<%BANNER%>

Vulnerability, Resource Use, and Market Access in South Africa

Permanent Link: http://ufdc.ufl.edu/UFE0043759/00001

Material Information

Title: Vulnerability, Resource Use, and Market Access in South Africa
Physical Description: 1 online resource (133 p.)
Language: english
Creator: Parent, Gregory David
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: africa -- econometrics -- kruger -- livelihood -- livestock -- management -- markets -- parks -- poverty -- resources -- rural -- vep -- vulnerability
Geography -- Dissertations, Academic -- UF
Genre: Geography thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The dry biomes of southern Africa are home to large numbers of charismatic megafauna. Animal biomass in these systems is limited by the metabolite production of the plants, and these plants are, in turn, limited by water. The natural system, typified by the diverse mix of brows- ers and grazers at varying levels of food selectivity, has been supplanted by the uniformity of ranching and agricultural systems. This has severely altered the dynamic nature of the ecosys- tem that has evolved between vegetation and high herbivore diversity, heavily contributing to desertification, bush encroachment and ultimately a reduction in yields of cattle and crops. Yet there exists few livelihood alternatives to ranching and rain fed agriculture. The lack of viable alternatives to rural households increase livelihood vulnerability as the local system be- comes progressively dryer and unpredictable in terms of rainfall. It is important to understand the interrelationship between covariant shocks and the local economic structure in order to design policy mechanisms that would decrease vulnerability to shocks and maximize benefit to commu- nities from their land while preserving its productivity. Markets, in addition to providing income generating activities and access to productive inputs, supply opportunities for ex ante risk mitiga- tion and ex post shock coping mechanisms. However, most studies to date that have looked at he influence between market access and household welfare or vulnerability, has defined "access" simply in terms of a households spatial relationship to a major market town. This research has two broad aims. The first is to go beyond traditional notions of the con- ceptualizing of market access of a spatial concept toward one that recognizes that households need the capacity to utilize and access the resources that markets provide (such as jobs, financial instruments, and insurance). The second broad goal is to build knowledge around the concept of household vulnerability as it relates to market access and household resource use decisions. Data was collected for this research from 6 communities bordering Kruger National Park in South Africa. A total of 489 households were interviewed from June of 2009 to June 2010. The first research segment used factor analysis to investigate the multi-dimensional quali- ties of market access. Findings suggest that market access is a function of place, social connec- tions, capacity and financial knowledge. The sub-dimensions were then utilized as independent variables in econometric models to see the impact of each sub-dimension on resource use deci- sions. For household decisions on the allocation of consumption towards local natural resources, each sub-dimension had a negative relationship. Hence as a household improved its score in relation to the place, social connection, household capacity, or financial knowledge sub-indices, the household shifts away from local natural resources to market substitutes. A model using the number of cattle a household owns as a dependent variable was also run. However, only an im- provement in household capacity would result in a reduction in cattle ownership, suggesting that market goods that are substitutes for cattle are only accessible via an increase in education and / or exposer to the cash economy. The next major portion of the research uses the above market access index to evaluate how household welfare and vulnerability are influenced by the market access sub-dimensions. The market access sub-components are then allowed to interact with vulnerability to evaluate if gains from market access to household resource use decisions are different between vulnerable and non-vulnerable households. Vulnerability is quantified via the econometric method vulnerability as Expected Poverty (VEP). Results indicate that market access does impact vulnerability, where vulnerability decreases as market access increases. While there is no significant interaction effect between market access and vulnerability, both have impacts on the decision to utilize natural resources. As the data needs of the VEP method goes beyond the capacity of SANPark Social Ecolo- gists in terms of funding and time, a short-form method is established that would allow SAN- Parks to identify vulnerable households. With such a tool, social ecologists will be able to ensure that policies and management plans will not unduly harm vulnerable populations. While several models were attempted, the best result was a model based on robust regression. This short form model agreed with the VEP model up to 78% of the time in the classification of households into groups based on vulnerability levels. Additionally the quantification of this short-form model only requires a maximum of 65 items, much less than the 1,200 items used in the VEP model.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Gregory David Parent.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Child, Brian.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0043759:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043759/00001

Material Information

Title: Vulnerability, Resource Use, and Market Access in South Africa
Physical Description: 1 online resource (133 p.)
Language: english
Creator: Parent, Gregory David
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: africa -- econometrics -- kruger -- livelihood -- livestock -- management -- markets -- parks -- poverty -- resources -- rural -- vep -- vulnerability
Geography -- Dissertations, Academic -- UF
Genre: Geography thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The dry biomes of southern Africa are home to large numbers of charismatic megafauna. Animal biomass in these systems is limited by the metabolite production of the plants, and these plants are, in turn, limited by water. The natural system, typified by the diverse mix of brows- ers and grazers at varying levels of food selectivity, has been supplanted by the uniformity of ranching and agricultural systems. This has severely altered the dynamic nature of the ecosys- tem that has evolved between vegetation and high herbivore diversity, heavily contributing to desertification, bush encroachment and ultimately a reduction in yields of cattle and crops. Yet there exists few livelihood alternatives to ranching and rain fed agriculture. The lack of viable alternatives to rural households increase livelihood vulnerability as the local system be- comes progressively dryer and unpredictable in terms of rainfall. It is important to understand the interrelationship between covariant shocks and the local economic structure in order to design policy mechanisms that would decrease vulnerability to shocks and maximize benefit to commu- nities from their land while preserving its productivity. Markets, in addition to providing income generating activities and access to productive inputs, supply opportunities for ex ante risk mitiga- tion and ex post shock coping mechanisms. However, most studies to date that have looked at he influence between market access and household welfare or vulnerability, has defined "access" simply in terms of a households spatial relationship to a major market town. This research has two broad aims. The first is to go beyond traditional notions of the con- ceptualizing of market access of a spatial concept toward one that recognizes that households need the capacity to utilize and access the resources that markets provide (such as jobs, financial instruments, and insurance). The second broad goal is to build knowledge around the concept of household vulnerability as it relates to market access and household resource use decisions. Data was collected for this research from 6 communities bordering Kruger National Park in South Africa. A total of 489 households were interviewed from June of 2009 to June 2010. The first research segment used factor analysis to investigate the multi-dimensional quali- ties of market access. Findings suggest that market access is a function of place, social connec- tions, capacity and financial knowledge. The sub-dimensions were then utilized as independent variables in econometric models to see the impact of each sub-dimension on resource use deci- sions. For household decisions on the allocation of consumption towards local natural resources, each sub-dimension had a negative relationship. Hence as a household improved its score in relation to the place, social connection, household capacity, or financial knowledge sub-indices, the household shifts away from local natural resources to market substitutes. A model using the number of cattle a household owns as a dependent variable was also run. However, only an im- provement in household capacity would result in a reduction in cattle ownership, suggesting that market goods that are substitutes for cattle are only accessible via an increase in education and / or exposer to the cash economy. The next major portion of the research uses the above market access index to evaluate how household welfare and vulnerability are influenced by the market access sub-dimensions. The market access sub-components are then allowed to interact with vulnerability to evaluate if gains from market access to household resource use decisions are different between vulnerable and non-vulnerable households. Vulnerability is quantified via the econometric method vulnerability as Expected Poverty (VEP). Results indicate that market access does impact vulnerability, where vulnerability decreases as market access increases. While there is no significant interaction effect between market access and vulnerability, both have impacts on the decision to utilize natural resources. As the data needs of the VEP method goes beyond the capacity of SANPark Social Ecolo- gists in terms of funding and time, a short-form method is established that would allow SAN- Parks to identify vulnerable households. With such a tool, social ecologists will be able to ensure that policies and management plans will not unduly harm vulnerable populations. While several models were attempted, the best result was a model based on robust regression. This short form model agreed with the VEP model up to 78% of the time in the classification of households into groups based on vulnerability levels. Additionally the quantification of this short-form model only requires a maximum of 65 items, much less than the 1,200 items used in the VEP model.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Gregory David Parent.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Child, Brian.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0043759:00001


This item has the following downloads:


Full Text

PAGE 1

1 VULNERABILITY, RESOURCE USE, AND MARKET ACCESS IN SOUTH AFRICA By GREGORY DAVID PARENT A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Gregory David Parent

PAGE 3

3 To Lizzy, who has made life so much more vibrant. To my parents, without your amazing support and inspiration, I would not be who I am today. And Maeve, for providing the constant diversion without which I may not have survived the dissertation process. I love you all.

PAGE 4

4 ACKNOWLEDGMENTS First and foremost, thank you to my advisor Dr Child for his continued support over the last several years. Additionally, a big thank you to my committee Dr. Swisher, Dr. Serra, and Dr. Carrion-Florress for all the help and advice throughout the process. This research would not have been possible without the help of Chunky Phiri, invaluable in helping me attain entry to the Mutale area and for the various forms of support throughout the process. Also a big thank you to Louise Swemmer for her help and knowledge of the area throughout the process. RIP Chief Mutale and thank you for allowing and backing me and my research in your communities. Finally a big thank you for may research enumerators. Without the Wildlife Conservation Society, Animal Health for the Environment and De velopment seed grant, none of this would have occurred. Thank you for giving me the chance to work in the area.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .............................................................................................................. 4 LIST OF TABLES .......................................................................................................................... 8 LIST OF FIGURES ...................................................................................................................... 10 LIST OF ABBREVIATIONS ....................................................................................................... 11 ABSTRACT .................................................................................................................................. 12 CHAPTER 1 INTRODUCTION ............................................................................................................... 15 What is Vulnerability? ........................................................................................................ 15 The Move to Vulnerability in Geography ............................................................................ 15 Vulnerability in Geography .................................................................................................. 16 Vulnerability from an Economic Perspective ...................................................................... 17 Research Goals ..................................................................................................................... 18 2 MARKET ACCESS, POVERTY, AND NATURAL RESOURCES IN SOUTH AFRICA: UNDERSTANDING THE MULTI-DIMENSIONALITY OF MARKET ACCESS AND ITS INFLUENCE ON RESOURCE USE DECISIONS. .......................... 20 Setting the Scene .................................................................................................................. 20 Objectives ....................................................................................................................... 22 Study Site ........................................................................................................................ 23 Methods ................................................................................................................................ 25 Data Collection ............................................................................................................... 25 Sampling ........................................................................................................................ 25 Data Analysis .................................................................................................................. 26 Factor analysis ........................................................................................................ 26 Index construction .................................................................................................. 29 Regression Modeling .............................................................................................. 29 Results .................................................................................................................................. 31 Market Access Index ...................................................................................................... 31 Validation with Consumption ......................................................................................... 33 Market Access and Resource Use ................................................................................... 34 Discussion ............................................................................................................................ 38

PAGE 6

6 3 TAKING THE UN OUT OF THE KNOWN UNKNOWNS OF HOUSEHOLD RESOURCE USE DECISIONS IN AFRICA: THE INFLUENCE BETWEEN RISK, MARKET ACCESS AND NATURAL RESOURCE EXTRACTION IN SOUTH AFRICA. .............................................................................................................................. 48 Prelude ................................................................................................................................. 48 Objectives ....................................................................................................................... 51 Study Area ...................................................................................................................... 51 Methodology ........................................................................................................................ 53 Questionnaire Design ..................................................................................................... 53 Sampling ........................................................................................................................ 54 Data Analysis .................................................................................................................. 55 Market access index ................................................................................................ 55 Vulnerability estimation ......................................................................................... 56 Results .................................................................................................................................. 58 Market Access ................................................................................................................. 58 Vulnerability ................................................................................................................... 59 Vulnerability and Market Access .................................................................................... 61 Interactions Between Vulnerability and Resource Use .................................................. 63 Discussion ............................................................................................................................ 65 4 CONSTRUCTION OF A SHORT-FORM VULNERABILITY MODEL FOR RAPID ASSESSMENT OF COMMUNITIES IN SOUTH AFRICA .............................................. 77 Why There is a Need ............................................................................................................ 77 Objectives ....................................................................................................................... 81 Study Site ........................................................................................................................ 81 Methodology ........................................................................................................................ 83 Data Collection ............................................................................................................... 83 Sampling ......................................................................................................................... 84 Data Analysis .................................................................................................................. 84 Factor analysis ........................................................................................................ 86 Regression modeling .............................................................................................. 87 The VEP model ....................................................................................................... 87 Results .................................................................................................................................. 88 Selection of model variables ........................................................................................... 88 Factor Analysis SFVP Model ......................................................................................... 90 Regression Based Models without Factor Analysis ....................................................... 93 Discussion ....................................................................................................................... 95 5 CONCLUSION .................................................................................................................. 106 APPENDIX A SURVEY INSTRUMENT ................................................................................................. 108

PAGE 7

7 B BASE REGRESSION FOR VULNERABILITY ANALYSIS .......................................... 121 C DESCRIPTIVE STATISTICS OF VARIABLES USED IN PROBIT MODELS ............. 122 LIST OF REFERENCES ............................................................................................................ 123 BIOGRAPHICAL SKETCH ...................................................................................................... 132

PAGE 8

8 LIST OF TABLES Table page 2-1 Household utilization of natural resources by study area .................................................... 43 2-2 Model Comparison Fit Results ............................................................................................. 43 2-3 Market access index construction, factor analysis ............................................................... 44 2-4 Correlation between total household consumption and the MAI index and sub-indices ..... 44 2-5 MAI and sub-dimensions by poverty status ......................................................................... 44 2-6 Descriptive of variables used in the models ......................................................................... 45 2-7 Regression Modeling Results (Dependent variable, Natural Resource Use Ratio) ............. 46 2-8 Regression Modeling Results (Dependent variable, Number of Cattle Owned) ................. 47 3-1 Market access index construction, factor analysis ............................................................... 69 3-2 MAI by village ..................................................................................................................... 70 3-3 Vulnerability within and between selected groups ............................................................... 71 3-4 Correlation between total household consumption, vulnerability, resource use, and the MAI index and sub-indices .................................................................................................. 72 3-5 Impact of market access and vulnerability on household natural resource use decisions .... 73 3-6 Impact of market access and vulnerability on household cattle ownership ......................... 75 4-1 Household utilization of natural resources by study area .................................................... 98 4-2 Factor analysis ...................................................................................................................... 98 4-3 Variables linked to vulnerability .......................................................................................... 99 4-4 Regression models for determining factor weights ............................................................ 101 4-5 Regression-only based models ........................................................................................... 103 4-6 List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data ................................................................................... 105

PAGE 9

9 B-1 Base regression for vulnerability analysis .......................................................................... 121 C-1 Descriptive Statistics of Variables Used in Probit Models ................................................. 122

PAGE 10

10 LIST OF FIGURES Figure page 2-1 Map of Study Communities ................................................................................................. 42 3-1 Predicted probabilities of local natural resource consumption by vulnerability ............................................................................... 74 MAI sub-dimensions ............................................................................................................ 76 4-1 Level of agreement between VEP and SFVP by 50th & 75th percentile groups and .50 level of vulnerability for factor based models .................................................................... 102 4-2 Level of agreement between VEP and SRVP by 50th & 75th percentile groups and .50 level of vulnerability .......................................................................................................... 104

PAGE 11

11 LIST OF ABBREVIATIONS FA Factor Analysis GLM Generalized Linear Model KNP Kruger National Park MAI Market Access Index NRR Natural Reource Use Ratio OLS Ordinary Least Square PCA Principal Component Analysis QLM Quasi Likelihood Method RSA Republic of South Africa SPI Place & Infrastructure Sub-Index SSC Social Connection Sub-Index SHC Household Capacity Sub-Index SFKA Financial Knowledge & Access Sub-Index VEP Vulnerability to Expected Poverty

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School Requirements for the Degree of Doctor of Philosophy VULNERABILITY, RESOURCE USE, AND MARKET ACCESS IN SOUTH AFRICA By Gregory David Parent May 2012 Chair:Brian Child Major: Geography The dry biomes of southern Africa are home to large numbers of charismatic megafauna. Animal biomass in these systems is limited by the metabolite production of the plants, and these ers and grazers at varying levels of food selectivity, has been supplanted by the uniformity of ranching and agricultural systems. This has severely altered the dynamic nature of the ecosys tem that has evolved between vegetation and high herbivore diversity, heavily contributing to Yet there exists few livelihood alternatives to ranching and rain fed agriculture. The lack of viable alternatives to rural households increase livelihood vulnerability as the local system be comes progressively dryer and unpredictable in terms of rainfall. It is important to understand the interrelationship between covariant shocks and the local economic structure in order to design nities from their land while preserving its productivity. Markets, in addition to providing income generating activities and access to productive inputs, supply opportunities for ex ante risk mitiga tion and ex post shock coping mechanisms. However, most studies to date that have looked at

PAGE 13

13 simply in terms of a households spatial relationship to a major market town. ceptualizing of market access of a spatial concept toward one that recognizes that households instruments, and insurance). The second broad goal is to build knowledge around the concept of household vulnerability as it relates to market access and household resource use decisions. Data was collected for this research from 6 communities bordering Kruger National Park in South Africa. A total of 489 households were interviewed from June of 2009 to June 2010. ties of market access. Findings suggest that market access is a function of place, social connec variables in econometric models to see the impact of each sub-dimension on resource use deci sions. For household decisions on the allocation of consumption towards local natural resources, each sub-dimension had a negative relationship. Hence as a household improved its score in the household shifts away from local natural resources to market substitutes. A model using the number of cattle a household owns as a dependent variable was also run. However, only an im provement in household capacity would result in a reduction in cattle ownership, suggesting that market goods that are substitutes for cattle are only accessible via an increase in education and / or exposer to the cash economy. The next major portion of the research uses the above market access index to evaluate how

PAGE 14

14 market access sub-components are then allowed to interact with vulnerability to evaluate if gains from market access to household resource use decisions are different between vulnerable and as Expected Poverty (VEP). Results indicate that market access does impact vulnerability, where between market access and vulnerability, both have impacts on the decision to utilize natural resources. As the data needs of the VEP method goes beyond the capacity of SANPark Social Ecolo gists in terms of funding and time, a short-form method is established that would allow SAN Parks to identify vulnerable households. With such a tool, social ecologists will be able to ensure that policies and management plans will not unduly harm vulnerable populations. While several models were attempted, the best result was a model based on robust regression. This short form only requires a maximum of 65 items, much less than the 1,200 items used in the VEP model.

PAGE 15

15 CHAPTER 1 INTRODUCTION What is Vulnerability? Weichselgartner (2001) stems from differing epistemological orientations. Three themes have emerged from the research that utilizes differing perspectives: vulnerability as exposure to risks; vulnerability as it relates to place; and vulnerability as social response (Cutter 1996). To a great extent, economics tends to approach vulnerability from the exposer to risk perspective, while raphy remains focused on the approach to venerability as it relates to place. However, there is a move towards the incorporation of the perspective of vulnerability as a habitat created by histori cal, cultural, and economic processes that affect individuals and societys capacity to cope to shocks (Kelly & Adger 2000; Cutter et al. 2003). As this research deals with answering questions in the context of vulnerability, it is important to understand what is meant by vulnerability, and how conceptualizations differ between geography and poverty economics as this research ap proaches vulnerability from the economic perspective. The Move to Vulnerability in Geography The origin of vulnerability as it exists in geography can be found in hazards research. In the early 1920s Borrows began an investigation into how people and society adjusts to environmen tal shocks (Kates & Burton 1986). This work focused on extreme events and the ex post changes that society makes to mitigate shock event impacts, with the emphasis on the applied aspects to

PAGE 16

16 and distribution of hazards; the adjustments available to individuals and society; and the percep tions and choices people make in regard to hazards events (Cutter et. al 2003). The critique began to emerge that hazard research was very applied with little theoretical underpinning (Lindell 1997) along with its focus on extreme hazards and the lack of international applications, (Cutter et. al 2003), change was initiated to alter course of the direction of hazards research. At about the same time hazards researchers started to turn towards anthropological work on human environmental relations that felt natural hazards were as much a cultural and social construction as they were a physical process (Blaikie et al. 1994; Cutter et. al 2003). This understanding that hazards are not only physical events, but have an element of social manifes tation began to be integrated into the research. Hazard research now recognizes the inherently complex interaction between physical and social phenomenon. Vulnerability in Geography Cutter takes to vulnerability can be thought of as the current dominating discourse in Geography. In this approach, vulnerability is examined in its causal structure, spatial variability and meth ods for reduction (Cutter et al. 2003). As stated prior, in other areas of research vulnerability is treated as a pre-existing condition in which a hazard takes place. Blaikie et. al (1994) and others have promoted a concept of social vulnerability in which vulnerability is dependent not on ge ography, but an underlying social condition of the system. Hazards are a given, while the social system gives rise to the condition that could result in losses of life or livelihood. in other words, vulnerability as potential exposer or social resilience. Geography, according to Cutter (2000), has approached the combination of physical, social and spatial forms by studying the dynamics of

PAGE 17

17 vulnerability in various place settings. The model that this gives rise to is called the hazards of place model. The model explains vulnerability in terms of its distributive hazard patterns and the underlying (social) processes that produce the event or shock. (Cutter 1996). But where does this lead us? The concept of vulnerability in Geography is still very much tied to place, and is often explain in terms of the spatial relationship of infrastructure (physical and social) that help reduce the impact of an event. The models that are used to ascertain vener ability mainly utilizes location in relation to social and physical infrastructure as main determi nant of vulnerability (Cutter et. al 2003). The terms used in this literature underpin this notion of spatiality as the primary driver, with terms such as riskscapes and hazardscapes bandied about. focused almost entirely on developed countries. Data is attained through census and extraction via GIS. Given the aggregation at the regional scale, it assumes homogeneity within the system, as such it will not capture differences within the counties and lose explanatory resolution. Vulnerability from an Economic Perspective Within the economic literature, location is viewed more as an element of poverty, in the sense that marginal areas have higher marginal costs of access (Adger 1999), and hence, alter the household production consumption system. While the origins of vulnerability in Geogra phy are located in hazards research, the origins are distinctly different in Economics. Cutter et al. (2000) claims that social vulnerability largely deals with, at the lowest scale, social groups, while individual (household) vulnerability is the purview of health and engineering disciplines. However, what is missing from Cutters discourse, and the discussion at large in Geography, are the insights that economics has made. Economic vulnerability has largely emerged from poverty

PAGE 18

18 as the development of household capabilities to utilize and access resources to be leveraged into growth. These capacities include health, education, and access to resources and markets. Economic vulnerability is the probability that a risk event would reduce the well-being of households or individuals, resulting in a state of poverty and destitution (Dercon 2005a). Es sentially, vulnerability alters perceived future utility in households exposed to uninsured risks, thus affecting the decision-making process (Dercon et al. 2005). Poverty cannot be viewed as being the sole causal effect of a households lack of income and/or factor endowments, but also a result of the decisions that households make in response to being exposed to possible risks (Sen, 1981; Dercon et al. 2005). The poor lack the necessary formal mechanisms that reduce risk, as such they respond to risk in ways such as diversifying income or choosing low-risk / low-reward income/consumption strategies (Kanbu & Squire 2001). Research has illustrated that households often choose lower levels of welfare (often in the form of production decisions) as a risk-mitigat ing response in the state of stress (Dercon et al. 2005). While many interpretation of vulnerability exist, the advantage in the of the economic approach is the ability to capture the heterogeneous nature of communities and shed light on how household decisions alters the likelihood of a shock Research Goals cess and how market access interacts with household resource use decisions. Is market access a multi-dimensional concept, where access is a function of place, social connection and household capacity? What impact does market access have on household resource use decisions? Does market access impact household natural resource consumption differently than cattle ownership decisions?

PAGE 19

19 related to market access and interaction with household resource use decisions. What households tend to be more vulnerable to poverty? Does market access reduce vulnerability and how does this differ amongst the various mar ket access dimensions? Does vulnerability interact with market access whereas additional gains to resource use decisions from market access are achievable? Can a short-form vulnerability method be established that is both accurate in classifying vulnerable households, while minimizing items necessary to quantify vulnerability?

PAGE 20

20 CHAPTER 2 MARKET ACCESS, POVERTY, AND NATURAL RESOURCES IN SOUTH AFRICA: UNDERSTANDING THE MULTI-DIMENSIONALITY OF MARKET ACCESS AND ITS INFLUENCE ON RESOURCE USE DECISIONS. Setting the Scene Improving market access has been found to increase household welfare (Jacoby 2000; Der con & Hoddinott 2005; Jacoby & Minten 2009; Mogues 2011) and this connection is at the heart of promoting road construction as key development initiative for poverty reduction (Gibson & Rozelle 2003; World Bank 2007; Jacoby & Minten 2009). Improving market access and house hold welfare will undoubtedly alter household behavior. Ultimately, increased welfare means an increase in consumption and natural resource use. This could be troubling for organizations con cerned with the conservation of natural resources, especially organizations tasked with the pro tection of fragile and high-valued ecosystems such as national parks throughout Southern Africa. Rural households are the primary agents of environmental change in the very area where parks and protected areas are attempting to conserve ecosystem diversity (Angelsen & Kaimowitz 1999; Gbetnekom 2005). In fact there have been several studies that have linked improved mar ket access market with increased local resource use. Reduced transport costs and greater access to productive inputs can lead to increased forest clearing, agricultural expansion, and increased extraction of natural resources (Rheardon & Vosti 1995; Nelson & Hellerstein 1997; Jacoby 2000; Gbtnekom 2005). But these studies classify market access as a spatial or infrastructure dependent concept. This assumes that all households at a given distance will be able to access which a household can travel to a market town. Market goods include more than just household food and non-food consumption inputs, but also a market to sell goods and, more importantly, la

PAGE 21

21 bor, in addition to accessing credit and insurance products. A household needs more than simply good roads or good transportation to access such goods. Drawing from Sens (1984) entitlement framework, households need the capacity to utilize and access these market resources. How a household responds to market access in regards to resource use behavior is linked tions that markets provide? First, markets provide crucial inputs for household production and are a main source of household consumables. Dercon & Hoddinott (2005) found that households in Ethiopia purchased more than half of consumables in market towns. In addition to cheaper productive inputs, households have greater access new technology (Xu et. al 2006). Having greater access to goods and assets from market towns could cause a shift away from local natural resources if the goods from a market town serve a similar purpose in livelihood structures (Rut tan 1992). Secondly, markets provide an outlet for goods and services produced in a household. These products can be natural resource based goods, like rubber in Indonesia (Miyamoto 2006) resource extraction do have negative consequences on the ecosystem, markets may provide other opportunities that remove the household from direct extraction. One such opportunity is access to jobs. Households may be able to re-allocate un-used or under-used human resources towards higher value added activities rather than subsistence agriculture and natural resource gathering activities (Dercon & Hoddinott 2005). Finally, and linked to the previous two points, markets provide opportunities for ex ante risk and ex post coping mechanisms to shocks (Christiaensen & Subbarao 2005). Much of a households behavior in developing countries in relation to natural resource use is conditioned on

PAGE 22

22 its livelihood structure. Where risk is uninsured, households diversify to spread risk (Rheardon et mechanism if a household attains access to services that function in the same way that diversify ing into natural resources do (Rheardon & Vosti 1995). tance from the household to a main market town (Christiaensen & Subbarao 2005; Dercon & Hoddinott 2005). As households have varying human and physical endowments that condition any access they have to a market, access is more than just a spatial relationship. While a reduc tion in travel time is important in helping a household in substituting certain consumables, such as food items and fuel, it could possibly have less impact in helping a household improve access to employment or other insurance goods that may be key in reducing reliance on natural resourc es (Mogues 2011). Bluffstone (1995) found, in modeling deforestation behavior in Nepal, that off-farm labor opportunities are key in limiting deforestation. Attaining access to employment requires greater skill, suggesting education. Given the role the patronage plays in Africa, hav ing family members in market towns may aid and improve the likelihood of accessing or selling certain goods. Stark (1991) found that remittances can substitute for missing credit markets in rural areas. Rheardon and Vosti (1995) found that where livestock is the only insurance available, as household welfare rises, so to does the demand for insurance, leading to further ecosystem degradation by livestock. Objectives concept, to one that that has multiple dimensions, where market access is a function of space and

PAGE 23

23 infrastructure, social capital, and household capacity. A households education, available labor, determining its access, and hence the goods and services it can derive from a market. Addition ally this study hypothesizes that a household will utilize fewer natural resources as market access increases, and that household capacity, as one dimension of market access, will have the great est impact on a households decision to allocate consumption to local resource use. cess using factor analysis and the extent to which the index can serve as a stand-alone develop ment indicator. Next it uses the aggregated and dis-aggregated form of the index as independent households decision to allocate consumption away from local natural resource products (such as crops, wildlife and livestock products, and collected wild natural resource goods) and (2) a households decision to keep cattle. Study Site The study was conducted in 6 communities bordering Kruger National Park (KNP) in Province, east of the Veterinary fence, were selected,while one community, Makoko, was select ed in the south. Makoko is also directly adjacent to KNP, located in the Mbombela Local Mu nicipality, Mpumalanga Province. Given the time and funding constraints, surveying at intervals along the entire length of KNP was deemed to be unfeasible. After consultation with KNP social scientists, these two areas were chosen as they represent the two extremes in terms of social and ecological conditions among communities bordering KNP (Swemmer 2009).

PAGE 24

24 In the north, the 5 communities studied were Bende Mutale (household population of 174, sample of 89), Tshikuyu (HH pop 71, sample 51), Duluthulu (HH pop 115, sample 71), Beleni (HH pop 23, sample 20), and Mutale B (HH pop 187, sample 92). In addition to the close proximity to one another, each village is dominated by the Vende ethnic group, with each con trolled by the same Traditional Authority. The southern community on the other hand is domi nated by the Swati ethnic group and is distinctly larger at about 1,100 households. Table 2-1 breaks down each region for various key characteristics in terms of resource use activities. In looking at household participation in various activities, the northern area has more diversity in the overall livelihood structure, while much of the village production in the south is concentrated in a smaller amount of activities. Given dryer conditions in the north where long-term mean annual rainfall is 400-500mm (Thomas et al. 2007), it is not surprising to see a greater diversity in agriculture and resource use activity as diversifying livelihood activities is a risk mitigation strategy (Francis 2002). Conversely, the mean annual rainfall around Makoko is 840-1670mm (Louw & Scholes 2006). In a similar vain, the greater reliance on livestock, espe cially cattle and goats, is also partly explained by the dryness of the north as livestock is more robust to dry conditions than cropping. The northern communities are much more remote than Makoko. The nearest market town, Thohoyandou (pop. 39,513), where most households travel to once a month for purchases of food and non-food consumables, is about a 100 km journey mainly over very harsh roads. Makoko, on the other hand, is within 30 km of two sizable towns, Hazyview (pop. 35,948) and White River (pop. 18,433), and within 45 km of the provincial capital of Nelspruit (pop. 94,714). The northern communities rely largely on minibus taxis, waiting an average of 88 minutes to be

PAGE 25

25 gin their journey and paying a mean fare of R41. Makoko, in addition to good roads, have fairly regular buses, with an average wait time of 18 minutes and a fee of R17. Methods Data Collection Data for this research was collected through household surveys. Preceding the data gath ering process, the researcher interviewed key informants and community members to establish the inclusion of relevant variables. After attaining regional and local permission, and training local enumerators, the questionnaire was piloted in 26 households which were subsequently not vocabulary and variable inclusion/exclusion to ensure data quality and participant interpretation. household demographics; water, market and health services access; detailed income, which looked at both the household production of products and formal and non-formal employment; comprehensive consumption, including questions on household food, non-food, and durable good consumption; and shocks and coping strategies. Sampling northern villages, the researcher either formed the sampling frame from preexisting village lists formants within each of the villages. Due to information constraints, a village list was not avail able for the southern village of Makoko, nor was it feasible to construct one because of time and

PAGE 26

26 size constraints. Unlike the northern villages, Makoko has an estimated household population of 1,100 spread throughout 4 blocks. A frame was established by modifying a vegetation transect the sample by blocks. The streets in each block were assigned a random number and then placed would be conducted. Interviews continued until reaching the end of that particular street edge. Upon the conclusion of a street edge, the next highest ordered street was selected. This contin ued until the sample quota was reached for each block. In total 489 households were interviewed from June 2009 to June 2010, from which 323 were conducted in the north with 166 from the south. Data Analysis Factor analysis The building of the market access index was achieved through factor analysis techniques. The goal is to reveal the internal structure of the dataset in order to disclose unique dimensions inherent in the indicators. While many studies utilize principal component analysis (PCA) for the Osborne 2005). Factor analysis is highly sensitive to sample size. Small samples lead to errors in the 1991). The size of the sample (cases) relates to the number of variables that one is attempting to

PAGE 27

27 factor. While there is no agreed upon case to variable ratio, several studies have attempted to as sertion an adequate ratio. Grossman et al. (1991) suggested that the case to variable ratio should the case to variables ratio should be 5 to 1. Hutcheson & Sofroniou (1999) suggested a case range from 150 to 300, where 300 cases would be optimal when highly correlated variables are few. Costello & Osborne (2005) found that errors were minimized with a 20 to 1 case to variable ratio, but that larger ratios would reduce errors further. In essence one needs a large sample to minimize errors in factor analysis. This study performed FA on 10 variables, with 489 cases to draw from, the case to variable ratio is about 49:1, easily surpassing the most stringent sampling There are many options in FA for the extraction of factors. When data is normally distrib 1996; Fabrigar et al., 1999). However, this is not always possible when dealing with social data. When dealing with data that violates the assumption of normality, as is the case with this study, the principal axis factors method is the preferred method (Fabrigar et al. 1999). The literature highlights several strategies for the extraction of factors that ultimately serve as the dimensions in the formation of multidimensional indexes. The most commonly used method is the Kaiser criterion in which all factors with an eigenvalue below 1.0 are dropped. Other potential extraction methods are parallel analysis (PA) and the scree test. While the Kaiser criterion is an attractive choice due to its simplicity of use and interpretation, it has been found to be the least reliable of the aforementioned methods (Hakstian et al., 1982; Zwick & Velicer, 1982), often resulting the the extraction of too many factors or components (Browne, 1968; Linn, 1968; Zwick & Velicer, 1986). Additionally, Jackson (1993) found that the scree plot approach

PAGE 28

28 consistently estimated one dimension too many. Parallel analysis has been found to be far more accurate (Zwick & Velicer 1986). In fact Velicer & Jackson (1990) concluded that the choice of how many factors to retain has far more impact on the conclusions drawn from studies utilizing FA than the selection of the extraction method. Hence, retaining the correct number of factors is highly important on both theoretical and empirical grounds. As such, this study uses parallel analysis in the decision of how many factors to retain. For the purpose of this study, four factors were retained, namely (1) place & infrastructure, (2) social connection, (3) household capacity, Once a decision is made on how many factors to retain, the next step is rotation. Rota tion is undertaken to improve the interpretability and the simplicity of the data. The two methods available are orthogonal and oblique rotation. With orthogonal rotation factors are produced that are uncorrelated, while oblique rotation permit the factors to correlate. While the orthogonal method is often used in the creation of indices, this study utilizes oblique rotation as social be havior is not commonly bound within phenomenon that function independently, but rather have some degree of correlation (Costello & Osborne 2005). If the factors are largely uncorrelated, the two methods should return the same results. Once the factors are rotated, the pattern matrix gives the rotated factor loadings. The loadings allow the researcher to identify what variables load on each factor. The variables that load above a predetermined cutoff point can be used to build that particular dimension or subindex. This study uses a cut-off of 0.32 for the selection of variables in the sub-indexes as sug gested by Tabachnick & Fidell (2001).

PAGE 29

29 Index construction sis, the sub-index (each dimension) and aggregated index must be constructed. At issue here is the weighting scheme of the individual variables and the aggregation method of the sub-indexes to form the market access index (MAI). Two issues need to be considered for making weight ing and aggregation decisions in the construction of an index, any method needs to minimize eclipsing (underestimating) and ambiguity (overestimating) (Swamee & Tygi 2000). Garriga and Foguet (2010) found that issues of eclipsing and ambiguity were minimized with equal weight ing of the variables for the construction of the sub-indexes, weighing each sub-index using the proportion of variance it explains in the data set, and using a geometric mean aggregation proce dure to form the main index: (1) where I is the aggregated index, N is the number of sub-indexes, s is the ith sub-index, and w is the weight applied to the ith sub-index. Additionally, by going to a geometric aggregation proce performance in one sub-index cannot be made up for in another. Regression Modeling To evaluate the impact that market access has on resource use, the MAI in its aggregated and disaggregated form were entered as independent variables in regression models. Two dif area to total consumption at the household level (NRR). Local natural resources consumed includes all agricultural products (this includes animal products) produced and natural resource

PAGE 30

30 goods collected by the household for self consumption, in addition to all agriculture and natural resource goods the household purchased that were grown from others in the area. Again, the goal is to see if, with other factors held constant, improved market access results in an allocation away from local natural resource consumption. Using a ratio as the dependent variable poses issues with standard OLS regression. At issue is the bounded nature of proportion data between 0 and 1. With bounded variables, the conditional variance approaches zero as the conditional mean draws closer to either 0 or 1 resulting in error distributions that are heteroskedastic. Papke and Wooldridge (1996) suggest using a robust quasi-likelihood method (QLM) where the mean is allowed to be a non-linear function of the regressors and the variance becomes a function of the mean. In practical terms, OLS regression with proportion data can result in biased parameters leading possibly to perverse policy recommendations, while models based on the QLM protect against this issue. The QLM method is achieved by using a generalized linear model (GLM) with a binomial family and a logit link, which Papke and Wooldridge (1996) found to be robust While animal products consumed are contained within the dependent variable above, households also own a substantial amount of cattle that serve often as a form of savings and in surance against risk. However, as stated above, cattle herds are another form of resource utiliza tion by the household. To look at whether or not the MAI alters the decision of a household to maintain herds, a regression model was used to regress heads of cattle owned against the MAI index and household characteristics (CATTLE). As the dependent variable was, again, highly non-normal with much of the data clustered around and truncated at zero, problems could arise

PAGE 31

31 and a log link. This is supported by the AIC statistic in Table 2-2. Results Market Access Index Table 2-3 displays the results from the factor analysis. While parallel analysis suggested hence it was decided to maintain four factors. While three dimensions were expected (place & infrastructure, social connections and capacity), the results from FA revealed four dimensions of were related to the spatial relationship of the household to the market or the quality of transpor tation infrastructure (market distance, transportation wait time, & roundtrip fare). The fourth variable that loaded captures the quality of the market (population divided by distance). Given structure sub-index (SPI). To capture the importance of social capital, two variables were included in assessing a which a family has a relative. A relative in another area may aid the household in accessing the resources in the location. Having a relative not only reduces the costs in visiting, but can also household goods and services. However, it is not only a function of the quantity of relatives a household has living in other areas, but also the nature of the connection. Similar to the market

PAGE 32

32 quality variable in SPI, the size of the town has a bearing on what is available to a household. For example, Johannesburg has both many more opportunities, when it comes to labor, and a greater diversity of household inputs. This quality feature can be degraded as distance increases. familiar and increasing dominated by different ethnic groups. As such, the cultural and language difference is likely to degrade the quality of a potential location as the differences increase. In a continent where patronage is a crucial factor, this will decrease social capital. Both of these vari ables loaded on the second dimension, or the Social Capital (SSC) sub-index. Household capacity was hypothesized to form the third dimension. The statistical analy sis suggested that the variables thought to comprise the third dimension be separated into two dimensions. Years of schooling and the ratio of cash to total household income loaded on factor three. Educational level is key in allowing a household to access market goods such as credit, insurance and employment. The second variable that loaded was an attempt to capture prior experience and exposer to the cash economy. The more a household is integrated into the cash economy, the more it may be able to access various services and goods. Factor three was identi Factor 4 appears similar to factor 3. The two variables loading above the 0.32 cutoff were hold. While questions were asked as to the total amount a household has in savings accounts, interest payment and outstanding loan principal, in addition to any insurance payments, many households were not comfortable with sharing this information. The survey was adjusted during

PAGE 33

33 possesses. This included all savings accounts, insurances, and loans. It is important to recognize included after consultation with key informants. Males have a greater potential to be hired for piecework (informal labor) and permanent jobs, especially jobs involving manual labor, which are the very type of jobs many of the households are restricted to given the lack of education. Validation with Consumption In addition to using the MAI and the sub-indices to evaluate whether or not increased market access reduces resource use, the secondary goal is to construct a stand-alone indicator to capture a concept of development that has yet to be measured. To evaluate whether or not household consumption and the MAI. Table 2-4 displays the results. The aggregate index, MAI, has a moderate strength of association with consumption in the direction one would expect; as market access increases, so to does household consumption. Additionally this association is SPI has the weakest association at rs = 0.12, while SFKA has the strongest association at rs = 0.40, a moderate strength of association. SSC at rs = 0.35 also illustrates a moderate level of as sociation with consumption, while SHC at rs = 0.21, is larger than SPI, but still small. As stated market access as a spatial relationship between the household and the nearest town (as captured in SPI), but this appears to only capture a small element of the relationship between access and

PAGE 34

34 welfare. The other sub-indices, and the aggregate, all illustrate higher strengths of association and highlight the multi-dimensionality of market access. Table 2-5 displays the breakdown of the MAI and its sub-indices by poverty status. The mean score for each of the sub-indices was lower for households in poverty, as is the mean of the composite index. Households in both groups scored the highest in the SPI dimension with a mean index score of 51.6 and 54.4 for households in and out of poverty respectively. The lowest scoring dimension for each group was SSC with the mean of impoverished households at 20.3 and at 24.6 for households not in poverty. The mean score in the composite index was 30.9 for households not in poverty, while impoverished households have a mean score of 36.5. Addition another with the surprise exception of SPI. The overall connection between welfare and market access illustrated above, and the link between poverty and market access in Table 2-5, reinforce the potential usefulness of the MAI as another tool in identifying poverty dynamics. Market Access and Resource Use Table 2-6 gives the descriptive statistics and describes each of the dependent and indepen dent variables used in the regression models. Two models were run for each dependent variable. disaggregated model with each of the sub-indices included as an independent variable. The other independent variables were used to control for other factors. While each variable is described in Table 2-6, a few bare further explanation. NORTH is a dummy variable to control for unobserved factors that differ by region. LIVILHD_DIVERS is a variable capturing the various products and ratio of local natural resources to total consumption to decrease a wealth increase due to a shift to

PAGE 35

35 higher quality substitutes. Households assets (ASSETS) was used a proxy for wealth to control for the effect that wealth has on resource use decisions. and standard errors, the marginal effects are given for each model run. The marginal effects are calculated at the sample means of the independent variables. Note that for the dummy variables, the marginal impact on the dependent variable from a change in the independent is for a dis This means that as market access increases, a household will allocate less of its consumption to local natural resources, all other factors held constant. Seven of the remaining nine inde Age of household head, household size, and the household dependent ratio are all statistically result in a household allocating less of its consumption towards natural resources, and contrary to what one would expect. One issue at play here, is the existence of governments grants. As of 2010, each household received R1050 per month for each individual who was older than 60, otherwise known as the Old Persons Grant. The added money for people over the age of 60 may suggest why the age of the household head causes a reduction in local resources use. Similarly, South Africa has what is called the Child Support Grant where a person can get R250 per month for each child they are responsible for who are younger than 18, up to a total of 6 children. The excess cash and increase in the dependency ratio is likely to cause less allocation of consumption

PAGE 36

36 north, all other factors held constant, consume a smaller proportion of local natural resources. The more diverse a households livelihood structure is, the more it allocates consumption towards The model incorporating the disaggregated MAI index displays similar results in re spect to the control variables. The only variables that illustrates a change is the dummy vari ables indicating whether or not a household is a member of the dominant local ethnic group (DOM_ETHN_GRP) and whether a household has at least one member permanently and locally cant in the disaggregated model. NRR. As such, as each sub-dimension of market access increases, a household will allocate less marginal effects can be utilized to see which dimension has a greater marginal impact on NRR. SHC has the greatest marginal impact on NRR at -0.0023, while SSC has the smallest marginal policy aimed at improving household capacity would have the greatest impact at reducing natural at the bottom of Table 2-7. While the aggregate MAI model has a lower AIC value, suggesting data. However, the added information that the disaggregated model provides by allowing the

PAGE 37

37 researcher to see how the MAI sub-dimensions impacts the decision to allocated consumption to local natural resources is preferred. The second set of models (Table 2-8) use number of cattle owned as the dependent vari able. While cattle products consumed are integrated into the NRR variable, cattle are mainly viewed as an asset by households and many households own a number of heads of cattle often improved market access would reduce the number of cattle a household owns is not supported appear that while improved market access provides suitable substitute goods in terms of natural for insurance products, again, the role cattle tends to play in rural African communities. The Disaggregated CATTLE model performs similar to the aggregated CATTLE model. While the AIC score is lower for the disaggregated model, the slight difference indicates each ciated with CATTLE, suggesting that people with secure jobs invest a portion of their income in

PAGE 38

38 a negative relationship with CATTLE. Again, when it comes to altering a households decision whether or not to own cattle, household capacity appears to have the greatest potential. Discussion As Dercon and Hoddinott (2005) point out, there are numerous links between rural areas and market towns. Households do not operate in a separate space, but are fully integrated into local and regional markets for the purchase and sale of goods and services, and to access risk mitigation mechanisms. However, access must be be thought of as being more than a spatial rela tionship to a market town. While improved infrastructure undoubtedly conveys improved access, households, given their individual capacity, may still be left disconnected from markets. Social ments, all potentially help households leverage market goods into productive gains. This study has suggested that market access is a function of space, social connections, and household capacity. Factor analysis suggested that MAI is comprised of four dimensions where the last 2 dimensions deal with household capacity and knowledge of markets. Interest ingly, when looking at how the individual dimensions correlate with total household consump tion, the sub-dimension that deals with infrastructure and distance, SPI, has the weakest correla tion with consumption. The small association between SPI and consumption and the lack of a factors that allow a household to access market goods. Other studies have suggested that improved access to markets, increases resource uti

PAGE 39

39 a smaller proportion of total consumption, suggesting that markets allow households to substi tute local resources with external goods and services. This applies to four of the dimensions by on household resource use behavior is household capacity, policies focusing on increasing the education level and experience with the cash economy, would result in the largest shift away from local natural resources. Theoretically, exchange (for which markets are required) creates wealth. This study economic and institutional factors, such as education, connectivity through family members, use one would expect that increased market access and capacity allows rural households to integrate themselves more into the urban / global economy, thereby reducing dependence on low-value, unreliable natural products. However, poorer, less connected households rely disproportionately on natural resources. Cattle play an important role in the livelihood structure in Southern Africa. They are viewed as an asset to be accumulated for the occasional sale or consumption, often as a risk detrimental impact on the local ecosystem. While the aggregated MAI index does not have a does not consider quality. The loans that most rural households have access to in South Africa, are very high interest with a low principal. Additionally, many of the households possessed either

PAGE 40

40 in order to access goods that serve a similar purpose as cattle, a household need higher educa can function in much the same way that cattle do, i.e for savings and risk mitigation. Households with greater capacity are able to access these market goods, while most households are not and maintain cattle as their primary insurance and savings mechanisms. This conforms to Reardon will lead to an increase need for insurance and, hence, an increase in livestock ownership. This study highlights that spatial market links do little to improve access to the necessary goods that would result in reduced cattle ownership through substitution, rather it is market access through improved household capacity that is key for this behavior change. As much of southern Africa is dry and populated by ethnic groups with a history of pastoralism, an approach to address other than simply a spatial market connection is key in the attainment of the dual goals of conservation and development. Undoubtedly poverty goes beyond a households command over market goods. The large quantity and diversity of multi-dimensional poverty indices point to the attention scientists have consumption or household assets, and the more holistic multi-dimension indices like the Physi cal Quality of Life Index and Human Development Index at the national level, or the Satisfaction with Life Scale or the Personal Wellbeing Index at the household level. The proposed Market Ac cess Index is not an attempt to establish a new multi-dimensional index, rather it is to apply Sens

PAGE 41

41 entitlement framework in looking at market access and to propose another tool for the develop ment toolbox.

PAGE 42

42 Figure 2-1. Map of Study Communities Figure 2-1. Map of Study Communities.

PAGE 43

43 Table 2-2. Model comparison fit results. Table 2-2. Model comparison fit results. Table 2-2. Model comparison fit results. CATTLE model CATTLE model CATTLE model NRR model NRR model NRR model Specification AIC1 AIC2 Specification AIC1 AIC2 Gamma 2.2578 0.0000 Binomial 0.6550 0.0000 Negative Binomial 3.3410 1.0832 Lognormal 8.0886 7.4336 Lognormal 6.1605 3.9027 Poisson 6.5638 4.3060 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2 A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. 1 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. Lower AIC values indicate better fit. 2A change of less than 2 AIC indicates fit equivalency. For more information please see Bolker 2008. Table 2-1. Household utilization of natural resources by study area. Table 2-1. Household utilization of natural resources by study area. Table 2-1. Household utilization of natural resources by study area. Activity Bende Mutale (%) Makoko (%) Grew maize last year 39.80 71.30 Grew groundnuts last year 33.40 11.90 Grew melon last year 38.90 1.30 Grew beans last year 36.60 3.12 Grew Sorghum last year 18.50 0.60 Grew cassava last year 0.00 13.12 Grew vegetables last year 12.42 29.38 Collected natural resources in the last year 1 94.90 69.40 HH owns cattle 26.11 22.50 HH own goats 33.76 15.00 HH owns other animal 2 58.12 45.54 Consumed self produced animal products in the last year 35.35 41.88 Sold animal products last year 28.40 6.60 Sold natural resource products last year 13.30 9.00 At least one person employed in secure wage work 39.20 35.50 At least one person employed in piece work 28.70 17.50 Received remittances last year 16.70 18.10 Received government grants 79.70 87.40 % of households below the poverty line 3 54.20 44.70 1 Includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area. 1 Includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area. 1 Includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area. 2 Includes mainly chickens, pigs, and sheep. 2 Includes mainly chickens, pigs, and sheep. 2 Includes mainly chickens, pigs, and sheep. 3 Poverty line established at R538 per adult equivalent per month (Statistics South Africa 2007) 3 Poverty line established at R538 per adult equivalent per month (Statistics South Africa 2007) 3 Poverty line established at R538 per adult equivalent per month (Statistics South Africa 2007)

PAGE 44

44 Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. Table 2-5. MAI and sub-dimension scores by poverty status. SPI Place & II II II II Infrastructure SPI, Place & II II II IIInfrastructure SSC Social II II II II Connection SSC, Social II II II IIConnection SHC Household II II Capacity SHC, Household II IICapacity S FKA Financial II II Knowledge & II IIAccess SFKA, Financial II IIKnowledge & II IIAccess MAI Composite MAI Composite Mean St. Dev. Mean St. Dev. Mean St. Dev. Mean St. Dev. Mean St. Dev. Not in poverty 54.35 1.31-1 24.57 15.96-1 35.14 25.37-1 28.86 16.21-1 30.96 17.78-1 In poverty 51.64 1.24-1 20.25 12.36-1 28.72 21.16-1 24.03 14.13-1 26.52 16.66-1 p-value 0.1341 0.1341 0.0011 0.0011 0.0031 0.0031 0.0011 0.0011 0.0071 0.0071 1 Difference in means established via a two tailed t-test 1 Difference in means established via a two tailed t-test 1 Difference in means established via a two tailed t-test 1 Difference in means established via a two tailed t-test 1 Difference in means established via a two tailed t-test Table 2-4. Correlation between total household consumption and the MAI index and subindices.1 Table 2-4. Correlation between total household consumption and the MAI index and subindices. 1 Table 2-4. Correlation between total household consumption and the MAI index and subindices.1 Table 2-4. Correlation between total household consumption and the MAI index and subindices.1 Table 2-4. Correlation between total household consumption and the MAI index and subindices.1 Table 2-4. Correlation between total household consumption and the MAI index and subindices.1 Table 2-4. Correlation between total household consumption and the MAI index and subindices.1 SPI, Place & II I IIInfrastructure SSC, Social II II IIConnection SHC, Household IICapacity S FKA, Financial I Knowledge II& Access MAI II II II II IIComposite Total consumption coef., rs 0.12 0.35 0.21 0.40 0.35 p-value 0.01 0.00 0.00 0.00 0.00 1 Spearman rank correlation was used due to non-normally distributed variables 1 Spearman rank correlation was used due to non-normally distributed variables 1 Spearman rank correlation was used due to non-normally distributed variables 1 Spearman rank correlation was used due to non-normally distributed variables 1 Spearman rank correlation was used due to non-normally distributed variables Table 2-3. Market access index construction, factor analysis. Table 2-3. Market access index construction, factor analysis. Table 2-3. Market access index construction, factor analysis. Table 2-3. Market access index construction, factor analysis. Table 2-3. Market access index construction, factor analysis. Description of variables Factor 1 II(Place & IIInfrastructure) Factor 2 II(Social II II I Capital) Factor 3 II(HH II II II II IICapacity) Factor 4 II(Financial II IIKnowledge II& Access) Inverse of the normalized distance to nearest town 0.91 Inverse of the normalized roundtrip paid to travel IIto nearest town 0.54 Inverse of the normalized transport waiting time 0.59 Sum of (population / distance) of each town the IIhousehold conducts market activities II(normalized) 0.78 Sum of each village, town, and city in which a IIhousehold has a family relative (normalized) 0.60 Sum of (population / distance) of each village, II town, and city in which a household has a family IIrelative (normalized) 0.56 Years of school of household head (normalized) 0.43 Ratio of cash to total household income II(normalized) 0.44 Sum of different financial services household IIpossesses (loans, savings, and insurance) II(normalized) 0.32 Total adult males in household (normalized) 0.35 Explained Variance 2.23 0.91 0.59 0.48 Sub-Dimension Weights 0.53 0.22 0.14 0.11

PAGE 45

45 Table 2-6. Descriptive of variables used in the models. Table 2-6. Descriptive of variables used in the models. Table 2-6. Descriptive of variables used in the models. Table 2-6. Descriptive of variables used in the models. Table 2-6. Descriptive of variables used in the models. Table 2-6. Descriptive of variables used in the models. Table 2-6. Descriptive of variables used in the models. Variable Obs. Mean St. Dev. Min Max Description MAI_COMPOSITE 462 28.68 17.84 0.00 73.76 Aggregate market access index II(MAI) MAI_ SPI467 52.95 19.48 12.49 93.91 MAI place & infrastructure sub-II II IIindex MAI_SSC 474 22.36 14.38 0.00 88.43 MAI social connection sub-index MAI_SHC 474 31.85 23.50 0.00 86.84 MAI household capacity sub-index MAI_SFKA469 26.39 15.36 0.00 91.67 MAI financial knowledge & access II sub-index NORTH 474 0.66 0.47 0.00 1.00 Geographic location of village (1 = II IINorth) AGE_HH_HEAD 474 43.79 17.01 16.00 99.00 Age of household head DOM_ETHN_GRP 474 0.84 0.36 0.00 1.00 Households ethnic affiliation is the IIvillage majority (1 = yes) FEMALE_HH_HEAD 474 0.58 0.49 0.00 1.00 The household is headed by a female II(1 = yes) HH_SIZE 474 5.11 2.39 1.00 17.00 Size of the household DEPND_RATIO 474 0.81 0.78 0.00 5.00 Ratio of dependents to total II II II II IIhousehold population LIVILHD_DIVERS 474 4.79 2.92 0.00 15.00 Sum of products and services from II IIwhich a HH derives consumption II IIor income ASSETS1470 4691.05 8525.66 0.00 79380.00 Value in Rand of all household assets II(does not include livestock) HH_PERM_EMPLOY 474 0.38 0.49 0.00 1.00 HH has at least one person II II II II IIpermanently employed in the local IIarea (1 = yes) NR_USE_RATIO1474 0.15 0.15 0.00 0.79 Total natural resources consumed II IIfrom local area to total household IIconsumption HEADS_CATTLE1474 2.37 6.31 0 52.00 Total heads of cattle owned by the II IIhousehold 1 Dependent variable

PAGE 46

46 Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). Table 2-7. Regression modeling results (dependent variable, natural resource use ratio). (N = 458) Aggregated model Aggregated model Aggregated model Disaggregated model Disaggregated model Disaggregated model (N = 458) Coef. II(Std. Err.1) Coef. II(Std. Err. 1 ) Marginal Effects Coef. II(Std. Err.1) Coef. II(Std. Err. 1 ) Marginal Effects MAI_COMPOSITE -0.0117*** -0.0014E(0.0037) M1_PLACE_INF -0.0137** -0.0017E(0.0062) M2_SOC_CONNECT -0.0104*** -0.0012E(0.004) M3_CAPACITY -0.0202*** -0.0023E(0.004) M4_INST_KNW_ACC -0.0111*** -0.0013E(0.004) NORTH -0.3398*** -0.0417E-0.7549*** -0.0950E(0.1200) (0.278) AGE_HH_HEAD -0.0103** -0.0012E-0.0172*** -0.0020E(0.0041) (0.0044) DOM_ETHN_GRP 0.1936 0.0217E0.2500* 0.0268E(0.1356) (0.1362) GEND_HH_HEAD -0.0687 -0.0081E-0.1426 -0.0165E(0.1181) (0.1156) HH_SIZE -0.0638** -0.0075E-0.0515* -0.0059E(0.0281) (0.0297) DEPND_RATIO -0.1446** -0.0170E-0.1944*** -0.0222E(0.0674) (0.0700) LIVILHD_DIVERS 0.1240*** 0.0146E0.1101*** 0.0126E(0.0199) (0.0195) ASSETS -2.25E-05** -2.65E-06 -1.48E-05* -1.70E-06 (9.04E-06) (8.58E-06) HH_PERM_EMPLOY -0.3193** -0.0366E0.1708 0.0198E(0.1609) (0.1899) Constant -0.8826*** 1.0298* (0.2885) (0.5364) AICc 2 0.6535 0.6535 0.6550 0.6550 *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008.

PAGE 47

47 Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). Table 2-8. Regression model results (dependent variable, number of cattle owned). (N = 458) Aggregated Model Aggregated Model Aggregated Model Disaggregated Model Disaggregated Model Disaggregated Model (N = 458) Coef. II(Std. Err.1) Coef. II(Std. Err. 1 ) Marginal Effects Coef. II(Std. Err.1) Coef. II(Std. Err. 1 ) Marginal Effects MAI_COMPOSITE -0.0170 -0.0195 (0.0120) M1_PLACE_INF -0.0067 -0.0074E(0.0266) M2_SOC_CONNECT -0.0055 -0.006 E(0.0119) M3_CAPACITY -0.0291** -0.0321E(0.0134) M4_INST_KNW_ACC 0.0025 0.0027E(0.0148) NORTH 0.7173** 0.7470 0.7566 0.7553E(0.3547) (0.9497) AGE_HH_HEAD 0.0129 0.0148 0.0073 0.0081E(0.0114) (0.0112) DOM_ETHN_GRP 0.8836* 0.7684 0.9284* 0.7675E(0.5141) (0.5198) GEND_HH_HEAD -0.7443* -0.9258 -0.8116** -0.9809E(0.3867) (0.3713) HH_SIZE 0.0130 0.0149 0.0035 0.0038E(0.0752) (0.0898) DEPND_RATIO -0.7284** -0.8351 -0.6562** -0.7240E(0.2998) (0.311) LIVILHD_DIVERS 0.3617*** 0.4147 0.3413*** 0.3765E(0.0640) (0.0634) ASSETS 2.80E-05 3.2E-05 3.11E-05* 3.43E-05 (2.04E-05) (2.04E-05) (1.92E-05) (1.92E-05) HH_PERM_EMPLOY 0.4978 0.6108 0.9649* 1.2400E(0.4731) (0.5473) Constant -2.2657** -1.3305 (1.1142) (2.2981) AICc 2 2.3214 2.3214 2.2578 2.2578 *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 1 Robust standard errors are used and are reported in parentheses under each Coefficient estimate. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008. 2 The Akaike Information Criteria calculated as AIC = (-2lnL + 2k)/N. For more information please see Bolker 2008.

PAGE 48

48 CHAPTER 3 TAKING THE UN OUT OF THE KNOWN UNKNOWNS OF HOUSEHOLD RESOURCE USE DECISIONS IN AFRICA: THE INFLUENCE BETWEEN RISK, MARKET ACCESS AND NATURAL RESOURCE EXTRACTION IN SOUTH AFRICA. Prelude ginal areas of productivity has stimulated many southern African countries to examine if wildlife utilization and ecotourism are a potential tool for both development and conservation (Child et al. in review). While countries have applied different strategies, the ultimate goal of the various policy approaches is to capture the biophysical and economic advantage of wildlife to generate a and the ecosystems in which they are present (Barnes et al. 2002; Suich and Child 2009). This conservation-development linkage has increased in prominence in recent years and underpins many national and regional programs, such as community based natural resource management. However, for any program to be successful, initiatives need to understand the links between vul nerability and resource use. While entitlements and factor endowments affect a rural households income level and constrain their coupled production-consumption decisions (Sen 1981; Dercon et al. 2005), household poverty cannot be explained by these parameters alone. Vulnerability is increasingly sions as uncertainty shifts production-consumption choices away from maximizing returns from a particular activity towards the mitigation of risk (Blarel et al. 1992; Ellis 1998; Alderman et al. 2003; Dercon et al. 2005). Most households in Africa do not have access to formal insurance mechanisms to maintain utility in the face of shocks (Alderman et al. 2003). The mechanisms

PAGE 49

49 employed to cope with risk are inherent in the households livelihood choices, both spatially and Poor households have been linked to increased wild product and fuelwood gathering as a way to diversify and cope with shock (Jodha 1992; Webb and Coppock 1992). Natural resource extrac tion requires labor and little capital, elements that are more typical of impoverished households. Conversely, cattle production is a strategy more often adopted by wealthier households given the high level of capital needed (Christensen 1989). Where there is not a suitable substitute, house holds increase stocks of cattle when welfare increases (Reardon & Vosti 1995). As consumption (or welfare for that matter) increases, so too do insurance needs of a household, and cattle pro vide such an insurance mechanism. What and how much a household decides to produce is, as ability. ture utility of households that are exposed to uninsured risks, thus affecting the decision-making process (Dercon et al. 2005). Poverty cannot be viewed as the sole outcome of a households lack of income and/or factor endowments, but is also a result of the decisions that households make in response to being exposed to possible risks (Sen 1981; Dercon et al. 2005). Several studies to liquid assets, to manage risk, in addition to selling such assets to cope with shocks (Eswaran and Kotwal 1989; Mordoch 1994; Kochar 1995). This risk has the potential to trap a household in poverty (Zimmerman and Carter 2003). To understand how and why a household chooses to

PAGE 50

50 consume local natural resource products or keep cattle, one also needs to understand vulnerabil measures of poverty that are so often used to inform policy design (Chaudhuri et al. 2002). mechanisms to shocks (Christiaenson and Subbarao 2005; Dercon et al. 2005), with some having integrated market access into their investigation of vulnerability. In a study in Vietnam, for in stance, Gaiha et al. (2007) found that households without access to markets are more vulnerable to poverty, moreover, the study indicated that markets reduced vulnerability primarily through a reduction of consumption volatility. Kochar (1995) found that a households ability to access credit and labor markets were key in insulating consumption from idiosyncratic shocks. In Ke nya, Christiaenson and Subbarao (2005) found that vulnerability increased as the time to market increased, largely acting via a decrease in household consumption. While markets play a role in mitigating vulnerability, most studies have conceptualized market access as a spatial concept where either distance or time to market were measured. This assumes that all households from the same village would have the same level of access. Given that the above discussion on vulnerability indicates that access to labor, credit and insurance products plays a crucial role in household decision-making, market access is more than a spatial concept. Accessing the labor market or formal insurance mechanisms likely requires more than decent transportation; it requires that a household has the capacity to utilize and access these resources, as suggested in Sens (1984) entitlement framework.

PAGE 51

51 Objectives This study attempts to advance the understanding of vulnerability through the building of a multi-dimensional market access index in which access is a combination of place, household social connection, and household capacity. The subcomponents are tested to see how they inter act with vulnerability. Additionally, vulnerability and market access are modeled to assess impact on household resource use decisions. It is important for regional policymakers to understand the dynamics between vulnerability and resource use. This will improve the design of mechanisms to reduce rural poverty or, in the case of conservation initiatives, do not unduly harm households by restricting access to crucial risk mitigation or shock-coping mechanisms. Finally, the models test whether there is an interaction effect between vulnerability and each of the market access sub-dimensions. The hypothesis here is that returns from changes in market access to household resource use decisions will be different between vulnerable and non-vulnerable groups. In other words, vulnerable households will shift away from natural resource consumption faster than nonvulnerable households if market access is improved. This has important implications for con servation policy in that improving welfare for vulnerable household by improving market access would also result in a shift away from local resource extraction as a primary coping strategy. Study Area and Duluthulu in the north, and Makoko in the south. The northern cluster of villages are in the Limpopo Province within the Mutale Municipality. Makoko lies near the Numbi entrance gate to Kruger National Park and is in Mbomela Municipality in Mpumalanga Province,. The Mutale region receives on average 400-500mm of rain annually (Thomas et al. 2007), while Makoko receives 840-1,670mm annually (Louw and Scholes 2006).

PAGE 52

52 tion of households are engaged in agricultural production in both provinces, particularly in Lim popo Province where 51.9% of all households are smallholder non-commercial producers, while in Mpumalanga 25.2% of households are non-commercial producers (Directorate of Agricultural Statistics 2008). The main constraints of production are the agro-ecological conditions, particu tance of identifying livelihood alternatives that are decoupled from primary production (Child 1989) to minimize the impacts of shocks on the current livelihood system. Within the Mutale village cluster (Mutale) the dominant ethnic group is the Venda who historically have been pastoralists. During the early 1900s the South African government began concentrating the Venda into homelands, forcing them into a village structure that was never there before, according to interviews conducted with Chief Mutale as part of this study. While terms of income or employment. As such, people extract revenue from land in the only other way for example). Livestock, especially cattle, have great cultural and economic importance to the Venda. Given the dryer conditions in the north, Mutale households rely more on cattle, as cattle are more robust to dry conditions. Makoko is a much larger community than any of the Mutale study villages. Yet, with about 1,100 households, it is not considered a large village in the south. Like Mutale, it is located adjacent to Kruger and is comprised of an ethnic group, the Swati, which have also historically been pastoralists. However, inhabitants of Makoko focus more on agriculture because rainfall

PAGE 53

53 is higher and so is population density. Makoko has the advantage of being within 45 minutes of three cities (Nelspruit, White River, and Hazyview), and several towns for providing household inputs, a viable labor market, and a greater diversity of coping mechanisms. If households are able to shift to opportunities in the surrounding areas in Makoko, they may not be forced into divesting assets or reducing consumption as main shock coping mechanisms. Methodology Questionnaire Design Preceding the data gathering process, key informants were interviewed to establish the attaining permission from the Traditional Authority in each village, the survey instrument was The information gained through the piloting process was used to improve question design and ensure adequate variables were entered to ensure data quality and uniform participant interpreta tion. domly selected household interviews were conducted, 323 from Mutale and 166 from Makoko. The detailed questionnaires consisted of 5 modules: household demographics; water, market and health services access; detailed income including both household production and formal and non-formal employment; comprehensive consumption, including questions on household food, non-food, and durable good consumption; and shocks and coping strategies. Data collected dur ing village household interviews was fed into an econometric model to establish vulnerability and risk.

PAGE 54

54 Sampling In each village a random sample was drawn from an established sampling frame. For three of the villages in the northern Mutale area, the researcher formed the sampling frame from preexisting village lists, but in Tshikuyu (71 households) and Beleni (23 households) villages, the into a computer, a random number generator assigned a number to each household. This number was used to order the households; the sample included households numbered from one until the random number and were included in ascending order if households from the original sample refused or were not present after two attempts. A village list was not available for the southern village of Makoko, nor was it feasible to construct one because of time, funding constraints and the size of the community (1,100 house hold in 4 blocks). A s established a sampling frame was by modifying vegetation transects ap sample by blocks. The researcher then numbered each street in the block and assigned each street a random number. Household interviews were conducted according to the number order gener of that particular street edge. Upon the conclusion of a street edge, the next highest ordered street was selected. This continued until the sample quota was reached for each block.

PAGE 55

55 Data Analysis Market access index From Chapter 2, it was established that the market access index was comprised of four dimensions or sub indices: Place & Infrastructure (SPI); Social Capital (SSC); household capac principal axis factors method, which is preferred to the maximum likelihood method with data that violates assumptions of normality (Fabrigar et al. 1999). It is also preferred to PCA, as PCA loads (Costello and Osborne 2005). Often the decision on how many factors to retain are made using the Kaiser criterion where factors are retained if the eigenvalues exceed unity (Joseph et al. 1998). However, this study utilized parallel analysis when deciding how many factors to retain as it has been found to be much more accurate (Zwick and Velicer 1986; Velicer and Jackson 1990). This lead to the the four factor solution given above. After the four factor decision, an oblique promax rotation was conducted on the FA solution to improve interpretability of the pattern matrix. The pattern matrix gives the item loadings on each factor that are used to construct the index. The pre-determined factor loading cut-off point was 0.32 (Tabachnick and Fidell 2001). dex. The sub-indices were weighted using the proportion of variance they explain in the dataset, and aggregated using a geometric mean aggregation procedure in forming the main MAI. The weighing and aggregation procedure in this study follows the suggestion by Garriga and Foguet to note that each sub-index and the composite was normalized to fall between 0 and 100.

PAGE 56

56 Vulnerability estimation tion, is an ex-ante measure. As such, measures of vulnerability based on static poverty levels fail to integrate risk into their measurement (Chaurdhuri 2003). To make inferences on vulnerability, one needs to understand future poverty, a forward looking measure. Three methods have been established in the econometric literature: vulnerability as expected poverty (VEP), vulnerability as uninsured exposure to risk (VER); and vulnerability as low expected utility (VEU). While each model is similar in that they attempt to predict a measure of future welfare, they all quantify vulnerability differently and have a slightly different conceptualization as the names suggest. consumption distributions would be directly observed (Hoddinott and Quisumbing 2003), time and funding constraints excluded this option. However, Chaudhuri et al. (2002) have established a method for establishing VEP using a single cross-sectional survey. The VEP model produces future, and this probability is calculated for each household. The main limitation of the VEP method is that it assumes that cross-sectional variability approximates inter-temporal variability (Chaudhuri et al. 2002). The model uses expected con sumption (the chosen welfare measure), and the variance of consumption, to see how consump (2002) approach is to assume that for a given household, h, consumption is a function of house

PAGE 57

57 hold characteristics, X h and a disturbance term, eh, that captures idiosyncratic shocks and other unobservable characteristics: : (1) The disturbance term, eh, is assumed to be: (2) where, in accordance with Tesliuc and Lindert (2002), Z h is the matrix of household character istics augmented by a vector of household shocks, and X h is a vector of parameters. Estimates of B and X h are attained via a three-step feasible generalized least squared (FGLS) procedure established by Amemiya (1977). The initial base regression is given in Appendix B. This allows one to estimate: (3) where vulnerability (V ht) is the probability that a household, with characteristics X h will be below the poverty line in the future (a detailed description and model derivation can be found in Chaudhuri et al. 2002). How effective is the VEP cross-sectional method at identifying vulnerable populations? cross-sectional method. In Chile and Argentina, Cruces, Gasparini, Bergolo, and Ham (2010) economic periods. However, the accuracy of the cross-sectional VEP method declined to 79% during the recession of 2001-2002, as the widespread economic shocks caused noise, producing increased error in the vulnerability estimates. Chuadhuri, Jalan, and Suryahadi (2002) used panel

PAGE 58

58 did a good job at predicting poverty and was very accurate at reproducing the true vulnerability distribution in the population. Results Market Access Chapter 2 goes into detail with respect to the construction and description of the MAI. The four dimensions or sub-indices are: Place & Infrastructure (SPI); the Social Connection (SSC); that comprise these factors and their loadings are given in Table 3-1. Table 3-2 breaks down the four dimensions of the MAI, and the composite, by village for the main descriptive statistics. As the sub-indices have been normalized, the lowest possible score is 0 with 100 being the highest, one would interpret a higher relative score as indicating better access. Makoko, the southern community, had the highest score in the SPI dimension, which is not surprising given its proximity to several large towns and a location within 45 min utes of the provincial capital of Nelspruit. Additionally, residents of Makoko have access to regular buses, with a lower cost relative to the north. Bende Mutale is the furthest from a main market town (that being Thoyondo for the northern communities), and has the lowest mean score in the SPI dimension at 29.73. Its distance from Thoyondo and location at the end of the road just before it enters KNP results in long wait times for transportation and high cost of transport when minibus taxis arrive. Tshikuyu has the lowest mean score on SSC at 16.93, followed by Mutale B (21.81). This low score in Mutale B may be related to the close proximity to a coal mine, with many house holds having at least one member working in the mine or in the staff town adjacent to the mine.

PAGE 59

59 Access to this resource may very well preclude the need to spatially diversify household labor as a risk mitigating strategy. Belleni has the highest mean of SSC at 33.76. Belleni has the lowest mean score in the SHC dimension at 24.35. Conversely Mutale B had the highest mean score for SHC at 40.37, presumably because proximity to the mine im proves access to educational opportunities and familiarity with the cash economy. Duluthulu, Belleni, and Makoko all have the highest mean score for SFKA at 28.87, 28.75 and 28.13 respectively. Duluthulu is within a 10 minute walk to a small secondary market town, Makoko is close to several large towns, and Bellenis is connected to town through rela households. Bende Mutale has the lowest mean of SFKA at 22.47. In terms of the aggregated MAI, Makoko has the highest mean score at 33.47, followed by Mutale B (31.23), Duluthulu (26.63), Tshikuyu (25.25), Belleni (24.41), with Bende Mutale attaining the lowest mean score at 21.42. Vulnerability In the examination of vulnerability, this study breaks down vulnerability into its proximate causes in addition to looking at the straight up vulnerability score. Vulnerability is produced both from low levels of welfare and from volatility in consumption streams. Households who are vulnerable due to persistent consumption below the poverty can be thought of as vulnerable to structural poverty, while households who would not be impoverished if consumption volatil below the poverty line and their vulnerability is above the 0.5 level (greater than 50% chance of

PAGE 60

60 predicted consumption above the poverty line and a mean vulnerability above the 0.5 threshold. Table 3-3 reports these results, in addition to mean vulnerability and an overall vulnerability nerability differs amongst households of varying characteristics. The results for vulnerability vulnerable to transitory poverty are given as proportions. The table also reports whether or not vulnerability levels and an equality of proportion test for the proportion data. Special attention is given to resource use characteristics of the household in Table 3-3 to examine the association between vulnerability and resource use activities. Households that own than non-owning households. Additionally, households that grow crops have a statistically sig crops. While crops and animals appear to reduce risk for a household, the opposite can be said of natural resources. Households who collect natural resources have a mean vulnerability level that households that make up a larger percentage of their consumption from local natural resources In looking at the proximate causes of vulnerability, a greater proportion of households who own cattle are vulnerable to transitory poverty (23%) than households who do not own animals. While cattle and other animals may serve as an insurance mechanism, it may not protect

PAGE 61

61 the household completely from consumption volatility. This is linked to households vulnerable to structural poverty, where 42% of non-cattle owning households are vulnerable to structural pov risk or cope with shocks is partly responsible for their lower level of mean consumption. Of all households that collect natural resources, 40% are vulnerable to structural poverty versus only 24% for households that do not collect natural resources. While 47% of households who con sume greater than 15% of their total consumption from local resources are vulnerable to struc tural poverty, only 32% of household that consume less than 15% are vulnerable to structural poverty. Vulnerability and Market Access To begin the investigation into the links between vulnerability, market access, and resource use, a spearman rank correlation was conducted to look at the relationships between the MAI, its sub-indices, and variables dealing with household resource use decisions, namely cattle (the total head of cattle owned by the household) and use of local natural resources or NRR as this study uses (Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption) (Table 3-4). Vulner the sub-dimensions except for SFKA. Each sub-dimension is negatively correlated with vulner ability, indicating that as SPI, SSC, or SHC increase vulnerability decreases, the same can be said for the aggregated index, MAI_C, as market access improves, vulnerability decreases. This reduction in vulnerability, is achieved through a combination of improved welfare and increased exposer to ex ante risk mitigation strategies, and/or ex post coping mechanisms that markets

PAGE 62

62 provide, as suggested by the literature. However, it bears mentioning that with each correlation Chapter 2 dealt more explicitly with consumption and the connection between market access and household use decisions. But table 3-4 also displays the correlation between con MAI sub-dimension and the composite index, MAI_C. With the exception of SPI, the strength of association is stronger to consumption as compared to vulnerability. SSC and SFKA have cor moderate strength of association at 0.34. As with Table 3-3, Table 3-4 tells a similar story in respect to household resource use decisions and vulnerability. Vulnerability has a positive association with NRR, meaning that as vulnerability increases, so does local natural resource use. This is consistent with Francis ral resource dependent activities such as the collection of wild resources, the raising of various livestock species, and the production of multiple crops. In a similar vein, NRR has a negative association to consumption, the poorer the household, the more it allocates consumption to local natural resources, likely both to maintain current consumption levels, spread risk, and possibly to cope with shocks. Cattle are an important savings and insurance mechanism to rural households, especially where other insurance options are non-existent (Reardon and Vosti 1994). If households are not able to access higher quality insurance options (such as formal credit and insurance instru ments), as welfare increases so to does the insurance needs of the households, hence the positive

PAGE 63

63 ship increases as consumption increases. However, the relationship is reverse when it comes to vulnerability, but also expected. Given the aforementioned reliance on cattle as an insurance is consistent. Households are able to leverage cattle assets to maintain welfare in times of shock through their sale. Furthermore, this alters the household decisions on what to produce and / or consume as they have insurance against possible shortfalls in welfare which future shocks may produce, potentially allowing the household to place greater emphasis on higher risk, higher reward activities. Interactions Between Vulnerability and Resource Use Tables 3-5 and 3-6 display the results of the probit regression modeling to investigate the able in these models if their probability of being impoverished in the future is greater than 0.50. descriptive statistics of dependent and independent variables are given in appendix C). While model. As such, the returns to local natural resource extraction for each of the MAI sub-dimen sions is the same for vulnerable and non-vulnerable groups, i.e., constant returns from the MAI need to be interpreted in conjunction. For example, with SPI at its mean (52.95), the effect of be the models.

PAGE 64

64 score, one can interpret that a vulnerable household has a greater probability of consuming local natural resource. In terms of SPI, as SPI improves, the probability of consuming local natural resources declines. Additionally in each of the model, SHC and SFKA are negatively associ ated with local natural resource use. As scores in any of these dimensions increase, local natural resource use declines. Figure 3-2 improves the interpretation of the models and the interaction between vulnerability and the MAI sub-dimensions by predicting the probability of local natural resource use as MAI and vulnerability change. The y axis displays the probability that a house hold will consume more than 15% of its total consumption from local natural resources, while the x axis displays scores of the MAI sub-dimension, all other variables in the model are held at mean values. By improving market access, the reliance on local natural resources is reduced for both vulnerable and non-vulnerable groups. Looking at it another way, vulnerable households rely much more on local natural resources than non-vulnerable groups. When formulating poli cies related to access to natural resources, this is important to understand as reducing access to natural resources without providing other options will impact the very households who have the least capacity to cope with reduced access to resources. Table 3-6 displays models using cattle ownership as the dependent variable. As the table constant returns to household cattle ownership for vulnerable and non-vulnerable households. Unlike the previous four models, vulnerability does not have an impact on cattle ownership when controlling for other factors. Given that table 3-3 illustrated households that are vulnerable to transitory poverty tended to have cattle, the lack of a relationship between vulnerable and nonvulnerable groups is not surprising as households that experience consumption volatility appear

PAGE 65

65 to mitigate risk by keeping cattle. In terms of the MAI sub-dimensions, SPI and SHC have an impact on cattle decisions. As SPI increases so to does the probability that a household will own cattle, the opposite impact to natural resource consumption decisions. As cattle serve a different purpose than other natural resource products (savings and insurance, rather than direct consump tion), improving the time to market town does not necessarily improve access to the type of substitute goods and activities that would cause a shift away from cattle. Keeping this relation expected. The type of activities and goods that would be an adequate cattle substitute are likely only accessible with higher education and exposer to the cash economy. Figure 3-3 predicts the probability of cattle ownership between vulnerable groups as MAI sub-dimension scores are al in probabilities of cattle ownership as SPI increases, while the probability that a household owns cattle decreases with an increase in SHC. It is important to note that Figure 3-2 and 3-3 report the Discussion Improved understanding of vulnerability is key in poverty dynamics. Poverty alleviation 15% of the households in the Mutale area are vulnerable to transitory poverty, while 21% of households in the south are vulnerable. These are households that are currently above the pov erty line that are in danger of falling below. Policy mechanisms using traditional poverty indices may fail to include these households. While most vulnerable households in these communities

PAGE 66

66 were vulnerable to structural poverty with consumption consistently below the poverty line, as Chaudhuri et al. (2002) pointed out, these households too are susceptible to risk. While they likely causes households vulnerable to structural poverty to choose low-risk, low-yield activi to structural as well as transitory poverty. As with the treatment of malaria in the public health realm, poverty should be approached through the treatment of both the occurrence and the risk of occurrence. As market access conveys improved exposer to household inputs, a market for household this study which illustrate that improved access reduces vulnerability is not surprising. However, looks at market access as a multidimensional concept that incorporates place, social connections, household capabilities, and institutional knowledge. While a method for aggregation into a single index is offered, keeping the sub-dimensions disaggregated adds more information to the mod role in reduced vulnerability. Additionally, the MAI sub-dimensions highlight that while distance and transportation infrastructure is important in the reduction of vulnerability, it does not tell the whole story as this study clearly shows that a household needs the capacity to leverage market goods. While the hypothesis that the MAI sub-dimensions interacts with vulnerability to increase the gains from MAI to local resource use is not supported, the models highlight the impact that both vulnerability and the MAI sub-dimensions have on resource use decisions. Improving SPI,

PAGE 67

67 SHC, or SFKA will reduce local resource consumption for both vulnerable and non-vulnerable households alike. Vulnerable households have a much greater probability of using local resources as a major part of total consumption. If policies and management plans aimed at conservation understand this link, focusing on vulnerable groups will have the greatest overall impact on local resource extraction. However, the structure of the policy is key. If a policy is restrictive in nature, such as an establishment of a new protected area or greater enforcement of a pre-existing one, the outcome is likely to push more people into or further into poverty. Essentially the households extraction partly to mitigate risk, the least resilient strata of society. As conservation is becoming increasing linked to rural development both around the world and within southern Africa, understanding factors related to household resource use deci sions is key. Natural resources are an important risk mitigation and shock coping mechanism to rural households as this study demonstrates. If initiative, such as community based natural re source management (CBNRM), that attempt to link development and conservation outcomes are to be successful, a greater understand of vulnerability must be integrated. It is not surprising that Shackelton et al. (2002) found in a case study of 13 CBNRM program in Africa that the negative trade-offs of program implementation were experienced by the poor. While CBNRM and similar mately are few in number and more accessible to individuals with higher education. Restricted resource access is felt mainly by the very households that a most dependent on such resources. Conservation programs need to understand the links between resource use and vulnerability. If policy is to be successful in achieving either its goal of conservation or development, steps need

PAGE 68

68 to be taken to ensure households at risk of poverty have access to substitute goods that would reduce reliance on natural resources.

PAGE 69

69 Table 3-1. Market access index construction, factor analysis. Table 3-1. Market access index construction, factor analysis. Table 3-1. Market access index construction, factor analysis. Table 3-1. Market access index construction, factor analysis. Table 3-1. Market access index construction, factor analysis. Description of Variables Factor 1 Factor 2 Factor 3 Factor 4 Inverse of the normalized distance to nearest town 0.9126 Inverse of the normalized roundtrip paid to travel to nearest IItown 0.5366 Inverse of the normalized transport waiting time 0.5908 Sum of (population / distance) of each town the household IIconducts market activities (normalized) 0.7750 Sum of each village, town, and city in which a household has a IIfamily relative (normalized) 0.6006 Sum of (population / distance) of each village, town, and city IIin which a household has a family relative (normalized) 0.5628 Years of school of household head (normalized) 0.4317 Ratio of cash to total household income (normalized) 0.4360 Sum of different financial services household possesses (loans, IIsavings, and insurance) (normalized) 0.3165 Total adult males in household (normalized) 0.3504 Explained Variance 2.2350 0.9150 0.5950 0.4850 Sub-Dimension Weights 0.5350 0.2250 0.1550 0.1150SPI Place & Infrastructure, SSC Social Connection, SHC Household Capacity, SFKA Financial Knowledge & Access SPI Place & Infrastructure, SSC Social Connection, SHC Household Capacity, SFKA Financial Knowledge & Access SPI Place & Infrastructure, SSC Social Connection, SHC Household Capacity, SFKA Financial Knowledge & Access SPI Place & Infrastructure, SSC Social Connection, SHC Household Capacity, SFKA Financial Knowledge & Access SPI Place & Infrastructure, SSC Social Connection, SHC Household Capacity, SFKA Financial Knowledge & Access

PAGE 70

70 Table 3-2. MAI by village. Table 3-2. MAI by village. Mean St. Dev. Min Max Belleni Belleni Belleni Belleni Belleni MAI_ SPI 39.86 8.08 28.22 57.85 MAI_SSC 33.76 18.43 12.50 88.43 MAI_SHC 24.35 22.12 0.00 68.44 MAI_SFKA 28.75 13.38 0.00 66.67 MAI_C 24.41 17.14 0.00 45.42 Bende Mutale Bende Mutale Bende Mutale Bende Mutale Bende Mutale MAI_ SPI 29.73 7.02 12.49 54.64 MAI_SSC 21.93 14.04 12.50 69.85 MAI_SHC 29.14 22.05 0.00 79.42 MAI_SFKA 22.47 13.89 0.00 83.33 MAI_C 21.42 10.57 0.00 40.39 Duluthulu Duluthulu Duluthulu Duluthulu Duluthulu MAI_ SPI 43.18 6.19 25.90 57.64 MAI_SSC 23.37 12.12 12.50 70.86 MAI_SHC 32.37 24.26 0.00 80.10 MAI_SFKA 28.87 15.93 0.00 83.33 MAI_C 26.63 15.34 0.00 46.24 Mutele B Mutele B Mutele B Mutele B Mutele B MAI_ SPI 47.66 6.69 29.06 62.46 MAI_SSC 21.81 10.32 12.50 76.63 MAI_SHC 40.37 22.16 0.00 81.30 MAI_SFKA 24.05 13.03 0.00 58.33 MAI_C 31.23 13.95 0.00 49.60 Tshikuyu Tshikuyu Tshikuyu Tshikuyu Tshikuyu MAI_ SPI 42.12 6.40 22.91 56.04 MAI_SSC 16.93 8.37 12.50 58.60 MAI_SHC 31.44 22.57 0.00 78.44 MAI_SFKA 26.63 13.54 8.33 58.33 MAI_C 25.26 13.10 0.00 42.24 Makoko Makoko Makoko Makoko Makoko MAI_ SPI 78.05 3.60 68.95 93.91 MAI_SSC 22.74 17.31 0.00 75.00 MAI_SHC 29.41 24.20 0.00 86.84 MAI_SFKA 28.13 17.28 0.00 91.67 MAI_C 33.47 22.97 0.00 73.76

PAGE 71

71 Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Table 3-3. Vulnerability within and between selected groups. Vulnerability Vulnerability Vulnerability classification (%) Vulnerability classification (%) Vulnerability classification (%) Vulnerability classification (%) Vulnerability classification (%) Vulnerability classification (%) Group Mean t-score1Vulnerable IIto IIpoverty z-score2 Vulnerable IIto IItransitory IIpoverty z-score2 Vulnerable IIto IIstructural IIpoverty z-score2N Households IIcurrently in IIpoverty 0.61 -11.227*** 73.4 -8.645*** 14.7 -1.448* 53.5 -7.439*** 243 Households IIcurrently not in IIpoverty 0.40 -11.227*** 33.5 -8.645*** 19.8 -1.448* 20.3 -7.439*** 227 Northern IIcommunities 0.54 -4.415*** 58.8 -2.920*** 15.3 -1.623* 23.9 -4.339*** 311 Southern IIcommunity 0.45 -4.415*** 44.7 -2.920*** 21.3 -1.623* 44.4 -4.339*** 159 Male headed IIhouseholds 0.46 -4.376*** 44.2 -3.651*** 14.9 -1.173* 31.0 -2.467** 197 Female headed IIhouseholds 0.55 -4.376*** 61.2 -3.651*** 19.0 -1.173* 42.1 -2.467** 273 Own Cattle 0.45 -3.287*** 44.3 -2.400** 22.9 -1.850* 23.5 -3.561*** 115 Do not own IIcattle 0.53 -3.287*** 57.1 -2.400** 15.4 -1.850* 42.0 -3.561*** 355 Own other IIanimals 0.48 3.029*** 51.8 -1.209 20.0 -1.909* 32.6 -2.646*** 282 Do not own IIother animals 0.55 3.029*** 57.4 -1.209 13.2 -1.909* 44.7 -2.646*** 188 Collect natural IIresources 0.52 -1.894* 56.4 -2.600*** 17.3 -0.026* 39.5 -2.314** 408 Do not collect IInatural IIresources 0.46 -1.894* 38.7 -2.600*** 17.2 -0.026* 24.2 -2.314** 62 Grow crops 0.45 6.157*** 65.9 -4.777*** 17.7 0.8* 26.5 -5.303*** 253 Do not grow IIcrops 0.58 6.157*** 43.9 -4.777*** 16.8 0.8* 50.2 -5.303*** 217 Household IIconsumes less IIthan 15% from IIlocal sources 30.48 -4.404*** 47.3 -3.877*** 15.8 -1.117* 32.3 -3.043*** 300 Household IIconsumes more IIthan 15% from IIlocal sources 30.57 -4.404*** 65.9 -3.877*** 19.9 -1.117* 46.5 -3.043*** 170 *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 1 T-score based on two group t-test. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 2 Z-score based on two group equality of proportion tests. All statistical tests conducted were two-tailed. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption. 3 Natural resource use ratio includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area over total household consumption.

PAGE 72

72 Table 3-4. Correlation between total household consumption, vulnerability, resource use, and the MAI index and sub-indices.1 Table 3-4. Correlation between total household consumption, vulnerability, resource use, and the MAI index and sub-indices. 1 Table 3-4. Correlation between total household consumption, vulnerability, resource use, and the MAI index and sub-indices.1 Vulnerability Consumption Vulnerability 2 1.2843*** Consumption -0.2843*** -1.2843*** Cattle -0.1742*** -0.2607*** NRR -0.2527*** -0.2388*** MAI_SPI -0.1816*** -0.1146*** MAI_SSC -0.1684*** -0.3423*** MAI_SHC -0.1861*** -0.2020*** MAI_SFKA -0.0446*** -0.3894*** MAI_C -0.1629*** -0.3390*** 1 Spearman rank correlation was used due to non-normally distributed variables 1 Spearman rank correlation was used due to non-normally distributed variables 2 Probability that a household will fall below the poverty line in the future 2 Probability that a household will fall below the poverty line in the future

PAGE 73

73 Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Table 3-5. Impact of market access and vulnerability on household natural resource use decisions. Variables Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Dependent variable: Natural resource use ratio (N = 458) Variables (1) (1) (2) (2) (3) (3) (4) (4) Variables Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) MAI_SPI*VULN -0.00484 (0.00686) MAI_SSC*VULN -0.00231 (0.00983) MAI_SHC*VULN -9.11e-05 (0.00628) MAI_SFKA*VULN -0.00776 (0.00963) VULN -1.103*** -0.897*** -0.849*** -0.666** (0.403) (0.276) (0.251) (0.281) MAI_ SPI -0.0187* -0.0218** -0.0218** -0.0215** (0.00968) (0.00859) (0.00860) (0.00860) MAI_SSC -0.00585 -0.00488 -0.00580 -0.00570 (0.00493) (0.00628) (0.00493) (0.00493) MAI_SHC -0.0203*** -0.0203*** -0.0204*** -0.0204*** -0.0202*** (0.00529) (0.00528) (0.00621) (0.00529) MAI_SFKA -0.0114** -0.0110** -0.0110** -0.0151** (0.00560) (0.00556) (0.00556) (0.00759) NORTH -1.199*** -1.219*** -1.221*** -1.216*** (0.361) (0.359) (0.359) (0.360) AGE_HH_HEAD -0.0111** -0.0109** -0.0110** -0.0111** (0.00515) (0.00516) (0.00515) (0.00516) DOM_ETHN_GRP -0.0922 -0.0931 -0.0969 -0.0800 (0.188) (0.188) (0.188) (0.189) GEND_HH_HEAD -0.129 -0.131 -0.126 -0.118 (0.144) (0.145) (0.145) (0.145) HH_SIZE -0.178*** -0.180*** -0.180*** -0.183*** (0.0409) (0.0408) (0.0409) (0.0411) DEPND_RATIO -0.00172 -0.00489 -0.00498 -0.00837 (0.0977) (0.0975) (0.0976) (0.0976) LIVILHD_DIVERS 0.162*** -0.160*** -0.161*** -0.163*** (0.0285) (0.0284) (0.0285) (0.0285) ASSETS -3.09e-07 -6.84E-08 -1.34e-08 -1.02e-06 (9.49e-06) (9.41e-06) (9.42e-06) (9.50e-06) HH_PERM_EMPLOY -0.316 -0.317 -0.316 -0.307 (0.211) (0.211) (0.211) (0.212) Constant -2.423*** -2.587*** -2.606*** -2.689*** (0.839) (0.801) (0.806) (0.804) Pseudo R 2 0.193 0.193 0.192 0.194 Joint hypotheses 1 25.02*** 24.60*** 24.56*** 25.05*** *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic.

PAGE 74

74 Predicted Pr (LNR c > 15% of Total c1) 0 0.2 0.5 0.7 0.9 0 25 50 75 100 0 0.2 0.5 0.7 0.9 0 25 50 75 100 0 0.2 0.5 0.7 0.9 0 25 50 75 100 Not Vulnerable Vulnerable Place & Infrastructure Sub-index (SPI) Household Capacity Sub-index (SHC) Financial Knowledge & Access Sub-Index (SFKA) Sub-Index Value Dotted lines indicate the 10th and 90th percentiles for each of the MAI sub-dimensions.1 Predicted probability of a household having consuming greater than 15% of its total consumption (c) from local natural resources (LNR). Figure 3-1. Predicted probabilities of local natural resource consumption by vulnerability classification and MAI sub-dimensions.

PAGE 75

75 Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Table 3-6. Impact of market access and vulnerability on household cattle ownership. Variables Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Dependent variable: Number of cattle owned (N = 458) Variables (5) (5) (6) (6) (7) (7) (8) (8) Variables Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) Coef. II(Std. Err.) MAI_SPI*VULN -0.00848 (0.00814) MAI_SSC*VULN -0.00732 (0.0116) MAI_SHC*VULN -0.00942 (0.00699) MAI_SFKA*VULN -0.00207 (0.0107) VULN -0.482 -0.184 -0.284 -0.0347 (0.481) (0.319) (0.271) (0.341) MAI_ SPI -0.0201* -0.0248** -0.0251** -0.0248** (0.0113) (0.0103) (0.0103) (0.0103) MAI_SSC -0.00189 -0.00433 -0.00244 -0.00192 (0.00568) (0.00690) (0.00576) (0.00569) MAI_SHC -0.0140** -0.0137** -0.0175*** -0.0136** (0.00594) (0.00593) (0.00662) (0.00592) MAI_SFKA -0.00717 -0.00711 -0.00667 -0.00803 (0.00619) (0.00618) (0.00618) (0.00799) NORTH -1.044** -1.058*** -1.086*** -1.063*** (0.406) (0.406) (0.407) (0.405) AGE_HH_HEAD -0.000812 -0.000999 -0.000921 -0.000705 (0.00580) (0.00581) (0.00582) (0.00581) DOM_ETHN_GRP -0.329 -0.332 -0.303 -0.322 (0.240) (0.240) (0.240) (0.240) GEND_HH_HEAD -0.173 -0.167 -0.193 -0.183 (0.164) (0.165) (0.164) (0.164) HH_SIZE -0.0222 -0.0202 -0.0203 -0.0202 (0.0408) (0.0409) (0.0410) (0.0411) DEPND_RATIO -0.305** -0.292** -0.287** -0.297** (0.142) (0.141) (0.142) (0.142) LIVILHD_DIVERS -0.236*** -0.240*** -0.246*** -0.237*** (0.0315) (0.0319) (0.0324) (0.0316) ASSETS -3.38e-05*** -3.38e-05*** -3.35e-05*** -3.35e-05*** -3.46e-05*** -3.46e-05*** -3.30e-05*** (1.04e-05) (1.04e-05) (1.05e-05) (1.05e-05) HH_PERM_EMPLOY -0.196 -0.185 -0.189 -0.194 (0.233) (0.233) (0.232) (0.233) Constant -3.521*** -3.772*** -3.725*** -3.845*** (1.015) (0.980) (0.981) (0.982) Pseudo R 2 -0.298 -0.297 -0.299 -0.296 Joint hypotheses 1 -1.10 -0.41 -1.83 -0.05 *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Standard errors are reported in parentheses under each Coefficient estimate. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic. 1 Displays the chi2 ( 2) test statistic.

PAGE 76

76 Predicted Pr (cattle > 01)Place & Infrastructure Sub-index (SPI) Household Capacity Sub-index (SHC) Sub-Index Value 0 0.2 0.4 0.5 0.7 0 25 50 75 100 0 0.1 0.2 0.3 0.4 0 25 50 75 100 Predicted Pr (cattle > 01) Not Vulnerable Vulnerable Dotted lines indicate the 10th and 90th percentiles for each of the MAI sub-dimensions.1 Predicted probability of a household owning at least one head of cattle.Figure 3-2. Predicted probabilities of household cattle ownership by vulnerability classification and MAI sub-dimensions.

PAGE 77

77 CHAPTER 4 CONSTRUCTION OF A SHORT-FORM VULNERABILITY MODEL FOR RAPID ASSESSMENT OF COMMUNITIES IN SOUTH AFRICA Why There is a Need In the developing world, many of the policy interventions are driven by the desire to reduce poverty. This is especially true in rural areas where policies are attempting to reduce the number of impoverished households. Increasingly, programs such as community based natural resource tion of who is in poverty has been crucial in the design and monitoring of these programs. While poverty analysis now considers poverty a multidimensional issue (Cruces et al. 2010), very few studies have integrated risk of future poverty into the analysis (Dercon 2005b). Poverty and vulnerability are linked but measure two very different concepts. Poverty is tion or lack of capabilities and resources preventing the household from meeting current needs (Chaudhuri 2003). Vulnerability, on the other hand, is an ex ante measure of anticipated future household welfare, where risk ensures that future welfare is uncertain (Dercon 2005b; Chaudhuri 2003). Essentially, households that are poor now, may not be poor in the future and, conversely, current non-poor households could fall into poverty into the future. There is a difference between the actual state of being poor and the threat of poverty. Vulnerability, or uninsured risk alters household decisions, as well as reduces household capacity to cope with shocks. The World Development Report (2001) highlighted the fact that households place a lot of time and effort in the attempt to deal with risk. Vulnerable households often chose low-risk, low reward activi ties when encumbered by uninsured risk (Blarel et al. 1992; Ellis 1998; Alderman et al. 2003). Ultimately risk has been found not only to reduce growth, but also to reduce the formation of

PAGE 78

78 capital stock, impacting future growth potential. In Zimbabwe, Elbers and Gunning (2003) found that exposer to risk resulted in a 40% reduction in capital stock for households. In an attempt to smooth consumption in the face of shocks, households may sell productive asset that could ulti mately result in transitory poverty becoming permanent (Rosenzweig and Wolpin 1993). Understanding vulnerability is necessary for the design of policy mechanisms aimed at reducing poverty. From a conservation standpoint, targeting vulnerable households is key to re ducing household use of natural resources. Parks and protected areas in Southern Africa are near by diversifying their livelihood activities. In addition to the growth of various crop types and the raising of multiple livestock species, vulnerable households have been linked to increased wild product and fuel wood gathering as a way to diversify livelihood activities and cope with shock (Jodha 1992; Webb & Coppock 1992). Reducing vulnerability would contribute to alleviating pressure that extractive resource activities are having in and near protected areas. Additionally, it is vital that protected area policies minimize impact on vulnerable households, as it is often the most vulnerable households that rely on natural resource activities and have the least capacity to weather reduced access to natural resources. Since the establishment of the Social Ecology Department in 1994, South African Na tional Parks (SANParks) has committed to meeting the needs of neighboring rural communities, in addition to the long standing goal of wildlife conservation. The Social Ecology Department education; cultural resource management; and research and monitoring (SANParks 2000). Social ecologists in Kruger National Park (KNP) are committed to performing these functions and con tinue the process of gathering information on the surrounding communities to aid in policy de

PAGE 79

79 sign. Social ecologists at SANParks recognize the important role that vulnerability has within the social-ecological system and are attempting to see how it is impacting local communities. The goal of such efforts is to ensure that policies will either aid in the reduction of vulnerability or The high population, over 1.5 million distributed over 120 villages (SANParks 2000), diverse department objective, and tight budget place a major constraint on the research and monitor ing that can realistically be achieved by the Social Ecology Department. While the research and monitoring shortfall can be mitigated by collaborating with outside researchers and institutions, as the research focus is largely set by the organization conducting and funding the research. This places an emphasis on using and creating instruments that can maximize information, while minimizing time and money requirements. As vulnerability is only one of a multitude of issues that Social Ecologists need to investigate and monitor, it is critical to establish a method that can be attached to surveys that may cover multiple topics, without placing an undo time burden on tool to establish household vulnerability to poverty should meet 3 criteria: Minimize the amount of time needed to collect data Reduce the quantitative burden to lower time burden and ensure capacity for data analysis While several econometric methods have been established for estimating vulnerability, each has major limitations in relation to the 3 points above. Two methods, vulnerability as uninsured exposer to risk (VER) and vulnerability as low expected utility (VEU), require panel or pseudo-panel data; and while vulnerability as expected poverty (VEP) has been adjusted to

PAGE 80

80 information in terms of household characteristics, shock exposure, and are most often quanti non-food, and durable goods a household consumed over a given period. While such data would provide the greatest accuracy in terms of identifying vulnerable households, it is extremely time consuming and costly. Many studies that have investigated vulnerability used pre-existing large scale data sets, such as the World Banks ongoing Living Standards Measurement Study (LSMS). The LSMS surveys vary in length, but the essential core is comprised of 11 modules over countless pages that require multiple-hour household interviews often over several visits (Grosh, et al. 2000). Additionally, while the datasets are large, often well exceeding 2,000 households, the samples are only adequate at the national level and not at the provincial or sub-provincial levels (Grosh & datasets are invaluable; however theyare only available for 32 countries. Furthermore, for KNP having a dual use survey. The survey designed and conducted that formed the base of building a short form vulnerability model for this study underpins this point. The survey had 5 modules consisting of a total of 1,521 items, with 916 items used to construct an aggregate consump tion index to measure welfare. While every question was not answered by each household, the survey still placed a major time constraint on the household and the survey enumerator, with a mean time of about 50 minutes per household with a max time of 132 minutes and a minimum of 19 minutes. It is also important to note that the wealthier the household the longer it took to

PAGE 81

81 complete the survey given a greater consumption of food, non-food items, and durable goods in wealthier households. While conducting a similar survey within most of the communities would provide SANParks a detailed and accurate picture of vulnerability, the time and money make it largely infeasible. Objectives tists to identify vulnerable households. The study uses vulnerability estimates from the econo part of this effort was used in previous work that looked at vulnerability in communities border ing KNP and how vulnerability impacts local resource use decisions. This previous work was used to inform the selection of key variables to include in the short model, based on both the impact it has on local vulnerability and the time and effort it took to collect that particular item. ing items used in model construction to ensure SANPark can collect reliable data while minimiz ing cost. Study Site After consultation with KNP Social Ecologists, two areas were selected that were thought to capture the broad differences that exists along the border of KNP (Figure 2-1). To the north, and Mutale B. In the south, Makoko was selected as the sampling community. These areas differ livelihood structure.

PAGE 82

82 Communities in the north tend to be much smaller than southern communities. In the north, Mutale B was the largest at 187 total households, followed closely by Bende Mutale at 174 at 71 households and Beleni with only 23 households. On the other hand Makoko had over 1,100 households. Makoko had the additional advantage of being within 45 km of three sizable cities of Hazyview, White River, and Nelspruit, with populations of about 36,000, 18,000 and 95,000 respectively. Along with access to semi-regular low-fair buses, good roads and electricity, the infrastructure and market access is much better in the south. Comparatively the Mutale area is about 100 km away from their nearest sizable city of Thohoyandou (pop. 39,513), with infre quent transportation over poor roads. While Mutale B has had electricity for several years, Bende access to electricity. Poverty is a major concern throughout Southern Africa, and this is no different in the two provinces in which Kruger lays, Mpumalanga and Limpopo. A study conducted in 2007 and us ing a poverty line of R 322 per adult equivalent per month, found that 45% of all South Africans lived below the poverty lines, while the 47% of all Mpumalanga households were in poverty (PROVIDE 2009a). Poverty is much more of an issue in Limpopo with of 67% of households below the poverty line (PROVIDE 2009b). Data from this study established that 55% of the sampled households in the northern communities were below the poverty line, while 45% of the sampled households in Makoko were below the poverty line established at R 539 per adult equiv alent per month (Oosthuizen 2008). As poverty and vulnerability has been linked to increased reliance on local natural resources, it is not surprising that 51.9% of all households in Limpopo

PAGE 83

83 are smallholder, non-commercial producers, while in Mpumalanga 25.2% of households are noncommercial producers (Directorate of Agricultural Statistics, 2008). While these households do not rely exclusively on agricultural items, diversifying into agriculture and natural resource extraction activities are a key risk mitigation strategy that need to be recognized and monitored to ensure policies do not hinder household risk management strategies without offering alternatives. Table 4-1 gives a breakdown on livelihood activities in which households participate for the Mutale area and Makoko. It is clearly apparent that house holds in the north have greater livelihood diversity and greater reliance on natural resources. Much of the agricultural production in Makoko was centered on Maize, while the Mutale area was much more diverse: four crop types being grown by over 30% of the households. Further more, there is greater participation in the north in natural resource collection and the livestock ownership. Methodology Data Collection Data for this research was collected from June 2009 to June 2010 via household oral ques tionnaires conducted in the six communities discussed above. Key informants were consulted with prior to pre-testing to improve question design and ensure relevant variables were included in the instrument. The survey was then piloted in 26 randomly selected households, from which of 13 pages, 1,521 items and 5 modules addressing the following (Appendix A): househol d demographics; access to water, market and health services;

PAGE 84

84 detailed income, which looked at both the household production of products and formal and non-formal employment; comprehensive consumption, including questions on household food, non-food, and du rable good consumption; and shocks and co ping strategies. Sampling A random sample was drawn from village lists in the northern communities. For the most part, each community had a preexisting list with the exception of Tshikuyu for which a list was given its larger size and lack of any household list. Time constraints made it infeasible to create and verify a household list given the number of Makoko households. The Makoko Traditional Makoko residents. To solve this issue, a sampling frame was established using a modifyed veg etation transect approach Interestingly Makoko has very regular street grid layout where each household has road frontage. Each street segment was numbered and randomly selected. The side an intersection was reached, at which point another segment was selected. This continued until the pre-determined sample size was reached. The total number of households surveyed was 489: 89 in Bende Mutale; 51 in Tshikuyu; 71 in Duluthulu; 20 in Beleni; 92 in Mutale B; and 166 in Makoko. Data Analysis Two methods were attempted and tested for the creation of a short form vulnerability mea sure. First, relevant variables were entered into factor analysis both to eliminate unnecessary and

PAGE 85

85 to statistically build an index. The normalized variables that loaded above a predetermined level where used to build the sub-indices. Each of the sub-indices were then placed into a regression model using the vulnerability score calculated using the long and robust VEP model to calibrate the sub-indices weights to the VEP vulnerability score. A second set of models were built without Before the establishment of the short-form model or the running of the VEP model, 125 observations were randomly selected and excluded from the dataset. The remaining 364 were used to create and calibrate the short form (hereafter called SFVP) model. Once established, the SFVP was tested on the remaining 125 households. As it is ultimately attempting to capture vulnerability to poverty, the same as VEP, accuracy was judged by comparing the SFVP to VEP. This approach to model calibration and testing is often used in remote sensing where research the actual land cover. Once the model is established, it is run on a much larger spatial scale (see Mertens & Lambin 2000 and Bockstael 1996 for a more in-depth discussion). Ultimately, this or often impossible, to know the true vulnerable state a household is currently in. Hence, most studies opt for the more robust and data intensive VEP measure. Again, this work is primarily concerned with the establishment of an instrument allowing for a rapid assessment of vulnerabil ity. As such, for the remaining 125 households, vulnerability is calculated using VEP and SFVP, model as compared to the VEP model.

PAGE 86

86 Factor analysis and deemed by researchers in this study to minimize time a participant would take to answer the item, were entered as inputs to factor analysis (FA). This method reveals the structure of the data that best explains the variance. This study utilized the FA output to reduce unnecessary variables and to reveal different dimensions of household vulnerability. The FA method employed was the principle axis method as it is preferred to the maximum likelihood method when data violates assumptions of normality (Fabrigar et al. 1999). While the Kaiser criterion is commonly used in social sciences for the decision on how many factors (dimensions) to keep, this study utilized parallel analysis as it has been found to be far more accurate (Zwick & Velicer 1986). This sug gested a 10 factor solution. These factors were than rotated using an oblique promax approach to improve interpretability and simplify the data. The study used a predetermined factor loading cut-off of 0.32 for each item (Tabachnick & Fidell 2001), i.e. all items on an individual factor that loaded 0.32 or above would be aggregated to comprise that dimension. Originally 23 vari led to an unstable factor structure with several factors having no item loadings of greater than 0.32. As such, variables were removed that did not load on any factor, as suggested by Costello and Osborne (2005), resulting in the factor output given in Table 4-2. The 10 dimensions measuring vulnerability as suggested by FA are then constructed by sion is standardized household population and total children added together). Once summed, each of the dimensions is normalized to 1 to ensure each is comparable to the other. The normal ized factors were regressed against the vulnerability scores determined via the VEP method to es

PAGE 87

87 tablish dimension weights. As stated previously, the resulting model was then applied to the 125 cases withheld from the model creation process (this includes FA for determining dimensions) when using categorical variables (Costello & Osborne 2005). As such, using variables like gen der of household head, is not possible. This is one reason why researchers also investigated using a model based on only regression. Regression modeling Regression modeling was used to establish factor weights and to establish a vulnerability standard error, robust regression, and truncated linear regression. Additionally, logit, ordered logit, and probit models were evaluated but ultimately rejected as as they were not as accurate in predicting vulnerable households and are not as straight forward in interpretation. The VEP model To evaluate vulnerability, the study used the econometric method Vulnerability as Ex pected Poverty (VEP) that allows the researcher to establish the likelihood that a household will fall below or further below the poverty line in the future. Like all models, VEP has strengths and certainty in the future welfare level of a household. If the VEP model utilizes panel data, deriv ing the temporal nature of the model is possible through observed distributions of consumption. However, in this study, the inclusion of panel data was not possible due to budgetary and time constraints.

PAGE 88

88 utilized that Chaudhuri et al. (2002) determined could be used with cross-sectional (ver sus panel) data. The main limitation is that it uses the assumption that cross-sectional variability proxies inter-temporal variability. The model Chaudhuri established is: Vht= Pr( Ch t + 1= c ( Xh,t + 1,h, eh t + 1) z Xh,t ,heht) (1) where vulnerability (Vht) is the probability that a households welfare in a future time pe riod (Ch,t+1) will be below the poverty line, with Ch,t+1, a function of: household characteristics (Xh) such as location, household population, access to resources, shock experienced, etc.; Bt, a parameter vector describing the state of the economy; ah, a household-level time invariant effect; and eh, a disturbance term with a mean of zero. The future level of consumption is established by estimating the distribution of consumption for the sample through a three-step feasible least squares process applied to the households that hold similar characteristics (a detailed description and model derivation can be found in Chapter 3 and in Chaudhuri et al. 2002). For the purpose of the study, welfare was established through a detailed consumption index that incorporated food and non-food items that the household either purchased or produced for home consumption. In addition to this models advantage of estimation via cross-sectional data, it also allows research ers to derive a baseline vulnerability score for each household permitting the researcher to iden tify vulnerable households and evaluate factors that coincide with vulnerability. The VEP model is is used to calibrate SFVPs and judge accuracy. Results Selection of model variables While the survey from which this study is based off had, again, over 1,500 items only a handful were selected to be included (the full survey is given in Appendix A). Table 4-3 reports

PAGE 89

89 the variables considered for use in the SFVP models with selection based on ease of collection formation on household characteristics, social connections, access to markets and water, agricul ture and natural resource based livelihood activities, cash earning activities, and shock impacts and coping behavior. In many cases, the variable listed in the table and used in this study can be tion is an example of this. Consumption is considered a very reliable measure of household wel fare and used as such in most of the studies cited in the table. However, the gathering of accurate ity it is key that the study includes a measure of it in the study. As such, income is included in the form of secure wage, piece work, government grants and remittances. Essentially wage income is considered a proxy for consumption which itself is a proxy for welfare. However, an accurate measure of consumption required about 916 items on the survey (items needed to quantify con sumption) and using income measures reduces it to, at most, 47 questions, a huge savings in time and, ultimately, money. Previous papers derived from this data utilized household GPS points to integrate dis tance information in the construction of a market access index. However, it should not be taken for granted that potential users of the proposed models will have access to the software and tech nology for such analysis. As such, all spatially derived data were excluded. Market and water ac cess has been linked to vulnerability so it is important to have a proxy for those concepts. These are covered via waiting time and fee for transportation to major market towns and water queue between time and accuracy.

PAGE 90

90 Factor Analysis SFVP Model The rotate factor matrix is given in Table 4-2 and used as the bases for constructing the factors have variables load that over 0.32 that each deal with household population characteris factor. Household population loaded on the second factor, with total adult females loading the of schooling of household head which loaded negatively on the factor. As such, this variable was subtracted from the other variable that loaded on this factor, total yearly value of government grants received by the household. The fourth factor dealt exclusively with livestock, with total cattle owned by the house hold loading the highest, followed by livestock ownership diversity (sum of livestock species ties, the variable with the highest loading was natural resource collection diversity (sum of all natural resource products household collects). Climate related shocks impacting crops and live Factor six had variables load that are related to transportation infrastructure (transporta tion wait time and trip fare) which can indicate household access to markets and health facilities. The seventh factor had variables load that are related to shocks and coping mechanism, household has indicated it has access to. Coping strategy diversity loaded negatively and was subtracted from the other variables for the establishment of this dimension.

PAGE 91

91 Total household yearly secure wage loaded the highest on factor eight, with household spatial diversity (Sum of each village, town, and city in which a household has a family relative) also loading. Household spatial diversity loaded on factor nine along with total yearly remittance income. Finally total household income derived from piecework (short-term, insecure, and inter mittent wage work) loaded on factor ten. The normalized factor scores were entered into a regression to establish factor weights. Table 4-4 displays the results. While all variables are used in the SFVP model, not all the vari and negative, meaning an increase in household cattle ownership or a higher diversity of live cating that an increase in crop diversity, natural resource collection activity and climate shock wait time and / or the trip fare increases, so to does vulnerability. The same is true for factor 7, have higher vulnerability. Additionally, the greater the access to different coping mechanisms, 0.01), indicating that as household secure wage increases, vulnerability decreases. While most of literature suggests that as natural resource activity diversity increases so too should vulnerability as such activities tend to be risk mitigation strategies for vulnerable household. The same can be said of climate shocks on crops and livestock. However, this variable is an aggregation of three

PAGE 92

92 the regressions establish the factor weights for each of the 10 factors. Independent of the above process to determine the sub-dimensions of vulnerability and regression to determine weights, the 10 factors are constructed for the 125 cases randomly extracted from the original analysis. The derived SFVP model was used to calculate vulnerability by inputing the factors into the equations derived from the regression modeling, i.e. for calculating the SFVP vulnerability score using the truncated regression weights, the factors created from the 125 cases are inputed into the equation: SFVP = 37.98 + 0.15 (Factor1norm) + 0.68 (Factor2norm) 0.01 (Factor3norm) 0.43 (Factor4norm) 0.32 (Factor5norm) + 0.33 (Factor6norm) + 0.29 (Factor7norm) 0.68 (Factor 8norm) + 0.03 (Factor9norm) 0.05 (Factor10norm). The SFVP model output is judged for accuracy against the full VEP model. For each out 75th percentiles to see if there is similarity in relative rankings of vulnerable household. Identi fying the top 50th vulnerable households would be effective an effective approach to monitor VEP model). In the VEP model this 0.5 indicates a household has a probability of 0.5 of falling with 0.5 and above indicating vulnerable households. As the SFVP model is constructed to gather

PAGE 93

93 the same information, 50 should have the same basic interpretation, i.e. a 50% chance of falling into poverty. comparable, the model based on robust regression appears to score marginally better across all measure developed that goes beyond a basic proportion measure of agreement by adjusting the observed proportion of agreement by the agreement one would expect to see by chance (Cohen 1960). This measure is often used in land-use, land-cover studies to adjust remote sensing based models of land cover to actual observed ground values. There is no agreed upon point of accept ability, with many guidelines to interpret strength of agreement in the literature. Landis and Koch (1977) suggest that a value below 0.2 has a poor strength of agreement, while a value 0.21 to 0.4 displays a fair strength of agreement, 0.41 to 0.6 is moderate, 0.61 to 0.8 is good, and 0.81 to 1 this is a moderate strength of agreement. The percent agreement is the same, 74.4% for the 75th result may suggest a reduction in model accuracy for values toward the tails. While the percent slightly to 0.34. Regression Based Models without Factor Analysis Models were run and tested without the running of factor analysis to allow for the inclu sion of discrete variables and to let each of the individual variables interact without aggregation

PAGE 94

94 with other variables, hitherto refered to as SRVP. This may improve clarity of model output and allowing the researcher to assess the direct impact that each of the variable has on vulnerability. independent variables. It is important to note, that this process was done for each of the three as each of the variables in these models are not transformed and, hence, in different units. The variable would have on vulnerability. Standard deviations of each of the variables are given in population, an increase in household population by one standard deviation (2.30) would increase vulnerability by 16 points (note that the dependent variable is the probability of being in poverty multiplied by 100). Secure wage has the largest negative relationship, increasing secure wage by one standard deviation would reduce vulnerability by 8 points. Many of the results are similar to the factor analysis models. However, total children and age of household head are now negative ly related to vulnerability. It is important to note that the variable, total household government for each child under 18, up to a total of 6 children, while each individual over the age of 60 receives R1050 each month. So all things being equal, older households or households with more children receive a decent amount of Rand each month which undoubtedly mitigates vulnerability.

PAGE 95

95 for the SRVP regression-only model approach. It is not surprising that the percent agreement and Again, the robust regression technique illustrates the highest accuracy in correctly classifying households rated the same, with the kappa score indicating a moderate strength of agreement at correctly, attaining a kappa of 0.39 which is close to indicating a moderate strength of associa sion only models, the factor model approach tends to assign lower vulnerability scores than the regression-only approach to only classifying 19.8% as not vulnerable when VEP indicates they are vulnerable. Discussion This study presented a model for the estimation of vulnerability achieved via a short-form model that reduces the data burden. While kappa values often exceed 0.90 in land cover studies, that level of agreement in social science is unlikely. But attaining an agreement above 71% for

PAGE 96

96 has done is attempted to create a model to predict another model. Unfortunately it is not easy to observe the actual vulnerability level of a household as it is to observe the actual land cover of a particular pixel. Instead this study distilled a long and complicated process into a quick, easy, and accurate package that would put a new tool in the SANPark Social Scientist toolbox. grasp on vulnerability. Policies and management structures need to be designed without placing undue strain on households that have the lowest capacity to handle such strain. Previous papers derived from this research have highlighted the link between natural resource use and vulnerabil ity. The most vulnerable households along the boarder of KNP tend to rely more on natural re sources. To prevent amplifying vulnerability in these households, access needs to be ensured for these resources, or households given an alternative that reduces natural resource reliance. KNP has a vested interest in conserving the very resources that households require. Any future plan that could alter a communitys access risks pushing households into further destitution. Before need to identify who is vulnerable. lists the variables used to calculate each of the short-form models and the number of questions needed to gather reliable data to build the variable. The survey used in this study and as the basis of other studies contained over 1,500 variables. While not all the variables were used in data analysis to date, the vast majority were included. At best 100 150 items could have culled or streamlined. The factor-based SFVp model required a total of 19 variables which were construct ed using 131 questionnaire items, of which 79 were discrete choice based questions (a choice between one of two categories). This question type lowers the cognitive burden on the part of

PAGE 97

97 the participant and reduces survey administration time. For the regression-only SRVP model, 12 variables are needed that require 65 survey items, with 43 of those items being discrete choice. While both approaches illustrated high levels of agreement and could be applied easily in the questionnaire items of over 50%. As many of these questions are likely to be included as part of other investigations, the addition of the remaining items would place a minimal time burden on any survey. It is estimated that the 65 items would take a household at most 15 minutes to an swer, but the majority of the households will not need to respond to many of the questions if they do not pertain to their situation. While this model will do a good job at quantifying vulnerability in rural communities along the border of KNP, perhaps even in the rural areas of Limpopo and Mpumalange, it could undoubtedly be improved. Ideally vulnerability is calculated using panel data that would allow the researcher to directly observe shortfalls in household welfare. As each household was georeferenced for this study, it is hoped that they will be revisited in the future if funding permits. issue is the need for occasional re-calibration. Changing national and regional factors could fun damentally alter the vulnerability space of a household. The method layout in this paper can be applied every so often to re-calibrate the model and to get a more robust picture of vulnerability.

PAGE 98

98 Table 4-2. Factor analysis (N = 347). Table 4-2. Factor analysis (N = 347). Table 4-2. Factor analysis (N = 347). Table 4-2. Factor analysis (N = 347). Variable Fact. 1 Fact. 2 Fact. 3 Fact. 4 Fact. 5 Fact. 6 Fact. 7 Fact. 8 Fact. 9 Fact. 10 HHPOP 0.6534 0.4449 TOTCHILD 0.9095 ADULTFEMALE 0.7793 GOVGRANTYR -0.565 YRSEDUHHEAD -0.5678 LIVSTCKDIVS 0.6113 TOTCATTLE 0.6399 CROPDIVERS 0.3435 NRCOLLECT 0.4501 CLIMATSHOCK5YRS 0.4163 WAITRANS 0.6357 TRANSFEE 0.5721 ILLNDEATHSHCK5YR -0.4993 INFLATSHCK5YRS -0.4562 COPESTRACCESS -0.5716 TOTSECWAGE 0.5984 FAMCONNCT 0.3528 0.5105 RMITTANCVALYR 0.5098 WAGEPW 0.3745 Table 4-1. Household utilization of natural resources by study area. Table 4-1. Household utilization of natural resources by study area. Table 4-1. Household utilization of natural resources by study area. Activity Mutale Area (%) Makoko (%) Grew maize last year 39.8 71.3 Grew groundnuts last year 33.4 11.9 Grew melon last year 38.9 1.3 Grew beans last year 36.6 3.1 Grew Sorghum last year 18.5 0.6 Grew cassava last year 0.0 13.1 Grew vegetables last year 12.4 29.4 Collected natural resources in the last year194.9 69.4 HH owns cattle 26.1 22.5 HH own goats 33.8 15.0 HH owns other animal258.1 45.5 Consumed self produced animal products in the last year 35.4 41.9 Sold animal products last year 28.4 6.6 Sold natural resource products last year 13.3 9.0 At least one person employed in secure wage work 39.2 35.5 At least one person employed in piece work 28.7 17.5 Received remittances last year 16.7 18.1 Received government grants 79.7 87.4 % of households below the poverty line 54.2 44.71 Includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area. 1 Includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area. 1 Includes fuelwood, thatching grass, reeds, and all other natural resource collected from the local area. 2 Includes mainly chickens, pigs, and sheep. 2 Includes mainly chickens, pigs, and sheep. 2 Includes mainly chickens, pigs, and sheep.

PAGE 99

99 Table 4-3. Variables linked to vulnerability. Table 4-3. Variables linked to vulnerability. Variables Studies indicating relationship with vulnerability Household characteristics Household characteristics Total household population Ligon & Schechter (2003) vulnerability increases with household size / Dercon et al. (2005) larger households tend to be more vulnerable Total children in the household Christiaenesn & Boisvert (2000) found that increased children increased vulnerability / Dercon & Krishnan (2000) the greater the dependency ratio, the more vulnerable the household / (Dercon et al. 2005) see previous Total adult female in the household Kochar (1995) households with few males were more vulnerable to shocks Female is household head Christiaensen & Subbarao (2005) female headed HH were more vulnerable through risk exposer / Dercon & Krishnan (2000) see pervious Age of household head Ninno & Marini (2005) found that as age of the HH head increased, so too does the probability of experiencing shocks / Ligon & Schechter (2003) found vulnerability increases with age Years of education by the household head Chaudhuri et al. (2002) increased education reduces mean vulnerability / Dercon et al. (2005) found HH with heads that had schooling were less vulnerable / (Ligon & Schechter 2003) see previous Social connections Social connections Household is part of the majority ethnic group The authors opinion that increasing cultural similarity may increase the likelihood that a household will receive help from community members Sum of each village, town, and city in which a household has a family relative See chapter 3 Access to markets and water Access to markets and water Total time required to wait for transportation Christiaensen & Subbarao (2005) decreased market access (measured via distance) increase vulnerability / Chaudhuri et al. (2002) access to improved transportation infrastructure reduced vulnerability Cost per trip to major market town (Christiaensen & Subbarao 2005) & (Chaudhuri et al. 2002) see above Total wait time in water queue Chaudhuri et al. (2002) found that households with better access to water had lower mean vulnerability. Fee for water (Chaudhuri et al. 2002) see above Agriculture and natural resource based livelihood activities Agriculture and natural resource based livelihood activities Sum of all crop types grown by the household Francis (2002) the poor diversify to survive while the rich diversify to thrive Sum of the various natural resources a household collects Dercon & Krishnan (1996) households diversify into NR as it has low entry costs, while it reduces risk it is usually an activity of vulnerable groups / Francis (2002) the poor diversify to survive Sum of the various species of livestock owned by the household Shewmake (2008) households that own livestock have lower vulnerability than households without livestock / Ninno & Marini (2005) owning livestock reduced probability of experiencing shocks / (Kurosaki 1995; Rosenzweig & Wolpin 1993) see previous Total cattle owned by the household Christiaensen & Subbarao (2005) found cattle ownership reduced vulnerability for households in arid & semiarid areas / (Ninno & Marini 2005) see previous / (Kurosaki 1995; Rosenzweig & Wolpin 1993) see previous

PAGE 100

100 Table 4-3. Continued Table 4-3. Continued Variables Studies indicating relationship with vulnerability Cash earning activities Cash earning activities Total yearly value of household secure wage Christiaensen & Subbarao (2005) Off-farm wages reduce risk & hence vulnerability / (Ninno & Marini 2005) see previous / Kochar (1995) shocks are mitigated through diversifying into wage work / Ligon & Schechter (2003) employment reduced vulnerability Total yearly value of household piece-work wage (Christiaensen & Subbarao 2005) & (Kochar 1995) & (Ligon & Schechter 2003) see above Total yearly value of government grants received by the household Ligon & Schechter (2003) found that government pensions reduced vulnerability / Dercon (2002) found that public safety nets can aid vulnerable households Total yearly value of remittances received by household Christiaenesn & Boisvert (2000) found that remittance income reduced vulnerability Shock impacts and coping behavior Shock impacts and coping behavior Total death or significant illness of a household member in last 5 years (self reported) Christiaensen & Subbarao (2005) illness increased vulnerability / Hoogeven (2005) found that household disablement increase vulnerability / (Dercon et al. 2005) see previous / Total inflation related events impacting the household in last 5 years (self reported) Heitzmann et al (2002) inflation has a detrimental impact on households Total of climate related shocks impacting crops or livestock in last 5 years (self reported) Harrower & Hooddinott (2005)Loss of cattle harmed households / Dercon et al. (2005) drought increased vulnerability / Total crime related events impacting the household in last 5 years (self reported) Most often not found significant, see Dercon et al. (2005) Diversity of coping behavior household has access to Kochar (1995) shocks are mitigated through diversifying into wage work

PAGE 101

101 Table 4-4. Regression models for determining factor weights (N = 347). Table 4-4. Regression models for determining factor weights (N = 347). Table 4-4. Regression models for determining factor weights (N = 347). Table 4-4. Regression models for determining factor weights (N = 347). Table 4-4. Regression models for determining factor weights (N = 347). Table 4-4. Regression models for determining factor weights (N = 347). Table 4-4. Regression models for determining factor weights (N = 347). Dependent variablevulnerability from VEP model1 M1 M1 M2 M2 M3 M3 Dependent variablevulnerability from VEP model1 Truncated regression Truncated regression Robust regression Robust regression OLS (robust) OLS (robust) Variable Coef. (St. Err.) Coef. (St. Err.) Coef. (St. Err.) Coef. (St. Err.) Coef. (St. Err.) Coef. (St. Err.) Factor1norm -0.1540 -0.0844 -0.1074 Factor1norm (0.1673) (0.1279) (0.1191) Factor2norm 0.6783*** 0.6783*** -0.5451*** -0.5043*** Factor2norm (0.1369) (0.1064) (0.0970) Factor3norm -0.0076 -0.0084 -0.0081 Factor3norm (0.0809) (0.0621) (0.0607) Factor4norm -0.4295*** -0.3266*** -0.3145*** Factor4norm (0.1306) (0.0870) (0.0914) Factor5norm -0.3219*** -0.2486*** -0.2398*** Factor5norm (0.0774) (0.0593) (0.0570) Factor6norm -0.3326*** -0.2653*** -0.2437*** Factor6norm (0.0744) (0.0542) (0.0544) Factor7norm -0.2902*** -0.2217*** -0.2121*** Factor7norm (0.0994) (0.0737) (0.0736) Factor8norm -0.6818*** -0.5143*** -0.5005*** Factor8norm (0.1105) (0.0861) (0.0696) Factor9norm -0.0276 -0.0331 -0.0327 Factor9norm (0.1387) (0.1086) (0.1049) Factor10norm -0.0498 -0.0286 -0.0407 Factor10norm (0.1044) (0.0800) (0.0752) Constant 37.98*** 40.90*** 41.25*** Constant (6.192) (4.898) (4.665) R2 -0.321 -0.314 -0.321 *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate.1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate.1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate.1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate.1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate.1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate.1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100.

PAGE 102

102 Figure 4-1. Level of agreement between VEP and SFVP by 50th & 75th percentile groups and .50 level of vulnerability for factor based models (N = 122). 0 20 40 60 80 Vuln in SFVP Same Vuln in VEP 13.2 73.6 13.2 Truncreg Vuln in SFVP Same Vuln in VEP 12.4 74.4 13.2 Rreg Vuln in SFVP Same Vuln in VEP 13.2 73.6 13.2 OLR Robust Vuln in SFVP Same Vuln in VEP 32.2 63.6 4.1 OLS Robust SFVP in Short Same Vuln in VEP 31.4 64.5 4.1 Rreg 0 18 35 53 70 Vuln in SFVP Same Vuln in VEP 32.2 63.6 4.1 Truncreg 0 20 40 60 80 Vuln in SFVP Same Vuln in VEP 12.4 75.2 12.4 Truncreg Vuln in SFVP Same Vuln in VEP 12.4 74.4 13.2 OLS Robust Vuln in SFVP Same Vuln in VEP 12.4 74.4 13.2 Rreg Kappa0.32 Kappa0.47 Kappa0.49 Kappa0.32 Kappa0.34 Kappa0.34 Kappa0.47 Kappa0.34 Kappa0.32 Classification by 75th percentile groups Classification by 50th percentile groups Classification by .50 vulnerability score

PAGE 103

103 Table 4-5. Regression-only based models (N = 347). Table 4-5. Regression-only based models (N = 347). Table 4-5. Regression-only based models (N = 347). Table 4-5. Regression-only based models (N = 347). Table 4-5. Regression-only based models (N = 347). Dependent variablevulnerability from VEP model1 RM1 RM1 RM2 RM2 RM3 RM3 Dependent variablevulnerability from VEP model 1 Truncated Regression Truncated Regression Robust Regression Robust Regression OLS (Robust) OLS (Robust) Variable Coef. (St. Err.) Stand. Coef. Coef. (St. Err.) Stand. Coef. Coef. (St. Err.) Stand. Coef. HHPOP (St. Dev., 2.30) --8.0674*** -18.587 --6.9524*** 16.017 6.3823*** 14.704 HHPOP (St. Dev., 2.30) -(1.0212) -18.587 -(0.6908) 16.017 -(0.7406) 14.704 TOTCHILD (St. Dev., 1.47) --5.0974*** --7.475 --4.7156*** -6.915 --4.1281*** --6.053 TOTCHILD (St. Dev., 1.47) -(1.2741) --7.475 -(1.0672) -6.915 -(1.0050) --6.053 AGEHHEAD (St. Dev., 17.38) --0.1679** --2.917 --0.1451** -2.521 --0.1362** --2.366 AGEHHEAD (St. Dev., 17.38) -(0.0765) --2.917 -(0.0634) -2.521 -(0.0610) --2.366 LIVSTCKDIVS (St. Dev., 1.06) --5.6351*** --6.000 --4.5957*** -4.893 --4.4174*** --4.703 LIVSTCKDIVS (St. Dev., 1.06) -(1.3312) --6.000 -(1.0591) -4.893 -(1.0257) --4.703 CROPDIVERS (St. Dev., 2.23) --3.1626*** --7.040 --2.6083*** -5.806 --2.538*** --5.650 CROPDIVERS (St. Dev., 2.23) -(0.5883) --7.040 -(0.4831) -5.806 -(0.4487) --5.650 WAITRANS (St. Dev., 62.40) --0.1097*** --6.848 0.0992*** -6.188 --0.0894*** --5.581 WAITRANS (St. Dev., 62.40) -(0.0217) --6.848 -(0.0165) -6.188 -(0.0171) --5.581 ILNDEATHSHCK5Y (St. Dev., 0.68) --3.832*** --2.616 4.0262** -2.748 --3.1466** --2.148 ILNDEATHSHCK5Y (St. Dev., 0.68) -(1.9303) --2.616 -(1.6110) -2.748 -(1.5786) --2.148 COPESTRACCESS (St. Dev., 1.08) --3.6403*** --3.939 --3.0093*** -3.256 --3.0969*** --3.351 COPESTRACCESS (St. Dev., 1.08) -(1.2021) --3.939 -(1.0207) -3.256 -(0.9705) --3.351 TOTSECWAGE (St. Dev., R18,701) --0.0006*** -11.362 --0.0005 -8.575 --0.0005*** --8.485 TOTSECWAGE (St. Dev., R18,701) -(0.0001) -11.362 -(0.0001) -8.575 -(0.00004) --8.485 RMITTANCVALYR (St. Dev., R6673) --0.0008*** --5.129 --0.0005 -3.380 --0.0006*** --3.765 RMITTANCVALYR (St. Dev., R6673) -(0.0003) --5.129 -(0.0002) -3.380 -(0.0002) --3.765 GENDHHEAD (St. Dev., 0.49) --6.938*** 3.416 --5.1387** -2.530 --5.9087*** --2.909 GENDHHEAD (St. Dev., 0.49) -(2.5948) 3.416 -(2.0862) -2.530 -(2.0509) --2.909 DOMETHNGRP (St. Dev., 0.37) -12.9736*** --4.745 -11.3861*** -4.164 -10.4567*** --3.824 DOMETHNGRP (St. Dev., 0.37) -(3.5397) --4.745 -(2.9592) -4.164 -(2.8004) --3.824 Constant -51.42*** -50.73*** -51.48*** Constant -(6.927) -(5.682) -(5.618) R2 --0.458 --0.458 --0.465 --0.465 --0.459 --0.459 *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100. *** Significant at the 1% level; ** Significant at the 5% level; Significant at the 10% level. Robust standard errors are reported in parentheses under each Coefficient estimate. 1 The VEP model calculates the probability of falling below the poverty line in the future for each household 0 to 1, to improve interpretability this is multiplied by 100.

PAGE 104

104 Figure 4-2. Level of agreement between VEP and SRVP by 50th & 75th percentile groups and .50 level of vulnerability (N = 122). 0 20 40 60 80 Vuln in SRVP Same Vuln in VEP 10.7 77.7 11.6 Truncreg Vuln in SRVP Same Vuln in VEP 11.6 76.9 11.6 Rreg Vuln in SRVP Same Vuln in VEP 10.7 77.7 11.6 OLS Robust Vuln in SRVP Same Vuln in VEP 11.6 76.9 11.6 OLS Robust 0 20 40 60 80 Vuln in SRVP Same Vuln in VEP 10.7 77.7 11.6 Truncreg Vuln in SRVP Same Vuln in VEP 10.7 78.5 10.7 Rreg 0 20 40 60 80 Vuln in SRVP Same Vuln in VEP 19.8 70.3 9.9 Truncreg Vuln in SRVP Same Vuln in VEP 19.0 71.1 9.9 OLS Robust Vuln in SRVP Same Vuln in VEP 18.2 71.1 10.7 Rreg Kappa0.39 Kappa0.55 Kappa0.54 Kappa0.42 Kappa0.39 Kappa0.38 Kappa0.55 Kappa0.41 Kappa0.38 Classification by 75th percentile groups Classification by 50th percentile groups Classification by .50 vulnerability score

PAGE 105

105 Table 4-6. List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data. Table 4-6. List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data. Table 4-6. List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data. Table 4-6. List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data. Table 4-6. List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data. Table 4-6. List of variables needed to quantify each SFV model and questions needed on survey instrument to collect necessary data. Variables Variable use in model Variable use in model Maximum items needed on survey Number of yes / no items1 Variables Factor based SFV Regr. only SFV Maximum items needed on survey Number of yes / no items1 Total household population X X 1 Total children in the household X X 1 Total adult female in the household X 1 Age of household head X 1 Female is household head X 1 1 Household is part of the majority ethnic group X 1 1 Years of education by the household head X 1 Sum of each village, town, and city in which a household has a family relative X 11 1 Total time required to wait for transportation X X 1 Cost per trip to major market town X 1 Sum of all crop types grown by the household X X 8 8 Sum of the various natural resources a household collects X 7 7 Sum of the various species of livestock owned by the household X X 6 6 Total cattle owned by the household X 1 Total yearly value of household secure wage X X 13 1 Total yearly value of household piece-work wage X 20 1 Total yearly value of government grants received by the household X 9 1 Total yearly value of remittances received by household X X 7 1 Total death or significant illness of a household member in last 5 years (self reported) X X 18 18 Total inflation related events impacting the household in last 5 years (self reported) X 3 3 Total of climate related shocks impacting crops or livestock in last 5 years (self reported) X 15 15 Diversity of coping behavior to which the household has access X X 7 7 1 Not necessarily yes / no, indicates a discrete question that is a choice between 2 categories that is often phrased yes / no, but could be male female for example. 1 Not necessarily yes / no, indicates a discrete question that is a choice between 2 categories that is often phrased yes / no, but could be male female for example. 1 Not necessarily yes / no, indicates a discrete question that is a choice between 2 categories that is often phrased yes / no, but could be male female for example. 1 Not necessarily yes / no, indicates a discrete question that is a choice between 2 categories that is often phrased yes / no, but could be male female for example. 1 Not necessarily yes / no, indicates a discrete question that is a choice between 2 categories that is often phrased yes / no, but could be male female for example. 1 Not necessarily yes / no, indicates a discrete question that is a choice between 2 categories that is often phrased yes / no, but could be male female for example.

PAGE 106

106 CHAPTER 5 CONCLUSION The study highlights that market access is a multidimensional concept going beyond a simple spatial relationship. While reducing time to market towns will increase access to house nancial instruments, and insurance. To access these items, households must have the capacity to utilize and access these resources. Improving market access within any of the dimensions will re duce household utilization of natural resources. However, the decision to own more cattle behave differently. Here, it is only the improvement in household capacity that results in a reduction in cattle numbers. Why is this? Cattle serves a drastically different purpose in the rural livelihood system then natural resources utilized for immediate consumption or sale, such as thatching grass, fuelwood, and mopani worms. While cattle is occasionally consumed, the primary purpose of cattle is as an insurance mechanism that can be leveraged (through sales) in times of stress to instruments. Accessing these products requires both exposer to the cash economy and education to navigate and understand these complex products. Given that these two variables comprise the household capacity sub-dimension, it follows that these households are able to shift away from cattle to other market goods. Households without adequate capacity are left to insure welfare via cattle ownership. As with household use decisions, improvements in market access can reduce household vulnerability. Improving the exposer to coping mechanism and other risk mitigation strategies that markets provide is key in reducing vulnerability. While reducing time and effort in travel ing to market towns results in a decline in vulnerability, one cannot ignore the impacts of the

PAGE 107

107 other sub-dimensions. Improvements in road infrastructure remains a key developmental strat egy in many countries. However, policies should also attempt to leverage the other components of market access that also show reductions in vulnerability or gains in consumption. Improving to vulnerability. Any potential policy intervention aimed at improving livelihoods, must understand the dynamic nature of livelihoods. As illustrated in this study and previous research, households do not make choices only to maximize income or minimize cost. In an environment of constant shocks they attempt to minimize risk through their household production consumption deci sions. If a policy does not recognize this dynamic and promotes and activity or a strategy that harms a households coping behavior without providing an alternative coping mechanism, the in tervention could have the unintended consequence of increasing vulnerability, potentially reduc ing future household welfare. Example, if a community tourism operation were promoted in an area and this operation needed land to provide wildlife habitat, households may no longer have livestock raising, this possesses a potentially serious problem if these households were no longer able to access livestock assets to mitigate a shock. Any policy needs to work to understand the multi-dimension nature of livelihoods to ensure that interventions either do not harm the risk me diating nature of livelihood choices or improves the capacity of the local system to mediate risk.

PAGE 108

108 APPENDIX A SURVEY INSTRUMENT Section A: Participant / Household Demographics 1) (Enumerator, check the gender of the survey participant ) 2) What year where you born? 3) What is your ethnic group? 4) Are you the head of the household? 4a) What is the gender of the household head? Section B: Water, Market & Health Services Access 1) Do you have a water pipe in your home? 2) If no to 11 where do you most collect water from? 3) Does this water source change throughout the year for at least 3 months? 6) Do you have to pay a fee to use this water source? Survey #: Surveyor: GPS Point Code: Village: Time Start: Time End:MALE (0)___ FEMALE (1)___ 19___ Dont Know (99)___ ____________________ Dont Know (99)___ Yes (1)___ No (0)___ __________ (The following 5 questions refer to family members who live within the house or compound for at least 3 months of the year) 6) How many total people live in this household?__________ 7) How many females? _____ 8) How many females under 16? _____ 9) How many males? _____ 10) How many males under 16? _____Yes (1)___ No (0)___ River (1)___ Public Borehole (2)___ Enumerator, please note the location of most used water source and assign a GPS point code GPS Point Code _____________ Yes (1)___ No (0)___4) How many trips do members of your family make per day? ________ 5) How long does it take to walk to the water source? _______(minutes)Yes (1)___ (Amount $________ ) No (0)___7) Where do you go to do the majority of your food shopping? _______________________ 8) If outside your village, how often do you travel in a typical month to make purchases?________ 9) Where do you go to do the majority of your non-food (clothing, pots, etc) shopping?_______________________________ 10) If outside your village, how often do you travel in a typical month to make purchases?________ 11) Where is your closest clinic?_____________ Hospi tal?_______________________ 12) If outside your village, how often did you have to travel in the last year to attain medical services?________ 13) How do you travel to the above destinations? Public Transport (1)_____ Personal Vehicle (2)_____ Vehicle of a family friend (3)_____ Government Vehicle (4)______ 14) How long did you have to wait for the transportation to your destina tion?________ 15) How much did you pay?__________Household Name:__________________________ MALE (0)___ FEMALE (1)___

PAGE 109

109 1) Did you grow maize in the last year? Yes (1)____ No (0)____ 2) # of bags_________ 3) Hectares planted ______ 4) Male____ Female____ 5) Did you grow sorghum in the last year? Yes (1)____ No (0)____ 6) # of bags_________ 7) Hectares planted ______ 8) Male____ Female____ 9) Did you grow beans in the last year? Yes (1)____ No (0)____ 10) # of bags_________ 11) Hectares planted ______ 12) Male____ Female____ 13) Did you grow watermelon in the last year? Yes (1)____ No (0)____ 14) # of bags_________ 15) Hectares planted ______ 16) Male____ Female____ 17) Did you grow groundnuts in the last year? Yes (1)___ No (0)___ 18) # of bags_________ 19) Hectares planted ______ 20) Male____ Female____ 21) Did you grow ______________ in the last year? Yes (1)___ No (0)___ 22) # of WM_________ 23) Hectares planted ______ 24) Male____ Female____ 25) Did you grow ______________ in the last year? Yes (1)___ No (0)___ 26) Yield___ ________ 27) Hectares planted ______ 28) Male____ Female____ 29) Did you grow ______________ in the last year? Yes (1)___ No (0)___ 30) Yield___ ________ 31) Hectares planted ______ 32) Male____ Female____ 33) How many cattle do you own? # _____ 34) # sold last year____ 35) # eaten last year____ 36) Male____ Female____ 37) How many goats do you own? # _____ 38) # sold last year____ 39) # eaten last year____ 40) Male____ Female____ 41) How many chickens do you own? # ______ 42) # sold last year____ 43) # eaten last year____ 44) Male____ Female____ 45) How many _________ do you own? # ______ 46) # sold last year____ 47) # eaten last year____ 48) Male____ Female____Livestock Agricultural Production (Enumerators, please tell interviewees that estimation is OK / Read questions across) 49) How many bundles of did you collect in the last month? #________ 50) # sold last year ____ 51) # sold to tourism businesses _____ 52) Male____ Female____ 53) How many bundles of thatching grass did you collect in the last 12 months? #________ 54) # sold last year ____ 55) # sold to tourism businesses _____ 56) Male____ Female____ 57) How many bundles of reeds did you collect in the last 12 months? #________ 58) # sold last year ____ 59) # sold to tourism businesses _____ 60) Male____ Female____Natural Resources Roughly, how many labor days did your family spend in each activity? 61) Do you make any type of traditional beer / alcohol? Yes (1)____ No (0)____ 62) buckets sold last year #_______63) Price? $_______ 64) Male____ Female____ 65) For each bucket, how much water is re quired?____________________________ 66) How much sorghum? _________________ 67) How much sugar? ______________________ 68) How much does it cost? _____________ 69) How many _____________ did you make in the last 12 months? #_______ 70) # sold last year ____ 71) Tourism Sales #_____ 72) Price? $_______ 73) Male____ Female____ Value Added Enterprises 74) Do you own any land? Yes (1)___ No(0)___ 75) If yes to 74, how much land do you own? ________Hectares 76) Do you rent land? Yes (1)___ No(0)___ 77) if yes to 76, how much land do you rent? _______Hectares 78) How much did you pay last year? R_____ 79 Do you have use of any other land? Yes (1)__ No(0)__ 80) if yes to 79, how much land do you use? ______Hectares 81) Do you irrigate? Yes (1)__ No(0)__Production and Income

PAGE 110

110 82) Are there any individuals living within the household who earns a wage? Yes (1)____ No (0)____ Relation to interviewee Work Location Position Wage per Month 1 R 2 R 3 R 95) Does anyone do piece work in the household? Yes (1)____ No (0)____ Relation to interviewee Position & Location # of Months Worked Approx. Wage per Month R R R R RIncome Sources 119) Does this household receive any other type of money or assistance from the Government? Yes (1)____ No (0)____ 120) If yes to 117 what type of assistance?(1)___________________(2)____________________(3)______________________(4)_______________________ 121) How much per month? 1)___________________(2)____________________(3)______________________(4)_______________________ 116) Did anyone living in another area, send money to this household? Yes (1)____ No (0)____ 117) If yes to 112 where does the person live who is sending money? __________________________ 118) How much per month does the person send? R_____________ Yes (1)____ No Yes (1)____ No (0)____ 123) Have there been any projects in your community funded through tourism income? Yes (1)____ No (0)____ 123) If yes, what?________________________ Wildlife (1)____ or Cattle (0)____ 125) Which land use is more sustainable? Wildlife (1)_____ or Cattle (0)_____ 126) Has the Kruger Park helped the community? Yes (1)____ No Yes (1)____ No (0)____

PAGE 111

111 Section C: Frequent Expenses2) What is the total amount of last weeks expenses for the purchased items? 4) What is the total value of meals eaten outside the home in the last 7 day? Item Yes No Currency Item Number Currency Cigarettes and other tobacco products Breakfast Newspapers / magazines Lunch Transportation expenses Dinner Other Daily Expenses (specify) Traditional beer Beer / Alcohol Snack and Beverages (not including alcohol or beer) 1) During the past 7 days have you or other household members purchased: 3) How many meals and snacks were eaten outside the household by household members within the last 7 days?

PAGE 112

112 Section D: Food Expense Purchases last 30 daysPurchases typical monthHome production Gifts2) Have members of your household bought any of the below items in the last 30 days? 3) How much did you pay in total? 5) How many months in the past 12 months did your household purchase food? 6) Is the amount spent in the last 30 days a normal month ( YES )? If not, how much do you spend in a normal month? 7) Typically, do you make the purchase of this item (1) in village, (2) out of village, but in district, (3) Out of district 8) In the last 12 months did your household consume (EATEN) food that you grew or produced at home? (number of months) 10) What was the value of the food you consumed in a typical month from your own production? 11) What is the total quantity of food given as gifts over the last 12 months? Food item No Yes Code Yes = 1 / No = 2 Currency Amt Unit Months Currency (1), (2), or (3) Months Amt Unit Currency Currency Bread 101 Maize (maizemeal) 102 rice 103 Sorghum 104 Groundnuts 105 melons 106 Cooking Oil 107 fresh milk 108 Baby formula 109 sugar 110 Beef 111 chicken 112 eggs 113 fish 114 vegetables 115 citrus fruits 116 caned fruits 117 beer 118 other alcoholic beverages 119 biscuits / cakes 120 tea 121 1) Has your household consumed (EATEN) the below food items over the last 12 months? Exclude items purchased for processing or resale. (Enumerator, please read off the item and if the interviewee answered YES continue across the column, if NO go to next item.) 4) How much did you buy? (QUANTITY) 9) How much did you consume in a typical month?

PAGE 113

113 Section D: Food Expense Purchases last 30 days Purchases typical month Home production Gifts 2) Have members of your household bought any of the below items in the last 30 days? 3) How much did you pay in total? 5) How many months in the past 12 months did your household purchase food? 6) Is the amount spent in the last 30 days a normal month ( YES )? If not, how much do you spend in a normal month? 7) Typically, do you make the purchase of this item (1) in village, (2) out of village, but in district, (3) Out of district 8) In the last 12 months did your household consume (EATEN) food that you grew or produced at home? (number of months) 10) What was the value of the food you consumed in a typical month from your own production? 11) What is the total quantity of food given as gifts over the last 12 months? Food item No Yes Code Yes = 1 / No = 2 Currency Amt Unit Months Currency (1), (2), or (3) Months Amt Unit Currency Currency coffee 122 Mopani Worms 123 Misc. other food: 124 Wheat 125 millet 126 127 firewood 128 charcoal 129 Paraffin 130 cooking gas 131 Electricity 132 Other: 133 134 135 136 137 138 139 1) Has your household consumed (EATEN) the below food items over the last 12 months? Exclude items purchased for processing or resale. (Enumerator, please read off the item and if the interviewee answered YES continue across the column, if NO go to next item.) 4) How much did you buy? (QUANTITY) 9) How much did you consume in a typical month?

PAGE 114

114 Section E: Non-Food Expenditure 12 Months 13) Have members of your household bought any of the below items in the last 30 days? 14) How much did you pay in total? 15) How much did your household spend for the items during the past 12 months? 16) Typically, do you make the purchase of this item (1) in village, (2) out of village, but in district, (3) Out of district 17) Did you receive any item as a gift during the past 12 months? 18) What is the value that you received as a gift during the past 12 months? Food item No Yes Code Yes = 1 / No = 2 Currency Currency (1), (2), or (3) Yes = 1 / No = 2 Currency men's clothing 201 Women's clothing 202 children's clothing 203 Cloth and sewing supplies 204 Tailoring expenses 205 cosmetics 206 personal care items 207 Traditional remedies 208 Over the counter remedies 209 Modern medicines 210 Health service fees 211 Books, stationary 212 Postal services 213 rentals 214 Household cleaning articles 215 Kitchen supplies 216 Electrical items 217 Repair and maintenance of households 218 Small kitchen appliances 219 Dishes / utensils 220 small electrical items 221 Toys 222 Sporting / hobby items 223 Vehicle repaired 224 Housing repair and maintenance 225 12) Has your household bought, spent money on or received gifts of any item below during the past 12 months? Exclude items purchased for processing or resale. (Enumerator, please read off the item and if the interviewee answered YES continue across the column, if NO go to next item.) Purchase 30 Days Gifts 12 Months

PAGE 115

115 Section E 12 Months 13) Have members of your household bought any of the below items in the last 30 days? 14) How much did you pay in total? 15) How much did your household spend for the items during the past 12 months? 16) Typically, do you make the purchase of this item (1) in village, (2) out of village, but in district, (3) Out of district 17) Did you receive any item as a gift during the past 12 months? 18) What is the value that you received as a gift during the past 12 months? Food item No Yes Code Yes = 1 / No = 2 Currency Currency (1), (2), or (3) Yes = 1 / No = 2 Currency Insurance (house, car) 226 Health Insurance 227 Donations / Charity 228 Income Tax 229 Land Tax 230 Housing / property tax 231 Cash loses 232 Deposits to savings accounts 233 Loan payments 234 Legal services 235 Marriage / birth / other ceremonies 236 Dowry 237Funeral insurance238 Funeral expenses 239 Other major expenses: 240 241 242 243 Farming Expenses: Fertilizer 244 Seeds 245 Cattle 246 Goats 247 Animal Feed 248 Animal Medicine / Vet service 249 Farm Labor 250 12) Has your household bought, spent money on or received gifts of any item below during the past 12 months? Exclude items purchased for processing or resale. (Enumerator, please read off the item and if the interviewee answered YES continue across the column, if NO go to next item.) Purchase 30 Days Gifts 12 Months

PAGE 116

116 Section F: Durable Goods21) How many years ago did you receive this item? 22) Was this time purchased by someone in your household (1) or was it given to you (2) ? 23) (If participant answered 1 in question 22) How much did you spent for it? 24) (If participant answered 2 in question 22) How much do you think it was worth? 25) If you were to sell it today, how much would you receive? ( Please estimate ) Item No Yes Code Item description Code Years ( 1) or (2) Currency ( 1) or (2) Currency Stove 301 Refrigerator 302 Washing Machine 303 Sewing Machine 304 Fan 305 Cell phone 306 Television 307 Video Player 308 Music player (tape / CD) 309 Film or video player 310 Bicycle 311 Motorcycle / Scooter 312 Car / Truck 313 Tractor 314 Plough 315 Other Main items: 316 317 318 319 Section G: Remittances Individuals Village / Town / City Currency Currency 1=yes/2=no 28) How much money has this household sent to this individual in the past 12 months? 29) What is the approximate value in cash of assistance given in food or other good? In the past 12 months? 30) Was this transfer in response to an illness or environment al shock? B. C. 19) Does your household own any of the following Items? (Enumerators, as yes or no for the below items and write all the yes answers in column 20)20) (List all the items below for which the respondent answered yes to the left, then proceed to question 3.) In the last 12 months has this household sent money or provided goods to people who do not live in this household? 26) What relation are these recipients of money to the household? (Please list below) A. 27) Where does this person reside?

PAGE 117

117 Section G: Shocks When Did these Occur?More than 3 year ago (list the year of greatest shocks)Shock Code Worst (C)A B A BDrought (lack of rain) 400 Too much rain 401 Erosion 402 Flooding 403 Timing of rain 404 Loss of livestock 405 Pest or diseases that affected crops before harvest 406 Pest or diseases that led to storage losses 407 Other (Specify): 408 409 410 Theft of tools or inputs for production 500 Theft of livestock 501 Theft of cash 502 Theft of stored crops 503 Destruction or theft of housing 504 Destruction or theft of consumer goods 505 Death of working adult household members 506 Death of other household members 507 Disablement of working adult household member 508 Disablement of other household members 509 Other (Specify): 510 511 Did These shocks result in : (1) Loss of household income (2) Loss of productive assets (3) A reduction in household consumption (4) No major change in above (Enumerators, write the letter for each of the above. It can be either one, all three or a combination of any 2) 2. Has there been civil conflict, banditry or crime shocks? Within the last {five years} and within the last 12 months, has this household been affected by a shock, an event that has resulted in a loss of household income, lead to a reduction in your assists, or significantly reduced household consumption? 1. Has there been any weather or climatic shocks? How widespread was this shock: (1) Only this household was affected (2) Affect some other households in this village (3) Affected all households in this village (4) Affected this village and some nearby villages (5) Affected the entire region (6) Affected the whole country Nature of Shock Shock SpecificsWithin the last 12 months (A) Between 1 and 3 years ago (B)C C

PAGE 118

118 Section G: Shocks When Did these Occur?More than 3 year ago (list the year of greatest shocks)Shock Code Worst (C)A B A BConfiscation of land 600 Confiscation of other assets 601 Land reform 602 Resettlement, villagization or forced migration 603 Bans on migration 604 Forced labor 605 Employment refusal based on social or ethnic reasons 606 Forced contributions or arbitrary taxation 607 Imprisonment for political reasons 608 Discrimination for political reasons 609 Discrimination for social or ethnic reasons 610 Contract dispute or default affecting access to land 611 Contract dispute or default affecting to other inputs 612 Contract dispute or default affecting sale of products 613 Other (Specify): 614 Lack of financing/capital 700 Lack of access to inputs 701 Increase in input prices 702 Decrease in output prices 703 Lack of demand or inability to sell agricultural products 704 Lack of demand or inability to sell nonagricultural products 705 Unemployment (loss of a job) 706 Inflation 707 Other (Specify): 708 709 4. Have there been economic shocks?Within the last 12 months (A) Between 1 and 3 years ago (B) 3. Have there been negative political, social or legal events? Nature of Shock Shock Specifics Did These shocks result in : (1) Loss of household income (2) Loss of productive assets (3) A reduction in household consumption (4) No major change in above (Enumerators, write the letter for each of the above. It can be either one, all three or a combination of any 2) How widespread was this shock: (1) Only this household was affected (2) Affect some other households in this village (3) Affected all households in this village (4) Affected this village and some nearby villages (5) Affected the entire region (6) Affected the whole country C C

PAGE 119

119 Section G: Shocks When Did these Occur?More than 3 year ago (list the year of greatest shocks)Shock Code Worst (C)A B A BDeath of husband 800 Death of wife 801 Other death (Specify relation ___________________________________) 802 Illness of husband 803 Illness of wife 804 Other illness (Specify relation ___________________________________) 805 Divorce 806 Abandonment 807 Disputes with extended family members regarding land 808 Disputes with extended family members regarding other assets 809 Other (Specify): 810 811 Pest or diseases that led to losses in livestock 411 Predation of livestock by wildlife 412 Other (Specify): 413 Other (Specify): 900 901 902 903 If yes to above, do you know what diseases harmed your livestock? (1)______________(2)______________(3)______________(4)______________ What were you're losses from each disease or total ( Cattle, Goats, Chickens)? (1) C__G__Ch__ (2)C__G__Ch__ (3)C__G__Ch__ (4)C__G__Ch__ If yes to above, do you know what animal harmed your livestock? (1)______________(2)______________(3)______________(4)______________ What were you're losses from each animal or total ( Cattle, Goats, Chickens)? (1) C__G__Ch__ (2)C__G__Ch__ (3)C__G__Ch__ (4)C__G__Ch__ Nature of Shock Shock Specifics Did These shocks result in : (1) Loss of household income (2) Loss of productive assets (3) A reduction in household consumption (4) No major change in above (Enumerators, write the letter for each of the above. It can be either one, all three or a combination of any 2)Within the last 12 months (A) Between 1 and 3 years ago (B)7. Have there been any other events or shocks we have not covered? 6. Have there been any events that have impacted livestock? 5. Have there been health and household cohesion events or shocks?C CHow widespread was this shock: (1) Only this household was affected (2) Affect some other households in this village (3) Affected all households in this village (4) Affected this village and some nearby villages (5) Affected the entire region (6) Affected the whole country

PAGE 120

120 21 122334455 6 7 8 11 10 9 14 17 18 19 20 15 16 12 13 22 23Code YearSold Items Consumption reduction Had help from individuals or organizationsAsked / Applied for a loan and mortgaged assetsApplied for a cash loan from a state bank Work Habits change Other, what?________________________ ____________________________________Most important 2nd most important 3rd most important Less than 1 year 1 to 3 years 3 to 5 years Never recoveredHad help from private businesses Had help from government organizations Had help from the family, neighbors, or friends Had help from NGOs or international entitiesMost important shocks / crises {Ask respondents to list shocks from the above list that has harmed their household the most. Use the code from the above list}Asked for a cash loan form work Sold other assets. What?________________ Reduce food consumption Stop consuming some products or services Mortgaged house or land Asked for a cash loan from a family member Asked for a cash loan from a friend Asked for a cash loan from a moneylenderWhat have been the main crises (shocks) that has impacted your household? What did your household do to compensate, resolve or address this loss of assets, loss of income and/or reduction in consumption? (Ask respondent to list five worst from the list used in the previous questions)? Other family members went to work Applied for a cash loan from a private bank Worked more, if already workingWhat did your household do to compensate or resolve this decrease or loss of income? {List the three most important activities in descending order of importance}Didn't do anythingHow much time did it take your household to the position you were in before the crisis? 1 Short term shock, sill recoveringSold animals Sold house or land Sold appliances, equipment, machines Sold the harvest in advance

PAGE 121

121 APPENDIX B BASE REGRESSION FOR VULNERABILITY ANALYSIS Table B-1. Base regression for vulnerability analysis (Dependent variable log(consumption)). Table B-1. Base regression for vulnerability analysis (Dependent variable log(consumption)). Table B-1. Base regression for vulnerability analysis (Dependent variable log(consumption)). Table B-1. Base regression for vulnerability analysis (Dependent variable log(consumption)). Table B-1. Base regression for vulnerability analysis (Dependent variable log(consumption)). Variables Coef. Std. Err. Description of Variables GEND_HH_HEAD -0.03*** -0.056 The household is headed by a female (1 = yes) HH_SIZE 0.07*** -0.012 Size of the household DEPND_RATIO 0.02*** -0.038 Ratio of dependents to total household population AGE_HH_HEAD 0.00*** -0.002 Age of household head DOM_ETHN_GRP 0.12*** -0.078 Households ethnic affiliation is the village majority (1 = yes) NORTH -0.24*** -0.074 Geographic location of village (1 = North) GOATS 0.02*** -0.007 Number of goats owned PIGS 0.01*** -0.016 Number of pigs owned LIVILHD_DIVERS 0.06*** -0.012 Sum of main sectors from which a HH derives consumption or income GOV_GRANT -0.29*** -0.078 Household received a government grant within the last year (1 = yes) ASSETS 0.00*** 0.000 Value in Rand of all household assets (does not include livestock) NRR -0.80*** -0.181 Total natural resources consumed from local area to total household consumption SH_RAIN -0.04*** -0.064 Experienced a rainfall shock in the last 12 months (1 = yes) SH_PESTDES_CL -0.03*** -0.100 Lost crops due to pest or disease in the last 12 months(1 = yes) SH_PREDDES_LVST_LOSS 0.15*** -0.065 Lost livestock from disease or predation in the last 12 months (1 = yes) SH_LOSS_HH_ASSTS 0.04*** -0.079 Lost household assets, animals or food due to theft in the last 12 months (1 = yes) SH_LACK_ACC_INP -0.06*** -0.069 Lack of access to inputs in the last 12 months (1 = yes) SH_INFLAT -0.18*** -0.065 Hurt by inflation in the last 12 months (1 = yes) SH_DEATH_HH -0.09*** -0.114 Death of household member in the last 12 months (1 = yes) SH_ILLN_HH -0.02*** -0.073 Illness of household member in last 12 months (1 = yes) SH_LAG_SHOCK_23 -0.04*** -0.022 Number of shocks experienced 2 to 3 years ago SH_LAG_SHOCK_35 -0.06*** -0.017 Number of shocks experienced 3 to 5 years ago Constant 9.74*** -0.147 Observations 470.06*** R-squared 0.39*** Adj. R-squared 0.36*** *** p<0.01, ** p<0.05, p<0.10 *** p<0.01, ** p<0.05, p<0.10

PAGE 122

122 APPENDIX C DESCRIPTIVE STATISTICS OF VARIABLES USED IN PROBIT MODELS Table C-1. Descriptive of variables used in the models. Table C-1. Descriptive of variables used in the models. Table C-1. Descriptive of variables used in the models. Table C-1. Descriptive of variables used in the models. Table C-1. Descriptive of variables used in the models. Table C-1. Descriptive of variables used in the models. Table C-1. Descriptive of variables used in the models. Variable Obs. Mean St. Dev. Min Max Description MAI_ SPI467 52.95 19.48 12.49 93.91 MAI place & infrastructure sub-II II IIindex MAI_SSC 474 22.36 14.38 0.00 88.43 MAI social connection sub-index MAI_SHC 474 31.85 23.50 0.00 86.84 MAI household capacity sub-index MAI_SFKA469 26.39 15.36 0.00 91.67 MAI financial knowledge & access II sub-index NORTH 474 0.66 0.47 0.00 1.00 Geographic location of village (1 = II IINorth) AGE_HH_HEAD 474 43.79 17.01 16.00 99.00 Age of household head DOM_ETHN_GRP 474 0.84 0.36 0.00 1.00 Households ethnic affiliation is the IIvillage majority (1 = yes) FEMALE_HH_HEAD 474 0.58 0.49 0.00 1.00 The household is headed by a female II(1 = yes) HH_SIZE 474 5.11 2.39 1.00 17.00 Size of the household DEPND_RATIO 474 0.81 0.78 0.00 5.00 Ratio of dependents to total II II II II IIhousehold population LIVILHD_DIVERS 474 4.79 2.92 0.00 15.00 Sum of products and services from II IIwhich a HH derives consumption II IIor income ASSETS1470 4691.05 8525.66 0.00 79380.00 Value in Rand of all household assets II(does not include livestock) HH_PERM_EMPLOY 474 0.38 0.49 0.00 1.00 HH has at least one person II II II II IIpermanently employed in the local IIarea (1 = yes) NR_USE_RATIO1474 0.15 0.15 0.00 0.79 Total natural resources consumed II IIfrom local area to total household IIconsumption HEADS_CATTLE1474 2.37 6.31 0 52.00 Total heads of cattle owned by the II IIhousehold 1 Dependent variable

PAGE 123

123 LIST OF REFERENCES Adger WN. 1999. Social vulnerability to climate change and extremes in coastal Vietnam. World Dev. 27(2):246-269. Alderman H, Hoddinott J. Kinsey B. 2003. Long-term Consequences of Early Childhood Malnu trition. Food Consumption and Nutrition Division Discussion Paper No. 168. Washington DC, International Food Policy Institute. Amemiya T. 1977. The maximum likelihood estimator and the nonlinear three-stage least squares estimator in the general nonlinear simultaneous equation model. Econometrica. 45:955968. Angelsen A, Kaimowitz D. 1999. Rethinking the causes of deforestation: lessons from economic models. World Bank Res Obs. 14(1):73. Aynsau A, De Graaff J. 2007. Determinants of adoption and continued use of stone terraces for soil and water conservation in an Ethiopian highland watershed. Ecol Econ. 61:294. within Namibias community wildlife use initiative. World Dev. 30(4):667-681. strategies in rural Africa: concepts, dynamics, and policy implications. Food Pol. 26:315331. Blaikie P, Cannon T, Davis I, Wisner B. 1994. At Risk: Natural Hazards, Peoples Vulnerability and Disasters, Routledge, London. Blarel B, Hazell P, Place F, Quiggin, J. 1992. The economics of farm fragmentation: Evidence from Ghana and Rwanda. World Bank Econ Rev. 6(2):233-254. Bockstael NE. 1996. Modeling economics and ecology: The importance of a spatial perspective. Am J Agr Econ. 78:1168-1180. Bolker BM. 2008. Ecological Models and Data in R. Princeton University Press : Princeton. Browne MW. 1968. A comparison of factor analysis techniques. Psychometrika. 33:267-333. factor analysis. In Grimm and Yarnold, Reading and understanding multivariate analysis. American Psychological Association Books.

PAGE 124

124 Bluffstone RA. 1995. The effect of labor market performance on deforestation in developing countries under open access: an example from rural Nepal. J Environ Econ Manag. 29:42 63. Chaudhuri S. 2003. Assessing Vulnerability to Poverty: Concepts, Empirical Methods and Illus trative Examples. The World Bank, Washington DC (mimeo). Chaudhuri S, Jalan J, Suryhadi A. 2002. Assessing Household Vulnerability to Poverty: A Meth odology and Estimates for Indonesia. Colombia University, New York. Child B. 1989. The Economic Use of Wildlife in Zimbabwe and its Effect on the Supply of Wild life Products. A paper presented at SADCC/GTZ workshop on Processing and Marketing of Wildlife Products in the SADCC region. Christensen G. 1989. Determinants of Private Investment in Rural Burkina Faso. Ph.D. [Disser tation], Cornell University. Christiaensen L, Boisvert R. 2000. On measuring Households Food Vulnerability: Case Evidence from Northern Mali. Working Paper, Department of Agricultural, Resource and Managerial Economics, WP 2000. Christiaensen LJ, Subbarao K. 2005. Towards an understanding of household vulnerability in rural Kenya. J Afr Econ. 14(4):520-558. Costello AB, Osborne JW. 2005. Best practices in exploratory factor analysis: Four recommenda tions for getting the most from your analysis. Practical Assess Res Eval. 10(7):1-9. Cruces G, Gasparini L, Bergolo M, Ham A. 2010. Vulnerability to Poverty in Latin America: Evidence from Cross-Section and Panel Data. Argentina, Chronic Poverty Research Centre, Univeridad Nacional de La Plata, Cutter SL. 1996. Vulnerability to environmental hazards. Progr Hum Geogr. 20: 529-539. Cutter SL, Boruff BJ, Shirley WL. 2003. Social vulnerability to environmental hazards. Soc Sci Q. 84(2):242-261.

PAGE 125

125 Cutter SL, Mitchell JT, Scott MS. 2000. Revealing the vulnerability of people and places: A case study of Georgetown County, South Carolina. Ann Assoc Am Geogr. 90(4):713. Dercon S. 2005a. Vulnerability: A Micro-perspective, paper presented at the Annual Bank Con ference on Development Economics, Amsterdam. Washington DC, The World Bank. Dercon S. 2005b. Risk, poverty, and vulnerability in Africa. J Afr Econ. 14(4):483-488. Dercon S, Hoddinott J, Woldehenna J. 2005. Shocks and consumption in 15 Ethiopian villages, 1999-2004. J Afr Econ. 14 (4):559-585. Dercon S, Krishnan P. 1996. Income portfolios in rural Ethiopia and Tanzania: Choices and con straints. J Dev Stud. 32:850-875. Dercon S, Krishnan P. 2000. Vulnerability, seasonality and poverty in Ethiopia. J Dev Stud. 36(6):25-53 Directorate of Agricultural Statistics. 2008. Abstract of Agricultural Statistics 2008. Pretoria. Elbers C, Gunning J. 2003. Growth and Risk: Methodology and Micro-evidence. Tinbergen Institute Discussion Papers, 03-068/2. Eswaran M, Kotwal A. 1989. Credit as insurance in agrarian economies. J Dev Econ. 31(1):3753. Fabrigar LR, Wegener DT, MacCallum RC, Strahan EJ. 1999. Evaluating the use of exploratory factor analysis in psychological research. Psychol Meth. 4(3):272-299. Francis, E. (2002). Rural livelihoods, institutions and vulnerability in North West Province, South Africa. J South Afr Stud. 28(3):531-550. Gaiha R, Imai K, Kang W. 2007. Vulnerability and poverty dynamics in Vietnam. Economics Discussion Paper. University of Manchester EDP 0708. Garriga RG, Foguet AP. 2010. Improved method to calculate a water poverty index at local scale. J Environ Eng. 136(11):1287-1298. Gbetnekom D. 2005. Deforestation in Cameroon: immediate causes and consequences. Environ Dev Econ. 10:557.

PAGE 126

126 Gibson J, Rozelle S. 2003. Poverty and access to roads in Papua New Guinea. Econ Dev Cult Change. 52(1):159. Gorsuch RL. 1983. Factor Analysis. Lawrence Erlbaum : Hillsdale, NJ. Gorsuch RL. 1997. Exploratory factor analysis: Its role in item analysis. J Pers Assess. 68(3):532-560. Grosh M, Glewwe P. 2000. Making decisions on the overall design of the survey, In: Designing Household Survey Questionnaires for Developing Countries: Lessons from Ten Years of LSMS Experience, Grosh M, Glewwe, P. (Eds.). pp 21-41. Grosh M, Glewwe P, Munoz J. 2000. Designing modules and assembling them into survey questionnaires, In: Designing Household Survey Questionnaires for Developing Countries: Lessons from Ten Years of LSMS Experience. Grosh M, Glewwe, P. (Eds.). pp 43-74. Grossman GD, Nickerson DM, Freeman MC. 1991. Principal component analyses of assemblage structure data: utility of tests based on eigenvalues. Ecology 72:341-347. Hakstian AR, Rogers WT, Cattell RB. 1982. The behavior of number of factors rules with simuated data. Multivariate Behav Res. 17:193-219. Harrower S, Hoddinott J. 2005 Consumption smooothing in the Zone Lacustre, Mali. J Afr Econ. 14(4):489-519. Heitzmann K, Canagarajah RS, Siegel PB. 2002. Guidelines for Assessing the Sources of Risks and Vulnerability, Social Protection Discussion Paper Number 0218. Washington DC, The World Bank. Hoddinott J, Quisumbing AR. 2003. Data Sources for Microeconometric Risk and Vulnerability Assessments. Manuscript, International Food Policy Research Institute, Washington DC. Hoogeveen JG. 2005. Measuring welfare for small but vulnerable groups: Poverty and disability in Uganda. J Afr Econ. 14(4):603-631. Hutcheson G, Sofroniou N. 1999. The multivariate social scientist: Introductory statistics using generalized linear models. Thousand Oaks, CA: Sage Publications. Jackson DA. 1993. Stopping Rules in Principal Components Analysis: A Comparison of Heuris tical and Statistical Approaches. Ecology. 74(8):2204-2214.

PAGE 127

127 89(1):28-38. Jodha NS. 1992. Agricultural growth and sustainability: perspectives and experiences from Himalayas, In: Agricultural sustainability, growth, and poverty alleviation: Issues and poli Joseph FH, Anderson RE, Tatham RL, Black WC. 1998. Multivariate data analysis. 5th ed. New Jersey : Prentice Hall. Kanbur R, Squire L. 2001. The evolution of thinking about poverty: exploring the interactions, in: Gerald M. Meier, Joseph E. Stiglitz eds. Frontiers of development economics : the fu ture in perspective, Oxford University Press. Kates R, Burton I. 1986. Geography, Resources, and Environment, vol. 1. Chicago: University of Chicago Press. Kelly PM, Adger WN. 2000. Theory and practice in assessing vulnerability to climate change and facilitating adaption. Clim Change. 47:325-352. Kieschnick R, McCullough BD. 2003. Regression analysis of variates observed on (0, 1): per centages, proportions and fractions. Stat Mod. 3:193-213. Kirsten J. 1996. The potential for creating additional rural livelihoods in agriculture and the rural nonfarm sector in semi-arid areas: A case study in the Northern Province. In Lipton M, Ellis F, Lipton M (eds), Land, Labour & Livelihoods in Rural South Africa, Vol 1: Western Cape. Indicator Press & LAPC : Johannesburg & Durban. Kochar A. 1995. Explaining household vulnerability to idiosyncratic income shocks, AEA Papers and Proceedings. Am Econ Rev. 85(2):159. Kurosaki T. 1995. Risk and insurance in a household economy: Role of livestock in mixed farm ing in Pakistan. Develop Econ. 33(4):464-485. Landis JR, Koch GG. 1977 The measurement of observer agreement for categorical data. Bio metrics 33:159-74. Ligon E, Schechter L. 2003. Measuring vulnerability. Econ J. 113:95-102.

PAGE 128

128 Linn RL. 1968. A Monte Carlo approach to the number of factors problem. Psychometrika. 33:37-71. Louw JH, Scholes MC. 2006. Site index functions using site descriptors for Pinus patula planta tions in South Africa. Forest Ecol Manag. 225:94-103. McArdle JJ. 1990. Principles Versus Principals of Structural Factor-Analyses. Multivariate Be hav Res. 25(1):81-87. Mertens B, Lambin EF. 2000. Land-Cover-Change trajectories in southern Cameroon. Ann Assoc Am Geogr. 90(3):467-494. Miyamoto M. 2006. Forest conversion to rubber around Sumatran villages in Indonesia: Com paring the impacts of road construction, transmigration projects and population. Forest Pol. 9:1. Mogues T. 2011. The Bang for the Birr: Public Expenditures and Rural Welfare in Ethiopia, J Dev Stud. 47(5):735-752 Morduch J. 1994. Poverty and vulnerability. Papers and Proceedings. Am Econ Rev. 84(2):22125. Nelson GC, Hellerstein D. 1997. Do roads cause deforestation? Using satellite images in econo metric analysis of land use. Am J Agr Econ. 79(1):80. Ninno C, Marini A. 2005. Households Vulnerability to Shocks in Zambia. World Bank Social Protection Discussion Paper 0536. Washington DC, World Bank. Oosthuizen M. 2008. Estimating Poverty Lines for South Africa, Discussion Document. Devel opment Policy Research Unit, University of Cape Town. Papke LE, Wooldridge JM. 1996. Econometric methods for fractional response variables with an application to 401(K) plan participation rates. J Appl Econometrics. 11:619. ity, and Unemployment form 2000 till 2007. Background Paper 2009:1(8). Elsenburg. and Unemployment form 2000 till 2007. Background Paper 2009:1(9). Elsenburg. Reardon T, Vosti SA. 1995. Links between rural poverty and the environment in developing countries: Asset categories and investment poverty. World Dev. 23(9):1495-1506.

PAGE 129

129 amongst farm households in Burkina Faso. J Dev Stud. 28(1):264. Rosenzweig M, Wolpin K. 1993. Credit market constraints, consumption smoothing and the ac cumulation of durable production assets in low-income countries: Investments in bullocks in India. J Polit Econ. 101(2):223-244. of agricultural Investments. Econ J. 103:56. Ruttan VW. 1992. Sustainable growth in agricultural production: poetry, policy, and science. In Vosti S, Reardon T, von Urff W. (Eds), Agricultural sustainability, growth, and poverty al native target species: a choice model. Can J Fish Aquat Sci. 61(3):374. SANParks. 2000. Visions of Change: Social Ecology and South African National Parks. Johan nesburg, South Africa: Development Communications Co. in association with South Afri can National Parks. Sen, A. 1981. Poverty and Famines: An Essay on Entitlements and Deprivation. Clarendon Press : Oxford. Sen A. 1984. Rights and capabilities. In: Sen A. (Ed.), Resources, Values and Development. Blackwell : Basil. Sen A. 1999. Development as Freedom. Anchor Books, New York. Shackleton S, Campbell B, Wollenberg E, Edmunds, D. 2002. Devolution and community-based perspectives (76). Overseas Development Institute, London. Shewmake S. 2008. Vulnerability and the Impact of Climate Change in South Africas Limpopo River Basin. IFPRI Discussion Paper 00804. International Food Policy Research Institute. Stark O. 1991. The migration of labor. Blackwell : Oxford. Statistics South Africa. 2007. A National Poverty Line for South Africa. National Treasury. Stiglitz J, Sen A, Fitoussi JP. 2009, Report by the Commission on the Measurement of Economic

PAGE 130

130 Swamee PK, Tyagi A. 2000. Describing water quality with aggregate index. J. Environ. Eng. 134(8):689. Swemmer L. 2009. Personal communications. Tabachnick BG, Fidell LS. 2001. Using Multivariate Statistics. Allyn and Bacon : Boston. Tesliuc E, Lindert K. 2002. Vulnerability: A quantitative and qualitative assessment, Guatemala Poverty Assessment Program. The World Bank, Washington, DC. Thomas DSG, Twyman C, Osbahr H, Hewitson B. 2007. Adaptation to climate change and variability: farmer responses to intra-seasonal precipitation trends in South Africa. Clim Change. 83:301-322. Velicer WF, Jackson DN. 1990. Component analysis versus common factor analysis: Some is sues in selecting an appropriate procedure. Multivariate Behav Res. 25(1):1-28. Webb P, Coppock DL. 1992. Prospects for pastoralism in semi-arid Africa. In Vosti S, Reardon T, von Urff W. (Eds), Agricultural sustainability, growth, and poverty alleviation: Issues and Weichselgartner J. 2001. Disaster mitigation: The concept of vulnerability revisited. Disast Prev Manag. 10(2):85-94. World Bank. 2001. World Development Report 2000/2001. Attacking Poverty, New York, Ox ford University Press. World Bank. 2007. A Decade of Action in Transport: An Evaluation of World Bank Assistance to the Transport Sector, 1995. The World Bank, Washington, DC. Xu J, Fox J, Melick D, Fujita Y, Jintrawet A, Jie Q, Thomas D, Weyerhaeuser H. 2006. Land use transition, livelihoods, and environmental services in montane mainland southeast Asia. Mt Res Dev. 26(3):278. Zimmerman FJ, Carter MR. 2003. Asset smoothing, consumption smoothing, and the reproduc tion of inequality under risk and subsistence constraints. J Dev Econ. 71(2):233. ponents to retain. Multivariate Behav Res. 17:253-269.

PAGE 131

131 nents to retain. Psychol Bull. 99:432-442.

PAGE 132

132 BIOGRAPHICAL SKETCH Gregory Parent was born and raised in New Hampshire. From 1997 to 2001 Gregory at tended McGill University in Montreal where he double majored in economics and international development studies. His interest in economics and development originated while at McGill and continues to this day. From 2002 to 2004, Gregory served in the Peace Corps in the capacity of a natural resource management advisor. He was stationed in the small remote village of Akloa, Wawa Prefecture, Togo. Gregory worked closely with villagers during his 2 year stay on the development of a village ecotourism project, a health clinic construction project, various projects demonstrating improved agricultural techniques, and health education. After Togo, Gregory was accepted into a masters program in the School of Forest Re sources and Conservation at the University of Florida. While here, Gregory worked with the Florida Division of Forestry to help quantify the impact of forest based recreation to the Florida economy using Input-Output analysis and econometric methods. Gregory successfully completed his Master of Science in the spring of 2007. At the conclusion of his MS, Gregory was offered an NSF IGERT fellowship in the Wa ter, Wetlands, Watershed UF IGERT program and was household in the Department of Geogra phy. At this time, Gregory was able to attain funding through various grants to conduct research in Africa. Over the last three years he has been the lead researcher on a project in South Africa, in conjunction with SANParks and regional NGOs, that has focused on the evaluation of vulner ability to poverty in rural community areas bordering Kruger National Park to help inform policy decisions in the Greater Limpopo region. For this project Gregory developed all aspects of the study methodology and was responsible for project management, data analysis, modeling, and

PAGE 133

133 report writing, amongst other tasks. The data collected as part of this study formed the core of his PhD research. In addition to the above, Gregory recently completed a project in the capacity of lead economist to evaluate the change to tourism demand in Kafue National Park, Zambia in response to various investment scenarios. To estimate demand change Gregory utilized a combination of revealed preference and contingent behavior models. Gregory further conducted research in Namibia, where through the development economic modeling, he investigated the impact of wildlife tourism versus that of traditional livelihood options (i.e. rain-fed agriculture and animal husbandry) on the rural village economic structure through the building a social accounting ma trix (SAM) model. Further, he evaluated the importance of water access to a households ability to participate in the market economy.