Download as pdf or txt
Download as pdf or txt
You are on page 1of 34

Accepted Manuscript

Quantifying social vulnerability to natural hazards in Botswana: An application of


cutter model

Kakanyo Fani Dintwa, Gobopamang Letamo, Kannan Navaneetham

PII: S2212-4209(18)31403-1
DOI: https://doi.org/10.1016/j.ijdrr.2019.101189
Article Number: 101189
Reference: IJDRR 101189

To appear in: International Journal of Disaster Risk Reduction

Received Date: 3 December 2018


Revised Date: 6 May 2019
Accepted Date: 17 May 2019

Please cite this article as: K.F. Dintwa, G. Letamo, K. Navaneetham, Quantifying social vulnerability
to natural hazards in Botswana: An application of cutter model, International Journal of Disaster Risk
Reduction (2019), doi: https://doi.org/10.1016/j.ijdrr.2019.101189.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Kakanyo Fani Dintwa1, Gobopamang Letamo2, Kannan Navaneetham3

Acknowledgements

The authors wish to thank Statistics Botswana for providing the 2011 Population and

Housing Census dataset used for analysis. Appropriate ethical considerations were followed

PT
as per the current laws of Botswana.

RI
U SC
AN
M
D
TE
C EP
AC

1
Corresponding Author’s Email: fdintwa@gmail.com Ph.D Scholar, University of Botswana, Department of
Population Studies
2
Department of Population Studies, University of Botswana, Email: Gobopamang.Letamo@mopipi.ub.bw
3
Department of Population Studies, University of Botswana, Email: Kannan.Navaneetham@mopipi.ub.bw
ACCEPTED MANUSCRIPT

1 Quantifying Social Vulnerability to Natural Hazards in Botswana: An


2 Application of Cutter Model
3

4 Abstract An assessment of social vulnerability to natural hazards poses an enormous challenge

PT
5 for disaster risk reduction efforts across the World. For successful implementation of risk
6 reduction and mitigation measures, it is important to perform coupled Geospatial and statistical
analysis of social vulnerability. The use of multiple indicators to define vulnerability is also

RI
7
8 important for identifying vulnerable people and vulnerable places. The study examined the
applicability of the Place Vulnerability Model in Botswana, as well as to analyse underlying

SC
9
10 factors contributing to social vulnerability to natural hazards. For data analysis, census district
11 level data from Botswana 2011 Population and Housing Census were mainly used. Principal

U
12 Component Analysis was conducted and three factors of social vulnerability accounted for
AN
13 81.06% of the variance among the social vulnerability indicator variables in the dataset. They
14 include: vulnerability preparedness, socio-economic status; and vulnerable population groups.
15 The Social Vulnerability Index (SoVI) scores reveal that the census districts with the highest
M

16 vulnerability scores are Ngamiland West and Central Tutume. The districts with the least
17 vulnerability are concentrated in the south (Jwaneng) and central (Orapa). There are urban-rural
D

18 differentials in terms of the distribution of socio-economic developments, hence high


TE

19 vulnerability in rural areas as compared to the urban centres. This study confirms that the Social
20 Vulnerability Index developed for the United States is applicable in Botswana. The application
21 of Botswana situation is meant to contribute towards the development of social vulnerability to
EP

22 natural hazards in sub-Saharan Africa.


23
C

24 Keywords: Social Vulnerability, Natural Hazards, Climate Change, Applicability, Botswana.


AC
ACCEPTED MANUSCRIPT
1 Quantifying Social Vulnerability to Natural Hazards in Botswana: An
2 Application of Cutter Model
3
4 1. Introduction
5 There is a growing awareness that the impacts of climate change on human population are
6 imminent, as well as potentially devastating. This has consequently urged some policymakers

PT
7 (Botswana’s National Policy on Disaster Management of 1996 and National Disaster Risk
8 Management Plan of 2009) and researchers (e.g. Cutter et al. 2003; Chipanshi et al. 2003;

RI
9 Cutter et al. 2008; Batisani, 2010; and Mogotsi et al. 2012) to call for more effective and pro-
10 active action to reduce the impacts. In today’s world, human vulnerability to natural hazards

SC
11 has attracted lots of attention because of the increased frequency and intensity of natural
12 hazard events and their impacts on the livelihoods of communities. The impacts of natural

U
13 hazards are becoming more pronounced in developing countries because of inadequate
14 preparedness, adaptation and mitigation measures. Social vulnerability to natural hazards has
AN
15 become a topical issue in the modern vulnerability research.
16 Recent literature advocates for a paradigm shift from the usual environmental vulnerability
M

17 to social vulnerability. For example, Chen et al. (2013) state that, in adopting the human
18 centered vulnerability concept, the social vulnerability paradigm (Hewitt 1997; Blaikie et al.
D

19 1994; Cutter et al. 2003) stresses that vulnerability is socially built and exhibits with
TE

20 stratification and inequality among different groups of people and different places. Therefore,
21 vulnerability reduction requires understanding the underlying social, economic, and political
22 context and then addressing the factors that increase risk and vulnerability (Chen et al. 2013).
EP

23 The demographic and socio-economic fabric of individuals or communities that are mostly
24 influenced by economic development, describes their risk and vulnerability to natural
C

25 hazards, while the environment acts as the agent of a disaster. Most of the individuals and
AC

26 communities in sub-Saharan Africa are vulnerable to natural hazards because of their poor
27 living standards and over-dependence on natural resources for their livelihoods. As a result
28 Africa is thought to be already on the verge of its natural coping ability (Sokona and Denton,
29 2001). Literature elsewhere (e.g. Wright, 2019) asserts that vulnerability of the impoverished
30 people to the impacts of climate change is on the increase, more especially those living in
31 coastal cities. In the study of vulnerability, it is thought that social processes generate unequal
32 exposure to risk by making some people more susceptible to disasters than others (Dunno,
33 2011).
ACCEPTED MANUSCRIPT
34 For effective public policy related to impacts of natural hazards, it is important to measure
35 social vulnerability of individuals or communities in order to establish who needs assistance
36 before, during and after a disaster has struck. According to George and Schensul (2013),
37 understanding and reducing climate vulnerability are at the centre of effective adaptation. A
38 comprehensive understanding of the drivers of social change is essential in order to avoid
39 superficial perspectives of the relationship between the ecosphere and the anthroposphere.

PT
40 Changes that affect the size, distribution and composition of human population also affect
41 both the nature of vulnerability and adjustments in natural or human systems, in response to

RI
42 actual or expected climatic stimuli (George and Schensul, 2013). It is important to collect and
43 compile social vulnerability data prior to making decisions on how best communities can

SC
44 adapt and recover from the impacts of climate change (Mavhura et al. 2017). Therefore this
45 data becomes handy as governments increase their attention towards planning for, responding
46 to, natural hazards, particularly those related to climate change (Stafford and Abramowitz,
47 2017).
U
AN
48 It is evident from the literature that most of research entities across the world have
49 identified the need to investigate social aspects of vulnerability to natural hazards as this has
M

50 been lacking in vulnerability research. The focus has been on biophysical vulnerability to
51 natural hazards instead. Disaster-related losses do not only emanate from the magnitude and
D

52 period of the natural hazard events which is referred to as biophysical vulnerability, but also
53 from lack of people’s ability to protect themselves, livelihoods and property (Chen et al.
TE

54 2013). This gap is also evident in Africa, particularly in the southern region, hence the
55 motivation to undertake studies like this one in order to fill this gap. The impacts of natural
EP

56 hazards events on the livelihoods of the communities in Botswana are significant in spite of
57 the Government of Botswana’s initiatives to sensitize the public about the potential and
C

58 actual impacts of natural hazards. Some of the initiatives include the National Policy on
59 Disaster Management of 1996, National Disaster Risk Management Plan of 2009, and Early
AC

60 Warning Systems. In 2010 a total of 418 households and 1,669 individuals were affected by
61 floods in the Ngamiland, North West, Kgatleng, and Ngwaketsi districts (Statistics Botswana,
62 2014). During the year 2012, a total of 329 households and 1,756 individuals were affected
63 by floods in Mahalapye sub-District alone. Even though there are low numbers of households
64 and individuals affected by natural hazards, this is a cause of concern because the poor
65 communities, children, elderly and those living with disabilities are the most affected. The
66 most vulnerable are considered those who are most exposed to perturbations, with limited
ACCEPTED MANUSCRIPT
67 coping capacity and least resilient to recovery (Vincent, 2004), and in most cases these are
68 people with low economic status.
69
70 1.1. Theoretical Models
71 An analysis of theoretical models of vulnerability is important in identifying underlying
72 components contributing to vulnerability to natural hazards. The selected models are then

PT
73 operationalized in the study in order to unpack and understand issues surrounding social
74 vulnerability.

RI
75
76 1.1.1. Pressure and Release Model

SC
77 Blaikie, Cannon, Davis and Wisner (1994) developed the Pressure and Release Model (PAR),
78 which explains the relationships that exist between disaster and vulnerability. The PAR
79 model recognizes a disaster as a meeting point between socio-economic pressure and physical
80

U
exposure. The focus of attention in the PAR model is on the conditions that make exposure
AN
81 unsafe which, as a result, leads to vulnerability and to the causes creating these conditions.
82 The model is used mainly situations where social groups are faced with hazard events, with
M

83 particular emphasis on exposure units such as social class, ethnicity, among others. Blaikie et
84 al. (1994) further note that the natural progression of vulnerability is threefold, that is, it
D

85 differentiates between three components on the social side: i) root causes, ii) dynamic
86 pressures and, iii) unsafe conditions, and natural hazard component on the natural side.
TE

87 Principal Root causes include ‘economic, demographic and political processes,’ affecting
88 how resources are shared and distributed among groups of people (Blaikie et al. 1994).
EP

89 According to Cutter et al. (2009) one of the limitations of the model is that it does not
90 fully address the role of proximity to the source of the threat and the interaction between the
C

91 social and natural environment in the production of the hazard. It is most valuable for
92 descriptive analyses. Dunno (2011) argues that the PAR model does not have a feedback loop
AC

93 for policy and mitigation interventions. Though the access model proposed by Blaikie and
94 colleagues (1994) has a provision for the integration of policy practices affecting
95 vulnerability, it still showed no ability to integrate mitigation strategies, as well as no signs
96 considering some multiple hazards analysis.
97
98 1.1.2. Hazard Dimensions Model
99 Tobin and Montz developed the Hazard Dimensions Model in 1997. Through a combination
100 of the physical, economic and social characteristics of hazards, the hazard dimensions model
ACCEPTED MANUSCRIPT
101 employs the integrative approach to studies of vulnerability. The physical characteristics
102 include magnitude, frequency, duration, spatial extent and seasonality. The economic and
103 political attributes capture closeness to the hazard event, structure type, and availability of
104 choice. While societal characteristics include but not limited to, distribution of wealth
105 (income inequality), age (15 years and below and 65 years and over), gender (percentage of
106 females and female-headed households, occupation (service provision). This model has not

PT
107 yet been widely operationalized, hence lack of literature on its application.
108

RI
109 1.1.3. Human Ecology of Endangerment
110 The “human ecology of endangerment” was started by Hewitt in 1997. The model looks at

SC
111 societal factors contributing to vulnerability together with their geospatial distribution. This
112 model considers the societal characteristics that influence vulnerability as well as their
113 geographic distribution. The model consists of three segments, which include: i) forms of
114

U
vulnerability, ii) loci of vulnerability, and iii) syndromes (i.e. assumed determinants) of
AN
115 vulnerability. Hewitt further asserts that there are three forms of vulnerability and these are
116 exposure, susceptibility and lack of aid, and lack of capabilities to respond to hazards event
M

117 and their impacts. The characteristics of a society that affect vulnerability are referred to as
118 ‘Loci of vulnerability’. Examples of loci of vulnerability include factors such as gender,
D

119 culture, ethnicity, and social fabric within a society. Furthermore, the political economy and
120 level of development of a society are also assumed to be determining factors of vulnerability.
TE

121 Within each component of vulnerability the geographic distribution aspect is infused.
122 The model lacks the feedback mechanism which can be used to inform policies and
EP

123 mitigation programmes (Dunno, 2011). There is no synergy between the end vulnerability
124 output and the potential mitigation and adaptive strategies. There was no evidence of the
C

125 application of the model.


126
AC

127 1.1.4 Place Vulnerability Model


128 The Place Vulnerability Model also known as ‘Hazards-of-Place Model of Vulnerability’ was
129 presented first in Cutter (1996). The model was later used in Cutter, Mitchell and Scott
130 (2000). The model pulls together both biophysical/environmental vulnerability and social
131 vulnerability in order to determine vulnerability of a place. According to Cutter (1996),
132 hazard potentials in the model are a result of the interaction between risk and mitigation
133 measures. This hazard potential is then reduced or worsened by the way the society interacts
134 (i.e. social vulnerability). Cutter and colleagues further asserted that the socio-economic and
ACCEPTED MANUSCRIPT
135 demographic factors and the societal ability (capacity) to respond to hazards constitute social
136 vulnerability. In the Place Vulnerability the biophysical vulnerability which is made up of
137 proximity to the source of event and situation of the place, among others, interacts with
138 hazard potential to show the level of vulnerability and its geographical location. There is a
139 feedback loop incorporated into the Place Vulnerability Model, leading to an increase or
140 decreases in risk, hence an enhanced or reduced vulnerability. This feedback mechanism is

PT
141 used to inform policy and mitigation-related strategies and programmes.
142 The Place Vulnerability Model is the only model that captures almost all the elements the

RI
143 afore-mentioned models have been criticized for. For example, the model can be used to
144 compare vulnerability among different geographic locations, as well as the ability to analyze

SC
145 how social and biophysical factors interact spatially. It is for the aforesaid strengths that the
146 Place Vulnerability was operationalized to realize the objectives of this study.
147
148 1.2. The conceptual framework of the study
U
AN
149 This study operationalized the Place Vulnerability Model to compare vulnerability among
150 different geographic locations in Botswana. The model has been used to guide several
M

151 researchers and scholars dealing with studies on vulnerability to natural hazards (Cutter et al.
152 2000; Cutter et al. 2008; O’Brien et al. 2010; Heinz Center for Science, Economics, and
D

153 Environment, 2002; Cutter et al. 2009; Dunno, 2011). This study focuses on the social
154 vulnerability component of the place vulnerability model. The Model has been accepted by
TE

155 many researchers as the best model for measuring social vulnerability (e.g. Holand and
156 Lujala, 2013). The model pulls together both biophysical/environmental vulnerability and
EP

157 social vulnerability in order to determine vulnerability of a place. According to Cutter (1996),
158 hazard potentials in the model are a result of the interaction between risk and mitigation
C

159 measures. This hazard potential is then reduced or worsened by the way the society interacts
160 (i.e. social vulnerability). The socio-economic and demographic factors and the societal
AC

161 ability (capacity) to respond to hazards constitute social vulnerability (Cutter et al. 2003).
162 There is a feedback loop incorporated into the Place Vulnerability Model, leading to an
163 increase or decreases in risk, hence an enhanced or reduced vulnerability. (see Figure 1)
ACCEPTED MANUSCRIPT

PT
RI
SC
164
165 Fig. 1 Cutter’s (1996) Place Vulnerability Model
166
167
U
Some of the factors constituting the social fabric influencing social vulnerability are
AN
168 discussed here. For example, children and the elderly are mostly affected by the impacts of
169 natural hazards because of their physical inability to respond to threats posed by natural
M

170 hazards. Cutter et al. (2003) confirm that extremes of the age spectrum affect the movement
171 out of harm’s way, and parents lose time and money caring for children when day-care
D

172 facilities are affected. On the other hand the elderly may have mobility constraints, which
173 increase the burden of care and lack of resilience. Both minors and elderly have insufficient
TE

174 monetary savings that enables them recover from floods, among other natural hazards
175 (Mavhura, 2019) therefore making them more vulnerable to natural hazards. The special
EP

176 needs (disabled) population are mostly ignored during the occurrence of a hazard, as well as
177 during recovery (Morrow, 1999). The population residing in rural areas is mostly associated
C

178 with poverty and in most cases are not able to recover from the impacts of natural hazards.
179 The urban residents on the other hand face a serious threat because highly built urban centres
AC

180 with high population densities presenting a difficult and confusing situation in the face of a
181 hazard. According to Cutter et al. (2003), “rural residents may be more vulnerable due to
182 lower incomes and more dependent on locally based resource extraction economies (e.g.
183 farming, fishing)” (p.247).
184 Education as a socio-economic variable, is also associated with high social vulnerability.
185 The less educated compared to the more educated people do not comprehend early warning
186 initiatives the same way, as thus their reaction is influenced by their level of understanding
187 (quick response= more educated, slow response= less educated). The Heinz Center for
ACCEPTED MANUSCRIPT
188 Science, Economics, and the Environment (2000), notes that education is associated with
189 socioeconomic status, that is, higher educational attainment results in greater lifetime
190 earnings. The Center further states that lower education constrains the ability to access
191 recovery information as well as understanding the warning information. Frigerio et al. (2018)
192 also adds that a higher educational status often leads to access to the high-skilled jobs and
193 high paying positions that usually enhances coping and recovery capabilities from the

PT
194 impacts of natural disasters.
195 The impacts of disasters are prominent on females than males. Females are most likely to

RI
196 be affected by the impacts of natural hazards events because most of them work in low
197 paying jobs and consequently have difficulties recovering from the impacts due to limited

SC
198 financial resources. This can be exacerbated by the fact that most of females are care-givers,
199 and have large family sizes compared to their male counterparts. Cutter et al. (2003) assert
200 that families with large numbers of dependents or single-parent households often have limited
201

U
finances to farm out care for dependents, and thus must juggle work responsibilities and care
AN
202 for family members. All these affect the resilience to and recovery from hazards.
203 Testing the applicability of place vulnerability model in Botswana definitely helps to
M

204 inform policies and programmes, more especially the review of the Disaster Risk Reduction
205 Strategy of 2013, as well as provide some Botswana-based literature on the assessment of
D

206 social vulnerability.


207
TE

208 1.3. Research objectives


209 The aim of this research is to fill an academic and policy demand for the first assessment of
EP

210 national level social vulnerability to natural hazards in Africa. It does that by investigating
211 whether the SoVI methodology can be used to appropriately assess the social vulnerability of
C

212 census districts of Botswana. The focus is on the list of variables that appropriately present
213 the social fabric of communities in the different census districts of Botswana, and why other
AC

214 districts are highly vulnerable than others.


215
216 2. Data and Methods
217 The study is based on a secondary data analysis of a subset of variables contributing to social
218 vulnerability to natural hazards. The majority of the variables were sourced from 2011
219 Population and Housing Census of Botswana. Thus, informed consent from individuals
220 participants were not obtained by the researcher but this was obtained as part of the original
ACCEPTED MANUSCRIPT
221 wider study. During the 2011 Census, informed consent was obtained from individuals who
222 voluntarily agreed to take part in the study.
223 2.1. Variables and Indicators Selection
224 The data were obtained mainly from the Statistics Botswana 2011 Population and Housing
225 Census. The total of 28 census districts from the 2011 Population and Housing Census data
226 were used for the analysis of this study (unit of analysis is the census district). This selection

PT
227 is justified by the fact that all the people in the districts are at risk of experiencing any tragedy
228 due to their varying coping ability (demographic and socio-economic factors) and recovery

RI
229 from the impacts of natural hazards events, hence the need to investigate their level of social
230 vulnerability.

SC
231 The data are secondary and obtained mainly from the Statistics Botswana 2011 Population
232 and Housing Census, in order to attempt to the aim of this study which is to test the
233 applicability of the Social Vulnerability Index (SoVI) methodology in the census districts of
234

U
Botswana. The original variables (Table 1) for the SoVI Model by Cutter, Boruff and Shirley
AN
235 (2003) were replicated in Botswana and compared with the results of the adjusted SoVI
236 Model. Careful selection and rationalization of the variables with the intention of retaining
M

237 only those most significant as indicators of social vulnerability was done. Moreover, the
238 selection of the original variables was informed by the rule of thumb for PCA from Taylor’s
D

239 (1977), with the ratio of cases to variables set at 3.5:1. This ratio yielded 8 variables to 28
240 study sites/cases.
TE

241
242 Table 1 Original Social Vulnerability Variables for Analysis
Abbreviation
EP

Variable Name code for PCA Potential Source


2011 Population and Housing
1. Percent Renter-occupied Housing Units MED_RENT Census
2011 Population and Housing
C

2. Percent of Population Under 5 Years Old UNDER_FVE Census


3. Percent of Civilian Labour Force 2011 Population and Housing
AC

Unemployed UNEMLYD Census


2011 Population and Housing
4. Average Number of Persons Per Household NUM_HH Census
5. Percent Living in Poverty (below US$1.90 a 2009/10 Botswana Core Welfare
day level) PERC_POV Indicator Survey
6. Percent of Population 25 Years or Older with 2011 Population and Housing
No Secondary Education NO_SEC Census
7. Percent of Households Receiving Social 2011 Population and Housing
Security SOC_SEC Census
8. Percent Female-Headed Households (single 2011 Population and Housing
parent) FEMHEAD Census
243
ACCEPTED MANUSCRIPT
244
245 Dunno (2011) adds that although the Social Vulnerability Index is a valued tool for
246 measuring social vulnerability within a community, it should be necessary to adjust the
247 original input variables so that they more accurately reflect the socio-economic structure of
248 the study area. This is done through including additional variables that can also help represent
249 the socio-economic and demographic structure of the system being studied. The adjustment
250 was done in this study and the adjusted model includes the original variables for the SoVI

PT
251 Model plus the three additional variables (Table 2) which best represent the socio-economic
252 and demographic structure of the census districts of Botswana. The list of variables used in

RI
253 this study was dependent on data availability at census district level.
254 One of the variables shown in Table 2 was included into the SoVI algorithm because it

SC
255 shows communities’ preparedness prior to natural hazards events and reflect Boruff and
256 Cutter’s (2007) results. Studies done elsewhere support this, for example, a study by Boruff

U
257 and Cutter (2007) cited in Dunno (2011) found that several preparedness indicators
AN
258 significantly influenced social vulnerability in the islands of Barbados and St. Vincent. These
259 variables included: percent of housing units cooking with electricity and lighting with
260 electricity (it is thought to decrease vulnerability). It is quicker and easier to reconnect
M

261 electricity into the main grid once it was switched off as a result of a natural event. Those
262 households using biomass for cooking and lighting are more vulnerable because it will be
D

263 difficult for them to fetch firewood nearer once they are struck by floods due to heavy rains,
TE

264 storms (hailstorms, windstorms or strong winds), as well as when they are faced with extreme
265 high temperatures. Botswana has various ethnic groups living with each other within the
same area and they mix during certain events, like wedding ceremonies, funerals, kgotla
EP

266
267 meetings, and workplaces. Just like in other regions (e.g. the U.S. Virgin Islands) each ethnic
268 group in Botswana holds on to its own institutional practices, beliefs, values, norms and roles
C

269 that revolve around family, religion, education and the economy. The variable added to best
AC

270 reflect the racial and cultural composition of the Botswana census districts include: Percent of
271 speaking a language at home other than Setswana. (Refer to Table 2)
272
273 Table 2 Additional Social Vulnerability Variables included for Analysis
Abbreviation
Variable Name code for PCA Potential Source
1. Percent of Households Using Any Source of
Biomass for Cooking (wood/charcoal & other 2011 Population and Housing
methods) BIOMAS_COOK Census
2. Percent Living With Disability (partially/fully 2011 Population and Housing
blind, hearing impaired & mentally retarded) DISABLED Census
ACCEPTED MANUSCRIPT
14. Percent of Speaking a Language at Home 2011 Population and Housing
other than Setswana NO_SETS Census
274
275
276
277
278 2.2. Selection of Participants
279 The social vulnerability assessment was conducted for Botswana using mainly census district

PT
280 level data obtained mainly from the 2011 Population and Housing Census. The
281 Administrative Districts of Botswana were further divided into Census Districts for the ease

RI
282 of undertaking population and housing census data collection. The census district
283 enumeration unit was chosen because it has the ability to show variation in vulnerability

SC
284 within and between each of the districts. This smallest unit of analysis (census district) has
285 sufficient data for this study. Statistical Package for the Social Science (SPSS) Version 22
286 was used for data analysis.

U
287
AN
288 2.3. Statistical Methods
289 The statistical methods section presents the measurement or computation of the Social
M

290 Vulnerability Index (SoVI).


291
D

292 2.3.1. PCA Analysis


293 In this study a social vulnerability index was computed from a technique of Principal
TE

294 Component Analysis (PCA), which is a statistical procedure that converts correlated variables
295 into a linearly uncorrelated variables referred to as principal components. PCA is also
EP

296 referred to as a reduction technique that group large number of analysis variables in the
297 dataset into a minimum number of factors or components explaining the variance existing in
the larger dataset. It is for this description that PCA was used in this. It is mathematically
C

298

299 specified as follows:


AC

PC 1 = β 11 Y1 + β 12 Y 2 + β 13 Y 3 + ... + β 1 n Y n
.
300 .
PCm = β m1Y1 + β m 2 Y 2 + β m 3 Y 3 + ... + β mn Y n
301

302 In the above, βmn is the weight for the mth Principal Component (PC) and the nth variable Y.
303 In the output the weights of the PCs are represented by the eigenvectors of the correlation
ACCEPTED MANUSCRIPT
304 matrix. However, if the data is standardized the eigenvectors would be of the co-variance
305 matrix. On the other hand, the variance of the PCs is given by the eigenvalues (Vyas and
306 Kumaranayake, 2006). According to Cutter, Boruff and Shirley (2003) the Social
307 Vulnerability Index requires that the Varimax Rotation and Kaiser Criterion be employed for
308 the selection of components with eigenvalues greater than 1.0. The varimax rotation has a
309 potential of loading each variable highly on one component as a result the varimax rotation

PT
310 was selected for this study. The analysis of correlations between the variables in each
311 component then followed and the component loadings were given in the loading matrix

RI
312 output of the principal component analysis.
313 Components were ordered according to their proportion of variation that they explain in

SC
314 the original data (total variance); with those in the top positions explaining larger amounts of
315 variation. The component groups were named based on their broad representation of social
316 vulnerability and directions (cardinal) were then assigned, positive (component increase
317

U
vulnerability), and negative (component decrease vulnerability). For those components which
AN
318 demonstrated negative factor loadings on variables that theoretically increase or positively
319 influence vulnerability, the cardinal directions were adjusted to represent their tendency to
M

320 increase vulnerability. Data normalization of the index score (SoVI) was done. For this study,
321 feature scaling was used to gather all scores into the range [0, 1].
D

322
323 2.3.2. Computation of Social Vulnerability Index (SoVI)
TE

324 Social Vulnerability Index (SoVI) was computed because it measures the social vulnerability
325 of different areas to environmental hazards. Furthermore, SoVI statistically assesses the
EP

326 underlying socio-economic and demographic characteristics of the population and how they
327 negatively contribute to people’s capacity to cope and recover from climate change-related
C

328 hazards (Emrich and Cutter, 2011). The adjusted SoVI index was computed as follows:
329
AC

330 SoVI = Ʃ [PC1, PC2, PC3 … PCnth]


331
332 In the above, SoVI is the Social Vulnerability Index which is computed through an additive
333 model from PC1 which is the Principal Component Number 1 to the nth Principal Component.
334 The number of the components is dependent on the number of components with eigenvalues
335 that are greater than 1.
ACCEPTED MANUSCRIPT
336 During the computation of the Social Vulnerability Index (SoVI), component scores were
337 created for all census districts in Botswana. Due to the fact that the SoVI model is replicated
338 and tested as part of this study; an additive model was selected based on assumption that each
339 component have an equal contribution to the impact of vulnerability, that is, no weights were
340 assigned to any of the components loaded in the equation because of lack of strong and
341 reliable method of assigning weights to the components. Therefore there is no priori

PT
342 assumption concerning the importance of individual components in the overall sum. Similar
343 approach was applied by Cutter, Boruff and Shirley (2003), hence it was replicated in this

RI
344 study.
345

SC
346 2.3.3. Mapping Social Vulnerability Index
347 The SoVI scores were mapped, with districts shaded according to the SoVI scores in order to
348 allow for recognition of the spatial patterns of social vulnerability within the Census Districts
349

U
of Botswana. The ARC INFO 10.2 mapping software was used to map the SoVI scores. The
AN
350 SoVI scores were mapped as standard deviations from the mean, to allow for a visual
351 recognition of the spatial patterns of social vulnerability within the Census Districts of
M

352 Botswana. Furthermore, the standard deviations (std. Dev.) were used as the classification
353 algorithm to highlight the extremes (low and high) in social vulnerability in the study area.
D

354 The study areas with low vulnerability (< -0.5 std. Dev. from the mean) and high
355 vulnerability (> 0.5 std. Dev. from the mean) were identified. According to Letsie (2015) this
TE

356 measure consents one to statistically see how similar or how different places are relative to
357 each other, therefore making the SoVI a comparative measure of vulnerability. The mapping
EP

358 exercise satisfies the vulnerability of place component of the conceptual framework.
359
C

360 3. Results
361 3.1. Original SoVI Model
AC

362 A total of three components were generated through Principal Component Analysis (PCA)
363 and they explained 84. 93% of the variance among variables in the data set. Variables with
364 both significant component loadings (> 0.5 or < -0.5) and those which loaded highly on
365 multiple components were identified after critically examining the three components. All the
366 communalities were over 0.50. Furthermore, the component loadings were examined and
367 eventually named as follows: Socio-Economic Insurance Deprivation; Vulnerable/Special
368 Needs Population; and Deprivation to Women Empowerment (Table 3).
ACCEPTED MANUSCRIPT
369 The components were then summed to produce an overall score of vulnerability and no
370 weights were assigned (Cutter, Borruf and Shirley, 2003) to any component (Table 3).
371
372 Social Vulnerability Index = Ʃ [(Component 1) + (Component 2) + (Component 3)]
373
374 Where, Component 1: Socio-Economic Insurance Deprivation, Component 2:

PT
375 Vulnerable/Special Needs Population, and Component 3: Deprivation to Women
376 Empowerment.

RI
377 The results show that the mean vulnerability score for all census districts in Botswana is
378 0.574. A total of 16 out of the 28 census districts had vulnerability scores above the average

SC
379 vulnerability score. It is evident from the table that the census districts with the highest
380 vulnerability scores comparatively to the rest of the districts are the Ngamiland West and
381 Central Serowe/Palapye with scores of 1.000 and 0.996, respectively.
382
U
AN
383 Table 3 Components of Social Vulnerability in the Original Model
Percent Component Expected
Name &
Component Variation Dominant Variables Loading Cardinality
Cardinality
M

Explained Scores
Average number of
persons per
D

household .914 +
Socio-Economic
Insurance
.798
TE

1 51.732 Percent unemployed +


Deprivation
(+) Percent Over 65
Years .737 +
Percent Living in
EP

Poverty .619 +
Percent Under Five
.651 +
C

Vulnerable/Special Percent Receiving


2 Needs Population 20.358 Social Security .631 +
AC

(+) Percent with No


Secondary Education
.629 +
Deprivation to
Women
3 12.835
Empowerment Percent Female-
(+) Headed Households .500 +
384
385
386 3.2. Adjusted SoVI Model
ACCEPTED MANUSCRIPT
387 The Social Vulnerability Index has been proven by several researchers (Cutter, 1996; Cutter,
388 Boruff and Shirley, 2003; Dunno, 2011) as a valid and appreciated method for measuring
389 social vulnerability in any given community. For this study it was found necessary to adjust
390 the variables inputted into the model so that it can accurately represent a true picture of the
391 underlying social fabric of communities in Botswana (refer to Table 2 for Additional
392 Variables), then a subsequent PCA was run.

PT
393 The results show that the mean vulnerability score for all census districts in Botswana is
394 0.564 and standard deviation of 0.324. A total of 16 out of the 28 census districts had

RI
395 vulnerability scores above the average vulnerability score. The census districts with the
396 highest vulnerability scores comparatively to the rest of the districts are Ngamiland West and

SC
397 Central Tutume with scores of 1.000 and 0.989, respectively. The lowest vulnerability scores
398 are observed in Jwaneng (0.000) and Sowa Town (0.026). Gaborone city comes fourth after
399 the mining towns of Jwaneng, Sowa Town, Orapa, and Selibe Phikwe in terms of the least
400 social vulnerability level.
U
AN
401
402 3.3. Adjusted Social Vulnerability Index (SoVI) Factors
M

403 A total of three components were generated through Principal Component Analysis. The
404 three components explained 81.06% of the variance among all the variables captured in the
D

405 dataset. Moreover, the component loadings were examined and named as follows:
406 Vulnerability Preparedness; Socio-Economic Status; and Vulnerable Population Groups
TE

407 (Table 4). Each of the aforesaid is briefly described below.


408
EP

409 SoVI = Ʃ [(Component 1) + (Component 2) + (Component 3)]


410
C

411 Where, Component 1: Vulnerability Preparedness, Component 2: Socio-Economic Status;


412 and Component 3: Vulnerable Population Groups.
AC

413
414 3.3.1. Vulnerability Preparedness
415 The vulnerability preparedness factor contributes 46.366% of the total variance among
416 variables in the data set. The vulnerability preparedness factor shows high values in the study
417 units having high percentage of homes cooking with any source of biomass (wood, charcoal,
418 etc.), and high percent receiving social security. According to Dunno (2011) “homes cooking
419 with alternative fuels would be more vulnerable than those cooking with electricity because
ACCEPTED MANUSCRIPT
420 the power grid would be restored much quicker in the aftermath of a disaster event” (p.125).
421 It is also driven by high percentage receiving social security. This factor increases social
422 vulnerability, therefore a positive sign was assigned.
423
424 3.3.2. Socio-Economic Status
425 The second factor identified high percentage of the unemployed, high percentage living in

PT
426 poverty, high average number of persons per household, and high percent of female-headed
427 households as the main drivers of social vulnerability. The aforesaid variables loaded

RI
428 positively on this factor. The socio-economic status factor explains 24.76% of variance. High
429 unemployment coupled with high poverty level reduces communities’ ability to absorb and

SC
430 recover from losses, ‘presenting a high risk to their belongings. The findings also show that
431 there is high percentage of female headed households comparatively to male-headed
432 households in the majority of the study units. As demonstrated in the literature, Chen et al.
433

U
(2013) confirms that females are more vulnerable in the face of a disaster than their male
AN
434 counterparts because of their biological make-up, psychological characteristics, and their
435 social roles in a society. Resultantly, this factor was given a positive sign.
M

436
437 3.3.3. Vulnerable Population Groups
D

438 The third factor identified high percent of both under five years and high percentage over 65
439 years as the main drivers of social vulnerability. The aforesaid variables loaded positively on
TE

440 this factor. The factor explains 9.93% of variance. Both high numbers of children and high
441 numbers of the elderly shows communities’ lack of ability to absorb, cope, and recover from
EP

442 losses: because they are physically and economically not able to cope.
443
444 Table 4 Components of Social Vulnerability in the Adjusted Model
C

Percent Component Expected


Component Name Variation Dominant Variables Loading Cardinality
AC

Explained Scores
Percent Receiving Social Security .913 +
Percent with No Secondary +
Vulnerability .896
Education
1 Preparedness 46.366
(+) Percent using Biomass for cooking .876 +
Percent Speaking Language Other +
.572
than Setswana
Socio- .855 +
Economic Percent Female-Headed Households
2 24.764
Status Percent Unemployed .787 +
(+) Percent Living in Poverty .719 +
ACCEPTED MANUSCRIPT
Percent Living with Disability .671 +
Average number of persons per +
.661
Household
Vulnerable Percent Over 65 Years .929 +
3 Population 9.926
Groups (+) Percent Under 5 Years .919 +
445

446 3.4. Spatial Distribution of Adjusted SoVI in the Census Districts of Botswana

PT
447 The results show that none of the study units fall within the high vulnerability category (> 1.5
448 std. Dev.). Those study units (2 out of 28) which had the social vulnerability index scores

RI
449 within < -1.5 std. Dev. were labelled as least vulnerable. They include in their descending
450 order, Jwaneng, and Sowa Town (Figure 4.1b). Both Jwaneng and Sowa Town have low

SC
451 social vulnerability index scores mainly because of their high economic status shown by high
452 scores of vulnerability preparedness (component 1), socio-economic status (component 2),

U
453 and vulnerable population groups (component 3).
454 A total of ten (36%) out of 28 census districts are categorized under the medium high level
AN
455 of vulnerability (0.5 – 1.5 std. Dev.). The majority of these census districts have high scores
456 of both socio-economic status (component 2) and vulnerable population groups (component
M

457 3). The location of these study units is in, Central Serowe/Palapye, Central Mahalapye,
458 Central Bobonong, Central Boteti, Central Tutume, Kweneng West, Kweneng East, and
D

459 Ngamiland East (Figure 4.1b). However, Kweneng West, Central Tutume and Central Boteti
460 are the only census districts in the medium high vulnerability category that are influenced by
TE

461 low scores of vulnerability preparedness (component 1).


462 About 25 percent of study units fall into the category of medium level of social vulnerability
EP

463 (0.5 - -0.5 std. Dev.). These study units are located in Ngwaketse West, South East, Kgatleng,
464 North East, Ghanzi, Kgalagadi North, and Kgalagadi South. The majority of the census
C

465 districts in this category have low vulnerability preparedness, and low socio-economic status.
AC

466 The study units (36%) constituting the medium low vulnerability (-1.5 - -0.5 std. Dev.)
467 include Gaborone, Francistown, Lobatse, Selibe Phikwe, Orapa, Chobe, and Okavango Delta.
468 These census districts are ranked medium low because of high scores of vulnerability
469 preparedness (component 1), and low scores of vulnerable population groups (component 3).
470 The majority of the study units in this category are cities and towns (Gaborone, Francistown,
471 Selebi Phikwe, Orapa, and Lobatse). As an exception, the Okavango Delta which is expected
472 to be the most economically deprived and least deprived in terms of unemployment
473 opportunities, is classified under the medium low vulnerability. This is attributable to the fact
ACCEPTED MANUSCRIPT
474 that there are not so many habitable sites in the Okavango delta because the area is swampy:
475 the majority of the people who resides within the Delta work in the Safari Camps that have
476 been erected in the Delta for boosting the tourism sector. (Refer to Figure 2)
477

PT
RI
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

478
479 Fig. 2 Spatial distribution of SoVI in the Census Districts of Botswana
ACCEPTED MANUSCRIPT
480 3.5. Geographical Spread of Underlying Factors Contributing to Social Vulnerability
481 The results of the geographical spread of underlying factors contributing to social
482 vulnerability are shown in Figure 3. This factor (component 1), vulnerability preparedness,
483 shows a clear geographic pattern, with the highest values showing lack of vulnerability
484 preparedness in the central (Central Kalahari Game Reserve), south (Ngwaketse West) and
485 south (Kweneng West). The lowest values are recorded in the south eastern part (Gaborone

PT
486 and Lobatse), North East (Francistown) and some parts of central (Selibe Phikwe, Orapa, and
487 Sowa Town) (Figure 3a) and all these are urban centres. Of study units mentioned above,

RI
488 Gaborone had the highest percentage not using any source of biomass for cooking followed
489 by Selebi Phikwe, and Jwaneng with 66.72%, 72.21%, and 78.21%, respectively. This result

SC
490 shows that aforementioned urban districts have high vulnerability preparedness compared to
491 the rural areas.
492 The socio-economic status factor (component 2) shows a clear disparity in the economic
493

U
status of residents in the urban districts (city and town districts, which are more developed)
AN
494 and urban-rural districts (less developed) (Figure 3b). The majority of urban districts
495 (Gaborone; Francistown; Jwaneng; Sowa Town; and Orapa); scored relatively low values
M

496 with regards to socio-economic status comparatively to all urban-rural districts. The
497 deprivation to women empowerment part confirms high values mainly in the north
D

498 (Ngamiland West, and North East), and central (Central Bobonong and Central Tutume)
499 (Figure 3b). It is evident from the results of this study that high percentage of the female-
TE

500 headed households was recorded in Ngamiland West, followed by North East then Central
501 Mahalapye, Central Bobonong and Central Tutume with 56.90%, 54.10%, 53.40%, 52.70%
EP

502 and 52.30% respectively.


503 The highest values (vulnerable population groups factor) are located in the south
C

504 (Kweneng East) and in the central (Central Serowe/Palapye, Central Mahalapye and Central
505 Tutume) (Figure 3c). On the other hand, the lowest values of this factor are mainly located in
AC

506 the central (Sowa Town, Orapa and Selibe Phikwe) and south (Lobatse).
507
508
509
510
511
512
513
ACCEPTED MANUSCRIPT
514
515
516 (a) FAC1- Vulnerability Preparedness (b) FAC2- Socio-Economic Status

PT
RI
U SC
AN
M

517
518
519 (c) FAC3- Vulnerable Population Groups
D
TE
C EP
AC

520
521
ACCEPTED MANUSCRIPT
522 Fig. 3 Scores of Factors Contributing to Social Vulnerability in Botswana437
523
524 3.6. Comparing SoVI scores between the Original and Adjusted Models
525 In comparing the SoVI scores between the original and adjusted models, the findings show
526 that there were no major changes of the social vulnerability index scores between the original
527 model and the adjusted model. In both the original and adjusted model, Ngamiland West
528 district remained the highest vulnerable district, while the lowest ranked vulnerable districts

PT
529 remained the same as well (Jwaneng followed by Sowa Town). The results also show an
530 insignificant difference in the mean scores of the two models. Furthermore, the same number

RI
531 (16) districts which scored above the average vulnerability scores in the two datasets were the
532 same. The indicator loadings on the components varied between the two datasets, hence

SC
533 different naming of the components. Consequently, it can be concluded that though the SoVI
534 may not be universally applicable, it is however applicable in Botswana based on the afore-

U
535 mentioned results.
536 Since the original model of social vulnerability developed by Susan Cutter is based on
AN
537 indicators that are best suited and available in the United States, the researcher found it
538 necessary to check for the availability of the same indicators in-country. This was done by
M

539 identifying and collecting proxy indicators and other indicators which best describe the socio-
540 economic and demographic structure of Botswana, hence an adjustment of the social
D

541 vulnerability model. It is in light of this justification that the adjusted model was found
TE

542 appropriate for measuring social vulnerability in the census districts of Botswana, hence the
543 subsequent analysis was based on the results of the adjusted social vulnerability index model.
544
EP

545 4. Discussion
546 The growing concern of the impacts of natural disasters on communities’ livelihoods and the
C

547 environment they depend on, has availed the need to test emerging vulnerability assessments
AC

548 methodologies and see how best they can inform policies locally. Though the Susan Cutter
549 SoVI model created for United States might not be universally applicable because it was
550 intended for use in that context, the results of this study show that it is applicable in
551 Botswana. However, an adjustment to the original model was made to best capture the socio-
552 economic and demographic structure of communities in Botswana. The results between both
553 the original and adjusted models were similar: the same number of districts which scored
554 above the mean scores were the same; the highly vulnerable and the least vulnerable districts
555 were the same; and the factors contributing to social vulnerability almost the same.
ACCEPTED MANUSCRIPT
556 The adjusted SoVI model produced a total of three factors: vulnerability preparedness;
557 socio-economic status; and vulnerable population groups. The majority of the social
558 vulnerability concepts produced from this study are consistent with those from previous
559 studies such as Cutter, Boruff and Shirley (2003), Boruff and Cutter (2007), Dunno (2011),
560 Letsie (2015), Frigerio et al. (2018), and Mavhura (2019). They include: socio-economic
561 status, vulnerable population groups, poverty, gender, age, employment, poverty, rural/urban,

PT
562 renters, occupation, education, family structure, population growth, medical services, and
563 social dependence.

RI
564 The adjusted SoVI model shows that social vulnerability is mainly driven by the following
565 factors: vulnerability preparedness; socio-economic status; and vulnerable population groups.

SC
566 The afore-mentioned factors are driven by these underlying indicators: high percentages of
567 population under 5 years and over 65 years old, high percentage using any source of biomass
568 for cooking, high percentage receiving social security nets, high percentage of female-headed
569

U
households, and high percentage unemployed and living in poverty. This result is consistent
AN
570 with the finding of the study by Chen et al. (2013) which revealed that employment and
571 poverty, education, and family size are drivers of social vulnerability in China. Moreover,
M

572 Cutter, Boruff and Shirley (2003) and Dunno (2011) also found that personal wealth, tenancy,
573 vulnerability preparedness, ageing and social dependence, culture, and social structure are the
D

574 main factors contributing to social vulnerability in the United States and the Virgin Islands,
575 respectively. The dominant variables contributing to social vulnerability factors in this study
TE

576 are consistent with the findings by Cutter et al. (2003) and Chen et al. (2013) which show the
577 following as dominant variables contributing to social vulnerability in the United States and
EP

578 China: Percent renting; Percent with no secondary; Percent of females; Percent females in
579 labour force; Percent under five; Percent over 65; Percent using any source of biomass for
C

580 cooking; Percent living in poverty.


581 The results further reveal that the dominant variable driving the ‘Vulnerability
AC

582 Preparedness Factor’ in Botswana is in line with those of the Caribbean Islands (Boruff and
583 Cutter, 2007). It include the percent of homes cooking with any source of biomass (e.g. wood
584 or charcoal and other methods), though Boruff and Cutter (2007) identified additional
585 indicators driving the vulnerability preparedness factor. They include: percent of homes using
586 electricity for cooking, percent of homes with no telephone service, and percent of homes not
587 owning a vehicle. The results of this study attempt to improve the original social vulnerability
588 model by including variables on vulnerability preparedness. The higher the percent of
589 households using alternative sources for cooking other than electricity, the more vulnerable
ACCEPTED MANUSCRIPT
590 the households are to the impacts of natural hazards. Dunno (2011) asserts that homes using
591 alternative fuels as source of energy for cooking would be more vulnerable than those using
592 electricity because the power grid would be restored quicker after a disaster event. On the
593 contrary, it can also be argued that it’s not all the time that power is restored quickly into the
594 power grid. The time taken to restore the power is dependent on the magnitude of the
595 damage. In this instance this situation can create a new vulnerability emanating from having

PT
596 no alternative sources of energy and longer period of waiting for restoration of power supply,
597 more especially in urbanized zones.

RI
598 A comparative examination of the levels of vulnerability between districts, as well as
599 establishing factors that most contribute to social vulnerability in the individual census

SC
600 district of Botswana show the following: The census districts with the highest vulnerability
601 scores are Ngamiland West in the north, followed by Central Tutume in the central. And the
602 districts with the least vulnerability are found in the south (Jwaneng), followed by Sowa
603

U
Town located in the central. Most of the urban districts (cities and towns) are ranked low and
AN
604 medium low in terms of the social vulnerability index scores as compared to the rest of the
605 districts (rural districts). This is expected because the socio-economic status of the urban
M

606 residents shown by access to medical services, employment, improved technology, and
607 transport and communication infrastructure, among other services, is better than that of the
D

608 rural dwellers, hence low vulnerability in the urban areas. This is consistent with literature
609 elsewhere (Cutter, Mitchell, and Scott, 2000), which state that medium to high socio-
TE

610 economic status enables communities to absorb and recover from losses more quickly due to
611 financial security (savings, insurance), social safety nets, decent jobs and entitlement
EP

612 programs, among others. Low socio-economic status associated with low incomes, high
613 unemployment, and low levels of education leads to construction of housing units from poor
C

614 building materials making them more vulnerable to floods (Mavhura, 2019). This increases
615 flood vulnerability.
AC

616 The results of this study reveal that different factors in individual census district contribute
617 to high vulnerability. For example, higher scores on Component 1 (Vulnerability
618 Preparedness) and Component 2 (Socio-Economic Status) contributed to high social
619 vulnerability scores of Ngamiland West and Central Tutume. Furthermore, factors such as
620 socio-economic status contribute to high vulnerability according to the results of this study
621 because most women work in low paying jobs, they are caregivers, and live in poverty; hence
622 their preparedness and coping ability in the face of a natural hazard event are compromised.
623 Cutter, Boruff and Shirley (2003) adds that, women face a more difficult time during
ACCEPTED MANUSCRIPT
624 recovery compared to males, mostly because of their sector-specific employment, lower
625 salaries, and family care responsibilities. Furthermore, females are usually linked to a higher
626 mortality rate during disasters compared to their male counterparts and this is attributable to
627 limited access to resources and information which in turn affects their physical and mental
628 health pre and post disasters (Frigerio et al. 2018). Statistics Botswana (2018) reports that
629 females had a higher prevalence rate of disability at about 4.7 percent compared to males at

PT
630 3.7, and these figures further show how vulnerable females are to the impacts of natural
631 hazards.

RI
632 Literature elsewhere (e.g. Cutter, 1996; Hewitt, 1997; Chen et al. 2013) also confirm that
633 the “She” people are the most vulnerable to natural hazards. For example, a study by Chen et

SC
634 al. (2013) reports that the high scores of the minorities factor (which is made of percentage of
635 female population) is situated in the Zhejiang Province which has one of the main habitations
636 of the female (She) people, who relocated to Zhejiang during Ming and Qing Dynasties. In
637

U
Botswana a study by Mogotsi, Nyangito and Nyariki (2012) revealed that the majority of
AN
638 households in Bobonong area are highly vulnerable (68%) and this could be attributable to
639 fact that most households are female-headed (64%). According to Fako and Molamu (1995)
M

640 these households are generally resource-poor; hence they are unable to mobilize enough
641 resources to cushion themselves against the impacts of drought, among other natural hazards.
D

642 Muyambo et al. (2017) asserts that gender and external support contribute highly to social
643 vulnerability to natural hazards, particularly drought: the impacts affects females the most,
TE

644 because they are disadvantaged in decision-making, even in the whole farming enterprise.
645 The decisions include among others, having to reduce the number of livestock prior to
EP

646 droughts, in order to cushion themselves against high mortality due to long dry spells. This is
647 applicable in the context of Botswana. The female population in Botswana followed an
C

648 update trend between the years 1998 (840,000) and 2017 (1,120,285) (Statistics Botswana,
649 2018), and this implies an increase in the number of population vulnerable to natural hazards.
AC

650 In Botswana, there is an unequitable distribution of resources like employment and


651 education opportunities, health services, and transport and communication infrastructure. The
652 urban areas of Botswana provide more of these social resources and services than rural areas.
653 This has a bearing on vulnerability, hence a significant spatial variability of individual factors
654 contributing to social vulnerability in the census districts of Botswana. For example, the
655 socio-economic status factor shows a clear geographic pattern, with the highest values in the
656 Ngamiland East, Kweneng West, North East, and Ngamiland West. The high vulnerability in
657 the aforementioned urban-rural districts is explained by high percentage living in poverty,
ACCEPTED MANUSCRIPT
658 high percentage of female-headed households and high percentage unemployed. Nonetheless,
659 it is worth noting that Botswana’s population is predominantly urban, with about 63.9 percent
660 of the population residing in urban areas i.e. cities, towns, and urban villages (Statistics
661 Botswana, 2018). It can therefore be assumed that slightly over half of the population of
662 Botswana is less vulnerable to natural hazards by virtue of them residing in urban areas with
663 accessible social resources and services. This result is in line with that of Chen et al. (2013)

PT
664 which reports that, in China, highly urbanized areas are less vulnerable as compared to the
665 rural areas, and social resources including among others, education opportunities, medical

RI
666 services, government financial aids, and job opportunities are not equally distributed between
667 urban and rural areas. Education and awareness of the impacts of natural hazards (e.g. floods)

SC
668 coupled with specific adaptive management policies can be the most effective preventers of
669 lots of tragedies (Wright, 2019). The worrying situation in Botswana is that, though the
670 national school enrolment is high rural areas still experience high proportions population who
671

U
had left school with about 46.7 percent reported in 2017 (Statistics Botswana, 2018).
AN
672 With regards to poverty, which constitutes the socio-economic status (component 2), it is
673 evident from the findings of this study that the percent living in poverty was highest in
M

674 Ngamiland West, Ngamiland East, Kweneng West, North East, and Ghanzi and these are the
675 districts which have high values of Component 2. Many of these study units have a lot of
D

676 rural settlements (rural villages, lands and cattle posts) which are harboured by the majority
677 of the population with low economic status. This shows unbalanced economic development,
TE

678 since the urban areas (in the south) are more developed than rural areas (in the west). The
679 results of the study by Cutter, Boruff and Shirley (2003) show that counties with high SoVI
EP

680 scores in the United States were labelled the most vulnerable, and they include a geographic
681 mix of highly urbanized counties, large Hispanic and/or Native American populations, and
C

682 socially dependent populations (those in poverty and lacking in education). Cutter, Boruff
683 and Shirley (2003) further adds that poverty is a main driver of social vulnerability as fewer
AC

684 individual and community resources for recovery are available, thus making them less
685 resilient to the impacts of natural hazards.
686 One of the factors that can exacerbate the existing problem of high social vulnerability to
687 natural hazards is that of the spread of HIV/AIDS, particularly in rural areas. High human
688 mortality due to HIV/AIDS can lead many households to extreme poverty. Loss of
689 economically active population more especially the caregivers or those commonly known as
690 breadwinners in poor households can leave the dependents in disarray, making them highly
691 vulnerable to natural hazards. Female-headed households are the hardest hit. UNDP (2008)
ACCEPTED MANUSCRIPT
692 asserts that in sub-Saharan Africa the translation of drought, as an example of a natural
693 hazard, into famine is mediated by internal displacements, HIV/AIDS, and economic crisis,
694 among others. This shows the potential impact of HIV/AIDS on social vulnerability. Not so
695 many studies have been done on the potential impact of HIV/AIDS on social vulnerability in
696 rural and urban areas of Botswana. Therefore this provides an opportunity for future research
697 in the field of social vulnerability. Statistics Botswana (2019) reports an increasing mortality

PT
698 due to HIV-related and other viral diseases mortality: the percentage of deaths related HIV
699 increased from 1.1 percent in 2013 to 4.0 percent in 2017. The figures shows how worrying

RI
700 the situation of HIV/AIDS is, in the face of increasing impacts of climate change.
701 The ‘vulnerable population groups’ factor shows high values in the study units with high

SC
702 percentages of children (under five years) and elderly (over 65 years old). These highest
703 values are located in the south (Kweneng East) and in the central (Central Serowe/Palapye
704 and Central Tutume). Hence the reason why Kgalagadi North and Kgalagadi South are least
705

U
vulnerable compared to a number of districts located in the central. This is consistent with the
AN
706 results of a study by Chen et al. (2013) which reveal that study units with high family size
707 factor scores are those with high percentages of the elderly. Cutter, Boruff and Shirley (2003)
M

708 add that “extremes of the age spectrum affect the movement out of harm’s way because
709 parents lose time and money caring for children when day-care facilities are affected while
D

710 the elderly may have mobility constraints or mobility concerns increasing the burden of care
711 and lack of resilience” (p.246). It is important for the humankind to anticipate and make
TE

712 better plans for the future climate change impacts particularly on low-income, and elderly
713 communities (Wright, 2019). Nonetheless, Dunno (2011) argues that being an elderly person
EP

714 does not necessarily mean that you are vulnerable to natural hazards, rather your economic
715 status is a major determinant in this regard. Old retirees (with retirement packages) who live
C

716 in urban areas are less vulnerable compared to those rural elderly with no retirement packages
717 because they are economically better-off to cope with the impacts of a natural hazard. For
AC

718 Botswana, this is only true for Gaborone district, which is the capital city and has a high
719 score of the vulnerable population groups’ factor. Generally the majority of the study sites
720 with high scores of social dependency factor are located in the rural districts, hence high
721 vulnerability. With the increase in the number of the elderly in Botswana, this presents a
722 worrying situation in the face of natural disasters and in terms of economic benefits. Statistics
723 Botswana (2018) reports an increase of the elderly from 5 percent in 2001 to 5.4 percent in
724 2017, and this increase also shows that a shift from young age structure to ageing structure.
725 This will eventually diminish the demographic dividend benefit.
ACCEPTED MANUSCRIPT
726
727 5. Conclusion
728 The study’s results reveal that both Ngamiland West and Central Tutume are the most
729 vulnerable due to high illiteracy levels and high poverty levels. Therefore it can be concluded
730 from this that low socio-economic status is associated with high vulnerability. The opposite is
731 true for urban centres, with middle to high socio-economic status. For example, Jwaneng

PT
732 Census District is the least vulnerable due to low percentage with no secondary education,
733 low unemployment, low percentage under social security, and less people living in poverty,

RI
734 among others.
735 In Botswana social vulnerability is mainly driven by: high percentage receiving social

SC
736 security, high poverty levels, high percentage of the unemployed, low literacy, high
737 percentage of children and elderly, and high percentage using biomass for cooking. Therefore
738 an important conclusion can be drawn from the fact there is no indicator that can be solely
739

U
used measure vulnerability, a combination of factors (a composite) can be used as a good
AN
740 measure of vulnerability. Hence the reason why this study yielded the following factors of
741 social vulnerability: vulnerability preparedness, socio-economic status; and vulnerable
M

742 population groups.


743 Further Susan Cutter Model created for United States can be applied in the setting of
D

744 Botswana, with additional variables (indicators) included in the original model in order to
745 appropriately assess the social fabric of the census districts of Botswana.
TE

746 Based on the analysis of the findings on the application of social vulnerability index model
747 to Botswana’s setting the following key recommendations can be made: That the government
EP

748 should formulate policies that avail more opportunities for women to be engaged in some
749 form of employment, as well as to encourage them to start up their own businesses. Most of
C

750 the women carry the burden of caring for large families on their own; this makes them more
751 vulnerable to the impacts of natural hazards. It is made worse by the fact the majority of them
AC

752 particularly in the rural areas work in low paying jobs, therefore they are not in a better place
753 to cope and adapt to the impacts of natural hazards. Since there is high percentage of the
754 elderly people in the rural areas and the trend is on the increase, the government should
755 improve its efforts in terms of the provision of the health services, social safety nets, and
756 other efforts which are intended to improving the socio-economic status and living conditions
757 of the elderly. This will reduce their vulnerability to natural hazards. Furthermore, the
758 emergence of the modern nuclear families as opposed to the traditional extended families,
759 disadvantages the elderly as they are left unattended to most of the time. The traditional
ACCEPTED MANUSCRIPT
760 support system is compromised, hence this worsens the elderly’s health and socio-economic
761 status.
762
763 6. Limitations
764 The use of secondary data limits the study to indicators collected by the survey. Lack of data in other
765 important indicators of social vulnerability presents another limitation. Some of the important

PT
766 indicators had no data at either national or district level, for example, per capita number of community
767 hospitals, per capita income and amount of money expended in removing the effects of natural
768 hazards events. This therefore provides an opportunities for future research in Botswana. Another

RI
769 opportunity for future include running correlation between social vulnerability results and livestock
770 mortality, low crop yields, among other drought-related variables.

SC
771
772
773

U
774 References
775
AN
776 Batisani, N., (2010). Dynamics of Cultivated Land & its Association with Rainfall Variability
777 in Botswana; Implications for Food Security under Climate Change, Climate Change
M

778 & Natural Resources Conflicts in Africa, Institute for Security Studies, Pgs 205 – 219.
779 Blaikie, P., Cannon, T., Davis, I., & Wisner, B., (1994). At Risk: Natural Hazards, People's
D

780 Vulnerability, and Disasters. London: Routledge.


TE

781 Boruff, B. J., Emrich, C., & Cutter, S.L., (2005). Hazard Vulnerability of U.S. Coastal
782 Counties, Journal of Coastal Research 21 (5): 932-942.
783 Central Statistics Office, (2009).Natural Disasters Digest, by Government printers.
EP

784 Central Statistics Office, (2011), 2011 Population and Housing Census; Preliminary Results
785 Brief, Pgs. 1-8. (www.cso.gov.bw)
C

786 Chen, W., Cutter, S.L., Emrich, C.T., & Shi, P., (2013). Measuring Social Vulnerability to
AC

787 Natural Hazards in the Yangtze River Delta Region, China. Int. J. Disaster Risk Sci.
788 2013, 4 (4): 169 – 181.
789 Chipanshi, A.C, Chanda, R., & Totolo, O., (2003). Vulnerability Assessment of the Maize
790 and
791 Sorghum Crops to Climate Change in Botswana, Climatic Change 61, 339 – 360.
792 Cutter, S. L. (1996). Vulnerability to Environmental Hazards, Progress in Human Geography
793 20 (4): 529-539.
794 Cutter, S. L., Mitchell, J.T., & Scott, M.S., (2000). Revealing the Vulnerability of People
ACCEPTED MANUSCRIPT
795 and Places: A Case Study of Georgetown County, South Carolina, Annals of the
796 Association of American Geographers 90(4): Pgs. 713–37.
797 Cutter, S. L., Boruff, B.J., & Shirley, W.L., (2003). Social Vulnerability to Environmental
798 Hazards. Social Science Quarterly 84(1): 242-261.
799 Cutter, S.L., & Finch C., (2008). Temporal and Spatial Changes in Social Vulnerability to
800 Natural Hazards, Proc Natl Acad Sci U S A. 105(7): Pages 2301–2306

PT
801 Cutter, S. L., Barnes, L., Berry, M., Burton, C., Evans, E. Tate, E., & Webb, J., (2008). A
802 Place-

RI
803 based Model for Understanding Community Resilience to Natural Disasters. Global
804 Environmental Change 18(4): 598-606.

SC
805 Cutter, S.L., Emrich C.T., Webb J.J., & Morath D., (2009). Social Vulnerability to Climate
806 Variability Hazards: A Review of the Literature. Hazards and Vulnerability Research
807 Institute, Final Report to Oxfam America, 1-44
808

U
Cutter, S. L., & Morath, D. P., (2014). The Evolution of the Social Vulnerability Index. In:
AN
809 Measuring Vulnerability to Natural Hazards: Towards Disaster Resilient Societies,
810 edited by J. Birkmann. Published by New York: United Nations University Press.
M

811 Dunno, C.H., (2011). Measuring Social Vulnerability to Natural Hazards: An Examination
812 of the United States Virgin Islands, Ph.D. Thesis, Directed by Dr. Rick Bunch. 207
D

813 pp.
814 Emrich, C.T., & Cutter, S.L., (2011).Social Vulnerability to Climate-Sensitive Hazards in
TE

815 the Southern United States. Wea. Climate Soc., 3, 193–208.


816 Fako, T.T., & Molamu, L., (1995). “The seven-year drought, household food security and
EP

817 vulnerable groups in Botswana.” Pula: Botswana Journal of African Studies, 9:


818 (pp. 48-70).
C

819 Frigerio, I., Carnelli, F., Cabinio, M., & De amicis, M. (2018). Spatiotemporal Pattern of
820 Social Vulnerability in Italy. International Journal of Disaster Risk Science. 9: 249-
AC

821 262 10.1007/s13753-018-0168-7.


822 George, M., & Schensul, D., (eds.) (2013). The Demography of Adaptation to Climate Change.
823 New York, London and Mexico City: UNFPA, IIED and El Colegiode México. Page XV
824 George, M., & Schensul, D., (eds.) (2013). The Demography of Adaptation to Climate
825 Change. New York, London and Mexico City: UNFPA, IIED and El Colegiode
826 México. Page XV
827 Heinz Center for Science, Economics, and the Environment. (2000). The Hidden Costs of
ACCEPTED MANUSCRIPT
828 Coastal Hazards: Implications for Risk Assessment and Mitigation. Covello, Cal.:
829 Island Press.
830 Heinz Center, (2002). Human Links to Coastal Disasters, Washington D.C.: The H. John
831 Heinz III Center for Science, Economics and the Environment.
832 Hewitt, K., (1997). Regions of Risk: A Geographical Introduction to Disasters. Addison
833 Wesley Longman Limited: London, England.

PT
834 Holand, I. S., & Lujala, P., (2013). Replicating and Adapting an Index of Social Vulnerability
835 to a New Context: A Comparison Study for Norway. The Professional Geographer 65

RI
836 (2): (pp. 312–328).
837 Letsie, M.M.A., (2015). An Assessment of Place Vulnerability to Natural Hazards in South-

SC
838 Western Lesotho (Quthing and Mohale’s Hoek Districts). A Ph.D dissertation
839 submitted to Faculty of Science of the University of the Witwatersrand. Pgs 1-341
840 Mavhura, E., (2019). Analysing drivers of vulnerability to flooding: a systems approach,
841 South
U
AN
842 African Geographical Journal, 101:1, 72-90, DOI:10.1080/03736245.2018.1541020
843 Mavhura, E., Manyena, B. S., & Collins, A. E. (2017). An Approach for H social
M

844 vulnerability in context: The case of flood hazards in Muzarabani district, Zimbabwe.
845 Geoforum, 86, 103–117.
D

846 Mogotsi, K., Nyangito, M.M., & Nyariki, D.M., (2012).Vulnerability of Rural Agro-Pastoral
847 Households to Drought in Semi-arid Botswana, Livestock Research for Rural
TE

848 Development, 24 (10)


849 Morrow, B.H., (1999). Identifying and Mapping Community Vulnerability, Disasters. 23(1):
EP

850 1-18.
851 Muyambo, F., Jordaan, A.J. & Bahta, Y.T., (2017). ‘Assessing social vulnerability to drought
C

852 in South Africa: Policy implication for drought risk reduction’, Jàmbá: Journal of
853 Disaster Risk Studies 9(1), a326. https://doi.org/10.4102/jamba.v9i1.326
AC

854 O’Brien, K, Quinlan, T, & Ziervogel, G., (2010). Vulnerability interventions in the context of
855 multiple stressors: lessons from the Southern Africa Vulnerability Initiative (SAVI).
856 Environ Science Policy, vol. 12, Pgs. 23–32.
857 Sokona, Y., & Fatma, D. (2001). Climate change impacts: can Africa cope with the
858 Challenges? Climate Policy 1: Pages 117-123.
859 Stafford, S., & Abramowitz, J. (2017). An analysis of methods for identifying social
ACCEPTED MANUSCRIPT
860 vulnerability to climate change and sea level rise: a case study of Hampton Roads,
861 Virginia. Nat. Hazards 85, 1089–1117.
862 Statistics Botswana, (2014). Botswana Environment Statistics: Human Settlements Report
863 2013. Published by Statistics Botswana.
864 Statistics Botswana, (2019). Botswana – Maternal Mortality Ratio 2017. Pgs. 1-12.
865 (www.statsbots.org.bw)

PT
866 Statistics Botswana, (2018). Botswana Demographic Survey Report 2017. Published by Statistics
867 Botswana. Pgs. 1-112. ISBN NO. 978-99968-2-046-5
868 Taylor, C.C., (1977). Principal Component Analysis and Factor Analysis, Exploring Data

RI
869 Structures (C.A. O’Muircheartaigh and C. Payne, editors), John Wiley and Sons, Inc.,
870 New York, pp. 89-124.

SC
871 UNDP (2008) “Bridging disaster and Development in Sri-lanka” GMSL-DMIP, UNDP
872 Vyas, S., & Kumaranayake, L., (2006). Constructing Socio-economic Status Indices: How to

U
873 Use Principal Component Analysis. Health Policy and Planning, Vol. 21, pp: 459-
AN
874 468.
875 Wright L.D. (2019). Future Societal Vulnerability, Risk and Adaptability. In: Wright L.,
876 Nichols C. (eds) Tomorrow's Coasts: Complex and Impermanent. Coastal Research
M

877 Library, vol 27. Springer, Cham


878
D

879
880
TE
C EP
AC

You might also like