hundred year perspective. Ambio 263-271.

Dietz, T., E. Ostrom, and P. C. Stern. 2003. The struggle to govern the commons. Science 302:1907-1912. https://doi.org/10.1126/science.1091015

Dionne, S., H. Sayama, C. Hao, and B. J. Bush. 2010. The role of leadership in shared mental model convergence and team performance improvement: an agent-based computational model. Leadership Quarterly 21(6):1035-1049. https://doi.org/10.1016/j.leaqua.2010.10.007

Duong, N. T. B., and W. T. De Groot. 2020. The impact of payment for forest environmental services (PFES) on community-level forest management in Vietnam. Forest Policy and Economics 113:102135. https://doi.org/10.1016/j.forpol.2020.102135

Eisenbarth, S., L. Graham, and A. S. Rigterink. 2021. Can community monitoring save the commons? Evidence on forest use and displacement. Proceedings of the National Academy of Sciences 118:e2015172118. https://doi.org/10.1073/pnas.2015172118

Faguet, J.-P. 2014. Decentralization and governance: a special issue of World Development, 2013. World Development 53:1-112. https://doi.org/10.1016/j.worlddev.2013.08.001

Ferraro, P. J., and A. Agrawal. 2021. Synthesizing evidence in sustainability science through harmonized experiments: community monitoring in common pool resources. Proceedings of the National Academy of Sciences 118:e2106489118. https://doi.org/10.1073/pnas.2106489118

Food and Agriculture Organization of the United Nations (FAO). 2020. Forests and poverty reduction. FOA, Rome, Italy. http://www.fao.org/forestry/livelihoods/en/

Fritzen, S. A. 2007. Can the design of community-driven development reduce the risk of elite capture? Evidence from Indonesia. World Development 35(8):1359-1375. https://doi.org/10.1016/j.worlddev.2007.05.001

Grossman, G., and W. W. Hanlon. 2014. Do better monitoring institutions increase leadership quality in community organizations? Evidence from Uganda. American Journal of Political Science 58(3):669-686. https://doi.org/10.1111/ajps.12071

Gupta, D., and T. M. Koontz. 2019. Working together? Synergies in government and NGO roles for community forestry in the Indian Himalayas. World Development 114:326-340. https://doi.org/10.1016/j.worlddev.2018.09.016

Hardin, G. 1978. Political requirements for preserving our common heritage. Wildlife and America 310-317.

Hermalin, B. 1998. Toward an economic theory of leadership: leading by example. American Economic Review 88:1188-1206.

Jayachandran, S., J. De Laat, E. F. Lambin, C. Y. Stanton, R. Audy, and N. E. Thomas. 2017. Cash for carbon: a randomized trial of payments for ecosystem services to reduce deforestation. Science 357(6348):267-273. https://doi.org/10.1126/science.aan0568

Kahsay, G. A., and E. Bulte. 2019. Trust, regulation and participatory forest management: micro-level evidence on forest governance from Ethiopia. World Development 120:118-132. https://doi.org/10.1016/j.worlddev.2019.04.007

Kahsay, G. A., and E. Bulte. 2021. Internal versus top-down monitoring in community resource management: experimental evidence from Ethiopia. Journal of Economic Behavior and Organization 189:111-131. https://doi.org/10.1016/j.jebo.2021.06.030

Kahsay, G. A., and H. Medhin. 2020. Leader turnover and forest management outcomes: micro-level evidence from Ethiopia. World Development 127:104765. https://doi.org/10.1016/j.worlddev.2019.104765

Kahsay, G. A., A. Nordén, and E. Bulte. 2021. Women participation in formal decision-making: empirical evidence from participatory forest management in Ethiopia. Global Environmental Change 70:102363. https://doi.org/10.1016/j.gloenvcha.2021.102363

Kling, J. R., J. B. Liebman, and L. F. Katz. 2007. Experimental analysis of neighborhood effects. Econometrica 75:83-119. https://doi.org/10.1111/j.1468-0262.2007.00733.x

Kolinjivadi, V., S. Charré, S., J. Adamowski, and N. Kosoy. 2019. Economic experiments for collective action in the Kyrgyz Republic: lessons for payments for ecosystem services (PES). Ecological Economics 156:489-498. https://doi.org/10.1016/j.ecolecon.2016.06.029

Kosfeld, M., and D. Rustagi. 2015. Leader punishment and cooperation in groups: experimental field evidence from commons management in Ethiopia. American Economic Review 105:747-783. https://doi.org/10.1257/aer.20120700

Kraus, S., J. Liu, N. Koch, and S. Fuss. 2021. No aggregate deforestation reductions from rollout of community land titles in Indonesia yet. Proceedings of the National Academy of Sciences 118(43):e2100741118. https://doi.org/10.1073/pnas.2100741118

Larson, A. M., P. Pacheco, F. Toni, and M. Vallejo. 2007. The effects of forestry decentralization on access to livelihood assets. Journal of Environment and Development 16:251-268. https://doi.org/10.1177/1070496507306220

Lund, J. F., and M. Saito-Jensen. 2013. Revisiting the issue of elite capture of participatory initiatives. World Development 46:104-112. https://doi.org/10.1016/j.worlddev.2013.01.028

Olken, B. A. 2007. Monitoring corruption: evidence from a field experiment in Indonesia. Journal of Political Economy 115:200-249. https://doi.org/10.1086/517935

Ostrom, E. 1990. Governing the commons: the evolution of institutions for collective action. Cambridge University Press, Cambridge, UK. https://doi.org/10.1017/CBO9780511807763

Ostrom, E. 2009. A general framework for analyzing sustainability of social-ecological systems. Science 325:419-422. https://doi.org/10.1126/science.1172133

Ostrom, E., M. A. Janssen, and J. M. Anderies. 2007. Going beyond panaceas. Proceedings of the National Academy of Sciences 104:15176-15178. https://doi.org/10.1073/pnas.0701886104

Ostrom, E., and H. Nagendra. 2006. Insights on linking forests, trees, and people from the air, on the ground, and in the laboratory. Proceedings of the National Academy of Sciences 103:19224-19231. https://doi.org/10.1073/pnas.0607962103

Pancel, L. 2015. Technical orientation of silviculture in the tropics. Pages 1-42 in L. Pancel and M. Köhl, editors. Tropical forestry handbook. Springer, Berlin, Germany. https://doi.org/10.1007/978-3-642-41554-8_122-2

Porter-Bolland, L., E. A. Ellis, M. R. Guariguata, I. Ruiz-Mallén, S. Negrete-Yankelevich, and V. Reyes-García. 2012. Community managed forests and forest protected areas: an assessment of their conservation effectiveness across the tropics. Forest Ecology and Management 268:6-17. https://doi.org/10.1016/j.foreco.2011.05.034

Raffler, P., D. N. Posner, and D. Parkerson. 2019. The weakness of bottom-up accountability: experimental evidence from the Ugandan health sector. Innovations for Poverty Action working paper. Innovations for Poverty Action, Washington, D.C., USA. https://poverty-action.org/sites/default/files/publications/The%20Weakness%20of%20Bottom-Up%20Accountability.pdf

Rustagi, D., S. Engel, and M. Kosfeld. 2010. Conditional cooperation and costly monitoring explain success in forest commons management. Science 330:961-965. https://doi.org/10.1126/science.1193649

Saunders, F. P. 2014. The promise of common pool resource theory and the reality of commons projects. International Journal of the Commons 8:636-656. https://doi.org/10.18352/ijc.477

Sedano, F., J. A. Silva, R. Machoco, C. H. Meque, A. Sitoe, N. Ribeiro, K. Anderson, Z. A. Ombe, S. H. Baule, and C. J. Tucker. 2016. The impact of charcoal production on forest degradation: a case study in Tete, Mozambique. Environmental Research Letters 11:094020. https://doi.org/10.1088/1748-9326/11/9/094020

Sikor, T., and T. Q. Nguyen. 2007. Why may forest devolution not benefit the rural poor? Forest entitlements in Vietnam’s central highlands. World Development 35(11):2010-2025. https://doi.org/10.1016/j.worlddev.2006.11.011

Slough, T., J. Kopas, and J. Urpelainen. 2021b. Satellite-based deforestation alerts with training and incentives for patrolling facilitate community monitoring in the Peruvian Amazon. Proceedings of the National Academy of Sciences 118:e2015171118. https://doi.org/10.1073/pnas.2015171118

Slough, T., D. Rubenson, R. Levy, F. Alpizar Rodriguez, M. Bernedo del Carpio, M. T. Buntaine, D. Christensen, A. Cooperman, S. Eisenbarth, P. J. Ferraro, L. Graham, A. C. Hartman, J. Kopas, S. McLarty, A. S. Rigterink, B. Seim, J. Urpelainen, and B. Zhang. 2021a. Adoption of community monitoring improves common pool resource management across contexts. Proceedings of the National Academy of Sciences 118:e2015367118. https://doi.org/10.1073/pnas.2015367118

Spicer, M. E., H. Mellor, and W. P. Carson. 2020. Seeing beyond the trees: a comparison of tropical and temperate plant growth forms and their vertical distribution. Ecology 101(4):e02974. https://doi.org/10.1002/ecy.2974

Tesfaye, Y., M. Bekele, M. Kebede, F. Tefera, and H. Kassa. 2015. Enhancing the role of forestry in Ethiopia: strategy for scaling up effective forest management practices in Oromia with emphasis on participatory forest management. CIFOR Ethiopia Office, Addis Ababa, Ethiopia.

Williams-Linera, G. 2002. Tree species richness complementarity, disturbance and fragmentation in a Mexican tropical montane cloud forest. Biodiversity and Conservation 11:1825-1843. https://doi.org/10.1023/A:1020346519085

Woody Biomass Inventory and Strategic Planning Project (WBISPP). 2001. A strategic plan for the sustainable development, conservation and management of woody biomass resources, Oromia Regional State. Oromia Regional State, Addis Ababa, Ethiopia.

Young, O. R., D. G. Webster, M. E. Cox, J. Raakjær, L. Ø. Blaxekjær, N. Einarsson, R. A. Virginia, J. Acheson, D. Bromley, E. Cardwell, C. Carothers, E. Eythórsson, R. B. Howarth, S. Jentoft, B. J. McCay, F. McCormack, G. Osherenko, E. Pinkerton, R. van Ginkel, J. A. Wilson, L. Rivers, III, and R. S. Wilson. 2018. Moving beyond panaceas in fisheries governance. Proceedings of the National Academy of Sciences 115(37):9065-9073. https://doi.org/10.1073/pnas.1716545115