Open Collections

UBC Undergraduate Research

Effects of community-based natural resource management on household welfare in Namibia Riehl, Brianne 2014-03

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
52966-Riehl_Brianne_ENVR_449_2014.pdf [ 2.85MB ]
Metadata
JSON: 52966-1.0074568.json
JSON-LD: 52966-1.0074568-ld.json
RDF/XML (Pretty): 52966-1.0074568-rdf.xml
RDF/JSON: 52966-1.0074568-rdf.json
Turtle: 52966-1.0074568-turtle.txt
N-Triples: 52966-1.0074568-rdf-ntriples.txt
Original Record: 52966-1.0074568-source.json
Full Text
52966-1.0074568-fulltext.txt
Citation
52966-1.0074568.ris

Full Text

!! EFFECTS OF COMMUNITY-BASED NATURAL RESOURCE MANAGEMENT ON HOUSEHOLD WELFARE IN NAMIBIA   by  BRIANNE RIEHL     A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE  REQUIREMENTS FOR THE DEGREE OF   BACHELOR OF SCIENCE (HONOURS)  in   THE FACULTY OF SCIENCE  (Environmental Sciences)      This thesis conforms to the required standard  ……………………………………… Supervisor    THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)  MARCH 2014    © Brianne Riehl, 2014!!!! ! !ii!ABSTRACT   Biodiversity conservation, as an environmental goal, is increasingly recognized to be connected to the socioeconomic well-being of local communities. The development of a widespread community-based natural resource management (CBNRM) program in Namibia makes it an ideal location to analyze the connection between conservation and socioeconomic well-being of local communities. Namibia’s CBNRM program involves the formation of communal conservancies within rural communities and is typically considered a success on both ecological and economic fronts. In order to broaden the understanding of the program’s impact to include social factors, we have conducted a comparative analysis to determine the effects of this program on household welfare outcomes. Data from two rounds of the Namibia Demographic and Health Surveys (2000 and 2006/07) and quasi-experimental statistical methods were used to evaluate changes in various health, education and wealth outcomes of those living in conservancies, relative to non-conservancy comparison groups. Regression results indicate mixed effects of the conservancy program at the household level. The program had positive effects on some health outcome variables, including bednet ownership, which was twice as likely to increase over time in conservancy compared to non-conservancy households. Program impacts were negative for education outcomes, with the proportion of school attendance of conservancy children being 45% less likely to increase over time than non-conservancy children. Wealth outcome results were inconclusive. Our findings highlight the importance of analyzing community conservation programs at the household level when evaluating overall impact, as community benefits do not necessarily extend down to the household level. We recommend further analysis of this program, as an extended time period, more outcome variables and larger sample sizes would provide a valuable addition to this analysis. !! ! !iii!TABLE OF CONTENTS   TITLE PAGE………………………………………………………………………….i ABSTRACT…………………………………………………………………………..ii TABLE OF CONTENTS…………………………………………………………….iii LIST OF FIGURES…………………………………………………………………...v LIST OF TABLES……………………………………………………………………vi LIST OF APPENDICES…………………………………………………………….vii ACKNOWLEDGMENTS…………………………………………………………..viii 1. INTRODUCTION………………………………………………………………….1 2. CONSERVATION AREAS AND LOCAL SOCIOECONOMIC IMPACTS…....2  2.1 Conservation and Development Links………………………………….....2  2.2 Biodiversity Conservation in Namibia……………………………….…....4   2.2.1 Characteristics of Study Site.........................................................4   2.2.2 CBNRM Program……………………………………………….6   2.2.3 Previous Socioeconomic Analyses of the CBNRM Program…..8 3. DATA AND METHODS………………………....……………………………....11  3.1 Demographic and Health Surveys Data……………………………….....11  3.2 Outcome Variables of Interest………………………………………..….12  3.3 Overview of Methods………………………………………………..…..14   3.3.1 Matching of Treatment and Comparison Groups…………...…14   3.3.2 Regression Analysis…………………………………...……….17 4. RESULTS…………………………………………………………………………19  4.1 Malaria Prevention………………………………………………….........19   4.1.1 Bednet Ownership…………………………………...…………19   4.1.2 Bednet Usage……………………………………...……….......21  4.2 Diarrhea Prevalence and Treatment………………………..…………….22   4.2.1 Diarrhea Prevalence……………………………...…………….22   4.2.2 Diarrhea Treatment……………………………...……………..23  4.3 School Attendance…………………………………………………….....24 !! ! !iv! 4.4 Household Wealth………………………………………………………..25 5. DISCUSSION…………………………………………………………………......29  5.1 Health Outcomes………………………………………………………....29  5.2 Education Outcomes…………………………………………………..…31  5.3 Wealth Outcomes……………………………………………………..….31  5.4 Limitations and Assumptions………………………………………..…..32 6. CONCLUSION……………………………………………………………………32 REFERENCES CITED………………………………………………………………34 Appendix 1…………………………………………………………………………...38 Appendix 2…………………………………………………………………………...39             !! ! !v!LIST OF FIGURES   Figure 1a. Location of Namibia in Africa…………………………………………………..…5 Figure 1b. Population of Namibia in 2000…………………………………………………….5 Figure 2. Communal conservancies and protected areas in relation to Namibia’s six major biomes………………………………………………………………………………....………6 Figure 3. Income benefits from the CBNRM program………………………………………..8 Figure 4. The range expansion of lion populations in northwest Namibia…………………....9 Figure 5. Map of DHS cluster locations for 2000 and 2006/07 surveys……………………..12 Figure 6. Flowchart summary of methods…………………………………………………...16 Figure 7. Trends from 2000 to 2006/07 for conservancy households versus 3 comparison groups for the proportion of households that own a bednet………………...……………..…19 Figure 8. Trends from 2000 to 2006/07 for conservancy residents versus 3 comparison groups for the proportion of respondents that slept under a bednet………………………….21 Figure 9. Trends from 2000 to 2006/07 for conservancy children versus 3 comparison groups for the proportion of children under 5 that had diarrhea……………………………………..23 Figure 10. Trends from 2000 to 2006/07 for conservancy children versus 3 comparison groups for the proportion of children under 5 that received medical treatment for diarrhea...25 Figure 11. Trends from 2000 to 2006/07 for conservancy children versus 3 comparison groups for the proportion of school-aged children that attended school…………..………...27 Figure 12. Trends from 2000 to 2006/07 for conservancy households versus 3 comparison groups for standardized household wealth factor scores…………………………………….28 !! ! !vi!LIST OF TABLES   Table 1. Summary of Namibian statistics……………………………………………………..5 Table 2. Logistic regression model results for household bednet ownership………………..20 Table 3. Logistic regression model results for bednet usage by the respondent during the previous night…………………………………………………………….…………………..22 Table 4. Logistic regression model results for diarrhea prevalence in children under 5 within the past 2 weeks……………………………………………………………………………...24 Table 5. Logistic regression model results for diarrhea treatment in children under 5 within the past 2 weeks……………………………………………………………………………...26 Table 6. Logistic regression model results for school attendance of children ages 6-16……27 Table 7. Linear regression model results for household wealth index factor scores………...28  !! ! !vii!LIST OF APPENDICES   Appendix 1. Summary statistics for outcome variables……………………………………...38 Appendix 2. Summary statistics and t-test results used to determine the best matching model for each outcome variable……………………………………………………………………39 !! ! !viii!ACKNOWLEDGEMENTS    I would like to thank my thesis supervisors, Dr. Hisham Zerriffi and Dr. Robin Naidoo, for the opportunity to work on this project under their guidance. Their advice, support and expertise have been invaluable. I am sincerely grateful for their patience in answering my questions and appreciate the time and effort they put into editing my thesis. I have learned so much from them. I would also like to thank Dr. Mary Lou Bevier for organizing this thesis course and for providing edits, support and encouragement. Thank you to the Department of Earth, Ocean and Atmospheric Science for their financial support.!!1!1. INTRODUCTION   Biodiversity conservation has become an important area of environmental policy, increasingly recognized to be connected to the socioeconomic well-being of local communities (Adams et al., 2004; Agrawal & Redford, 2006; Pelser et al., 2013). The social impact of conservation programs on communities is an intensely debated topic, fostering a wide spectrum of views regarding whether or not human development and conservation can be achieved simultaneously (Adams et al., 2004; Agrawal & Redford, 2006; Pelser et al., 2013). Those most vulnerable to the deterioration of natural systems are typically rural populations of the developing world, who commonly lack a minimum standard of living (Millennium Ecosystem Assessment, 2005). The strong dependence of these populations on natural resources for their livelihoods leads to a complex relationship between conservation and human development (Pelser et al., 2013). Four common perspectives on this relationship include (Adams et al., 2004):  (1) Socioeconomic development and conservation are separate policy realms (2) Socioeconomic development is a critical constraint on conservation (3) Conservation efforts should not compromise socioeconomic development (4) Socioeconomic development depends upon conservation   Despite sound reasoning and evidence to support each of these perspectives, research is increasingly citing evidence to support the final statement that the future of biodiversity conservation and the socioeconomic needs of rural communities are intricately connected (Agrawal & Redford, 2006; Millennium Ecosystem Assessment, 2005; Pelser et al., 2013). The Millennium Ecosystem Assessment (2005) provides a compelling argument to explain the dependence of human well-being on the services provided by nature, suggesting that threats to natural assets must be addressed as part of an effective strategy for human development. The strong body of opinion maintaining that socioeconomic development and conservation can coexist provides increasing evidence of successful conservation and !! ! !2!development projects, with many international conservation organizations and development agencies strongly emphasizing the links between rural poverty and environmental degradation (Adams et al., 2004; Agrawal & Redford, 2006; Frost & Bond, 2008; Pelser et al., 2013; Timmer & Juma, 2005). These ideas are still challenged, however, due to a lack of thorough understanding of this link (Adams et al., 2004; Agrawal & Redford, 2006; Pelser et al., 2013). The purpose of this thesis is to investigate the argument that conservation and local development can be jointly achieved. The aim is to determine, through analysis of an existing conservation effort in Namibia, whether conservation programs can successfully contribute to improved social well-being in local communities. Our analysis focuses on households and individuals within communities to determine whether the benefits of conservation programs aimed at the community-level extend down to the household level.  We begin by introducing global examples of research focusing on conservation and development, and then explain the current state of a community-based natural resource management (CBNRM) program in Namibia, including findings from previous analyses of the program. We then determine appropriate comparison groups between conservancy and non-conservancy households, which allow us to run regression models to compare temporal trends between groups on a variety of socioeconomic outcomes. Finally, we offer possible explanations for our findings and make recommendations for future research in this area.    2. CONSERVATION AREAS AND LOCAL SOCIOECONOMIC IMPACTS   2.1 Conservation and Development Links  Arguments in support of biodiversity conservation increasingly cite its positive contribution to the well-being of local communities (Johnson et al., 2013; Palmer & Di Falco, 2012). These arguments often refer to the fundamental dependence of humans on services derived from natural ecosystems, expressing an increasing concern for the potential health and welfare impacts of continued ecosystem degradation (Johnson et al., 2013; Palmer & Di Falco, 2012). Ecosystems provide provisioning, regulating, cultural and supportive services, which are particularly important for rural communities who often rely directly on !! ! !3!these services for their livelihood (Johnson et al., 2013; Millennium Ecosystem Assessment, 2005; Roe et al., 2009). Links between biodiversity conservation and human well-being include food security, health improvements, income generation, reduced vulnerability to climate and resource changes, ecosystem services, and cultural value (Timmer & Juma, 2005). Emphasis in recent literature on these linkages suggests a momentum towards approaching environmental conservation and human development in an integrated way (Agrawal & Redford, 2006; Palmer & Di Falco, 2012; Roe et al., 2009; Timmer & Juma, 2005). Previous analyses of community-based resource management initiatives have determined that, because of their knowledge and direct dependence on the land being protected, local communities can often undertake conservation more effectively and cost efficiently than a centralized government agency (Agrawal & Redford, 2006; Millennium Ecosystem Assessment, 2005). A key principle underlying these community-based initiatives is to align long-term conservation with the short-term needs of local people, ensuring that community members gain some benefit for their participation in conservation efforts (Adams et al., 2004; Agrawal & Redford, 2006). It is important to analyze these initiatives in order to further understand the relationship between conservation and human well-being (Adams et al., 2004; Agrawal & Redford, 2006; Pelser et al., 2013). Following are a few examples of community-based conservation initiatives, which consider both monetary and non-monetary benefits of conservation at national and community levels. Examples of national-level monetary benefits of conservation include a study of wildlife resources in Botswana, which determined that wildlife conservation can contribute significantly to national income and economic development of the country, particularly through non-consumptive tourism on high-quality wildlife land (Barnes, 2001). Community-level economic benefits of conservation have also been reported from the Communal Areas Management Program for Indigenous Resources (CAMPFIRE) in Zimbabwe. This program is a long-term approach to rural development that allows communities to manage, use and benefit from wildlife and other natural resources. Between 1989 and 2006, CAMPFIRE income totaled nearly US$ 30 million, 52% of which was allocated to districts and villages to provide community and household benefits (Frost & Bond, 2008). !! ! !4!In terms of non-monetary benefits of conservation, studies by Kumar and Hotchkiss (1988) and Cooke (1998) analyzed the consequences of deforestation in Western Nepal and found that the loss of forest ecosystems requires women to allocate more time towards fuel collection. This provides evidence that deforestation causes a shift in time allocation away from agriculture, reducing household incomes and adversely affecting food consumption and eventually the nutritional status of local communities. Another non-monetary benefit of conservation is improvement in the health and nutrition of local communities. A recent case study conducted by Johnson et al. (2013) demonstrates a relationship between forest cover and the health and nutritional outcomes of local people. This study found that increased forest cover is associated with increased dietary diversity and consumption of vitamin A rich foods, as well as reduced risk of diarrheal disease. These findings suggest that the protection of natural ecosystems and their services are important components in improving the health and nutrition of local communities.   2.2 Biodiversity Conservation in Namibia  To further test the relationship between conservation and human well-being, demographic survey data in Namibia was analyzed in order to understand the impacts of the country’s CBNRM program. As discussed further below, the combination of an extensive conservation program geared towards community benefits, the demographics, economics and natural resource dependence of the country, as well as the availability of household level survey data, make it an ideal case study.  2.2.1 Characteristics of Study Site  Namibia is a country located in southern Africa, with a population of 2.259 million people (Figure 1) (World Bank Group, 2013). The majority of this population (62%) lives in rural areas and depends on natural resources for their livelihood (NACSO, 2013; UNICEF, 2011). Although ranked as a middle-income country, the distribution of income in Namibia is highly skewed with a 51% unemployment rate and a 38.2% incidence of poverty in rural areas (Table 1) (NACSO, 2013; World Bank Group, 2013). Most rural Namibians generate income through farming (livestock and crop production in north-central and eastern areas, mainly livestock production in the arid north-western and southern areas) (NACSO, 2013). !! ! !5!The Namibian climate ranges from arid and semiarid in the west, including temperate coastal desert, to more subtropical in the northeast (Barnard, 1998; Thuiller et al., 2006). This broad range of ecosystem types can be summarized into six major biomes (Figure 2) based on similar plant life and climatic characteristics (NACSO, 2013). These ecosystems are home to remarkable biodiversity, including more than 4,500 plant taxa, almost 700 of which are endemic to the country, as well as 217 species of mammals, 26 of which are endemic (Thuiller et al., 2006). This incredible special richness and endemism makes Namibia a critical location for conservation programs that can offer protection of this biodiversity while promoting the social and economic well-being of local communities.  Table 1. Summary of Namibian statistics (World Bank Group, 2013; UNICEF, 2011). GDP Gini coefficient HDI Adult literacy rate 13.07 billion (2012) 0.5971 (2010) 0.608 (2011) 76.5% (2011)     A  B Figure 1. Location of Namibia in Africa (A) and population density of Namibia in 2000, smoothed to 10km resolution (B) (Mendelsohn et al., 2002).   !! ! !6! Figure 2. Communal conservancies and protected areas in relation to Namibia’s six major biomes (NACSO, 2013).   2.2.2 CBNRM Program  Namibia experiences high hunting and poaching pressures which, in combination with disenfranchisement, civil war, and drought, led to a decrease in populations of many large wildlife species and large-scale animal migrations in the 1970s and 1980s (NACSO, 2013). In response to these threats to biodiversity, the Nature Conservation Act was passed in 1996, allowing for the formation of communal conservancies, i.e. areas of customary land tenure in which rights to benefits derived from natural resources are devolved to local communities. Various community-based conservation activities had begun in 1991, but the passing of this legislation was the official beginning of Namibia’s CBNRM program (NACSO, 2013; Naidoo et al., 2011a; Roe et al., 2009).  By the end of 2013, the CBNRM program included 79 registered conservancies covering 19.4% of Namibia’s land surface and bringing the total land surface under conservation management to 43% (NACSO, 2013). These conservancies vary greatly in size, !! ! !7!environment, human population density, wildlife resources and tourism potential, providing significant variation in income gains and challenges for conservancy management (Naidoo et al., 2011a,b,c; Silva & Mosimane, 2012; Suich, 2010). A main focus of conservancies is wildlife management, as healthy populations of indigenous wildlife are central to unlocking the value of natural resources in the region. The sustainable use of wildlife through tourism, trophy hunting and own-use consumption is particularly valuable in communal areas where agricultural land uses are limited by low, erratic rainfall and infertile soils (NACSO, 2013). It is recognized that these wildlife-based benefits are heavily reliant on foreign visitors to the country, and ecosystem services provided by conservancies are increasingly diversifying through the sustainable harvest of indigenous plant products, fishing, and craft sales (NACSO, 2013; Naidoo et al., 2011a).  The CBNRM program operates with three main goals: natural resource management and conservation, rural development, and empowerment and capacity building (NACSO, 2013). The program is generally considered a success, gaining national and international recognition for making an important contribution to both environmental and socioeconomic development goals (NACSO, 2013; Roe et al., 2009). From 1991-2011 the program contributed more than N$ 2.4 billion (US$ 240 million in 2011) in total economic value to Namibia’s net national income, with CBNRM activities generating almost N$ 50 million (US$ 5 million) in 2011 (Figure 3) (NACSO, 2013). The program has generated 1,512 full time and 11,223 part time jobs between 1991 and 2011, as well as contributed to dramatic increases in wildlife numbers and range expansion, one example being lions (Panthera leo) in the northwest (Figure 4) (NACSO, 2013). Another noteworthy impact of this program has been a major attitudinal shift towards natural resources, as wildlife previously perceived as a detriment to livelihoods are increasingly seen as an asset and regarded with great pride by conservancy members (NACSO, 2013). Despite these examples of socioeconomic benefits, there remains a level of discontent with CBNRM as a development strategy. One of the main sources of discontent is the issue of equity in the distribution of conservancy benefits, as well as arguments that the indirect benefits (such as improved infrastructure, communal soup kitchens, waterpoints, schools and clinics) expected to promote development for all conservancy residents have not yet materialized (Bandyopadhyay et al., 2011; Silva & Mosimane, 2012). !! ! !8! Figure 3. Income benefits from the CBNRM program. Incomes are shown in two categories: benefits to conservancies and benefits from CBNRM activities outside of conservancies (NACSO, 2013; taken from http://www.nacso.org.na).   2.2.3 Previous Socioeconomic Analyses of the CBNRM Program Although often cited as a CBNRM success, previous analyses and evaluations of Namibia’s CBNRM program focus only on its benefits, often using cross-sectional and highly aggregated data with limited geographic scope (Bandyopadhyay et al., 2011; Suich, 2010). Findings of previous analyses focus on economic benefits (both at the community and household level), with few analyzing other socioeconomic impacts of the conservancy program.  An analysis of national monetary benefits of five conservancies was done by Barnes et al. (2002) and determined that conservancies provide positive annual contributions to gross and net national income, contributing positively to national economic well-being. At the community level, this study found that financial returns for communities from CBNRM initiatives exceed levels of investments. Another analysis of financial benefits to communities was done by Naidoo et al. (2011b) and explored the link between the diversity of large wildlife assemblages and the economic success of conservancies. This study demonstrated that the diversity of wildlife has an important positive effect on financial  !! ! !9!  Figure 4. The range expansion of lion populations in northwest Namibia, shown between 1995 and 2007 (NACSO, 2013; taken from http://www.nacso.org.na).   returns of the CBNRM program, since greater numbers of wildlife species provide a greater range of hunting options and tourist viewing opportunities. A similar study by Naidoo et al. (2011c) examined how biodiversity in communal areas of Namibia is related to financial benefits derived from two ecosystem services, trophy hunting and ecotourism. This study also found a positive relationship between biodiversity, as represented by large wildlife !! ! !10!species, and economic benefits generated by conservancies, suggesting a major financial incentive for conservation in the region. An analysis of conservancy welfare benefits by Bandyopadhyay et al. (2011) looked at seven conservancies and found that conservancies provide employment opportunities, dividends and improved meat distribution, with welfare benefits apparent at both the community and household level. This study found mean household incomes to be 24% higher in more established conservancies, suggesting a positive and cumulative impact of CBNRM activities. The study also found conservancies to be poor-neutral in one of the surveyed regions, and pro-poor in another, refuting the concern that conservancies are disproportionately beneficial to the wealthy. A contrasting study by Silva and Mosimane (2012) analyzed conservancies in the same two regions and came to the conclusion that although conservancy members that participate in CBNRM initiatives do receive direct economic benefits, these do not adequately compensate for practices that are considered unjust or unfair by community residents. These practices include an optional membership policy, which leads to differing degrees of involvement in the program, as well as hunting restrictions on certain game species, increased occurrence of crop raiding by wild animals and unequal distribution of conservancy benefits. This study stresses the importance of improving participation of conservancy members in the program, and paying attention to local concepts of equitable development. A similar study of these two regions by Suich (2010) found that conservancies led to increased food availability in certain areas due to distribution of legal game meat, but decreased food availability and threatened household food security in other areas due to increases in wildlife.   Aside from the financial benefits provided by conservancies, Naidoo and Johnson (2013) analyzed the behavioural impact of an HIV/AIDS outreach and mainstreaming effort associated with the CBNRM program. The study found a 50% decrease in the mean reported number of sexual partners for men in conservancies relative to men outside of conservancies. Considering the substantial economic and social impacts of HIV, these findings suggest potentially significant improvements in the socioeconomic well-being of communities within conservancies.  Prior analyses of Namibia’s CBNRM program have generally focused on income generation as an outcome using a sub-sample of conservancies, often lacking empirical !! ! !11!evidence. Only a few of these evaluations used quasi-experimental methods (Bandhyopadhyay, Naidoo, Johnson), and those that have, lack evaluation of many aspects of livelihood that have yet to be examined quantitatively. Our analysis uses large-scale datasets at the household level, employing similar techniques to Naidoo and Johnson but with a broader range of welfare outcomes.      3. DATA AND METHODS   3.1 Demographic and Health Surveys Data We used Demographic and Health Surveys (DHS) data from 2000 and 2006/07 to evaluate the effect of conservancies on various health, education and wealth outcome variables. The DHS are nationally and sub-nationally representative surveys, implemented using a stratified 2-stage cluster sampling design. They contain detailed demographic and socioeconomic data at both the individual and household level, obtained by interviewing women and men aged 15-49 on a variety of issues related to household assets, reproductive health, family planning and child health. In Namibia, 6392 households participated in the 2000 survey (household response rate: 97%; individual response rate: 92%) and 9200 households participated in the 2006/07 survey (household response rate: 98%; individual response rate: 95%) (MoHSS, 2003, 2008). This included 429 households in 13 conservancies in 2000 and 581 households in 22 conservancies in 2006/07 (Figure 5). DHS surveys do not provide panel data, meaning that households interviewed in 2006/07 are different than those sampled in 2000.  The DHS are globally recognized as a key source of comparative quantitative data across developing countries (Chan et al., 2010). DHS surveys in Namibia were conducted by the Ministry of Health and Social Services (MoHSS) in collaboration with the Central Bureau of Statistics and with technical assistance provided by ICF Macro through the MEASURE DHS project (MoHSS, 2003, 2008). Survey design and implementation passed a national ethics review panel and participation was voluntary with informed consent obtained from all survey respondents. !! ! !12! !Figure 5. Map of DHS cluster locations for the 2000 and 2006/07 surveys (MoHSS, 2003, 2008).   3.2 Outcome Variables of Interest The outcome variables we chose to analyze cover four main categories related to socioeconomic well-being: (1) disease prevention, (2) disease prevalence and duration, (3) education and (4) wealth. When choosing outcome variables, we were limited to those asked in both the 2000 and 2006/07 DHS surveys with a good response rate. The dependent variables chosen to best represent the four socioeconomic categories of interest are: bednet ownership and usage (1), diarrhea prevalence and treatment (2), school attendance (3) and wealth index (4) (Appendix 1).  1. Bednet ownership and usage: These variables were chosen to represent disease prevention (1) as an indicator of socioeconomic well-being. Malaria is the 10th most !! ! !13!common cause of death in Namibia and bednets treated with insecticide are an effective way to prevent the disease (CDC, 2013). Survey respondents were asked whether the household owned a bednet, and whether she slept under a bednet the previous night. We coded households with a bednet as “1” and households without a bednet as “0” for the ownership outcome variable. For the bednet usage variable, we coded respondents who slept under the bednet as “1” and those who didn’t as “0”.  2. Diarrhea prevalence and duration: These variables represent disease prevalence and duration (2) as an indicator of socioeconomic well-being. Diarrheal diseases are the 5th most common cause of death in Namibia and were therefore important to capture in our outcome variables (CDC, 2013). In the DHS surveys, mothers of children under age 5 were asked whether their child had experienced diarrhea in the past 2 weeks. We coded children whose mothers answered yes with a “1” and the remaining children with “0”. The surveys also asked mothers who answered yes whether the child had received any treatment for this bout of diarrhea. We coded children that had received medical treatment with a “1” and those that hadn’t with a “0”.  3. School attendance: This variable represents education (3) as an indicator of socioeconomic well-being. DHS data contains information on the school attendance of each household member. We chose school-aged children (ages 6 to 16) as our unit of analysis and coded children who were currently attending school with a “1” and all others with a “0”.  4. Wealth Index: This variable represents wealth (4) as an indicator of socioeconomic well-being. Relative household wealth is included in DHS data as an asset-based wealth index, reported as both a standardized factor score and as a quintile. Data on household asset ownership collected during the survey are dichotomized and entered into a principal component analysis (PCA), which assigns weights to each asset. The asset values are then weighted accordingly and summed for each household, yielding the household factor score (Rutstein & Johnson, 2004). We used this standardized factor score as a continuous variable in our analysis of household wealth.     !! ! !14!3.3 Overview of Methods   In order to compare conservancy (treatment) and non-conservancy (comparison) households/individuals, comparison groups were created with the goal of having non-statistically different treatment and comparison groups for each outcome variable in 2000. This ensures that any difference in temporal trends between the two groups can be attributed to the presence or absence of a conservancy program. Three comparison groups were created for each treatment group. A statistically matched comparison group was used, since quasi-experimental matching models are regarded as one of the best alternatives when random experiment design is not possible (Rubin, 1973). The other two comparison groups included all households/individuals in the geographically nearest sampling cluster and all households/individuals outside of conservancies. Comparison groups were then evaluated based on mean differences from the treatment group for each outcome variable. The best comparison group was chosen for each outcome variable, applied to the 2006/07 data, and then used in a regression analysis. The purpose of the regression was to determine the effect of conservancy residence over time on each socioeconomic outcome (Figure 6).  3.3.1 Matching of Treatment and Comparison Groups   We compared trends in conservancy households with three non-conservancy comparison groups: (1) all surveyed households outside of conservancies; (2) all surveyed households in the nearest DHS sampling cluster outside of each surveyed conservancy; and (3) a matched comparison group determined using Mahalanobis distance matching (Stuart & Rubin, 2008). The quasi-experimental comparison group (3) was matched to be similar to in-conservancy households in terms of characteristics that may confound the conservancy impact on socioeconomic indicators of interest. The variables of socioeconomic well-being we were interested in analyzing included both household level variables and individual level variables. For this reason, the three comparison groups were created for four different units of analysis: (a) households; (b) children under 5; (c) school-aged children; and (d) female respondents. For the quasi-experimental comparison group (3), conservancy households/individuals were matched with households/individuals outside of conservancies using the following variables for all four units of analysis: !! ! !15!(1) Number of household members (2) Urban/rural residence (3) Gender of household head (4) Education of household head (5) Language of household respondent (6) Distance to main roads (7) Geographical region (8) Precipitation (9) Altitude (10)  Biome Variables used only for the individual units of analysis are as follows: (11)  Age (b, c, d) (12)  Gender (b, c) (13)  Number of household members aged 5 and under (b) (14)  Number of household members between 6 and 16 years of age (c) Some variables were used in the matching model for only the relevant outcome variables. These include: (15) Distance to nearest health clinic (16) Source of drinking water (17) Wealth Index Distance to a health clinic was used only for outcome variables related to health and disease, source of drinking water only for diarrhea-related outcome variables, and household wealth index was included only for health and education outcome variables. Matching variables chosen are those that may determine both economic and social well-being as they can affect access to economic opportunities, health care and education. In cases where the inclusion of all matching variables resulted in many comparison group matches being used multiple times, and thereby weakening the matching model, some confounding variables of interest were removed from the matching model and analyzed as covariates within the regression model instead. These cases will be made clear in the results below. !! ! ! 16! Figure 6. Flowchart summary of methods. !! ! !17!The matching model used in the creation of the matched comparison groups (3) was 1-to-1 nearest neighbour matching with replacement, using a Mahalanobis distance metric. This model was found to be the best at producing comparison groups with matching variable distributions that were similar to those of the treatment (conservancy) groups. The Mahalanobis distance is a descriptive statistic, measuring the distance of a point from a data distribution by calculating the Euclidian distance while taking into account covariance in the data (Stuart & Rubin, 2008). The matching was implemented using the ‘Matching’ library of the statistical software R (Sekhon, 2011). Using a model with replacement means that a single non-conservancy household/individual could be matched with multiple in-conservancy households. In cases where more than one non-conservancy household/individual was tied with a conservancy household as the best match (meaning they were the same Mahalanobis distance away), both of these non-conservancy households/individuals were used and weighted correspondingly.   For each outcome variable, we evaluated which comparison group provided the smallest difference between conservancy and non-conservancy households/individuals in 2000. This was done using a two-sample t-test to compare conservancy households to each of the three non-conservancy comparison groups for each outcome variable. The comparison group that yielded the smallest mean difference when compared to the conservancy group in 2000 was considered to be the best matching model. T-tests were used to ensure that this difference was not statistically different from zero and that the comparison group was therefore not statistically different from the treatment group in 2000 (Appendix 2). The best matching model was then applied to the 2006/07 data to produce a comparison group and complete the regression analysis described below. This assumes that because the matching model produced statistically similar treatment and comparison groups in 2000, when applied in 2006/07 it would produce identical outcomes between treatment and comparison households/individuals in the absence of an impact of conservancies on the outcome variable.   3.3.2 Regression Analysis We used generalized linear models (GLMs) to evaluate the statistical significance of the effects of year and conservancy residence on each outcome variable. A binary logistic !! ! !18!regression model was used for all binary outcome variables and a linear regression model for continuous variables: Yi = β0 + β1conservancyi + β2yeari + β3conservnacyi × yeari + εi  (1) where Yi is the response variable of the ith household, conservancyi is a binary variable indicating whether the ith household is within a conservancy, yeari is a binary variable indicating the year of the response by the ith household (if year of survey=2006/07, yeari=1), β0 is the intercept of the regression model, β1 and β2 are the coefficients on the main effects of conservancy and year, β3 is the interaction effect between conservancy and year, and εi is the error term, which is assumed to be independent and normally distributed. In cases of binary outcome variables, the logit function is: Pi = 1/(1+e-Y)     (2) where P is the probability that the household/individual experiences an increase in the outcome variable, e is a constant, and Y is the log odds of the dependent variable, given by Yi=ln(Pi/(1-Pi)). In the case of binary outcome variables, β coefficients in the linear model are interpreted as eβ=odds ratio, where the odds ratio expresses the likelihood of change in the dependent variable given a unit of change in the independent variable (Kutner et al., 2004). In the case of continuous outcome variables, β coefficients in the linear model represent how much the dependent variable is expected to increase when that independent variable increases by one, holding all other independent variables constant. The main determinant as to whether conservancy residence has an impact on the temporal trends in each outcome variable was whether the conservancy-year interaction term was significantly different than zero, where a significantly positive coefficient means that the temporal trend in conservancies was significantly greater than the trend in the comparison group.         !! ! !19!4. RESULTS   4.1 Malaria Prevention 4.1.1 Bednet Ownership  In our analysis of household bednet ownership, the comparison group that produced the smallest difference between conservancy and non-conservancy households in 2000 was the nearest geographical cluster matching model, with a p-value of 0.347 between treatment and comparison households (Figure 7). Our final dataset, including all conservancy households and geographically nearest non-residents, contained 750 households in 2000 (400 in conservancy, 350 non-conservancy) and 1170 households in 2006 (581 in conservancy, 589 non-conservancy).   Figure 7. Trends from 2000 to 2006/07 for conservancy households (filled squares, solid line) versus 3 comparison groups (dashed lines, circles = quasi-experimental match; triangles = nearest geographical cluster; diamonds = entire non-conservancy population) for the proportion of households that own a bednet.    !! ! !20!Logistic regression results indicate that conservancy households are significantly more likely to increase the proportion of bednet ownership from 2000 to 2006/07, relative to those in the comparison group (Table 2). The odds ratio indicates that in-conservancy households are more than twice as likely to have increased bednet ownership over this time period as non-conservancy households. The second set of regression results reported in Table 2 are produced by the same regression model, but with the inclusion of two additional covariates that we expected to have an important effect on the proportion of bednet ownership and which we were interested in observing directly. The incorporation of these variables improved the pseudo R-squared value of the model and produced similar results, with in-conservancy households being about twice as likely to have increased bednet ownership over time as non-conservancy households. The interaction term remained significant (though not at the same level) indicating that the regression model is robust. When including the covariates, the year effect on bednet ownership becomes stronger and marginally significant, and the conservancy residence coefficient remains insignificant. The education of the household head has a significant and positive effect on bednet ownership, and household wealth is also significant, with a fairly strong negative effect on bednet ownership.   Table 2. Logistic regression model results for household bednet ownership. Bednet Ownership  Coefficient Std. Error Z-value p Odds Ratio No additional covariates      Intercept -1.441 0.136 -10.60 <0.001*** 0.237 Year 0.218 0.168 1.302 0.193 1.244 Conservancy residence -0.181 0.191 -0.945 0.345 0.835 Year:conservancy interaction 0.774 0.232 3.335 <0.001*** 2.168  N = 1920      Pseudo R2 = 0.040     With additional covariates      Intercept -2.740 0.190 -14.41 <0.001*** 0.065 Year 0.311 0.177 1.760 0.078・  1.364 Conservancy residence -0.135 0.199 -0.680 0.497 0.874 Education of household head 0.129 0.014 9.160 <0.001*** 1.137 Wealth index -0.998 0.094 -10.60 <0.001*** 0.368 Year:conservancy interaction 0.687 0.242 2.837 0.005** 1.988  N = 1908      Pseudo R2 = 0.160     !! ! !21! 4.1.2 Bednet Usage  The quasi-experimental matched group was the best comparison group when analyzing the proportion of respondents that slept under a bednet during the previous night, with a p-value of 0.754 between treatment and comparison groups in 2000 (Figure 8). Our final dataset, including all in-conservancy female respondents and matched non-conservancy respondents from households that own a bednet, contained 139 women in 2000 (69 conservancy residents, 70 non-conservancy) and 410 women in 2006/07 (202 conservancy residents, 208 non-conservancy).  Although logistic regression results suggest that women in conservancies are approximately 20% more likely than non-conservancy women to have increased bednet usage over time, these results were not statistically significant (Table 3).   Figure 8. Trends from 2000 to 2006/07 for conservancy residents (filled squares, solid line) versus 3 comparison groups (dashed lines, circles = quasi-experimental match; triangles = nearest geographical cluster; diamonds = entire non-conservancy population) for the proportion of respondents that slept under a bednet during the previous night.   !! ! !22! Table 3. Logistic regression model results for bednet usage by the respondent during the previous night Bednet Usage  Coefficient Std. Error Z-value p Odds Ratio Intercept -0.616 0.256 -2.405 0.016* 0.540 Year 0.025 0.295 0.084 0.933 1.025 Conservancy residence 0.112 0.357 0.316 0.752 1.119 Year:conservancy interaction 0.179 0.411 0.435 0.663 1.196  N = 547      Pseudo R2 = 0.006       4.2 Diarrhea Prevalence and Treatment 4.2.1 Diarrhea Prevalence  The comparison group that produced the smallest difference in 2000 between conservancy and non-conservancy children who experienced diarrhea was the quasi-experimental matched group, with a p-value of 0.709 between treatment and comparison groups (Figure 9). Four covariates were removed from the matching model, as their inclusion resulted in a worse match between treatment and comparison groups in 2000. This is caused by the large number of matching variables limiting the number of good matches between conservancy and non-conservancy children in 2000. Variables left out of the matching model were those we were interested in observing the effect of and were included as covariates in the regression model instead. The final dataset used for the analysis of this variable contained 447 children in 2000 (217 conservancy members, 230 non-conservancy) and 614 children in 2006 (284 conservancy residents, 330 non-conservancy). Logistic regression results indicate a non-significant decreasing effect of conservancy residency on the rate of change of the proportion of children (ages 5 and under) to have diarrhea in the past two weeks, with in-conservancy children being approximately 20% less likely to experience an increase in diarrhea occurrence, but to a non-significant level (Table 4). The inclusion of covariates in the model increased the pseudo-R-squared value slightly and shows that education of the household head, household wealth, distance to health clinic and type of water source all have a non-significant effect on diarrhea prevalence in children. Including covariates changed the interaction term, questioning the robustness of the model, and indicates that conservancy and non-conservancy children are about equally as likely to experience an increase in diarrhea occurrence over time. The interaction term remained  !! ! !23! Figure 9. Trends from 2000 to 2006/07 for conservancy children (filled squares, solid line) versus 3 comparison groups (dashed lines, circles = quasi-experimental match; triangles = nearest geographical cluster; diamonds = entire non-conservancy population) for the proportion of children under 5 that had diarrhea within the past 2 weeks.   insignificant, however, and therefore no statistically significant conservancy impacts on this variable were found.  4.2.2 Diarrhea Treatment  The best comparison group for the proportion of children who received medical treatment for diarrhea in the past two weeks was the quasi-experimental matched group, with a p-value of 0.931 between treatment and comparison groups in 2000 (Figure 10). Our final dataset, including all conservancy children and matched non-conservancy children, included 69 children in 2000 (36 conservancy residents, 33 non-conservancy) and 89 children in 2006 (39 conservancy residents, 50 non-conservancy). The reason for the small sample size is that only children whose mothers’ responded yes to them having had diarrhea in the past two weeks were eligible to answer this question. Four additional covariates were removed from   !! ! !24!Table 4. Logistic regression model results for diarrhea prevalence in children under 5 within the past 2 weeks. Diarrhea Prevalence  Coefficient Std. Error Z-value p Odds Ratio No additional covariates      Intercept -1.714 0.191 -8.992 <0.001*** 0.180 Year 0.005 0.249 0.020 0.984 1.005 Conservancy residence 0.099 0.264 0.374 0.709 1.104 Year:conservancy interaction -0.228 0.354 -0.644 0.520 0.796  N = 1061      Pseudo R2 = 0.001     With additional covariates      Intercept -1.479 0.295 -5.01 <0.001 *** 0.228 Year -0.144 0.264 -0.55 0.585 0.866 Conservancy residence -0.096 0.276 -0.35 0.728 0.908 Education of household head -0.037 0.023 -1.58 0.113 0.964 Wealth index 0.086 0.162 0.53 0.596 1.090 Distance to health facility 0.000 0.000 0.26 0.797 1.000 Improved water source 0.031 0.226 0.14 0.892 1.031 Year:conservancy interaction 0.086 0.370 0.23 0.817 1.089  N=995      Pseudo R2 = 0.006       the matching model, as they weakened the match between conservancy and non-conservancy children in 2000 for the same reasons given for diarrhea prevalence above. Regression results suggest a non-significant and slight increase in the likelihood of receiving medical treatment for diarrhea within conservancies compared to non-conservancy children (Table 5). Including additional covariates in the regression model improved the pseudo-R-squared value, but all covariates as well as the interaction term remained insignificant.  4.3 School Attendance  In our analysis of school attendance by children (ages 6-16), the best comparison group was the quasi-experimentally matched model, with a p-value of 0.307 between conservancy and comparison children in 2000 (Figure 11). Our final dataset contained 877 children in 2000 (427 conservancy residence, 450 non-conservancy) and 1239 children in 2006 (601 conservancy residents, 638 non-conservancy).  !! ! !25! Figure 10. Trends from 2000 to 2006/07 for conservancy children (filled squares, solid line) versus 3 comparison groups (dashed lines, circles = quasi-experimental match; triangles = nearest geographical cluster; diamonds = entire non-conservancy population) for the proportion of children under 5 that received medical treatment for their diarrhea in the past 2 weeks.   Logistic regression results indicate that school attendance of conservancy children is stable between 2000 and 2006/07, while school attendance of the matched comparison group increases over time to a significant degree. Conservancy children have a significantly lower rate of growth in school attendance over time, with the odds ratio indicating that the proportion of conservancy children attending school is approximately 45% less likely to increase over time than non-conservancy children (Table 6).  4.4 Household Wealth  For household wealth index factor scores, none of the matching models we tried were able to reduce mean differences between conservancy and non-conservancy households to a statistically insignificant level in 2000. This is depicted in Figure 12, where the difference between treatment and comparison groups is statistically different from zero for all matching models. The best comparison group was the quasi-experimental matching model, with a p-  !! ! !26! Table 5. Logistic regression model results for diarrhea treatment in children under 5 within the past 2 weeks. Diarrhea Treatment  Coefficient Std. Error Z-value p Odds Ratio No additional covariates      Intercept -0.154 0.352 -0.438 0.661 0.857 Year 0.285 0.460 0.620 0.535 1.329 Conservancy residence 0.043 0.485 0.088 0.929 1.044 Year:conservancy interaction 0.406 0.659 0.617 0.537 1.501  N = 158      Pseudo R2 = 0.026     With additional covariates      Intercept -0.231 0.571 -0.40 0.686 0.794 Year 0.435 0.498 0.87 0.382 1.545 Conservancy residence 0.295 0.522 0.57 0.572 1.343 Education of household head 0.005 0.046 0.11 0.915 1.005 Wealth index -0.411 0.347 -1.18 0.237 0.663 Distance to health facility -0.002 0.001 -1.63 0.104 0.998 Improved water source -0.242 0.446 -0.54 0.588 0.785 Year:conservancy interaction 0.296 0.709 0.42 0.677 1.344  N = 144      Pseudo R2 = 0.090       value of 0.031. In this case, our dataset contained 645 households in 2000 (289 in conservancy, 356 non-conservancy) and 1233 households in 2006/07 (578 in conservancy, 655 non-conservancy). Bearing in mind our inability to identify an appropriate comparison group model, linear regression results are reported in Table 7. No conclusions can be drawn from the model since we were unable to match the treatment and comparison groups sufficiently in 2000. A t-test of conservancy versus matched comparison groups in 2006/07 produces a p-value of 0.015, indicating that households remain statistically different in 2006/07 with no sign of convergence between groups.   !! ! !27! Figure 11. Trends from 2000 to 2006/07 for conservancy children (filled squares, solid line) versus 3 comparison groups (dashed lines, circles = quasi-experimental match; triangles = nearest geographical cluster; diamonds = entire non-conservancy population) for the proportion of school-aged children that attended school during the current year.     Table 6. Logistic regression model results for school attendance of children ages 6-16. School Attendance  Coefficient Std. Error Z-value p Odds Ratio Intercept 0.998 0.109 9.149 <0.001*** 2.713 Year 0.774 0.159 4.865 <0.001*** 2.167 Conservancy residence 0.161 0.157 1.021 0.307 1.174 Year:conservancy interaction -0.595 0.220 -2.710 0.007* 0.551  N = 2116      Pseudo R2 = 0.020       !! ! !28! Figure 12. Trends from 2000 to 2006/07 for conservancy households (filled squares, solid line) versus 3 comparison groups (dashed lines, circles = quasi-experimental match; triangles = nearest geographical cluster; diamonds = entire non-conservancy population) for standardized household wealth factor scores.     Table 7. Linear regression model results for household wealth index factor scores. Wealth Index  Coefficient Std. Error t-value p Intercept -0.703 0.038 -18.54 <0.001*** Year 0.035 0.046 0.757 0.449 Conservancy residence 0.111 0.054 2.062 0.040• Year:conservancy interaction -0.011 0.066 -0.165 0.869  N = 1878     Multiple R2 = 0.006 Adjusted R2 = 0.005          !! ! !29!5. DISCUSSION   5.1 Health Outcomes  The presence of the CBNRM program in Namibia has had positive household-level effects on health outcomes related to malaria prevention. Bednet ownership inside conservancies is significantly more likely to increase over time than outside conservancies, and bednet usage in conservancies is also more likely to improve over time, although not to a significant degree. Our analysis accounted for wealth and urban/rural residence, as well as geographic location, between conservancy and non-conservancy groups, implying that this finding is not due to targeting of bednet dispersal to rural, low income communities or to a higher risk of malaria in conservancy regions. This leads us to believe that increased bednet ownership and usage can be attributed to the presence of conservancies and considered a benefit of the CBNRM program. The community structure provided by conservancies may make it easier for government distribution of bednets and improve the effectiveness of current bednet distribution and education programs (MoHSS, 2010). Although there are no specific malaria-prevention education programs associated with conservancies, other health-related education programs such as the HIV/AIDS outreach and education program may encourage conservancy residents to educate themselves and be more active in preventing disease. Training and capacity building programs are provided by tourism enterprises to employees and by NGOs to conservancy staff, and one of the effects of this training may be a better understanding of the importance of disease prevention and treatment, as well as increased confidence of women in sharing their ideas and opinions (Suich, 2010). Our results indicate that a more educated head of household improves the likelihood of owning a bednet, both in and out of conservancies. This is expected based on previous analysis of bednet ownership in Sub-Saharan Africa (Njau et al., 2013). Our results also indicate that wealthier households are less likely to own a bednet, both in and out of conservancies. This is an unexpected result and may be due to three possible reasons: wealthier homes may use different malaria prevention methods such as secure screens, repellant creams and antimalarial medications; bednet distribution and promotion programs may preferentially target lower income households (MoHSS, 2010); or this may be due to a !! ! !30!structural difference if poorer households live in regions with a higher risk of malaria. Analysis of wealth and malaria distributions in the country suggest that this may be the case, as lower income households and higher risk of malaria both occur in the most northern Namibian provinces (Central Bureau of Statistics, 2011; Tatem et al., 2014).  Our analysis of bednet usage did not provide statistically significant results, although it may suggest a higher likelihood of increased bednet usage over time in conservancies. The best matching model for this variable was the quasi-experimentally matched group, but we note that if the geographically nearest matching model was chosen (which is also not statistically different to the treatment group in 2000, p-value=0.447), the results would have been very different and statistically significant (Figure 8). This demonstrates the importance of the matching model when determining effects of conservancies over time. It is also important to note that our analysis of this variable was limited by the 2000 DHS dataset to female respondents only. This means that if the respondent reported no to this question, there is still the possibility that a child or other female household member used the bednet during the previous night.  Trends in prevalence and treatment of diarrhea in conservancy children are not statistically different from non-conservancy children. Results indicate that conservancy children may be less likely to have diarrhea and more likely to receive medical treatment over time, but conclusions cannot be drawn with statistical certainty. Analysis of the diarrhea treatment variable was done using a small sample size, since only a fraction of children were eligible for this question, and may have compromised the robustness of the regression model. Previous analyses in different regions (such as Johnson et al. (2013)), the expected benefits of ecosystem conservation (such as natural water purification processes) and expected benefits of conservancy structure (such as community taps to provide safe drinking water), all predict that Namibia’s CBNRM program would have a positive effect on diarrhea prevalence in conservancies (NACSO, 2011; O’Gorman, 20116). Our results were unable to verify this prediction, but the use of a longer time period and larger sample size in future analysis may improve statistical significance.  A limitation of the 2000 DHS survey, in terms of health outcomes, was a lack of relevant variables allowing for analysis of nutrition. Ordinarily our analysis of the health impacts of conservancies would include a nutrition outcome variable, but due to the apparent !! ! !31!universal nature of the nutrition-related variables collected in this survey (such as vitamin A supplements), we were unable to draw any conclusions in this regard.   5.2 Education Outcomes  Our results indicate that school attendance of children living in conservancies is less likely to increase over time compared to non-conservancy children. The school attendance of conservancy children is not decreasing over time, but instead is remaining the same while the school attendance of non-conservancy children increases. Under the assumption that conservancy children would match the school attendance of non-conservancy children in the absence of conservancies, our results suggest that the presence of conservancies is responsible for this lack of improvement in school attendance over time. These results support the findings of Silva and Mosimane (2012), which suggest that indirect benefits of the CBNRM program, including improved infrastructure such as schools, have not yet been realized for many communities.  5.3 Wealth Outcomes  We are unable to draw any conclusions from our analysis of household wealth due to an insufficient match between treatment and comparison groups in 2000. It is unclear from our analysis, as well as previously published literature, whether it is expected for conservancies to have a direct positive impact on household wealth. Although previous research has demonstrated community-level financial benefits due to conservancies, it is not clear whether these positive impacts will be seen at the household level (Barnes, 2002; Naidoo, 2011b,c). Given the debate surrounding the distribution of economic benefits in conservancies and whether these benefits extend down to households, we strongly recommend further analysis of conservancy residence on household wealth. A better matching model that is able to find non-statistically different households in 2000 would be better able to determine household wealth trends over time. The small R-squared value of our regression model also suggests that many of the factors affecting household wealth are not captured in this model, and a more inclusive model may provide better insight into the effects of conservancy residence on household wealth.   !! ! !32!5.4 Limitations and Assumptions The use of DHS data means that surveys were designed to be nationally and regionally representative, but did not take into account the location of conservancies. We assumed that the matching of households dealt with this issue of unequal survey distribution inside and outside of conservancies. The DHS are not panel surveys, meaning that the households interviewed in 2006/07 were not the same as in 2000, causing difficulties in comparison between the two years. Our analysis was also limited to the questions asked by the DHS surveys, which inevitably lack some of the confounding factors impacting the outcomes we were interested in. Our analysis assumes that statistically identical households in 2000 would remain the same in 2006/07 in the absence of conservancies. The validity of this assumption depends on the strength of the matching model used, and improvements in the matching model for certain variables may provide a more robust analysis of socioeconomic trends over time. We did not incorporate year of conservancy registration in our analysis, which may have important impacts on the results, as more established conservancies are better indicators of the long-term effects of this program.  We also note that the R-squared and pseudo-R-squared values for most regression models were small, suggesting that the models did not capture all the variability in the data for most outcome variables. Since most outcome variables were binary, the pseudo-R-squared values were not expected to be close to one, but were considerably low given that a value between 0.2 and 0.4 indicates a good fit (Kutner et al., 2004).    6. CONCLUSION   Results of this thesis indicate mixed effects of Namibia’s CBNRM program at the household level, given the data available. We found that Namibia’s conservancy program has a positive effect on malaria prevention in conservancy households, a negative effect on school attendance of conservancy children, and recommend further analysis of diarrhea prevalence and household wealth trends over time. For community conservation programs, !! ! !33!such as this one, it is important to analyze household and individual level effects when determining the overall impact of the program, as community benefits do not necessarily extend to the household level. We recommend further analysis of these trends, specifically with the release of the 2013 DHS, from which an extended time period, more variables and larger sample sizes will allow for a valuable addition to this analysis. Use of the 2013 DHS will also capture any impacts of conservancies that have occurred within the past 6 years, which are not captured in this analysis. It is important to note that this study did not consider the issue of equity in the distribution of benefits and costs associated with conservancy activities. Future work should address these issues for an improved understanding of the effects of conservancies on Namibians.   !! ! !34!REFERENCES CITED   Adams, W. M., Aveling, R., Brockington, D., Dickson, B., Elliott, J., Hutton, J., Roe, D.,  Vira B. & Wolmer, W. (2004). Biodiversity conservation and the eradication of poverty. Science, 306(5699), 1146-1149.  Agrawal, A., & Redford, K. (2006). Poverty, development, and biodiversity conservation:  Shooting in the dark? Wildlife Conservation Society, 1-56.  Bandyopadhyay, S., Humavindu, M., Shyamsundar, P. & Wang,  L. (2009). Benefits to local communities from community conservancies in Namibia: an assessment, Development Southern Africa, 26(5), 733-754.   Barnard, P. (1998). Biological diversity in Namibia: a clash of sea and land, fog and dust. Biodiversity and Conservation, 7(4), 415-417.  Barnes, J. I., Macgregor, J., & Chris Weaver, L. (2002). Economic efficiency and incentives for change within Namibia's community wildlife use initiatives.World development, 30(4), 667-681.  Chan, M., Kazatchkine, M., Lob-Levyt, J., Obaid, T., Schweizer, J., Sidibe, M., Veneman,  A., & Yamada, T. (2010). Meeting the demand for results and accountability: a call for action on health data from eight global health agencies. PLoS Med, 7(1), e1000223.  Centers for Disease Control and Prevention (CDC). (2013). CDC in Namibia. Atlanta, GA.   Central Bureau of Statistics. (2011). An Atlas of poverty in Namibia. Central Bureau of Statistics, Windhoek.   Cooke, P. A. (1998). Intrahousehold labor allocation responses to environmental good  scarcity: a case study from the hills of Nepal. Economic Development and Cultural Change, 46(4), 807-830.  Frost, P. G., & Bond, I. (2008). The CAMPFIRE programme in Zimbabwe: payments for  wildlife services. Ecological Economics, 65(4), 776-787.  Johnson, K. B., Jacob, A., & Brown, M. E. (2013). Forest cover associated with  improved child health and nutrition: evidence from the Malawi Demographic and Health Survey and satellite data. Global Health: Science and Practice, 1(2), 237-248.  Kumar, S. K., & Hotchkiss, D. (1988). Consequences of deforestation for women's time  allocation, agricultural production, and nutrition in hill areas of Nepal (Vol. 69). Washington DC: International Food Policy Research Institute. !! ! !35! Kutner, M. H., Neter, J., Nachtsheim, C. J., & Li, W. (2004). Applied linear statistical  models. New York, NY: McGraw Hill Higher Education.  Mendelsohn, J., A. Jarvis, C. Roberts, and T. Robertson. (2002). Atlas of Namibia: a portrait  of the land and its people. David Philip Publishers, Cape Town, South Africa.  Millennium Ecosystem Assessment. (2005). Living beyond our means: natural assets and  human well-being. Statement from the Board. UNEP, Paris, France.   Ministry of Health and Social Services (MoHSS) [Namibia] and Macro International Inc.  (2008). Namibia Demographic and Health Survey 2006-07. Windhoek, Namibia and Calverton, Maryland, USA: MoHSS and Macro International Inc.   Ministry of Health and Social Services (MoHSS) [Namibia]. (2003). Namibia  Demographic and Health Survey 2000. Windhoek, Namibia: MOHSS.   Ministry of Health and Social Services (MoHSS). (2010). Malaria Strategic Plan (2010- 2016). Windhoek: Republic of Namibia.  NACSO. (2013). Namibia’s communal conservancies: a review of progress and  challenges in 2011. NACSO, Windhoek.  Naidoo, R., & Johnson, K. (2013). Community-based conservation reduces sexual risk factors for HIV among men. Globalization and health, 9(27), 1-5.  Naidoo, R., Stuart-Hill, G., Weaver, L. C., Tagg, J., Davis, A., & Davidson, A. (2011b).  Effect of diversity of large wildlife species on financial benefits to local communities in northwest Namibia. Environmental and Resource Economics, 48(2), 321-335.  Naidoo, R., Weaver, L. C., De Longcamp, M., & Du Plessis, P. (2011a). Namibia's  community-based natural resource management programme: an unrecognized payments for ecosystem services scheme. Environmental Conservation, 38(4), 445-453.  Naidoo, R., Weaver, L. C., Stuart‐Hill, G., & Tagg, J. (2011c). Effect of biodiversity on  economic benefits from communal lands in Namibia. Journal of Applied Ecology, 48(2), 310-316.  Njau, J. D., Stephenson, R., Menon, M., Kachur, S. P., & McFarland, D. A. (2013). Exploring the impact of targeted distribution of free bed nets on households’ bed net ownership, socio-economic disparities and childhood malaria infection rates: analysis of national malaria survey data from three sub-Saharan Africa countries. Malaria journal, 12(1), 245.  O’Gorman, T. L. (2006). Species and People: Linked futures. World Wildlife Fund, Swedish  !! ! !36!International Cooperation Agency.  Palmer, C., & Di Falco, S. (2012). Biodiversity, poverty, and development. Oxford   Review of Economic Policy, 28(1), 48-68.  Pelser, A., Redelinghuys, N., & Velelo, N. (2013). Protected areas as vehicles in  population development: lessons from rural South Africa. Environment, Development and Sustainability, 1-22.  Roe, D., Nelson, F., Sandbrook, C. (2009). Community management of natural resources in  Africa: impacts, experiences and future directions. London, UK: IIED.  Rubin, D. B. (1973). The use of matched sampling and regression adjustment to remove  bias in observational studies. Biometrics, 185-203.  Rutstein S., & Johnson K. The DHS wealth index. Calverton, MD: ORC Macro; 2004. DHS  Comparative Reports No. 6. Available from: http://www.measuredhs.com/pubs/pdf/CR6/CR6.pdf   Sekhon, J. S. (2011). Multivariate and propensity score matching software with automated balance optimization: the matching package for R. Journal of Statistical Software, 42(7), 1-52.  Silva, J. A., & Mosimane, A. W. (2013). Conservation-Based Rural Development in Namibia A Mixed-Methods Assessment of Economic Benefits. The Journal of Environment & Development, 22(1), 25-50.  Stuart, E. A., & Rubin, D. B. (2008). Best practices in quasi-experimental designs:  Matching methods for causal inference. In J. Osborne, Best practices in quantitative methods (pp. 155-176). USA: Sage Publications.   Suich, H. (2010). The livelihood impacts of the Namibian community based natural resource management programme: a meta-synthesis. Environmental Conservation, 37(1), 45-53.  World Bank Group. (2013). Namibia. Retrieved from  http://www.worldbank.org/en/country/namibia  Tatem, A., Huang, Z., Narib, C., Kumar, U., Kandula, D., Pindolia, D., Smith, D., Cohen, J.,  Graupe, B., Uusiku, P., & Lourenco, C. (2014) Predicted probability of malaria cases in January-May 2011 for northern Namibia (with named regions marked). Malaria Journal, 13(52), 1-15.  Thuiller, W., Midgley, G. F., Hughes, G. O., Bomhard, B., Drew, G., Rutherford, M. C.,  & Woodward, F. (2006). Endemic species and ecosystem sensitivity to climate change in Namibia. Global Change Biology, 12(5), 759-776. !! ! !37! Timmer, V., & Juma, C. (2005). Taking root: Biodiversity conservation and poverty  reduction come together in the tropics. Environment: Science and Policy for Sustainable Development, 47(4), 24-44.  UNICEF. (2011). Namibia. Retrieved from  http://www.unicef.org/infobycountry/namibia_statistics.html !! ! ! 38!Appendix 1. Summary statistics for outcome variables.   In conservancy Outside conservancy Year Variable Best Matching Model Obs Mean Std. Dev. Min Max Obs1 Mean Std. Dev. Min Max Bednet ownership Geographically nearest 400 0.165 0.372 0 1 350 0.191 0.394 0 1 Bednet usage Matched (with covariates) 69 0.377 0.488 0 1 67 0.351 0.481 0 1 Diarrhea prevalence Matched (no covariates) 217 0.166 0.373 0 1 212.8 0.153 0.361 0 1 Diarrhea treatment Matched (no covariates) 36 0.472 0.506 0 1 32.5 0.462 0.506 0 1 School attendance Matched (with covariates) 427 0.761 0.427 0 1 427 0.731 0.444 0 1 2000 Wealth index Matched (with covariates) 289 -0.593 0.694 -1.147 1.706 289 -0.703 0.522 -1.090 1.485 Bednet ownership Geographically nearest 581 0.348 0.477 0 1 589 0.228 0.420 0 1 Bednet usage Matched (with covariates) 202 0.426 0.496 0 1 202 0.356 0.480 0 1 Diarrhea prevalence Matched (no covariates) 284 0.137 0.345 0 1 300 0.153 0.361 0 1 Diarrhea treatment Matched (no covariates) 39 0.641 0.486 0 1 46 0.533 0.504 0 1 School attendance Matched (with covariates) 601 0.792 0.406 0 1 602 0.855 0.353 0 1 2006/07 Wealth index Matched (with covariates) 578 -0.568 0.771 -1.452 1.720 578 -0.668 0.617 -1.452 1.639 1Values reported for matched comparison groups are weighted according to the output from ‘Matching’ in R. !! ! ! 39!Appendix 2. Summary statistics and t-test results used to determine the best matching model for each outcome variable. Bolded values indicate the matching model chosen.    In conservancy Out of conservancy   Mean Variance Mean Variance Mean difference p-value Matched (no covariates) 0.189 0.154 0.109 0.098 -0.080 0.009 Matched (with covariates) 0.189 0.154 0.133 0.116 -0.056 0.079 Nearest 0.165 0.138 0.191 0.155 0.026 0.347 Bednet ownership All 0.165 0.138 0.120 0.105 -0.045 0.018 Matched (no covariates) 0.377 0.238 0.326 0.223 -0.051 0.536 Matched (with covariates) 0.377 0.238 0.351 0.231 -0.026 0.754 Nearest 0.377 0.238 0.441 0.250 0.064 0.447 Bednet usage All 0.377 0.238 0.293 0.207 -0.084 0.170 Matched (no covariates) 0.166 0.139 0.153 0.130 -0.013 0.709 Matched (with covariates) 0.150 0.128 0.133 0.116 -0.017 0.639 Nearest 0.166 0.139 0.109 0.098 -0.057 0.087 Diarrhea prevalence All 0.166 0.139 0.140 0.120 -0.026 0.316 Matched (no covariates) 0.472 0.256 0.462 0.256 -0.011 0.931 Matched (with covariates) 0.500 0.259 0.800 0.167 0.300 0.019 Nearest 0.472 0.256 0.435 0.257 -0.037 0.783 Diarrhea treatment All 0.472 0.256 0.497 0.251 0.024 0.782 Matched (no covariates) 0.761 0.182 0.794 0.164 0.033 0.243 Matched (with covariates) 0.761 0.182 0.731 0.197 -0.030 0.307 Nearest 0.757 0.184 0.811 0.154 0.054 0.025 School attendance All 0.757 0.184 0.835 0.138 0.079 0.000 Matched (no covariates) -0.593 0.481 -0.705 0.283 -0.113 0.029 Matched (with covariates) -0.593 0.481 -0.703 0.272 -0.111 0.031 Nearest -0.642 0.401 -0.482 0.676 0.160 0.002 Wealth index All -0.642 0.401 0.046 1.011 0.688 0.000 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.52966.1-0074568/manifest

Comment

Related Items