Community forests, timber production, and certification: success factors in the African context

. In Africa, forest tenure reform to decentralize forest management from central governments to local communities has been occurring since the 1990s to promote forest conservation, poverty alleviation, and sustainable forest-based livelihoods. African governments and donor organizations continue to invest in community forestry, raising the question of what contributes to "success." The present study examines social, economic, and biophysical factors that contribute to success, or lack thereof, in community forestry in southeastern Tanzania. There, community forest enterprises produce commercial timber from natural stands in community forests certified by the Forest Stewardship Council. We compare success across 14 certified community forests using local criteria, emphasizing total revenue earned because half of this revenue is invested in creating community benefits, which incentivize sustainable forest management. Our methods include financial analysis, analysis of potential linkages between success outcomes and attributes of villages and community forests, key informant interviews, and a survey of forest managers. We found that for community forestry with community forest enterprises that produce commercial timber to be most successful, success factors for both community forestry and small-scale, forest-based enterprises must be present and co-occur. In particular, we highlight the importance of having large community forests with a high natural endowment of merchantable timber species for success. A favorable national policy environment, good governance and support at the community and district government levels, secure tenure, tangible benefits, and long-term technical and financial support from a local organization were also important. In addition, we examine the role of forest certification in contributing to success, finding it does so by reinforcing many other success factors. We draw insights from the Tanzanian case that may be relevant for community forestry elsewhere in Africa where community forest enterprises producing commercial timber operate or are desired.


INTRODUCTION
Since the late 1970s, many low-and middle-income countries worldwide have decentralized forest governance to address deforestation, forest degradation, and poverty among forest peoples (Charnley and Poe 2007). Often referred to as "community forestry," these efforts entail formal devolution of some degree of responsibility and authority for forest management by governments to local communities, with the goals of creating benefits for communities while fostering ecologically sustainable forest use and management. In Africa, tenure reform to decentralize forest management has been occurring since the 1990s (Barrow et al. 2016) to promote conservation of existing forests, restoration of degraded forests, poverty alleviation, and sustainable forest-based livelihoods. Nevertheless, only ca. 7.4% of forests in low-and middle-income African countries (n = 11) were designated for or owned by communities and/or indigenous peoples in 2017, compared with 32.6% in Asia (n = 13) and 36.2% in Latin America (n = 9) (Ginsburg and Keene 2020). Given the potential for expansion and ongoing investment in community forestry by African governments and donor organizations, of interest are the factors that contribute to "success." To date, most research on this topic comes from Latin America and Asia.
Community forestry takes many forms and may entail establishing community forest enterprises (CFEs). We define CFEs as collective community-based businesses that are directed by community governance bodies, are managed for multiple goals, and produce commercial products or services from community forests to benefit community members (Antinori and Bray 2005). Community forest enterprises ideally generate profits, are financially and ecologically sustainable, and create both monetary and non-monetary benefits for communities (Peredo and Chrisman 2006, Soviana 2015. Revenue from sustainable production of forest goods and services creates community benefits via diverse pathways (e.g., direct payments to community members, income earnings for CFE workers and shareholders, reinvestment of profits in community development projects) (Antinori and Bray 2005). These benefits in turn incentivize community forestry.
This study examines factors that contribute to success in community forestry where CFEs produce commercial timber from natural forest stands. To our knowledge, the only countries in Africa with this model are Tanzania and Cameroon. Nevertheless, some countries have provisions to help local communities benefit from commercial timber production on state lands, including Mozambique (Sitoe and Guedes 2015), Ghana (Agyei and Adjei 2017), and Ethiopia (Ameha et al. 2014, Gatiso 2019. There is potential to expand this model to incentivize sustainable forest management through community forestry in Africa. There is also potential to adapt and expand forest certification in Africa in the community forestry context. Forest certification is a voluntary tool that producers use to demonstrate that their forests are managed in a socially, economically, and environmentally sustainable manner (Vogt et al. 2000). They do so by labeling their products using the name of the organization whose certification standards and criteria they meet. Certification incentivizes sustainable forest management by providing financial Secure tenure rights over land and trees Pagdee et al. 2006, Brooks et al. 2012, Baynes et al. 2015, Gilmour 2016, Hajjar and Oldekop 2018, Hajjar et al. 2020 Supportive government policies and programs Blomley and Ramadhani 2006, Baynes et al. 2015, Gilmour 2016, Hajjar and Oldekop 2018, Frey et al. 2019 Meaningful devolution of forest management responsibility and authority from state governments to local communities Charnley and Poe 2007 Effective community-level forest governance institutions Pagdee et al. 2006, Oldekop et al. 2010, Brooks et al. 2012, Baynes et al. 2015, Brooks 2017 Education and training to develop individual, organizational, and community capacity for effective forest management Pagdee et al. 2006, Brooks et al. 2012, Brooks 2017 Tangible community benefits that are equitably distributed among community members Pagdee et al. 2006, Brooks et al. 2012, Baynes et al. 2015 A "community of practice" consisting of community members and external entities (e.g., governments, development partners, conservation organizations) with common interests who work together to achieve goals Pagdee et al. 2006, Arts and de Koning 2017, Segura Warnholtz et al. 2020 or reputational benefits (Ehrenberg-Azcárate and Peña-Claros 2020). To date, Tanzania is the only African country where CFEs produce timber from community-managed natural forest stands that have been certified by the Forest Stewardship Council (FSC). In Cameroon, national certification standards for community forests were developed by the FSC in 2010 but have not been applied (Minang et al. 2019). Forest certification in contexts other than community forestry has waxed and waned in Africa since the 1990s and been more prevalent in southern than eastern Africa (Kalonga et al. 2019). As of 2016, there were 17 FSC certificates for the management of natural tropical forests in Africa, representing 5.45 million ha of forest area, occurring mainly in Gabon and Republic of the Congo (Ehrenberg-Azcárate and Peña-Claros 2020). Most of these certificates were held by private companies (Ehrenberg-Azcárate and Peña-Claros 2020).
In this article, we investigate what success in community forestry means as locally defined; the social, economic, and biophysical factors that enable it; and the role of forest certification in contributing to it. We draw insights from the Tanzanian case that may be relevant elsewhere in Africa.

Success Factors in Community Forestry and Community Forest Enterprises
Case studies of community forestry in low-and middle-income countries have been the object of numerous comparative and meta-analyses to identify factors, or enabling conditions, contributing to success (Table 1). Success is typically evaluated in terms of achieving environmental goals such as forest conservation and carbon sequestration, economic goals associated with poverty alleviation, and social goals such as increasing land and resource tenure security and equity. It is important that these factors co-occur; if only some are present, success will likely be more limited (Baynes et al. 2015, de Jong et al. 2016. A recent meta-analysis of the outcomes of community forestry finds that it often meets environmental and incomerelated goals but has more negative outcomes for forest resource access rights (Hajjar et al. 2020).
Another body of work examines aspects of CFEs that lead to revenue generation and prosperity (Hajjar and Oldekop 2018).
Community forest enterprises engaged in commercial timber production usually generate more income than CFEs based on other forest products (de Jong et al. 2016). As with community forestry, researchers have undertaken comparative and metaanalyses of individual CFEs, and small and medium-size forest enterprises (not run by community forest groups), to identify factors underlying success (Table 2). Some are the same as those identified for community forestry: secure tenure over forest resources, a supportive national policy environment, and effective community organizations for managing forest resources. Again, these conditions operate together; if only some are present, success is likely to be compromised (Sanchez Badini et al. 2018).
A related literature pertains to governance institutions for common-pool resource management. Community forests are common-pool resources (it is difficult to exclude people from benefiting from them, and resource consumption by one person subtracts from resource benefits to others) (Ostrom 1990). Since Ostrom's (1990) seminal work on this topic, numerous researchers have investigated design principles that lead to enduring institutions for sustainably governing common-pool resources (see Agrawal 2001 for a review). Community-scale resource governance institutions are an important success factor for community forestry and CFEs, thus these design principles add insight. However, this literature does not consider the broader suite of conditions that underlie successful community forestry or CFEs.  Blomley and Ramadhani 2006, Macqueen 2013, Anderson et al. 2015, Cubbage et al. 2015, de Jong et al. 2016, Sanchez Badini et al. 2018, Segura Warnholtz et al. 2020 The ability to produce diverse forest products Foundjem-Tita et al. 2018 A positive business environment in which to operate Cubbage et al. 2015, de Jong et al. 2016, Sanchez Badini et al. 2018, Gnych et al. 2020 Favorable market conditions for the product being sold and access to markets Gilmour 2016, Sanchez Badini et al. 2018, Segura Warnholtz et al. 2020 High-capacity community-level organizations for managing the enterprise Macqueen 2013, Gilmour 2016, Sanchez Badini et al. 2018  occurs when communities enter into a joint forest management agreement with the central government or local district council to co-manage a government-designated forest reserve. Community-based forest management occurs when communities designate "village land forest reserves" on village lands. Reserve management is conducted by village governments or other specified entities for production and/or protection based on sustainable management objectives and forest management plans. We focus on the second approach. We use the terms "community forestry" instead of community-based forest management, and "community forest" instead of village land forest reserve, because these terms are common in the international literature and reflect the same concepts as the Tanzanian terminology.
As of 2012, an estimated 546,000 ha of forestland in 381 villages throughout Tanzania had been designated as community forests (Pailler et al. 2015). Community forestry in Tanzania appears to function better than in several other African countries . One reason is favorable national laws and policies that support secure forest tenure, effectively devolve forest management authority and responsibility to the community level, and enable villages to fully retain financial benefits from CFEs (Alden Wily 2011Wily , 2018. Blomley and Ramadhani (2006) highlight three such laws. The Local Government Act of 1982 gave Village Councils (elected village government bodies) the authority to make bylaws regulating natural resource access and use within village boundaries. The Village Land Act of 1999 called for villages to survey their boundaries and formally register as villages, giving them official tenure rights over village lands and resources, and authority to manage them. The Forest Act of 2002 allowed registered villages to gazette community forests within village boundaries to be managed by the Village Council under a forest management plan, and villagers to harvest forest products there consistent with the plan. Importantly, villages do not have to pay royalties or taxes to the district, regional, or national governments on products harvested and sold, including timber, which allows them to retain all earnings (Blomley and Ramadhani 2006). As of 2017, an estimated 45.6% of Tanzania's forestlands (22 million ha) were owned by or designated for local communities, compared with 5.2% of forestlands averaged across 11 African countries (including Tanzania) (Ginsburg and Keene 2020).
Nevertheless, community forestry in Tanzania is not without challenges. These include limited financial and technical capacity for sustainable forest management at the village level; lack of adequate staffing and resources in district government forestry offices to provide technical support to villages ; complex bureaucratic process requirements for initial establishment (Gross-Camp 2017); unsustainable levels of tree harvesting for charcoal production close to urban areas (Treue et al. 2014); potential land shortages outside of community forests to accommodate community needs (Brockington 2007); and encroachment by livestock into community forests in some areas . Villages may also have limited ability to generate revenue from community forests owing to lack of merchantable timber, or high biodiversity or watershed protection values that create restrictions on forest products harvesting Ramadhani 2006, Gross-Camp 2017).

The Mpingo Conservation and Development Initiative
This study focuses on community forestry in 14 villages located across three districts and three regions of southeastern Tanzania (Fig. 1). Forests in the project area are predominantly miombo woodland and East African Coastal Forest. Numerous commercially valuable hardwoods occur there, the most important of which is East African Blackwood (Dalbergia melanoxylon), known as "mpingo" in Kiswahili. Mpingo is the national tree of Tanzania, has wood qualities that make it valuable for manufacturing woodwind instruments such as clarinets, oboes, and bagpipes, and is one of the most expensive timber species in the world by volume (Ball 2004). Because of its high value and slow growth rate (70-100 yr to reach a merchantable size), the only places in Africa where a commercial trade in mpingo remains viable are southern Tanzania and Mozambique (Ball 2004).
Community forestry in southeastern Tanzania arose partly to conserve mpingo and encourage its sustainable harvest (Gross-Ecology and Society 27(3): 6 https://www.ecologyandsociety.org/vol27/iss3/art6/ . The region's forests have been under threat from improvements in transportation infrastructure that increased connectivity to Dar es Salaam (Tanzania's largest city) and Mtwara (the regional center in southeastern Tanzania), forest conversion to agriculture, illegal logging, charcoal production, and fire (Ahrends et al. 2010, Miya et al. 2012, Dokken et al. 2014. In 2004, these drivers led to establishment of the Mpingo Conservation and Development Initiative (MCDI), a Tanzanian non-governmental organization (NGO), with support from international donor organizations. The MCDI's mission is "to advance forest conservation and community development in Tanzania by facilitating sustainable and socially equitable utilization of forest resources," with timber serving as an economic tool for advancing conservation (MCDI 2021).
Between 2006 and 2018, MCDI facilitated establishment of community forests encompassing 413,700 ha in 43 villages in southeastern Tanzania following the process set forth in Tanzanian's Forest Act. As of 2018, 14 of these (totaling 197,435 ha) were certified by the FSC under a group certificate, with standards developed for Tanzania as part of FSC's initiative for small or low intensity managed forests (SLIMF). The MCDI obtained the group certificate in 2009 and is responsible for managing it. Village natural resource committees (VNRCs), part of the Village Council and comprising 17-20 elected members serving 3-yr terms, are responsible for community forest management. The MCDI provides training to VNRCs to build capacity for forest management and ensure it meets FSC standards. It also pays all costs associated with forest certification, subsidizes the cost of forest management, and supports all aspects of the CFEs, including marketing wood products.
At the time of this study (2017)(2018), CFEs in the MCDI project area operated by selling roundwood to timber buyers, who undertook harvests with oversight from VNRC members. Since then, CFEs themselves have conducted some harvests and processed the wood using a portable sawmill purchased by MCDI. Other sources of revenue (non-timber forest products, tourism) largely have not developed. The potential for payments for carbon offsets generated by implementing fire management practices designed to reduce forest losses to wildfire was evaluated under a REDD+ pilot project but deemed unrealistic (Khatun et al. 2015).
To incentivize villagers to support community forestry, about 50% of community forest revenues go to the Village Council. The Village Council proposes how to invest the earnings, subject to approval by the Village Assembly (village residents 18 yr and older). Funds are usually invested in community development projects or community assistance programs. Examples include construction of community buildings, health care and water projects, support for education expenses and facilities, and financial assistance to people in need, such as expectant mothers.
Ecology and Society 27(3): 6 https://www.ecologyandsociety.org/vol27/iss3/art6/ Villages with high population density may be unable to create large community forests and exert greater pressure to convert community forests to alternative land uses.

Proportion of village land in community forest
The more village land in a community forest, the larger the community forest, which may reflect a strong commitment to community forestry. But, higher proportions of community forest land leave less land available for alternative forest uses-increasing potential for illegal activity and pressure for conversion to other uses. District in which village is located District governments play an important role in supporting community forests and assisting with enforcement actions when bylaws are violated.

Distance of village to MCDI
The closer the village to MCDI, the easier it is for MCDI staff to provide support for community forestry.

Community forest attributes
Year community forest created Older community forests might have more experienced VNRCs, increasing capacity for good governance; better market connections; and have conducted more timber sales and community development activities. Community forest size Larger community forests likely have more diverse and greater volumes of merchantable timber species, increasing potential for revenue generation. Harvesting in large community forests is more efficient for buyers who can purchase multiple sales. Size of community forest per capita Community forest assets and revenue earned per capita can be meaningful indicators of success. Productive area of community forest The larger the productive area of a community forest, the greater the potential for revenue generation from timber sales.

Merchantable volume of mpingo
Mpingo has the highest market demand and commercial value and is preferred among buyers. Merchantable volume of eight merchantable hardwood species † The more merchantable timber in a community forest, the greater the potential for revenue generation. Of the remaining funds, 40% fund forest management activities by VNRCs, 5% are voluntarily distributed to the local District Council, and 5% go to MCDI as partial compensation for technical assistance.

METHODS
We examine success through a comparative analysis of the 14 community forests participating in MCDI's FSC group certificate in 2017-2018, when we conducted fieldwork. Field visits to the MCDI office in Kilwa Masoko, the Kilwa District Office, and villages with certified community forests in Kilwa District occurred in October-November 2017 and July 2018. We solicited criteria for evaluating success, identified which community forests were successful or unsuccessful relative to others, and explored factors contributing to success or lack thereof. We also examined the role of FSC certification in contributing to success. We used four methods: (1) analysis of village-level financial data pertaining to community forest management and CFE operations; (2) analysis of linkages between success outcomes and social and biophysical attributes of community forests and associated villages; (3) key informant interviews with MCDI staff and partner organizations; and (4) a survey of VNRC members.

Financial Analysis
Key informant interviews led us to focus our financial analysis on how much revenue community forests generate as a key measure of success. Appendix 1 provides a detailed description of methods used for the financial analysis. In short, we compiled, categorized, and analyzed revenue data obtained from MCDI for the 14 community forests for fiscal years 2011-2012 through 2017-2018, the years for which complete data were available at the time of fieldwork. Revenues mainly come from timber sales, but include other sources like fees, fines, and auction of confiscated timber or equipment from illegal loggers. We used the Green Value Tool (https://www.green-value.org/home) to compile and analyze the data. This tool was developed to help forest initiatives track costs and revenues, evaluate financial viability, and identify areas for improvement (Humphries and Holmes 2015). We used total revenue earned rather than average annual earnings because total revenue reflects a community forest's performance over time and a longer history of performance.

Village and Community Forest Attributes
We obtained data from the MCDI office for 12 social and biophysical attributes of the 14 certified community forests, and the villages where they are located, that might be linked to success (Table 3). We chose these attributes based on review of literature, discussions with MCDI staff during field visits, and data availability. We analyzed the attributes by developing matrices in Excel that enabled cross-village comparison by attribute, and by looking for those that were consistently associated with the successful and unsuccessful community forests, as defined by revenue earned. We estimated the Pearson correlation to identify potential linkages between success outcomes and ten of the community forest/village attributes. We did not correlate revenue outcomes with percentage of land in community forest because this variable is calculated from other variables (village size, size of community forest) that were tested individually. Location in

Survey
We conducted a survey of VNRC members from the 14 villages with FSC-certified forests to investigate their perceptions of the costs and benefits of certification. Members of the VNRCs are more familiar with community forest management and certification than the general village population, so were most qualified to respond. We developed the survey in consultation with MCDI staff, pilot tested it in four villages, and trained MCDI staff to administer it during our 2018 field visit. It was administered between July and November 2018 in Kiswahili. Roughly 50% of VNRC members per village were randomly chosen to be surveyed, resulting in a sample size of 132 (87 men, 40 women, 5 missing gender data). Respondents were paid to participate, per MCDI protocol. Survey methods are described in full in Append. 1.
The survey contained both quantitative and qualitative questions that enabled respondents to scale their answers and explain their reasoning behind them. Questions solicited respondents' perceptions of a range of activities required by certification, the benefits and drawbacks of these activities, the benefits and challenges of certification, and whether it should be continued (Append. 1). Quantitative data were entered into Open Data Kit software in the field, downloaded to Excel, and analyzed using STATA. Qualitative data were recorded by hand in the field, entered into Excel, and analyzed by identifying response categories for each question, and quantifying how many respondents mentioned each response category per question. For purposes of this study, we examined the data to gain insight into the role of certification in relation to success of community forestry and CFEs.

Limitations
Limitations of our methodological approach include the following: . We did not compare success between certified and uncertified community forests in the MCDI project area because of challenges controlling for differences between them. Uneven record keeping and financial accounting among uncertified community forests also make it harder to get reliable data for them. In addition, we were concerned about selection bias. The MCDI may choose community forests with the most potential for success to include in the FSC group certificate. These community forests may also receive more support from MCDI, causing them to perform better.
. Only 14 community forests were included in the FSC group certificate. This small sample size limited our ability to perform statistical tests that might reveal significant correlations between success (total revenue earned) and social and biophysical factors contributing to it. However, this smaller sample size allowed us to explore the cases more in depth using a mixed-methods approach.
. We did not interview or survey community members at large about our research topics because of limited resources. Instead, we focused our investigation on individuals having the greatest familiarity with community forestry in the MCDI project area. Therefore, our results do not represent the full range of views held by community members.

Evaluating Success
The criterion for success most commonly cited by interviewees (8 of 9) was high levels of revenue generation from community forests. Their reasoning was that revenues invested in tangible community projects benefit communities, creating an incentive for village residents to support community forests and for VNRCs to enforce bylaws and manage them well. Only one interviewee mentioned financial self-sustainability as a criterion for success.
Five interviewees identified the number of community development projects as another important criterion for judging success. Seven of nine cited good forest management practicesespecially patrolling and bylaw enforcement-as an additional criterion. Five interviewees viewed good governance as an important criterion for evaluating success because it underlies the ability of villages to manage community forests and income from them well. They cited effective leadership by VNRCs and Village Councils, transparent operations and decision-making processes, good financial management, high levels of engagement by village residents in community forest meetings and management, and low levels of internal political conflict within villages as indicators of good governance. Interviewees described unsuccessful community forests as having the opposite characteristics: low revenue generation, few if any community projects, weak governance, illegal logging activity, and poor forest conditions.
It was outside the scope of this study to independently evaluate forest management practices and governance. We rely instead on interviewees' perceptions of how these variables influenced success. Our focus is on revenue generation and community payments as indicators of success.

Revenue generation and community payments
We grouped the 14 villages with certified community forests into four categories based on CFE performance, measured by total earnings between 2011 and 2018 ( Table 4). The most successful community forest in total revenue generation and community payments to Village Councils was Nanjirinji A, earning almost eight times as much as the next highest income earner (Fig. 2). At the other end of the spectrum, four community forests (Mchakama, Nyamwage, Sautimoja, Tawi) earned little if any income. Namatewa, also in the "poor" category, was in the process of establishing a community forest at the time of our study, and no timber sales had occurred there yet. We did not assess how revenues were invested in community projects. Community forests were established in different years, suggesting older forests may have higher total revenue. Table 5 compares total and average annual revenue earned by the 14 CFEs since year of establishment, or 2011 if older. The five highest total revenue earners (in consecutive order) were also the top five in average annual income earned (consecutively), with the exception of Machemba-now third in average annual income-one of our anomalous community forests (see below), and one of the newest. The four lowest total revenue earners also had the four lowest annual average earnings, despite large differences in year of creation. Thus, age of community forest did not appear to be linked to total revenues. Comparing revenue earned per capita and per hectare of productive community forest area provides an alternative view of success (Fig. 3). Although Nanjirinji A had the highest total revenue, it has a large population and large community forest (Table 6). Therefore, in Nanjirinji A, revenue per capita totaled $69 per person between 2011 and 2018. Liwiti, by comparison, has a very small population, so total revenue per capita over this time period was much higher, at $180 per person, followed by Nainokwe ($83 per person) (both classified as good performers in total revenue generation). Revenue per hectare demonstrates success at creating income from community forests independent of size. The two most successful community forests using this measure were Kisangi ($28 per hectare) and Kikole ($26 per hectare) (good and fair performers, respectively).

Interviews
The greatest number of interviewees identified Nanjirinji A (excellent, 9 of 9) and Likawage (good, 7 of 9) as having the most successful community forests of the group, followed by Nainokwe  Table 3). § MCDI headquarters is located in Kilwa Masoko, Tanzania; distances are estimates obtained from MCDI driving records. | This community forest was new at the time of our fieldwork and a forest inventory had not yet been completed, explaining the lack of data for some variables.
(good), which 8 of 9 said had a successful community forest (but not "the most" successful). Criteria they used to judge were high levels of revenue generation from community forests, high levels of investment in community projects as a result, and good governance. Interviewees identified Machemba, Tawi, and Nyamwage as having unsuccessful community forests using the same indicators: low revenue, low community benefits, and low governance capacity. These perceptions were largely consistent with the results of our financial analysis (Table 4), except for Machemba, ranked fair instead of poor. Opinion was mixed about the success of the remaining community forests.

Community forest and village attributes
The correlation test found that volume of mpingo, volume of eight merchantable hardwood species, size of community forest, and productive area of community forest were positively correlated with total revenue and community payments (Table 7). Larger community forests likely contain more merchantable timber. Villages with the largest community forests were not necessarily the largest villages; the percentage of village land allocated to community forests varied widely (from 5% to 85%; Table 6). Nevertheless, the moderate correlation between village size and total revenue and community payments suggests that larger villages have more land to allocate to community forests, favoring larger community forest size and potential success. Variables with no apparent correlation to success were village population density, year community forest was created, community forest size per capita, and distance to MCDI. These findings underscore the importance of having high endowments of forest assets for CFEs to be successful.
Nevertheless, favorable forest assets do not ensure success. Timber must be marketed and purchased by buyers to produce financial benefits. Most species harvest volumes were far below the allowable cut specified in community forest 5-yr management plans (Table 8). With the exception of Bobgunnia madagascariensis (having an extremely low allowable cut volume), the volume discrepancies in 2017-2018 were so large that they were unlikely to be substantially reduced within the lifetime of existing management plans. Thus, total merchantable timber volume, while important for success, did not always produce high earnings, as Table 6 reveals.
Distance to MCDI (in Kilwa District) did not predict success. Although we did not include location in Kilwa District in the correlation test, three of the four villages classified as poor performers (excluding Namatewa, too new to have begun producing timber) were located outside of Kilwa District. Kilwa District was the initial focus of the MCDI project, and villages there have received the greatest investment of resources from MCDI. Additionally, the Kilwa District government has been supportive of community forestry, and according to four interviewees, has occasionally intervened to help enforce bylaws and counteract village-level corruption. These factors likely contributed to success.

Interviews
Interview results were consistent with the above findings and added to them. Factors most frequently cited by interviewees as being associated with successful community forests were large size and high and diverse stocks of merchantable timber species. Associated high timber revenues that Village Councils prioritized investing in community development projects were also frequently cited success factors. These created a strong incentive among villagers to engage in community forestry activities and decision making, and to support their community forest. Another frequently cited success factor was good governance by VNRCs, https://www.ecologyandsociety.org/vol27/iss3/art6/  Burkea Africana 156 10.0 (6.4%) † Namatewa is not included because the community forest was established in 2017 and the forest inventory had not yet been completed at the time of fieldwork. ‡ These numbers represent the total allowable cut from the villages' most recent 5-yr management plans, combined.
Conversely, the reason interviewees most frequently cited for lack of success was small community forests without adequate stocks of merchantable timber that limited income generation and, consequently, motivation for community forestry. There were additional influencing factors, however. For example, the community forests in Tawi and Nyamwage (poor performers) were among the first established (2006) and certified (2010) but were established under the auspices of a World Wildlife Fund project. The MCDI subsequently inherited them, but would not have prioritized them for inclusion in the FSC group certificate, and has provided them with minimal support. Additionally, these are the only two villages located in Rufiji District, the closest district in the project area to Dar es Salaam. Although this geography may be beneficial for market access, it has subjected both community forests to illegal logging and charcoal production. These activities have left them with few mpingo trees, and Tawi with the lowest volume of merchantable timber species in the group of 14. Consequently, little revenue is generated from timber sales, and there is low investment in community projects. The Nyamwage community forest also suffers from illegal livestock grazing by pastoralists searching for dry season forage, threatening its ability to remain in the group certificate. These factors combine to create little incentive for good forest management, and low levels of engagement by village residents. Low governance capacity was another problem cited. Illegal activities are difficult to control, calling for enforcement assistance from the Rufiji District office, which has not been forthcoming.

Anomalies
Despite the foregoing results, two community forests presented anomalies. One was Sautimoja, in the poor performance cluster but with the third largest community forest and productive area of the 14, the third highest merchantable timber volume, and ranking sixth in mpingo volume. The other is Machemba, identified by three interviewees as being among the least successful villages, although it was in the fair performance cluster and has the fourth highest merchantable volume of timber (but lacks mpingo). Both villages are located in Tunduru District. Villages in Tunduru District are 500 km or more from MCDI headquarters in Kilwa Masoko (the most distant of the group), making it challenging for MCDI staff to travel there and provide support. Few buyers or difficulty accessing timber were mentioned three times by interviewees, suggesting market access may be challenging in this district. Two interviewees reported that Tunduru District government officials don't support community forestry, leading to lack of bylaw enforcement and illegal activities, including illegal logging and livestock incursion into the Sautimoja community forest. Two interviewees explained that livestock herders are attracted to Sautimoja owing to the presence of dry season water resources there. But 85% of village lands are part of the community forest, creating a potential shortage of grazing land and contributing to the problem of illegal grazing, which is hard to control. In Machemba, one interviewee identified poor governance and corruption at the local level as problems. These findings underscore the importance of having strong governance institutions at the village and district levels, and support from MCDI, for success.

The Role of Forest Stewardship Council Certification
Quantitative and qualitative survey results provide insight into how FSC certification may contribute to the success of community forestry. Community forests that are part of MCDI's group certificate are classified as "high conservation value forests" (FSC 2020). As such, the certificate contains standards that require VNRCs to carry out specific conservation-oriented management activities: forest patrols, tree harvest supervision, Ecology and Society 27(3): 6 https://www.ecologyandsociety.org/vol27/iss3/art6/

Benefits of conservation-oriented management activities
Forest patrols Protect the forest and forest products, and prevent illegal harvesting, criminal activity, and forest destruction 60% Monitor overall forest conditions, including human activity, wildlife populations, tree growth, and when trees are ready to harvest 37%

Harvest supervision
Ensure that the correct trees are harvested 28% Ensure that the correct amount of timber is harvested 26% Ensure customer follows harvest rules to prevent damage to the forest, ensure a sustainable harvest 23% Ensure community benefits from timber harvests 8% Ensure safety of the harvest, such as planning the direction in which trees are felled 6% Controlled burning Protect the CF from wildfire; prevent fire from entering the forest 62% Reduce wildfire severity and impacts by causing fires to occur less often, spread more slowly, burn less hot, and become easier to control 12% Protect biodiversity, including insects, wildlife, and habitat 11% Stimulate grass growth, beneficial to animals for forage 5% Help seeds germinate and promote growth of small trees 5% Biodiversity monitoring Monitoring forest health 28% Monitoring forest growth 22% Observing and monitoring bird populations 19% Observing overall forest conditions 11% Ensuring that the forest is being protected 11% Monitoring animal populations and biodiversity 6% Benefits of other required management activities Road building and maintenance Make it easier and faster for people and vehicles to travel through the forest 34% Make it easier for forest monitors to patrol, helping to reduce crime 30% They are good for business, i.e., make harvesting timber and transporting logs easier 20% There are environmental benefits, i.e., less erosion, less tree damage 6% Record keeping Having a record of past activities 58% Having information readily available for responding to questions, reporting, providing guidance, and other needs 18% Providing information that can be used by others in the future 14% Training Learning how to manage the CF and conduct activities required for certification 30% Learning new things and new skills 26% Becoming more educated 25% † The list does not represent verbatim responses, but shows similar answers grouped together post-survey by the research team. ‡ Percentages indicate percentage of total responses for each benefit category. Responses with less than 5% frequency are not reported. ecological monitoring, and early burning. The survey asked VNRC members whether they considered these activities worthwhile. On average, they said yes . They cited numerous conservation benefits of these activities, such as forest protection, preventing illegal activity, promoting awareness of forest conditions, ensuring proper harvest practices, and wildfire prevention (Table 9). They also identified conservation benefits of road construction and maintenance, another required activity (Table 9). Thus, VNRC members surveyed perceive some activities required by certification as helping to protect community forests from degradation.
Certification also requires VNRCs to keep records of management activities and participate in training activities. Quantitative survey results found that respondents also thought these activities were worthwhile , citing several ways in which they increase capacity for community forest and CFE management (Table 9). Doing so improves administration by the VNRC, contributing to good governance and building business capacity.
Finally, informal interviews with MCDI staff indicated that FSC certification helps attract funding from donor organizations because meeting certification standards increases the likelihood that community forestry will contribute to forest conservation, social equity, and community development goals. This support comes from organizations that share these goals (e.g., World Wildlife Fund, Fauna and Flora International, the Royal Norwegian Embassy, the U.S. Forest Service). The MCDI is entirely reliant on donor funding for its operations, and communities in turn are reliant on MCDI for technical and financial support. The survey asked VNRC members whether they could, as of 2018, manage their community forests without assistance from MCDI. In response, 81% said no, 17% said yes, and 2% were uncertain. Among those who said no, the main reasons were lack of knowledge, experience, and capacity (38%), and lack of sufficient funds for forest management and training (25%). Most who said yes believed they would not do a good job and would face many challenges. Ecology and Society 27(3): 6 https://www.ecologyandsociety.org/vol27/iss3/art6/

DISCUSSION
Our analysis of success factors in community forestry in southeastern Tanzania began by identifying measures of success as defined by MCDI staff and local partners interviewed for this study. Interviewees identified high levels of total revenue generation from community forests, leading to investments in tangible community projects that village residents benefit from, as the most important measure of success. Success at total revenue generation does not imply financial sustainability, however. Villages use 40% of their earnings to pay forest management costs, but actual costs are much higher. The MCDI paid 83% of total forest management costs for the 14 villages between 2015-2016). Thus, financial success can be evaluated using different measures, with different results. These measures include total earnings since a community forest was established, financial self-sustainability, average profits or portion of years with an operating profit, activities undertaken at lower cost, amount of revenue earned per village resident, or revenue generated per hectare of community forest. The best way to evaluate financial success depends on project goals and how financial benefits are distributed.
In the MCDI case, typically 50% of community forest revenues are transferred to Village Councils for community investment. Most socioeconomic benefits of community forestry are experienced at the village-not household-level (Gross-Camp 2017). Total earnings are therefore an appropriate financial indicator of success for this study: the more total revenue, the more community-scale development investment. Because a fixed proportion of revenues, rather than profits, are transferred to Village Councils and management costs are subsidized by external donors via MCDI, there is little reason for communities to perceive cost as a barrier to success. If external subsidies were reduced in the future, there would be pressure to change the financial model and prioritize profits (i.e., low cost and high revenue) over high revenue alone.
We found that size of community forest, size of its productive area, volume of mpingo, and volume of eight merchantable hardwoods within community forests were the local-level community forest or village characteristics most highly correlated with financial success of those we measured. Although this finding is unsurprising, it emphasizes the importance of locating community forests with CFEs in locations that have a robust endowment of forest resources upon which the CFE is based, and where it is possible to delineate a large community forest area. The most successful community forest in our study was 83,538 ha, and those that were in the good performance category averaged 13,098 ha. By comparison, law in Cameroon stipulates that community forests with timber-producing CFEs may not exceed 5,000 ha, which can be a barrier to both production and certification (Alemagi et al. 2011). Moreover, community forestry is often implemented in forests having low commercial value, leaving communities to manage "leftovers" (Anderson et al. 2015). Poor forest stocks that take a long time or high investment to restore and regain value likely limit the success of CFEs and create low incentive to engage in community forestry (Anderson et al. 2015). Good governance at the village level, support from district-level government, tangible benefits from community forests, and MCDI technical and financial support (ongoing since 2004) were also important success factors in the MCDI case.
Limiting financial success of the CFEs were unfavorable market conditions and limited access to markets, lack of high capacity business organizations at the village level, and limited ability to diversify forest products beyond mpingo, despite availability of numerous merchantable hardwood species . Competition with timber harvested from government forest reserves and illegally harvested timber also depressed prices for FSC-certified wood from the project area, which was sold at government floor prices and did not bring a price premium . Our findings are consistent with the broader literature regarding conditions that limit the success of CFEs. These include poor market development and market access, low business and management capacity, an unfavorable policy environment (Hajjar andOldekop 2018, Segura Warnholtz et al. 2020), and illegal timber harvesting that degrades forest resources and drives down prices for legally harvested wood (Blomley and Ramadhani 2006, de Jong et al. 2016, Sanchez Badini et al. 2018).
This study focused on local (community and district) level factors influencing success by comparing 14 community forests that participate in one initiative located in southeastern Tanzania. But multi-scalar factors influence success in community forestry. As discussed earlier, important national-scale factors that make community forestry in Tanzania successful relative to other African countries are national laws and policies that support secure forest tenure, effectively devolve forest management to the community level, and enable villages to retain financial benefits from community forests. These features operated in all 14 villages studied and were important in contributing to the potential for success.
Much of what we know about success factors in community forestry and CFEs comes from case studies conducted in Asia and Latin America. Success factors for community forestry identified in this literature were applicable in southeastern Tanzania (Table 10). Our study also considered the role of certification in contributing to success. We found that certification reinforces many other success factors for community forestry and small-scale forest-based enterprises (Table 10).
On the ecological side, certification standards are designed to promote sustainable forest management, practices to maintain biodiversity, timber harvest operations that take place in an environmentally appropriate manner, monitoring of forest conditions and enforcement of forest management bylaws, and annual third-party audits to ensure certification standards are met. Interviewed VNRC members identified numerous forest conservation benefits associated with the management activities they perform, as required by certification. Certification thus helps ensure that community forests maintain endowments of diverse and quality forest products that can be harvested and sold. Kalonga et al. (2016) found that FSC-certified forests in the MCDI project area performed better at conserving biodiversity than nearby government forest reserves and village forestlands located outside of community forests. They attribute this finding to good governance, including the presence of forest guards and enforcement of bylaws; effective forest management plans that reduce harvest intensity by stipulating selective harvest practices, allowable harvest quotas, and minimum diameters for tree harvest; and regulation of forest uses.

Ecological
High endowment of forest products to harvest and sell X X X Ability to produce diverse forest products X X Socioeconomic Secure tenure over land and forest resources X X X X Supportive government policies and programs X X X X Meaningful devolution of forest management responsibility and authority from state to communities X X Effective community-level forest governance institutions X X X X Education and training to develop capacity for forest management X X X Tangible community benefits equitably distributed among community members X X X A "community of practice" working together to achieve common goals X X X Positive business environment X X Favorable market conditions and access to markets X X X High-capacity community organizations for managing the enterprise X X X Access to development assistance or affordable financial capital X X X Compatibility with the sociocultural characteristics of the community X X X Clustering of producers, organizations, networks, or associations X X X Socioeconomically, VNRC members surveyed indicated that certification requirements such as training and administrative tasks contribute to capacity building for effective forest management, which likely contributes to effective forest governance institutions. The MCDI FSC group certificate can potentially facilitate clustering and networking among producers and help create a community of practice to achieve common goals.  found that certification in the MCDI project area also supports some sociocultural characteristics of communities, access to development assistance, secure tenure rights over forestland, improved market access, and increased recognition and support from the Tanzanian government. Other research from the MCDI project area indicates that villages with certified community forests experienced less corruption and more equitable distribution of income from community forests (Kalonga et al. 2015, Gross-Camp et al. 2019. Elite capture and inequitable distribution of benefits were not topics of concern in most MCDI villages where they worked (Khatun et al. 2015, Gross-Camp 2017. Research has also found that in MCDI villages with FSC-certified community forests, average annual household income from forests is higher, community benefits in the form of development projects are greater, and forest governance is better than in villages without community forests (Kalonga and Kulindwa 2017). These researchers did not compare villages with certified and uncertified community forests, however.
Finally, certification may contribute to "investment readiness" among CFEs, attracting financial investment by private or government entities that might in turn increase their profitability (Gnych et al. 2020). Certification standards help mitigate the risks of over-harvesting and forest degradation, and poor forest and business management, decreasing business risk and increasing assurances that certified operations are a good financial investment. Because MCDI and the FSC (via annual audits) oversee community forests to ensure they are meeting certification standards, investing organizations don't have to, reducing their transaction costs. Consequently, certification can help overcome barriers to financial investment faced by CFEs (Gnych et al. 2020). Although external investment can increase access to capital, other resources, and market connections for CFEs, it may threaten the independence of CFEs and lead to inequitable benefit sharing (Gilmour 2016).
External investment is now happening in the MCDI case. In 2017, Yamaha Corporation-in the musical instrument businessbegan taking measures to develop private business operations in southeastern Tanzania in cooperation with MCDI and the Japan International Cooperation Association (Yamaha nd). Forest Stewardship Council certification attracted Yamaha to the project area because it helps ensure a sustainable supply of mpingo, which Yamaha uses to manufacture some woodwind instruments (Japan International Cooperation Agency (JICA) and Yamaha 2019). In 2018, Sound and Fair Limited completed construction of a sawmill in Nanjirinji A. Certification drew investment because Sound and Fair's mission is to promote the fair and ethical trade of timber from community forests in Tanzania, which FSC certification promotes.
We conclude that, despite high transaction costs , FSC certification holds potential to promote the success of community forestry with CFEs that produce timber when other success factors are present by strengthening many of these factors. Whether certification in and of itself is a success factor is unclear. We did not compare certified and uncertified community forests in the MCDI project area. Nor is there sufficient literature from other countries to support such a finding. The literature that does exist suggests that the benefits of certification for community forestry and CFEs are either uncertain or mixed, may be outweighed by its costs, and depend on other factors influencing successful operation . Additionally, certification often requires long-term technical and financial support from external organizations to be successful (Markopoulos 2003, Hajjar 2013).
To our knowledge, the model of community forestry that includes CFEs based on commercial timber production has yet to develop in Africa other than in Tanzania and Cameroon. Although there have been benefits, overall the Cameroon experience has largely been considered a failure (Movuh and Schusser 2012, Oyono et al. 2012, Lescuyer et al. 2016). The MCDI case offers insights for community forestry elsewhere in Tanzania, in Cameroon, and in other African countries seeking to pursue this model. These include: . establishing large community forests where possible; . ensuring community forests contain a substantial quantity of merchantable timber; . enacting policies that promote secure tenure over community forests, exclusive rights to manage them, and retention of benefits; . providing early, tangible, and locally supported social and economic benefits to community members; . engaging partner organizations over the long term to provide financial and technical assistance; . providing support from government entities to backstop village institutions in enforcing rules; and . adopting certification, where possible and appropriate, to reinforce success factors.

CONCLUSIONS
This article investigates success factors for community forestry with timber-producing CFEs in southeastern Tanzania through a comparison of 14 FSC-certified community forests. It finds that to optimize success, this model of community forestry must exhibit features associated with both successful community forestry and successful small-scale forest-based enterprises. If only some success factors are operative, outcomes are likely to be compromised. Success factors for community forestry and smallscale forest-based enterprises identified in the literature-largely from Asia and Latin America-also apply to the Tanzanian case. Forest Stewardship Council certification reinforces several of these success factors and contributes to investment readiness, helping draw financial investors and improve community forestry outcomes.
Community forestry and CFEs in southeastern Tanzania continue to develop. Despite challenges, the 14 community forests we studied have been relatively successful in achieving desired outcomes where there is broad community support. This community support is heavily dependent on current and future prospects for generating revenue through timber-based CFEs to fund community development projects that provide tangible benefits to community members. Revenue generation in turn relies on having a high natural endowment of timber, good governance, effective forest management institutions, secure forest tenure, a strong support organization, a favorable policy environment, and markets for products.
This article contributes to the literature on success factors in community forestry and CFEs in six ways. First, it uses a locally defined measure of success for evaluating it and points to the importance of critically assessing the use of financial measures of success. Second, it takes a comprehensive look at success factors rather than focusing on one type (e.g., governance institutions, business operations), as is common in the literature (Baynes et al. 2015). Third, it examines both biophysical and social factors influencing success (much of the literature focuses only on social factors; Baynes et al. 2015, Hajjar et al. 2016

A. Financial analysis methods
To conduct the financial analysis, we sought comprehensive income and expense data related to the 14 FSC-certified community forests for the five most recent fiscal years (FY 2013(FY -14 to 2017. FY 2017-18 was an appropriate cut-off as it was the last year before MCDI began fullscale implementation of a portable sawmill and kiln for adding value to standing timber. Up through 2017-18, all of the community forest enterprises primarily sold standing timber to third parties. Data were available from before 2013-14 for some, but not all, community forests. Although five years is a relatively short period of time for trend analysis, it is sufficient for understanding general status. This study reports revenue data but not expense data; the following methods pertain to revenue calculations. We examined financial data regarding revenues received by Village Natural Resource Committees (VNRCs) from timber harvests and other sources (income), and transfer payments from the VNRC to the Village Council for the benefit of the community (community transfer payments). All VNRCs record their expenses and revenues in ledgers maintained on site in village offices, in the Kiswahili language, using Tanzanian Shillings as the monetary unit. For auditing purposes, MCDI annually compiles these expense and revenue data from each community forest by activity type and translates them to English. We used these annual MCDI compilations for the financial analysis. We obtained the compilations during site visits to the MCDI office in Kilwa Masoko during October/November 2017 and July 2018. A few years had missing information or questionable data; in these cases MCDI staff returned to the villages to compare the compilations to the original ledgers for verification and to fill in data gaps. We obtained additional data to fill in gaps and clarified data quality issues during late 2018 and 2019 via email and skype calls with MCDI staff. Three of the community forests were not in the FSC group certificate for the entire five-year period, and were excluded from the analysis in those years: Machemba and Sautimoja (2013-14), and Namatewa (2013-14 through 2015-16). Other community forests had no accounted VNRC costs or revenues for entire years that they were in the FSC group certificate. MCDI staff indicated that the VNRCs most likely had no revenues in these years. For years when revenue was generated, the financial compilations recorded the sources, which were divided into two main categories: timber sales and other. Timber sales accounted for 96% of total revenues. These were reported by sales volume in cubic meters, price per cubic meter, and total sales price. Other revenue sources included fines, fees, and sales of other products (such as non-timber forest products or auction of tools or timber confiscated from illegal loggers).
Financial compilations also included the community transfer payments from the community forest account to the village account for community development projects. Generally, these were a fixed percentage (often 50%, but variable by village) of the revenue received. Occasionally, the VNRCs would make ad hoc additional payments for community projects, which were recorded in the financial compilations and added to community payments for our analysis.

C. Survey methods and instrument
The process of becoming and staying certified by the FSC involves numerous activities, and creates potential costs and benefits, some of which are monetary in nature and some of which are not. We conducted a survey of Village Natural Resource Committee (VNRC) members to obtain information about their perceptions of certification and its advantages and disadvantages. The survey targeted VNRC members rather than village residents as a whole because these individuals are most familiar with certification, being responsible for community forest management. Consultations with MCDI staff led us to believe that most village residents would have only superficial, if any, understanding of certification and its implications.
The survey had both quantitative and qualitative components and was developed jointly with MCDI. To meet certification standards, VNRCs must conduct specified forest management activities designed to encourage and monitor sustainable forest management and promote safe working conditions (Table A1.1). For each of these activities, we asked respondents whether they thought they were worthwhile or not by providing one of four responses: 1) cost and difficulty of the activity outweighs the benefits of the activity; 2) is equal to the benefits; 3) benefits are greater than the costs and difficulty; or 4) don't know. We tested a 5-point Likert scale for responses during our pilot survey work, but found it was difficult for people to make finer distinctions in their responses, rendering them not meaningful. Conduct forest patrols to monitor activities occurring in community forest and ensure no prohibited activities are taking place Supervise timber harvest Supervise tree harvesting to ensure that harvests comply with specifications laid out in timber sale contracts First aid provider Ensure first aid provider is on site during tree harvesting in case of injury Early burning Perform burning of understory vegetation in community forest early in the dry season to prevent high-intensity wildfire later in the season Bird monitoring Monitor specific bird species populations and distribution as an indicator of biodiversity Road repair and maintenance Repair and maintain community forest road system to facilitate access and transportation of wood products. Keep roads accessible to prevent diversions around obstacles Record keeping Maintain administrative records associated with community forest activities and finances, up to the standard prescribed by government rules and the certification body Training Participate in trainings on forest administration, planning, management, and certification 5 A second series of statements asked respondents about their perceptions of potential benefits and challenges of FSC certification. The potential benefits and challenges were identified in advance through review of the literature , Molnar 2004, Quaedvlieg et al. 2014, Romero et al. 2017, Wiersum et al. 2013) and discussion with MCDI staff and others (Table A1.2). Potential positives and negatives were asked in the order shown in Table A1.2, which mixed positives and negatives. We asked survey respondents to rate the frequency of the various potential benefits and challenges using a 4-point scale (never, rarely, mostly, always, or unsure/don't know). Because of certification, we are better recognized by the government (district, regional, or national). Complexity It is difficult to understand the requirements of certification.
The qualitative survey questions aimed to get respondents to explain what they viewed as the costs and benefits of the forest management activities required to meet certification standards; to probe their understanding of FSC certification and how certified VLFRs differ from those that are not certified; and whether they saw value in continuing to have a FSC-certified VLFR, and why. We also asked for their suggestions about how VLFRs could become more financially selfsustaining, and how VNRCs could obtain more money to cover the costs of forest management and certification, of which MCDI currently pays the majority.
The survey was conducted in all 14 villages with VLFRs that are part of the FSC group certificate. Our goal was to survey 50% of VNRC members in each village. We obtained lists of VNRC members' names from MCDI or on site in the villages, numbered them, and then randomly selected names from each village list using a random numbers table. If someone was unavailable while researchers were in the field or did not give their informed consent to be surveyed, we randomly selected alternates. Each person who participated was compensated for 6 their time (5,000 T Sh), per MCDI protocol. Altogether, we surveyed 132 VNRC members from the 14 villages; 87 men and 40 women (5 missing gender data). The survey was administered by two MCDI staff members in Kiswahili. The Forest Service researchers worked with MCDI to develop the survey, pilot test it in four villages (two in November 2017, two in July 2018), and train MCDI staff on implementation in July 2018. A team of two MCDI staff members administered the survey between July and November 2018. While the survey was being administered, one team member entered responses to the quantitative questions into Open Data Kit software that was loaded onto their cell phone, while the other wrote responses to the qualitative questions in a notebook.
The quantitative data were downloaded to Excel and analyzed for statistically-significant differences in mean response levels between variables using STATA's Wilcoxon signed-rank test. We also used STATA's ordered logistic regression to test for influences of various respondent-and village-level factors on the responses. We tested for potential effects of more successful or less successful villages (more successful = Nanjirinji A, Likawage, Liwiti, and Nainokwe; less successful = Tawi, Sautimoja, Nyamwage, Mchakama), between male and female VNRC members, between VNRC officers and non-officers, and between newer (less than 5 years) and longer-term (5 or more years) VNRC members.
The qualitative data were translated into English by MCDI staff who administered the survey, and entered into Excel for analysis. We reviewed the responses to each question, identified response categories/key points, and quantified how many respondents mentioned each point. This was possible because many people gave similar responses to the questions. We then synthesized and interpreted the results.