Vulnerability is generally conceptualised as encompassing a combination of ‘external’ determinants relating to exposure to shocks and stress and ‘internal’ determinants relating to the ability to respond to these (Rakodi, 1999; Chambers, 1989). This division between external and internal determinants is increasingly seen as superficial and insufficient, as vulnerability emerges from a combination of these aspects working in unison (Chapter 3; Shackleton & Shackleton, 2012).
To understand vulnerability in a specific context requires an exploration how stressors interact with factors that make some groups of people disproportionately vulnerable to different stressors (O’Brien et al., 2009). This context should incorporate social, economic and ecological spheres and include different spatial scales to capture the multitude of causal relationships (Turner et al. 2003; Holling, 2001). Incorporating broader spatial and temporal scales across socio-ecological systems also contributes to understanding the extent to which vulnerability is emerging beyond the sphere of control of the household, which may imply a limit to the household’s capacity to respond (Adger et al., 2009; Brooks et al. 2005). The previous chapter has illustrated the importance of considering temporal dimensions, as many aspects contributing to present-day vulnerability are deeply rooted in the past (Chapter 5).
Factors such as poverty and gender are well documented as being factors that increase vulnerability to HIV/Aids, climate change and a multitude of additional stressors (Eriksen & O’Brien, 2007;
Whiteside, 2002; O’Brien et al., 2009; Chapter 3). Marginalised groups frequently lack access to quality goods and services valuable for building resilience to shocks and stressors or for responding to shock and stress in a way that minimises long-term harm (Eriksen & O’Brien, 2007; Chapter 3).
Various contextual factors and household characteristics influence vulnerability by constraining options or by increasing exposure, and different localities and groups also have diverse personal values, ethics and aspirations which influence their perceptions of vulnerability (Davis, 2010; Sen, 1999; Adger & Kelly, 1999). These different personal values and perceptions are also partly defined by contextual factors such as culture or gender roles (Eisler et al., 2003). Identities are important considerations in understanding perceptions of vulnerability, with implications for motivation to adapt (Eakin et al., 2011). This study incorporates a variety of groups in different localities which may have their own distinct perceptions and experiences: men and women, rural and peri-urban and high and low income levels.
65 6.1.2 Rationale, key questions and hypotheses
In order to address this thesis’s broad objective of understanding how capital stocks are used to create livelihoods and respond to stress within the context of HIV/Aids and climate change, an initial understanding of differential and disproportionate experiences of shocks and stressors is needed to understand the vulnerability context. This chapter explores this aspect by answering the question:
What defines and shapes vulnerability in the two sites?
This question can be elaborated into the following sub-questions:
1.) At what scales do stressors act and interact?
2.) What do different groups perceive as being the main drivers of vulnerability?
3.) How do factors which may render households vulnerable over-lap?
4.) Do households characterised by different locality, income brackets and gendered headship experience differential shocks and stress?
This chapter will explore different experiences of vulnerability across the two sites. In order to answer questions 1 and 2 above, the chapter starts with a more in-depth look at the current vulnerability context through the use of participatory mental modelling exercises to illustrate the links between multiple problems and vulnerabilities in the two sites as perceived by men and women respectively.
To answer question 3, the chapter goes on to explore the distributions of the income and gender groups derived from the survey data across the two sites to assess how important contextual factors over-lap. To answer question 4, additional survey data is used to describe differential impacts of HIV/Aids based on proxy indicators, as well as the different types and frequencies of shocks and stresses experienced in the two sites. Perceptions of household-level food security and climate change impacts are analysed amongst the different sites, gender and income groups. This variety of
contextual information helps to understand disproportionate and differential drivers of vulnerability.
Based on the heightened vulnerability of rural and low-income households, and women, to HIV/Aids, climate change and other stressors (see 1.1.1; 1.1.2; 3.2.2; 3.2.3) it was hypothesised that households in the rural site of Gatyana, those headed by woman, and/or those with low incomes, will be worse affected by various shocks and stressors as it was assumed these households would be less able to respond. It is assumed that this heightened vulnerability will be reflected in perceptions of stressors.
6.1.3 Methods
This chapter uses a variety of statistical and participatory methods to capture distinct vulnerabilities of different groups based on site, the gender of the household’s head, and income levels, whilst
66 considering how vulnerability is defined by multiple drivers operating at different spatial scales (Chapter 3). This chapter thus incorporates perspectives around the drivers of vulnerability at a community-level as well as household-level experiences and impacts of vulnerability. Aside from participatory exercises, this chapter draws from a survey which interviewed 340 households split across two sites (see Chapter 4 for survey design and implementation).
a) Participatory mental modelling and problem trees to map multiple stressors
Mental modelling was used in both sites to record local perceptions of linkages between multiple stressors present in the area. In both Gatyana and Lesseyton, small focus group workshops were held with men and women separately. These focus groups were based on the methods developed by Ozemsi and Osemzi (2004) and used by Bunce et al. (2010). Bunce et al. (2010) used mental models to analyse perceptions of climate change amongst marginal African coast communities, and defined mental models as “qualitative representations of a system consisting of variables and the causal relationship between them” (Bunce et al., 2010: 414). Mental models have also been used to understand causes and consequences of climate change (Tschakert & Sagoe, 2009). Mental models are useful to quickly ascertain perceived causes and effects, and thereby feedback loops, amongst stressors operating across different spatial and temporal scales.
The mental models took the form of a spider-gram, with a key stressor being linked via lines showing directional causal relationships to other stressors. HIV/Aids was suggested to participants as an initial item to discuss and participants agreed that it was a major problem in the area that should be
discussed. The causes and effects of HIV/Aids were added to the spider-gram. Once participants felt they had contributed enough towards this stressor, they were asked to list another stressor in their community, and this process was repeated until participants felt they had addressed all of the main stressors in their community. Participants were then asked to reflect on the stressors and their causes and effects, and to identify what they perceived to be the main or key driving stressor. This key stressor was then further analysed in a problem tree.
Problem trees are useful in exploring causes and effects of problems across scales in a systematic way to help identity interventions. A core problem is broken down to its local, regional, national and global causes and effects. The problem trees used in these focus groups were adapted so as to explore causes and actual and possible responses across scales. After identifying perceived responses to what was perceived to be a core problem in the area, participants were asked to reflect on the extent to which responding in these ways to the core problem would be effectual in addressing the other stressors identified in the mental models.
b) Using household survey data to compare the vulnerability context of the sites in relation to gender and income
67 To contribute to a fuller understanding of the present-day local vulnerability context, the data from the household survey were used to compare the distributions of various household types categorised by gender and income groups (Box 6.1.3 a). Within each site, the distribution of different gender headship groups and income quartiles were compared (see Box 6.1.3 a and 4.4 for more detailed explanations of these groups).
c) Experiences of household shocks and stresses by different households
Household-level experiences of HIV/Aids impacts were analysed by site, gender and income whilst experiences of shocks and stressors were analysed by site only (Appendix 1).
The HIV/Aids proxy indicator categories were defined differently to other studies, where a count of the met proxy indicators is used to categorise the household as either affected or non-affected (Kaschula, 2009; McGarry, 2008; see 4.2.1). These proxy indicator data were disaggregated
differently to try to capture different experiences of HIV/Aids. Chronic illness, illness-related deaths and the presence of orphans were reasoned to be qualitatively different experiences of HIV/Aids.
Chronic illness increases the household’s health-related expenses and diverts work towards the care of the ill person, while the ill person is not able to contribute to income. An illness-related death is a shock which combines the expense of a funeral with the loss of an income. The presence of orphans in the household adds a financial burden. During the interview process, respondents would sometimes report that one of a child’s parents had passed away, while the other parent could not be accounted for or was described dismissively as “away”. For this analysis, such children were considered in effect to be orphans.
Households were thus categorised as non-affected (not meeting any proxy indicators), chronic illness and receiving free care (amongst people aged 0 – 59 years old), an illness related death in the last ten years, and households with de facto orphans (children whose parents were deceased, or who had lost a parent and whose other parent was away) (Box 6.1.3 b). As many households experience more than Box 6.1.3 a: Explanations of household types used in analyses
Household head (see 4.4.3)
Male only Households with only adult males; i.e. male headed Male with female* Male headed households with adult females Female with male* Female headed households with adult males
Female only Households with only adult females; i.e. female headed Income
quartiles (see 4.4.4)
Lowest income Quarterly income in cash and kind from R0.00 to R3480.37 Low income Quarterly income in cash and kind fromR3354.85 to R6106.08 Moderate income Quarterly income in cash and kind from R6179.03 to R9600.78 High income Quarterly income in cash and kind fromR9617.31 to R54 074.36
*Although often the case, these may not necessarily indicate a married couple, but could at times rather represent a mother/father living with adult children
68 one category, there was over-lap between categories. Households which were categorised as having experienced an illness-related death may or may not have also had the presence of chronic illness.
Households with de facto orphans may or may not also have had an illness-related death and/or may or may not have been experiencing chronic illness. Including households into these categories when they displayed other characteristics was necessary to create large enough groups for comparison.
A figure for comparison representing the degree to which HIV/Aids is impacting the household was derived by summing the amount of HIV/Aids experiences the household had (Box 6.1.3). A figure of 0 thus indicates no affliction, while 3 indicates that the household has chronic illness, has experienced an illness related death, and looks after de facto orphans (Box 6.1.3 b).
In the household survey, respondents were asked whether the household experienced any of a list of shocks and stresses, which included serious illness or death of a productive adult, loss of major household assets, loss of employment, or a costly social event such as a wedding or initiation (see Appendix 1). Households were also given space to specify other shocks not on the list. The severity of the shock’s impact was ranked with 0 indicating no crisis, 1 indicating moderate crisis, and 2
indicating severe crisis. It is debatable whether these shocks and stresses should be considered as a
‘shock’, that is, an unpredictable and irregular disturbance or a ‘stress’, that is, a regular and predictable disturbance (King, 2011). For instance, serious illness or an expensive event may be either, depending on the household’s experience and whether it was anticipated. The shocks and stresses here refer to large expenses or losses of income which disturbed the household’s wellbeing or functioning.
Different types of shocks and stresses experienced by households were disaggregated by site only.
The percentages of households experiencing a variety of shocks and stresses (see 4.2.3) were Box 6.1.3 b: Summarised explanations of HIV/Aids categories and numerical values
HIV/Aids categories
Non-affected: Household did not meet any proxy indicators
Chronic illness: Presence of chronic illness amongst at least one household member aged 0-59 years old. Person also receives free health care for the illness.
Illness related death in previous ten years: A household member had passed away in the ten years prior to the survey, following an illness of longer than 3 months
De facto orphans: Household contains children whose parents are both deceased, or who had lost one parent while the other parent is absent
Numerical values
HIV/Aids impact: Sum of HIV/Aids experiences in household
Shocks: Frequency of shock x weight of severity of shock’s effect Food security: Weighted response ranging from 0 – 2
Climate change impact: Sum of weighted responses
69 tabulated, together with a score reflecting the impact of the shock or stress. These scores were derived by multiplying the frequency of a shock in a household by the respondent’s rating of the shock’s effect. If the shock caused no crisis in the household, it scored zero. If the shock caused moderate crisis, it scored 1. If the shock caused severe crisis within the household, it scored 2.
e) Alcohol purchasing amongst different household types
Alcohol emerged as a driver of vulnerability in both sites during participatory exercises (see 5.2.1;
5.2.2; 6.2). A question in the survey asked about household expenses, including expenditure on alcohol (see Appendix 2). The percentages of households purchasing alcohol was analysed categorically by site, and by gender headship types and income quartiles.
d) Food security and climate change perceptions
Households’ food security and climate change perceptions were analysed by site, gender and income (see 4.2.3; Box 6.1.3 b; Appendix 1).
Food security perceptions were based on a question that asked whether the household’s food
production and income has been sufficient over the last 12 months. A food security score was created for each household by weighting their response to whether their food is sufficient, with ‘no’ = 0;
‘reasonable’ = 1; and yes = ‘2’.
Questions around climate change perceptions included ranking the impact of weather on aspects of the household’s production and consumption. These included: the ability of crops to survive, the ability of livestock to survive, the abundance of useful plant and animal species in the area, the availability of water for livestock and crops, the availability of water for the household, food security, human health, and damage caused by extreme events. Respondents were given four scaled response options, which were weighted with ‘no impact’ = 0, ‘low impact’ = 1, ‘moderate impact’ = 2 and
‘high impact’ = 3. These weighted responses were summed for the eight weather impact questions to create a score out of 24.
f) Determining statistical differences between groups
Significant differences in categorical data derived from the survey were compared across groups using Pearson’s chi-squared test, whereas continuous data derived from the survey were analysed for significance using t-tests when comparing the two sites, and ANOVA for analysis of differences between the gender and income groups. Where the data were not normality distributed, Mann-Whitney U and Kruskal-Wallis H tests were used for site, and gender and income comparisons respectively. Summarised explanations for how numerical scores were derived can be found in Box 6.1.3 b. Significant differences (where P < 0.05) are highlighted in the tables of results.
70 6.2 Results: Mapping the links between multiple, interacting stressors